Diabetes Care - lib.ajaums.ac.irlib.ajaums.ac.ir/booklist/382463.pdf · Management of Type 2...

Diabetes Care

December 2007, Volume 30, Number 12, PP 2989-3151,e128-e143

Clinical Care/Education/Nutrition/Psychosocial Research:

Recruitment to a Clinical Trial Improves Glycemic Control in Patients With Diabetes

Edwin A.M. Gale, Scott D. Beattie, Jinghui Hu, Veikko Koivisto, and Meng H. Tan Diabetes Care 2007 30: 2989-2992.

Lipid, Glycemic, and Insulin Responses to Meals Rich in Saturated, cis-Monounsaturated, and Polyunsaturated (n-3 and n-6) Fatty Acids in Subjects With Type 2 Diabetes

Meena Shah, Beverley Adams-Huet, Linda Brinkley, Scott M. Grundy, and Abhimanyu Garg Diabetes Care 2007 30: 2993-2998.

Burden of Comorbid Medical Conditions and Quality of Diabetes Care Jewell H. Halanych, Monika M. Safford, Wendy C. Keys, Sharina D. Person, James M. Shikany, Young-Il Kim, Robert M. Centor, and Jeroan J. Allison Diabetes Care 2007 30: 2999-3004.

Diabetes, Depression, and Death: A randomized controlled trial of a depression treatment program for older adults based in primary care (PROSPECT)

Hillary R. Bogner, Knashawn H. Morales, Edward P. Post, and Martha L. Bruce Diabetes Care 2007 30: 3005-3010.

Glycemic Effects of Moderate Alcohol Intake Among Patients With Type 2 Diabetes: A multicenter, randomized, clinical intervention trial

Iris Shai, Julio Wainstein, Ilana Harman-Boehm, Itamar Raz, Drora Fraser, Assaf Rudich, and Meir J. Stampfer Diabetes Care 2007 30: 3011-3016.

Management of Type 2 Diabetes in Treatment-Naive Elderly Patients: Benefits and risks of vildagliptin monotherapy

Richard E. Pratley, Julio Rosenstock, F. Xavier Pi-Sunyer, Mary Ann Banerji, Anja Schweizer, Andre Couturier, and Sylvie Dejager Diabetes Care 2007 30: 3017-3022.

Continuous Home Monitoring of Glucose: Improved glycemic control with real-life use of continuous glucose sensors in adult subjects with type 1 diabetes

Satish K. Garg, William C. Kelly, Mary K. Voelmle, Peter J. Ritchie, Peter A. Gottlieb, Kim K. McFann, and Samuel L. Ellis Diabetes Care 2007 30: 3023-3025.

Targeting Glucose in Acute Myocardial Infarction: Has glucose, insulin, and potassium infusion missed the target?

Ajay Chaudhuri, Michael Miller, Richard Nesto, Noah Rosenberg, and Paresh Dandona Diabetes Care 2007 30: 3026-3028.

Diabetes or Impaired Glucose Tolerance: Does the label matter? Carmen Lara, Sergio Ponce de Leon, Hector Foncerrada, and Martin Vega Diabetes Care 2007 30: 3029-3030.

Flexible Intensive Versus Conventional Insulin Therapy in Insulin-Naive Adults With Type 2 Diabetes: An open-label, randomized, controlled, crossover clinical trial of metabolic control and patient preference

Christof Kloos, Alexander Sämann, Thomas Lehmann, Anke Braun, Barbara Heckmann, and Ulrich A. Müller Diabetes Care 2007 30: 3031-3032.

Perception of Offspring Risk for Type 2 Diabetes Among Patients With Type 2 Diabetes and Their Adult Offspring

Masakazu Nishigaki, Koji Kobayashi, Takako Hitomi, Taeko Yokomura, Mitsunao Yokoyama, Naoto Seki, and Keiko Kazuma Diabetes Care 2007 30: 3033-3034.

Epidemiology/Health Services Research:

Trends in Hospitalizations for Diabetes Among Children and Young Adults: United States, 1993–2004

Joyce M. Lee, Megumi J. Okumura, Gary L. Freed, Ram K. Menon, and Matthew M. Davis Diabetes Care 2007 30: 3035-3039.

Influence of Family History of Diabetes on Incidence and Prevalence of Latent Autoimmune Diabetes of the Adult: Results from the Nord-Trøndelag Health Study

Sofia Carlsson, Kristian Midthjell, and Valdemar Grill Diabetes Care 2007 30: 3040-3045.

Oral Health Knowledge, Attitude, and Practices and Sources of Information for Diabetic Patients in Lahore, Pakistan

Kamran Masood Mirza, Ayyaz Ali Khan, Munawar Manzoor Ali, and Saima Chaudhry Diabetes Care 2007 30: 3046-3047.

Pathophysiology/Complications:

Influence of Flickering Light on the Retinal Vessels in Diabetic Patients Aleksandra Mandecka, Jens Dawczynski, Marcus Blum, Nicolle Müller, Christoph Kloos, Gunter Wolf, Walthard Vilser, Heike Hoyer, and Ulrich Alfons Müller Diabetes Care 2007 30: 3048-3052.

Atrophy of Foot Muscles in Diabetic Patients Can Be Detected With Ultrasonography

Kaare Severinsen, Annette Obel, Johannes Jakobsen, and Henning Andersen Diabetes Care 2007 30: 3053-3057.

Neurovascular Factors in Wound Healing in the Foot Skin of Type 2 Diabetic Subjects

Singhan T.M. Krishnan, Cristian Quattrini, Maria Jeziorska, Rayaz A. Malik, and Gerry Rayman Diabetes Care 2007 30: 3058-3062.

Evaluation of Polyneuropathy Markers in Type 1 Diabetic Kidney Transplant Patients and Effects of Islet Transplantation: Neurophysiological and skin biopsy longitudinal analysis

Ubaldo Del Carro, Paolo Fiorina, Stefano Amadio, Luisa De Toni Franceschini, Alessandra Petrelli, Stefano Menini, Filippo Martinelli Boneschi, Stefania Ferrari, Giuseppe Pugliese, Paola Maffi, Giancarlo Comi, and Antonio Secchi Diabetes Care 2007 30: 3063-3069

Incidences, Treatments, Outcomes, and Sex Effect on Survival in Patients With End-Stage Renal Disease by Diabetes Status in Australia and New Zealand (1991–2005)

Emmanuel Villar, Sean Haw Chang, and Stephen Peter McDonald Diabetes Care 2007 30: 3070-3076.

Microvascular and C-Fiber Function in Diabetic Charcot Neuroarthropathy and Diabetic Peripheral Neuropathy

Neil Baker, Alistair Green, Singhan Krishnan, and Gerry Rayman Diabetes Care 2007 30: 3077-3079.

C-Reactive Protein in Diabetic and Nondiabetic Patients With Acute Myocardial Infarction

Wolfgang Otter, Michael Winter, Wittich Doering, Eberhard Standl, and Oliver Schnell Diabetes Care 2007 30: 3080-3082.

Pregnancy-Associated Plasma Protein-A Levels Are Related to Glycemic Control but Not to Lipid Profile or Hemostatic Parameters in Type 2 Diabetes

Silvia Pellitero, Jordi L. Reverter, Eduarda Pizarro, Maria Cruz Pastor, Maria Luisa Granada, Dolors Tàssies, Juan-Carlos Reverter, Isabel Salinas, and Anna Sanmartí Diabetes Care 2007 30: 3083-3085.

Low Serum Angiogenin Concentrations in Patients With Type 2 Diabetes Janusz Siebert, Magdalena Reiwer-Gostomska, Zofia Babinska, Jolanta Mysliwska, Andrzej Mysliwski, Ewa Skopinska-Rózewska, Ewa Sommer, and Piotr Skopinski Diabetes Care 2007 30: 3086-3087.

Latent Autoimmune Diabetes in Adults in a South Asian Population of the U.K. Abigail C. Britten, Karen Jones, Carina Törn, Magnus Hillman, Birgitte Ekholm, Sudhesh Kumar, Anthony H. Barnett, and Marilyn Ann Kelly Diabetes Care 2007 30: 3088-3090.

GAD Antibody in Multiplex Diabetic Pedigrees of Chinese Tao Chen, Yan Ren, Yang Long, Xiangxun Zhang, Honglin Yu, and Haoming Tian Diabetes Care 2007 30: 3091-3092.

Cardiovascular and Metabolic Risk:

Intrahepatic Fat Accumulation and Alterations in Lipoprotein Composition in Obese Adolescents: A perfect proatherogenic state

Anna M.G. Cali, Tosca L. Zern, Sara E. Taksali, Ana Mayra de Oliveira, Sylvie Dufour, James D. Otvos, and Sonia Caprio Diabetes Care 2007 30: 3093-3098.

Metabolic Syndrome and Incident End-Stage Peripheral Vascular Disease: A 14-year follow-up study in elderly Finns

Jianjun Wang, Sanna Ruotsalainen, Leena Moilanen, Päivi Lepistö, Markku Laakso, and Johanna Kuusisto Diabetes Care 2007 30: 3099-3104.

Does Waist Circumference Predict Diabetes and Cardiovascular Disease Beyond Commonly Evaluated Cardiometabolic Risk Factors?

Peter M. Janiszewski, Ian Janssen, and Robert Ross Diabetes Care 2007 30: 3105-3109.

Gene Expression of Adiponectin Receptors in Human Visceral and Subcutaneous Adipose Tissue Is Related to Insulin Resistance and Metabolic Parameters and Is Altered in Response to Physical Training

Matthias Blüher, Catherine J. Williams, Nora Klöting, Alex Hsi, Karen Ruschke, Andreas Oberbach, Mathias Fasshauer, Janin Berndt, Michael R. Schön, Alicja Wolk, Michael Stumvoll, and Christos S. Mantzoros Diabetes Care 2007 30: 3110-3115.

Waist Circumference Thresholds Provide an Accurate and Widely Applicable Method for the Discrimination of Diabetes

Obesity in Asia Collaboration Diabetes Care 2007 30: 3116-3118.

Change in Albuminuria Is Predictive of Cardiovascular Outcome in Normotensive Patients With Type 2 Diabetes and Microalbuminuria

Adrienne A.M. Zandbergen, Liffert Vogt, Dick de Zeeuw, Steven W.J. Lamberts, Rob J.T.H. Ouwendijk, Marinus G.A. Baggen, and Aart H. Bootsma Diabetes Care 2007 30: 3119-3121.

Simvastatin Reduces Plasma Osteoprotegerin in Type 2 Diabetic Patients With Microalbuminuria

Birgitte Nellemann, Lars C. Gormsen, Jens Dollerup, Ole Schmitz, Carl E. Mogensen, Lars M. Rasmussen, and Søren Nielsen Diabetes Care 2007 30: 3122-3124.

Lifestyle Intervention and Adipokine Levels in Subjects at High Risk for Type 2 Diabetes: The Study on Lifestyle intervention and Impaired glucose tolerance Maastricht (SLIM)

Eva Corpeleijn, Edith J.M. Feskens, Eugène H.J.M. Jansen, Marco Mensink, Wim H.M. Saris, and Ellen E. Blaak Diabetes Care 2007 30: 3125-3127.

Impaired Postprandial Blood Flow in Adipose Tissue May Be an Early Marker of Insulin Resistance in Type 2 Diabetes

George Dimitriadis, Vaia Lambadiari, Panayota Mitrou, Eirini Maratou, Eleni Boutati, Demosthenes B. Panagiotakos, Theofanis Economopoulos, and Sotirios A. Raptis Diabetes Care 2007 30: 3128-3130.

Reviews/Commentaries/ADA Statements:

Review Articles:

Diabetes, the Metabolic Syndrome, and Ischemic Stroke: Epidemiology and possible mechanisms

Ellen L. Air and Brett M. Kissela Diabetes Care 2007 30: 3131-3140.

Commentary:

Lessons From the Avandia Controversy: A new paradigm for the development of drugs to treat type 2 diabetes

Robert I. Misbin Diabetes Care 2007 30: 3141-3144.

Perspectives on the News:

Diabetes and Obesity: Part 1 Zachary T. Bloomgarden Diabetes Care 2007 30: 3145-3151.

Online Letters: Observations: Effects of Maternal Diabetes on Visual Evoked Potentials and Early Psychomotor Development of the Offspring

Mario Brinciotti, Maria Matricardi, Antonietta Colatrella, Francesco Torcia, Francesco Fallucca, and Angela Napoli Diabetes Care 2007 30: e128.

Metformin in Heart Failure Silvio E. Inzucchi, Frederick A. Masoudi, and Darren K. McGuire Diabetes Care 2007 30: e129.

Islet Autotransplantation Restores Normal Glucose Tolerance in a Patient With Chronic Pancreatitis

Severine Illouz, M'Balu Webb, Cristina Pollard, Patrick Musto, Kieran O'Reilly, David Berry, and Ashley Dennison Diabetes Care 2007 30: e130.

Possible Relevance of HLA-DRB1*0403 Haplotype in Insulin Autoimmune Syndrome Induced by -Lipoic Acid, Used as a Dietary Supplement

Tetsuya Yamada, Junta Imai, Yasushi Ishigaki, Yoshinori Hinokio, Yoshitomo Oka, and Hideki Katagiri Diabetes Care 2007 30: e131.

Online Letters: Comments and Responses: The Role of Iron in Diabetes and Its Complications: Reponse to Swaminathan et al.

Giovanni Targher, Massimo Franchini, Martina Montagnana, and Giuseppe Lippi Diabetes Care 2007 30: e132.

On Real-Time Estimates of Blood Glucose Levels: Response to Treviño Oliver J. Gibson, Andrew J. Farmer, Patrick E. McSharry, and Lionel Tarassenko Diabetes Care 2007 30: e133.

On Real-Time Estimates of Blood Glucose Levels: Response to Gibson et al. George Treviño Diabetes Care 2007 30: e134.

Vitamin D, Parathyroid Hormone Levels, and the Prevalence of Metabolic Syndrome in Community-Dwelling Older Adults: Response to Reis et al.

Giuseppe Lippi, Martina Montagnana, Giovanni Targher, and Gian Cesare Guidi Diabetes Care 2007 30: e135.

Vitamin D, Parathyroid Hormone Levels, and the Prevalence of Metabolic Syndrome in Community-Dwelling Older Adults: Response to Lippi et al.

Jared P. Reis and Denise von Mühlen Diabetes Care 2007 30: e136.

Strong Association Between Time Watching Television and Blood Glucose Control in Children and Adolescents With Type 1 Diabetes: Response to Margeirsdottir et al.

Alessandro Giannattasio, Francesca Lugani, Angela Pistorio, Nicola Minuto, Renata Lorini, and Giuseppe d'Annunzio Diabetes Care 2007 30: e137.

Effect of Periodontitis on Overt Nephropathy and End-Stage Renal Disease in Type 2 Diabetes: Response to Shultis et al.

Frank N. Varon Diabetes Care 2007 30: e138.

Effect of Periodontitis on Overt Nephropathy and End-Stage Renal Disease in Type 2 Diabetes: Response to Varon

Wendy A. Shultis, E. Jennifer Weil, Helen C. Looker, Jeffrey M. Curtis, Marc Shlossman, Robert J. Genco, William C. Knowler, and Robert G. Nelson Diabetes Care 2007 30: e139.

Metabolic Syndrome in Hypertensive Patients: Correlation Between Anthropometric Data and Laboratory Findings: Response to Bulhões and Araújo

Ticiana C. Rodrigues, Caroline K. Kramer, Thais Steemburgo, Valesca Dall'Alba, and Mirela J. Azevedo Diabetes Care 2007 30: e140.

Consensus Statement on the Worldwide Standardization of the Hemoglobin A1C Measurement: the American Diabetes Association, European Association for the Study of Diabetes, International Federation of Clinical Chemistry and Laboratory Medicine, and the International Diabetes Federation: Response to the Consensus Committee

George Treviño Diabetes Care 2007 30: e141.

Corneal Sensitivity Is Reduced and Relates to the Severity of Neuropathy in Patients with Diabetes: Response to Tavakoli et al.

Sandip Kumar Dash Diabetes Care 2007 30: e142.

Corneal Sensitivity Is Reduced and Relates to the Severity of Neuropathy in Patients With Diabetes: Response to Dash

Mitra Tavakoli, Panagiotis A. Kallinikos, Nathan Efron, Andrew J.M. Boulton, and Rayaz A. Malik Diabetes Care 2007 30: e143

Recruitment to a Clinical Trial ImprovesGlycemic Control in Patients With DiabetesEDWIN A.M. GALE, MD

1

SCOTT D. BEATTIE, PHD2

JINGHUI HU, PHD2

VEIKKO KOIVISTO, MD3

MENG H. TAN, MD2

OBJECTIVE — We assessed the effect upon A1C of recruitment to a clinical trial in patientswith diabetes who had been screened and interviewed to determine eligibility but whose therapywas otherwise unchanged.

RESEARCH DESIGN AND METHODS — Eligible trials were selected from the globalprogram of an insulin manufacturer. Included were studies in which patients were seen on asingle screening visit, pharmaceutical therapy was not altered before randomization, and A1Cwas measured in a central laboratory at both screening and randomization. Three trials involvingpatients with type 1 diabetes (n � 429) and three trials involving patients with type 2 diabetes(n � 611) were identified for analysis. The main outcome measure was change in A1C. Separateregression equations on the change in A1C were fitted for type 1 and type 2 diabetes and includedeffects of baseline A1C and the interval between the screening and randomization visits.

RESULTS — A1C changed by �0.13% (range �0.09 to �0.26%) in those with type 1diabetes at a median of 28 days and by �0.16% (�0.14 to �0.27%) for those with type 2diabetes at a median of 14 days. The mean change in A1C in those with an interval of �28 dayswas �0.24% for those with type 1 diabetes and �0.23% for those with type 2 diabetes. Thereduction was proportional to initial A1C, with large decreases in those with the poorest initialcontrol but no overall change in those at or below the 10th percentile of A1C.

CONCLUSIONS — Recruitment to a clinical trial, independent of any therapeutic interven-tion, produces improvements in glucose control.

Diabetes Care 30:2989–2992, 2007

D iabetes management centers on thepatient, who assumes direct re-sponsibility for all aspects of his or

her care. This includes day-to-day man-agement of finger-stick glucose measure-ments, diet, exercise, and glucose-lowering medications (oral tablets and/orinsulin injections). Successful integrationof these variables is demanding and re-quires unremitting attention. Behavioralinterventions have been shown to im-prove glucose control (1), but it is noteasy to distinguish between the specificbenefit of such interventions and the non-specific effects of study participation,which include increased patient attentionand motivation. There is some evidence

that patients’ glycemic control will benefitsimply from participation in a clinicalstudy (2).

We wanted to estimate the influenceof study participation on glucose controlby retrospective analysis of the effect ofthe single screening visit that precedes al-location to treatment in a clinical trial. Pa-tients potentially eligible for such trialsmeet a study representative, usually thestudy nurse. In the course of this visit, thenature of the study is explained, writtenconsent to participation is obtained, andbaseline clinical and laboratory measure-ments are made. Advice about aspects ofmanagement, such as blood glucose mon-itoring, may also be offered. No other in-

tervention was offered in the trials weconsidered. Eligibility having been con-firmed, the patient is brought back on asecond occasion and randomized to newtherapy. We set out to analyze the differ-ence in glycemic control, as measured byA1C, between the two visits. Since theanalysis was retrospective, the patientsand clinical teams participating in thesetrials were unaware that differences inglucose control might be considered overthe period between screening and ran-domization, thus allowing us to examinethe influence of recruitment to a clinicaltrial upon glycemic control in isolationfrom any change in therapy.

RESEARCH DESIGN ANDMETHODS — Eli Lilly and Companydesigns and carries out randomized con-trolled trials of new therapies and/or newregimens for type 1 and type 2 diabetesand therefore has a large database foranalysis. Eligible trials were those inwhich 1) the only contact before random-ization had been a single visit in which thepurpose of the study had been explained,informed consent had been obtained, andblood had been taken for testing; 2) phar-maceutical therapy was unchanged be-tween screening and randomization; and3) A1C had been measured at the samelaboratory on each occasion.

Of the clinical trials conducted by thesponsor in its global development pro-gram for insulin lispro, three trials in pa-tients with type 1 diabetes (n � 429) andthree trials in patients with type 2 diabetes(n � 611), conducted between 1994 and2001, met the inclusion criteria. The pa-tients in the three trials in type 1 diabetescame from six European countries, andthe three trials in type 2 diabetes wereconducted in the U.S. Selection criteriafor the type 1 diabetic (studies A, B, andC) and type 2 diabetic (studies D, E, andF) study groups differed in that those inthe type 1 diabetes studies were in rela-tively satisfactory glucose control,whereas those in the type 2 diabetes stud-ies were identified on the basis of poorglucose control.

Patients recruited for the trials in type1 diabetes were aged 18–75 years, had aclinical diagnosis of type 1 diabetes, andall were on four daily injections of human

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Diabetes and Metabolism, University of Bristol, Bristol, U.K.; 2Eli Lilly andCompany, Indianapolis, Indiana; and 3Lilly Research Laboratories, Hamburg, Germany.

Address correspondence and reprint requests to Edwin A.M. Gale, MD, Southmead Hospital, Departmentof Diabetes and Metabolism, Medical School Unit, Bristol BS10 5NB, U.K. E-mail: [email protected].

Received for publication 31 January 2007 and accepted in revised form 21 August 2007.Published ahead of print at http://care.diabetesjournals.org on 28 August 2007. DOI: 10.2337/dc07-

0155.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion

factors for many substances.© 2007 by the American Diabetes Association.The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby

marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

C l i n i c a l C a r e / E d u c a t i o n / N u t r i t i o n / P s y c h o s o c i a l R e s e a r c hO R I G I N A L A R T I C L E

DIABETES CARE, VOLUME 30, NUMBER 12, DECEMBER 2007 2989

insulin. Patients with poor glucose con-trol were excluded; this was defined as anA1C �1.5 times the upper limit of thenondiabetic range. Participants in thethree trials were subsequently random-ized to open or blinded comparisons ofinsulin lispro (Humalog) with conven-tional human insulin. The age criteria fortrials in type 2 diabetes varied from 40 to85 years (study D), 25 to 85 years (studyE), and 18 to 75 years (study F). All weretaking oral glucose-lowering therapy forhyperglycemia and were in suboptimalglucose control. Study D involved pa-tients with sulfonylurea failure, as definedby an A1C �8.5%; study E involvedpatients with A1C �8.0%, despite treat-ment with at least one oral glucose-lowering agent; and study F involvedpatients with A1C �8.0%, despite treat-ment with at least two oral glucose-loweringagents. Subsequent randomization was tocombination oral therapy, once-dailyNPH insulin plus glyburide, or thrice-daily insulin lispro plus glyburide (studyD); to twice- or thrice-daily insulin (studyE); and to once-daily NPH insulin plusmetformin or a thrice-daily insulin mix-ture plus metformin (study F). Of six tri-als included in this analysis, studies A, B,and D have been published in full (3–5),and studies E and F have been publishedin abstract form (6,7). Free oral medica-tion, but not glucose-monitoring equip-ment, was offered to patients in the threestudies in type 2 diabetes during therun-in period.

Statistical analysisData from all six studies were initiallycombined, irrespective of diabetes type. Alinear model was used to explore the ef-fects on change in A1C during the timeinterval between screening and random-

ization, A1C at screening, diabetes type,and the interactions between these fac-tors. Regressions were constrained to passthrough zero, so that the effects includeda linear change over time (interval), anadjustment of this slope for screeningA1C (A1C-by-interval interaction), an ad-justment of the slope for diabetes type (di-abetes type-by-interval interaction), andthe three-way interaction. All terms with aP value �0.05 from the type III sums ofsquares F statistic were included in a hi-erarchical fashion in the model. As a sig-nificant differential slope betweendiabetes types was found, subsequentanalyses were performed separately fortype 1 and type 2 diabetes. Sensitivityanalyses were conducted by restrictingthe regression to patients whose intervalwas at least 28 days. The analyses includ-ing all patients produced similar results tothese, so the final models used data fromall patients. Confirmatory unadjustedmeans were summarized for each individ-ual study and diabetes type and werecompared with the predictions from thefitted models. Data analysis was per-formed using SAS version 8.2 statisticalsoftware (SAS Institute, Cary, NC).

RESULTS — The median interval be-tween screening and randomizationranged from 9 to 42 days in the studiesanalyzed, with an overall median of 28days for patients with type 1 diabetes andof 14 days for patients with type 2 diabe-tes. The mean change in A1C in this in-terval was �0.13% for type 1 diabetes(range �0.09 to �0.26%) and �0.16%for type 2 diabetes (�0.14 to �0.27%).Among patients with an interval of at least28 days (median: 31 days for type 1 dia-betes, 28 days for type 2 diabetes), meanchanges in A1C were �0.24% (type 1 di-

abetes) and �0.23% (type 2 diabetes)(Table 1).

In both patient populations, the re-gression equations for change in A1Cdemonstrated significant effects for thetime interval between screening and ran-domization and the interaction of base-line A1C with this time interval. Thefollowing were the final models: A1Cchange � (�0.0061 base l ine �0.0418) � interval for type 1 diabetes andA1C change � (�0.0074 baseline �0.0640) � interval for type 2 diabetes.The proportion of variation explainedwas 11.6% in patients with type 1 diabe-tes and 11.7% in patients with type 2 di-abetes. In both groups, the observeddecline in A1C was directly proportionalto the screening visit value (P � 0.001 forthe effect of baseline A1C on the degree ofchange), such that those with the best ini-tial control showed little response tostudy recruitment, whereas those withworst control showed the greatest im-provement (Figs. 1 and 2). Predictionsfrom the fitted models indicated an over-all decline in A1C of 0.16% within 28days of study recruitment for type 1 dia-betes and 0.14% within 14 days for type 2diabetes (Table 1). The models show thatpatients with control around the 90thpercentile would be expected to improvetheir control within 45 days by 0.67%(type 1 diabetes) and 1.08% (type 2 dia-betes), respectively. The models alsoshow that no change in glycemic controlwould be expected in those who startaround the 10th percentile.

CONCLUSIONS — Many diabetesspecialists believe that there are nonspe-cific benefits of participation in a clinicaltrial. Possible explanations include morefrequent contact with the clinical team,

Table 1—A1C (%) change during the interval between screening and randomization

Study(reference)

Diabetestype n

Mean atscreening

Mean atrandomization

Mean � SDchange

Medianinterval

Predictedmean atmedianinterval

n with aninterval of�28 days

Meanchange

A 1 196 7.56 7.33 �0.24 � 0.72 28 7.44 155 �0.24B 1 93 7.78 7.52 �0.26 � 0.67 42 7.54 91 �0.27C 1 140 7.96 8.05 0.09 � 0.69) 21 7.81 40 �0.15

All type 1 429 7.74 7.60 �0.13 � 0.71 28 7.58 286 �0.24D 2 130 10.41 10.26 �0.15 � 0.76 21 10.14 42 �0.12E 2 117 9.94 9.67 �0.27 � 0.77 28 9.68 73 �0.29F 2 364 9.90 9.77 �0.14 � 0.45 9 9.82 2 *

All type 2 611 10.02 9.85 �0.16 � 0.60 14 9.88 115* �0.23*

*Study F not included since there are only two patients in this category.

Trial recruitment improves glycemic control

2990 DIABETES CARE, VOLUME 30, NUMBER 12, DECEMBER 2007

better information, and increased motiva-tion to do well. We set out to examine theeffect of recruitment to a controlled clin-ical trial upon glycemic control, as judgedby changes in A1C between the time ofthe first visit, in which patients arescreened for eligibility, and the subse-quent visit, in which those eligible arerandomized to therapy. A1C levels reflectmean glycemia over 4–6 weeks but aremost strongly influenced by the periodimmediately preceding the test (8). In ouranalysis, the typically brief interval be-tween screening and randomization wassufficient to register an effect, althoughthis was not fully expressed in the greatmajority of patients. Our analysis showedthat A1C fell in time-dependent fashionfrom the time of the screening visit, withan overall observed reduction of just un-der �0.25% in those waiting �28 daysfor randomization. The fall was propor-tional to initial A1C, and the observedchanges might well be judged of thera-peutic relevance when evaluating a newtreatment for diabetes.

How can these observations be ex-plained? One possibility is regression tothe mean (9), since all patients had ele-vated A1C at study entry. Patients with

type 1 diabetes were, however, selectedbecause of relatively good glucose con-trol, while those with type 2 diabetes wereselected for relatively poor control; there-fore, one would, on this argument, expectthe two groups to change in opposite di-rections. Hence, regression to the meancannot fully account for these observa-tions. A second possible explanation re-lates to the educational content of thescreening visit. This typically involves anexplanation concerning the nature andpurpose of the study, collection of per-sonal data from the patient, brief physicalexamination, collection of blood andurine samples, and collection of signedconsent to participation. Although proto-cols frequently indicate that patientsshould be given advice about “optimiza-tion of therapy,” the time available for thisis typically limited in practice. Since nochange in pharmaceutical therapy was of-fered at the time of the screening visit, anysubsequent benefit must be attributed to al-tered patient behavior, whether as a result ofspecific advice and education given in thecourse of the screening visit or simply be-cause of increased interest and motivation.The latter appears more plausible.

The classic example of the nonspe-

Figure 1—The change in A1C (%) (black dots) in type 1 diabetic patients (n � 429) between thescreening and randomization visits. The regression lines are derived from the following equation:A1C change � (�0.0061 � baseline A1C value � 0.0418) � duration in days. The change in A1Cwas inversely related to the value at the screening visit (P � 0.001). Predicted lines are shown forfive percentiles of baseline A1C: 10th percentile (P10), 25th percentile (P25), 50th percentile(P50), 75th percentile (P75), and 90th percentile (P90).

Figure 2—The change in A1C in type 2 diabetic patients (n � 611) between the screening andrandomization visits. The regression lines are derived from the following equation: A1C change �(�0.0074 � baseline A1C value � 0.0640) � duration in days. The change in A1C was inverselyrelated to the value at the screening visit (P � 0.001). Predicted lines are shown for five percentilesof baseline A1C: 10th percentile (P10), 25th percentile (P25), 50th percentile (P50), 75th per-centile (P75), and 90th percentile (P90).

Gale and Associates


cific effect of participation in research de-rives from a study performed in theHawthorne plant of the Western ElectricCompany in Chicago in 1924. Althoughthe original study reports were lost, andthe only contemporary record derivesfrom a few paragraphs in a trade journal,this study forms the basis of the fre-quently cited “Hawthorne effect.” Thisstems from the observation that peoplealter their behavior when they know thatit is being studied in ways that may influ-ence study outcomes. The original studyshowed that productivity on the factoryfloor increased in both test and controlgroups during experiments regardless ofwhether ambient lighting was adjustedupwards or downwards (10–12). Thereare several examples of a similar effect inthe medical literature. Medical residentsin a U.S. hospital participated in a trial oftwo methods designed to reduce the fre-quency with which they ordered labora-tory tests and X-rays: financial incentiveor chart discussion. One-third of the res-idents acted as control subjects. The chartreview group made a 47% reduction, thefinancial incentive group made a 29% re-duction, and the control group made a36% reduction (13). An example in thearea of diabetes relates to measurement ofblood glucose by patients. When thistechnique was introduced, it was as-sumed that patients who monitor theirblood would achieve better control thanthose who tested their urine. Contrary toexpectation, a controlled trial showedthat patients randomized to blood orurine tests did equally well when equaltime and attention was provided by theirhealth care providers (14).

In conclusion, our study shows thatrecruitment to a clinical trial involving asingle screening visit can produce a clini-cally useful improvement in glycemiccontrol, especially in those with relativelypoor control at study outset. This has im-plications for the conduct of clinical trials,which should allow for the likelihood thatA1C will already be falling in most partic-ipants by the time any other interventionis introduced. The conventional baselinemeasures, recorded at the time of treat-

ment allocation, are therefore not a stablepoint of reference, and the benefits of anyintervention are likely to be overesti-mated. This underscores the need for con-trolled comparisons, with a comparablelead-in period and an equal balance ofnonspecific interventions. The effect maybe of particular importance in the eval-uation of washout studies, since the be-havioral responses to stopping onemedication and starting another mayvary in opposite directions; the intervalbetween changes of therapy could alsoinfluence the outcome. Clinical trials ofpatients with diabetes should thereforereport A1C at screening, as well as atrandomization (to be used as baselineA1C value), and should detail the inter-val between the two samplings.

Finally, there are implications for pa-tient care. Good clinicians provide vari-ety, novelty, and constant encouragementfor their patients, who thus derive non-specific benefits that may outweigh theadvantages of any specific therapy theyreceive. This may help to explain whythere are so many conflicting, yet appar-ently effective, recipes for improved gly-cemic control. Voltaire once commentedthat the main function of doctors was tokeep the patient amused “while the dis-ease runs its inevitable course,” but thisanalysis has shown that keeping the pa-tient interested can be a very effectiveform of therapy.

References1. Ismail K, Winkley K, Rabe-Hesketh S:

Systematic review and meta-analysis ofrandomised controlled trials of psycho-logical interventions to improve glycae-mic control in patients with type 2diabetes. Lancet 363:1589–1597, 2004

2. Devries JH, Snoek FJ, Kostense PJ, HeineRJ: Improved glycaemic control in type 1diabetes patients following participationper se in a clinical trial: mechanisms andimplications. Diabetes Metab Res Rev 19:357–362, 2003

3. Holleman F, Schmitt H, Rottiers R, ReesA, Symanowski S, Anderson JH: Reducedfrequency of severe hypoglycemia andcoma in well-controlled IDDM patientstreated with insulin lispro: the Benelux-UK

Insulin Lispro Study Group. Diabetes Care20:1827–1832, 1997

4. Gale EAM: A randomised controlled trialcomparing insulin lispro with soluble in-sulin in patients with type 1 diabetes onintensified insulin therapy. Diabet Med17:209–214, 2000

5. Bastyr EJ 3rd, Stuart CA, Brodows RG,Schwartz S, Graf CJ, Zagar A, RobertsonKE: Therapy focused on lowering post-prandial glucose, not fasting glucose, maybe superior for lowering A1C: IOEZ StudyGroup. Diabetes Care 23:1236–1241,2000

6. Bastyr EJ III, Zagar A, Graf CJ, FletcherAL: Insulin lispro (LP) versus humulin70/30 following secondary oral agent fail-ure in primary care (Abstract). Diabetes 49(Suppl. 1):A97, 2000

7. Holcombe JH, ZagarAJ, Pinaire JA, PrinceMJ: Superiority of insulin lispro 75/25compared with NPH when used in com-bination with metformin (MET) (Ab-stract). Diabetes 51 (Suppl. 2):A101, 2002

8. Rohlfing CL, England JD, WiedmeyerH-M, Tennill A, Little RR, Goldstein DE:Defining the relationship between plasmaglucose and A1C. Diabetes Care 25:275–278, 2002

9. Barnett AG, van der Pols JC, Dobson AJ:Regression to the mean: what it is andhow to deal with it. Int J Epidemiol 34:215–220, 2005

10. Gillespie R: Manufacturing Knowledge: AHistory of the Hawthorne Experiments.Cambridge, U.K., Cambridge UniversityPress, 1991

11. Roethlisberger FJ, Dickson WJ: Manage-ment and the Worker. Cambridge, Massa-chusetts, Harvard University Press, 1939

12. Gale EAM: The Hawthorne studies: a fa-ble for our times? Quart J Med 97:439–449, 2004

13. Martin AR, Wolf MA, Thibodeau LA,Dzau V, Braunwald E: A trial of two strat-egies to modify the test-ordering behaviorof medical residents. N Engl J Med 303:1330–1336, 1980

14. Worth R, Home PD, Johnston DG, Ander-son J, Ashworth L, Burrin JM, Appleton D,Binder C, Alberti KGMM: Intensive atten-tion improves glycaemic control in insu-lin-dependent diabetes without furtheradvantage from home blood glucosemonitoring: results of a controlled trial. BrMed J 285:1233–1240, 1982

Trial recruitment improves glycemic control


Lipid, Glycemic, and Insulin Responses toMeals Rich in Saturated, cis-Monounsaturated, and Polyunsaturated(n-3 and n-6) Fatty Acids in Subjects WithType 2 DiabetesMEENA SHAH, PHD

1,2,3

BEVERLEY ADAMS-HUET, MS4,5

LINDA BRINKLEY, RD5

SCOTT M. GRUNDY, MD, PHD2,5

ABHIMANYU GARG, MD1,2,5

OBJECTIVE — The recommendations for dietary fats in patients with type 2 diabetes arebased largely on the impact of fatty acids on fasting serum lipid and glucose concentrations. Howfatty acids affect postprandial insulin, glucose, and triglyceride concentrations, however, re-mains unclear. The objective of this study was to study the effect of fatty acids on postprandialinsulin, glucose, and triglyceride responses.

RESEARCHDESIGNANDMETHODS — Test meals rich in palmitic acid, linoleic acid,oleic acid, and eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA) and containing1,000 kcal each were administered in a randomized crossover design to 11 type 2 diabeticsubjects. Serum insulin, glucose, and triglyceride concentrations were measured for 360 min. Allsubjects received an isoenergetic diet of constant composition throughout the study.

RESULTS — According to repeated-measures ANOVA, the insulin (P � 0.0002) but notglucose (P � 0.10) response was significantly different between meals. The insulin response waslower to meals rich in oleic acid or EPA and DHA than to meals rich in palmitic acid or linoleicacid (P � 0.01). The triglyceride response did not reach statistical significance (P � 0.06) buttended to be lower with EPA and DHA than with the other fatty acids. Similar trends were seenfor area under the curve (AUC) and incremental AUC for serum insulin and triglycerides, but thedifferences were not significant.

CONCLUSIONS — In comparison with palmitic acid and linoleic acid, oleic acid or EPA andDHA may modestly lower insulin response in patients with type 2 diabetes without deterioratingthe glucose response. EPA and DHA may also reduce the triglyceride response.

Diabetes Care 30:2993–2998, 2007

T he dietary recommendations forfatty acid intakes to manage dyslip-idemia and glycemia in patients

with type 2 diabetes are largely based on

the findings from studies on the impact offatty acids on fasting serum lipid and glu-cose concentrations (1). How the differ-ent types of fatty acids affect postprandial

lipid, glucose, and insulin concentra-tions, however, is not clearly understood.This information is important becausemost individuals in Western countries arein a postprandial state for most of the day(2,3). Postprandial triglyceride concen-trations are associated with cardiovascu-lar disease (CVD) (4,5), and this fact maybe even more relevant in patients withtype 2 diabetes, given that these individ-uals have higher postprandial triglycerideresponses than individuals without type 2diabetes (6,7), even when baseline tri-glyceride concentrations are normal (7).How the different types of fatty acids af-fect postprandial glucose and insulin re-sponse also needs to be further examinedespecially because poor glycemic controlis linked to diabetes complications in-cluding CVD, and hyperinsulinemia is arisk factor for CVD (1).

The acute effect of different types offats on postprandial insulin response insubjects with type 2 diabetes has been ex-amined in only two studies (8,9). Thesestudies compared meals rich in butter andolive oil and reported no difference in in-sulin levels. Studies in subjects withoutdiabetes showed either no difference ininsulin response to meals rich in satu-rated, monounsaturated, or polyunsatu-rated (n-3 or n-6) fatty acids (10–18) or ahigher insulin response to meals rich ineicosapentaenoic acid (EPA) and docosa-hexaenoic acid (DHA) or linoleic acidthan to a meal rich in beef fat (19). A pos-sible explanation in part for the results ofthe latter study (19) may be that the mealsrich in polyunsaturated fatty acids con-tained about 10% more carbohydrate, adeterminant of postprandial insulin re-sponse (3), than the meal rich in beef fat.

Three studies have looked at the acuteeffect of different fatty acids on postpran-dial glucose response in subjects withtype 2 diabetes and either showed no dif-ferences in glucose response to meals richin butter and olive oil (9) and to mealsrich in oleic acid and oleic acid plus EPAand DHA (20) or showed a lower glucose

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Division of Nutrition and Metabolic Diseases, University of Texas Southwestern Medical Centerat Dallas, Dallas, Texas; the 2Center for Human Nutrition, University of Texas Southwestern Medical Centerat Dallas, Dallas, Texas; the 3Department of Kinesiology, Texas Christian University, Fort Worth, Texas; the4Department of Clinical Sciences, University of Texas Southwestern Medical Center at Dallas, Dallas, Texas;and the 5Department of Internal Medicine, University of Texas Southwestern Medical Center at Dallas,Dallas, Texas.

Address correspondence and reprint requests to Abhimanyu Garg, MD, Professor and Chief, Division ofNutrition and Metabolic Diseases, Center for Human Nutrition, University of Texas Southwestern MedicalCenter, 5323 Harry Hines Blvd., Dallas, TX 75390. E-mail: [email protected].

Received for publication 29 May 2007 and accepted in revised form 29 August 2007.Published ahead of print at http://care.diabetesjournals.org on 5 September 2007. DOI: 10.2337/dc07-

1026. Clinical trial reg. no. NCT00479791, clinicaltrials.gov.Abbreviations: AUC, area under the curve; iAUC, incremental AUC; CVD, cardiovascular disease; DHA,

docosahexaenoic acid; EPA, eicosapentaenoic acid.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





response to a meal rich in butter com-pared with a meal rich in olive oil (8). Amajority of the studies in subjects withoutdiabetes showed no difference in glucoseresponse to meals rich in different fattyacids (10 –15,18,21). Gatti et al. (16),however, found a lower glucose responseto meals rich in olive oil than to meals richin corn oil or butter, and Lardinois et al.(19) found a higher glucose response to ameal rich in fish compared with a mealrich in corn oil or beef in individuals with-out diabetes.

The number of studies examining theacute effect of meals on postprandial tri-glyceride response in type 2 diabetic sub-jects is also limited. West et al. (20)reported a lower triglyceride response to ameal rich in both oleic acid and EPA andDHA compared with a meal rich in justoleic acid but only in subjects with hightriglyceride concentrations. This differ-ence may be due to decreased chylomi-cron production or secretion (22) andreduced synthesis of VLDLs (23) associ-ated with very long-chain n-3 fatty acids.Comparison of meals rich in butter andolive oil in subjects with type 2 diabetesrevealed either no difference (8) or ahigher triglyceride response to the mealrich in butter (9). The results from studiesin subjects without diabetes have alsobeen controversial, with some studiesshowing no difference in postprandial tri-glyceride response to different fatty acids(11–15,21,24–27) and others showingeither a lower response to meals rich inn-3 (17), n-6 (10), and n-9 (10,18) fattyacids or a higher response to n-6 (28,29)and n-9 (28,29) fatty acids comparedwith responses to saturated fatty acids.

Possible reasons for the conflicting re-sults could be the fact that the order inwhich the meals were tested was not ran-domized in a number of studies (13,14,19,24), and in several studies aconstant background diet was not pro-vided (10 –17,19 –21,24,27,28). It has

been reported that the fats and carbohy-drates in the daily diet may influence thepostprandial response to a meal (3,30). Inaddition, some researchers limited the en-ergy content of the test meal (300–500kcal) (16,19) and the time during whichthe response was measured (180 min)(16,19,26), and these limitations maymake it difficult to detect significant dif-ferences between the different test meals.The controversial results may also be dueto the use of butter as a source of saturatedfat (8–10,16,18,21,25). Nearly 15% ofthe saturated fat content in butter is ac-counted for by medium-chain fatty acidsknown to improve insulin sensitivity andglycemic control (31).

The above studies have a number ofdesign issues, which make it difficult toclearly interpret the results. Also, there isa paucity of data in patients with type 2diabetes. We addressed these issues inour study in which we compared thepostprandial triglyceride, glucose, and in-sulin response to meals rich in palmiticacid, oleic acid, linoleic acid, and EPA andDHA in subjects with type 2 diabetes. Thepostprandial response was measured for360 min, and the meals were adminis-tered in a randomized crossover design.Each test meal was designed to provide1,000 kcal, and the percent energy fromcarbohydrate, protein, and fat was heldconstant. In addition, the subjects werefed an isoenergetic diet of constant com-position throughout the study. We hy-pothesized that there will be no differencein postprandial insulin and glucose re-sponse to meals rich in different fatty ac-ids. The secondary hypothesis was thatthe postprandial triglyceride responsewill be lower to meals rich in very long-chain n-3 fatty acids compared with theother fatty acids.

RESEARCH DESIGN ANDMETHODS — Eleven men with type 2diabetes who had fasting blood glucose

concentrations �200 mg/dl and were notreceiving insulin therapy were studied atthe General Clinical Research Center ofthe University of Texas SouthwesternMedical Center at Dallas. The protocolwas approved by the University of TexasSouthwestern Institutional Review Board,and each participant gave informed con-sent. Mean � SD age was 54.6 � 12.2years, and mean BMI was 33.2 � 3.7 kg/m2. Six of the subjects were non-Hispanicwhites, three were African American, andone each was Asian and Hispanic. None ofthe subjects had thyroid, renal, or hepaticdisease, uncontrolled hypertension, ane-mia, or a history of ketosis.

All subjects received an isoenergeticdiet of constant composition throughoutthe study duration of 12–15 days. Thesubjects were instructed to maintain aconstant level of physical activity.

At intervals of 3–4 days, after an over-night fast, each subject consumed a mixedtest meal on four occasions in a random-ized manner. The type of fat in the testmeal varied on each occasion, and themeal was rich in palmitic acid, oleic acid,linoleic acid, or EPA and DHA.

Daily diet and test mealsDaily energy intake of the subjects wasestimated using the Harris Benedict equa-tion (32). The subjects received an isoen-ergetic background diet containing 15%of total energy as protein, 35% as fat, and50% as carbohydrate throughout thestudy period, which started 3–4 days be-fore the first test meal was evaluated. Thesubjects picked up their daily meals every3–4 days from the General Clinical Re-search Center metabolic kitchen. Energyintake was adjusted to maintain a con-stant body weight. Alcohol was not al-lowed during the entire study period.Coffee was limited to one serving of re-constituted freeze-dried coffee (2 g drycoffee) at breakfast, and tea was limited toone serving of reconstituted instant tea atlunch and one at dinner. Sugar-free softdrinks were allowed but only as a replace-ment for tea. No deviations from theabove guidelines were reported.

Each test meal provided 1,000 kcalwith 15% energy as protein, 35% as car-bohydrate, and 50% as fat. The test mealscontained farina, egg substitute, ham with5% fat, white bread, skim milk, orangejuice, and 51 g of test oil. The test mealsrich in palmitic acid, oleic acid, linoleicacid, and EPA and DHA were made usingpalm oil, olive oil, safflower oil, and

Table 1—Fatty acid composition of the various test oils

Fatty acids (%)

Saturated Cis-mono-unsaturated

n-9

Polyunsaturated

OthersMedium

chainLongchain n-6 n-3

Palm oil 0.1 51.3 38.9 9.6 0.2 0Olive oil 0 13.8 77.1 8.3 0.6 0.2Safflower oil 0 6.5 15.0 78.0 0 0.5Salmon oil 0 17.6 29.4 2.3 38.6 12.1

Postprandial response to fatty acids in diabetes


salmon oil, respectively. The fat contentof the four test oils is shown in Table 1.

Meal tolerance testThe meal tolerance test was conductedafter a 12-h overnight fast. An intrave-nous cannula was placed in a forearm veinfor blood sampling. After collection ofthree baseline blood samples at �30,�15, and 0 min, patients were asked toconsume the test meal in a 15-min period,and blood was collected every 30 min forthe next 360 min for measurement ofplasma glucose, insulin, and triglycerideconcentrations.

Biochemical analysisPlasma glucose concentrations were as-sayed by the glucose oxidase method(Beckman Glucose Analyzer; BeckmanInstruments, Fullerton, CA). Plasma insu-lin concentrations were measured using aradioimmunoassay kit (Linco Research,St. Louis, MO). Plasma triglycerides weremeasured enzymatically (Sigma Diagnos-tics, St. Louis, MO).

Statistical methodsA repeated-measures ANOVA model wasused to assess the effect of the differentmeals on plasma glucose, insulin, and tri-glyceride responses after log transforma-tions. The main effects and meal � timeinteraction effects were evaluated. Pair-wise contrasts were made by comparingthe least-square mean estimates, and Pvalues were adjusted for multiple com-parisons using the Bonferroni Holmmethod (33). Repeated-measuresANOVA was also used to compare the ef-fect of the meals following rank transfor-mation of the glucose, insulin, andtriglyceride response values after subtrac-tion of the respective baseline values.Peak response and peak time were com-pared across meals by a single-factor re-peated-measures ANOVA model.

Area under the curve (AUC) and in-cremental AUC (iAUC), i.e., the areaabove baseline, were calculated using thetrapezoidal rule. The respective naturallog values were compared by a single-factor repeated-measures ANOVA model.

Similar trends were seen even after weadjusted our analyses for treatment withlipid- or glucose-lowering medications orexcluded the four men who were receiv-ing one of these medications. All statisti-cal analyses were performed using SAS(version 9.13; SAS Institute, Cary, NC).

RESULTS — Body weight was main-tained throughout the study period withthe carefully controlled isoenergetic back-ground diet of constant composition. Ac-cording to the repeated-measuresANOVA test after log transformation, thepostprandial insulin response (Fig. 1A)was significantly different among the var-ious meals (P � 0.0002), whereas thepostprandial glucose (Fig. 1C) and tri-glyceride (Fig. 1E) responses did not dif-fer significantly by the type of meal given(P � 0.10 and P � 0.06, respectively).Post hoc analyses, adjusted for multiplecomparisons, showed that the insulin re-sponse was significantly higher to themeal rich in palmitic acid than to themeals rich in oleic acid (P � 0.002) orEPA and DHA (P � 0.002) and to themeal rich in linoleic acid than to the mealsrich in oleic acid (P � 0.007) or EPA andDHA (P � 0.006). There was no differ-ence in insulin response between mealsrich in oleic acid and EPA and DHA andbetween meals rich in palmitic acid andlinoleic acid. The meal � time interactionwas not significant, indicating that thepeak insulin concentration was reached ataround the same time across meals andthat the difference in the insulin responseto the meals was due to the difference inthe magnitude of the response. There wasno difference in peak time (P � 0.62),which was reached at �60 min for eachmeal. Peak insulin concentration was sig-nificantly different (P � 0.01) by meals(Fig. 1A). It was higher after the meal richin linoleic acid than after the meals rich inoleic acid (P � 0.04) or EPA and DHA(P � 0.02) and not different between themeals rich in oleic acid and EPA and DHAor between the meals rich in palmitic acidand the other fatty acids.

Repeated-measures ANOVA follow-ing rank transformation after subtractionof the baseline values showed a signifi-cantly different effect of the meals on thepostprandial triglyceride (P � 0.004) andinsulin (P � 0.006) responses but not onthe glucose (P � 0.58) response. Post hocanalyses, adjusted for multiple compari-sons, showed that the insulin responsewas higher to the meals rich in linoleicacid than to the meals rich in oleic acid(P � 0.05) or EPA and DHA (P � 0.02)and also higher to the meal rich inpalmitic acid compared with EPA andDHA (P � 0.05). The postprandial tri-glyceride response tended to be lower af-ter the meals rich in EPA and DHA thanafter the meals rich in other fatty acids,but the difference was significant only be-

tween EPA and DHA and oleic acid (P �0.003). Because the triglyceride responsewas delayed by 120 min, we conductedan additional analysis after excluding thepostprandial data obtained during thefirst 120 min and found that the responsewas significantly higher after the meal richin oleic acid than after the other meals.However, because the study was designedto look at the postprandial response for360 min and because the typical responselasts for �120 min, our focus will be onthe entire postprandial period.

AUC for insulin (Fig. 1B) was higherfor meals rich in palmitic acid or linoleicacid than for meals rich in EPA and DHAor oleic acid, and that for triglycerides(Fig. 1F) tended to be lower for the mealrich in EPA and DHA than for the othermeals, but the differences did not reachstatistical significance. AUC for glucose(Fig. 1D) did not differ by meals. Similarresults were seen for iAUC (data notshown).

CONCLUSIONS — We examinedthe effects of different fatty acids on post-prandial triglyceride, glucose, and insulinconcentrations in subjects with type 2 di-abetes. According to repeated-measuresANOVA, the insulin response was signif-icantly different by the type of fatty acidconsumed. It was significantly higher inresponse to the meals rich in palmitic acidor linoleic acid than to meals rich in oleicacid or EPA and DHA. A similar trend wasseen for AUC and iAUC, but the differ-ences did not reach statistical significancepossibly because of the small sample size.These results contradict most of the ear-lier studies that showed no difference ininsulin response to meals rich in differenttypes of fatty acids in subjects with (8,9)or without (10–18) diabetes.

A possible mechanism for the insulinresponse observed in our study is the dif-ferent insulinotropic potency of the dif-ferent fatty acids. Stein et al. (34) studiedthe influence of fatty acids on insulin se-cretion in the perfused rat pancreas andfound that glucose-stimulated insulin re-lease was higher with palmitic acid thanwith oleic acid, which in turn was higherthan that with linoleic acid. Although wealso found an increased postprandial in-sulin response with palmitic acid com-pared with oleic acid, our observation ofhigher insulin response to linoleic acidthan to oleic acid is not consistent withthe data of Stein et al. (34). Holness et al.(35) reported that acute replacement ofsome dietary saturated fatty acids with

Shah and Associates


EPA and DHA, in rats made insulin resis-tant by high–saturated fat feeding for 4weeks, reversed insulin hypersecretion invivo and during glucose perifusion of iso-lated islets. The lowered insulin secretion,however, was not accompanied by im-

proved insulin action, and glucose toler-ance was adversely affected. In our study,the insulin-lowering effect of EPA andDHA and oleic acid was not associatedwith deterioration in glucose tolerance, asindicated by the lack of difference in post-

prandial glucose response to the differentfatty acids, and may suggest an increase ininsulin sensitivity. In our study, we didnot observe an improvement in insulinsensitivity after the meal rich in linoleicacid compared with that after the meal

Figure 1— Postprandial insulin, glucose, and triglyceride responses to meals. Shown are median values for postprandial insulin (A), glucose (C),and triglyceride (E) responses to meals rich in palmitic acid (�), oleic acid (�), linoleic acid (‚), and EPA and DHA (E). The baseline values arethe means of the values collected at �30, �15, and 0 min. To convert insulin values from picomoles per liter to microunits per milliliter, divide by6.0; to convert glucose and triglyceride values from millimoles per liter to milligrams per deciliter, divide by 0.05551 and 0.01129, respectively. Alsoshown are AUC values (medians and 25th and 75th percentiles) for insulin (B), glucose (D), and triglycerides (F) for different meals.



rich in palmitic acid. This finding con-flicts with data from a 5-week study (36)in which improved insulin sensitivity wasreported with a diet rich in linoleic acidcompared with a diet rich in palmiticacid. Energy intake in the latter study,however, was lower during the linoleicacid–rich phase compared with the otherdiet phase, which may partly explain theresults.

The difference in insulin response inour study may be due to secretion of in-cretin hormones, glucagon-like-peptide-1,and glucose-dependent insulinotropicpolypeptide. However, how the incretinhormones respond to different fatty acidsremains to be studied.

Our results on glucose response aresimilar to the results from most studies insubjects with (9,20) and without (10–15,18,21) type 2 diabetes, which showedno difference in glucose response to dif-ferent fatty acids. It is important to notethat although the insulin-lowering effectsof EPA and DHA and oleic acid did notadversely affect glucose response, the lat-ter remained in the diabetes range. Thisresult indicates a need for more aggressivecontrol of postprandial glucose levels us-ing several treatment strategies.

According to the repeated-measuresANOVA with rank-transformed values af-ter subtraction of the baseline values, thetriglyceride response tended to be lowerto the meal rich in EPA and DHA than tothe other meals, but the difference wasonly significant between the meals rich inEPA and DHA and oleic acid. A similar,albeit not significant, trend was seen forAUC or iAUC. These results are corrobo-rated by studies examining the acuteeffect of different fatty acids on postpran-dial triglyceride concentrations in type 2diabetic subjects with high baseline tri-glyceride concentrations (20) and in indi-viduals without diabetes (17). Otheracute studies in healthy subjects (15,24),however, showed no difference in post-prandial triglyceride response after mealsrich in oleic acid or EPA and DHA. Ourdata are similar to those from a long-termintervention study (37), which showedreduced postprandial triglyceride con-centrations in healthy subjects when dietsrich in saturated fatty acids or monoun-saturated fatty acids were supplementedwith fish oil. The composition of the testmeals was similar to that of the diets towhich the subjects were assigned (37).

It has been reported that the post-prandial triglyceride response may de-pend on insulin status (38). We looked at

the influence of insulin resistance, esti-mated using the homeostasis model assess-ment insulin resistance calculator 2.2(39), on triglyceride responseby repeated-measures ANOVA and found no evidenceof an interaction between insulin resis-tance status and triglyceride response tomeals. Nevertheless, a larger sample mayhelp to better examine this relationship.

Possible mechanisms by which EPAand DHA lower triglyceride levels includedecreased chylomicron production or se-cretion (22) and reduced synthesis of he-patic VLDLs (23) seen after chronicfeeding of EPA and DHA. The reducedVLDL concentrations would result in lesscompetition for lipoprotein lipase for hy-drolysis of chylomicrons. It is not knownwhether acute consumption of EPA andDHA would result in the above mecha-nisms, however.

To accurately distinguish the effect ofdifferent types of fatty acids on the post-prandial responses, we tested meals thatcontained 1,000 kcal and 50% of energyas fat. Although these meals are more en-ergy dense than the diet that is typicallyconsumed by Americans (3,40), we be-lieve that the test meal rich in oleic acidmay be acceptable over the long termbased on the strong adherence that wehave observed in our earlier studies (3,41)testing high monounsaturated fat diets for6–12 weeks. Whether large doses of fishoil are acceptable over the long term re-mains to be studied. Also, whether mealswith a lower fat content would lead toreduced or similar postprandial responsescompared with the meals administered inour study will require additional studies.Our test meals also contained some car-bohydrate sources such as white bread,which has a high glycemic index. Thisshould not preclude us from distinguish-ing the effect of the different fatty acids onpostprandial response, however, becausethe type and amount of carbohydratewere held constant across the test meals.

In summary, our study shows thatmeals containing a high percentage of en-ergy from oleic acid or EPA and DHA, thefatty acids that patients with type 2 diabe-tes are encouraged to consume by theAmerican Diabetes Association (1), maybe beneficial in lowering postprandial in-sulin response in comparison with mealsrich in palmitic acid or linoleic acid with acomparable postprandial glucose re-sponse. Meals containing a high percent-age of energy from EPA and DHA mayalso be beneficial in lowering the post-prandial triglyceride response.

Acknowledgments— This study was fundedby General Clinical Research Center U.S. Pub-lic Health Service Grant M01-RR00633 and bythe Southwestern Medical Foundation.

Parts of this study were presented in ab-stract form at the 48th annual meeting of theAmerican College of Nutrition, Orlando, Flor-ida, 27–30 September 2007.

References1. Bantle JP, Wylie-Rosett J, Albright AL,

Apovian CM, Clark NG, Franz MJ, Hoog-werf BJ, Lichtenstein AH, Mayer-Davis E,Mooradian AD, Wheeler ML: Nutritionrecommendations and interventions fordiabetes—2006: a position statement ofthe American Diabetes Association. Dia-betes Care 29:2140–2157, 2006

2. Anderson JW, O’Neal DS, Riddell-MasonS, Floore TL, Dillon DW, Oeltgen PR:Postprandial serum glucose, insulin, andlipoprotein responses to high- and low-fiber diets. Metabolism 44:848–854, 1995

3. Garg A, Bantle JP, Henry RR, CoulstonAM, Griver KA, Raatz SK, Brinkley L,Chen YD, Grundy SM, Huet BA, et al.:Effects of varying carbohydrate content ofdiet in patients with non-insulin-depen-dent diabetes mellitus. JAMA 271:1421–1428, 1994

4. Simons LA, Dwyer T, Simons J, BernsteinL, Mock P, Poonia NS, BalasubramaniamS, Baron D, Branson J, Morgan J, et al.:Chylomicrons and chylomicron remnantsin coronary artery disease: a case-controlstudy. Atherosclerosis 65:181–189, 1987

5. Bansal S, Buring JE, Rifai N, Mora S, SacksFM, Ridker PM: Fasting compared withnonfasting triglycerides and risk of car-diovascular events in women. JAMA 298:309–316, 2007

6. Curtin A, Deegan P, Owens D, Collins P,Johnson A, Tomkin GH: Elevated triglyc-eride-rich lipoproteins in diabetes: astudy of apolipoprotein B-48. Acta Diabe-tol 33:205–210, 1996

7. Rivellese AA, De Natale C, Di Marino L,Patti L, Iovine C, Coppola S, Del Prato S,Riccardi G, Annuzzi G: Exogenous andendogenous postprandial lipid abnormal-ities in type 2 diabetic patients with opti-mal blood glucose control and optimalfasting triglyceride levels. J Clin EndocrinolMetab 89:2153–2159, 2004

8. Rasmussen O, Lauszus FF, ChristiansenC, Thomsen C, Hermansen K: Differentialeffects of saturated and monounsaturatedfat on blood glucose and insulin re-sponses in subjects with non-insulin-de-pendent diabetes mellitus. Am J Clin Nutr63:249–253, 1996

9. Thomsen C, Storm H, Holst JJ, Her-mansen K: Differential effects of saturatedand monounsaturated fats on postpran-dial lipemia and glucagon-like peptide 1responses in patients with type 2 diabetes.

Shah and Associates


Am J Clin Nutr 77:605–611, 200310. Mekki N, Charbonnier M, Borel P, Leo-

nardi J, Juhel C, Portugal H, Lairon D:Butter differs from olive oil and sunfloweroil in its effects on postprandial lipemiaand triacylglycerol-rich lipoproteins aftersingle mixed meals in healthy young men.J Nutr 132:3642–3649, 2002

11. Roche HM, Zampelas A, Jackson KG, Wil-liams CM, Gibney MJ: The effect of testmeal monounsaturated fatty acid: satu-rated fatty acid ratio on postprandial lipidmetabolism. Br J Nutr 79:419–424, 1998

12. Cortes B, Nunez I, Cofan M, Gilabert R,Perez-Heras A, Casals E, Deulofeu R, RosE: Acute effects of high-fat meals enrichedwith walnuts or olive oil on postprandialendothelial function. J Am Coll Cardiol 48:1666–1671, 2006

13. Raitakari OT, Lai N, Griffiths K, McCredieR, Sullivan D, Celermajer DS: Enhancedperipheral vasodilation in humans after afatty meal. J Am Coll Cardiol 36:417–422,2000

14. Blum S, Aviram M, Ben-Amotz A, Levy Y:Effect of a Mediterranean meal on post-prandial carotenoids, paraoxonase activ-ity and C-reactive protein levels. Ann NutrMetab 50:20–24, 2006

15. Burdge GC, Powell J, Calder PC: Lack ofeffect of meal fatty acid composition onpostprandial lipid, glucose and insulin re-sponses in men and women aged 50–65years consuming their habitual diets. Br JNutr 96:489–500, 2006

16. Gatti E, Noe D, Pazzucconi F, Gianfrance-schi G, Porrini M, Testolin G, Sirtori CR:Differential effect of unsaturated oils andbutter on blood glucose and insulin re-sponse to carbohydrate in normal volun-teers. Eur J Clin Nutr 46:161–166, 1992

17. Zampelas A, Murphy M, Morgan LM,Williams CM: Postprandial lipoproteinlipase, insulin and gastric inhibitorypolypeptide responses to test meals of dif-ferent fatty acid composition: comparisonof saturated, n-6 and n-3 polyunsaturatedfatty acids. Eur J Clin Nutr 48:849–858,1994

18. Thomsen C, Rasmussen O, Lousen T,Holst JJ, Fenselau S, Schrezenmeir J, Her-mansen K: Differential effects of saturatedand monounsaturated fatty acids on post-prandial lipemia and incretin responses inhealthy subjects. Am J Clin Nutr 69:1135–1143, 1999

19. Lardinois CK, Starich GH, Mazzaferri EL,DeLett A: Polyunsaturated fatty acids aug-ment insulin secretion. J Am Coll Nutr6:507–515, 1987

20. West SG, Hecker KD, Mustad VA, Nichol-son S, Schoemer SL, Wagner P, Hinder-liter AL, Ulbrecht J, Ruey P, Kris-Etherton

PM: Acute effects of monounsaturatedfatty acids with and without omega-3 fattyacids on vascular reactivity in individualswith type 2 diabetes. Diabetologia 48:113–122, 2005

21. Freese R, Mutanen M: Postprandialchanges in platelet function and coagula-tion factors after high-fat meals with dif-ferent fatty acid compositions. Eur J ClinNutr 49:658–664, 1995

22. Harris WS, Muzio F: Fish oil reducespostprandial triglyceride concentrationswithout accelerating lipid-emulsion re-moval rates. Am J Clin Nutr 58:68–74,1993

23. Nestel PJ, Connor WE, Reardon MF, Con-nor S, Wong S, Boston R: Suppression bydiets rich in fish oil of very low densitylipoprotein production in man. J Clin In-vest 74:82–89, 1984

24. Harris WS, Connor WE, Alam N, Illing-worth DR: Reduction of postprandial tri-glyceridemia in humans by dietary n-3fatty acids. J Lipid Res 29:1451–1460,1988

25. Bellido C, Lopez-Miranda J, Blanco-ColioLM, Perez-Martinez P, Muriana FJ, Mar-tin-Ventura JL, Marin C, Gomez P, Fu-entes F, Egido J, Perez-Jimenez F: Butterand walnuts, but not olive oil, elicit post-prandial activation of nuclear transcrip-tion factor �B in peripheral bloodmononuclear cells from healthy men.Am J Clin Nutr 80:1487–1491, 2004

26. Rueda-Clausen CF, Silva FA, LindarteMA, Villa-Roel C, Gomez E, Gutierrez R,Cure-Cure C, Lopez-Jaramillo P: Olive,soybean and palm oils intake have a sim-ilar acute detrimental effect over the en-dothelial function in healthy youngsubjects. Nutr Metab Cardiovasc Dis 17:50–57, 2007

27. Salomaa V, Rasi V, Pekkanen J, JauhiainenM, Vahtera E, Pietinen P, Korhonen H,Kuulasmaa K, Ehnholm C: The effects ofsaturated fat and n-6 polyunsaturated faton postprandial lipemia and hemostaticactivity. Atherosclerosis 103:1–11, 1993

28. Gradek WQ, Harris MT, Yahia N, DavisWW, Le NA, Brown WV: Polyunsaturatedfatty acids acutely suppress antibodies tomalondialdehyde-modified lipoproteinsin patients with vascular disease. Am JCardiol 93:881–885, 2004

29. Tholstrup T, Sandstrom B, Bysted A,Holmer G: Effect of 6 dietary fatty acidson the postprandial lipid profile, plasmafatty acids, lipoprotein lipase, and choles-terol ester transfer activities in healthyyoung men. Am J Clin Nutr 73:198–208,2001

30. Weintraub MS, Zechner R, Brown A,Eisenberg S, Breslow JL: Dietary polyun-

saturated fats of the W-6 and W-3 seriesreduce postprandial lipoprotein levels.Chronic and acute effects of fat saturationon postprandial lipoprotein metabolism.J Clin Invest 82:1884–1893, 1988

31. Eckel RH, Hanson AS, Chen AY, BermanJN, Yost TJ, Brass EP: Dietary substitutionof medium-chain triglycerides improvesinsulin-mediated glucose metabolism inNIDDM subjects. Diabetes 41:641–647,1992

32. Harris J, Benedict G: A Biometric Study ofBasal Metabolism in Man. Washington,DC, Carnegie Institutes of Washington,1919 (publ. no. 279)

33. Holm S: A simple sequentially rejectivemultiple test procedure. Scand J Stat 6:65–70, 1979

34. Stein DT, Stevenson BE, Chester MW, Ba-sit M, Daniels MB, Turley SD, McGarryJD: The insulinotropic potency of fatty ac-ids is influenced profoundly by theirchain length and degree of saturation.J Clin Invest 100:398–403, 1997

35. Holness MJ, Smith ND, Greenwood GK,Sugden MC: Acute -3 fatty acid enrich-ment selectively reverses high–saturatedfat feeding–induced insulin hypersecre-tion but does not improve peripheral in-sulin resistance. Diabetes 53 (Suppl. 1):S166–S171, 2004

36. Summers LK, Fielding BA, Bradshaw HA,Ilic V, Beysen C, Clark ML, Moore NR,Frayn KN: Substituting dietary saturatedfat with polyunsaturated fat changes ab-dominal fat distribution and improves in-sulin sensitivity. Diabetologia 45:369–377, 2002

37. Rivellese AA, Maffettone A, Vessby B,Uusitupa M, Hermansen K, Berglund L,Louheranta A, Meyer BJ, Riccardi G: Ef-fects of dietary saturated, monounsatu-rated and n-3 fatty acids on fastinglipoproteins, LDL size and post-prandiallipid metabolism in healthy subjects. Ath-erosclerosis 167:149–158, 2003

38. Wu CJ, Yu ZR: Effects on blood glucose,insulin, lipid and proatherosclerotic pa-rameters in stable type 2 diabetic subjectsduring an oral fat challenge. Lipids HealthDis 3:17, 2004

39. Wallace TM, Levy JC, Matthews DR: Useand abuse of HOMA modeling. DiabetesCare 27:1487–1495, 2004

40. Briefel RR, Johnson CL: Secular trends indietary intake in the United States. AnnuRev Nutr 24:401–431, 2004

41. Garg A, Bonanome A, Grundy SM, ZhangZJ, Unger RH: Comparison of a high-carbohydrate diet with a high-monoun-saturated-fat diet in patients with non-insulin-dependent diabetes mellitus.N Engl J Med 319:829–834, 1988



Burden of Comorbid Medical Conditionsand Quality of Diabetes CareJEWELL H. HALANYCH, MD, MSC

1,2

MONIKA M. SAFFORD, MD1,2

WENDY C. KEYS, MPH1

SHARINA D. PERSON, PHD1

JAMES M. SHIKANY DRPH1

YOUNG-IL KIM, PHD1

ROBERT M. CENTOR, MD2,3

JEROAN J. ALLISON, MD, MSC2,3

OBJECTIVE — With performance-based reimbursement pressures, it is concerning thatmost performance measurements treat each condition in isolation, ignoring the complexitiesof patients with multiple comorbidities. We sought to examine the relationship betweencomorbidity and commonly assessed services for diabetic patients in a managed careorganization.

RESEARCH DESIGN AND METHODS — In 6,032 diabetic patients, we determinedthe association between the independent variable medical comorbidity, measured by theCharlson Comorbidity Index (CCI), and the dependent variables A1C testing, lipid testing,dilated eye exam, and urinary microalbumin testing. We calculated predicted probabilitiesof receiving tests for patients with increasing comorbid illnesses, adjusting for patientdemographics.

RESULTS — A1C and lipid testing decreased slightly at higher CCI: predicted probabilitiesfor CCI quartiles 1, 2, 3, and 4 were 0.83 (95% CI 0.70–0.91), 0.83 (0.69–0.92), 0.82 (0.68–0.91), and 0.78 (0.61–0.88) for A1C, respectively, and 0.82 (0.69–0.91), 0.81(0.67–0.90),0.79 (0.64–0.89), and 0.77 (0.61–0.88) for lipids. Dilated eye exam and urinary microalbumintesting did not differ across CCI quartiles: for quartiles 1, 2, 3, and 4, predicted probabilities were0.48 (0.33–0.63), 0.54 (0.38–0.69), 0.50 (0.34–0.65), and 0.50 (0.34–0.65) for eye exam,respectively, and 0.23 (0.12–0.40), 0.24 (0.12–0.42), 0.24 (0.12–0.41), and 23 (0.11–0.40)for urinary microalbumin.

CONCLUSIONS — Services received did not differ based on comorbid illness burden. Be-cause it is not clear whether equally aggressive care confers equal benefits to patients with varyingcomorbid illness burden, more evidence confirming such benefits may be warranted beforewidespread implementation of pay-for-performance programs using currently available “onesize fits all” performance measures.

Diabetes Care 30:2999–3004, 2007

D elivering high-quality medical careis a major focus in today’s healthcare market. To achieve the desired

gains in quality, performance measuresrooted in guideline-recommended carehave been widely implemented and arebeing publicly reported (1). Accumulat-ing reports suggest that these practices are

having measurable effects, but progressmay not be sufficiently rapid (2). Thiscommitment to quality has spawned anew direction in accountability, withclear movement toward tightening thelink between reimbursements and high-quality care (3).

The growing enthusiasm for pay-for-

performance (P4P) programs may alsousher in a new set of problems. Most per-formance measures focus on the quality ofcare provided for a single disease (4). Yet,as the U.S. population ages, the numberof patients with a high burden of chronicmedical conditions is increasing. In 1999,48% of Medicare enrollees aged �65years had at least three chronic medicalconditions, and 21% had five or more (5).Patients having multiple conditions cre-ate considerable management complex-ity, forcing the clinician to consider andprioritize a large array of recommendedinterventions and preventive services.Market forces may encourage physiciansto “play to the test” (6), possibly replacingvaluable time in the office visit that couldbe spent addressing issues that have agreater impact on quality of life. Ulti-mately, how we should adjust perfor-mance measurement to reflect thiscomplexity presents a major challenge.

The forces of quality measurement,performance-based reimbursement, andmultiple comorbidities dramatically con-verge for patients who have diabetes,which affects 20.8 million Americans (7).Many patients with diabetes are older andhave several other medical problems,forcing the busy clinician to balance therelative benefits of multiple competingrecommendations (7,8). Some of theserecommendations, such as closely moni-toring and controlling blood glucose orblood lipids, take years to deliver benefitsin terms of risk reduction (9–11). Formany older patients with multiple comor-bid illnesses and limited life expectancy,the benefits of routinely recommendeddisease monitoring may not be as great asthose for younger patients. The extent towhich clinicians forego routinely recom-mended screening practices because ofillness burden and life expectancy is notknown.

We examined the relationship be-tween illness burden and receipt of diabe-tes services in a population of olderpatients with diabetes enrolled in a Medi-care managed care health plan, whichspans three Southern states. We assessedillness burden with a commonly usedindex of comorbid illness burden thatpredicts mortality, the Charlson Co-morbidity Index (CCI). We studied four

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Division of Preventive Medicine, Department of Medicine, University of Alabama at Birmingham,Birmingham, Alabama; the 2Deep South Center on Effectiveness at the Birmingham Veterans Affairs MedicalCenter, Birmingham, Alabama; and the 3Division of General Internal Medicine, Department of Medicine,University of Alabama at Birmingham, Birmingham, Alabama.

Address correspondence and reprint requests to Jewell H. Halanych, MD, MT 639, 1530 3rd Ave. South,Birmingham, AL 35294-4410. E-mail: [email protected].

Received for publication 31 August 2006 and accepted in revised form 19 August 2007.Published ahead of print at http://care.diabetesjournals.org on 23 August 2007. DOI: 10.2337/dc06-

1836.Abbreviations: CCI, Charlson Comorbidity Index; P4P, pay for performance.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





commonly assessed services for diabeticpatients: A1C, lipid, and urinary mi-croalbumin testing and dilated eyeexam. We hypothesized that testing fre-quency would diminish as comorbid ill-ness burden rose.

RESEARCH DESIGN ANDMETHODS — Study participants wereenrolled in a Medicare managed carehealth insurance plan providing coveragein Alabama, Florida, and North Carolina.We included patients who were aged atleast 65 years, enrolled with the healthplan continuously from 1 Januarythrough 31 December 2003, and alive on31 December 2003. Eligible patients hadone of the following: 1) at least one phar-macy claim for diabetes medication or 2)an ICD-9 code for diabetes from an in-person visit with a clinician in the outpa-tient or inpatient setting. We also mergeddata from the Center for Medicare andMedicaid Services with the health plan’sclaims and pharmacy data, allowing us toascertain race/ethnicity. Only patientsidentified as African American or Euro-pean American were included in our anal-ysis. The Western Institutional ReviewBoard approved this study. All analyseswere conducted using SAS (version 9.1;SAS Institute, Cary, NC).

Comorbid conditions were identifiedusing ICD-9 codes from encounters in thehealth insurance claims data. We usedRomano’s modification of the CCI (12).The CCI includes 17 specific illnesses andweights them according to severity (1, 2,3, or 6). The CCI for each patient is cal-culated by summing the weighted num-ber of points for each diagnosis carried bythe patient. CCI (range 0 –15) reflectsmortality risk at 1 and 10 years, with lowscores representing lowest risk (13).

For our main analyses, we used fourseparate commonly recommended per-formance measures for diabetes (14). Weused data collected as part of the managedcare organization’s reporting activities tothe National Committee on Quality As-surance Health Plan Employer Data andInformation Set. Based on claims data, wedetermined for the calendar year 2003whether patients had A1C testing, dilatedeye exam, and urinary microalbumin test-ing in the past year and lipid testing in thepast 2 years. Because screening for mi-croalbumin is not clinically indicatedonce a patient has renal disease, we ex-cluded the 780 participants with knownrenal disease from the urinary microalbu-min analyses.

Initially, we examined performancefor each of these measures at every level ofthe CCI: 0–1, 2, 3, 4, 5, 6, 7, 8, 9, 10, and11�. Because only eight patients had aCCI of 0, we constructed a single categoryfor patients with 0 and 1 comorbidity.Likewise, because of small numbers, wecombined all patients with comorbidityindexes �11 into a single category. Be-cause the data were not normally distrib-uted, we also classified CCI by quartile,which represented a clinically reasonableapproach. For multivariable modeling,the main independent variable was quar-tile of CCI.

Using the proc logistic command inSAS, we developed separate multivariablelogistic regression models for each of thefour dependent variables, adjusting forage, sex, race/ethnicity, and diabetesmedications. Diabetes medications weredefined by presence or absence of phar-macy claims for either oral hypoglycemicmedications or insulin. These covariatesrepresented important confounders of therelationship between CCI and the out-comes. All patients were included in themodels testing the associations betweenCCI and A1C and lipid testing. As in theunadjusted analyses, for the microalbu-min screening model we excluded the780 individuals with known renal dis-ease, for whom microalbuminuria screen-ing is no longer indicated. All patientswere included in the dilated eye exammodel, because the presence or absence ofdiabetic retinopathy does not change therecommendation for at least annual di-lated eye exams. However, to account fordifferential treatment in patients with rec-ognized diabetic retinopathy, we in-cluded a variable that reflected presenceor absence of diabetic retinopathy in thedilated eye exam model.

Next, we calculated unadjusted andadjusted predicted probabilities for eachmodel. More specifically, for each of thefour models, we calculated the averageprobability of the outcome having a pos-itive response within each comorbidityquartile (15). For the adjusted predictedprobabilities, we entered the mean valuesof all remaining covariates from each co-morbidity quartile. Based on the individ-ual variance estimates for each modelparameter, we calculated an overall SD forthe predicted logit of the outcome. Fi-nally, we transformed each predicted logitalong with the upper and lower boundsfor the 95% CIs into probabilities. To testfor bias from the possible misidentifica-tion of patients with diabetes based on

ICD-9 codes, we repeated all analyses onthe subset of patients taking hypoglyce-mic medications (n � 2,472) (16).

RESULTS — For the 6,032 patients in-cluded in the study, the average age was74.5 years; 56.8% were female, and39.8% were African American (Table 1).The subset of patients without renal fail-ure, which made up the study populationfor the urinary microalbumin analyses(n � 5,252), was similar demographicallyto the overall cohort (data not shown).CCI ranged from 0 to 15, with a mean �SD of 3.2 � 2.2. The mean age and theproportion of patients with diabetes com-plications increased as CCI increased. Fig.1 shows the unadjusted percentage of pa-tients who received each diabetes serviceby category of CCI. A1C and lipid testingwere somewhat lower in patients withhigher CCI. Rates of dilated eye exam andurinary microalbumin were similar acrossCCI.

In the multivariable analysis, resultswere similar after adjustment for age, sex,race/ethnicity, and receipt of diabetesmedications. The models showed littledifference in predicted probability of test-ing as comorbid disease burden increased(Table 2). Repeating all analyses for the2,472 patients with pharmacy claims fororal diabetes medications or insulin pro-vided results of similar direction andmagnitude (data not shown).

CONCLUSIONS — Contrary to ourexpectations, we found few differences inroutine testing by comorbidity burden forthis elderly population with diabetes en-rolled in a Medicare managed care healthplan. While rates of A1C and lipid testingwere slightly lower in patients with highercomorbidity than in those with the leastcomorbidity, rates of dilated eye examand urinary microalbumin testing did notdiffer by comorbid burden. These resultssuggest that physicians are not adjust-ing the provision of routine diabetesservices for patients with varying levelsof comorbidity, although our studydoes not inform the appropriateness ofthis observation.

In acute settings, tight glycemic con-trol has been shown to improve outcomes(17,18). But in chronic care management,tight glycemic or lipid control yields clin-ically meaningful benefits only after sev-eral years of intervention (9–11). Distinctfrom lipid and glucose control, detectionof proliferative diabetic retinopathy andtreatment with laser photocoagulation

Comorbid conditions and quality of care


have been shown to significantly reducesevere visual loss over as short a time pe-riod as 2 years (19). Therefore, dilated eyeexams should be performed in all patientswith diabetes; yet only half of patients inthis managed care organization, regard-

less of comorbid burden, received dilatedeye exams.

Patients with multiple comorbid con-ditions have often been excluded fromthe evidence-generating randomizedcontrolled trials that form the basis for

performance measures (4,20 –23). Ran-domized clinical trials, evidence-basedguidelines, and quality measures are re-markably silent on the inevitable trade-off decisions that must be made duringthe routine clinical care of medically

Figure 1—Unadjusted percentage of patients with diabetes receiving testing by CCI quartile, 2003. Urinary microalbumin excludes patients withrenal disease (n � 780). Q1–Q4: first quartile (lowest comorbidity) through fourth quartile (highest comorbidity) of CCI.

Table 1—Characteristics of patients with diabetes from a managed care organization by CCI* quartile, 2003

Total1st quartile(CCI 0–1)

2nd quartile(CCI 2)

3rd quartile(CCI 3–4)

4th Quartile(CCI �5)

n (%) 6,032 1,514 (25.1) 1,356 (22.5) 1,792 (29.7) 1,370 (22.7)Mean age (years) 74.5 � 6.2 73.3 � 5.6 74.2 � 6.1 74.9 � 6.2 75.7 � 6.6Female 3,428 (56.8) 912 (60.2) 843 (62.2) 965 (53.9) 708 (51.7)Race/ethnicity

African American 2,403 (39.8) 584 (38.6) 516 (38.1) 718 (40.1) 585 (42.7)European American 3,629 (60.1) 930 (61.4) 840 (62.0) 1,074 (60.0) 785 (57.3)

Diabetes complications†Retinopathy 863 (14.3) 99 (6.5) 199 (14.7) 274 (15.3) 291(21.2)Neuropathy 785 (13.0) 0 151 (11.1) 287 (16.0) 347 (25.3)Nephropathy 264 (4.4) 0 23 (1.7) 55 (3.1) 186 (13.6)

Data are means � SD or n (%). *CCI includes 17 specific illnesses, weights them according to severity (1, 2, 3, or 6), and sums the weighted conditions into an indexthat reflects risk for 1- and 10-year mortality. †ICD-9 physician claims.

Halanych and Associates


complex, older patients. In the absenceof evidence regarding the benefit orharm of applying these guidelines to pa-tients with multiple comorbid condi-tions, we are unable to definitivelydetermine whether our finding of al-most constant testing across comorbid-ity quartiles represents appropriate orinappropriate care.

In the P4P era, physicians may feelpressure to adopt a “one size fits all” ap-proach and order tests to improve theirperformance on quality indicators devel-oped from trials that excluded patientswith multiple comorbid conditions (24).In patients with limited life expectancy,the appropriate clinical course may be todecrease testing associated with delayedbenefit and to focus on interventions withhigh short-term potential for improvingquality of life. Until evidence is availablethat confirms current performance mea-sures as appropriate in patients with mul-tiple comorbidities or advanced age, P4Pprograms and managed care organiza-tions may wish to limit accountability toonly those populations for whom strongevidence of benefit exists.

Currently, there is only limited evi-dence that the health care–quality indus-try acknowledges life expectancy. Forexample, the Veterans Health Adminis-tration/Department of Defense ClinicalPractice Guideline for the Management ofDiabetes includes the recommendation toconsider life expectancy in setting treat-ment targets. Similarly, the American Ge-riatric Society and the CaliforniaHealthcare Foundation have endorsedguidelines that recognize the complexityof managing older patients with diabetesand multiple comorbid conditions (8).However, these acknowledgments havenot been translated into measures de-signed to assess quality of care in theseindividuals. Indeed, the National Com-mittee for Quality Assurance applies an

upper age limit (75 years) for its HealthPlan Employer Data and Information Setmeasures, effectively excluding older in-dividuals from quality assessments (14).Rather than exclude such patients, vali-dated measures that assess their quality ofcare and directly incorporate life expect-ancy are needed.

With increasing implementation ofperformance-based reimbursement, theintersection of multiple comorbidity andquality measurement has become a high-profile topic. Two recent articles, one byBoyd et al. (24) and the other by Tinetti etal. (4), used hypothetical patients to con-sider possible concerns for applying cur-rent guidelines to patients with multiplecomorbid conditions. Both articles con-cluded that guideline-concordant caremay result in great expense and marginalbenefits, and the authors cautioned thatenforcing quality measurement for oldercomplex patients may result in unin-tended harm unless future quality mea-sures consider chronic conditions, lifeexpectancy, and patient preferences.

Higashi et al. (25) and Min et al. (26)examined the association of comorbiditywith quality of care in two articles basedon three adult cohorts, the CommunityQuality Index Study, the Assessing Careof Vulnerable Elders Study, and the Vet-erans Health Administration quality ofcare project. They examined 236–439quality indicators covering 22–30 clinicalareas and found, “contrary to [their] ex-pectations,” that as the number of comor-bid conditions increased, adherence toquality measures increased. Both articlesconcluded that multiple comorbid condi-tions result in better quality of care.

In an editorial accompanying thestudy by Min et al., Ritchie (27) arguesthat the study raises as many questions asit answers. Although the Assessing Care ofVulnerable Elders Study attempted to ac-count for patient preferences by including

desire for hospitalization or surgery, it didnot balance the added patient burden ofguideline adherence with goals of re-duced patient/caregiver burden andsymptom control. In fact, current qualitymeasures focus on reducing mortality andmainly ignore functional status and qual-ity of life, which may be more importantin older populations. Ritchie also notesthat the patient and provider may placedifferent values on symptom relief versuscontrol of asymptomatic risk factors fordisease progression. Consistent with ourmain thesis, Ritchie and others concludethat more research must be done to assistproviders in making evidence-based de-cisions that reflect multiple competingclinical factors for patients with high co-morbidity (27–29).

This study has important limitations.First, it is well documented that ICD-9codes in claims data have variable sensi-tivity and specificity for the actual pres-ence of disease (16,30). To increase thesensitivity of our algorithm for identifyingpatients with diabetes, we required onlyone ICD-9 code from a face-to-face phy-sician encounter. Hebert et al. (30) foundthat a single diagnosis of diabetes in Medi-care claims data from a face-to-face phy-sician encounter had a sensitivity of57.9% and a specificity of 96.9%. TheICD-9 code used to determine testing forurine protein is specific to urinary mi-croalbumin; thus, we did not captureother methods of proteinuria screeningsuch as 24 h urine collection, which mayhave contributed to the low observed test-ing rate. Using this ICD-9 code allowed usto be specific for microalbumin and to notoverestimate testing by including routineurinalysis. The analysis of patients onlyon diabetes medications confirmed ourfindings in the larger group. Likewise,claims data may not adequately docu-ment patient comorbidity. However, inan analysis of older men with diabetes,

Table 2—Adjusted predicted probabilities (PPs)* with 95% CIs for diabetes performance measures by CCI quartile for patients in a medicaremanaged care plan, 2003

1st quartile(CCI 0–1)

2nd quartile(CCI 2)

3rd quartile(CCI 3–4)

4th quartile(CCI �5)

A1C testing† 0.83 (0.81–0.85) 0.83 (0.81–0.85) 0.82 (0.80–0.84) 0.78 (0.75–0.80)Lipid testing 0.82 (0.80–0.84) 0.81 (0.79–0.83) 0.79 (0.77–0.82) 0.77 (0.74–0.79)Dilated eye exam‡ 0.48 (0.44–0.51) 0.54 (0.50–0.58) 0.50 (0.47–0.54) 0.50 (0.46–0.54)Urinary microalbumin§ 0.23 (0.21–0.26) 0.24 (0.22–0.27) 0.24 (0.22–0.27) 0.23 (0.20–0.26)

Data are PP (95% CI). Separate logistic regression models were constructed for each test. For all models, the main independent variable was CCI quartile, with quartile1 as the reference. PPs were calculated from the logistic regression parameter estimates. *Models adjusted for age, sex, race/ethnicity, and receipt of diabetesmedication (oral hypoglycemic medications or insulin). †P � 0.05 for overall association of CCI across all quartiles. ‡Model adjusted for above variables and thepresence of diabetic retinopathy. §Model excludes patients with renal disease (n � 780).



there was a 38% increase in 1-year mor-tality for each point increase in the CCIscore, suggesting that CCI score couldserve as an important factor for temperingguideline adherence (31). We note thatmany patients did not participate in thishealth plan’s pharmacy benefit; thus, welacked complete medication data. Ourstudy population included patients of aMedicare managed care health organiza-tion, so we lacked data on other elderlypopulations. We were unable to examinetreatment rates or patient preferences. Fi-nally, unlike recent studies on the topic ofthe association between multiple comor-bid conditions and quality measurement,we focused on quality measures associ-ated with only one disease, diabetes,rather than multiple disease-associatedquality measures (4,24–26).

Providers in this Medicare managedcare health plan had similar rates of ad-herence to diabetes performance mea-sures across all quartiles of comorbidillness burden. It is not clear whetherequally aggressive care confers equal ben-efits to patients with varying comorbid ill-ness burden. Empiric research shouldevaluate the optimal screening intervalsfor A1C, lipid, and urinary microalbumintesting in patients with high comorbid ill-ness burden and decreased life expect-ancy, as well as possible unintendedconsequences of such testing. Finally, pa-tient outcomes in the P4P era should becarefully monitored to assure that healthoutcomes and quality of life in complexpatients are not compromised by pres-sures to perform well on “required” tests.

Acknowledgments— This work was sup-ported by National Center on Minority Healthand Health Disparities/National Institutes ofHealth (NIH) Grant P60MD00502-01. J.H.H.is also supported by a Minority InvestigatorResearch Supplement from the NationalHeart, Lung, and Blood Institute/NIH (N01-HC-48047, modification no. 42).

We thank Nelda Wray, MD, and CatarinaKiefe, MD, PHD, for insightful comments andsuggestions on this manuscript.

References1. National Committee for Quality Assur-

ance: National Committee for Quality As-surance interactive tool: NCQA healthplan report card [tool online]. Availablefrom http://hprc.ncqa.org/. Accessed 4October 2007

2. Jha AK: Measuring hospital quality: whatphysicians do? How patients fare? Orboth? JAMA 296:95–97, 2006

3. Christianson JB, Knutson DJ, Mazze RS:Physician pay-for-performance: imple-mentation and research issues. J Gen In-tern Med 21 (Suppl 2):S9–S13, 2006

4. Tinetti ME, Bogardus ST Jr, Agostini JV:Potential pitfalls of disease-specific guide-lines for patients with multiple condi-tions. N Engl J Med 351:2870–2874, 2004

5. Anderson G, Horvath J: Making the casefor ongoing care: Robert Wood JohnsonFoundation’s partnership for solutions[article online], 2002. Available fromhttp://www.rwjf.org/programareas/resources/product.jsp?id�141978pid�11428gsa�1. Accessed 4 October 2007

6. Wachter RM: Expected and unanticipatedconsequences of the quality and informa-tion technology revolutions. JAMA 295:2780–2783, 2006

7. National Institute of Diabetes and Di-gestive and Kidney Diseases: Nationaldiabetes statistics fact sheet: general in-formation and national estimates on di-abetes in the United States, 2005 [articleonline], 2005. Bethesda, MD, U.S. De-partment of Health and Human Ser-vices, National Institutes of Health.Available from http://diabetes.niddk.nih.gov/dm/pubs/statistics/index.htm#7.Accessed 4 August 2006

8. Brown AF, Mangione CM, Saliba D, Sarki-sian CA, the California Healthcare Foun-dation/American Geriatrics SocietyPanel on Improving Care for Elders withDiabetes: Guidelines for improving thecare of the older person with diabetesmellitus. J Am Geriatr Soc 51 (Suppl. 5):S265–S280, 2003

9. UK Prospective Diabetes Study (UKPDS)Group: Intensive blood-glucose controlwith sulphonylureas or insulin comparedwith conventional treatment and risk ofcomplications in patients with type 2 di-abetes (UKPDS 33). Lancet 352:837–853,1998

10. Randomised trial of cholesterol loweringin 4444 patients with coronary heart dis-ease: the Scandinavian Simvastatin Sur-vival Study (4S). Lancet 344:1383–1389,1994

11. American Diabetes Association: Stan-dards of medical care in diabetes–2006(Position Statement). Diabetes Care 29(Suppl. 1):S17–S26, 2006

12. Romano PS, Roos LL, Jollis JG: Adapting aclinical comorbidity index for use withICD-9-CM administrative data: differingperspectives. J Clin Epidemiol 46:1075–1079, 1993

13. Charlson M, Szatrowski TP, Peterson J,Gold J: Validation of a combined comor-bidity index. J Clin Epidemiol 47:1245–1251, 1994

14. NCQA Programs: The Health Plan Em-ployer Data and Information Set (HEDIS)[article online]. Available from http://web.ncqa.org/tabid/59/Default.aspx. Ac-cessed 4 October 2006

15. Sofroniou N, Hutcheson GD: Confidenceintervals for the predictions of logistic re-gression in the presence and absence of avariance-covariance matrix. Understand-ing Statistics 1:3, 2002

16. Miller DR, Safford MM, Pogach LM: Whohas diabetes? Best estimates of diabetesprevalence in the Department of VeteransAffairs based on computerized patientdata. Diabetes Care 27 (Suppl 2):B10–B21, 2004

17. Langley J, Adams G: Insulin-based regi-mens decrease mortality rates in criticallyill patients: a systematic review. DiabetesMetab Res Rev 23:184–192, 2007

18. Reed CC, Stewart RM, Sherman M, MyersJG, Corneille MG, Larson N, Gerhardt S,Beadle R, Gamboa C, Dent D, Cohn SM,Pruitt BA Jr: Intensive insulin protocolimproves glucose control and is associ-ated with a reduction in intensive careunit mortality. J Am Coll Surg 204:1048–1054, 2007

19. Fong DS, Aiello L, Gardner TW, King GL,Blankenship G, Cavallerano JD, Ferris FL,Klein R: Retinopathy in diabetes (PositionStatement). Diabetes Care 27 (Suppl 1):S84–S87, 2004

20. Gross CP, Mallory R, Heiat A, KrumholzHM: Reporting the recruitment process inclinical trials: who are these patients andhow did they get there? Ann Intern Med137:10–16, 2002

21. Fortin M, Dionne J, Pinho G, Gignac J,Almirall J, Lapointe L: Randomized con-trolled trials: do they have externalvalidity for patients with multiple co-morbidities? Ann Fam Med 4:104 –108,2006

22. Starfield B: Threads and yarns: weavingthe tapestry of comorbidity. Ann Fam Med4:101–103, 2006

23. Sachdev M, Sun JL, Tsiatis AA, NelsonCL, Mark DB, Jollis JG: The prognosticimportance of comorbidity for mortalityin patients with stable coronary arterydisease. J Am Coll Cardiol 43:576 –582,2004

24. Boyd CM, Darer J, Boult C, Fried LP,Boult L, Wu AW: Clinical practice guide-lines and quality of care for older patientswith multiple comorbid diseases: impli-cations for pay for performance. JAMA294:716–724, 2005

25. Higashi T, Wenger NS, Adams JL, FungC, Roland M, McGlynn EA, Reeves D,Asch SM, Kerr EA, Shekelle PG: Relation-ship between number of medical condi-tions and quality of care. N Engl J Med356:2496–2504, 2007

26. Min LC, Wenger NS, Fung C, Chang JT,Ganz DA, Higashi T, Kamberg CJ, Ma-cLean CH, Roth CP, Solomon DH, YoungRT, Reuben DB: Multimorbidity is associ-ated with better quality of care among vul-nerable elders. Med Care 45:480–488,2007

27. Ritchie C: Health care quality and multi-

Halanych and Associates


morbidity: jury still out. Med Care 45:477–479, 2007

28. Rijken M, van Kerkhof M, Dekker J,Schellevis FG: Comorbidity of chronicdiseases: effects of disease pairs on phys-ical and mental functioning. Qual Life Res14:45–55, 2005

29. Fried LP, Bandeen-Roche K, Kasper JD,

Guralnik JM: Association of comorbiditywith disability in older women: Women’sHealth and Aging Study. J Clin Epidemiol52:27–37, 1999

30. Hebert PL, Geiss LS, Tierney EF, Engel-gau MM, Yawn BP, McBean AM: Identify-ing persons with diabetes using Medicareclaims data. Am J Med Qual 14:270–277,

199931. Kahler KH, Rajan M, Rhoads GG, Safford

MM, Demissie K, Lu SE, Pogach LM: Im-pact of oral antihyperglycemic therapy onall-cause mortality among patients withdiabetes in the Veterans Health Adminis-tration. Diabetes Care 30:1689–1693,2007



Diabetes, Depression, and DeathA randomized controlled trial of a depression treatment program for olderadults based in primary care (PROSPECT)

HILLARY R. BOGNER, MD, MSCE1

KNASHAWN H. MORALES, SCD2

EDWARD P. POST, MD, PHD3,4

MARTHA L. BRUCE, PHD, MPH5

OBJECTIVE — We sought to test our a priori hypothesis that depressed patients with dia-betes in practices implementing a depression management program would have a decreased riskof mortality compared with depressed patients with diabetes in usual-care practices.

RESEARCH DESIGN AND METHODS — We used data from the multisite, practice-randomized, controlled Prevention of Suicide in Primary Care Elderly: Collaborative Trial(PROSPECT), with patient recruitment from May 1999 to August 2001, supplemented with asearch of the National Death Index. Twenty primary care practices participated from the greatermetropolitan areas of New York City, New York; Philadelphia, Pennsylvania; and Pittsburgh,Pennsylvania. In all, 584 participants identified though a two-stage, age-stratified (aged 60–74or �75 years) depression screening of randomly sampled patients and classified as depressedwith complete information on diabetes status are included in these analyses. Of the 584 partic-ipants, 123 (21.2%) reported a history of diabetes. A depression care manager worked withprimary care physicians to provide algorithm-based care. Vital status was assessed at 5 years.

RESULTS — After a median follow-up of 52.0 months, 110 depressed patients had died.Depressed patients with diabetes in the intervention category were less likely to have died duringthe 5-year follow-up interval than depressed diabetic patients in usual care after accounting forbaseline differences among patients (adjusted hazard ratio 0.49 [95% CI 0.24–0.98]).

CONCLUSIONS — Older depressed primary care patients with diabetes in practices imple-menting depression care management were less likely to die over the course of a 5-year intervalthan depressed patients with diabetes in usual-care practices.

Diabetes Care 30:3005–3010, 2007

D iabetes and depression are two of themost common problems seen in pri-mary care settings. Epidemiologic

data indicate that diabetes and depressionare intimately related. Depression is a riskfactor for diabetes (1), and risk of depres-sion is increased by a factor of two in pa-tients with diabetes (2). Depression is notonly common in patients with diabetes

but also contributes to poor adherence tomedication and dietary regimens, poor gly-cemic control, reduced quality of life, andincreased health care expenditures (3).Depression has been specifically linked toprognostic variables in diabetes such asmicro- and macrovascular complications(4). Evidence from intervention trialsshows that treatment of depression in pa-

tients with diabetes improves depression(5–7), but findings regarding improve-ment in glycemic control have beenmixed (5,8,9). Although cohort studiesdocument that depression is associatedwith increased risk of death among in-dividuals with diabetes (10 –13), noknown intervention study has evaluatedwhether treatment for depression mod-ifies this increased risk of mortalityamong older primary care patients withdiabetes.

We investigated the relationship be-tween diabetes, depression treatment,and all-cause mortality using data fromthe multisite, randomized trial, Preven-tion of Suicide in Primary Care Elderly:Collaborative Trial (PROSPECT), supple-mented with a search of the NationalDeath Index (NDI) Plus. The study inter-vention was implemented at the practicelevel and involved a depression care man-ager working with physicians to providealgorithm-based treatment and ongoingcare management. Overall, interventionpatients had a more favorable course of de-pression in both degree and speed of symp-tom reduction compared with usual-carepatients (14). We took the opportunityafforded by PROSPECT to evaluate the ef-fect of diabetes and of depression caremanagement on all-cause mortality forthe following reasons. While multiplemedical conditions are of interest in thisintervention trial, depression associatedwith diabetes has been shown to increasethe risk of death (10–13). Furthermore,the demonstrated morbidity, mortality,and health services implications of diabe-tes and depression separately (15–17)support both understanding of the enor-mous public health significance of the co-occurrence of diabetes and depressionand the urgency to finding evidence-based solutions to reduce the burden ofthese conditions. We hypothesized thatdepressed older adults with diabetes inpractices randomized to interventionwould be less likely to die over a 5-yearfollow-up interval compared with de-pressed older adults with diabetes inusual care.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Family Medicine and Community Health, University of Pennsylvania, Philadel-phia, Pennsylvania; the 2Center for Clinical Epidemiology and Biostatistics, School of Medicine, Universityof Pennsylvania, Philadelphia, Pennsylvania; the 3Veterans Affairs Health Services Research and Develop-ment and National Serious Mental Illness Treatment Research and Evaluation Center, Ann Arbor, Michigan;the 4Department of Internal Medicine, University of Michigan, Ann Arbor, Michigan; and the 5Departmentof Psychiatry, Weill Medical College of Cornell University, White Plains, New York.

Address correspondence and reprint requests to Hillary R. Bogner, MD, Department of Family Practiceand Community Medicine, University of Pennsylvania, 3400 Spruce St., 2 Gates Building, Philadelphia, PA19104. E-mail: [email protected].

Received for publication 21 May 2007 and accepted in revised form 18 August 2007.Published ahead of print at http://care.diabetesjournals.org on 23 August 2007. DOI: 10.2337/dc07-

0974. Clinical trial reg. no. NCT00000367, clinicaltrials.gov.Abbreviations: CES-D, Centers for Epidemiologic Studies Depression scale; NDI, National Death Index;

PROSPECT, Prevention of Suicide in Primary Care Elderly: Collaborative Trial.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





RESEARCH DESIGN ANDMETHODS

PROSPECTPROSPECT compared a primary care–based intervention with usual care in im-proving the outcomes of depression. Allstudy procedures were implemented withwritten informed consent, and the studyprotocols were approved by the institu-tional review boards of Cornell University(Ithaca, NY), the University of Pittsburgh(Pittsburgh, PA), and the University ofPennsylvania (Philadelphia, PA) schoolsof medicine. Details of the study design ofPROSPECT are available elsewhere (14).In brief, 20 primary care practices fromthe greater metropolitan areas of NewYork City, New York, and Philadelphiaand Pittsburgh, Pennsylvania, partici-pated in the study from May 1999 to Au-gust 2001, with individual patientsclinically followed for 2 years. Practicesranged in size (solo to medium sized), set-ting (sparsely populated, suburban, andurban), population type (including twoserving primarily African-American pa-tients), and affiliation (16 community-based and 4 academic practices).Practices were paired by region (urban vs.suburban/sparsely populated), affiliation,size, and population type. Within the 10pairs, practices were randomly assignedby coin flip to intervention or usual care(described below). A two-stage samplingdesign was used to recruit patients. First,an age-stratified (aged 60–74 or � 75years), random sample of patients with anupcoming appointment was obtained.The sampled patients were mailed a letterallowing patients to decline. Second,trained lay interviewers telephoned thepatients who did not decline. Patientswho gave oral consent were assessed forenrollment using the Centers for Epide-miologic Studies Depression scale(CES-D) (18). All patients with a CES-Dscore �20 were invited into the study, aswere those from a 5% random sample ofpatients with lower scores. Patients with aCES-D score �20 and who were not ran-domly selected were also recruited if theyresponded positively to supplementalquestions about mood, prior depressiveepisodes, or treatment. A positive re-sponse to the supplemental questionstriggered a diagnostic assessment.

The intervention, described in detailelsewhere (14), consisted of trained de-pression care managers offering guide-line-concordant recommendations to theprimary care physicians and helping pa-

tients with treatment adherence. The caremanagers monitored psychopathology,treatment adherence, response, and sideeffects and provided follow-up care atpredetermined intervals or when clini-cally necessary. Patients who refused anti-depressants were offered interpersonalpsychotherapy by the depression caremanagers. In the intervention, a first-lineantidepressant (citalopram, a selective se-rotonin reuptake inhibitor) and the inter-personal psychotherapy were provided atno cost. In usual care, physicians wereinformed of patients’ depression diag-noses. Physicians also received informa-tional materials and treatment guidelinesfor geriatric depression. No specific rec-ommendations were given to these physi-cians regarding individual patients exceptfor psychiatric emergencies. The typesand proportions of treatment receivedover time by individuals in practices ran-domized to intervention or usual carehave previously been published (14,19).

Measurement strategyTrained research assistants assigned de-pression diagnoses to patients using theStructured Clinical Interview for DSM-IVAxis I Disorders (SCID-I) diagnoses (20).Severity of depression was assessed usingthe 24-item Hamilton Depression RatingScale (21).

Individuals were classified as having amedical comorbidity and as having diabe-tes by self-report. The questionnaire usedwas based on the Charlson ComorbidityIndex (22). To assess for diabetes, partic-ipants were asked, “Have you ever beentold you have diabetes or high blood glu-cose?” For the current analysis, patientswere considered to have diabetes if theyreported having been told that they haddiabetes or high blood glucose.

We used standard questions to obtaininformation from the respondents on age,level of educational attainment, sex, mar-ital status, and self-reported ethnicity.Smoking status was based on report ofsmoking within 6 months of interview.The Philadelphia Multidimensional As-sessment Instrument assessed instrumen-tal activities of daily living and mobility(23). The Scale for Suicidal Ideation mea-sured presence of suicidal ideation (24).The Mini-Mental State Examination is ashort standardized mental status exami-nation that has been widely employed forclinical and research purposes (25).

Ascertainment of vital statusVital status in this investigation was basedon follow-up of participants using the Na-tional Center for Health Statistics NDI(26). Because obtaining vital status re-quires that we provide personally identi-fiable information to the National Centerfor Health Statistics for NDI searches,confidentiality safeguards warrant discus-sion here. We did not transmit any PROS-PECT data (e.g., information aboutdepression status, physical disorders, orfunctional status) with identifying data,nor did we transmit identifying data viae-mail. Upon obtaining vital status data,the University of Pennsylvania Data Coresent the data to the sites for verification.Study sites then sent the data file—stripped of any identifying data—to theUniversity of Pennsylvania Data Core forfinal production of the study data linkedto vital status for analysis. The time framefor the ascertainment of vital status wasthe period of 5 years from overall com-mencement of PROSPECT.

Analytic strategyOur analysis involved sorting patientsinto four groups according to whether pa-tients self-reported diabetes at baselineand practice assignment (intervention orusual care). We carried out survival anal-ysis adjusting for within-practice cluster-ing (27). The Cox proportional hazardsmodel for clustered data was used to ex-plore the effect of variables on survival.Point estimates and associated 95% CIsare provided for the unadjusted and ad-justed hazard ratios (HRs) (as presentedin previous studies [14,28]). Survivalcurves were prepared using the method ofKaplan and Meier (29) to illustrate themortality of each group defined by patientdiabetes status at baseline and to practicerandomization assignment to interven-tion or usual care. We began by exploringpotential confounding variables usingunivariate models with baseline charac-teristics as predictors of time to death.Our final model included influential co-variates identified by their association(P � 0.10) with the outcome of interest,time to death. The final model includedterms to adjust for baseline differences inage, sex, education, ethnicity, smokingstatus, number of medical conditions,number of disabilities, and cognition.

We have been guided by publishedcriteria for performing and reporting sub-group analyses (30,31). Evaluating ourprespecified study hypothesis required atest for effect modification of intervention

Diabetes, depression, and death


assignment on the risk of death by base-line diabetes status. The formal test foreffect modification involved introducingterms representing interaction into theCox proportional hazards model, in addi-tion to main effects for diabetes status andintervention. Consistent with the litera-ture (32), we set � at 0.10 to denote sta-tistical significance for the interactionterm in the Cox proportional hazardsmodel. SAS was used to carry out analyses(version 9.1; SAS Institute, Cary, NC).

RESULTS

Study sampleThe CONSORT (Consolidated Standardsfor Reporting of Trials) flow diagram forPROSPECT has previously been pub-lished (14). In brief, the study screened9,072 older individuals, 1,888 of whomwere invited to participate. Of the 1,888individuals invited to participate, 1,238(65.8%) agreed to a baseline interview.Our study sample included 599 de-pressed patients, of whom 396 (66.1%)met Diagnostic and Statistical Manual ofMental Disorders-IV criteria for major de-pression. Fifteen people were excludeddue to missing data on baseline diabetesstatus, leaving a sample size of 584 for thisanalysis.

Sample characteristicsThe mean � SD age of our study samplewas 70.3 � 7.9 years. The age range was

60 –94 years. Women comprised 422(72.3%) of the participants. The self-identified ethnic groups of the partici-pants consisted of 407 Caucasians(69.7%), 161 African Americans (27.6%),and 16 American Indians, Hispanics, orAsians (2.7%). Of all 584 participants,123 (21.2%) reported a history of diabe-tes. Table 1 compares baseline character-istics between patients in the interventionand usual-care practices, stratified by di-abetes status. After 5 years, 110 depressedpatients had died. Only one documentedsuicide occurred during the study in a de-pressed patient with diabetes in an inter-vention practice. The median length offollow-up in ascertainment of vital statuswas 52.0 months (range 0.8–67.7).

Mortality risk according to diabetesstatusDepressed patients with diabetes in theintervention practices experienced a mor-tality rate of 68.2/1,000 person-years(95% CI 41.0–106.5), whereas depressed

patients with diabetes in usual care expe-rienced a mortality rate of 103.4/1,000person-years (63.2–159.7). Individualswithout diabetes experienced similarmortality rates in the intervention andusual-care practices (36.0/1,000 person-years [95% CI 25.3–49.6] vs. 38.2/1,000[26.4–53.3], respectively).

Table 2 provides unadjusted and ad-justed HR estimates according to diabetesstatus. In the univariate model, depressedpatients with diabetes in the interventionpractices were less likely to have died dur-ing the 5-year follow-up interval than de-pressed patients with diabetes in usualcare (unadjusted HR 0.66 [95% CI 0.36–1.21]), but the 95% CIs included the null.The final model accounted for baselineimbalances in age, sex, education, ethnic-ity, smoking status, number of medicalconditions, number of disabilities, andcognition among patients. Depressed pa-tients with diabetes in the interventionpractices were significantly less likely tohave died during the 5-year follow-up in-

Table 1—Characteristics of the study sample according to randomization assignment of primary care practice and diabetes status at baseline

Depressed patients Intervention Usual care No intervention No usual careTest of equality

across groups (P)*

n 70 53 241 220Sociodemographic characteristics

Age (years) 71 � 8.5 67 � 6.8 71 � 7.6 71 � 8.1 0.0004Education (years) 12 � 3.0 12 � 3.2 13 � 3.3 13 � 3.3 0.0006Women 50 (71) 37 (70) 167 (69) 168 (76) 0.3944Ethnic minority 23 (33) 27 (51) 58 (24) 69 (31) 0.0802Married 22 (31) 21 (40) 91 (38) 81 (37) 0.5074

Medical conditionsCurrent smoker 9 (13) 3 (6) 26 (11) 12 (5) 0.2472Medical conditions 5 � 2.9 5 � 2.4 2 � 1.9 2 � 1.9 �0.0001Number of disabilities (MAI score) 3 � 2.3 3 � 2.2 2 � 1.9 2 � 1.8 0.0040

Baseline depression and cognitive statusDepression severity (HDRS score) 18 � 5.3 19 � 5.6 18 � 6.3 17 � 5.8 0.1789Suicidal ideation (SSI score �0) 23 (33) 12 (23) 67 (28) 44 (20) 0.0997Cognitive function MMSE score) 27 � 4.2 27 � 2.6 28 � 2.4 27 � 2.5 0.0848

Data are means � SD or n (%), with percentages based on the total number in the corresponding column, unless otherwise indicated. Test of equality across groupsbased on regression models. Data gathered from PROSPECT. The ranges of scores are 0–30 for the Mini-Mental State Examination (MMSE), with high scoresindicating less severe cognitive impairment.; 0–76 for the Hamilton Depression Rating Scale (HDRS), with high scores indicating greater depressive symptoms; and0–38 for the Scale for Suicidal Ideation (SSI), with high scores indicating greater suicidal ideation. *Univariate logistic or linear regression model with random effects.MAI, Multidimensional Assessment Instrument.

Table 2—Relationship of practice random assignment, patient baseline diabetes status, andmortality during a 5-year follow-up interval

Intervention and usual-carepatients’ diabetes status

Unadjusted HR(95% CI)

Adjusted HR(95% CI)*

Diabetes 0.66 (0.36–1.21) 0.49 (0.24–0.98)No diabetes 0.94 (0.58–1.52) 0.79 (0.42–1.47)

Data gathered from PROSPECT.

Bogner and Associates


terval than depressed patients with diabe-tes in usual care (adjusted HR 0.49 [0.24–0.98]). In contrast, depressed patientswithout diabetes in the intervention prac-tices were not at decreased risk comparedwith those without diabetes in usual care(adjusted HR 0.79 [0.42–1.47]). The in-teraction of randomization group by dia-betes status was statistically significant(P � 0.04). We have provided Kaplan-Meier curves according to diabetes statusfor intervention versus usual care (Fig. 1).

CONCLUSIONS — Depressed olderadults with diabetes who were in prac-tices randomized to intervention were lesslikely to have died at the end of the 5-yearfollow-up interval than depressed indi-viduals with diabetes in usual care afteradjustment for baseline differences. De-pressed patients without diabetes in theintervention were not at decreased riskcompared with depressed patients with-out diabetes in usual care. Interventionattenuates the influence of diabetes onmortality risk among older adults withdepression. We believe these findingssupport the integration of depressionevaluation and treatment with diabetesmanagement in primary care.

Before discussing our findings, the re-sults must first be considered in the con-text of potential study limitations. First,we obtained our results from primary caresites in the greater metropolitan areas ofNew York City, Philadelphia, and Pitts-

burgh, whose patients may not be repre-sentative of other primary care practicesin the U.S. However, the participatingsites were diverse practices of varying sizelocated in urban, suburban, and sparselypopulated areas. Second, diabetes diag-nosis was based on self-report alone;however, relying on medical records mayalso be incomplete because many individ-uals receive health care from providers inmultiple systems. Studies have shownthat self-reported data on diabetes as wellas other chronic diseases is reliable (33).Third, the question regarding diabetesmight have included patients with im-paired glucose tolerance who did not havediabetes. However, misclassification ofsome individuals with impaired glucosetolerance as having diabetes would lead toa conservative bias toward the null (i.e.,no intervention effect for patients with di-abetes on mortality). Fourth, the mortal-ity reduction among depressed patientswith diabetes randomized to interventionmay be due to factors other than the spe-cific effects of a depression managementprogram. For example, we only have alimited ability to address whether patientswith diabetes in the intervention practiceswere seen more frequently for reasonsother than depression by their physicians;similarly, we do not have information onspecific diabetes outcomes such as A1C.Fifth, misclassification of vital status was apotential limitation. However, overallsensitivity of the NDI for ascertainment of

vital status has generally been well over90% in most studies (34).

We selected patients with diabetes asa subgroup from the larger interventiontrial (14,35), realizing that we must pro-ceed with caution about the inferences wemake. Statisticians are wary of subgroupanalyses, but clinicians must make deci-sions about individual patients (30,36,37).At the same time, large-scale interventionstudies carried out in primary care prac-tice are limited, so we need to make themost of the data we have from interven-tion studies. Guided by published criteriafor performing and reporting subgroupanalyses (30,31), we have identified agroup— older individuals with diabe-tes—for whom risk of death has been re-ported to be increased (10 –13). Inaddition, the link between diabetes, de-pression, and the outcome (mortality)may have common pathophysiologicmechanisms (38,39).

Finally, uncertainty persists aboutthe influence of treatment of depressionon outcomes for diabetes and other med-ical comorbidity (8,9). Consistent withrecommendations regarding subgroupanalyses, we reported the statistical signif-icance of the interaction between inter-vention assignment and the condition ofinterest on the outcome (32,40) and ad-justed our estimates for potential imbal-ances in covariates across treatmentgroups (41).

Despite some limitations, our studywarrants attention because older de-pressed primary care patients with diabe-tes in practices implementing depressioncare management were significantly lesslikely to die over the course of a 5-yearinterval than depressed patients with dia-betes in usual-care practices. To ourknowledge, this is the first study to reporton the relationship between diabetes andmortality in a depression interventiontrial. A formal test of the interaction be-tween intervention assignment and dia-betes on the outcome of interest, all-causemortality, was significant (32,40). Thissuggests that individuals with diabeteswere more likely to benefit from interven-tion than individuals without diabetes.Adjustment of the HR for imbalance in thedistribution of baseline covariates as-sessed at baseline can be expected to yieldestimates of the hazard closer to the trueestimate of the treatment effect (40,41).Because our sample was derived from pri-mary health care, the public health signif-icance of these findings is high.

The combination of clinical evalua-

Figure 1—Survival curves for patients with (n � 53) and without (n � 220) diabetes in practicesrandomized to usual care and patients with (n � 70) and without (n � 241) diabetes randomizedto intervention practices. Data gathered from PROSPECT.



tion and monitoring, pharmacotherapy,and, in some cases, interpersonal psycho-therapy in the PROSPECT interventionappears to be effective in depressed pa-tients with diabetes in reducing all-causemortality risk. We realize that our studydoes not examine potential mechanismsunderlying the relationship between thePROSPECT intervention and a decreasedmortality risk among depressed patientswith diabetes. Both physiologic factors,such as increased inflammation (38,39)and poor glucose regulation (3,4), and be-havioral processes, such as poor adher-ence (3), may link depression withincreased mortality in patients with dia-betes. The potential mediators betweentreatment assignment and outcomes forpatients with diabetes deserve furtherstudy.

Our results add to the literature onclinical trial outcomes from treatment ofdepression in patients with diabetes. Spe-cifically, the collaborative care model fordepression, of which PROSPECT is oneexample, has been found to improve de-pression care and depression outcomes inpatients with diabetes. The IMPACT (Im-proving Mood-Promoting Access to Col-laborative Treatment) trial found thatdepressed older adults with diabetes in adepression care management interven-tion had better depression outcomes at 1year compared with depressed olderadults with diabetes in usual care, al-though A1C levels were unaffected by theintervention (9). The authors point outthat because patients had good glycemiccontrol at baseline, power to detect smallbut clinically important improvements inglycemic control was limited. The Path-ways Study randomized 329 patients withdiabetes and comorbid major depressionor dysthymia to depression care manage-ment or usual care and found that al-though depression outcomes wereimproved, no differences in A1C levelswere observed (8). However, these au-thors also point out that the patients in thePathways Study had good glycemic con-trol at baseline.

In summary, our investigation addsnew evidence to the literature on depres-sion and diabetes by examining whetherthe PROSPECT intervention influencedsurvival among depressed older primarycare patients with diabetes. Specifically,these results indicate that a depressioncare management intervention can signif-icantly reduce all-cause mortality amongdepressed patients with diabetes. Theseresults should propel the development

and dissemination of models of care thatbetter integrate depression managementfor individuals with diabetes.

Acknowledgments— PROSPECT was a col-laborative research study funded by the Na-tional Institute of Mental Health (NIMH). Themortality follow-up of PROSPECT partici-pants was funded by the NIMH (R01MH065539). Participation of H.R.B., E.P.P.,and M.L.B. was also supported by NIMHawards (K23 MH67671, K23 MH01879, andK02 MH01634, respectively).

References1. Eaton WW, Armenian H, Gallo J, Pratt L,

Ford DE: Depression and risk for onset oftype II diabetes: a prospective population-based study. Diabetes Care 19:1097–1102, 1996

2. Eaton WW: Epidemiologic evidence onthe comorbidity of depression and diabe-tes. J Psychosom Res 53:903–906, 2002

3. Lustman PJ, Clouse RE: Depression in di-abetic patients: the relationship betweenmood and glycemic control. J DiabetesComplications 19:113–122, 2005

4. de Groot M, Anderson R, Freedland KE,Clouse RE, Lustman PJ: Association of de-pression and diabetes complications: ameta-analysis. Psychosom Med 63:619–630, 2001

5. Lustman PJ, Griffith LS, Clouse RE,Freedland KE, Eisen SA, Rubin EH, Car-ney RM, McGill JB: Effects of nortriptylineon depression and glycemic control in di-abetes: results of a double-blind, placebo-controlled trial. Psychosom Med 59:241–250, 1997

6. Lustman PJ, Griffith LS, Freedland KE,Kissel SS, Clouse RE: Cognitive behaviortherapy for depression in type 2 diabetesmellitus: a randomized, controlled trial.Ann Intern Med 129:613–621, 1998

7. Lustman PJ, Freedland KE, Griffith LS,Clouse RE: Fluoxetine for depression indiabetes: a randomized double-blind pla-cebo-controlled trial. Diabetes Care 23:618–623, 2000

8. Katon WJ, Von Korff M, Lin EH, SimonG, Ludman E, Russo J, Ciechanowski P,Walker E, Bush T: The Pathways Study:a randomized trial of collaborative carein patients with diabetes and depres-sion. Arch Gen Psychiatry 61:1042–1049, 2004

9. Williams JW, Katon W, Lin EHB, NoelPH, Worchel J, Cornell J, Harpole L, FultzBA, Hunkeler E, Mika VS, Unutzer J; IM-PACT Investigators: The effectiveness ofdepression care management on diabetes-related outcomes in older patients. AnnIntern Med 140:1015–1024, 2004

10. Zhang X, Norris SL, Gregg EW, Cheng YJ,Beckles G, Kahn HS: Depressive symp-toms and mortality among persons with

and without diabetes. Am J Epidemiol 161:652–660, 2005

11. Katon WJ, Rutter C, Simon G, Lin EH,Ludman E, Ciechanowski P, Kinder L,Young B, Von Korff M: The association ofcomorbid depression with mortality inpatients with type 2 diabetes. DiabetesCare 28:2668–2672, 2005

12. Black SA, Markides KS, Ray LA: Depres-sion predicts increased incidence of ad-verse health outcomes in older MexicanAmericans with type 2 diabetes. DiabetesCare 26:2822–2828, 2003

13. Egede LE, Nietert PJ, Zheng D: Depres-sion and all-cause and coronary heart dis-ease mortality among adults with andwithout diabetes. Diabetes Care 28:1339–1345, 2005

14. Bruce ML, Ten Have TR, Reynolds CF3rd, Katz II, Schulberg HC, Mulsant BH,Brown GK, McAvay GJ, Pearson JL,Alexopoulos GS: Reducing suicidal ide-ation and depressive symptoms in de-pressed older primary care patients: arandomized controlled trial. JAMA 291:1081–1091, 2004

15. Ezzati M, Jamison DT, Lopez AD, MathersCD, Murray C, Eds.: Global Burden of Dis-ease and Risk Factors. Washington, DC,Oxford University Press and World Bank,2006

16. Gallo JJ, Bogner HR, Morales KH, Post EP,Ten Have T, Bruce ML: Depression, car-diovascular disease, diabetes, and two-year mortality among older, primary-carepatients. Am J Geriatr Psychiatry 13:748–755, 2005

17. Carter GM, Bell RM, Dubois RW, Gold-berg GA, Keeler EB, McAlearney JS, PostEP, Rumpel JD: A clinically detailed riskinformation system for cost. Health CareFinanc Rev 21:65–91, 2000

18. Radloff LS: The CES-D scale: a self-reportdepression scale for research in the gen-eral population. Appl Psych Meas 1:385–401, 1977

19. Schulberg HC, Post EP, Raue PJ, Have TT,Miller M, Bruce ML: Treating late-life de-pression with interpersonal psychother-apy in the primary care sector. Int J GeriatrPsychiatry 22:106–14, 2007

20. Spitzer RL, Gibbon M, Williams JB: Struc-tured Clinical Interview for Axis I DSM-IVDisorders (SCID). Washington, DC, Ameri-can Association Press, 1995

21. Hamilton M: A rating scale for depression.J Neurol Neurosurg Psychiatry 23:56–62,1960

22. Charlson ME, Pompei P, Ales KL, Mac-Kenzie CR: A new method of classifyingprognostic comorbidity in longitudinalstudies: Development and validation.J Chronic Dis 40:373–383, 1987

23. Lawton MP, Moss M, Fulcomer M, KlebanMH: A research and service oriented mul-tilevel assessment instrument. J Gerontol37:91–99, 1982

24. Beck A, Brown G, Steer R: Psychometric

Bogner and Associates


characteristics of the scale for suicide: ide-ation with psychiatric outpatients. BehavRes Ther 35:1039–1046, 1997

25. Folstein MF, Folstein SE, McHugh PR:“Mini-Mental State”: a practical methodfor grading the cognitive state of patientsfor the clinician. J Psychiatr Res 12:189–198, 1975

26. Doody MM, Hayes HM, Bilgrad R: Com-parability of national death index plusand standard procedures for determiningcauses of death in epidemiologic studies.Ann Epidemiol 11:46–50, 2001

27. Lee E, Wei L, Amato D: Cox-type regres-sion analysis for large numbers of smallgroups of correlated failure time observa-tions. In Survival Analysis. Norwell, MA,Kluwer Academic Publishers, 1992

28. Bogner HR, Ford DE, Gallo JJ: The role ofcardiovascular disease in the identifica-tion and management of depression byprimary care physicians. Am J Geriatr Psy-chiatry 14:71–78, 2006

29. Kaplan E, Meier P: Nonparametric esti-mation from incomplete observations.J Am Stat Assoc 53:457–481, 1958

30. Rothwell PM: Treating individuals. 2.Subgroup analysis in randomised con-trolled trials: importance, indications,and interpretation. Lancet 365:176–186,2005

31. Oxman AD, Guyatt GH: A consumer’sguide. Ann Intern Med 116:78–84, 1992

32. Fleiss JL: Analysis of data from multi-clinic trials. Control Clin Trials 7:267–75, 1986

33. Bowlin SJ, Morrill BD, Nafziger AN, LewisC, Pearson TA: Reliability and changes invalidity of self-reported cardiovasculardisease risk factors using dual response:the behavioral risk factor survey. J ClinEpidemiol 49:511–517, 1996

34. Sathiakumar N, Delzell E, Abdalla O: Us-ing the National Death Index to obtainunderlying cause of death codes. J OccupEnviron Med 40:808–813, 1998

35. Gallo JJ, Bogner HR, Morales KH, Post EP,Lin JY, Bruce ML: The effect on mortalityof a practice-based depression interven-tion program for older adults in primarycare: a cluster randomized trial. Ann In-tern Med 146:689–698, 2007

36. Kravitz RL, Duan N, Braslow J: Evidence-based medicine, heterogeneity of treat-ment effects, and the trouble withaverages. Milbank Q 82:661–687, 2004

37. Senn S: Individual response to treatment:BMJ 329:966–968, 2004

38. Musselman DL, Betan E, Larsen H, Phil-lips LS: Relationship of depression to di-abetes types 1 and 2: epidemiology,biology, and treatment. Biol Psychiatry 54:317–329, 2003

39. Joynt KE, Whellan DJ, O’Connor CM: De-pression and cardiovascular disease:mechanisms of interaction. Biol Psychiatry54:248–261, 2003

40. Lu M, Lyden PD, Brott TG, Hamilton S,Broderick JP, Grotta JC: Beyond subgroupanalysis: improving the clinical interpre-tation of treatment effects in stroke re-search. J Neurosci Methods 143:209–216,2005

41. Buyse M: Analysis of clinical trial out-comes: some comments on subgroupanalysis. Control Clin Trials 10 (Suppl. 4):187S–194S, 1989



Glycemic Effects of Moderate AlcoholIntake Among Patients With Type 2DiabetesA multicenter, randomized, clinical intervention trial

IRIS SHAI, RD, PHD1

JULIO WAINSTEIN, MD2

ILANA HARMAN-BOEHM, MD3

ITAMAR RAZ, MD4

DRORA FRASER, PHD1

ASSAF RUDICH, MD, PHD5

MEIR J. STAMPFER, MD, DRPH6

OBJECTIVE — In a randomized controlled trial, we assessed the effect of daily moderatealcohol intake on glycemic control in the fasting and postprandial states in patients with type 2diabetes who previously had abstained from alcohol.

RESEARCH DESIGN AND METHODS — We randomly assigned 109 patients (41–74years old) with established type 2 diabetes who abstained from alcohol to receive 150 ml wine(13 g alcohol) or nonalcoholic diet beer (control) each day during a 3-month multicenter trial.The beverages were consumed during dinner. Diet and alcohol consumption were monitored.

RESULTS — During the intervention, 17% of participants (12% from the alcohol group)dropped out, leaving 91 who completed the trial. Within the alcohol group, fasting plasmaglucose (FPG) decreased from 139.6 � 41 to 118.0 � 32.5 mg/dl after 3 months compared with136.7 � 15.4 to 138.6 � 27.8 mg/dl in the control subjects (Pv � 0.015). However, alcoholconsumption had no effect on 2-h postprandial glucose levels (difference of 18.5 mg/dl in thecontrol group vs. 17.7 mg/dl in the alcohol group, Pv � 0.97). Patients in the alcohol group withhigher baseline A1C levels had greater reductions in FPG (age-adjusted correlation �0.57, Pv �0.001). No significant changes were observed in the levels of bilirubin, alkaline phosphatase,alanine aminotransferase, or aspartate aminotransferase, and no notable adverse effects werereported. Participants in the alcohol group reported an improvement in the ability to fall asleep(Pv � 0.001).

CONCLUSIONS — Among patients with type 2 diabetes who had previously abstained fromalcohol, initiation of moderate daily alcohol consumption reduced FPG but not postprandialglucose. Patients with higher A1C may benefit more from the favorable glycemic effect of alcohol.

Further intervention studies are needed toconfirm the long-term effect of moderate alco-hol intake.

Diabetes Care 30:3011–3016, 2007

A s summarized in a recent editorial(1), proving the beneficial effect ofmoderate alcohol intake awaits re-

sults of randomized controlled interven-tion trials. In observational studies,moderate alcohol intake is associated withlower incidence of type 2 diabetes, withan apparent J-shape association (2–4).Also, a recent meta-analysis of patientswith type 2 diabetes (5) suggested thatmoderate alcohol consumption is associ-ated with a lower risk of mortality andcoronary heart disease.

Successful long-term control of hy-perglycemia decreases diabetes complica-tions (6) and is therefore a major goal indiabetes management. Ethanol metabo-lism increases the hepatic cytosolicNADH-to-NAD� ratio that inhibits glu-coneogenesis, a process that is elevated intype 2 diabetes, particularly when impair-ment of glucose homeostasis is advanced.The decline in hepatic glucose productioncan provoke hypoglycemia when alcoholis ingested in the fasting state (7). Becauseethanol does not appear to affect insulinsecretion or glucose disposal directly, ahypoglycemic effect of ethanol is likely tobe highly dependent on nutritional state(8). In several small, short-term studies of5–20 patients with type 2 diabetes, a de-crease in plasma glucose concentrations(9–11) with moderate alcohol consump-tion was reported. However, other studiesshowed no effect of alcohol on glycemiccontrol (12–14). Inhibition of hepaticglucose production is the major therapeu-tic effect of established antidiabetic med-ications, such as metformin, so thepotential impact of moderate alcohol con-sumption on glycemic control in diabeticsubjects remains intriguing, but un-proven. Therefore, we conducted a3-month multicenter randomized con-trolled intervention study of alcohol (150

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1S. Daniel Abraham International Center for Health and Nutrition, Department of Epidemiology,Faculty of Health Sciences, Ben-Gurion University of the Negev, Beer-Sheva, Israel; the 2Diabetes Unit,Wolfson Medical Center, Holon, Israel; the 3Department of Internal Medicine C and the Diabetes Unit,Soroka University Medical Center, Beer-Sheva, Israel; the 4Diabetes Unit, Hadassah Hebrew UniversityMedical Center, Jerusalem, Israel; the 5S. Daniel Abraham International Center for Health and Nutrition,Department of Biochemistry, Faculty of Health Sciences, Ben-Gurion University of the Negev, Beer-Sheva,Israel; and the 6Channing Laboratory, Department of Medicine, Brigham and Women’s Hospital and HarvardMedical School and Departments of Epidemiology and Nutrition, Harvard School of Public Health, Boston,Massachusetts.

Address correspondence and reprint requests to Iris Shai, RD, PhD, S. Daniel Abraham InternationalCenter for Health and Nutrition, Department of Epidemiology and Health Systems Evaluation, Ben-GurionUniversity of the Negev, P.O. Box 653, Beer-Sheva 84105, Israel. E-mail: [email protected].

Received for publication 11 June 2007 and accepted in revised form 8 September 2007.Published ahead of print at http://care.diabetesjournals.org on 11 September 2007. DOI: 10.2337/dc07-

1103. Clinical trial reg. no. NCT00295334, clinicaltrials.gov.I.S. and J.W. contributed equally to this work.Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/

dc07-1103.Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; FPG, fasting plasma

glucose.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





ml wine; 13 g alcohol/day) or a controlnonalcoholic beer among 109 patientswith type 2 diabetes who had abstainedfrom alcohol and assessed the effect onfasting and postprandial glycemia.

RESEARCH DESIGN ANDMETHODS — We enrolled patientsfrom three diabetes units in academicmedical centers in Israel (Hadassah He-brew University Medical Center, Jerusa-lem; Wolfson Medical Center, Holon; andSoroka University Medical Center, Beer-Sheva). Inclusion criteria were 1) estab-lished diagnosis of type 2 diabetes, 2)abstinence from alcohol (not �1 drink/week), 3) age between 40 and 75 years,and 4) clinically stable condition, with nohistory of stroke or myocardial infarctionor major surgery within the previous 3months. Exclusion criteria were 1) �2 in-sulin injections/day or insulin pump ther-apy, 2) triglycerides �500 mg/dl, 3) A1C�10%, 4) serum creatinine �2 mg/dl, 5)liver dysfunction (�2-fold elevation ofalanine aminotransferase [ALT] or aspar-tate aminotransferase [AST]), 6) evidenceof severe diabetes complications (such asproliferative retinopathy or overt ne-phropathy), 7) autonomic neuropathymanifested as postural hypotension or hy-poglycemia unawareness, 8) use of drugsthat might significantly interact with alco-hol such as sedatives, antihistamines, oranticoagulants, 9) the presence of activecancer or chemotherapy within the past 3years, 10) major illness that may requirehospitalization, 11) a high potential foraddictive behavior based on physician’sassessment or personal or family historyof addiction, alcoholism, or alcoholabuse, 12) pregnancy or lactation(women), or 13) participation in anothertrial with active intervention.

The study was coordinated by the In-ternational Center for Health and Nutri-tion, Ben-Gurion University, Beer-Sheva,Israel, and was independently approvedby the institutional review boards of eachof the three medical centers. All volun-teers gave written informed consent anddid not receive compensation for theirparticipation.

We screened 201 patients with type 2diabetes, of whom 126 were eligible. Ofthese, we randomly assigned 109 and 91completed the study (see Fig. 1 of the on-line appendix available at http://dx.doi.org/10.2337/dc07-1103). Therandomization design used a 2:1 ratio (in-tervention-to-control), to obtain betterestimates of any adverse effects of the al- T

able

1—L

evel

sof

mea

sure

men

tsbe

fore

and

afte

rth

e12

-wee

kal

coho

lint

erve

ntio

n

Var

iabl

e

Alc

ohol

grou

pC

ontr

olgr

oup

Tim

e0

Tim

e12

Diff

eren

ceT

ime

0T

ime

12D

iffer

ence

FPG

(mg/

dl)

139.

57�

41.0

411

8.01

�32

.46*

�21

.55

�41

.18†

136.

71�

15.4

138.

64�

27.8

1.92

�25

.69

2h

post

mea

lglu

cose

(mg/

dl)

128.

76�

28.0

114

6.45

�33

.04

17.6

8�

34.4

013

5.55

�39

.48

150.

02�

52.5

018

.46

�66

.68

Wei

ght

(kg)

83.6

4�

15.5

783

.66

�15

.54

0.02

�1.

9381

.88

�17

.03

81.4

0�

17.1

6�

0.48

�2.

07W

aist

(cm

)98

.84

�12

.41

96.4

0�

12.5

2*�

2.44

�5.

0110

1.53

�14

.92

100.

03�

15.2

3�

1.50

�4.

67Sy

stol

icbl

ood

pres

sure

(mm

/Hg)

133.

82�

15.4

813

1.32

�16

.89

�2.

50�

14.4

712

7.73

�14

.23

127.

94�

15.6

90.

21�

16.1

4

Dia

stol

icbl

ood

pres

sure

(mm

/Hg)

77.6

6�

10.8

976

.84

�8.

76�

0.82

�11

.42

70.2

1�

9.25

72.3

1�

9.26

2.10

�9.

79

A1C

(%)

7.37

�1.

217.

07�

00.9

1‡�

0.29

�00

.99

7.08

�00

.96

6.84

�0.

75�

0.24

�00

.86

Tri

glyc

erid

es(m

g/dl

)13

5.53

�54

.48

158.

58�

85.7

823

.05

�81

.10

146.

86�

63.1

157.

26�

120.

9510

.40

�87

.86

HD

Lch

oles

tero

l(m

g/dl

)48

.95

�12

.81

46.6

1�

12.3

0�

2.34

�9.

8549

.60

�12

.68

46.6

0�

11.2

5‡�

3.00

�4.

55LD

Lch

oles

tero

l(m

g/dl

)96

.65

�29

.23

85.1

1�

28.3

1*�

11.5

3�

29.2

292

.49

�30

.62

92.6

1�

29.7

80.

12�

24.0

3Bi

lirub

in(m

g/dl

)0.

54�

0.20

0.59

�0.

260.

04�

0.18

0.61

�0.

330.

61�

0.24

0.00

2�

0.25

Alk

alin

eph

osph

atas

e(u

nits

/l)60

.34

�19

.94

56.8

4�

18.7

0�

3.50

�15

.54

62.0

0�

9.45

59.9

0�

10.1

8�

2.09

�6.

62

ALT

(uni

ts/l)

23.4

4�

10.3

033

.71

�55

.93

10.2

6�

54.0

729

.23

�15

.57

26.3

8�

11.4

2�

2.85

�6.

37A

ST(u

nits

/l)21

.23

�1.

1030

.92

�7.

699.

68�

47.4

023

.92

�8.

4223

.00

�8.

14�

0.92

�3.

79

Dat

aar

em

eans

�SD

.n�

91.*

P v�

0.01

,wit

hin-

grou

pdi

ffer

ence

,pai

red

ttes

t,co

mpa

red

wit

hti

me

0.†P

v�

0.05

,bet

wee

n-gr

oup

diff

eren

ces,

grou

ptt

est

of�

s;‡P

v�

0.05

,wit

hin-

grou

pdi

ffer

ence

.

Moderate alcohol intake and diabetes: randomized trial


cohol intervention. Participants met withthe nurse study coordinator in the diabe-tes center on eight occasions during thetrial and with the physicians and the die-titians at weeks 1, 7, and 12 (Table 1 ofthe online appendix). Three months afterthe end of the study, we interviewed par-ticipants who completed the alcohol armby telephone to assess voluntary continu-ation of alcohol consumption as well asadverse effects.

InterventionAll participants received individual di-etary counseling by registered dietitianstrained to work with type 2 diabetic pa-tients. Each dietitian reinforced identicalnutritional strategies to achieve glycemiccontrol in both study groups but did notspecifically try to promote weight loss.Reinforcement of dietary counseling forboth groups was based on the AmericanDiabetes Association recommendationsfor patients with type 2 diabetes, whichinclude 45–60% calories from carbohy-drates, up to 30% from fat (with restric-tion of saturated fat to �7% of totalcalories and minimization of trans fat),and 15–20% from protein. Caloric intakewas calculated according to age, sex, andlevel of physical activity. Based on thesecalculations, patients were instructed toconsume an isocaloric diet. To compen-sate for the calories in the assigned bever-ages, the alcohol group was instructed to

reduce carbohydrates by 100 kcal, butnot at dinner, to decrease the likelihood ofalcohol-induced hypoglycemia (6). Thecontrol group was instructed to deduct 30kcal from carbohydrates. Participantscompleted 3-day food diaries and drinkpattern questionnaires before each visit toenable the dietitians to monitor adher-ence to the diet and alcohol intake. Pa-tients assigned to consume alcohol wereinstructed to start drinking gradually(over a 2-week period) 150 ml of wine(13% alcohol, 13 g) that we provided, us-ing a standard measured glass, duringdinner. The patients could choose eitherdry red (Merlot) or white (SauvignonBlanc) wine; 75% chose red wine. Partic-ipants randomly assigned to the controlgroup were instructed to drink 150 ml ofthe nonalcoholic diet malt beer we pro-vided, using the same standard measuredglass, during dinner. Every other week,the study coordinator provided eitherthree bottles of wine (750 ml each) or twobottles of nonalcoholic diet malt beer (1.5liters each), after the return of empty bot-tles from the previous fortnight.

Blood and clinical measurementsBaseline and 12-week blood sampleswere drawn in the morning, after an 8-hfast. All biochemical determinations wereperformed in the central laboratories ofthe medical centers using Olympus ana-lytical equipment and reagents. LDL cho-

lesterol was calculated by the Friedewaldformula (15). A1C was determined usingCOBAS INTEGRA reagents and analyticalequipment. A value �5.8% is considerednormal. Blood pressure was measuredwith the subject sitting, after 5 min of rest,using an Omron M41 digital apparatus.Waist circumference was measured half-way between the last rib and the iliaccrest. The patients were instructed tomeasure glucose, preprandially and 2 hpostprandially at dinnertime, three timesweekly using their own self-glucose mon-itoring device. The glucometers usedwere Accutrend Sensor (Roche Diagnos-tics), Elite (Bayer Diagnostics), or Free-style (Thera Sense, Alameda, CA).

Statistical analysisWe used �2 analyses to determine differ-ences between categorical variables andpaired t tests to compare changes in mea-surements within the two groups. In themain analyses, we compared the mean ofthe individual changes, from baseline to12 weeks, in the two arms of the trial. Wealso calculated age-adjusted correlationsand performed interaction tests betweenthe intervention groups and strata of sex,median BMI, and median age. The levelsof individual postmeal glucose representan average of three reports, taken 2 h afterdinner in the same week. We comparedthe proportion of positive responders inboth groups to the following question:“Do you think that, since the beginning ofthis study, the addition of the drink toyour dinner was associated with an in-crease in the following symptoms/adverseeffects?” Statistical analyses were per-formed using SPSS software (version14.0).

RESULTS — The randomly assignedpatients, 61 men and 48 women, rangedin age from 41 to 74 years, had an averagefasting plasma glucose (FPG) level of144.5 mg/dl, A1C of 7.39%, blood pres-sure of 133.7/76.5 mm/Hg, and BMI of30.1 kg/m2. These characteristics weresimilarly distributed between the ran-domized groups, as were other parame-ters such as duration of the disease,smoking status, physical activity, regularconsumption of nutritional supplements,waist circumference, and years of educa-tion (Table 2 of the online appendix). Af-ter random assignment but before theintervention began, 12 participants with-drew from the trial, 5 (7%) from the alco-hol group and 7 (21%) from the controlgroup. During the intervention, 4 addi-

Figure 1—Individual changes in FPG and 2-h postmeal glucose after 12 weeks of moderatealcohol intervention among patients with type 2 diabetes. Vertical lines indicate means � SD.

Shai and Associates


tional participants withdrew from the al-cohol group and 2 from the controlgroup. The 18 patients who withdrewfrom the study (12% of the interventionand 26% of the control groups) weregenerally similar to the 91 patients whocompleted the study, but they had sig-nificantly higher baseline levels of FPG(167 vs. 140 mg/dl) and were younger(Table 3 of the online appendix).

The individual changes in FPG and2-h postmeal glucose between baselineand at the end of the trial are shown in Fig.1. The alcohol group experienced a signif-icant 9.2% decrease in FPG levels, drop-ping from 139.6 � 41 at baseline to118.0 � 32.5 mg/dl after 3 months (Pv �0.001), whereas there was no materialchange in FPG levels in the control group(136.7 � 15.4 at baseline and 138.6 �27.8 mg/dl at week 12, Pv � 0.783). Thedifference between the groups was signif-icant (Pv � 0.015). The postprandial val-ues represent an average of three self-measurements that were taken afterdinner, at baseline, and during weeks 11–12. We observed nonsignificant increasesin the 2-h postmeal glucose levels of sim-ilar magnitude in both groups (18.5 in thecontrol group vs. 17.7 in the alcoholgroups, P for difference � 0.97). Withinthe alcohol group, but not among controlsubjects (Fig. 2), we found a significantinverse correlation between baseline lev-els of A1C and changes of FPG levels (age-adjusted correlation �0.567, Pv �0.001), suggesting that patients with type2 diabetes with higher baseline A1C levelshad greater reductions in FPG after mod-erate alcohol consumption. We found nomodification of the alcohol effect by sex,age, BMI, or specific medical center (datanot shown), although the statistical powerto observe such interactions was limited.

We observed significant decreases(Table 2 of the online appendix) in levelsof A1C, LDL cholesterol, and waist cir-cumference in the alcohol group and anunexpected significant reduction in HDLcholesterol in the control group after 12weeks compared with baseline levels.However, none of these changes differedsignificantly between the two groups. Wefound no significant changes in weight,blood pressure, or triglycerides amongpatients in either group and no materialchanges in levels of bilirubin, alkalinephosphatase, ALT, or AST.

We elicited reports of symptoms (Fig.2 of the online appendix) that participantsattributed to the intervention. In the alco-hol group, one woman dropped out be-

cause of gastric pain and 5% reportedepisodes of hypoglycemia, headaches, ormuscle weakness, symptoms that werenot reported in the control group. Noother adverse effects were reported. Par-ticipants in the alcohol group (8%) butnone in the control group reported in-creased sexual desire. The only item thatdiffered significantly was improved abil-ity to fall asleep in the alcohol group com-pared with control subjects (Pv � 0.001).

Three months after the study ended,61% of the participants in the alcoholgroup reported that they thought that thealcohol was beneficial to them, and 49%reported continuing to drink alcohol inmoderation (frequency ranging from onedrink a week to one drink a day). Nonereported an increase of the quantity of al-cohol consumed.

CONCLUSIONS — In the presentrandomized trial among patients withtype 2 diabetes who had previously ab-stained from alcohol, we showed thatmoderate alcohol consumption signifi-cantly decreased fasting but not postpran-dial glucose levels. Those with higherbaseline A1C levels appeared to benefitmore. Initiating moderate daily alcoholconsumption in type 2 diabetic patients

aged �40 years who had previously beenabstainers caused no notable adverse ef-fects or changes in liver function biomar-kers during the 3-month intervention.

Several limitations of the study war-rant consideration. Neither the partici-pants nor the diabetes clinic staff could beblinded to the intervention (alcohol ver-sus control nonalcoholic beverage), andalthough adherence was good, the drop-out rates were not negligible. However,participants who dropped out generallyhad clinical profiles similar to those ofparticipants completing the study, and, infact, two-thirds of the dropouts occurredimmediately after the randomization andbefore the intervention began, and rateswere higher among control subjects, sug-gesting that adverse effects of alcoholcaused few if any dropouts. The 3-monthintervention period, although longer thanmost alcohol trials, could not capture allof the possible long-term adverse or ben-eficial effects of alcohol, limiting our abil-ity to draw conclusions about the long-term risks and benefits. We believe ourinclusion criteria (particularly age �40years and screening for past addictive be-havior) largely limited the danger of pro-moting alcohol addictive behavior. In atelephone interview 3 months after the

Figure 2—Correlations between baseline levels of A1C and change of FPG levels after 12 weeksof moderate alcohol intervention among patients with type 2 diabetes. *Age-adjusted correlationamong the alcohol group.



end of the trial, all participants reportedalcohol consumption of one drink a dayor less. The alcohol dose of 13 g/day maybe a less than optimal dose to achievemaximal effects in patients with type 2diabetes. Red and white wine presumablycontain different amounts of polyphe-nols, possibly confounding the effects ofthe alcohol per se. Finally, we assayedfasting and postmeal glucose levels andA1C but have no data on levels of insulinand glucagon, degree of insulin resis-tance, or hepatic glucose output. Thislack of data limits our ability to dissect outthe relevant importance of mechanismsmediating the alcohol-induced decreasein FPG and the differential effects on FPGand postprandial glucose.

There are several strengths to thisstudy: The number of participants islarger than that of most other alcohol in-tervention trials and adherence to the in-tervention protocols was high. Mostimportantly, the randomized trial designpermitted assessment of the independenteffect of initiating moderate alcohol con-sumption in abstainers. Nutritional coun-seling to both groups of participants wasadjusted for the added calories, but newdietary instructions aimed to promoteweight loss were not introduced.

Moderate alcohol consumption hasbeen associated with a lower risk of car-diovascular disease (16) and type 2 diabe-tes (2,17). The apparent beneficial effectsfor cardiovascular disease are probablymediated via effects on lipid metabolism(18), coagulation, fibrinolysis (19), andinsulin sensitivity (20,21). We have pre-viously shown (22) that among �700men with type 2 diabetes, moderate alco-hol intake was associated with decreasinglevels of inflammatory biomarkers (solu-ble tumor necrosis factor receptor-2, sol-uble intercellular adhesion molecule-1,and fibrinogen) as well as elevated circu-lating levels of adiponectin. Prospectivestudies showed an inverse relationshipbetween alcohol consumption and diabe-tes incidence, with moderate drinkershaving a 43–46% reduction in risk fordiabetes compared with nondrinkers(23–25). In addition, alcohol is linked tolower cardiovascular risk among patientswith type 2 diabetes (26). In a recentmeta-analysis of cohort studies among pa-tients with diabetes (4), alcohol consum-ers had a 21–36% lower total mortalityrate and a 25–66% lower rate of total andfatal coronary heart disease than abstain-ers. The magnitude of these associations is

stronger than that seen in the general pop-ulation (27).

Although beneficial effects of moder-ate alcohol consumption have beenstrongly suggested by observational stud-ies, data from randomized trials of alco-hol, especially among patients with type 2diabetes, are sparse. In a randomized con-trolled crossover trial (28) of 63 healthypostmenopausal women over 8 weeks,consumption of 30 g/day of alcohol (twodrinks per day) reduced insulin and tri-glyceride concentrations and improvedinsulin sensitivity in these nondiabeticwomen, but fasting glucose concentra-tions were not materially affected. In a trialamong patients with diabetes after a firstmyocardial infarction (29), red wine takenwith meals significantly reduced oxidativestress and proinflammatory cytokines.

The major glycemic effect in our trialwas a decrease in fasting, but not post-meal, plasma glucose levels. The mecha-nisms for this effect probably involveenhanced insulin secretion (3) and thewell-documented effect of alcohol metab-olism, which, by increasing the hepaticcytosolic NADH-to-NAD� ratio, inhibitsgluconeogenesis, a process largely con-trolling fasting, rather than postmeal, gly-cemia. The nonsignificant increase ofpostprandial glucose levels could be aconsequence of increased consumption ofsimple carbohydrates in the eveningmeal. The contribution of increased fluxthrough the gluconeogenesis pathway tohyperglycemia is a characteristic of dys-regulated glucose homeostasis in diabe-tes. Thus, it is plausible that patients withhigher A1C have elevated gluconeogenicflux and, hence, exhibit more pro-nounced fasting hypoglycemic effectswhen moderate alcohol consumption isstarted.

In our study, patients in the alcoholgroup significantly reduced their waistcircumference and LDL cholesterol andA1C levels, but these changes were notstatistically significant compared with thechange in these parameters in the placebogroup. Intriguingly, we observed that di-abetic subjects consuming 13 g alcoholdaily for 3 months showed no increase inHDL. The likely explanations for this ob-servation are related to the alcohol dose orduration or to unique characteristics ofthe study population. Significant in-creases in HDL could be observed inhealthy men as early as 17 days after ini-tiation of 40 g/day of alcohol (30). Alter-natively, it is possible that the HDL-elevating effect of alcohol is less readily

detectable among diabetic subjects, par-ticularly when they are also treated withglucose- and lipid-lowering medications.This notion is supported by observationsmade during a previously mentioned trialamong diabetic subjects after a myocar-dial infarction, in which a significant in-crease in HDL was observed only after 9months of alcohol intake (R. Marfella,personal communication). Thus, in thediabetic population, alcohol apparentlyexerts a more rapid glucose-lowering ef-fect, whereas the elevation in HDL re-quires more prolonged intervention. Indoses shown in epidemiological studies toconfer cardiovascular disease and glyce-mic benefits, not all metabolic changes at-tributed to alcohol can be captured within3 months in patients with type 2 diabetes.Longer intervention studies are needed todetermine the long-term efficacy andsafety of initiating moderate alcohol in-take among abstainers with type 2 diabe-tes, with assessment of clinical orintermediate outcomes.

Acknowledgments— We thank T i shb iWines, Israel, and Admiral Wine Imports,U.S., for providing the wine for this study. Wethank the following physicians, dietitians,nurses, and researchers for their valuable con-tributions to the study: Dr. Joseph Glassman,Dr. Mariella Glant, Orit Shemesh, and EtiAbutbul of Hadassah Medical Center; Dr. LeaChananshvili, Dr. Gila Dovinski, Dr. Lisy Lud-mila, Naomi Mor, Tami Uzer, and NaomiMevorach of Wolfson Medical Center; Dr. Ta-tiana Shuster, Dr. Natalya Shapiro, Dr. IditLiberty, Dr. Max Mayzlus, and Shula Witkowof Soroka University Medical Center; Prof.Shimon Weitzman, Prof. Yaakov Henkin,Rachel Golan, and Osnat Tanji-Rozental ofBen-Gurion University.

References1. Matthew SF, Samet JH: Alcohol and cor-

onary heart disease: the answer awaits arandomized controlled trial. Circulation112:1379–1381, 2005

2. Stampfer MJ, Colditz GA, Willett WC,Manson JE, Arky RA, Hennekens CH,Speizer FE: A prospective study of mod-erate alcohol drinking and risk of diabetesin women. Am J Epidemiol 128:549–558,1988

3. Koppes Ll, Dekker JM, Hendriks HF,Bouter LM, Heine RJ: Moderate alcoholconsumption lowers the risk of type 2: ameta-analysis of prospective observa-tional studies. Diabetes Care 28:719–725,2005

4. Howard AA, Arnsten JH, Gourevitch MN:Effect of alcohol consumption on diabetes

Shai and Associates


mellitus: a systematic review. Ann InternMed 140:211–219, 2004

5. Koppes LL, Dekker JM, Hendriks HF,Bouter LM, Heine RJ: Meta-analysis ofthe relationship between alcohol con-sumption and coronary heart diseaseand mortality in type 2 diabetic pa-tients. Diabetologia 49:648 – 652, 2006

6. UK Prospective Diabetes Study (UKPDS)Group. Intensive blood-glucose controlwith sulphonylureas or insulin comparedwith conventional treatment and risk ofcomplications in patients with type 2 di-abetes (UKPDS 33). Lancet 352:837–853,1998

7. van de Wiel A: Diabetes mellitus and al-cohol. Diabetes Metab Res Rev 20:263–267, 2004

8. Arky RA, Veverbrand E, Abramson EA:Irreversible hypoglycemia: a complica-tion of alcohol and insulin. JAMA 206:575–578, 1968

9. Koivisto VA, Tulokas S, Toivonen M,Haapa E, Pelkonen R: Alcohol with a mealhas no adverse effects on postprandialglucose homeostasis in diabetic patients.Diabetes Care 16:1612–1614, 1993

10. Walsh CH, O’Sullivan DJ: Effect of mod-erate alcohol intake on control of diabe-tes. Diabetes 23:440–442, 1974

11. Burge MR, Zeise TM, Sobhy TA, RassamAG, Schade DS: Low-dose ethanol predis-poses elderly fasted patients with type 2diabetes to sulfonylurea-induced lowblood glucose. Diabetes Care 22:2037–2043, 1999

12. McMonagle J, Felig P: Effects of ethanolingestion on glucose tolerance and insulinsecretion in normal and diabetic subjects.Metabolism 24:625–632, 1975

13. Christiansen C, Thomsen C, RasmussenO, Hauerslev C, Balle M, Hansen C, Her-mansen K: Effect of alcohol on glucose,insulin, free fatty acid and triacylglycerolresponses to a light meal in non-insulin-dependent diabetic subjects. Br J Nutr 71:449–454, 1994

14. Christiansen C, Thomsen C, RasmussenO, Glerup H, Berthelsen J, Hansen C, Or-skov H, Hermansen K: Acute effects ofgraded alcohol intake on glucose, insulin

and free fatty acid levels in non-insulin-dependent diabetic subjects. Eur J ClinNutr 47:648–652, 1993

15. Friedewald WT, Levy RI, Fredrickson DS:Estimation of the concentration of low-density lipoprotein cholesterol in plasma,without use of the preparative ultracentri-fuge. Clin Chem 18:499–502, 1972

16. Rimm EB, Williams P, Fosher K, CriquiM, Stampfer MJ: Moderate alcohol intakeand lower risk of coronary heart disease:meta-analysis of effects on lipids and hae-mostatic factors. BMJ 319:1523–1528,1999

17. Carlsson S, Hammar N, Grill V, Kaprio J:Alcohol consumption and the incidenceof type 2 diabetes: a 20-year follow-up ofthe Finnish Twin Cohort Study. DiabetesCare 26:2785–2790, 2003

18. Clevidence BA, Reichman ME, Judd JT,Muesing RA, Schatzkin A, Schaefer EJ, LiZ, Jenner J, Brown CC, Sunkin M, Camp-bell WS, Taylor PR: Effects of alcoholconsumption on lipoproteins of pre-menopausal women: a controlled dietstudy. Arterioscler Thromb Vasc Biol 15:179 –184, 1995

19. Pellegrini N, Pareti FI, Stabile F, Brusamo-lino A, Simonetti P: Effects of moderateconsumption of red wine on platelet ag-gregation and haemostatic variables inhealthy volunteers. Eur J Clin Nutr 50:209–213, 1996

20. Kiechl S, Willeit J, Poewe W, Egger G,Oberhollenzer F, Muggeo M, Bonora E:Insulin sensitivity and regular alcoholconsumption: large, prospective, crosssectional population study (Bruneckstudy). BMJ 313:1040–1044, 1996

21. Cooper DE, Goff DC, Bell RA, Zaccaro D,Mayer-Davis EJ, Karter AJ: Is insulin sen-sitivity a causal intermediate in the rela-tionship between alcohol consumptionand carotid atherosclerosis? The insulinresistance and atherosclerosis study. Dia-betes Care 25:1425–1431, 2002

22. Shai I, Rimm EB, Schulze MB, Rifai N,Stampfer MJ, Hu FB: Moderate alcohol in-take and markers of inflammation and en-dothelial dysfunction among diabeticmen. Diabetologia 47:1760–1767, 2004

23. Conigrave KM, Hu BF, Camargo CA Jr,Stampfer MJ, Willett WC, Rimm EB: Aprospective study of drinking patterns inrelation to risk of type 2 diabetes amongmen. Diabetes 50:2390–2395, 2001

24. Ajani UA, Hennekens CH, Spelsberg A,Manson JE: Alcohol consumption andrisk of type 2 diabetes mellitus among USmale physicians. Arch Intern Med 160:1025–1030, 2000

25. Hu FB, Manson JE, Stampfer MJ, ColditzG, Liu S, Solomon CG, Willett WC: Diet,lifestyle, and the risk of type 2 diabetesmellitus in women. N Engl J Med 345:790–797, 2001

26. Solomon CG, Hu FB, Stampfer MJ, Cold-itz GA, Speizer FE, Rimm EB, Willett WC,Manson JE: Moderate alcohol consump-tion and risk of coronary heart diseaseamong women with type 2 diabetes mel-litus. Circulation 102:494–499, 2000

27. Bagnardi V, Zambon A, Quatto P, CorraoG: Flexible meta-regression functions formodeling aggregate dose-response data,with an application to alcohol and mor-tality. Am J Epidemiol 159:1077–1086,2004

28. Davies MJ, Baer DJ, Judd JT, Brown ED,Campbell WS, Taylor PR: Effects of mod-erate alcohol intake on fasting insulin andglucose concentrations and insulin sensi-tivity in postmenopausal women: a ran-domized controlled trial. JAMA 287:2559–2562, 2002

29. Marfella R, Cacciapuoti F, Siniscalchi M,Sasso FC, Marchese F, Cinone F, Musac-chio E, Marfella MA, Ruggiero L, Chi-orazzo G, Liberti D, Chiorazzo G,Nicoletti GF, Saron C, D’Andrea F, Am-mendola C, Verza M, Coppola L: Effect ofmoderate red wine intake on cardiacprognosis after recent acute myocardialinfarction of subjects with type 2 diabetesmellitus. Diabet Med 23:974–981, 2006

30. Beulens JW, Sierksma A, van Tol A,Fournier N, van Gent T, Paul JL, HendriksHF: Moderate alcohol consumption in-creases cholesterol efflux mediated byABCA1. J Lipid Res 45:1716–1723, 2004



Management of Type 2 Diabetes inTreatment-Naive Elderly PatientsBenefits and risks of vildagliptin monotherapy

RICHARD E. PRATLEY, MD1

JULIO ROSENSTOCK, MD2

F. XAVIER PI-SUNYER, MD3

MARY ANN BANERJI, MD4

ANJA SCHWEIZER, PHD5

ANDRE COUTURIER, MSC6

SYLVIE DEJAGER, MD, PHD6

OBJECTIVE — The purpose of this study was to evaluate the efficacy and safety of vildaglip-tin in elderly patients with type 2 diabetes.

RESEARCH DESIGN AND METHODS — Efficacy data from five double-blind, ran-domized, placebo- or active-controlled trials of �24 weeks’ duration were pooled. Effects of24-week vildagliptin monotherapy (100 mg daily) were compared in younger (�65 years, n �1,231) and older (�65 years, n � 238) patients. Safety data from eight controlled clinical trialsof �12-weeks’ duration were pooled; adverse event profiles in younger (n � 1,890) and older(n � 374) patients were compared.

RESULTS — Mean baseline A1C and fasting plasma glucose (FPG) were significantly lower inolder (70 years: 8.3 � 0.1% and 9.6 � 0.1 mmol/l, respectively) than in younger (50 years: 8.7 �0.0% and 10.5 � 0.1 mmol/l, respectively) patients. Despite this, the adjusted mean change frombaseline (AM�) in A1C was �1.2 � 0.1% in older and �1.0 � 0.0% in younger vildagliptin-treated patients (P � 0.092), and the AM� in FPG was significantly larger in older (�1.5 � 0.2mmol/l) than in younger (�1.1 � 0.1 mmol/l, P � 0.035) patients. Body weight was significantlylower at baseline in older (83.4 � 1.0 kg) than in younger (92.0 � 0.6 kg) patients. Weightdecreased significantly in the older subgroup (AM� �0.9 � 0.3 kg, P � 0.007), whereas smaller,nonsignificant decreases occurred in younger patients (AM� �0.2 � 0.1 kg). Adverse eventrates were slightly higher in older than in younger subgroups but were lower among older,vildagliptin-treated subjects (63.6%) than in the pooled active comparator group (68.1%).Vildagliptin treatment did not increase adverse events among older patients with mild renalimpairment (62.0%). Hypoglycemia was rare (0.8%) in the elderly patients, and no severe eventsoccurred.

CONCLUSIONS — Vildagliptin monotherapy was effective and well tolerated in treatment-naive elderly patients.

Diabetes Care 30:3017–3022, 2007

T ype 2 diabetes is among the mostcommon chronic conditions inolder adults. Nearly 20% of individ-

uals aged �65 years are affected, al-though in nearly half of them diabetes isundiagnosed (1). Management of type 2diabetes in elderly individuals can be par-ticularly challenging for a number of rea-sons (2). First, hypoglycemia is morecommon in older than in younger peopletaking oral antidiabetic drugs (OADs), isoften more severe, and can precipitate se-rious events such as falls and hip frac-tures. This higher incidence is due in partto higher rates of conditions such as de-pression, cognitive dysfunction, poor ap-petite, and irregular eating habits thatpredispose to hypoglycemia. Age-associated abnormalities in counterregu-lation (3) can also impair the patient’sability to recognize and respond to hypo-glycemia. Second, elderly patients withtype 2 diabetes have a high prevalence ofcomorbidities (4) and, accordingly, con-comitant use of multiple medications isvery common. Further, undiagnosed re-nal impairment may be present in �50%of elderly patients with type 2 diabetes(4). These issues may limit therapeuticchoices and can lead to inappropriate, lessaggressive treatment goals. Thus, fewerthan half of patients aged �65 yearsachieve recommended levels of glycemiccontrol (A1C �7.0%) (5). Collectively,these data highlight a substantial unmetmedical need for safe and effective thera-peutic agents for elderly patients withtype 2 diabetes.

Vildagliptin is a potent and selectivedipeptidyl peptidase IV (DPP-4) inhibitorthat improves glycemic control in patientswith type 2 diabetes through incretin-hormone–mediated increases in both �-and �-cell responsiveness to glucose (6).In studies enrolling OAD-naive patientswith type 2 diabetes, 24 weeks’ treatmentwith vildagliptin monotherapy (50 or 100mg daily) was reported to decrease A1Cby 0.9–1.1% (7,8).

Because the effects of incretin hor-mones to increase insulin secretion (9)and of glucagon-like peptide-1 (GLP-1)to suppress glucagon secretion (10) are

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Vermont College of Medicine, Burlington, Vermont; the 2Dallas Diabetes and Endocrine Center,Dallas, Texas; 3St. Lukes-Roosevelt Hospital, New York, New York; 4State University of New York DownstateMedical Center, Brooklyn, New York; 5Novartis Pharma AG, Basel, Switzerland; and 6Novartis Pharmaceu-ticals Corporation, East Hanover, NJ.

Address correspondence and reprint requests to Anja Schweizer, PhD, Novartis Pharma AG, Postfach,CH-4002 Basel, Switzerland. E-mail: [email protected].

Received for publication 21 June 2007 and accepted in revised form 12 September 2007.Published ahead of print at http://care.diabetesjournals.org on 18 September 2007. DOI: 10.2337/dc07-

1188. Clinical trial reg. nos. NCT00099905, NCT00099866, NCT00099918, NCT00101673,NCT00101803, and NCT00120536, clinicaltrials.gov.

R.E.P. has received research grants and consulting fees from Novartis. J.R. has received research grants andconsulting fees from Novartis. F.X.P.-S. has received research grants and consulting fees from Novartis.M.A.B. has received grant support and honoraria from Novartis.

Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/dc07-1188.

Abbreviations: AM�, adjusted mean change; DPP-4, dipeptidyl peptidase IV; FPG, fasting plasma glu-cose; GFR, glomerular filtration rate; GLP-1, glucagon-like peptide-1; OAD, oral antidiabetic drug; SAE,serious adverse event.

A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversionfactors for many substances.

© 2007 by the American Diabetes Association.The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby




glucose dependent, DPP-4 inhibitorssuch as vildagliptin are associated with avery low risk of hypoglycemia. Further,experience thus far with vildagliptin indi-cates that it is well tolerated, as demon-strated in placebo-controlled (7,11) andactive-controlled studies with metformin(12) and thiazolidinediones (8,13).Hence, vildagliptin appears to possessmany characteristics that could make it auseful therapeutic option for treatment oftype 2 diabetes in elderly individuals.

The purpose of the present analysiswas to ascertain the efficacy and tolerabil-ity of vildagliptin monotherapy in elderlypatients with type 2 diabetes. Thus, datafrom vildagliptin monotherapy trials werepooled, and the efficacy and safety ofvildagliptin in patients aged �65 yearswere compared with those in patients�65 years of age.

RESEARCH DESIGN ANDMETHODS — Studies were multi-center, randomized, double-blind, paral-lel-group, placebo- or active-controlledtrials of 12–52 weeks’ duration, with oneor more vildagliptin monotherapy arms.Twenty-four-week efficacy data from allcompleted trials of �24 weeks’ durationwere pooled (patients receiving 100 mgvildagliptin daily as monotherapy, either50 mg b.i.d. or 100 mg q.d.) from twoplacebo-controlled and three active-controlled studies (n � 1,469). To pro-vide the most comprehensive informationavailable, safety data from all completedtrials of �12 weeks’ duration (i.e., theaforementioned five trials, two placebo-controlled 12-week studies, and one ac-tive-controlled 12-week study) werepooled from patients receiving 50 mgq.d., 50 mg b.i.d., or 100 mg q.d. vilda-gliptin (n � 2,264), all active comparators(up to 1,000 mg b.i.d. metformin, 30 mgq.d. pioglitazone, or 8 mg q.d. rosiglita-zone, n � 735), and placebo (n � 347).Details about study designs and inclusionand exclusion criteria are summarized inTable A1 of the online appendix (availableat http://dx.doi.org/10.2337/dc07-1188)and are also provided in the individualstudy publications (7,8,11–13).

Study assessmentsA1C, fasting plasma glucose (FPG), bodyweight, fasting lipid levels (triglyceridesand total, LDL, HDL, non-HDL, andVLDL cholesterol), and sitting systolicand diastolic blood pressure were mea-sured periodically, and the changes frombaseline to week 24 are reported as effi-

cacy parameters. Changes in A1C werealso assessed in the prespecified sub-groups of patients with lower (�8.0 or�9.0%) and higher (�8.0 or �9.0%)baseline A1C levels and of patients withlower (�30 or �35 kg/m2) and higher(�30 or �35 kg/m2) baseline BMI.Changes in body weight were assessed inthe same BMI subgroups. In addition, re-sponder analyses were performed to de-termine the percentage of patientsachieving A1C �7.0% in the overall pop-ulation and in the prespecified subgroupsof patients with a baseline A1C �8%.

Glomerular filtration rate (GFR) wasestimated with the Modification of Diet inRenal Disease study method (14), and pa-tients were classified according to criteriapreviously specified in guidelines pub-lished by the Food and Drug Administra-tions into a group with normal renalfunction (GFR �80 ml/min 1.73/m2)and a group with mild renal impairment(GFR �80 and �50 ml/min 1.73/m2).All adverse events were recorded and as-sessed by the investigator as to the sever-ity and possible relationship to the studymedication. Patients were provided withglucose monitoring devices and suppliesand instructed on their use. Hypoglyce-mia was defined as symptoms suggestiveof low blood glucose, confirmed by self-monitoring of blood glucose measure-ment of �3.1 mmol/l plasma glucoseequivalent. Severe hypoglycemia was de-fined as any episode requiring the assis-tance of another party.

All laboratory assessments were per-formed by central laboratories: Bioana-lytical Research Corporation-US (LakeSuccess, NY), Bioanalytical Research Cor-poration-EU (Ghent, Belgium), DiabetesDiagnostics Laboratory (Columbia, MO),Covance-US (Indianapolis, IN), or Medi-cal Research Laboratories International(Zaventem, Belgium). A1C was measuredby high-performance liquid chromatog-raphy (ion exchange or boronate affinity).All laboratories were either National Gly-cohemoglobin Standardization Programcertified or National GlycohemoglobinStandardization Program network labora-tories, thus allowing traceability to the Di-abetes Control and Complications Trialreference method of A1C measurement.Data analysisThe safety population comprised all pa-tients receiving vildagliptin monotherapy(50 or 100 mg daily) for whom at leastone postbaseline safety assessment wasavailable. The efficacy population com-prised all patients receiving vildagliptin

(100 mg daily: 50 mg b.i.d. or 100 mgq.d.) for whom both a baseline and post-baseline efficacy assessment were avail-able. Changes from baseline in efficacyparameters were analyzed using anANCOVA model containing treatment,study, age-group, treatment age-groupinteraction, and baseline value as a covari-ate. Within-group comparisons (endpoint versus baseline) and between-group comparisons (patients aged �65years vs. patients aged �65 years) weremade using two-sided tests at a signifi-cance level of 0.05. Safety data are sum-marized for the overall safety populationand for the younger and older subgroups;statistical comparisons of safety data werenot made.

Ethics and good clinical practiceAll participants provided written informedconsent. All protocols were approved by theindependent ethics committee/institutionalreview board at each study site. All studieswere conducted using good clinical practiceand in accordance with the Declaration ofHelsinki.

RESULTS — Table A2 of the online ap-pendix summarizes the baseline anthro-pometric and disease characteristics of theoverall population and of the younger(mean age 50 years) and older (meanage 70 years) subgroups of patients inthe safety population. Patients aged �65years represented 17% of the pooledsafety database. The majority of all pa-tients were Caucasian and obese, withmean A1C of 8.6% and mean FPG of 10.1mmol/l. Minorities represented a largerproportion of the younger subgroup,whereas the older subgroup was on aver-age less obese (with only about half theprevalence of severe obesity than theyounger subgroup) and had better glyce-mic control while receiving no OAD, de-spite a somewhat longer mean diseaseduration. More than 85% of the oldersubgroup had one or more additional car-diovascular risk factors (vs. 62% of theyounger subgroup), and nearly two-thirds of the older patients had undiag-nosed mild renal impairment (vs. 28%of the younger subgroup). More than75% of the older subgroup had hyperten-sion, about half had dyslipidemia, andnearly 25% had coronary artery disease,whereas these conditions were, as ex-pected, much less prevalent in theyounger subgroup. Furthermore, the el-derly patients were taking an average of9.8 concomitant medications at study en-

Vildagliptin in elderly patients


rollment compared with 4.4 in theyounger subgroup, so twice as many el-derly patients were taking �5 concomi-tant medications as their youngercounterparts. The baseline characteristicsof patients in the efficacy population weresimilar to those of patients in the safetypopulation.

EfficacyTable 1 summarizes all efficacy parame-ters, responder analyses, and subgroupanalyses of A1C and body weight in theoverall efficacy population and in theyounger and older subgroups. In theoverall population, vildagliptin signifi-cantly decreased A1C by 1.0% from amean baseline of 8.6%. The decrease in

the elderly subgroup (adjusted meanchange [AM�] �1.2%) tended to begreater (P � 0.092) than that in theyounger subgroup (AM� �1.0%) despitehaving a significantly lower baseline A1C(8.3 vs. 8.7%). Because the majority of theavailable data derived from active-controlled trials, there were only 26 el-derly patients receiving placebo. BaselineA1C was 8.2 � 0.2% in these patientswith AM� of �0.5 � 0.3%, and a similarreduction (0.3 � 0.1%) was also seen inthe younger subgroup (n � 156), drivenprimarily by a single study (11).

For FPG, the difference betweenolder and younger vildagliptin-treatedpatients achieved statistical significance.In the older subgroup, vildagliptin de-

creased FPG by a significantly greater de-gree (AM� �1.5 mmol/l, P � 0.035)from a significantly lower baseline value(9.6 vs. 10.5 mmol/l).

In the elderly subgroup, 47% of thepatients achieved the American DiabetesAssociation recommended target A1C(�7.0%) versus 36% of the younger sub-group (P � 0.002 for younger versusolder). In patients with baseline A1C�8.0% (mean of 7.6% in both sub-groups), the percentage of patientsachieving the target was also significantlygreater for the elderly (63%) than for theyounger patients (52%, P � 0.034younger versus older).

Vildagliptin did not significantly af-fect body weight relative to baseline in the

Table 1—Efficacy parameters in patients receiving vildagliptin (100 mg daily)

All Aged �65 years Aged �65 years

BL AM� BL AM� BL AM�

n 1,469 1,231 238A1C (%) 8.6 � 0.0 �1.0 � 0.0* 8.7 � 0.0 �1.0 � 0.0* 8.3 � 0.1† �1.2 � 0.1*FPG (mmol/l) 10.4 � 0.1 �1.1 � 0.1* 10.5 � 0.1 �1.1 � 0.1* 9.6 � 0.1† �1.5 � 0.2*†Body weight (kg) 90.6 � 0.5 �0.3 � 0.1 92.0 � 0.6 �0.2 � 0.1 83.4 � 1.0† �0.9 � 0.3*

Responder analyses(achieving A1C �7.0%)

n‡ n (%)responders



Overall 1,462 548 (37.5) 1,226 438 (35.7) 236 110 (44.6)†Baseline A1C �8.0% 526 286 (54.4) 405 210 (51.9) 121 76 (62.8)†

Fasting lipids (mmol/l) AM%� AM%� AM%�Triglycerides 2.4 � 0.1 �3.3 � 1.3* 2.4 � 0.1 �2.8 � 1.4* 2.1 � 0.1 �6.3 � 2.9*Total cholesterol 5.3 � 0.0 �2.2 � 0.4* 5.3 � 0.0 �2.0 � 0.5* 5.3 � 0.1 �3.0 � 1.0*LDL 3.1 � 0.0 �0.7 � 0.8 3.1 � 0.0 �0.3 � 0.8 3.1 � 0.1 �2.5 � 1.7HDL 1.2 � 0.0 4.5 � 0.6* 1.1 � 0.0 4.5 � 0.6* 1.3 � 0.0 4.8 � 1.3*Non-HDL 4.1 � 0.0 �3.3 � 0.6* 4.2 � 0.0 �3.0 � 0.6* 4.0 � 0.1 �4.9 � 1.3*VLDL 0.95 � 0.01 �3.4 � 1.1* 1.0 � 0.0 �3.0 � 1.3* 0.9 � 0.0 �5.3 � 2.4*

Blood pressure (mmHg) mean � mean � mean �Diastolic 81.3 � 0.3 �1.4 � 0.2* 81.5 � 0.2 �1.3 � 0.2* 80.1 � 0.5 �2.0 � 0.5*Systolic 132.1 � 0.3 �2.2 � 0.3* 130.8 � 0.4 �2.2 � 0.4* 138.5 � 0.8 �2.2 � 1.0*

Subgroup analyses BL (n) AM� BL (n) AM� BL (n) AM�A1C (%)

BL A1C �8.0% 7.6 (533) �0.6 � 0.0* 7.6 (410) �0.6 � 0.1* 7.6 (123) �0.7 � 0.1*BL A1C �8.0% 9.2 (936) �1.3 � 0.1* 9.2 (821) �1.2 � 0.1* 9.0 (115) �1.4 � 0.1*BL A1C �9.0% 8.1 (995) �0.8 � 0.0* 8.1 (806) �0.7 � 0.0* 7.9 (189) �0.9 � 0.1*BL A1C �9.0% 9.9 (474) �1.6 � 0.1* 9.9 (425) �1.6 � 0.1* 9.7 (49) �1.7 � 0.2*BL BMI �30 kg/m2 8.7 (613) �1.2 � 0.1* 8.8 (487) �1.2 � 0.1* 8.4 (126) �1.3 � 0.1*BL BMI �30 kg/m2 8.6 (855) �0.9 � 0.1* 8.7 (743) �0.9 � 0.1* 8.2 (112) �1.0 � 0.1*BL BMI �35 kg/m2 8.7 (1,034) �1.1 � 0.0* 8.8 (838) �1.1 � 0.1* 8.3 (196) �1.2 � 0.1*BL BMI �35 kg/m2 8.6 (434) �0.9 � 0.1* 8.6 (392) �0.9 � 0.1* 8.2 (42) �0.9 � 0.2*

Body weight (kg)BL BMI �30 kg/m2 75.6 (613) �0.0 � 0.1 75.8 (487) 0.1 � 0.1 74.5 (126) �0.5 � 0.3†BL BMI �30 kg/m2 101.3 (855) �0.6 � 0.2* 102.6 (743) �0.5 � 0.2* 93.3 (112) �1.3 � 0.4*†BL BMI �35 kg/m2 82.1 (1,034) �0.2 � 0.1 82.8 (838) �0.1 � 0.1 79.4 (196) �0.8 � 0.2*†BL BMI �35 kg/m2 110.7 (434) �0.6 � 0.3* 111.7 (392) �0.6 � 0.3* 101.8 (42) �1.4 � 0.7*

Data are means � SE unless otherwise indicated. *P � 0.05 vs. baseline (within group); †P � 0.05 vs. younger subgroup. ‡Patients with both baseline A1C �7%and an end point value. BL, baseline.

Pratley and Associates


overall population (AM� �0.3 kg) or inthe younger subgroup (AM� �0.2 kg). Incontrast, in older patients, vildagliptinsignificantly decreased body weight(AM� �0.9 kg) from a baseline (83.4 kg)that was significantly lower than that inyounger patients (92.0 kg). In both theyounger and the older subgroups, weightloss was more substantial in the moreobese patients (Table 1).

In the overall efficacy population aswell as in both subgroups, vildagliptinproduced modest but statistically signifi-cant improvements in the fasting lipidprofile. Although there were no signifi-cant differences between the responsesobserved by age-groups, the most sub-stantial changes were observed in theelderly subgroup. Very modest reduc-tions in blood pressure were seen in theoverall population, and these did notdiffer between older and younger sub-groups (Table 1).

Subgroup analysesBoth baseline A1C and baseline BMI ap-peared to influence the magnitude of theresponses to vildagliptin, and the efficacyof vildagliptin was consistently of slightlygreater magnitude in elderly patientscompared with younger patients acrossall prespecified subgroups (Table 1). Al-though reductions in A1C were some-what larger in the leaner subgroups, theenhanced efficacy of vildagliptin in olderversus younger patients was not ex-plained by their lesser degree of obesity.Thus, when analyses of covariance wereperformed to adjust for baseline BMI, thesame differential effect remained for bothA1C (between-group difference in AM��0.13 � 0.09%, P � 0.140) and FPG(between-group difference in AM��0.4 � 0.2 mmol/l, P � 0.041).

Safety and tolerabilityFigure 1 depicts adverse event profiles inthe overall safety population (Fig. 1A) andin the younger (Fig. 1B) and older (Fig.1C) subgroups. Adverse events wereslightly more frequent in older (63.6%)than in younger (60.6%) patients receiv-ing vildagliptin, but a more substantialdifference was seen for the pooled activecomparator group (68.1% in the elderlysubgroup vs. 63.0% in the younger sub-group). Further, no excess of adverseevents in elderly versus younger patientswith mild renal impairment receivingvildagliptin (62.0 vs. 62.1%) and no ex-cess in older patients with mild renal im-pairment compared with older patients

with normal renal function (62.0 vs.64.3%) were noted, whereas the adverseevent rate in renally impaired patients re-ceiving an active comparator was higherfor both older (74.6%) and younger(71.7%) patients. Adverse events sus-pected to be drug related were more com-mon in both older (18.1%) and younger(17.9%) patients receiving an active com-parator than in older (12.3%) or younger(9.2%) patients receiving vildagliptin. Se-rious adverse events (SAEs) were reportedby a somewhat higher percentage of older(6.4%) than of younger (3.1%) vildaglip-tin-treated patients or of older patients re-

ceiving an active comparator (2.6%); thisrepresented a total of 24 patients withSAEs, distributed across 13 “primary sys-tem organ classes” (Medical Dictionaryfor Regulatory Affairs categories), with nocluster of events within any specific pre-ferred term. None of the SAEs in vilda-gliptin-treated elderly patients wassuspected to be drug related. A possibledrug-related SAE was reported by one pa-tient receiving vildagliptin in the youngersubgroup and by one elderly patient re-ceiving an active comparator. Discontinu-ations due to an adverse event wereslightly more frequent in older (3.7%)

Figure 1— Adverse events (AE) in patients receiving vildagliptin monotherapy, patients receivingmonotherapy with any active comparator, and patients receiving placebo in the overall safetypopulation (A), the subgroup of patients aged �65 years (B), and patients aged �65 years (C).D/C, discontinued; RI, renal impairment.



than in younger (2.6%) vildagliptin-treated patients but were more frequent inboth older (6.0%) and younger (5.3%)patients receiving an active comparator.

A summary of the most commonly re-ported specific adverse events occurringin elderly patients and the incidence ofthose specific adverse events in the overallsafety population and in the younger sub-group is provided in Table A3 of the on-line appendix. The frequency of anyspecific adverse event in vildagliptin-treated elderly patients was similar to thatin younger patients receiving vildagliptin.In elderly patients receiving vildagliptin,the frequencies of upper respiratory tractinfection (6.4%), dizziness (5.3%), andsinusitis (2.4%) were somewhat higherthan in the pooled active comparators(3.4, 2.6, and 1.7%, respectively), where-as the frequencies of diarrhea (11.2%),nausea (6.0%), peripheral edema (6.0%),and nasopharyngitis (7.8%) were higherin elderly patients receiving an activecomparator than in elderly patients re-ceiving vildagliptin (7.0, 2.9, 1.9, and1.9%, respectively).

Confirmed hypoglycemia was rare,reported by 9 of 2,264 patients (0.4%)receiving vildagliptin monotherapy, ofwhich 3 were �65 years of age (0.8% ofthe elderly subgroup). All hypoglycemicevents in elderly patients were mild in se-verity; none of the hypoglycemic eventsled to discontinuation of therapy, andnone occurred at night. No severe hypo-glycemia occurred in any treatmentgroup. Two of 735 patients (0.3%) receiv-ing an active comparator reported con-firmed hypoglycemia, and no patientreceiving placebo had a hypoglycemicevent.

Four deaths occurred during treat-ment with vildagliptin (0.2%); two werein the elderly subgroup. In elderly pa-tients receiving vildagliptin, one deathwas due to ischemic stroke and the otherto postoperative bleeding and septicshock after surgery for a small bowel ob-struction. Two patients receiving an ac-tive comparator died, both of whom werein the younger subgroup; no deaths oc-curred with placebo.

CONCLUSIONS — The main find-ings of the present pooled analyses of theefficacy and safety of vildagliptin are thatthis DPP-4 inhibitor is both effective andwell tolerated in elderly patients with type2 diabetes. Although the elderly popula-tion was on average less obese than theyounger subgroup, comorbid conditions

were much more common; in particular,the older subgroup had a poorer cardio-vascular risk profile and higher preva-lence of coronary artery disease, as well asa high prevalence of undiagnosed mildrenal impairment. These factors and theuse of multiple comedications make themanagement of type 2 diabetes consider-ably more difficult in elderly individuals.Despite these potential problems, theoverall adverse event profile was similarin older and younger patients receivingvildagliptin. It is noteworthy that in pa-tients with mild renal impairment, therewas no increase in the incidence of ad-verse events in older compared withyounger patients receiving vildagliptin.Additionally, in older patients, the inci-dence of adverse events in patients withmild renal impairment was similar to thatin patients with normal renal functionwith vildagliptin treatment. In contrast,the adverse event rate in younger andolder patients with mild renal impairmentreceiving an active comparator was higherthan that in patients with normal renalfunction. Because mild renal impairmentis common in elderly patients with type 2diabetes, although frequently undiag-nosed, its impact on the tolerability of anyOAD is important to assess and to takeinto consideration in the choice and in-tensity of treatment.

In view of the greater propensity forhypoglycemia (and severe hypoglycemia)in elderly patients (2), another importantfinding is the fact that the incidence ofhypoglycemia was very low (0.8%) in el-derly patients receiving vildagliptin; nosevere hypoglycemia occurred. Althoughhypoglycemia was even less frequent inpatients receiving an active comparator(two patients, 0.3%), it is important tonote that the pooled dataset did not in-clude studies with a sulfonylurea or insu-lin as an active comparator. With regardto hypoglycemia, a recent study of vilda-gliptin added to insulin therapy is rele-vant. During 24 weeks of treatment withvildagliptin (100 mg daily) versus pla-cebo added to a stable insulin treatmentregimen, it was found that hypoglycemiawas significantly less frequent and less se-vere with vildagliptin than with placebo,and the same trend held in the subgroupof patients aged �65 years (15).

Overall, the present safety analysisshowed that in elderly patients receivingvildagliptin, there was a slightly lower in-cidence of any adverse event, drug-related adverse events, and adverse eventsin those with mild renal impairment than

in elderly patients receiving an activecomparator. Although there was a slightlyhigher incidence of SAEs in elderly pa-tients receiving vildagliptin than in thosereceiving an active comparator, none wassuspected to be drug related. Some spe-cific adverse events, such as peripheraledema, nausea, or diarrhea, were less fre-quently reported with vildagliptin thanwith the active comparators (metforminand thiazolidinediones).

A relatively benign adverse event pro-file is an important consideration fortreatment of type 2 diabetes in older pa-tients in whom metformin should be usedwith caution in case altered renal functionis present, sulfonylureas present a well-documented risk of hypoglycemia, andthiazolidinediones raise concerns aboutcongestive heart failure. With a new classof OAD, however, particularly one thatacts by inhibiting a ubiquitous enzymesuch as DPP-4, long-term monitoringwith much broader patient exposure willbe crucial to further ascertain its safety inelderly patients.

The influence of vildagliptin mono-therapy on all efficacy parameters in drug-naive elderly patients with type 2 diabeteswas consistently as robust, if not more so,than that in younger patients. Despitelower baseline levels of A1C, FPG, andbody weight, in patients aged �65 years,the decrease in A1C tended to be greater(� �1.2%) than that in patients �65years of age (� �1.0%); the decrease inFPG was significantly greater in the older(� �1.5 mmol/l) than in the younger (��1.1 mmol/l) subgroup, and bodyweight decreased significantly from base-line only in the older subgroup (� � 0.9kg). Further, relative to the younger sub-group, a significantly higher percentage ofelderly patients achieved the AmericanDiabetes Association recommended tar-get A1C (�7.0%), both in the whole el-derly subgroup (which began with asomewhat lower mean baseline value)and in the population of patients withbaseline A1C within 1% of target (inwhich the elderly and younger subgroupshad the same mean baseline A1C of7.6%).

In view of a report that DPP-4 activityis reduced in elderly subjects (both non-diabetic and those with type 2 diabetes)and the prediction arising from this find-ing that DPP-4 inhibitors would be lesseffective in elderly than in younger pa-tients (16), the present efficacy resultsmay be particularly noteworthy andclearly refute that hypothesis. There are at

Pratley and Associates


least two possible explanations for thetrend toward enhanced efficacy of vilda-gliptin in older patients. It may be that themechanisms underlying development oftype 2 diabetes in older patients are moreamenable to treatment with a DPP-4 in-hibitor. Thus, islet dysfunction, includinghyperglucagonemia (17) and postpran-dial hyperglycemia (18), may play a moresignificant role in elderly patients withtype 2 diabetes, especially when insulinsecretion is considered in the context ofthe prevailing degree of insulin resistance(17). Because vildagliptin acts via GLP-1–mediated improvements in both �- and�-cell function (6) and nutrient intake isthe primary stimulus for GLP-1 release,vildagliptin has a pronounced effect to re-duce postprandial hyperglycemia (13).This unique mechanism of action couldunderlie the maintenance of robust effi-cacy of vildagliptin in elderly patientswith type 2 diabetes. Further, the glucose-dependent insulinotropic polypeptide re-sponse to nutrient intake is exaggerated inelderly patients with type 2 diabetes (16),which may compensate for the impaired�-cell responsiveness to GIP seen in el-derly individuals (19) and in patients withtype 2 diabetes (20).

The mechanism by which vildagliptintreatment leads to modest weight loss inelderly individuals is unclear but is notattributable to gastrointestinal upset be-cause gastrointestinal adverse events werereported by few patients and somewhatless frequently in the elderly than in theyounger subgroup (e.g., nausea incidenceof 1.9 vs. 2.7%, respectively). Moreover,subgroup analyses established that moreweight loss was generally seen in moreobese subjects, whereas the elderly sub-jects were on average less obese than theyounger subjects. Although vildagliptintreatment does not seem to influence therate of gastric emptying (21) or satiety inthe general population, selective effects ofvildagliptin in the elderly on these poten-tial mechanisms or on DPP-4 substratesother than GLP-1 cannot be ruled out.

In summary, although much remainsto be understood about the mechanismsunderlying some unique aspects of DPP-4inhibitors in the elderly, vildagliptinmonotherapy is effective and appears tobe well tolerated in OAD-naive patientsaged �65 years. Accordingly, the presentfindings strongly support the continuedassessment of vildagliptin to more fullyascertain its safety and efficacy in elderlypatients with type 2 diabetes.

Acknowledgments— All of the phase IIIstudies pooled in the manuscript were fundedby Novartis Pharmaceuticals Corporation.

The authors gratefully acknowledge the in-vestigators and staff at participating sites for allthe studies, as well as Beth Dunning Lower,PhD, for helpful discussion and editorial assis-tance.

References1. Halter JB: Diabetes mellitus in older

adults: underdiagnosis and undertreat-ment. J Am Geriatr Soc 48:340–341, 2000

2. Rosenstock J: Management of type 2 di-abetes mellitus in the elderly: specialconsiderations. Drugs Aging 18:31– 44,2001

3. Matyka K, Evans M, Lomas J, Cranston I,Macdonald I, Amiel SA: Altered hierarchyof protective responses against severe hy-poglycemia in normal aging in healthymen. Diabetes Care 20:135–141, 1997

4. Del Prato S, Heine RJ, Keilson L, GuitardC, Shen SG, Emmons RP: Treatment ofpatients over 64 years of age with type 2diabetes: experience from nateglinidepooled database retrospective analysis.Diabetes Care 26:2075–2080, 2003

5. Selvin E, Coresh J, Brancati FL: The bur-den and treatment of diabetes in elderlyindividuals in the U.S. Diabetes Care 29:2415–2419, 2006

6. Mari A, Sallas WM, He YL, Watson C,Ligueros-Saylan M, Dunning BE, DeaconCF, Holst JJ, Foley JE: Vildagliptin, adipeptidyl peptidase-IV inhibitor, im-proves model-assessed �-cell function inpatients with type 2 diabetes. J Clin Endo-crinol Metab 90:4888–4894, 2005

7. Pi-Sunyer FX, Schweizer A, Mills D, De-jager S: Efficacy and tolerability of vil-dagliptin monotherapy in drug-naivepatients with type 2 diabetes. Diabetes ResClin Pract 76:132–138, 2007

8. Rosenstock J, Baron MA, Dejager S, MillsD, Schweizer A: Comparison of vildaglip-tin and rosiglitazone monotherapy in pa-tients with type 2 diabetes: a 24-week,double-blind, randomized trial. DiabetesCare 30:217–223, 2007

9. Kreymann B, Williams G, Ghatei MA,Bloom SR: Glucagon-like peptide-1 7–36:a physiological incretin in man. Lancet2:1300–1304, 1987

10. Nauck MA, Heimesaat MM, Behle K,Holst JJ, Nauck MS, Ritzel R, Hufner M,Schmiegel WH: Effects of glucagon-likepeptide 1 on counterregulatory hormoneresponses, cognitive functions, and insu-lin secretion during hyperinsulinemic,stepped hypoglycemic clamp experi-ments in healthy volunteers. J Clin Endo-crinol Metab 87:1239–1246, 2002

11. Dejager S, Razac S, Foley JE, Schweizer A:Vildagliptin in drug-naive patients withtype 2 diabetes: a 24-week, double-blind,

randomized, placebo-controlled, multi-ple-dose study. Horm Metab Res 39:218–223, 2007

12. Schweizer A, Couturier A, Foley JE, De-jager S: Comparison between vildagliptinand metformin to sustain reductions inHbA1c over one year in drug-naıve pa-tients with type 2 diabetes. Diabet Med24:955–961, 2007

13. Rosenstock J, Baron MA, Camisasca RP,Cressier F, Couturier A, Dejager S: Effi-cacy and tolerability of initial combina-tion therapy with vildagliptin andpioglitazone compared to componentmonotherapy in patients with type 2 dia-betes. Diabetes Obes Metab 9:175–185,2007

14. Rigalleau V, Lasseur C, Perlemoine C,Barthe N, Raffaitin C, Liu C, Chauveau P,Baillet-Blanco L, Beauvieux MC, CombeC, Gin H: Estimation of glomerular filtra-tion rate in diabetic subjects: Cockcroftformula or Modification of Diet in RenalDisease study equation? Diabetes Care 28:838–843, 2005

15. Fonseca V, Schweizer A, Albrecht D,Baron MA, Chang I, Dejager S: Additionof vildagliptin to insulin improves glycae-mic control in type 2 diabetes. Diabetolo-gia 50:1148–1155, 2007

16. Korosi J, McIntosh CH, Pederson RA, De-muth HU, Habener JF, Gingerich R, EganJM, Elahi D, Meneilly GS: Effect of agingand diabetes on the enteroinsular axis. JGerontol A Biol Sci Med Sci 56:M575–M579, 2001

17. Basu R, Breda E, Oberg AL, Powell CC,Dalla MC, Basu A, Vittone JL, Klee GG,Arora P, Jensen MD, Toffolo G, Cobelli C,Rizza RA: Mechanisms of the age-associ-ated deterioration in glucose tolerance:contribution of alterations in insulin se-cretion, action, and clearance. Diabetes52:1738–1748, 2003

18. Chang AM, Halter JB: Aging and insulinsecretion. Am J Physiol 284:E7–E12, 2003

19. Meneilly GS, Ryan AS, Minaker KL, ElahiD: The effect of age and glycemic level onthe response of the �-cell to glucose-de-pendent insulinotropic polypeptide andperipheral tissue sensitivity to endog-enously released insulin. J Clin EndocrinolMetab 83:2925–2932, 1998

20. Elahi D, McAloon-Dyke M, FukagawaNK, Meneilly GS, Sclater AL, MinakerKL, Habener JF, Andersen DK: The in-sulinotropic actions of glucose-depen-dent insulinotropic polypeptide (GIP)and glucagon-like peptide-1 (7–37) innormal and diabetic subjects. Regul Pept51:63–74, 1994

21. Vella A, Bock G, Giesler PD, Burton DB,Serra DB, Ligueros SM, Dunning BE,Foley JE, Rizza RA, Camilleri M: Effects ofdipeptidyl peptidase 4 inhibition on gas-trointestinal function, meal appearance,and glucose metabolism in type 2 diabe-tes. Diabetes 56:1475–1480, 2007



Continuous Home Monitoring of GlucoseImproved glycemic control with real-life use of continuous glucose sensorsin adult subjects with type 1 diabetes

SATISH K. GARG, MD1,2,3

WILLIAM C. KELLY, BS1

MARY K. VOELMLE, MS, FNP, CDE1,3

PETER J. RITCHIE, BA1

PETER A. GOTTLIEB, MD1,2,3

KIM K. MCFANN, PHD1,4

SAMUEL L. ELLIS, PHARMD, CDE1

Improving metabolic control reducesmicro- and macrovascular complica-tions of diabetes. However, intensive

insulin therapy increases severe hypogly-cemia more than threefold (1–3). Contin-uous glucose monitoring (CGM) is beingintroduced into routine clinical care de-spite a lack of reimbursement. Registra-tion studies for the Food and DrugAdministration (FDA) documented thatsubjects using real-time CGM improveglucose excursions, reduce variability, de-crease time spent in hypoglycemia andhyperglycemia, and improve A1C values(4–9). Despite these reports, there aredata unsupportive of new technologiessuch as CGM (10) or personal digital as-sistants (11) for reducing hypoglycemia.This study evaluates glucose control andits relationship with glucose target rangeswith continuous home monitoring of glu-cose (CHMG).

RESEARCH DESIGN ANDMETHODS — Inclusion criteria lim-ited analysis to subjects with A1C valuesand downloaded CHMG data at baselineand 3 months, as well as software todownload receivers (not available for thefirst 9 months). Patients who were preg-

nant or planning a pregnancy wereexcluded.

A total of 24 subjects on CHMG wereincluded in this analysis. All patients inthis study used the DexCom STS sensor(DexCoM, San Diego, CA). Subjects werecomputer matched for baseline A1C (�0.3%), sex, age, and duration of diabetesexcept for one subject in the CHMGgroup, who had diabetes for 57 years.Baseline demographics were similar be-tween groups (Table 1). This protocol wasinstitutional review board approved.

Subjects initiating CHMG attended asession on glucose trends, features of theCHMG receiver, and proper insertiontechniques conducted by certified diabe-tes educators. All subjects were instructednot to change treatment based on theirfirst week of CHMG use.

All subjects had baseline and 12-week A1C measurements (DCA 2000;Bayer, Tarrytown, NY). The CHMG datawere downloaded prospectively at base-line and at 6 (� 2) and 12 (� 2) weeks,except for one subject who did not have6-week data. Subjects wore sensors asthey felt necessary. Subjects were taughtto override the receiver every 3 days anduse the same sensor for an additional 3

days. All subjects in the comparisongroup received similar diabetes care. No6-week data or fingerstick SMBG mea-surements were available for the compar-ison group.

CHMG data were analyzed for within(WTRs) (60–150 mg/dl), above (ATRs)(�150 mg/dl), and below (BTRs) (�60mg/dl) target ranges of blood glucose. TheBTR of �60 mg/dl was used as a result ofclinical observations that subjects usingCHMG are more likely to treat glucosevalues of 60 mg/dl as opposed to 70 mg/dl, which was used for BTR in our previ-ous self-monitoring of blood glucose(SMBG) publication (12). The ATR read-ings were further analyzed for 151–240and �240 mg/dl. The percentages ofreadings within each target range werecompared among baseline and 6- and 12-week data. The number of subjects reach-ing target A1C values was also analyzed.No subject had severe hypoglycemianeeding glucagon or emergency roomvisits.

Statistical analysisAnalyses of A1C change from baselineand time within glycemic ranges wereperformed using SAS software (version9.1; SAS, Cary, NC). Two-tailed testswere used unless otherwise stated. Base-line characteristics were compared usingindependent-samples t tests. Fisher’s ex-act tests were performed on the number ofsubjects reaching target A1C values atbaseline. Logistic regression, with base-line A1C target as a covariate, was used toexamine whether the experimental groupwas more likely than the comparisongroup to reach A1C targets by 3 months.Mixed-model repeated-measures analysiswas used to evaluate the change over timein A1C, insulin dose, and the number ofpatients WTR, BTR, and ATR of bloodglucose within the CHMG group.

RESULTS — Mean � SD sensor useper subject was 17.6 � 8.4 days permonth. Subjects extended the use (de-spite 3-day approval and now FDA ap-proval for 7 days) of sensors to 6.8 � 1.6days.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Barbara Davis Center for Childhood Diabetes, University of Colorado Health Sciences Center,Aurora, Colorado; the 2Department of Internal Medicine, University of Colorado at Denver, Aurora, Colo-rado; the 3Department of Pediatrics, University of Colorado at Denver, Aurora, Colorado; and the 4Depart-ment of Preventive Medicine and Biometrics, University of Colorado at Denver, Aurora, Colorado.

Address correspondence and reprint requests to Satish K. Garg, MD, Barbara Davis Center for ChildhoodDiabetes, University of Colorado at Denver, 1775 North Ursula St., Aurora, CO 80045. E-mail:[email protected].

Received for publication 24 July 2007 and accepted in revised form 31 August 2007.Published ahead of print at http://care.diabetesjournals.org on 11 September 2007. DOI: 10.2337/dc07-

1436.Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/

dc07-1436.Abbreviations: ATR, above target range; BTR, below target range; CGM, continuous glucose monitoring;

CHMG, continuous home monitoring of glucose; WTR, within target range.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion



C l i n i c a l C a r e / E d u c a t i o n / N u t r i t i o n / P s y c h o s o c i a l R e s e a r c hB R I E F R E P O R T


Changes in A1CA1C values at baseline were 7.43 � 1.0and 7.39 � 1.0% for the CHMG and com-parison groups, respectively (P � 0.896)(Table 1). There was a significant decreasein A1C in the CHMG group (0.4 � 0.5%;P � 0.047, mixed repeated-measuresanalysis) at 12 weeks, with a nonsignifi-cant increase in A1C (0.3 � 1.1%; P �0.0710) in the comparison group. Also, at12 weeks there was a difference in A1Cvalues between groups (P � 0.0385) de-spite the fact that there was no change ininsulin dose. The number of subjectsachieving A1C values �7.5% was higherin the CHMG group at 12 weeks (OR7.229; P � 0.0234) (13).

Glucose target rangesSubjects using CHMG increased WTRglucose readings by 6.5 � 15.0% (P �0.0353) and reduced mean ATR glucosereadings by 5.6 � 16.7% (P � 0.0355) at12 weeks compared with baseline. ATRglucose values also showed a significantreduction in readings �240 mg/dl by6.4 � 14.0% (P � 0.0351) at 12 weeks.Results were similar for subjects usingmultiple daily injections or insulinpumps. Pie charts for glucose ranges areavailable in an online appendix at http://dx.doi.org/10.2337/dc07-1436.

CONCLUSIONS — This study dem-onstrates that use of real-time CHMG isassociated with improved metabolic con-trol over 12 weeks in adults with type 1diabetes, as previously documented(8,14–19). This study supports previousfindings carrying over to real-life use ofCHMG in subjects with reasonable glu-cose control (A1C �7.4%). The modestimprovement in A1C of 0.4% could bedue to the subject population, the short-term nature of the study, and near-targetbaseline A1C values of 7.43%.

Improvements in metabolic controlwith CHMG were not associated with in-creased hypoglycemia, supporting earlierfindings (8,14–19). The mean increase inWTR glucose readings of 6.5% and de-crease in ATR glucose readings of 5.6% at3 months corresponded with a 0.4% de-cline in A1C, which is lower than was ex-pected based on our previous self-monitoring of blood glucose data (12).This could be due to lower A1C values atbaseline.

Limitations of this study include asmall sample size, shorter follow-up, andlack of a randomized control group. How-ever, the data show that CHMG use re-sults in a small A1C reduction withoutincreasing hypoglycemia, most likely dueto behavioral changes.

We conclude that use of CHMG canfurther improve glucose control in sub-jects with relatively well-controlled type 1diabetes, with no increase in hypoglyce-mia. Prospective randomized clinical tri-als using CHMG with a large sample sizeneed to be conducted.

Acknowledgments— This study was spon-sored in part by grant 08 FLA 00250 from theState of Colorado Public Health and Environ-ment; grant P30 DK575616 from the DiabetesEndocrine Research Center, National Insti-tutes of Health (NIH); grant M01 RR00069from the General Clinical Research CentersProgram, NIH; and grants R01 HL61753, RO1HL079611, and RO1 DK32493 from the Chil-dren’s Diabetes Foundation (Denver, CO).

References1. The effect of intensive treatment of diabe-

tes on the development and progressionof long-term complications in insulin-de-pendent diabetes mellitus: The DiabetesControl and Complications Trial Re-search Group. N Engl J Med 329:977–986,1993

2. UK Prospective Diabetes Study (UKPDS)Group: Intensive blood-glucose controlwith sulphonylureas or insulin comparedwith conventional treatment and risk ofcomplications in patients with type 2diabetes (UKPDS 33). Lancet 352:837–

Table 1—Demographics and results

Baseline 3 months

CHMG Comparison P CHMG Comparison P

n 24 23 NSAge (years) 45.8 � 13.2 44.3 � 13.4 0.703Duration (years) 27.2 � 16.6 24.0 � 15.8 0.513Sex (male/female) 11/13 10/13 0.871BMI (kg/m2) 26.1 � 4.1 26.9 � 4.8 0.565Treatment

MDI 18 16 0.677CSII 6 7 0.677

A1C (%) 7.43 � 1.0* 7.39 � 1.0 0.896 7.06 � 0.8* 7.73 � 1.4 0.039Target A1C

�7.5% 14 of 24 13 of 23 0.900 20 of 24 12 of 23 0.023�7.0% 7 of 24 6 of 23 0.814 12 of 24 6 of 23 0.211�6.5% 4 of 24 4 of 23 1.000 4 of 24 3 of 23 1.000

Insulin dose 51.9 � 31.4† 45.7 � 28.6† 0.413 50.1 � 31.4† 49.0 � 33.4† 0.310Glucose target ranges (%)

WTR‡ 42.6 � 19.5 NA NA 49.1 � 16.7 NA 0.0353§ATR 53.2 � 20.4 NA NA 47.6 � 17.0 NA 0.0355§BTR 4.2 � 3.5 NA NA 3.4 � 6.7 NA 0.638§

Data are means � SD or n unless otherwise indicated. *There was a significant decrease (P � 0.047) in A1C in the CHMG group from baseline to 3 months. †Therewas no significant change in total insulin dose in the CHMG or comparison group from baseline to 3 months. ‡WTR glycemia was defined as 60–150 mg/dl (3.3–8.3mmol/l). §These P values represent differences in glucose target ranges from baseline to 3 months in the CHMG group using mixed-model repeated-measuresanalysis. CSII, continuous subcutaneous insulin infusion; MDI, multiple daily injection; NA, not applicable; NS, not significant.

Real-life use of CHMG


853, 19983. Writing Team for the Diabetes Control

and Complications Trial/Epidemiology ofDiabetes Interventions and Complica-tions Research Group: Sustained effect ofintensive treatment of type 1 diabetesmellitus on development and progressionof diabetic nephropathy: the Epidemiol-ogy of Diabetes Interventions and Com-plications (EDIC)study. JAMA290:2159–2167, 2003

4. Garg SK, Schwartz S, Edelman SV: Im-proved glucose excursions using an im-plantable real-time continuous glucosesensor in adults with type 1 diabetes. Di-abetes Care 27:734–738, 2004

5. Garg S, Jovanovic L: Relationship of fast-ing and hourly blood glucose levels toHbA1c values: safety, accuracy, and im-provements in glucose profiles obtainedusing a 7-day continuous glucose sensor.Diabetes Care 29:2644–2649, 2006

6. Diess D, Bolinder J, Riveline JP, BattelinoT, Bosi E, Tubiana-Rufi N, Kerr D, PhillipM: Improved glycemic control in poorlycontrolled patients with type 1 diabetesusing real-time continuous glucose mon-itoring. Diabetes Care 29:2730–2732,2006

7. Bailey T, Zisser H, Garg S: Reduction inhemoglobin A1c with real-time continu-ous glucose monitoring: results from a12-week observational study. DiabetesTechnol Ther 8:203–210, 2007

8. Chase HP, Kim LM, Owen SL, MacKenzieTA, Klingensmith GJ, Murtfeldt R, GargSK: Continuous subcutaneous glucosemonitoring in children with type 1 diabe-tes. Pediatrics 107:222–226, 2001

9. Kaufman FR, Austin J, Neinstein A, Jeng L,Halvorson M, Devoe DJ, PitukcheewanontP: Nocturnal hypoglycemia detected withthe continuous glucose monitoring systemin pediatric patients with type 1 diabetes.J Pediatr 141:625–630, 2002

10. Hirsch I, Bode B, Abelseth J, Fischer J,Kaufman F, Mastrototaro J, Wolpert H,Buckingham B: Sensor augmented pumptherapy: results of the first treat-to-targetstudy (Abstract). Diabetes 56 (Supp1. 1):A24, 2007

11. Ellis S, Beatson C, Gottlieb P, Gutin R,Bookout T, Figal C, Snyder B, Garg S: Im-proved glycemic control in intensivelytreated subjects with type 1 diabetes usingAccu-Chek® Advisor insulin guidancesoftware (Abstract). Diabetes 56 (Suppl.1):A8, 2007

12. Brewer KW, Chase HP, Owen S, Garg SK:Slicing the pie. Correlating HbA1c valueswith average blood glucose values in a piechart form. Diabetes Care 21:209–212,1998

13. American Diabetes Association: Stan-dards of medical care in diabetes–2006.Diabetes Care 29 (Suppl. 1):S4–S42, 2006

14. Bode BW, Gross TM, Thornton KR, Mas-trototaro JJ: Continuous glucose monitor-

ing used to adjust diabetes therapyimproves glycosylated hemoglobin: a pi-lot study. Diabetes Res Clin Pract 46:183–190, 1999

15. Chase HP, Roberts MD, Wightman C,Klingensmith G, Garg SK, Van Wyhe M,Desai S, Harper W, Lopatin M, BartkowiakM, Tamada J, Eastman RC: Use of the Glu-coWatch Biographer in children with type 1diabetes. Pediatrics 111:790–794, 2003

16. Ludvigsson J, Hanas R: Continuous sub-cutaneous glucose monitoring improvedmetabolic control in pediatric patientswith type 1 diabetes: a controlled cross-over study. Pediatrics 111:933–938, 2003

17. Schaepelynck-Belicar P, Vague P, Simo-nin G, Lassmann-Vague V: Improvedmetabolic control in diabetic adolescentsusing the continuous glucose monitoringsystem (CHMGS). Diabet Metab 29:608–612, 2003

18. Schiaffini R, Ciampalini P, Fierabracci A,Spera S, Borrelli P, Bottazzo GF, Crino A:The continuous glucose monitoring sys-tem (CHMGS) in type 1 diabetic childrenis the way to reduce hypoglycemic risk.Diabetes Metab Res Rev 18:324–329, 2002

19. Tanenberg R, Bode B, Lane W, Levetan C,Mestman J, Harmel AP, Tobian J, Gross T,Mastrototaro J: Use of the continuous glu-cose monitoring system to guide therapyin patients with insulin-treated diabetes: arandomized controlled trial. Mayo ClinProc 79:1521–1526, 2004

Garg and Associates


Targeting Glucose in Acute MyocardialInfarctionHas glucose, insulin, and potassium infusion missed the target?

AJAY CHAUDHURI, MD1

MICHAEL MILLER, PHD2

RICHARD NESTO, MD3

NOAH ROSENBERG, MD2

PARESH DANDONA, MD, PHD1

P roinflammatory mechanisms maycontribute to hyperglycemia-associated adverse outcomes in

acute myocardial infarction (AMI) (1–3).Insulin exerts anti-inflammatory effects inST elevation myocardial infarction andcoronary artery bypass graft patients (4–6). Inflammation plays an important rolein the pathogenesis of atherosclerosis andthrombosis (7,8). Thus, insulin infusionin AMI should be beneficial. However,glucose, insulin, and potassium (GIK) in-fusion was neutral in its benefit in theClinical Trial of Metabolic Modulation inAcute Myocardial Infarction TreatmentEvaluation–Estudios Cardiologicos Latino-america (CREATE-ECLA) study (9).

The GIK regimen (fixed combinationof 1 l of 25% dextrose with 50 units reg-ular insulin and 80 meq/l potassium)used in CREATE-ECLA is known to lowerserum free fatty acid concentrations (10)and to reduce mortality by 18% in AMI, aspreviously reviewed (11). However, theinfusion of 30 g/h glucose without titra-tion of glucose or insulin in CREATE-ECLA led to a significant increase in bloodglucose concentrations. Mehta et al. sug-gested that the adverse effect of GIK-induced hyperglycemia may haveneutralized any potential benefit of insu-lin in the GIK regimen. However, Mehta

et al. did not comment on the magnitudeof the potential contribution of hypergly-cemia to neutralize benefits of GIK. Sinceadmission hyperglycemia was predictiveof mortality in AMI in CREATE-ECLA, wehave estimated the potential effects ofGIK-induced hyperglycemia on mortalityin this study.

RESEARCH DESIGN ANDMETHODS — Based on the admissionblood glucose–related mortality rates incontrol subjects in CREATE-ECLA, wehave constructed a model: 30-day % mor-tality � 100% � [1 � c � exp(�d � BG)]predicting 30-day mortality as a functionof admission blood glucose, where exp �exponential function and BG � bloodglucose. The constants c and d were esti-mated using a regression involving 3points from this constructed curve. Thismodel was then applied to the blood glu-cose at admission and at 6 and 24 h forboth the GIK and control groups, assum-ing that the relationship between glyce-mia and mortality is maintained even afteradmission in AMI. This assumption isprobably valid because for every 0.6-mmol/l reduction in glucose postadmis-sion, there is an 8% reduction in mortalityin patients with AMI (12) and because

glucose levels after admission predictmortality in AMI (13,14).

The projected mortality at 0, 6, and24 h (Table 1) yielded trapezoids of whichthe areas were calculated. After dividingby 24 h, the following weighted averageformula for mortality was obtained: %mortality (average) � (0.125) � % mor-tality at 0 h � (0.5) � % mortality at 6 h� (0.375) � % mortality at 24 h.

RESULTS — In our model, the esti-mated 30-day mortality for control sub-jects based on blood glucose achievedduring 24 h is 9.9%, which is similar tothe observed mortality of 9.7% for thecontrol subjects in CREATE-ECLA. How-ever, the estimated mortality rate for theGIK group on the basis of the GIK-induced hyperglycemia during 24 h was12.2%, which was 2.2% higher than theobserved mortality of 10% for the GIKgroup.

CONCLUSIONS — We suggest thatthe insulin in the GIK infusion used inCREATE-ECLA might have neutralizedthe 2.2% (12.2% [expected] � 10% [ob-served]) increase in mortality that shouldhave been observed in the GIK group be-cause of the effect of hyperglycemia in-duced by this infusion. Thus, i fhyperglycemia was not induced by theGIK infusion used in CREATE-ECLA, theadministration of insulin in this trialcould have resulted in a 2.2% absoluteand a 22% relative reduction in mortalityin the GIK group. Indeed, CREATE-ECLAinvestigators have recently reported anexcess mortality and congestive cardiacfailure in the GIK group in the first 3 days,when GIK-induced hyperglycemia waspresent and probably had its maximal ef-fect. In contrast, there was a reduction inmortality and congestive cardiac failurebetween 4 and 30 days (15), when glu-cose levels had probably approximatedthe levels in the control group. In a caninemodel of AMI, low-dose insulin alone re-duced the infarct size, while glucose andpotassium (16) caused hyperglycemiaand increased infarct size. These find-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Endocrinology, State University at New York at Buffalo and Kaleida Health, Buffalo,New York; 2Sanofi-Aventis Pharmaceuticals, Bridgewater, New Jersey; and the 3Department of Cardiovas-cular Medicine, Lahey Clinic and Harvard Medical School, Boston, Massachusetts.

Address correspondence and reprint requests to Paresh Dandona, MD, PHD, Diabetes-EndocrinologyCenter of WNY, 3 Gates Cir., Buffalo, NY 14209. E-mail: [email protected].

Received for publication 28 May 2007 and accepted in revised form 4 September 2007.Published ahead of print at http://care.diabetesjournals.org on 11 September 2007. DOI: 10.2337/dc07-

1220.P.D. is currently affiliated with the Diabetes-Endocrinology Center of Western New York, Buffalo, New

York.Abbreviations: AMI, acute myocardial infarction; CREATE-ECLA, Clinical Trial of Metabolic Modulation

in Acute Myocardial Infarction Treatment Evaluation–Estudios Cardiologicos Latinoamerica; GIK, glucose,insulin, and potassium.






ings and the proinflammatory, pro-thrombotic effects of hyperglycemiamay explain how GIK-induced hyper-glycemia can neutralize the potentialbenefits of insulin (2).

A limitation of our study is that ouranalysis is based on the published data inCREATE-ECLA. Because of the absence ofdetailed data and the derivational natureof our methodology, which arrives at12.2% expected mortality, we are not in aposition to provide a P value or SEs. How-ever, based on the level of the expectedprecision in a large study like CREATE-ECLA, a potential absolute reduction inmortality of 2.2% or relative reduction of22% would be statistically significant andwell outside the confidence bounds (95%CI 9.4–10.6) of the observed mortality of10% for the GIK. Although our model ac-curately predicted the death rate for thecontrol subjects, it is possible that wecould have overestimated the death ratefor the GIK patients. We have also specu-lated that the adverse effects of the reac-tive hyperglycemia observed followingAMI are equivalent to the iatrogenic hy-perglycemia induced by the GIK infusion.Iatrogenic hyperglycemia is known to in-duce proinflammatory cytokines and en-dothelial dysfunction (17,18). Thesemechanisms may be responsible for theadverse cardiovascular outcomes associ-ated with hyperglycemia. The importantpoint for discussion is not whether themodel is fundamentally wrong butwhether, if these assumptions are “essen-tially correct,” is there then a factor (insu-lin) that protected the GIK patients fromthe toxic effects of hyperglycemia? Usingthe model based on the observations inCREATE-ECLA, the answer is probably inthe affirmative. Thus, hyperglycemianeeds to be avoided when designing stud-ies investigating whether insulin adminis-tration is beneficial in AMI.

We have now designed a trial to testthe hypothesis that insulin is cardio-protective in AMI because of its anti-

inflammatory, profibrinolytic, antioxidant,antiapoptotic, vasodilatory, and antiaggre-gatory actions and that these effects are en-hanced by lowering glucose into thenormoglycemic range. We are using intra-venous insulin infusion to lower glucose to90–130 mg/dl in ST elevation myocardialinfarction patients. To allow us to infusea minimal dose of 2.5 units/h (theanti-inflammatory dose of insulin) in thistrial, �7 g/h dextrose appropriately ti-trated will be infused simultaneously tomaintain euglycemia (R.N., P.D., personalcommunication). With sucn an insulin in-fusion regimen, the anti-inflammatory andpotentially cardioprotective effect of insulinis likely to be observed.

References1. The ACE/ADA Task Force on Inpatient

Diabetes: American College of Endocri-nology and American Diabetes Associa-tion consensus statement on inpatientdiabetes and glycemic control: a call toaction. Diabetes Care 29:1955–1962, 2006

2. Dandona P, Mohanty P, Chaudhuri A,Garg R, Aljada A: Insulin infusion in acuteillness. J Clin Invest 115:2069–2072, 2005

3. Vanhorebeek I, Langouche L, Van denBerghe G: Glycemic and nonglycemic ef-fects of insulin: how do they contribute toa better outcome of critical illness? CurrOpin Crit Care 11:304–311, 2005

4. Chaudhuri A, Janicke D, Wilson MF, Tri-pathy D, Garg R, Bandyopadhyay A, Cali-eri J, Hoffmeyer D, Tufail S, Ghanim H,Aljada A, Dandona P: Anti-inflammatoryand pro-fibrinolytic effect of insulin inacute ST-elevation myocardial infarction.Circulation 109:849–854, 2004

5. Visser L, Zuurbier CJ, Hoek FJ, OpmeerBC, de Jonge E, de Mol BA, van Wezel HB:Glucose, insulin and potassium applied asperioperative hyperinsulinaemic normo-glycaemic clamp: effects on inflammatoryresponse during coronary artery surgery.Br J Anaesth 95:448–457, 2005

6. Dandona P, Aljada A, Mohanty P, GhanimH, Hamouda W, Assian E, Ahmad S: In-sulin inhibits intranuclear nuclear factor

kappaB and stimulates IkappaB in mono-nuclear cells in obese subjects: evidencefor an anti-inflammatory effect? J Clin En-docrinol Metab 86:3257–3265, 2001

7. Libby P, Simon DI: Inflammation andthrombosis: the clot thickens. Circulation103:1718–1720, 2001

8. Ross R: Atherosclerosis: an inflammatorydisease. N Engl J Med 340:115–126, 1999

9. Mehta SR, Yusuf S, Diaz R, Zhu J, Pais P,Xavier D, Paolasso E, Ahmed R, Xie C,Kazmi K, Tai J, Orlandini A, Pogue J, LiuL: Effect of glucose-insulin-potassiuminfusion on mortality in patients withacute ST-segment elevation myocardialinfarction: CREATE-ECLA randomizedcontrolled trial. JAMA 293:437– 446,2005

10. Rackley CE, Russell RO Jr, Rogers WJ,Mantle JA, McDaniel HG, Papapietro SE:Clinical experience with glucose-insulin-potassium therapy in acute myocardial in-farction. Am Heart J 102:1038–1049,1981

11. Yusuf S, Mehta SR, Diaz R, Paolasso E,Pais P, Xavier D, Xie C, Ahmed RJ,Khazmi K, Zhu J, Liu L: Challenges in theconduct of large simple trials of importantgeneric questions in resource-poor set-tings: CREATE and ECLA trial programevaluating GIK (glucose, insulin and po-tassium) and low-molecular-weight hep-arin in acute myocardial infarction. AmHeart J 148:1068–1078, 2004

12. Goyal A, Mahaffey KW, Garg J, NicolauJC, Hochman JS, Weaver WD, TherouxP, Oliveira GB, Todaro TG, Mojcik CF,Armstrong PW, Granger CB: Prognosticsignificance of the change in glucoselevel in the first 24 h after acute myo-cardial infarction: results from the CAR-DINAL study. Eur Heart J27:1289 –1297, 2006

13. Cheung NW, Wong VW, McLean M: TheHyperglycemia: Intensive Insulin Infu-sion in Infarction (HI-5) Study: a random-ized controlled trial of insulin infusiontherapy for myocardial infarction. Diabe-tes Care 29:765–770, 2006

14. van der Horst IC, Nijsten MW, VogelzangM, Zijlstra F: Persistent hyperglycemia isan independent predictor of outcome inacute myocardial infarction (Letter). Car-diovasc Diabetol 6:2, 2007

15. Diaz R: Glucose-insulin-potassium inSTEMI patients: results of OASIS-6 studyand a combined analysis with CREATEECLA trial. Paper presented at the Euro-pean Society of Cardiology ScientificCongress, 3 September 2006, Fira GranVia M2, Barcelona, Spain

16. Zhang HX, Zang YM, Huo JH, Liang SJ,Zhang HF, Wang YM, Fan Q, Guo WY,Wang HC, Gao F: Physiologically tolera-ble insulin reduces myocardial injury andimproves cardiac functional recovery inmyocardial ischemic/reperfused dogs.

Table 1—Hyperglycemia-related mortality in CREATE-ECLA based on relationship of admis-sion glucose to mortality in control subjects

Time

Control group GIK group

Blood glucose(mmol/l) Mortality

Blood glucose(mmol/l) Mortality

0 9 11.4 9 11.46 h 8.2 10.2 10.4 13.424 h 7.5 9.1 8.6 10.8

Data are percentages unless otherwise indicated.

Otter and Associates


J Cardiovasc Pharmacol 48:306–313, 200617. Esposito K, Nappo F, Marfella R, Giugli-

ano G, Giugliano F, Ciotola M, QuagliaroL, Ceriello A, Giugliano D: Inflammatorycytokine concentrations are acutely in-

creased by hyperglycemia in humans: roleof oxidative stress. Circulation 106:2067–2072, 2002

18. Srinivasan M, Herrero P, McGill JB, Ben-nik J, Heere B, Lesniak D, Davila-Roman

VG, Gropler RJ: The effects of plasmainsulin and glucose on myocardialblood flow in patients with type 1 diabe-tes mellitus. J Am Coll Cardiol 46:42–48,2005

Targeting glucose in acute myocardial infarction


Diabetes or Impaired Glucose ToleranceDoes the label matter?

CARMEN LARA, MD1,2

SERGIO PONCE DE LEON, MD3

HECTOR FONCERRADA, MD1

MARTIN VEGA, MD4

D iagnosis has been considered a pro-cess of “labeling” with conse-quences that can be both positive

(access to treatment) and negative (socialrejection). The ultimate goal of making adiagnosis is to adequately inform the pa-tient, thereby enhancing knowledge ofthe disorder, adherence to therapeuticadvice, and the ability to manage illnesseffectively.

Impaired glucose tolerance and di-abetes are terms that differentiate twometabolic carbohydrate abnormalities.Establishing the optimum diagnosticlevels for glycemic thresholds, however,depends on balancing the medical, so-cial, and economic costs of labeling apatient who is not at substantial risk fordeveloping complications versus thecorresponding costs of not diagnosing“true” diabetes cases (1).

Among patients who have alterationin metabolizing glucose, varying (in acontrolled setting) the diagnostic criteriafor diabetes and related carbohydrate dis-orders offers an opportunity to assesswhether labeling, either as having “im-paired glucose tolerance” or “diabetes,”influences patients’ knowledge of theirdisease, adherence to therapy, or mecha-nisms for coping and metabolic control.We used a clinical trial design to assess theeffects of diagnostic labeling among par-ticipants randomly assigned to be in-formed that they had either impairedglucose tolerance or diabetes.

RESEARCH DESIGN ANDMETHODS — Participants were adultsattending a primary care clinic, and inclu-

sion criteria were fasting glucose level100–140 mg/dl and glucose level �140but �200 mg/dl 2-h post–75-g glucose.Glucose tests were obtained from routineevaluation. Patients had no previous diag-nosis of impaired glucose tolerance or di-abetes or any additional comorbidity;they also had not ingested medicines thataffect glucose metabolism. All patientsvolunteered to participate in the study.Written consent was not requested be-cause the maneuver being tested involvedonly what was being said to the patient.The study was approved by the local eth-ics committee.

A sample size of 52 patients was cal-culated to detect a difference in compli-ance between the groups (1:1 ratio) of30% (from 2.4 to 3.2 points in the scale),assuming 20% loss to follow-up, with atwo-tailed significance test and � � 0.05with 80% power. Diagnostic labeling wasstandardized via a leaflet containing gen-eral information on problems of glucosemetabolism and how to manage them.Half of the leaflets (randomly) used thelabel “diabetes” and the other half used“impaired glucose tolerance;” the rest ofthe information was identical. The leafletsadvised readers that problems associatedwith glucose are dynamic and could resultin changes in diagnosis, depending on thephase of the illness. The specific variablesstudied were knowledge about the dis-ease (2), compliance with treatment (2),quality of life (3), emotional functioning(3,4), coping mechanisms (4), and glu-cose control.

The study was presented as a “Pro-gram for Glucose Problems.” Randomiza-

tion was carried out in blocks of 13subjects. The informational leaflets werein envelopes numbered from 1 to 52, cor-responding to participants’ sequence insigning up for the study. One of the au-thors, not involved with clinical care ofpatients, did the “labeling” and was alsoresponsible for interviewing the partici-pants and giving them the appropriateleaflet. No recommendations for drugtherapy were given. Family physicianscontinued with patients’ treatment andwere trained not to emphasize either di-agnostic label.

At the first postlabeling visit, 8 weeksafter the baseline visit, the researcher rein-forced the “label” as described in the leaf-let. The final evaluation was carried out16 weeks after baseline. Both postlabelingevaluations were done by an evaluatorblind to group assignment. Differencesbetween groups were analyzed with Stu-dent’s t test and changes within groupswith paired t tests.

RESULTS — We recruited 52 partici-pants: 50 patients (25 in each group) re-mained until the end of the study, ofwhich 42 were women. The mean � SDlevel of glycemia postload was 164 � 15.4mg/dl (range 141–193), and the fastingglucose level at the end of the study was107 � 8.6 mg/dl. No statistically signifi-cant differences (data not shown) werefound between the groups compared at thebaseline and follow-up evaluations. In thiscontext, our primary hypothesis—to detectenhanced compliance in the diabetes-labeled group—was not confirmed.

In comparisons within each group,however, patients who were labeled with“impaired glucose tolerance” significantlyincreased their knowledge about theirdisease on the composite scale (see Table1). Patients labeled with “diabetes” alsoincreased their knowledge of the disease,as measured by both of the evaluationsused. In addition, patients labeled with“diabetes” had decreased scores on emo-tional impact, avoidance distraction, andthe integration subscale (with lower rat-ings showing that a patient is more likelyto accept the prospect of living with dia-betes). For example, the mean score foravoidance distraction decreased from

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1School of Medicine, Autonomous University of Puebla, Puebla, Mexico; the 2National Institute ofPsychiatry, Mexico City, Mexico; the 3National Institute of Medical Sciences and Nutrition, Mexico City,Mexico; and the 4Mexican Institute of Social Security, Puebla, Mexico.

Address correspondence and reprint requests to Carmen Lara, Psychiatry, 13 Sur 2702, CP 72000,Puebla, Mexico. E-mail: [email protected].

Received for publication 20 November 2006 and accepted in revised form 21 August 2007.Published ahead of print at http://care.diabetesjournals.org on 28 August 2007. DOI: 10.2337/dc06-






0.32 to 0.21 (P � 0.002) among the dia-betes-labeled group.

CONCLUSIONS — Telling patientsthey had diabetes had a greater impact inchanging some of their views about thedisease than telling them they had im-paired glucose tolerance. When treatingpatients, physicians should recognize thatdiagnosis (labeling) is a critically impor-tant component of the therapeutic pro-cess. In fact, patients can identify with adiagnostic label despite a lack of under-standing regarding the details of theirillness. Much of the tendency to adhereto treatment depends on whether pa-tients accept the label placed on theirillness (5).

This study has several strengths, in-cluding its randomized trial design,blinded evaluations, and measurementsusing previously validated instruments.Study limitations included a short timeinterval and relatively small sample size,and although information provided to pa-

tients was standardized, it was not possi-ble to control all information patients mayhave received. Overall, the results high-light the potential impact of how diagnos-tic information is provided to patients.

In summary, diagnosis, like measure-ment, assigns subjects to a category ac-cording to defined rules. Feinstein (6) haspointed out that the measurement of clin-ical phenomena must be done after an-swering key questions, e.g., what is theobjective of the measurement? Does treat-ment differ depending on the diagnosisassigned? More recently, the Standardsfor Reporting of Diagnostic Accuracy(STARD) Initiative (7) provided criteriafor evaluation diagnostic accuracy. Thecurrent study emphasizes the additionalconsideration of how patients react to di-agnostic information.

References1. Aguilar Salinas CA, Gomez Perez FJ, Rull

JA: Limitaciones de los criterios de diag-

nostico de la diabetes tipo 2 y la intoler-ancia a la glucosa. Rev Invest Clin 52:177–184, 2000 [article in Spanish]

2. Perez SLM: Apego Terapeutico y ControlMetabolico en el Paciente Diabetico. Dis-sertation. Mexico City, Mexico, Univer-sidad Nacional Autonoma de Mexico,1997

3. Lara Munoz MC, Arcega Domınguez A,Soriano S, Romero T: Caracterısticas psi-cometricas del Cuestionario de ImpactoEmocional de la Diabetes. Psiquiatrıa 18:127–130, 2002 [article in Spanish]

4. Welch GW, Jacobson AM, Polonsky WH:The Problem Areas in Diabetes Scale: anevaluation of its clinical utility. DiabetesCare 20:760–766, 1997

5. Tudge C: In the end is the word. New Sci85:37–38, 1980

6. Feinstein AR: Clinimetrics. New Haven,Connecticut, Yale University Press, 1985

7. Bossuyt P, Reitsma J, Bruns D, Gatsonis C,Glasziou P, Irwig LM, Lijmer JG, MoherD, Rennie D, de Vet HC: Towards com-plete and accurate reporting of studies ofdiagnostic accuracy: the STARD initiative.BMJ 326:41–44, 2003

Table 1—Mean � SD values among patients labeled as having either “impaired glucose tolerance” or “diabetes”

Impaired glucose tolerance (n � 25) Diabetes (n � 25)

1st evaluation 2nd evaluation *P 1st evaluation 2nd evaluation *P

Glucose 105.0 � 8.43 106.1 � 9.84 0.29 102.8 � 9.67 107.7 � 11.23 0.07GHb 6.49 � 0.44 6.46 � 0.36 0.62 6.38 � 0.45 6.60 � 0.42 0.05PAID (total score) 31.9 � 24.4 27.6 � 21.1 0.32 38.6 � 17.8 32.3 � 18.8 0.03Quality of life 3.53 � 0.37 3.57 � 0.29 0.59 3.35 � 0.31 3.44 � 0.31 0.23Avoidance distraction. 0.34 � 0.25 0.26 � 0.24 0.08 0.32 � 0.19 0.21 � 0.19 0.002Tackling spirit 0.81 � 0.10 0.77 � 0.1l 0.26 0.80 � 0.10 0.80 � 0.13 0.62Passive acceptance 0.36 � 0.22 0.29 � 0.18 0.13 0.31 � 0.18 0.25 � 0.16 0.15Integration subscale 0.39 � 20.2 0.33 � 0.20 0.11 0.46 � 0.20 0.35 � 0.21 0.006Knowledge of the disease (composite) 5.08 � 1.36 6.01 � 1.60 0.009 5.50 � 0.83 6.16 � 1.36 0.03Global knowledge of the disease 4.20 � 2.94 5.04 � 3.26 0.21 5.40 � 3.40 7.08 � 2.20 0.01Treatment compliance 2.04 � 0.73 2.20 � 0.82 0.44 1.84 � 0.75 1.92 � 0.86 0.65Global treatment compliance 8.62 � 2.31 8.87 � 1.66 0.84 8.08 � 2.56 8.67 � 1.58 0.27

*P value for paired t test of the differences between 1st and 2nd evaluations within each group. PAID, Problem Areas in Diabetes (ref. 4).

Diabetes or impaired glucose tolerance?


Flexible Intensive Versus ConventionalInsulin Therapy in Insulin-Naive AdultsWith Type 2 DiabetesAn open-label, randomized, controlled, crossover clinical trial of metaboliccontrol and patient preference

CHRISTOF KLOOS, MD1

ALEXANDER SAMANN, MD1

THOMAS LEHMANN, DSC2

ANKE BRAUN, MD3

BARBARA HECKMANN4

ULRICH A. MULLER, MD1

Improving metabolic control can re-duce complications in type 2 diabetes(1–4). Conventional insulin therapy

(CIT) and flexible intensive insulin ther-apy (FIT) are treatment options in insu-lin-dependent type 2 diabetic patients. InCIT, participants inject premixed humaninsulin (30% regular insulin, 70% NPHinsulin) before breakfast and dinner andfollow individually adjusted diet planswith fixed amounts of carbohydrates (5).In FIT, human regular insulin is adjustedbefore main meals according to currentblood glucose readings and desired car-bohydrate intake. When necessary, NPHinsulin is added at bedtime. CIT can beeasy to handle and requires less active di-abetes self-management. In FIT, patientsbenefit from dietary freedom and im-provement in quality of life (6). In pilotstudies, FIT has shown good metaboliccontrol and low risk of hypoglycemia (7).FIT may have additional advantages dueto better postprandial blood glucose con-trol (8).

RESEARCH DESIGN ANDMETHODS — We tested the hypothe-sis that FIT and CIT in insulin-naiveadults with type 2 diabetes are equally ef-

fective in regard to metabolic outcomes.We hypothesized that younger partici-pants, in employment, would prefer FIT.

The trial was designed as a clinical,prospective, randomized, open-label, sin-gle-center, crossover study. The primaryend point was glycosylated hemoglobinA1c (GHb); secondary end points weremild and severe hypoglycemia, insulindosage, blood pressure, BMI, and therapypreference. Participants started insulintherapy either with CIT (group A) or FIT(group B), randomly. Individual insulindosage and carbohydrate intake were de-termined over a 4-week run-in periodwith weekly visits. All visits were the samein both groups and were held in the studycenter. In CIT, daily blood glucose self-control was performed before breakfastand dinner; in FIT, this was done beforemain meals. Oral antidiabetes drugs werenot permitted. The participants com-pleted an outpatient Diabetes Treatmentand Teaching Program with five lessons(90–120 min) during run-in. The run-inwas followed by an 8-week study se-quence until crossover. At crossover, par-ticipants were given one structuredteaching session for refreshing andswitched from CIT to FIT or from FIT to

CIT. After a 1-week run-in period for in-sulin dose adjustment, participants com-pleted the second 8-week studysequence. At trial end, therapeutic prefer-ence was investigated with a structuredinterview.

Eligible participants who failed ther-apeutic goals under oral antidiabetes ther-apy were consecutively recruited(outpatient clinic, Friedrich-Schiller-University). Exclusion criteria were nothaving type 2 diabetes, diabetes duration�2years, not insulin naive, ineffectiveoral antidiabetes therapy �3 months,GHb �7 or �11%, and age �40 or �65years.

GHb was measured using high-performance liquid chromatography(normal range 1.6–5.9%; mean 5.2 �0.33%; TOSOH-Glykohamoglobin-Analyzer-HLC-723-GHbV; Tosoh, To-kyo, Japan). Mild hypoglycemia wasdefined as symptomatic neuroglycopeniaor blood glucose readings �3.3 mmol/l.Severe hypoglycemia required intrave-nous glucose or subcutaneous/intramus-cular glucagon injection.

To have a 90% chance of detecting assignificant (at the two-sided 5% level), a0.5% difference in GHb between the twogroups with an assumed SD of 0.8%, 38participants were required. Intent-to-treat analysis was carried out according toa preestablished analysis plan. Means andSDs were calculated. The t test for pairedand nonpaired samples was used whereappropriate. For statistical analysis of as-sociations of GHb and participant’s char-acteristics, linear mixed-effects modelswere used. P � 0.05 was regarded as sta-tistically significant. Statistical analysiswas performed with SPSS (version 15;SPSS, Chicago, IL).

The study was approved by the localethics committee and was performed ac-cording to the principles of the Declara-tion of Helsinki. Written informedconsent was obtained before participants

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Internal Medicine III, Friedrich-Schiller-University, Jena, Germany; 2Industrie undHandelskammer (Chamber of Commerce) Thuringia, Erfurt, Germany; the 3Bethanien-Hospital, Heidel-berg, Germany; and the 4Hospital-Fulda, Fulda, Germany.

Address correspondence and reprint requests to Christof Kloos, MD, Klinik fur Innere Medizin, ErlangerAllee 101, Friedrich-Schiller-Universitat, 07740 Jena, Germany. E-mail: [email protected].

Received for publication 27 February 2007 and accepted in revised form 19 August 2007.Published ahead of print at http://care.diabetesjournals.org on 23 August 2007. DOI: 10.2337/dc07-

0397. Clinical trial reg. no. NCT00440284, clinicaltrials.gov.U.A.M. has received lecture and other fees from Takeda, Novo Nordisk, Berlin Chemie, Roche Diagnos-

tics, Deutsche BKK, Diabeteszentrum Thuringen, Merck, and Reha aktiv 2000.Abbreviations: CIT, conventional insulin therapy; FIT, flexible intensive insulin therapy.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





took part. The trial was registered withclinicaltrials.gov (NCT00440284).

RESULTS — Baseline clinical data be-tween groups were not significantly dif-ferent (Table 1, group A [first CIT, n �20] vs. group B [first FIT, n � 19]): age56.6 � 7.3 vs. 54.7 � 6.6 years, diabetesduration 7.5 � 4 vs. 8.2 � 4.2 years, totalcholesterol 5.3 � 1.1 vs. 5.3 � 1.1mmol/l, creatinine 82 � 18.4 vs. 77 �16.3 �mol/l, peripheral neuropathy 9 vs.4 participants, and number of partici-pants employed 12 vs. 15, respectively.

All participants completed both studysequences except for one person who re-fused to switch. After initiation of insulintherapy, GHb declined from 8.9 � 1.7 to7.3 � 0.9% (P � 0.0001), and BMI in-creased from 29.4 � 4.4 to 30 � 4.2kg/m2 (P � 0.01). Blood pressure andlipid profiles remained unchanged. Therewas no significant difference between CITand FIT regarding GHb, BMI, blood pres-sue, insulin dosage, and hypoglycemia(Table 1). In linear mixed-effects models,GHb after CIT and GHb after FIT werenot associated with age, sex, diabetes du-ration, initial GHb, blood pressure, lipidprofiles, comorbidity, and occupation.

Twenty participants preferred to con-tinue CIT, whereas 18 opted for FIT. Rea-sons to opt for FIT were therapy flexibility(n � 9), easier therapy (n � 1), and met-abolic control (n � 1). Arguments in favorof CIT were easier therapy (n � 8), fewerinjections (n � 8), and metabolic control(n � 3). In a binary logistic regressionmodel, 88% of therapy decisions were ex-plained by the last therapy option. Themodel was not improved when additionalvariables like age, sex, diabetes duration,GHb after CIT or FIT, initial GHb, bloodpressure, lipid profiles, and occupationwere added.

CONCLUSIONS — Initiation of in-sulin therapy in type 2 diabetes was safeand effective. Metabolic control improvedduring the first study sequence in FIT andCIT but did not further improve in thesecond sequence. Participants did notreach GHb levels below 7%. After havingpracticed CIT and FIT for 8 weeks each,participants preferred their last therapy.This indicates that clinical advantages ofCIT or FIT were of minor importance forparticipants irrespective of age or whetherthey were employed. Interestingly, incontrast to patients with type 1 diabetes,gaining more dietary freedom seems notto be a prevalent motive in type 2 diabetes(6).

There are several limitations to con-sider. The sample size of this randomizedcontrolled trial was too small and thestudy period too short to detect minor dif-ferences. Carry-over effects of the cross-over design can reduce the effect of thesecond trial sequence. The primary out-come measure of this trial (GHb) was asurrogate parameter. Comparison of sideeffects may be limited by different defini-tions of mild and severe hypoglycemia.Participants were young and had earlymanifestation of type 2 diabetes com-pared with the general population (9).Older patients with impaired cognitivefunction might be unable to practice ef-fective diabetes self-management usinginsulin therapy or might be disinterestedin the clinical advantages of FIT and CIT(10).

References1. UK Prospective Diabetes Study (UKPDS)

Group: Intensive blood-glucose controlwith sulphonylureas or insulin comparedwith conventional treatment and risk ofcomplications in patients with type 2 di-abetes (UKPDS 33). Lancet 352:837–853,

19982. Gaede P, Vedel P, Parving HH, Pedersen

O: Intensified multifactorial interventionin patients with type 2 diabetes mellitusand microalbuminuria: the Steno type 2randomised study. Lancet 353:617–622,1999

3. Gaede P, Vedel P, Larsen N, Jensen GV,Parving HH, Pedersen O: Multifactorialintervention and cardiovascular disease inpatients with type 2 diabetes. N Engl J Med348:383–393, 2003

4. Shichiri M, Kishikawa H, Ohkubo Y,Wake N: Long-term results of the Kum-amoto Study on optimal diabetes controlin type 2 diabetic patients. Diabetes Care23 (Suppl. 2):B21–B29, 2003

5. Gruesser M, Hartmann P, Schlottmann N,Joergens V: Structured treatment andteaching programme for type 2 diabeticpatients on conventional insulin treat-ment: evaluation of reimbursement pol-icy. Patient Educ Couns 29:123–130, 1996

6. DAFNE Study Group: Training in flexi-ble, intensive insulin management to en-able dietary freedom in people with type 1diabetes: dose adjustment for normal eat-ing (DAFNE) randomised controlled trial.BMJ 325:746, 2002

7. Kalfhaus J, Berger M: Insulin treatmentwith preprandial injections of regular in-sulin in middle-aged type 2 diabetic pa-tients: a two year observational study.Diabetes Metab 26:197–201, 2000

8. Robertson C: Physiologic insulin replace-ment in type 2 diabetes: optimizing post-prandial glucose control. Diabetes Educ32:423–432, 2006

9. Expert Committee on the Diagnosis andClassification of Diabetes Mellitus: Reportof the Expert Committee on the Diagnosisand Classification of Diabetes Mellitus.Diabetes Care 20:1183–1197, 1997

10. Braun A, Muller UA, Muller R, Leppert K,Schiel R: Structured treatment and teach-ing of patients with type 2 diabetes melli-tus and impaired cognitive function: theDICOF trial. Diabet Med 21:999–1006,2004

Table 1—Main outcomes (intent-to-treat analysis)

Baseline First 8-week sequence Second 8-week sequence

Group A Group B Group A CIT Group B FIT Group A FIT Group B CIT

GHb (%) 8.9 � 1.5 9.2 � 2.1 7.4 � 1 7.3 � 1.1 7.3 � 0.9 7.2 � 1BMI (kg/m2) 29 � 4.1 29.3 � 4.8 29.9 � 3.7 29.7 � 4.5 29.9 � 3.8 30.1 � 4.5SBP (mmHg) 137.3 � 17 137.6 � 17.3 138.3 � 13.3 133.2 � 14.9 137.7 � 17.7 134.4 � 15.1DBP (mmHg) 81 � 10 80.9 � 8.9 81.5 � 8.9 77.9 � 7.3 81.4 � 8.5 77.3 � 7.4Insulin (IU/day) 0 0 34 � 14 37.8 � 15.6 37.1 � 22.5 40.9 � 16.5Mild hypoglycemia 0 0 2 7 4 5Severe hypoglycemia 0 0 0 1 0 0

Data are means � SD or n. Participants were randomized to start insulin therapy either with CIT (Group A) or FIT (Group B); differences were not significant. DBP,diastolic blood pressure; SBP, systolic blood pressure.

Insulin therapy and type 2 diabetes


Perception of Offspring Risk for Type 2Diabetes Among Patients With Type 2Diabetes and Their Adult OffspringMASAKAZU NISHIGAKI, RN, MHLTHSC

1

KOJI KOBAYASHI, RN, PHD1

TAKAKO HITOMI, RN, MHLTHSC1

TAEKO YOKOMURA, RN2

MITSUNAO YOKOYAMA, MD, PHD2

NAOTO SEKI, MD, PHD2

KEIKO KAZUMA, RN, PHD1

R elatives of type 2 diabetic patientsare at higher risk for type 2 diabe-tes, as they are likely to share ge-

netic predisposition and have similarlifestyle habits (1). To actively involvehigh-risk individuals in prevention, rec-ognition of disease risk is crucial (2).Some studies have suggested that Japa-nese are genetically predisposed to diabe-tes (3,4), so recognition of risk by high-risk Japanese individuals is particularlyimportant. In addition, patients with dis-eases involving genetic predispositionneed to advise and warn their offspringabout risks associated with the disease(5,6). Diabetic parents must therefore rec-ognize the high-risk status of their offspring,but no studies have examined risk percep-tion of parents and their offspring in Japan.

This self-administered questionnairesurvey was conducted to clarify percep-tions of Japanese type 2 diabetic patientsregarding risk of diabetes in their off-spring, as well as perceptions of adult off-spring about their own risk for diabetes.

RESEARCH DESIGN ANDMETHODS — Subjects in the presentstudy comprised 164 pairs of type 2 dia-betic patients (aged �75 years) receivingtreatment at a general hospital’s diabetesclinic located in the suburbs of Tokyo,Japan, and their offspring (aged �20 but�50 years with no diabetes or glucoseintolerance). After obtaining written in-formed consent to participate in the study,

the patient and offspring completed ananonymous questionnaire separately, witha unique ID to identify each parent/offspring pair. The present study was con-ducted from October to December 2005.The ethical committee of the University ofTokyo approved all study protocols.

The perception of offspring risk fortype 2 diabetes was assessed among bothpatients and offspring as “the likelihoodof your offspring/you developing diabetesin comparison to the general Japanesepopulation.” The likelihood was evalu-ated from three perspectives: risk due tocurrent lifestyle, risk due to family his-tory, and overall risk. Response categoriesfor each ranged from 1 (“very likely”) to 5(“very unlikely”). Results were tabulated;then, interperspective comparison ineach group and pairwise comparison ineach parent/offspring pair for risk percep-tion were conducted.

RESULTS — Backgrounds of subjectswere as follows: male ratio in patients andoffspring 54.3 and 40.2% (P � 0.01, �2

test), respectively; mean � SD age 64.0 �6.5 and 33.4 � 7.6 years; mean BMI24.0 � 3.6 and 22.9 � 3.8 kg/m2 (P �0.01, t test); mean educational years12.6 � 2.6 and 14.3 � 1.9 years; and58.5% of pairs were living together.Among patients, 23.2% were receiving in-sulin therapy, and 26.2% reported diabe-tes-related complications.

About 40% of patients stated that their

offspring were more likely to develop dia-betes from the perspective of lifestyle habitsand about one-half from the perspective offamily history and an overall view. No inter-perspective differences in risk perceptionwere seen. Among offspring, about one-halfrecognized that they were at higher risk fordiabetes compared with the general popu-lation from the perspective of lifestyle habitsand 63.5% from an overall view. A higherrisk from the perspective of family historywas recognized by 74% of offspring, repre-senting significantly higher risk perceptioncompared with the other two perspectives(vs. lifestyle, P � 0.001; vs. overall view,P � 0.01; Steel-Dwass test for multiplecomparisons) (Table 1).

Pairwise comparison showed that off-springs’ perception of their risk related tofamily history and their overall risk wassignificantly higher than their parents’(P � 0.001, Wilcoxon’s sign-rank sumtest), but no difference was found in per-ception of lifestyle-related risk.

CONCLUSIONS — In the presentstudy, perceptions of risk for diabetesamong offspring were higher comparedwith previous research involving both pa-tients and offspring (7–9). Two possiblecauses may contribute to this higher riskperception. First, optimistic biases aboutrisk perception and cross-cultural varia-tions might exist between current and pre-vious research. Many studies havedescribed people underestimating risks ofunfavorable events, representing optimisticbias (10). In addition, some research hasshown cultural variations in optimism, withWestern people more optimistic than Ori-ental people (11). These factors were relatedto lower risk perception in previous West-ern research. However, a previous study inKorea showed much lower risk perceptionthan that seen in the present research (12),suggesting that the present subjects still dis-play relatively high risk perception even af-ter considering possible pessimistic trendsin Asian countries. Secondly, an increasingawareness of diabetes may have affected thisresult. The present results are comparablewith figures obtained from patients edu-cated about genetic risks (13). In Japan,studies on genetic predispositions for diabe-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Adult Nursing/Palliative Care Nursing, School of Health Sciences and Nursing,Graduate School of Medicine, The University of Tokyo, Tokyo, Japan; and the 2Social Insurance FunabashiCentral Hospital, Chiba, Japan.

Address correspondence and reprint requests to Masakazu Nishigaki, Department of Adult Nursing,7-3-1, Hongo, Bunkyo-ku, Tokyo, Japan. E-mail: [email protected].

Received for publication 9 April 2007 and accepted in revised form 24 August 2007.Published ahead of print at http://care.diabetesjournals.org on 5 September 2007. DOI: 10.2337/dc07-






tes have been conducted as part of nationalprojects since 2000 (4,14), and the massmedia has been actively reporting lifestylediseases such as metabolic syndrome. Thepresent results could indicate that these na-tional projects have successfully raisedawareness of diabetes among the Japanesepopulation.

This study also clarified marked differ-ences in risk perception between diabeticparents and offspring from both quantita-tive and qualitative perspectives. Offspringdisplayed greater recognition of diabetesrisks than parents, contradicting a percep-tion by health care professionals that off-spring view diabetes as something that doesnot concern them (15). More interestingly,interperspective differences were found be-tween patients and offspring as qualitativedifferences in risk perception. Patients per-ceived diabetes risks for offspring as equallylow for lifestyle-related and hereditary risk,whereas offspring perceive hereditary riskas much higher than lifestyle-related risk.This may reflect self-serving thinkingamong both patients and offspring. Parentswould feel guilty about passing susceptibil-ity for a disease to their children (16), thusprobably downplaying hereditary risk to re-duce feelings of guilt about genetic predis-position toward diabetes. Conversely,offspring downplay their own responsibil-ity by emphasizing hereditary risks morethan lifestyle-related risks. These findingshave some implications: if patients are toinform their offspring about disease risksmore effectively, their own potential feel-ings of guilt need to be tackled first. If off-

spring are to take a more active stancetoward prevention, realization of the impor-tance of their own actions is particularly im-portant in terms of risk.

Whether these findings are specific toJapanese diabetic patients and their off-spring remains unclear, due to a limita-tion of this study: not containing a controlgroup of nondiabetic adults. Further re-search investigating risk perceptions inthe Japanese general population is neededto confirm and clarify these findings.

References1. Kuzuya T, Matsuda A: Family histories of

diabetes among Japanese patients withtype 1 (insulin-dependent) and type 2(non-insulin-dependent) diabetes. Diabe-tologia 22:372–374, 1982

2. Rosenstock IM: Why people use healthservices. Milbank Mem Fund Q 44 (Suppl.):94–127, 1966

3. Matsuoka K: Genetic and environmentalinteraction in Japanese type 2 diabetics.Diabetes Res Clin Pract 50 (Suppl. 2):S17–S22, 2000

4. Kadowaki T, Hara K, Yamauchi T, TerauchiY, Tobe K, Nagai R: Molecular mechanismof insulin resistance and obesity. Exp BiolMed (Maywood) 228:1111–1117, 2003

5. Weil J: Psychosocial Genetic Counseling.New York, Oxford University Press, 2000

6. Wilson BJ, Forrest K, van Teijlingen ER,McKee L, Haites N, Matthews E, SimpsonSA: Family communication about geneticrisk: the little known. Community Genet7:15–24, 2004

7. Farmer AJ, Levy JC, Turner RC: Knowl-edge of risk of developing diabetes melli-

tus among siblings of type 2 diabeticpatients. Diabet Med 16:233–237, 1999

8. Pierce M, Harding D, Ridout D, Keen H,Bradley C: Risk and prevention of type IIdiabetes: offspring’s views. Br J Gen Pract51:194–199, 2001

9. Pierce M, Hayworth J, Warburton F, KeenH, Bradley C: Diabetes mellitus in thefamily: perceptions of offspring’s risk.Diabet Med 16:431–436, 1999

10. Weinstein ND: Optimistic biases about per-sonal risks. Science 246:1232–1233, 1989

11. Heine SJ, Lehman DR: Cultural variationin unrealistic optimism: does the west feelmore vulnerable than the east? J Pers SocPsychol 68:595–607, 1995

12. Kim J, Choi S, Kim CJ, Oh Y, Shinn SH:Perception of risk of developing diabetesin offspring of type 2 diabetic patients.Korean J Intern Med 17:14–18, 2002

13. Gnanalingham MG, Manns JJ: Patientawareness of genetic and environmentalrisk factors in non-insulin-dependent di-abetes mellitus–relevance to first-degreerelatives. Diabet Med 14:660–662, 1997

14. Nishigaki M, Kobayashi K, Shibayama T,Kadowaki T, Kazuma K: [Attitude of dia-betes care specialists to prevention of di-abetes to relatives of patients with type 2diabetes]. J Jpn Diabetes Soc 49:669–676,2006 [in Japanese]

15. Haga H, Yamada R, Ohnishi Y, NakamuraY, Tanaka T: Gene-based SNP discoveryas part of the Japanese Millennium Ge-nome Project: identification of 190,562genetic variations in the human genome:single-nucleotide polymorphism. J HumGenet 47:605–610, 2002

16. Chapple A, May C, Campion P: Parentalguilt: the part played by the clinical genet-icist. J Genet Couns 4:179–191, 1995

Table 1—Patients’ and offsprings’ perception of risk for diabetes from three perspectives

Very unlikely UnlikelySame as general

population Likely Very likely

Interperspective comparison

P P

PatientCurrent lifestyle habits 3 (1.8) 24 (14.6) 65 (39.6) 54 (32.9) 18 (11.0) 0.56* 0.93†Family history 8 (4.9) 26 (15.9) 44 (26.8) 57 (34.8) 29 (17.7) 0.56‡ 0.74†Overall view 5 (3.0) 22 (13.4) 59 (36.0) 59 (36.0) 18 (11.0) 0.93‡ 0.74*

OffspringCurrent lifestyle habits 3 (1.8) 17 (10.4) 62 (37.8) 59 (36.0) 23 (14.0) �0.001* 0.07†Family history 1 (0.6) 10 (6.1) 31 (18.9) 74 (45.1) 48 (29.3) �0.001‡ 0.01†Overall view 2 (1.2) 10 (6.1) 48 (29.3) 78 (47.6) 26 (15.9) 0.07‡ 0.01*

Offsprings’ riskperception is:

Lower thanhis/her parent

Same ashis/her parent

Higher thanhis/her parent

Pairwisecomparison (P)

Current lifestyle habits 44 (26.8) 60 (36.6) 60 (36.6) 0.07Family history 25 (15.2) 51 (31.1) 88 (53.7) �0.001Overall view 27 (16.6) 69 (42.3) 67 (41.1) �0.001

Data are n (%) unless otherwise indicated. Total number of subjects is 164; however, total number in �Overall view� perspective is 163 due to a patient’s missing value.*Compared with �Family history� perspective; †compared with �Overall view� perspective; ‡compared with �Current lifestyle habits� perspective.

Diabetes risk perception of patient and offspring


Trends in Hospitalizations for DiabetesAmong Children and Young AdultsUnited States, 1993–2004

JOYCE M. LEE, MD, MPH1,2

MEGUMI J. OKUMURA, MD3

GARY L. FREED, MD, MPH2

RAM K. MENON, MD1

MATTHEW M. DAVIS, MD, MAPP2,4,5

OBJECTIVE — The purpose of this study was to examine national trends in hospitalizationsassociated with diabetes for U.S. children and young adults.

RESEARCH DESIGN AND METHODS — The study included hospital discharges forindividuals aged 0–29 years with a diagnosis of diabetes (250.xx) in the Nationwide InpatientSample (1993–2004). Outcomes were weighted, nationally representative estimates of the fre-quency of population-adjusted hospital discharges and hospital charges (2004 $U.S.).

RESULTS — Among individuals aged 0–29 years, population-adjusted rates of hospitaliza-tions associated with diabetes over the 12-year period increased by 38% (99.1 of 100,000 in1993 and 136.4 of 100,000 in 2004; P � 0.001 for curvilinear trend). Age-specific increases inannual hospitalizations for diabetes occurred primarily among individuals aged 20–24 years(152.6 of 100,000 in 1993 and 222.2 of 100,000 in 2004) and 25–29 years (224.9 of 100,000in 1993 and 331.2 of 100,000 in 2004). Trends in hospitalizations among younger individualsshowed no significant patterns. Hospitalization rates were consistently higher for females thanfor males, with a greater rate of increase for females (42%) than for males (29%) (P � 0.001).Inflation-adjusted total charges for diabetes hospitalizations increased 130%, from $1.05 billionin 1993 to $2.42 billion in 2004.

CONCLUSIONS — The number of young adults hospitalized with diabetes in the U.S. hasincreased significantly over the last decade. Sex-specific differences in hospitalization rates andtrends in obesity among U.S. children may amplify future trends in diabetes hospitalizations andcorresponding rapid growth in associated health care expenditures.

Diabetes Care 30:3035–3039, 2007

S tudies indicate that the burden ofdiabetes, type 1 and type 2, is sub-stantial (1) and rising among U.S.

children (2– 4). Diabetes registries inPhiladelphia (5), Pittsburgh (6), and Chi-cago (3) have reported increasing rates oftype 1 diabetes throughout the 1990s,and recent data from the Colorado InsulinDependent Diabetes Mellitus Registry

showed a 2.3% increase per year in inci-dence of type 1 diabetes over the last twodecades (2). The Chicago Childhood Di-abetes Registry also reported significantincreases in rates of type 2 diabetes amongAfrican-American and Latino childrenduring 1985–2001 (3), presumably dueto trends in childhood obesity.

Given the increasing numbers of chil-

dren with diabetes and the considerablemorbidity and associated health care ex-penditures, we wished to evaluate na-tional trends in hospitalizations associatedwith diabetes for children and young adults.One previous study evaluated trends inhospital discharges associated with diabe-tes for children from the National Hospi-tal Discharge Survey (7), reporting aprevalence rate of 1.43% during 1979–1981 and 2.36% during 1997–1999.However, that study only included diabe-tes discharges associated with an obesitydiagnosis, which is uncommonly coded,did not evaluate population-adjustedtrends for specific age strata, and did notinclude hospital charge data (7).

We used the Nationwide InpatientSample (NIS), a nationally representativeannual sample of discharges from nonfed-eral, short-term, general, and other spe-cialty hospitals in the U.S., to assesstrends in hospitalizations and hospitalcharges associated with a diabetes diagno-sis from 1993 to 2004. Based on trends indiabetes reported from previous studies,we hypothesized that there would be in-creasing prevalence rates of discharges as-sociated with diabetes in children andyoung adults, with corresponding in-creases in hospital charges.

We chose to include individuals aged20–29 years because one study from the1990s suggested that increases in diabetesamong young adults in the U.S. weremarked (8). We wished to evaluate trendsin diabetes over the early life course, per-mitting a comparison of trends amongchildren with concurrent trends amongyoung adults.


Data sourcesThe NIS is a publicly available, deidenti-fied annual database of hospital inpatientstays, sponsored by the Agency forHealthcare Research and Quality (9),which includes data on ICD-9 codes andhospital charges. The NIS represents alldischarges from an approximate 20%stratified sample of U.S. community hos-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Pediatric Endocrinology Unit, University of Michigan, Ann Arbor, Michigan; the 2Child HealthEvaluation and Research (CHEAR) Unit, University of Michigan, Ann Arbor, Michigan; the 3Division ofGeneral Pediatrics, University of California, San Francisco, San Francisco, California; the 4Department ofInternal Medicine, University of Michigan, Ann Arbor, Michigan; and the 5Gerald Ford School of PublicPolicy, University of Michigan, Ann Arbor, Michigan.

Address correspondence and reprint requests to Joyce Lee, MD, MPH, 300 NIB, Room 6E05, Ann Arbor,MI 48109-5456. E-mail: [email protected].

Received for publication 20 April 2007 and accepted in revised form 22 August 2007.Published ahead of print at http://care.diabetesjournals.org on 28 August 2007. DOI: 10.2337/dc07-

0769.Abbreviations: CPI, Consumer Price Index; NIS, Nationwide Inpatient Sample; PPV, positive predictive

value.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion



E p i d e m i o l o g y / H e a l t h S e r v i c e s R e s e a r c hO R I G I N A L A R T I C L E


pitals, with stratification based on hospi-tal region, urban/rural location, teachingstatus, number of beds, and ownership.NIS data from 1993 to 2004 were in-cluded in this analysis.

Study populationWe searched for diabetes diagnoses (ICD-9-CM codes 250.xx) in any of 15 diagnos-tic positions associated with dischargesfor individuals aged 0–29 years. We con-sidered inclusion of other diagnosticcodes including gestational diabetes butchose to focus on the 250.xx diabetescodes alone to make our analysis morecomparable with other clinical studies.Codes specific for type 1 diabetes are in-dicated by a fifth digit of 1 (250.x1) or 3(250.x3) and for type 2 diabetes are indi-cated by a fifth digit of 0 (250.x0) or 2(250.x2). One recent study in childrenand young adults showed a positive pre-dictive value (PPV) of 97% for type 1 di-abetes diagnostic codes but only 16% fortype 2 diabetes codes (10) (i.e., whereas97% of individuals with diagnostic codesfor type 1 diabetes had type 1 diabetesverified by medical chart review, only16% of individuals with diagnostic codesfor type 2 diabetes had type 2 diabetes).The low PPV of type 2 diabetes diagnosticcodes was mostly due to the misclassifica-tion of type 1 for type 2 diabetes. Because74% of individuals with type 2 diabetesdiagnostic codes were verified as havingeither type 1 or type 2 diabetes (PPV 74%)(10), we evaluated discharges for bothtypes combined. Among females, dis-charges associated with childbirth wereidentified using diagnosis-related groupcodes 370–376 (cesarean/vaginal deliv-ery) and 383 (other antepartum diag-noses with medical complications).

Data analysisDischarge-level weights for each yearwere used to estimate national hospital-ization rates, total charges, and corre-sponding SE estimates. To obtainhospitalization rates standardized to theconcurrent national population, U.S.Census data for each respective year(1993–2004) were used to calculate ratedenominators for the overall populationand for sex- and age-stratified analyses.Standardization to the U.S. populationwith diabetes was not performed, as dia-betes prevalence data among individuals0–29 years for each specific year are un-available. The SEARCH (Search for Dia-betes in Youth) study estimated theprevalence of diabetes among children

but only for 2001 (1). Rates of missingdata by sex (0%) and insurance type(1.2%) were low, but trends by race/ethnicity could not be determined be-cause of significant missing data (�25%).We were unable to look at trends in age-specific death rates associated with diabe-tes because of the small number of deathsin age-specific strata.

All analyses were conducted usingSTATA 9.0, with application of appropri-ate weights to account for the complexsampling design and to allow for extrap-olation to national population estimates.Taylor series linearization was used forvariance estimation. All results presentedare weighted estimates.

For assessing trends in health carespending by payer type (Medicaid versusprivate), total hospital charges were esti-mated for each year using the appropriatedischarge weights and standardized to2004 U.S. dollars using the ConsumerPrice Index (CPI). We chose to use theoverall CPI because of concerns that themedical care CPI does not accurately cap-ture the cost of health care for third-partypayers who would be paying the vast ma-jority of claims for hospital services (11).The charge data were analyzed for outli-ers, and trend analyses were performedafter removing the top 1% of totalcharges, with similar results (data notshown).

Because of the complex survey designof the NIS, sampling weights are changedannually to reflect increases in the num-ber of states participating. Therefore,combining data from each of the 12 yearsinto a single dataset would lead to inaccu-rate point and variance estimates. To de-termine whether there were significant

increases in hospitalizations and totalcharges, variance-weighted tests for linearand curvilinear trends were performed,which do not assume homogeneity ofvariance and incorporate the standard er-rors of the estimates for each year. To testthe hypothesis that the rate of change var-ied by age, two different variance-weighted least-squares regression modelswere run: 1) a model for linear trend, in-cluding age-group, year, and the interac-tion between age-group and year; and 2) amodel for curvilinear trend, includingage-group, year squared, and the interac-tion between age-group and year squared.Similar regression models were run to testfor sex-specific differences in rate ofchange.

RESULTS — For individuals aged0–29 years during the years 1993 and2004, the NIS sample included data for2,112,556 and 2,305,258 unweighteddischarges, respectively, representing11,143,316 and 11,099,327 dischargesannually. Table 1 shows sample charac-teristics of discharges associated with di-abetes by age-group, sex, and insurancetype.

Annual rates of hospitalizationsassociated with diabetesFigure 1 presents population-adjustedannual rates of hospitalizations associatedwith diabetes, type 1 and type 2 com-bined, for individuals aged 0–29 yearsover the 12-year study period. Overall,there was a 38% increase in the number ofpopulation-adjusted hospitalizations as-sociated with diabetes, with 99.1 of

Table 1—Sample characteristics for 1993 and 2004

1993 2004

Weighted no. of diabetes discharges(n � unweighted)

111,313 (20,867) 166,509 (34,517)

Age-groups0–9 years 8.0 7.0

10–14 years 11.9 9.615–19 years 16.4 16.620–24 years 25.2 28.025–29 years 38.6 38.9

Female sex 60.4 62.5Insurance type

Medicaid 31.2 37.9Private 43.3 38.0Other 25.5 24.2

Data are %.

Diabetes trends in children and young adults


100,000 in 1993 and 136.4 of 100,000 in2004 (P � 0.001 for curvilinear trend).

Age-specific patternsFigure 2 illustrates the population-adjusted annual rates of hospitalizationsassociated with diabetes across agestrata. There were no significant in-creases in hospitalization rates for dia-betes for children aged �20 years.However, there were significant increasesin hospitalizations among individualsaged 20–24 and 25–29 years (P � 0.001for curvilinear trend).

Sex-specific patternsSex-specific hospitalization rates, whichincreased over the 12-year period for bothsexes, are also shown in Fig. 1 (P � 0.001for curvilinear trend). Rates of hospital-izations were consistently higher for fe-males than for males throughout the timeperiod (122.1 of 100,000 for females vs.77.0 of 100,000 for males in 1993 and174.1 of 100,000 for females vs. 99.3 of100,000 for males in 2004). Even afterremoving hospitalizations associated withchildbirth, which accounted for �15% ofdischarges overall throughout the study

period, rates for females remained higher(data not shown). Furthermore, increasesin hospitalization rates were significantlygreater for females (42%) than for males(29%) (P � 0.001).

Concurrent increases in hospitalchargesInflation-adjusted annual aggregate hos-pital charges for diabetes over the studyperiod increased 130%, from $1.05 bil-lion in 1993 to $2.42 billion in 2004, withsignificant increases in total charges forMedicaid and private payers (P � 0.001for curvilinear trend). Figure 3 illustratesthe proportion of total charges by Medic-aid, private insurers, and “other” insur-ance category (including uninsured) foreach year. In 2004, aggregate estimatedhospital charges were $924 million forMedicaid and $849 million for privatepayers.

CONCLUSIONS — This is the firstanalysis of which we are aware to use na-tionally representative hospital dischargedata to document statistically significantincreasing prevalence rates of hospitaliza-tions and associated hospital charges forindividuals with diabetes aged 0–29 yearsover a recent 12-year period. Strengths ofthis study include the representative na-ture of the NIS with sample sizes largeenough to examine year-to-year trends inhospitalizations, the time period coveringover a decade, and the inclusion of indi-viduals over the early life course, fromchildhood through young adulthood.

We found significant increases inpopulation-adjusted hospitalization ratesamong individuals aged 0–29 years over-all, but the increases were chiefly attrib-utable to significant increases in annualhospitalization rates among young adultsaged 20–29 years. Given that type 2 dia-betes accounts for 90–95% of incidentcases in adulthood (12), the increase inyoung adults is consistent with an in-crease in the prevalence of type 2 diabe-tes. We speculate that this increase indiabetes among young adults reflects thegrowing prevalence of childhood obesityin the U.S. and the physiological connec-tion between obesity and type 2 diabetes.

Obesity is the hypothesized criticalrisk factor contributing to the elevatedrisk of type 2 diabetes in individuals (13).Although severity of obesity is an impor-tant risk factor for development of type 2diabetes (14), duration of obesity is also acritical risk factor, with studies showinghigher incidence rates of diabetes in indi-

Figure 1—Estimated annual hospitalizations associated with diabetes (type 1 and type 2 com-bined) among U.S. children and young adults aged 0–29 years overall and by sex. F, all individ-uals aged 0–29 years; �, females; Œ, males. Error bars indicate 95% CI.

Figure 2—Estimated annual hospitalizations associated with diabetes (type 1 and type 2 com-bined) by age strata among U.S. children and young adults aged 0–29 years. F, 0–9 years; f,10–14 years; Œ, 15–19 years; E, 20–24 years; �, 25–29 years. Error bars indicate 95% CI.

Lee and Associates


viduals with a longer duration of obesi-ty.(15,16) Because progression fromobesity to development of type 2 diabetesis hypothesized to occur over a period ofseveral years (17), latency between thedevelopment of obesity in childhood andthe onset of clinically diagnosed type 2diabetes years later could contribute tothe increase in hospitalization rates inyoung adulthood observed in these data.

Specifically, we found increases inhospitalizations among young adults oc-curring after the year 2000. Individualsaged 20–29 years after 2000 represent abirth cohort who experienced childhoodand adolescence in the late 1970s andearly 1980s, at the leading edge of thechildhood obesity epidemic (18). There-fore, the cohort effects of childhood obe-sity from the 1970s to the 1980s could bepresenting just now as the first manifesta-tion of a related “diabetes epidemic”among young adults. Other studies maymore precisely elucidate the impact ofchildhood obesity on type 2 diabetes ratesin young adults and how the changingdynamics of obesity in children will affectfuture type 2 diabetes rates in the U.S.population. Of great concern are the factsthat studies have documented that theprevalence of obesity among U.S. chil-dren has doubled over the last two de-cades (18), obesity onset is occurring atyounger ages (18), and the severity ofobesity in children has increased overtime (19). Therefore, further increases inrates of type 2 diabetes among young

adults, beyond what is evident by 2004,appear very likely and perhaps inevitable.

Contrary to our hypothesis, increasesin diabetes were not observed for the pe-diatric age strata. We know of one otherstudy in which trends in diabetes preva-lence among children were evaluated onthe basis of diagnostic codes from outpa-tient as well as inpatient claims data (20).In contrast to the findings of our study,that analysis showed significant increasesin diabetes rates, in particular type 1 dia-betes. Because the NIS tracks inpatienthospitalizations and not outpatient visits,our study may have underestimatedtrends in diabetes, given that an increas-ing number of children with new-onsettype 1 diabetes are no longer admitted tothe hospital but are managed as outpa-tients and that individuals who presentwith new-onset type 2 diabetes maypresent less often with ketosis comparedwith individuals with new-onset type 1diabetes and can often be managed ini-tially with lifestyle management and oralmedications as outpatients.

Nevertheless, our study is the first todocument increasing trends in aggregatenational hospital charges for diabetesamong children and young adults. Previ-ous studies have evaluated hospitaliza-tions and costs primarily for older adults(21), and one pediatric study did not eval-uate total hospital charges but rather ap-plied a standard average hospital costmultiplied by the number of diabetes dis-charges (7). We found evidence of in-

creasing trends in hospital charges forboth Medicaid and private insurers, ap-proaching $1 billion in annual charges foreach group. With the continuing epi-demic of childhood obesity and increas-ing trends in type 2 diabetes amongyoung adults, the economic burden at-tributable to diabetes will probably con-tinue to rise, affecting public andprivate insurance plans alike. This eco-nomic reality may serve as a critical impe-tus for payers to consider coveringservices that may reduce or otherwise ad-dress obesity among their younger enroll-ees before the onset of diabetes andrelated hospitalizations.

Our finding of higher rates of hospi-talizations among females than amongmales, even after exclusion of childbirth-related discharges, is consistent with pre-vious studies of state-based dischargedata from California (22) and North Caro-lina (23), which also documented similarsex differences in hospital discharges.This finding may be related to the higherprevalence (1) and incidence (4) of diabe-tes in U.S. females versus males. Ourstudy is unique in that we found a largerrate of increase in diabetes hospitaliza-tions for females (42%) than for males(29%) over the 12-year period. We spec-ulate that larger increases among femalesover this period may be related to greatermorbidity, as one recent U.S. study re-ported a doubling of all-cause mortalityamong adult females with diabetes be-tween 1971 and 2000, in contrast with a43% decrease in all-cause mortality inadult males with diabetes (24). However,that study reported mortality only amongindividuals with diabetes who were �35years (24), suggesting the need for furtherstudies to better understand sex differ-ences in diabetes hospitalization rates foryounger individuals.

LimitationsThere are limitations to our study. Dis-charges associated with diabetes do notnecessarily represent a new diagnosis ofdiabetes but may represent multiple re-peat hospitalizations for individuals,which could lead to overestimates of thetrends in diabetes based on hospital dis-charges. However, the fact that the likeli-hood of hospitalizations for diabetes hasdecreased over the study period due tochanges in clinical practice means thatobserved increases in hospitalization ratesare all the more remarkable.

Type 2 diabetes has received greaterattention during the study period, which

Figure 3—Estimated total charges associated with hospitalizations for diabetes among childrenand young adults aged 0–29 years by payer type. f, Medicaid; z, private payers; p, other payers.Error bars indicate 95% CI.

Diabetes trends in children and young adults


may have led to increased provider aware-ness of diabetes and a higher likelihood ofdocumenting diabetes diagnoses associ-ated with hospital discharges over time.Furthermore, the proportion of undiag-nosed to diagnosed diabetes has de-creased over time (25), and the definitionof diabetes based on fasting plasma glu-cose was lowered to 126 mg/dl in 1997,leading to possible year-specific biases inestimation of diabetes trends. Increases indiabetes hospitalizations may also be dueto an increased burden of type 1 ratherthan type 2 diabetes, as some studies haveshown increases in the incidence of type 1diabetes (2,6). Finally, rates of dischargesassociated with diabetes in states whoparticipated in the NIS may have been dif-ferent from those of states that did notparticipate.

ImplicationsIncreasing rates of hospitalizations andassociated expenditures among youngadults with diabetes over the last decadesuggest the need for further studies to ex-amine trends in diabetes prevalenceamong young adults and to understandhow the childhood obesity epidemic inthe U.S. may further amplify these trends.Increases in hospitalizations and expendi-tures may present third-party payers witha strong impetus to cover services that ad-dress prevention and treatment of obesityin younger generations.

Acknowledgments— J.M.L. was supportedby National Institutes of Health (National In-stitute of Child Health and Human Develop-ment) Pediatric HSR Training Grant T32HD07534-05 and the Clinical Sciences ScholarsProgram. This work used the biostatistics coreof the Michigan Diabetes Research and Train-ing Center funded by the National Institute ofDiabetes and Digestive and Kidney DiseasesGrant 5P60 DK20572.

We thank Achamyeleh Gebremariam for histechnical assistance.

This study was presented in abstract form atthe annual meeting of the Pediatric AcademicSocieties, Toronto, Ontario, Canada, 5–8 May2007.

References1. Liese AD, D’Agostino RB Jr, Hamman RF,

Kilgo PD, Lawrence JM, Liu LL, Loots B,Linder B, Marcovina S, Rodriguez B, Stan-

diford D, Williams DE: The burden of di-abetes mellitus among US youth:prevalence estimates from the SEARCHfor Diabetes in Youth Study. Pediatrics118:1510–1518, 2006

2. Vehik K, Hamman RF, Lezotte D, NorrisJM, Klingensmith G, Bloch C, Rewers M,Dabelea D: Increasing incidence of type 1diabetes in 0- to 17-year-old Coloradoyouth. Diabetes Care 30:503–509, 2007

3. Lipton RB, Drum M, Burnet D, Rich B,Cooper A, Baumann E, Hagopian W:Obesity at the onset of diabetes in an eth-nically diverse population of children:what does it mean for epidemiologists andclinicians? Pediatrics 115:e553–e560,2005

4. Dabelea D, Bell RA, D’Agostino RB Jr, Im-peratore G, Johansen JM, Linder B, LiuLL, Loots B, Marcovina S, Mayer-Davis EJ,Pettitt DJ, Waitzfelder B: Incidence of di-abetes in youth in the United States. JAMA297:2716–2724, 2007

5. Lipman TH, Chang Y, Murphy KM: Theepidemiology of type 1 diabetes in chil-dren in Philadelphia 1990–1994: evi-dence of an epidemic. Diabetes Care 25:1969–1975, 2002

6. Libman I: Was there an epidemic of dia-betes in non-white adolescents in Allegh-eny County, Pennsylvania. Diabetes Care21:1278–1281, 1998

7. Wang G, Dietz WH: Economic burden ofobesity in youths aged 6 to 17 years:1979–1999. Pediatrics 109:e81–e81,2002

8. Mokdad AH, Ford ES, Bowman BA, Nel-son DE, Engelgau MM, Vinicor F, MarksJS: Diabetes trends in the U.S.: 1990–1998. Diabetes Care 23:1278–1283,2000

9. Health Care Cost and Utilization inProject (HCUP): Nationwide InpatientSample (NIS). Rockville, MD, Agency forHealthcare Research and Quality, 2003

10. Rhodes ET, Laffel LM, Gonzalez TV, Lud-wig DS: Accuracy of administrative cod-ing for type 2 diabetes in children,adolescents, and young adults. DiabetesCare 30:141–143, 2007

11. Consumer Price Index: Cost-of-Living Con-cepts and the Housing and Medical CareComponents. Washington, DC, Govern-ment Accounting Office, 1996

12. Engelgau MM, Geiss LS, Saaddine JB,Boyle JP, Benjamin SM, Gregg EW, Tier-ney EF, Rios-Burrows N, Mokdad AH,Ford ES, Imperatore G, Narayan KM: Theevolving diabetes burden in the UnitedStates. Ann Intern Med 140:945–950,2004

13. Knowler WC, Pettitt DJ, Saad MF, CharlesMA, Nelson RG, Howard BV, Bennett PH:Obesity in the Pima Indians: its magni-

tude and relationship with diabetes.Am J Clin Nutr 53 (Suppl. 6):1543S–1551S, 1991

14. Fox CS, Pencina MJ, Meigs JB, Vasan RS,Levitzky YS, D’Agostino RB Sr: Trends inthe incidence of type 2 diabetes mellitusfrom the 1970s to the 1990s: the Fra-mingham Heart Study. Circulation 113:2914–2918, 2006

15. Everhart JE: Duration of obesity increasesthe incidence of NIDDM. Diabetes41:235–240, 1992

16. Wannamethee SG, Shaper AG: Weightchange and duration of overweight andobesity in the incidence of type 2 diabetes.Diabetes Care 22:1266–1272, 1999

17. Mokdad AH, Ford ES, Bowman BA, Nel-son DE, Engelgau MM, Vinicor F, MarksJS: The continuing increase of diabetes inthe U.S. Diabetes Care 24:412, 2001

18. Ogden CL, Carroll MD, Curtin LR, Mc-Dowell MA, Tabak CJ, Flegal KM: Preva-lence of overweight and obesity in theUnited States, 1999–2004. JAMA 295:1549–1555, 2006

19. Jolliffe D: Extent of overweight among USchildren and adolescents from 1971 to2000. Int J Obes 28:4–9, 2004

20. Kemper AR, Dombkowski KJ, Menon RK,Davis MM: Trends in diabetes mellitusamong privately insured children, 1998–2002. Ambul Pediatr 6:178–181, 2006

21. Aubert RE, Geiss LS, Ballard DJ, Coca-nougher B, Herman WH: Diabetes-related hospitalization and hospitalutilization. In Diabetes in America. 2nd ed.Bethesda, MD, National Institute of Dia-betes and Digestive and Kidney Diseases,1995, p. 553–569

22. California Department of Health Services:Diabetes and Diabetic Complications: Deathsand Hospitalizations in California, 1983–1987. Sacramento, CA, California Chronicand Sentinel Diseases Surveillance Program,1992 (Tech. rep. no. 9)

23. North Carolina Department of Environ-ment, Health, and Natural Resources: Di-abetes Surveillance in North Carolina FinalEvaluation Report, FY90-FY93. Raleigh,NC, State Center for Health and Environ-mental Statistics, 1993

24. Gregg EW, Gu Q, Cheng YJ, Narayan KM,Cowie CC: Mortality trends in men andwomen with Diabetes, 1971–2000. AnnIntern Med 147:149–155, 2007

25. Gregg EW, Cadwell BL, Cheng YJ, CowieCC, Williams DE, Geiss L, Engelgau MM,Vinicor F: Trends in the prevalence andratio of diagnosed to undiagnosed diabe-tes according to obesity levels in the U.S.Diabetes Care 27:2806–2812, 2004

Lee and Associates


Influence of Family History of Diabetes onIncidence and Prevalence of LatentAutoimmune Diabetes of the AdultResults from the Nord-Trøndelag Health Study

SOFIA CARLSSON, PHD1

KRISTIAN MIDTHJELL, MD, PHD2

VALDEMAR GRILL, MD, PHD3

OBJECTIVE — The aim of this study was to investigate the association between family historyof diabetes (FHD) and prevalence and incidence of latent autoimmune diabetes of the adult(LADA), type 1 diabetes, and type 2 diabetes.

RESEARCH DESIGN AND METHODS — The results were based on cross-sectionaldata from 64,498 men and women (aged �20 years) who were in the Nord-Trøndelag HealthStudy, which included 128 cases of LADA, 1,134 cases of type 2 diabetes, and 123 cases of type1 diabetes. In addition, prospective data on 46,210 subjects, which included 80 incident casesof LADA, observed between 1984 and 1986 and 1995 and 1997 were available. Patients withLADA had antibodies against GAD and were insulin independent at diagnosis.

RESULTS — FHD was associated with a four times (odds ratio [OR] 3.92 [95% CI 2.76–5.58]) increased prevalence of LADA. Corresponding estimates for type 2 and type 1 diabeteswere 4.2 (3.72–4.75) and 2.78 (1.89–4.10), respectively. Patients with LADA who had FHDhad lower levels of C-peptide (541 vs. 715 pmol/l) and were more often treated with insulin (47vs. 31%) than patients without FHD. Prospective data indicated that subjects with siblings whohad diabetes had a 2.5 (1.39–4.51) times increased risk of developing LADA during the 11-yearfollow-up compared with those without.

CONCLUSIONS — This study indicates that FHD is a strong risk factor for LADA and thatthe influence of family history may be mediated through a heritable reduction of insulinsecretion.

Diabetes Care 30:3040–3045, 2007

L atent autoimmune diabetes of theadult (LADA) is a common form ofdiabetes, but the risk factors, includ-

ing the impact of family history of diabe-tes (FHD), are less well understood thanthose for type 1 and type 2 diabetes (1).Familial clustering of diabetes is believed

to be due to a combination of shared ge-netic and environmental factors. For type1 diabetes, the genetic influence has beenlocated to the histocompatibility (HLA)region of chromosome 6 (2), whereas thegenetic background for type 2 diabetesremains largely unknown. Studies indi-

cate that LADA has the same genetic fea-tures characteristic of type 1 diabetes,including an increased frequency of HLA-DQB1 genotypes (3,4). On the otherhand, results from a British study indi-cated that 33% of patients with LADAhave relatives with type 2 diabetes (5).These findings suggest that LADA mayshare inherited features with both type 1and type 2 diabetes.

Epidemiological studies indicate athree to four times increased risk of type 2diabetes in subjects with close relativeswith diabetes (6–8). For type 1 diabetes,a 15 times increased risk has been re-ported in siblings of diabetic patients (2).The risk of type 1 and type 2 diabetes isknown to increase with an increasingnumber of affected relatives (6,7,9). It hasalso been shown that the risk varies, de-pending on which relative(s) has diabetes.For type 1 diabetes, several studies haveshown that having a father with diabetesis associated with a higher risk than hav-ing a mother with diabetes (10). For type2 diabetes, on the other hand, some stud-ies have suggested a preferential maternaleffect (7,11). To what extent the risk ofLADA is influenced by family history ofdiabetes is largely unknown.

The Nord-Trøndelag Health Survey(HUNT) is a large, population-basedstudy in which cases of diabetes havebeen classified according to clinical his-tory and the presence or absence of GADantibodies. We used these data to investi-gate the influence of FHD on the preva-lence and incidence of LADA comparedwith those for type 1 and type 2 diabetes.


HUNT 1From 1984 to 1986, all inhabitants of theNorwegian county of Nord-Trøndelagwho were aged �20 years were invited totake part in HUNT 1 (n � 85,100) (12).The survey featured a clinical examina-tion, including measurements of height,weight, and blood pressure and question-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Division of Epidemiology, Stockholm Centre of Public Health and Department of Epidemiology,Institute of Environmental Medicine, Karolinska Institutet, Stockholm, Sweden; the 2Nord-TrøndelagHealth Study Research Centre, Department of Community Medicine and General Practice, NorwegianUniversity of Science and Technology; and the 3Norwegian University of Science and Technology Instituteof Cancer Research and Molecular Medicine, Norwegian University of Science and Technology and Depart-ment of Molecular Medicine, Karolinska Institutet, Stockholm, Sweden.

Address correspondence and reprint requests to Sofia Carlsson, Division of Epidemiology, Norrbacka,S-171 76 Stockholm, Sweden. E-mail: sofia.carlsson@ ki.se .

Received for publication 12 April 2007 and accepted in revised form 12 September 2007.Published ahead of print at http://care.diabetesjournals.org on 18 September 2007. DOI: 10.2337/dc07-

0718.Abbreviations: FHD, family history of diabetes; HUNT, Nord-Trøndelag Health Survey; LADA, latent

autoimmune diabetes of the adult.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion



E p i d e m i o l o g y / H e a l t h S e r v i c e s R e s e a r c hO R I G I N A L A R T I C L E


naires with questions on current health,diabetes, and lifestyle factors such assmoking and alcohol consumption. Ofthose invited, 90.3% participated (n �76,885).

HUNT 2Between 1995 and 1997 a second healthsurvey (n � 92,703) was conducted inNord-Trøndelag (HUNT 2), again includ-ing all inhabitants aged �20 years. Theoverall response rate in this follow-up in-vestigation was 71.3% (n � 65,258) (13).The clinical investigation includedheight, weight, blood pressure, waist andhip circumference, HDL cholesterol, cho-lesterol, triglycerides, and glucose. In ad-dition to the questions used in HUNT 1,this questionnaire included more detailedinformation on FHD.

Study populationThe analyses in this article were based oncross-sectional data from 64,833 men andwomen who participated in HUNT 2 forwhom complete information on FHD,age, and sex was available. In addition,prospective data from 41,548 subjectswho participated in both investigationsand were free from diabetes at the base-line investigation were included.

FHDDetailed information on FHD was avail-able from the HUNT 2 questionnaire, in-cluding separate questions on diabetes inmother, father, brothers and sisters, andchildren together with age at onset foreach relative. In addition, information ondiabetes in siblings was available from thebaseline questionnaire.

BMI and smokingBased on measures of height and weighttaken at the clinical investigations inHUNT 2, we calculated BMI as weight inkilograms divided by the square of heightin meters. Information on current andprevious smoking was used to classifysubjects as never, former, and currentsmokers.

Identification of diabetes casesThe HUNT 2 questionnaire identified1,951 cases of diabetes. These subjectswere given an appointment to have theirfasting blood glucose measured togetherwith levels of C-peptide and anti-GAD.Information on treatment was also col-lected. Altogether 1,454 (74.5%) patientscompleted this second investigation.

Patients starting insulin treatmentwithin 6 months of diagnosis were classi-fied as having type 1 diabetes, if, in addi-tion, they were anti-GAD� or had fastingC-peptide levels �150 pmol/l (n � 123).Patients were classified as having LADA ifthey were anti-GAD� and had not beentreated with insulin within 12 months ofdiagnosis (n � 128). Type 2 diabetic sub-jects were anti-GAD� and had not re-ceived insulin treatment within 1 year ofdiagnosis (n � 1,134). Of the 1,454 cases,845 were incident cases of diabetes, i.e.,subjects diagnosed during the follow-upperiod between HUNT 1 and HUNT 2.Among these were 80 cases of LADA, 744cases of type 2 diabetes, and 21 cases oftype 1 diabetes.

Biochemical analysesAnti-GAD and fasting C-peptide were an-alyzed at the Hormone Laboratory of AkerUniversity Hospital, Oslo, Norway. Anti-

GAD was analyzed by an immunoprecipi-tation radioligand assay based on apreviously validated method (14), andthe results are expressed as antibody in-dex. The latter was calculated as (countsin the patient sample � counts in a neg-ative reference serum)/(counts in a refer-ence antibody-containing serum �counts in a negative reference serum).The assay was tested for proficiency in acurrent diabetes autoantibody standard-ization program. At the cutoff level of�0.08, sensitivity was 0.64 and specific-ity was 1.00. Analysis of C-peptide wasdone by a radioimmunoassay (DiagnosticSystem Laboratories, Webster, TX).

Statistical analysesData for characteristics of the participantsare expressed as means � SD. C-peptidewas not normally distributed and there-fore is expressed as median and interquar-tile range. P values were calculated withStudent’s t test (means), Kruskall-Wallistest (medians), and with a �2 test (propor-tions). Analyses of FHD and the preva-lence of LADA, type 1 diabetes, and type 2diabetes were performed on the basis ofcross-sectional data from HUNT 2. In ad-dition, we investigated the influence ofhaving siblings with diabetes on the cu-mulative incidence of diabetes, i.e., therisk of developing diabetes during the 11-year follow-up period between HUNT 1and HUNT 2. To assess the associationbetween FHD and prevalence and inci-dence of LADA, type 1 diabetes, and type2 diabetes, we calculated odds ratios(ORs) together with 95% CIs using mul-tiple logistic regression analysis (Proc Lo-gistic, SAS/STAT; SAS Institute, Cary,NC). Confounding was adjusted for by

Table 1—Characteristics of participants in HUNT 2, 1995–1997

No known diabetes LADA Type 2 diabetes Type 1 diabetes

n 63,113 128 1,134 123Men (%) 46.8 53.1 49.3 59.4FHD (%) 14.6* 43.0 44.6 30.9†Age (years) 49.4 � 17.1* 68.2 � 11.8 68.1 � 11.1 48.7 � 16.2†BMI (kg/m2) 26.3 � 4.05* 28.5 � 4.7 29.6 � 4.8‡ 26.1 � 3.9†Waist-to-hip ratio 0.84 � 0.08* 0.89 � 0.07 0.90 � 0.08 0.85 � 0.07†Systolic blood pressure (mmHg) 131.5 � 19.3* 147.7 � 23.3 148.3 � 22.2 134.1 � 19.1†Diastolic blood pressure (mmHg) 82.6 � 11.0* 89.8 � 12.3 90.5 � 10.9) 83.6 � 9.8†HDL cholesterol (mmol/l) 1.38 � 0.39* 1.24 � 0.47 1.19 � 0.38 1.57 � 0.45†Cholesterol (mmol/l) 5.89 � 1.26 5.87 � 1.28 6.25 � 1.28‡ 5.55 � 1.16†Triglycerides (mmol/l) 1.74 � 1.11* 2.29 � 1.44 2.60 � 1.55‡ 1.34 � 0.71†

Data are means � SD, unless indicated otherwise. *P �0.05 for difference between subjects with LADA and subjects without known diabetes. †P � 0.05 fordifference between subjects with LADA and subjects with type 1 diabetes. ‡P � 0.05 for difference between subjects with LADA and subjects with type 2 diabetes.

Carlsson, Midthjell, and Grill


inclusion of age and sex in the regressionmodel. Additional adjustment for BMIand smoking did not change the ORs(change �10%), and, therefore, thesevariables were not included in the finalmodel.

The regional ethical committee forMedical Research and the NorwegianData Inspectorate approved these studies.All participants gave informed consent.

RESULTS — Subjects with LADA andtype 2 diabetes were similar in most of thecharacteristics recorded (Table 1). Theywere on average 20 years older; hadhigher BMI, waist-to-hip ratio, and bloodpressure; and had less favorable levels ofblood lipids than subjects with type 1 di-abetes and those without diabetes. Morethan 40% of the subjects with LADA andtype 2 diabetes reported FHD comparedwith 31% of type 1 diabetic subjects and15% of subjects without diabetes.

There were no clear differences be-tween patients with LADA with and with-out FHD with regard to age atinvestigation, age at onset, diabetes dura-tion, BMI, blood pressure, or lipids (Table2). However, patients with LADA whohad FHD had lower titers of anti-GADthan those without FHD. In addition,they seemed to have lower levels of C-peptide and more often were beingtreated with insulin. Subjects with type 2diabetes and FHD were marginallyyounger and leaner, were more oftentreated with insulin, and had lower C-peptide levels than those without FHD.With regard to type 1 diabetes, subjectswith FHD seemed to be younger at onsetand have higher BMI and anti-GAD thanthose without FHD.

Subjects with a family member withdiabetes were almost four times as likelyto have LADA (Table 3) compared withsubjects without FHD. Similar resultswere seen for type 2 and type 1 diabetes.There was no indication of sex differencesin the influence of FHD on the occurrenceof LADA or type 2 diabetes. However, fortype 1 diabetes, men with diabetes in thefamily had an OR of 3.75 (95% CI 2.29–6.14), whereas the corresponding esti-mate in women was 1.81 (0.96–3.42).

Having any family member with dia-betes was associated with increased prev-alence of LADA and type 2 diabetes (Table3). In contrast, type 1 diabetes was muchmore common in subjects with diabetesin siblings than in those with parents withdiabetes. The occurrence of LADA wastwice as high in subjects with male rela- T

able

2—C

hara

cter

isti

csof

subj

ects

wit

hL

AD

A,t

ype

2di

abet

es,a

ndty

pe1

diab

etes

byF

HD

:HU

NT

2,19

95–1

997

LAD

AT

ype

2di

abet

esT

ype

1di

abet

es

No

FHD

FHD

Pva

lue

No

FHD

FHD

Pva

lue

No

FHD

FHD

Pva

lue

n(%

)73

(57.

0)55

(43.

0)62

8(5

5.4)

506

(44.

6)85

(69.

1)38

(30.

9)M

en(%

)56

.249

.10.

4273

51.6

46.4

0.08

4756

.565

.80.

3309

Age

(yea

rs)

68.2

�12

.368

.1�

11.1

0.43

9769

.0�

11.3

67.1

�10

.50.

0185

47.1

�12

.352

.4�

16.8

0.58

78BM

I(k

g/m

2)

28.0

�4.

429

.1�

5.0

0.31

4029

.7�

4.8

29.4

�4.

80.

8777

25.5

�3.

627

.4�

4.2

0.27

25W

aist

-to-

hip

rati

o0.

89�

0.07

0.90

�0.

080.

3896

0.90

�0.

080.

89�

0.08

0.47

790.

83�

0.06

0.87

�0.

080.

0527

Dia

stol

icbl

ood

pres

sure

(mm

Hg)

90.5

�12

.389

.0�

12.3

0.98

4691

.2�

11.0

89.5

�10

.60.

3522

84.4

�10

.382

.0�

8.7

0.30

72Sy

stol

icbl

ood

pres

sure

(mm

Hg)

149.

1�

24.2

145.

8�

22.1

0.49

1715

0.0

�23

.014

6.4

�20

.90.

0290

134.

7�

19.8

132.

8�

17.8

0.53

75H

DL

chol

este

rol(

mm

ol/l)

1.25

�0.

471.

24�

0.48

0.92

981.

18�

0.39

1.21

�0.

350.

0197

1.59

�0.

471.

52�

0.42

0.40

08C

hole

ster

ol(m

mol

/l)5.

77�

1.22

6.00

�1.

360.

3854

6.26

�1.

326.

24�

1.23

0.11

415.

53�

1.14

5.61

�1.

210.

6734

Tri

glyc

erid

es(m

mol

/l)2.

17�

1.32

2.45

�1.

590.

1417

2.70

�1.

622.

46�

1.44

0.00

461.

28�

0.70

1.49

�0.

720.

7840

Age

aton

set

(yea

rs57

.4�

14.3

57.4

�13

.40.

6243

59.5

�14

.858

.3�

12.8

0.00

1629

.5�

15.6

26.1

�18

.60.

2031

Mea

ndu

rati

on(y

ears

)10

.6�

12.5

10.5

�10

.30.

1812

9.0

�12

.18.

6�

10.2

0.00

0217

.4�

12.3

24.7

�14

.70.

1852

Insu

lintr

eatm

ent

(%)

31.5

47.3

0.06

9319

.323

.50.

0815

100

100

C-p

epti

de(p

mol

)58

8.0

(798

.0)

440.

0(6

06.0

)0.

1110

783.

50(6

62.0

)72

8.0

(556

.0)

0.02

58—

*—

*A

nti-

GA

D(a

ntib

ody

inde

x)0.

43�

0.60

0.26

�0.

34�

0.00

010.

01�

0.02

0.01

�0.

020.

4655

0.29

�0.

450.

44�

0.60

0.03

13

Dat

aar

em

eans

�SD

orm

edia

n(i

nter

quar

tile

rang

e)(f

orC

-pep

tide

),un

less

othe

rwis

ein

dica

ted.

*Sub

ject

sw

ith

type

1di

abet

esdi

dno

tha

vem

easu

rabl

eC

-pep

tide

leve

ls.

Family history and LADA


tives having diabetes than for those withfemale relatives having diabetes (OR 1.98[95% CI 1.07–3.66]). For type 2 and type1 diabetes, there seemed to be no system-atic differences between having female ormale relatives with diabetes.

The prospective data (Table 4)showed that the risk of developing LADAand type 2 diabetes during the 11-yearfollow-up was 2.5 and 2.1 times in-creased, respectively, in subjects who atbaseline reported having a sibling with di-abetes. The risk of type 1 diabetes wasmore than seven times increased in thosewith siblings with diabetes.

CONCLUSIONS — We found byanalysis of cross-sectional data that LADAwas four times more common in subjectswith FHD. In addition, prospective datashowed that subjects who had siblingswith diabetes were twice as likely to de-velop LADA during the 11-year follow-upcompared with those without FHD. To-gether, these findings demonstrate thatFHD is a risk factor for LADA of the samemagnitude as for type 2 diabetes.

With regard to type 2 diabetes, ourstudy confirms previous findings indicat-

ing a four times increased prevalence insubjects with FHD (6–8). For type 1 di-abetes, the association with FHD wasweak compared with previous data (2).One reason may be that the majority ofour type 1 diabetic subjects (66%) hadonset at age �20. The genetic back-ground may be stronger in subjects withearly-onset type 1 diabetes (15). Accord-ingly, we found that 40% of subjects withonset of type 1 diabetes before the age of20 had FHD compared with 25% of thosewith onset during adulthood.

Previous reports have shown that therisk of both type 1 and type 2 diabetesincreases with number of affected rela-tives (6,7,9). Our study extends thesefindings to LADA by showing a six timesincreased prevalence in subjects withmore than one relative with diabetes.

Subjects with LADA and type 2 dia-betes with FHD had lower levels of C-peptide and were more often treated withinsulin than those without FHD, despitehaving the same duration of diabetes.This finding suggests that FHD influencesthe risk of LADA by way of reduced insu-lin secretion and that this situation isshared with subjects with type 2 diabetes.

Notably, the differences in C-peptide andinsulin treatment were substantial butonly significant for type 2 diabetes andnot for LADA: Still our findings are con-sistent with previous reports of impairedinsulin secretion in the offspring of pa-tients with LADA (16).

Previous information on the role ofFHD in the etiology of LADA is sparse. Ina British study, 33% of patients withLADA had close relatives with diabetes(5). This figure was somewhat lower thanthat in our study (43%), which may beexplained by the fact that Castleden et al.(5) excluded subjects whose relatives hadtype 1 diabetes. As far as we know, FHDin subjects with type 2 diabetes and LADAhave been compared in only two previousstudies, and the results of these studieswere not consistent; Castleden et al. (5)reported a higher proportion of type 2 di-abetic patients with FHD than of patientswith LADA, whereas the opposite wasfound in an Icelandic study (17). Unfor-tunately, these studies did not includetype 1 diabetes. Our findings suggest thatthe role of FHD is equally strong forLADA as for type 2 diabetes but strongerthan for type 1 diabetes, because only

Table 3—FHD and ORs of prevalent LADA, type 2 diabetes, and type 1 diabetes: HUNT 2, 1995–1997

FHD

Subjects notreportingdiabetes

LADA Type 2 diabetes Type 1 diabetes

Cases OR (95% CI) Cases OR (95% CI) Cases OR (95% CI)

No 53,926 73 1.0 628 1.0 85 1.0Yes 9,187 55 3.92 (2.76–5.58) 506 4.20 (3.72–4.75) 38 2.78 (1.89–4.10)One family member with diabetes 8,100 42 3.51 (2.40–5.14) 361 3.51 (3.07–4.0) 28 2.32 (1.51–3.57)Two or more family members with diabetes 1,087 13 6.29 (3.46–11.44) 145 8.33 (6.84–10.15) 10 6.60 (3.38–12.88)Mother with diabetes 3,965 18 3.34 (1.99–5.62) 238 5.17 (4.42–6.06) 10 1.67 (0.86–3.23)Father with diabetes 2,678 15 5.66 (3.21–9.99) 96 4.29 (3.42–5.38) 7 1.68 (0.78–3.65)Parents with diabetes 6,370 29 4.07 (2.62–6.31) 284 4.62 (3.99–5.36) 13 1.36 (0.75–2.43)Sister with diabetes 1,264 12 3.59 (1.93–6.69) 117 4.01 (3.24–4.95) 13 7.79 (4.20–14.44)Brother with diabetes 1,420 17 5.13 (3.00–8.75) 131 4.76 (3.89–5.81) 15 7.52 (4.25–13.31)Siblings with diabetes 1,827 17 3.54 (2.07–6.05) 121 2.92 (2.37–3.58) 21 8.51 (5.14–14.09)Children with diabetes 458 4 4.06 (1.47–11.21) 21 2.40 (1.53–3.78) 3 4.74 (1.48–15.22)Mother or sister with diabetes 4,554 19 2.65 (1.60–4.40) 257 4.16 (3.57–4.84) 19 2.83 (1.71–4.68)Father or brother with diabetes 3,472 23 5.15 (3.21–8.27) 137 3.58 (2.95–4.34) 16 3.02 (1.77–5.17)

ORs were adjusted for age and sex of the participants.

Table 4—Baseline information on diabetes in siblings (HUNT 1, 1984–1986) and ORs of incident LADA, type 2 diabetes, and type 1 diabetesduring 11 years of follow-up (HUNT 2, 1995–1997)

Siblings withdiabetes

Subjects notreportingdiabetes

LADA Type 2 diabetes Type 1 diabetes

Cases OR (95% CI) Cases OR (95% CI) Cases OR (95% CI

No 38,759 65 1.0 617 1.0 17 1.0Yes 1,944 15 2.51 (1.39–4.51) 127 2.21 (1.80–2.71) 4 7.24 (2.22–23.64)

ORs were adjusted for age and sex of the participants.



31% of type 1 diabetic patients had FHDcompared with 43% of patients withLADA.

Subjects with male relatives with dia-betes were twice as likely to have LADA asthose with female relatives with diabetes.This observation corresponds to previousfindings in type 1 diabetes (10). Severalexplanations behind this phenomenonhave been proposed, including a higherrate of miscarriage in women with type 1diabetes (18). Notably, no difference be-tween maternal or paternal diabetes onthe risk of type 1 diabetes was seen in thepresent study. This result may be ex-plained by the fact that we, for the mostpart, investigated patients with onset dur-ing adulthood for which the associationwith parental diabetes was weak.

For type 1 diabetes, having a siblingwith diabetes carried a much greater riskthan having parents with diabetes. Thisobservation is in accordance with resultsfrom the Diabetes Prevention Trial 1 (19).It could mean that the family environ-ment shared by siblings is particularly im-portant for evolution of type 1 diabetes.However, it should be noted that our type1 diabetic subjects were on average al-most 20 years younger than our subjectswith LADA and type 2 diabetes. Hence,the parents may not have developed dia-betes yet.

We found that patients with LADAwho had FHD had lower levels of anti-GAD than subjects without FHD. This ob-servation confirms findings of Castledenet al. (5) who reported that FHD was lesscommon in patients with LADA who hadanti-GAD levels in the highest tertile. Atentative explanation could be that lessautoimmune activity is required to causeLADA in individuals with genetic suscep-tibility to diabetes (i.e., genetic suscepti-bility of a kind that is unrelated toautoimmunity). In this context we notethat patients with LADA with high anti-GAD (highest 50%) were on average 6years younger at onset of diabetes thanthose with low anti-GAD (results notshown), indicating that age-related insu-lin resistance was strongest in those withlow anti-GAD.

Recent studies have shown that type 1and type 2 diabetes often occur in thesame families, indicating in part a com-mon genetic background (20). In thepresent study, we did not have informa-tion on the type of diabetes in relatives.We did find that none of the parents ofour LADA patients had onset of diabetesbefore the age of 40 even though this does

not exclude autoimmune diabetes. Wealso found that subjects with LADA andtype 2 diabetes were similar in phenotypeand in associations with FHD. These ob-servations indicate at least in part a com-mon genetic background for LADA andtype 2 diabetes.

The main results of this study may beaffected by recall bias as they were basedon cross-sectional data, which may havecaused an overestimation of the associa-tion between FHD and diabetes. Ourfindings were, however, supported byprospective data indicating increased in-cidence of diabetes in subjects reportingsiblings with diabetes. The 64% sensitiv-ity of our anti-GAD assay means thatsome cases of LADA could be classified ascases of type 2 diabetes. Still, even thoughthis misclassification would reduce thepower of our LADA analyses, it would notbias the relative risk estimates for LADAas long as the underdiagnosis was notrelated to FHD. Finally, it should bementioned that FHD reflects not solelygenetic influences but also a combina-tion of shared genetic and environmen-tal effects.

In summary, the results of this studydemonstrate that FHD is a strong risk fac-tor for LADA and indicates a genetic back-ground that may have more in commonwith type 2 diabetes than with type 1 di-abetes. Further, the influence of familyhistory may be mediated through a heri-table reduction of insulin secretion.

Acknowledgments— HUNT is a collabora-tion between the HUNT Research Centre,Norwegian University of Science and Technol-ogy; the Norwegian Institute of Public Health,and the Nord-Trøndelag County Council. Gl-axoSmithKline Norway and the NorwegianDiabetes Association supported HUNT. Thisparticular study was supported by a grant fromthe Swedish Council for Working Life and So-cial Research.

References1. Fourlanos S, Dotta F, Greenbaum CJ,

Palmer JP, Rolandsson O, Colman PG,Harrison LC: Latent autoimmune diabe-tes in adults (LADA) should be less latent.Diabetologia 48:2206–2212, 2005

2. Field LL: Genetic linkage and associationstudies of type I diabetes: challenges andrewards. Diabetologia 45:21–35, 2002

3. Turner R, Stratton I, Horton V, Manley S,Zimmet P, Mackay IR, Shattock M, Bot-tazzo GF, Holman R: UKPDS 25: autoan-tibodies to islet-cell cytoplasm andglutamic acid decarboxylase for predic-

tion of insulin requirement in type 2 dia-betes: UK Prospective Diabetes StudyGroup. Lancet 350:1288–1293, 1997

4. Tuomi T, Carlsson A, Li H, Isomaa B, Mi-ettinen A, Nilsson A, Nissen M, Ehrn-strom BO, Forsen B, Snickars B, Lahti K,Forsblom C, Saloranta C, Taskinen MR,Groop LC: Clinical and genetic character-istics of type 2 diabetes with and withoutGAD antibodies. Diabetes 48:150–157,1999

5. Castleden HA, Shields B, Bingley PJ, Wil-liams AJ, Sampson M, Walker M, GibsonJM, McCarthy MI, Hitman GA, Levy JC,Hattersley AT, Vaidya B, Pearson ER:GAD antibodies in probands and their rel-atives in a cohort clinically selected fortype 2 diabetes. Diabet Med 23:834–838,2006

6. Grill V, Persson P-G, Carlsson S, Alvars-son M, Norman A, Svanstrom L, EfendicS, the Stockholm Diabetes PreventionProgram Group: Family history of dia-betes in middle-age Swedish men is agender unrelated factor which associ-ates with insulinopenia in newly diag-nosed diabetes subjects. Diabetologia42:15–23, 1999

7. Meigs JB, Cupples LA, Wilson PW: Paren-tal transmission of type 2 diabetes: theFramingham Offspring Study. Diabetes49:2201–2207, 2000

8. Hariri S, Yoon PW, Qureshi N, Valdez R,Scheuner MT, Khoury MJ: Family historyof type 2 diabetes: a population-basedscreening tool for prevention? Genet Med8:102–108, 2006

9. Bonifacio E, Hummel M, Walter M,Schmid S, Ziegler AG: IDDM1 and multi-ple family history of type 1 diabetes com-bine to identify neonates at high risk fortype 1 diabetes. Diabetes Care 27:2695–700, 2004

10. Gale EA, Gillespie KM: Diabetes and gen-der. Diabetologia 44:3–15, 2001

11. Kim DJ, Cho NH, Noh JH, Lee MS, LeeMK, Kim KW: Lack of excess maternaltransmission of type 2 diabetes in a Ko-rean population. Diabetes Res Clin Pract65:117–124, 2004

12. Midthjell K, Bjørndal A, Holmen J, KrugerØ, Bjartveit K: Prevalence of known andpreviously unknown diabetes mellitusand glucose intolerance in an adult Nor-wegian population: indications of an in-creasing diabetes prevalence: the Nord-Trøndelag Diabetes Study. Scand J PrimHealth Care 13:229–235, 1995

13. Midthjell K, Kruger O, Holmen J, TverdalA, Claudi T, Bjorndal A, Magnus P: Rapidchanges in the prevalence of obesity andknown diabetes in an adult Norwegianpopulation. Diabetes Care 22:1813–1820,1999

14. Petersen JS, Hejnaes KR, Moody A,Karlsen AE, Marshall MO, Høier-MadsenM, Boel E, Michelsen BK, Dyrberg T. De-tection of GAD65 antibodies in diabetes

Family history and LADA


and other autoimmune diseases using asimple radioligand assay. Diabetes 43:459–467, 1994

15. Graham J, Hagopian WA, Kockum I, LiLS, Sanjeevi CB, Lowe RM, Schaefer JB,Zarghami M, Day HL, Landin-Olsson M,Palmer JP, Janer-Villanueva M, Hood L,Sundkvist G, Lernmark A, Breslow N,Dahlquist G, Blohme G, the Diabetes In-cidence in Sweden Study Group, theSwedish Childhood Diabetes StudyGroup: Genetic effects on age-dependentonset and islet cell autoantibody markers

in type 1 diabetes. Diabetes 51:1346–1355, 2002

16. Vauhkonen I, Niskanen L, Knip M, IlonenJ, Vanninen E, Kainulainen S, UusitupaM, Laakso M: Impaired insulin secretionin non-diabetic offspring of probandswith latent autoimmune diabetes mellitusin adults. Diabetologia 43:69–78, 2000

17. Jonsdotter AM, Aspelund T, SigurdssonG, Gudnason V, Benediktsson R: Latentautoimmune diabetes in adults in Iceland:prevalence, phenotype and relatedness.Laeknabladid 91:909–914, 2005

18. Harjutsalo V, Reunanen A, Tuomilehto J:Differential transmission of type 1 diabe-tes from diabetic fathers and mothers totheir offspring. Diabetes 55:1517–1524,2006

19. Yu L, Cuthbertson DD, Eisenbarth GS,Krischer JP: Diabetes Prevention Trial 1:prevalence of GAD and ICA512 (IA-2) au-toantibodies by relationship to proband.Ann NY Acad Sci 958:254–258, 2002

20. Tuomi T: Type 1 and type 2 diabetes:what do they have in common? Diabetes54 (Suppl. 2):S40–S45, 2005



Oral Health Knowledge, Attitude, andPractices and Sources of Information forDiabetic Patients in Lahore, PakistanKAMRAN MASOOD MIRZA, BDS

1

AYYAZ ALI KHAN, PHD2

MUNAWAR MANZOOR ALI, BDS2

SAIMA CHAUDHRY, BDS1

Sustained hyperglycemia affects al-most all tissues in the body (1), in-cluding those in the oral cavity (2).

Oral complications of diabetes includexerostomia, opportunistic infections,greater accumulation of plaque, delayedwound healing, susceptibility to periodon-tal disease, oral paresthesia, and alteredtaste (2). Studies suggest a bidirectional ad-verse relationship between diabetes andperiodontal disease; diabetes can aggravateperiodontitis, and periodontitis can neg-atively affect control of diabetes (3,4).Therefore, preventive behaviors likebrushing, flossing, and periodic dentalvisits, which have a positive correlationwith better periodontal health (5), be-come paramount for diabetic patients (6).Oral hygiene behavior and seeking oralhealth care depend on a number of fac-tors. Patients comply better with oralhealth care regimens when informed andpositively reinforced. Lack of informationis among the reasons for nonadherence tooral hygiene practices. Further, oralhealth attitudes and beliefs are significantfor oral health behavior (7). A higher like-lihood of seeking preventive dental care isfound to be associated with dental knowl-edge (8). The motives prompting peopleto seek preventive dental care include thebeliefs that one is susceptible to dentaldisease, that dental problems are serious,and that dental treatment is beneficial.Those who believe that they are highlysusceptible to dental disease make morepreventive dental visits (9). Health educa-tion attempts to change behaviors by al-tering an individual’s knowledge,attitudes, and beliefs about health matters

(9). The present study aimed to gatherbaseline information on knowledge, atti-tude, and practices of diabetic patients re-garding their oral health with the view ofenhancing dental health education forthis population, which would upgradetheir knowledge and understanding. Thisis believed to improve the oral health sta-tus of the diabetic patients, in turn con-trolling diabetes and, ultimately, qualityof life.

RESEARCH DESIGN ANDMETHODS — This study was a cross-sectional descriptive survey of 240 diabeticpatients visiting the Diabetic Clinic ofShaikh Zayed Medical Complex, Lahore,Pakistan. Inclusion criteria for samplingwere the fulfillment of all three of the fol-lowing conditions: that the patient 1) be suf-fering from type 1 or type 2 diabetes, 2)have at least one natural tooth, and 3) bediagnosed with diabetes for at least 6months. Any diabetic medical personnel orpatients with apparent physical or mentalhandicap were excluded. Patients of all age-groups were included in the sample.

A questionnaire was designed to as-sess the knowledge, attitude, and prac-tices of diabetic patients along withcorresponding demographic variables(Table 1). The questionnaire was pilotedin 30 patients to determine its validity.The study was approved by the ethicalcommittee of Shaikh Zayed Medical Com-plex. Informed verbal consent was takenfrom each eligible participant before ad-ministration of the questionnaire. Willingparticipants were informed in detail bythe investigators about the research

project and its consequences. The inves-tigators asked the questions verbally inUrdu and filled out the form. Privacy ofthe patients was ensured during filling ofquestionnaires. At the end of questioning,patients were informed about the impactof their systemic condition on oral health.

RESULTS — The mean � SD (range)age of the sample was 49 � 11.05 years(17–80). The male-to-female ratio was1:1.4. The results show that 35.4% of thepatients had knowledge about the oralcomplications of diabetes. Only 17.7% ofthis group knew about this issue fromtheir treating physicians. Fifty-seven per-cent did not know that diabetes predis-posed them to oral disease, and 7.6%denied any existence of a link betweendiabetes and oral health. Sources ofknowledge included treating physician,self-experience, diabetic patients’ familymembers and friends, dentists, and, veryrarely, printed media.

According to 28% of respondents,self-remedy was the solution to dentalproblems. Forty-five percent of subjectsalso said that if told of their predispositionto oral disease, they would increase theirbrushing frequency; 31.5% said that thisinformation would not affect their rou-tine, while 23% said that they would con-sult a dentist. Two percent of theparticipants brushed their teeth threetimes a day, and 22% brushed twice daily.

Knowledge regarding oral complica-tions of diabetes that was imparted by phy-sicians was significantly related to brushingfrequency (P � 0.005); 53.4% of counseledpatients brushed two or three times daily,while only 22.3% of uncounseled patientsbrushed two or three times per day.

CONCLUSIONS — The pr ima ryfinding of this study is a lack of knowl-edge about the relationship of diabeteswith oral complications. The results areconsistent with studies conducted world-wide (10–12). However, most diabeticpatients knew about various medicalcomplications of diabetes like nephropa-thy, retinopathy, and diabetic foot be-cause their physicians had laid emphasison these topics. This may indicate lack oforal health counseling on the part of phy-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Shaikh Zayed Federal Post Graduate Medical Institute, Lahore, Pakistan; and the 2Department ofDentistry, Shaikh Zayed Medical Complex, Lahore, Pakistan.

Address correspondence and reprint requests to Ayyaz Ali Khan, Dentistry, Shaikh Zayed Medical Com-plex, Lahore, Pakistan. E-mail: [email protected].

Received for publication 12 March 2007 and accepted in revised form 14 August 2007.Published ahead of print at http://care.diabetesjournals.org on 21 August 2007. DOI: 10.2337/dc07-0502.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion



E p i d e m i o l o g y / H e a l t h S e r v i c e s R e s e a r c hB R I E F R E P O R T


sicians, as evidenced by other studies(13–15). On the other hand, patients feltthat they would be more careful aboutoral hygiene if they were informed. Over-all oral hygiene measures in diabetic pa-tients were found to be deficient.

We found an association betweencounseling by physicians and positivepractices toward oral health by patients.Diabetic patients who claimed to knowabout the oral complications of diabetesthrough sources other than their physi-cian showed no significant difference intheir brushing habits compared withthose who never knew about the systemiceffects of diabetes (P � 0.225).

Diabetic patients who smoke need tobe informed that smoking adversely af-fects their periodontium 10-fold morethan that of normal individuals (16). Thiscalls for a targeted effort in motivating di-abetic patients against smoking by healthcare providers. Further studies are recom-

mended on a larger scale to confirm theassociation indicated in the present study.

Acknowledgments— This study was sup-ported by Pakistan Medical Research CouncilGrant 4-22-15/05/RDC/SZPGMI.

Findings of this study were presented inpart at the 28th Asia Pacific Dental Congress,Karachi, Pakistan, 23–27 February 2006.

References1. Mealey B: Diabetes mellitus. In Burket’s

Oral Medicine Diagnosis & Treatment. 10thed. Glick M, Greenberg M, Eds. Hamil-ton, BC Decker, 2003, p. 563–577

2. Matthews DC: The relationship betweendiabetes and periodontal disease. J CanDent Assoc 68:161–164, 2002

3. Loe P: Periodontal disease: the sixth com-plication of diabetes mellitus. DiabetesCare 16:329–334, 1993

4. Grossi SG, Geneo RJ: Periodontal diseaseand diabetes mellitus: a two way relation-

ship. Ann Periodontol 3:51–61, 19985. Lang WP, Ronis DL, Farghaly MM: Pre-

ventive behaviors as correlates of peri-odontal health status. J Public Health Dent55:10–17, 1995

6. Maung MT: Dental care for diabetics [ar-ticle online], 2001. Pretoria, South Africa,South African Diabetes Association.Available from http://home.intekom.com/buildlink/ips/sada/dental.htm. Accessed16 July 2006.

7. Kneckt M: Psychological Features Charac-terizing Oral Health Behavior, Diabetes Self-Care and Health Status Among IDDMPatients [article online]. Dissertation.Oulu, Finland, Institute of Dentistry, Uni-versity Of Oulu, 2000, p. 16–18. Availablefrom http://herkules.oulu.fi/isbn9514256301/isbn9514256301.pdf. Accessed 18July 2006

8. Tash RH, O’Shea MM, Cohen K: Testing apreventive-symptomatic theory of dentalhealth behaviour. Am J Public Health 59:514–521, 1969

9. Kegeles SS: Some motives for seeking pre-ventive dental care. J Am Dent Assoc 67:110–118, 1963

10. Sandberg GE, Sundberg HE, Wikblad KF:A controlled study of oral self-care andself-perceived oral health in type 2 dia-betic patients. Acta Odontol Scand 59:28–33, 2001

11. Taiwo JO: Oral health education needs ofdiabetic patients in Ibadan. Afr J Med MedSci 29:269–274, 2000

12. Kamel NM, Badawy YA, el-Zeiny NA,Merdan IA: Sociodemographic determi-nants of management behaviour of dia-betic patients. II. Diabetics’ knowledge ofthe disease and their management behav-iour. East Mediterr Health J 5:974–983,1999

13. Morgan R, Tsang J, Harrington N, Fook L:Survey of hospital doctors’ attitudes andknowledge of oral conditions in older pa-tients. Postgrad Med J 77:392–394, 2001

14. Institute for Healthcare Improvement:Better oral health for mothers and chil-dren [article online], 2006. Availablefrom http://www.ihi.org/IHI/Topics/ChronicConditions/AllConditions/ImprovementStories/FSBetterOral-HealthforMothersandChildren.htm.Accessed 13 January 2007.

15. Mouradian WE, Reeves A, Kim S, LewisC, Keerbs A, Slayton RL, Gupta D, Osk-ouian R, Schaad D, Kalet T, Marshall SG:A new oral health elective for medical stu-dents at the University of Washington.Teach Learn Med 18:336–342, 2006

16. Moore PA, Weyant RJ, Mongelluzzo MB,Myers DE, Rossie K, Guggenheimer J,Block HM, Huber H, Orchard T: Type 1diabetes mellitus and oral health: assess-ment of periodontal disease. J Periodontol70:409–417, 1999

Table 1—Questions regarding oral health knowledge, attitude, and practices of the sample

Did your physician tell you about the oral problems related to diabetes?Yes 6.3No 77.9Don’t know 15.8

Is a diabetic more prone to oral diseases?Yes 35.4No 7.6Don’t know 57

Do you have any dental/oral problems?Yes 66.6No 31.3Don’t know 2.1

Is it because of diabetes?*Yes 53.1No 28.8Don’t know 18.1

If there is an oral problem, what should be done?Consult a physician 20Consult a dentist 47Self-remedy 28Ignore it 5

If someone tells you that you are more prone to oral diseases, what would you do?Increase brushing frequency 45Decrease brushing frequency 0.5Same as normal routine 31.5Consult a dentist 23

Do you smoke?Yes 17.1No 70Occasionally (�1 daily) 12.9

Is smoking more injurious to the gums of a diabetic than those of a nondiabetic?Yes 38No 14.3Don’t know 47.7

Data are percent. *Asked only of those who answered “yes” to the question, “Do you have any dental/oralproblems?” (66%).

Oral health knowledge and practices in Pakistan


Influence of Flickering Light on the RetinalVessels in Diabetic PatientsALEKSANDRA MANDECKA, MD

1

JENS DAWCZYNSKI, MD2

MARCUS BLUM, MD3

NICOLLE MULLER, DNUTR1

CHRISTOPH KLOOS, MD1

GUNTER WOLF, MD1

WALTHARD VILSER, MS4

HEIKE HOYER, MS5

ULRICH ALFONS MULLER, MD, MSC1

OBJECTIVE — Stimulation of the retina with flickering light increases retinal vessel diame-ters in humans. Nitric oxide is a mediator of the retinal vasodilation to flicker. The reduction ofvasodilation is considered an endothelial dysfunction. We investigated the response of retinalvessels to flickering light in diabetic patients in different stages of diabetic retinopathy.

RESEARCHDESIGNANDMETHODS — We studied 53 healthy volunteers, 68 type 1diabetic patients, and 172 type 2 diabetic patients. The diameter of retinal vessels was measuredcontinuously online with the Dynamic Vessel Analyzer (DVA). Diabetic retinopathy was classi-fied using Early Treatment Diabetic Retinopathy Study criteria. Changes in vasodilation areexpressed as percent change over baseline values.

RESULTS — After adjustments for age, sex, and antihypertensive treatment, the response ofretinal arterioles to diffuse luminance flicker was significantly diminished in patients with type1 diabetes compared with healthy volunteers. The vasodilation of retinal arterioles and venulesdecreased continuously with increasing stages of diabetic retinopathy. The retinal arterial diam-eter change was 3.6 � 2.1% in the control group, 2.6 � 2.5% in the no diabetic retinopathygroup, 2.0 � 2.7% in the mild nonproliferative diabetic retinopathy (NPDR) group, 1.6 � 2.2%in the moderate NPDR group, 1.8 � 1.9% in severe NPDR group, and 0.8 � 1.6% in proliferativediabetic retinopathy group.

CONCLUSIONS — Flicker responses of retinal vessels are abnormally reduced in diabeticpatients. This decreased response deteriorated with increasing stages of retinopathy. The re-sponse was already reduced before clinical appearance of retinopathy. The noninvasive testing ofretinal autoregulation with DVA might prove to be of value in early detection of diabetic vesselpathological changes.

Diabetes Care 30:3048–3052, 2007

H yperglycemia, dyslipidemia, hyper-tension, and diabetes duration arethe main risk factors for the devel-

opment and progression of diabetic reti-nopathy (1–3). However, the exactpathogenesis of this disease remains in-completely understood. There is evidencethat endothelial dysfunction may play an

important role in the pathogenesis of di-abetic retinopathy (4). Markers of endo-thelial dysfunction such as solubleintercellular adhesion molecule-1 andsoluble vascular cell adhesion molecule-1are elevated in patients with diabetic ret-inopathy. However, the association be-tween markers of endothelial dysfunction

and diabetic retinopathy has not alwaysbeen consistent, presumably because ofthe considerable biological variation inthe measurement of such markers (5).Therefore, the use of other parameters isdesirable to assess the regulation ability ofretinal vessels.

Endothelial cells regulate vascular re-activity by responding to mechanicalforces and neurohumoral mediators withthe release of a variety of relaxing andcontracting factors (6). One of the mostimportant endothelium-derived vasodila-tors is nitric oxide (NO), the bioavailabil-ity of which is decreased in diabetes (7).In addition, NO appears to be a mediatorof the retinal vasodilator response toflicker light (8). Several human studiesshowed an increase in retinal vessel diam-eter during stimulation with diffuse lumi-nance flicker (8–10).

In the current study, we investigatedthe response of retinal arterial and venousvessels to flickering light in patients withdiabetes. All retinal vessels are by defini-tion vessels of microcirculation. For ourpurpose, a recently developed provoca-tion test was used. Diffuse luminanceflicker was applied, and retinal vessel di-ameters were measured with a DynamicVessel Analyzer (DVA) (Imedos, Jena,Germany). We determined the endotheli-um-derived vasodilation of retinal arteriesin different stages of diabetic retinopathy.

RESEARCH DESIGN ANDMETHODS — The study was per-formed on a group of healthy volunteersand type 1 and 2 diabetic patients whowere being treated in a large outpatientdiabetes clinic at a tertiary university hos-pital. A total of 53 healthy subjects, 68patients with type 1 diabetes, and 172 pa-tients with type 2 diabetes were investi-gated. All subjects were of Caucasianorigin. Of the diabetic patients, 83.3%were treated with insulin, 31.6% with oralantidiabetic agents, and 80.4% with anti-hypertensive drugs. All examinationswere performed after the patients had re-ceived oral and written information aboutthe study and had given their consent toparticipate. The examinations were per-formed in accordance with the Declara-tion of Helsinki and were approved by thelocal ethics committee.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Internal Medicine III, Friedrich-Schiller University, Jena, Germany; the 2Depart-ment of Ophthalmology, Friedrich-Schiller University, Jena, Germany; the 3Department of Ophthalmology,Helios Klinikum, Erfurt, Germany; 4IMEDOS, Jena, Germany; and the 5Institute of Medical Statistics,Computer Sciences and Documentation, Friedrich-Schiller-University, Jena, Germany.

Address correspondence and reprint requests to Aleksandra Mandecka, Department of Internal MedicineIII, Friedrich-Schiller University, Bachstrasse 18, 07743, Jena, Germany. E-mail: [email protected].


0927.Abbreviations: DVA, Dynamic Vessel Analyzer; ETDRS, Early Treatment Diabetic Retinopathy Study;

NPDR, nonproliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion



P a t h o p h y s i o l o g y / C o m p l i c a t i o n sO R I G I N A L A R T I C L E


In all subjects, the left eye was stud-ied. Volunteers were taking no medica-tion at the time of the study. The healthyparticipants were nonsmokers, had noprevious history of arterial hypertensionor metabolic or cardiovascular disease,and did not receive any medication onprescription. All subjects had no historyof epilepsy or ocular disease other thandiabetic retinopathy and carotid arteryobstruction and were nonsmokers. Everypatient had undergone measurement ofintraocular pressure within 1 year beforeenrollment. Patients with increased in-traocular pressure were excluded. Allsubjects were asked to refrain from use ofalcohol, nicotine, and caffeine for at least1 h before the examination.

The clinical data of the patients exam-ined are shown in Table 1. Diabetic pa-tients were significantly older and hadhigher mean arterial blood pressure com-pared with the healthy control group.

Study protocolAt the start of the study, fundus examina-tion was performed after induction of my-driasis with 1% tropicamide eye drops.Diabetic retinopathy was classified usingEarly Treatment Diabetic RetinopathyStudy (ETDRS) criteria (11) as no diabeticretinopathy (ETDRS level 10), mild non-proliferative diabetic retinopathy (NPDR)(ETDRS level �20), moderate NPDR (ET-DRS level �43), severe NPDR (ETDRSlevel �53), and proliferative diabetic ret-inopathy (PDR) (ETDRS level �61).

The DVA was used for digital fundusimaging for conventional fundus exami-nations and for retinal vessel analysis. Ret-

inal vessel analysis with the DVA allowsnoninvasive diagnosis of microvascularfunction by measuring the diameter of ar-terial and venous retinal vessels continu-ously 25 times/s and by using stimulationtests of vessel functions. By interruptingof the green measuring light, the DVAgenerates flicker light with a frequency of12.5 Hz and with a bright-to-dark ratio of25:1 for a stimulation test. Diameter re-sponses can be recorded by use of flickerlight periods during the vessel diametermeasurements. The dilatation of vessel di-ameter caused by flickering light can beused as a function diagnostic parameterfor the endothelium-derived vasodila-tion. Details of the DVA and the processesof diameter measurements and flickerstimulations are described elsewhere(12–15).

After the baseline vessel diameter wasmeasured for 50 s, provocation withflicker light was performed for 20 s, andthe response was observed for 80 s afterthe end of the flicker exposure. The cycle

was then repeated two times. An arterialsegment of �1.5 mm was evaluated ineach eye. Selection criteria for the seg-ment were location within a circular areaof two disc diameters, no crossing or bi-furcation in the measuring segment, cur-vature of not �30°, a distance toneighboring vessels of at least one vesseldiameter, and sufficient contrast to thesurrounding fundus. The position of thevessel edges, the vessel course, the vesseldiameter, and correction for ocular move-ments were calculated automaticallyonline.

Blood pressure measurementsThe mean systemic arterial blood pres-sures of the groups are listed in Table 1.No significant increase in blood pressureoccurred during the examination. Themean arterial blood pressure (mean RR)was calculated as mean RR � RR diastole� 1⁄3 � (RR systole � RR diastole) mmHg,where RR systole is systolic blood pres-sure and RR diastole is diastolic bloodpressure.

Statistical analysesChanges in ocular hemodynamic param-eters are expressed as the percent changeover baseline values. Retinal vessel diam-eters were calculated as an average of thelast 30 s of the baseline of each cycle. Ves-sel diameter during flicker was calculatedas an average of the last 3 s of light stim-ulation and the following 3 s after stimu-lation. All variables obtained weredescribed by adequate statistical mea-sures. Differences between groups werestatistically evaluated by t test, the Mann-Whitney U test, or �2 test as appropriate.To adjust for imbalances of age, sex, andantihypertensive treatment, ANCOVA wasapplied. Contrasts were defined to testdifferences between type 1 or type 2 dia-betic patients compared with the controlgroup and to analyze the linear trend of

Table 1—Characteristics of the study groups

ParameterControlgroup

Type 1diabetes

Type 2diabetes

n 53 68 172Sex

Male 14 (26) 30 (44)* 97 (56)*Female 39 (74) 38 (56) 75 (44)

Age (years) 41.9 � 16.6 47.5 � 15.3* 61.7 � 10.1*Mean arterial pressure

(mmHg)87.8 � 8.5 92.9 � 8.2* 100.7 � 9.9*

A1C (%) — 7.9 � 1.2 7.4 � 1Duration of diabetes

(years)— 17.7 � 10.3 11.4 � 7.8

Antihypertensivetreatment

0 (0) 33 (49)* 156 (92)*

Missing data 2 1 2

Data are n (%) or means � SD. A1C normal range: 4.6–5.9. *Significantly different compared with thecontrol group.

Table 2—Mean diameter change of retinal arteries and veins to flicker in healthy subjects anddiabetic patients

ParameterControlgroup

Type 1diabetes

Type 2diabetes

Arterial vasodilation (%) 3.6 � 2.0 2.1 � 2.3 2.2 � 2.5P* �0.001 �0.001

Arterial vasoconstriction (%) �1.4 � 1.7 �1.0 � 1.7 �0.6 � 1.4P 0.135 0.001

Venous diameter change (%) 4.5 � 2.4 4.0 � 2.3 3.5 � 2.1P 0.212 0.005

Data are means � SD. *P values from unadjusted comparison with control group.

Mandecka and Associates


vasodilatation depending on the severityof diabetic retinopathy. P � 0.05 wasconsidered to be statistically significant.Statistical analysis was performed withSPSS (version 13.0.1; SPSS, Chicago, IL).

RESULTS — In retinal arterioles, theresponse to stimulation with luminanceflicker was diminished in diabetic pa-tients compared with healthy volunteers(Table 2). In healthy control subjects,flicker stimulation increased the retinalarterial diameter by 3.6 � 2.0%, in type 1diabetic patients by 2.1 � 2.3%, and intype 2 diabetic patients by 2.2 � 2.5%.The response was significantly decreasedregardless of type of diabetes. The con-striction of the retinal arteries, as well asthe response of retinal venous diameters,was also diminished in diabetic patientscompared with control subjects, but dif-ferences were significant only in type 2diabetic patients compared with controlsubjects.

Association of retinal vessel flickerresponse with age and duration ofthe diseaseAge and duration of diabetes were signif-icantly associated with arterial diameterresponse in diabetic subjects. The vasodi-lation of the arteries decreased signifi-cantly with increasing age and duration ofdisease. With increasing age, there was atendency toward smaller dispersion of thedilation.

The age versus arterial diameterchange scatterplot shows a decreasingflicker response and increasing dispersionof the measured values in subjects of mid-dle to advanced age. The small coefficientof correlation (r � 0.22) indicates a weakcorrelation between the two parameters(data not shown).

To account for confounding by age,by antihypertensive treatment, and prob-ably by sex, vasodilatation was further an-alyzed by ANCOVA (Table 3). Afteradjustment, the difference between dia-betic patients and the control group re-mained significant for the arterialdiameter change in type 1 diabetic pa-tients. The difference in venous diameterchange in type 2 diabetic patients com-pared with control subjects was morepronounced after adjustment but failed toreach statistical significance at the globalsignificance level.

Association of retinal vessel flickerresponse with mean arterial bloodpressure and A1CThere was no significant association be-tween arterial retinal flicker response andmean arterial blood pressure or A1C indiabetic patients (multiple regressionanalyses). The flicker response of retinalarteries in diabetic patients deterioratedbut not significantly with increasing A1C(data not shown).

Retinal vessel flicker response indifferent stages of diabeticretinopathyThe retinal arterial diameter changes were3.6 � 2.1, 2.6 � 2.5, 2.0 � 2.7, 1.6 �2.2, 1.8 � 1.9, and 0.8 � 1.6% in thecontrol group (n � 53), no diabetic reti-nopathy group (n � 145), mild NPDRgroup (n � 36), moderate NPDR group(n � 27), severe NPDR group (n � 18),and PDR group (n � 14), respectively

Figure 1—Arterial diameter changes at stages of diabetic retinopathy (DR).

Table 3—Age-, antihypertensive treatment–, and sex-adjusted mean differences of diameterchange comparing type 1 and type 2 diabetic patients with the control group of healthy subjects

GroupAdjusted

difference (%)* 95% CIP value

(global test)†

Arterial vasodilatationControl Reference 0.024Type 1 diabetes �1.1 �2.0 to �0.2Type 2 diabetes �0.3 �1.4 to 0.8

Arterial vasoconstrictionControl Reference 0.823Type 1 diabetes 0.2 �0.4 to 0.9Type 2 diabetes 0.2 �0.5 to 0.9

Venous diameter changeControl Reference 0.063Type 1 diabetes �0.7 �1.6 to 0.2Type 2 diabetes 1.2 �2.2 to �0.2

Groups: type 1 diabetic (n � 68), type 2 diabetic (n � 170), and control (n � 53). *Difference(type 1�control)

or difference(type2�control). †P values from covariance analyses; deviation from symmetry of CIs because ofrounding).

Vasodilation and diabetic retinopathy


(Fig. 1). There was a significant trend ofdecreasing retinal arterial response alongthe groups (age-, antihypertensive treat-ment–, and sex-adjusted trend test P �0.002). The venous diameter changeswere 4.6 � 2.4, 3.9 � 2.3, 3.7 � 2.2,3.5 � 2.1, 2.7 � 2.2, and 3.1 � 2.0% inthe control group, no diabetic retinopa-thy group, mild NPDR group, moderateNPDR group, severe NPDR group, andPDR group, respectively (Fig. 2). The ad-justed retinal venous response was alsosignificantly decreased (trend test P �0.007). No significant trend could be ob-served for constriction of the retinalarteries.

CONCLUSIONS — In this in vivostudy, we compared endothelial functionunder physiological flow conditions andin the presence of the diabetic milieu.Noninvasive testing of the function of au-toregulation of retinal arterioles is possi-ble with the DVA (13–16). There is muchevidence for abnormal autoregulation ofretinal vessels in diabetic patients. Usingthe laser Doppler technique, Grunwald etal. (17) reported reduced retinal arterialand venous blood velocity as well as en-larged retinal veins in patients with dia-betes with background retinopathy.Moreover, retinal blood flow is reduced inpatients with diabetes with no diabeticretinopathy compared with patients with-

out diabetes (18,19). There is also evi-dence that the early stages of diabeticretinopathy are associated with increasedretinal blood flow and retinal vasodila-tion, abnormal retinal vascular responseto hyperoxia, and abnormal retinal auto-regulation (20–23). The intrinsic abnor-mality in diabetic retinopathy appears tobe endothelial cell dysfunction (24–26).The present study focuses on the retinaldiameter changes of major temporal reti-nal vessels of diabetic patients to diffuseluminance flicker. In humans, the flickerlight–induced vasodilation is mediated byNO (27). Hence, this test could be used asan estimate of the capacity of endothelialcells of retinal vessels to release NO inresponse to a physiological stimulus indisease states.

We demonstrated that the retinal ves-sel flicker response is diminished in dia-betic patients compared with that innormal control participants. This findingis in agreement with a previous report in-dicating reduced flicker response in type1 diabetic patients (28). We have also re-ported abnormal autoregulation in pa-tients with type 2 diabetes.

The present study demonstrates thatthe vasodilation of retinal arteries andveins under the flickering light decreasescontinuously with increasing stages of di-abetic retinopathy. Furthermore, auto-regulation was also found to be abnormal

in diabetic patients without retinopathyand deteriorated continuously in patientswith retinopathy, suggesting that the dis-turbance is involved in the disease patho-genesis. The venous retinal response wasreduced in diabetic patients without anyvisible signs of diabetic retinopathy com-pared with the control group. This find-ing is in agreement with several previousreports indicating impairment of bloodflow regulation in the retina before theclinical appearance of retinopathy (21,29). In our study we showed an associa-tion between flicker response and age;however, the coefficient of correlationwas weak, which is in agreement withprevious reports (30). For example,Jeppesen et al. (31) reported significantlyreduced diameter response in normal in-dividuals aged �40 years.

Most of the diabetic patients were re-ceiving antihypertensive treatment at thetime of testing. To rule out the possibleconfounding effect of drugs, we adjustedthe data for imbalances of antihyperten-sive medication. The adjusted response ofretinal vessels to flickering light decreasedsignificantly with increasing stages of di-abetic retinopathy. This finding suggeststhat diabetes has a deteriorating effect perse on the flow regulation in response toflicker stimulation.

In summary, this study demonstrateda decreased retinal vessel flicker responsein patients with diabetes. This decreasedresponse deteriorated with increasingstages of diabetic retinopathy. Indeed, theresponse was already low before the clin-ical appearance of retinopathy. The pre-dictive value of this method for detectingdiabetic patients at risk for the develop-ment of diabetic retinopathy needs tobe tested with long-term observationalstudies.

References1. Van Leiden HA, Dekker JM, Moll AC, Ni-

jpels G, Heine RJ, Bouter LM, StehouwerCD, Polak BC: Blood pressure, lipids andobesity are associated with retinopathy:the Hoorn study. Diabetes Care 25:1320–1325, 2002

2. Nagi DK, Pettitt DJ, Bennett PH, Klein R,Knowler WC: Diabetic retinopathy as-sessed by fundus photography in Pima In-dians with impaired glucose toleranceand NIDDM. Diabetes Med 14:449–456,1997

3. Stratton IM, Kohner EM, Aldington SJ,Turner RC, Holman RR, Manley LE, Mat-thews DR: UKPDS 50: risk factors for in-cidence and progression of retinopathy in

Figure 2—Venous diameter changes at stages of diabetic retinopathy (DR).

Mandecka and Associates


type II diabetes over 6 years from diagno-sis. Diabetologia 44:156–163, 2001

4. Stehouwer CD, Lambert J, Donker AJ, vanHinsbergh VW: Endothelial dysfunctionand pathogenesis of diabetic angiopathy.Cardiovasc Res 34:55–68, 1997

5. van Hecke MV, Dekker JM, Nijpels G,Moll AC, Heine RJ, Bouter LM, Polak BCP,Stehouwer CDA: Inflammation and endo-thelial dysfunction are associated withretinopathy: the Hoorn study. Diabetolo-gia 48:1300–1306, 2005

6. Furchgott RF, Vanhoutte PM: Endotheli-um-derived relaxing and contracting fac-tors. FASEB J 3:2007–2018, 1989

7. Hink U, Li H, Mollnau H, Oelze M,Matheis E, Hartmann M, Skatchkov M,Thaiss F, Stahl RA, Warnholtz A, MeinertzT, Griendling K, Harrison DG, Forster-mann U, Munzel T: Mechanisms underly-ing endothelial dysfunction in diabetesmellitus. Circ Res 88:E14–E22, 2001

8. Dorner GT, Garhofer G, Kiss B, Polska E,Polak K, Riva CE, Schmetterer L: Nitricoxide regulates retinal vascular tone inhumans. Am J Physiol 285:H631–H636,2003

9. Polak K, Schmetterer L, Riva CE: Influ-ence of flicker frequency on flicker in-duced changes of retinal vessel diameters.Invest Ophthalmol Vis Sci 43:2721–2726,2002

10. Dorner GT, Garhofer G, Huemer KH,Riva CE, Woltz M, Schmetterer L: Hyper-glycemia affects flicker-induced vasodila-tion in the retina of healthy subjects. VisRes 43:1495–500, 2003

11. The effect of intensive diabetes treatmenton the progression of diabetic retinopathyin insulin-dependent diabetes mellitus:the Diabetes Control and ComplicationsTrial. Arch Ophthalmol 113:36–51, 1995

12. Vilser W, Nagel E, Fuhrmann G, Riemer T:Retinale Gefaessanalyse-Neue Moeglich-keiten zur Untersuchungen von Netzhaut-

gefaessen. In Fortbildung Glaukom Band 3.Schmidt KG, Pillunat LE, Eds. Stuttgart,Enke Verlag, 2000, p. 73–91

13. Polak K, Dorner G, Kiss B, Polska E, FindlO, Rainer G, Eichler HG, Schmetterer L:Evaluation of the Zeiss retinal vessel anal-yser. Br J Ophthalmol 84:1285–1290,2000

14. Blum M, Bachmann K, Wintzer D, RiemerT, Vilser W, Strobel J: Noninvasive mea-surement of the Bayliss effect in retinalautoregulation. Graefes Arch Clin ExpOphthalmol 237:296–300, 1999

15. Blum M, Bachmann K, Pietscher S,Braeuer-Burchhardt C, Vilser W, StrobelJ: [Online measurement of retinal arterybranches in type II diabetic patients. Ini-tial clinical trials before and after laser co-agulation]. Ophthalmologe 94:724–727,1997 (article in German)

16. Vilser W, Nagel E, Lanzl I: Retinal vesselanalysis—new possibilities. Biomed Tech(Berl) 47:682–685, 2002

17. Grunwald JE, Riva CE, Sinclair SH: LaserDoppler velocimetry study of retinal cir-culation in diabetes mellitus. Arch Oph-thalmol 104:991–996, 1986

18. Grunwald JE, Brucker AJ, Schwartz SS,Braunstein SN, Baker L, Petrig BL, RivaCE: Diabetic glycemic control and retinalblood flow. Diabetes 39:602–607, 1990

19. Bursell SE, Clermont AC, Kinsley BT, Si-monson DC, Aiello LM, Wolpert UA: Ret-inal blood flow changes in patients withinsulin-dependent diabetes mellitus andno diabetic retinopathy. Invest OphthalmolVis Sci 37:886–897, 1996

20. Falck A, Laatikainen L: Retinal vasodila-tion and hyperglycaemia in diabetic chil-dren and adolescents. Acta OphthalmolScand 73:119–124,1995

21. Grunwald JE, DuPont J, Riva CE: Retinalhaemodynamics in patients with early di-abetes mellitus. Br J Ophthalmol 80:327–331, 1996

22. Patel V, Rassam SM, Chen HC, KohnerEM: Oxygen reactivity in diabetes melli-tus: effect of hypertension and hypergly-caemia. Clin Sci 86:689–695, 1994

23. Rassam SM, Patel V, Kohner EM: The ef-fect of experimental hypertension on ret-inal vascular autoregulation in humans: amechanism for the progression of diabeticretinopathy. Exp Physiol 80:53–68, 1995

24. Colwell JA, Winocour PD, Lopes-VirellaM, Haluschka PV: New concepts aboutthe pathogenesis of atherosclerosis in di-abetes mellitus. Am J Med 75:67–80,1983

25. Kohner EM, Porta M: Vascular abnormal-ities in diabetes and their treatment. TransOphthalmol Soc UK 100:440–444, 1980

26. Almer LA, Pandolfi M: Fibrinolysis anddiabetic retinopathy. Diabetes 25 (Suppl.2):807–810, 1976

27. Delles C, Michelson G, Harazny J, Oeh-mer S, Hilgers KF, Schmieder RE: Im-paired endothelial function of the retinalvasculature in hypertensive patients.Stroke 35:1289–1293, 2004

28. Garhofer G, Zawinka C, Resch H, KothyP, Schmetterer L, Dorner GT: Reduced re-sponse of retinal vessel diameters toflicker stimulation in patients with diabe-tes. Br J Ophthalmol 88:887–891, 2004

29. Feke GT, Buzney SM, Oqasawara H, FujioN, Goger DG, Spack NP, Gabbay KH: Ret-inal circulatory abnormalities in type 1 di-abetes. Invest Ophthalmol Vis Sci 35:2968–2975, 1994

30. Nagel E, Vilser W, Lanzl I: Age, bloodpressure, and vessel diameter as factorsinfluencing the arterial retinal flicker re-sponse. Invest Ophthalmol Vis Sci 45:1486–1492, 2004

31. Jeppesen P, Gregersen PA, Bek T: The age-dependent decrease in the myogenic re-sponse of retinal arterioles as studied withthe Retinal Vessel Analyzer. Graefes ArchClin Exp Ophthalmol 242:914–919, 2004

Vasodilation and diabetic retinopathy


Atrophy of Foot Muscles in DiabeticPatients Can Be Detected WithUltrasonographyKAARE SEVERINSEN, MD

1

ANNETTE OBEL, MD2

JOHANNES JAKOBSEN, MD, PHD1

HENNING ANDERSEN, MD, PHD1

OBJECTIVE — To establish a bedside test with ultrasonography for evaluation of foot muscleatrophy in diabetic patients.

RESEARCH DESIGN AND METHODS — Thickness and cross-sectional area (CSA) ofthe extensor digitorum brevis muscle (EDB) and of the muscles of the first interstitium (MILs)were determined in 26 diabetic patients and in 26 matched control subjects using ultrasonog-raphy. To estimate the validity, findings were related to the total volume of all foot musclesdetermined at magnetic resonance imaging (MRI-FMvol). Furthermore, the relations ofultrasonographic estimates to nerve conduction, sensory perception thresholds, and clinicalcondition were established.

RESULTS — In diabetic patients, the ultrasonographic thickness of EDB (U-EDBt) was(means � SD) 6.4 � 2.1 vs. 9.0 � 1.0 mm in control subjects (P � 0.001), the thickness of MIL(U-MILt) was 29.6 � 8.3 vs. 40.2 � 3.6 mm in control subjects (P � 0.001), and the CSA of EDB(U-EDBCSA) was 116 � 65 vs. 214 � 38 mm2 in control subjects (P � 0.001). The MRI-FMvol

was directly related to U-EDBt (r � 0.77), U-MILt (r � 0.71), and U-EDBCSA (r � 0.74). U-EDBt

and U-MILt were thinner in neuropathic than in nonneuropathic diabetic patients (5.8 � 2.1 vs.7.5 � 1.7 mm [P � 0.05] and 28.3 � 8.8 vs. 35.6 � 4.3 mm [P � 0.03], respectively).

CONCLUSIONS — Atrophy of intrinsic foot muscles determined at ultrasonography is di-rectly related to foot muscle volume determined by MRI and to various measures of diabeticneuropathy. Ultrasonography seems to be useful for detection of foot muscle atrophy in diabetes.

Diabetes Care 30:3053–3057, 2007

Motor dysfunction is an establishedpart of diabetic polyneuropathy,resulting in distal atrophy and

weakness. At the clinical examination,foot deformities clearly indicate muscleatrophy, whereas detection of atrophy atearlier stages is difficult.

Atrophy of small foot muscles hasbeen reported using magnetic resonanceimaging (MRI) (1–3). Due to excellentsoft-tissue contrast, MRI enables detec-

tion of even subtle changes in size andstructure of foot muscles (1–3). In a pre-vious study (1), we found substantial at-rophy in neuropathic patients withoutany foot deformity, whereas muscle vol-ume was preserved in nonneuropathic di-abetic patients. Recently, a study using31P MRI at 3 Tesla observed minor loss ofmuscle tissues in nonneuropathic pa-tients (3).

MRI is the gold standard for visualiza-

tion of soft-tissue structures in the footdue to its high spatial resolution enablingidentification of the individual small footmuscles (2). However, MRI is time con-suming, cannot be performed bedside,and is more expensive than ultrasonogra-phy. Ultrasonography is an establishedmethod for examination of various mus-culoskeletal structures in children andadults with chronic neuromuscular dis-eases and traumatic muscle injuries (4–6). Also, animal experiments in acutemuscle denervation indicate consistencybetween MRI, electromyography, and ul-trasonography 1–64 days after denerva-tion (7).

In the present study, the size of indi-vidual foot muscles was examined withultrasonography in diabetic patients withand without neuropathy and in matchedcontrol subjects compared with the totalvolume of all foot muscles determined byMRI.

RESEARCH DESIGN ANDMETHODS — Twenty-six diabeticpatients (22 with type 1 and 4 with type 2diabetes) and 26 control subjectsmatched for age, sex, height, and weightwere included in the study. Demographicand baseline clinical data are shown inTable 1. Patients were recruited from theoutpatient diabetes clinic, and controlsubjects were recruited among hospitalstaffs. All patients were able to walk un-supported, and none had a history of footsurgery or had symptoms or signs of arte-rial insufficiency of the lower extremities.

Patients with severe cardiac or lungdisease, cancer, alcoholism, acute orchronic musculoskeletal disease, otherneurological disease, other endocrine dis-orders, or symptomatic peripheral arterydisease were excluded. All subjects gaveinformed consent to the study, which wasapproved by the local ethics committee.

All ultrasonographic examinationswere made by the same examiner (K.S.)using a scanner with a linear array real-time ultrasonic probe (Toshiba Powervi-sion 6000 duplex). The subjects wereplaced in a supine position with the non-dominant foot placed on a plastic ramp tokeep the ankle joint in a neutral position.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Neurology, Aarhus University Hospital, Aarhus, Denmark; and the 2Departmentof Neuroradiology, Aarhus University Hospital, Aarhus, Denmark.

Address correspondence and reprint requests to Henning Andersen, Department of Neurology, AarhusUniversity Hospital, 8000 Aarhus C, Denmark. E-mail: [email protected].

Received for publication 25 January 2007 and accepted in revised form 17 August 2007.Published ahead of print at http://care.diabetesjournals.org on 23 August 2007. DOI: 10.2337/dc07-


dc07-0108.Abbreviations: CSA, cross-sectional area; EDB, digitorum brevis muscle; MIL, muscle of the first inter-

stitium; MNCV, motor conduction velocity; MRI, magnetic resonance imaging; NIS, neurological impair-ment scale; SNCV, sensory nerve conduction velocity.






At ultrasonographic evaluation, theextensor digitorum brevis muscle (EDB)and the muscle group between the firstand second metatarsal bone (MIL), in-cluding the first dorsal interosseus mus-cle, the adductor hallucis muscle, and thefirst lumbrical muscle, could be unambig-iously identified. The EDB thickness (U-EDBt) and cross-sectional area (U-EDBCSA) was determined by scanningtransverse to the muscle fibers (Fig. 1),whereas the MIL (U-MILt) was scannedlongitudinal to the fibers (Fig. 1). The fre-quency of the ultrasonic beam was 15MHz for the EDB and 8 MHz for the MIL.In each case, the position of the ultra-sound probe was marked externally onthe skin using easily defined bone land-marks.

For evaluation of the EDB muscle, aline drawn perpendicular to the midpointof a straight line between the lateral mal-leolus and the tuberositas of the fifth

metatarsal bone defined the scanningplane. The exact position along this linefor maximum cross-sectional musclethickness differs between individuals andwas defined at each scanning procedure.For examination of the MIL, the distal po-sition of the ultrasound probe was de-fined by a line between the first andsecond metatarso-phalangeal joint andtwo lines marking the first and secondmetatarsal bone.

The ultrasonographic measurementswere performed with the ultrasound probeperpendicular to the muscle surface, gentlyplaced on the skin to avoid any pressure-induced alterations of muscle tissue dimen-sion using generous amounts of gel(ULTRA/PHONIC conductivity gel; Phar-maceutical Innovations, Newark, NJ). Dur-ing ultrasonographic scanning, thepatients were initially asked to perform avoluntary contraction of the muscles fa-cilitating the definition of the borders of

Figure 1—A: Ultrasonographic image of CSA of the EDB muscle (U-EDB-CSA) and thickness ofthe EDB muscle (U-EDBt). B: Ultrasonographic image of thickness of the first dorsal interosseusmuscle, the adductor hallucis muscle, and the first lumbrical muscle (U-MILt).

Tab

le1—

Cli

nica

land

elec

trop

hysi

olog

icfi

ndin

gsin

diab

etic

pati

ents

and

cont

rols

ubje

cts

nA

ge(y

ears

)W

eigh

t(k

g)H

eigh

t(c

m)

Mal

e/fe

mal

e

Typ

e1/

type

2di

abet

es

Dia

bete

sdu

rati

on(y

ears

)A

1C(%

)

Vib

rato

rype

rcep

tion

thre

shol

d(fi

rst

toe)

(JN

D)

NIS

Pero

neal

MN

CV

(m/s

)

Pero

neal

CM

AP

(mV

)

Dia

beti

cpa

tien

ts26

50(2

6–64

)73

(56–

104)

176

(158

–190

)16

/10

21/4

32(8

–49)

8.9

(6.4

–11.

3)22

(14–

25)

13(0

–40)

36(2

3–48

)3

(0–1

0)W

ith

diab

etic

neur

opat

hy17

47(2

6–64

)72

(60–

95)

175

(158

–190

)10

/715

/232

(8–4

6)9.

1(6

.4–1

1.3)

22(1

4–25

)20

(2–4

0)35

(23–

41)

2(0

–4)

Wit

hout

diab

etic

neur

opat

hy9

49(3

7–63

)72

(56–

104)

175

(167

–187

)6/

37/

231

(14–

49)

8.1

(6.5

–9.5

)19

(15–

21)

2(0

–17)

44(4

3–48

)6

(1–1

0)C

ontr

olsu

bjec

ts26

49(2

5–67

)78

(54–

102)

176

(160

–185

)16

/10

——

——

——

—

Dat

aar

em

edia

n(r

ange

).C

MA

P,co

mpo

und

mus

cle

acti

onpo

tent

ial;

JND

,jus

tno

tice

able

diff

eren

ce.

Ultrasonography of atrophic diabetic foot muscles


the muscle. Then, the patient was askedto relax while the ultrasonographic imagewas recorded. For each parameter, fivemeasurements were made. The imageswere saved on a magneto-optical mediafor digital storage for later analysis on apersonal computer. Afterward, the lowestand highest values were excluded, and theaverage of the remaining values was usedfor further analysis. The average timespent for ultrasonography was 5 min forpreparation of the scanning session andidentification and demarcation of thebone landmarks followed by 10 –15min for making the 15 ultrasonographicmeasurements.

The nondominant foot of all patientsand control subjects was visualized byMRI using a 1.5-Tesla scanner (SigmaGE), adopting the principles for MRI es-timation of foot muscle size as describedin earlier studies (2). All magnetic reso-nance scans were obtained with a conven-tional T1-weighted Spin-Echo sequence(echo time � 20 ms, repetition time �540 ms) using cross-sectional magneticresonance images with a slice thickness of1.5 mm and an intersection interval of 10mm. A 256 � 256 matrix and two excita-tions were used. The images were storedon a personal computer and transferredfrom the 256 � 256 matrix to a 512 �512 bitmap color picture. The identity ofthe magnetic resonance images wasblinded to the observer. Within the mus-cle compartments, an upper level of signalintensity for muscle tissues was definedfor each patient by the examiner because afixed limit could not be applied due toautoscaling. Signal intensities above theupper level were defined as signal inten-sities of fat, allowing separation of muscletissues from “nonmuscle tissues” withinthe muscle compartments. Muscle fas-ciae, tendons, and blood vessels were ex-cluded. At each image, the cross-sectionalarea (CSA) of all muscles was estimated

using a stereological point-counting tech-nique described elsewhere (1,8,9). Thetotal volume of all foot muscles (MRI-FMvol) was calculated by multiplying thedistance (10 mm) between the sections bythe total CSA, the first section being ran-domly placed within the first interslice in-terval. It was not possible to study thesame muscles at MRI and ultrasonogra-phy because the largest CSA could not bedetermined beforehand at MRI. Further-more, it was difficult to obtain enoughslices to ensure a reliable estimate usingthe point-counting technique (9).

All patients were clinically evaluatedaccording to a neuropathy symptomscore (10) and a neurological impairmentscale (NIS) (11). Vibratory perceptionthresholds were evaluated at the dorsumpart of the dominant index finger and thenondominant great toe using the 4, 2, and1 stepping algorithm (12) (CASE IV; WRMedical Electronics, Stillwater, MN). Theperception thresholds for each patientwere compared with results from a largegroup of healthy control subjects (CASEIV) (P. J. Dyck, unpublished data).

Nerve conduction studies were per-formed using an electromyograph (Key-Point; Medtronic, Skovlunde, Denmark)and standard methods as described else-where (13,14). Motor conduction veloc-ity (MNCV) and amplitude of thecompound muscle action potential weremeasured of the nondominant peronealand tibial nerve. Sensory nerve conduc-tion velocity (SNCV) and sensory nerveamplitude of the sensory nerve action po-tential were measured of the nondomi-nant sural nerve, skin temperaturesranging between 31 and 34 centigrades. Zscores reflecting the degree of deviationfrom the expected mean were calculatedfor MNCV and SNCV using values ofhealthy volunteers obtained with similartechniques (13,14).

Patients were defined as neuropathic

in accordance with the minimal criteriafor diabetic neuropathy (15). For quanti-fication of severity of neuropathy, a neu-ropathy rank-sum score was calculatedfor each patient, including rank scores ofthe neuropathy symptom score, the NIS,the vibratory perception thresholds, andthe average of the rank scores of theMNCVs and SNCVs (1).

Statistical comparisons of muscle sizedetermined at ultrasonography and atMRI between groups were made with un-paired t tests, and correlations weresought for using linear regression analy-sis. Microsoft Excel was applied for thestatistical comparisons using a signifi-cance level of 0.05. Reproducibility anal-yses of the five ultrasononographicmeasurements were performed withANOVA using STATA.

RESULTS — According to the minimalcriteria for diabetic neuropathy, 17 pa-tients were neuropathic and 9 patientswere nonneuropathic. Among the 17neuropathic patients, 13 patients weresymptomatic. Patients with and withoutneuropathy had a median diabetes dura-tion of 32 years (range 8–46) and 31years (14–49), respectively.

For all diabetic patients, the NIS has amedian of 13 (range 0–40). The neuro-pathic and nonneuropathic patients hadan NIS of 20 (2–40) vs. 2 (0–17), respec-tively (P � 0.001). Furthermore, the neu-ropathic patients had significantly highervibratory perception threshold, lowerperoneal MNCV, and lower peronealcompound muscle action potential com-pared with the nonneuropathic patients(Table 1).

In patients, mean U-EDBCSA was(means � SD) 116 � 65 vs. 214 � 38mm2 in control subjects (P � 0.001), U-EDBt was 6.4 � 2.1 vs. 9.0 � 1.0 mm incontrol subjects (P � 0.001), and U-MILtwas 29.6 � 8.3 vs. 40.2 � 3.6 mm incontrol subjects (P � 0.001). For the neu-ropathic and nonneuropathic patients, U-EDBCSA, U-EDBt, and MRI-FMvo lexpressed as a percentage of individuallymatched control subjects were signifi-cantly reduced (Table 2). The reductionof U-MILt reached significance in the neu-ropathic patients only (Table 2). Compar-ing the neuropathic and nonneuropathicpatients, U-EDBCSA, U-MILt, and MRI-FMvol were significantly reduced in theneuropathic patients compared with thenonneuropathic diabetic patients (Table2).

Close correlations were found be-

Table 2—Muscle size (%) in diabetic patients with and without neuropathy relative to musclesize of individually matched control subjects

Diabetic patientswith neuropathy

Diabetic patientswithout neuropathy

P value (diabetic patientswith and without

neuropathy are compared)

U-EDBCSA (mm2) 50 � 33 79 � 27 0.05U-EDBt (mm) 66 � 26 85 � 21 NSU-MILt (mm) 71 � 22 90 � 8.2 0.01MRI-FMvol (mm3) 42 � 30 74 � 34 0.05

Data are means � SD. U-EDBCSA, CSA of EDB muscle; U-EDBt, thickness of EDB muscle; U-MILt, thicknessof MIL; MRI-FMvol, total volume of all foot muscles.

Severinsen and Associates


tween the distally placed U-MILt and theproximally placed U-EDBCSA and U-EDBtin the neuropathic group and in the non-neuropathic group as well (online appen-dix Table 3 [available at http://dx.doi.org/10.2337/dc07-0108]).

For diabetic patients and healthy con-trol subjects, close relationships could beestablished between MRI-FMvol on theone side and U-EDBCSA (r � 0.74 and r �0.64), U-EDBt (r � 0.77 and r � 0.76),and U-MILt (r � 0.71 and r � 0.58) onthe other side (Tables 3 and 4 and Fig. 2 ofthe online appendix). Reproducibilityanalysis of the five repeated ultrasono-graphic measurements using ANOVAshowed a coefficient of variation of U-EDBCSA, U-EDBt, and U-MILt, amountingto 0.031, 0.034, and 0.015, respectively,in control subjects and 0.046, 0.059, and0.026, respectively, in diabetic patients.

In all patients, close correlations werefound between the neuropathy rank-sumscore and the ultrasonographic measure-ments of U-EDBCSA (r � �0.76), U-EDBt(r � �0.73), and U-MILt (r � �0.71).Results of regression analyses performedfor the neuropathic and nonneuropathicsubgroups are shown in Fig. 2 and Table 3of the online appendix.

CONCLUSIONS — In this study, ul-trasonography could detect atrophy of in-dividual foot muscles in a group ofdiabetic patients. Muscular atrophy wasmore pronounced in diabetic patientswith clinical neuropathy. In addition, asignificant reduction of muscle size wasobserved in nonneuropathic diabetic pa-tients. Close relationships were found be-tween ultrasonographic estimates of footmuscle size and MRI-determined volumeof all foot muscles. Furthermore, ultra-sonographic measurements of foot mus-cle size were closely related to the clinicalseverity of neuropathy in diabetic patientswhen expressed with a neuropathy rank-sum score.

In previous studies (8,9), substantialmuscular atrophy has been found in dia-betic patients with neuropathy. There wasa proximal to distal gradient of atrophy atthe leg (8), and pronounced atrophy ofthe foot muscles in diabetic subjects withneuropathy have been found in recentstudies (1–3) using different MRI tech-niques. Brash et al. (16) observed in-creased fatty infiltration as well asindications of muscle fiber depletion ofthe intrinsic foot muscles at the first meta-tarsal joint in 19 patients suffering fromdiabetic neuropathy using MRI and mag-

netization transfer sequences. The quan-titative techniques in that study are notcomparable with those used in thepresent study because the magnetizationtransfer method provides an indirect esti-mate of muscle size. The method, how-ever, has the potential to discover evensubtle tissue changes in non-neuropathicdiabetic patients. Using standard MRItechniques and analysis of single-pixel re-laxation times, Bus et al. (2) found atro-phy of the intrinsic foot muscles at thelevel of the metatarsal heads in eight dia-betic patients with neuropathy, amount-ing to a 73% reduction of muscle CSA. Inour study, less pronounced atrophy wasobserved, the U-MIL and MRI-FMvolmuscle size amounting to 71 and 42% ofthe matched control subjects, respec-tively. In a previous study, we found pro-nounced atrophy of all foot muscles withreduction in the total muscle volume of15 patients suffering from diabetic neu-ropathy using traditional MRI techniquecombined with unbiased stereologicalmethods (1). Greenman et al. (3) used 31PRARE MRI and found significant footmuscle atrophy at the level of the meta-tarsal joint not only in neuropathic pa-tients but also in diabetic patients withoutclinical neuropathy. Their finding sug-gests that muscular atrophy may occurvery early in the neuropathic process evenbefore the minimal criteria of diabeticneuropathy are fulfilled. In accordancewith their observation, we found a statis-tically significant reduction in total footmuscle volume in non-neuropathic pa-tients. These observations suggest thateven subtle changes in nerve functionmay lead to muscle loss and that the ap-plied clinical criteria for neuropathy aretoo insensitive for detection of the earliestneuropathic changes.

In this study we have introduced ul-trasonography as a new method to esti-mate the size of foot muscles. We foundthat the ultrasonographic method hadhigh reproducibility with a coefficient ofvariation �0.06 for all muscles evaluated.Ultrasonography has been used in evalu-ation of muscular dystrophies (17–20)and other neuromuscular diseases inadults (21–23). Comparative studies ofultrasonography and MRI indicate thatthe lower spatial resolution in ultrasonog-raphy is in part compensated for by itsbedside availability and higher cost-effectiveness (20,24,25). In a study byKullmer et al. (7) in experimental dener-vated rabbits, MRI and sonography wereequally informative.

In our study, the foot muscles wereevaluated differently applying MRI andultrasonography. At ultrasonography thethickness and CSA of the whole muscle ofinterest was determined, whereas at MRIthe tissue with increased signal intensitywithin the muscles reflecting degenera-tion was excluded from the analysis of thevolume of the foot muscles. Therefore, theultrasonographic measurements mightunderestimate the degree of loss of mus-cle tissue.

In Fig. 2 of the online appendix, thereis a close correlation between MRI-FMvoland U-EDBCSA; however, this does notnecessarily imply that there is a high levelof agreement comparing these two meth-ods (26). In our study, we applied twodifferent visualization techniques on dif-ferent muscle structures including a sin-gle foot muscle and all foot muscles,respectively. Since different muscularstructures were evaluated, a direct analy-sis of limits of agreement was not per-formed.

A frequent objection to quantificationof structures using ultrasonography is theoperator dependency. However, in astudy by Bargfrede et al. (21) investigatingfocal neuropathies, an interobserver cor-relation of 0.85 of measurements of mus-cle size was obtained after scanning ofseveral muscles. In accordance with thisobservation, Maurits et al. (27) found aninterobserver correlation of 0.845 at de-termination of the thickness of the bicepsbrachii muscles. Furthermore, Maurits etal. found an intraobserver correlation co-efficient of 0.93, confirming the observa-tions made by Reimers et al. in idiopathicinflamatory myopathies (test-retest) (28).

In the present study, five measure-ments at each of the various scanningpositions were performed to obtain infor-mation about the variation of the esti-mate. In the clinical setting, however, oneor two standardized measurements aresufficient for an experienced examiner.The costs for ultrasonographic evaluationare low because the equipment is afford-able and needs one operator only.

In the present study, we evaluated thesize of small foot muscles at a proximaland at a distal position. The two musclesevaluated (EDB and MIL) are innervatedfrom branches of the peroneal and tibialnerves. The examination of the two mus-cles with different nerve supply ensuresthat atrophy of the EDB muscle was notdue to compression of the peroneal nerveat the fibular head. The close correlationsbetween the size of the two muscle groups

Ultrasonography of atrophic diabetic foot muscles


and the volume of all foot muscles suggestthat all measures reflect the same patho-physiological process.

During ultrasonographic scanning ofthe MIL muscles, segmentation of indi-vidual muscles was not feasible. However,strong correlations were established be-tween MRI-FMvol and the MIL musclegroup, indicating that ultrasonographicestimates reflect atrophy of individualmuscles. The ultrasonography techniquehas the advantage that it can be performedin almost all patients. The technique isnoninvasive and requires some experi-ence of the examiner to obtain robust andreproducible results. The present studysuggests that ultrasonography is a reliabletechnique for screening and monitoringof muscle atrophy in the diabetic foot. De-spite its lower resolution, this methodsupplies the examiner with sufficient an-atomical and functional information forevaluation of muscle size.

Acknowledgments— The Danish DiabetesAssociation (Diabetesforeningen) and theDanish Foundation of Neurological Research(Neurologisk Forskningsfond) are acknowl-edged for economic support.

References1. Andersen H, Gjerstad MD, Jakobsen J: At-

rophy of foot muscles: a measure of dia-betic neuropathy. Diabetes Care 27:2382–2385, 2004

2. Bus SA, Yang QX, Wang JH, Smith MB,Wunderlich R, Cavanagh PR: Intrinsicmuscle atrophy and toe deformity in thediabetic neuropathic foot: a magnetic res-onance imaging study. Diabetes Care 25:1444–1450, 2002

3. Greenman RL, Khaodhiar L, Lima C, DinhT, Giurini JM, Veves A: Foot small muscleatrophy is present before the detection ofclinical neuropathy. Diabetes Care 28:1425–1430, 2005

4. Peetrons P: Ultrasound of muscles. EurRadiol 12:35–43, 2002

5. Maurits NM, Beenakker EA, van SchaikDE, Fock JM, van der Hoeven JH: Muscleultrasound in children: normal values andapplication to neuromuscular disorders.Ultrasound Med Biol 30:1017–1027, 2004

6. Walker FO, Cartwright MS, Wiesler ER,

Caress J: Ultrasound of nerve and muscle.Clin Neurophysiol 115:495–507, 2004

7. Kullmer K, Sievers KW, Reimers CD,Rompe JD, Muller-Felber W, Nagele M,Harland U: Changes of sonographic, mag-netic resonance tomographic, electro-myographic, and histopathologic findingswithin a 2-month period of examinationsafter experimental muscle denervation.Arch Orthop Trauma Surg 117:228–234,1998

8. Andersen H, Gadeberg PC, Brock B, Ja-kobsen J: Muscular atrophy in diabeticneuropathy: a stereological magnetic res-onance imaging study. Diabetologia 40:1062–1069, 1997

9. Gadeberg P, Andersen H, Jakobsen J: Vol-ume of ankle dorsiflexors and plantarflexors determined with stereologicaltechniques. J Appl Phys 86:1670–1675,1999

10. Dyck PJ, Sherman WR, Hallcher LM, Ser-vice FJ, O’Brien PC, Grina LA, PalumboPJ, Swanson CJ: Human diabetic endo-neurial sorbitol, fructose, and myo-inosi-tol related to sural nerve morphometry.Ann Neurol 8:590–596, 1980

11. Dyck PJ: Quantitating severity of neurop-athy. In Peripheral Neuropathy. Dyck PJ,Thomas PK, Griffin JW, Low PA, PodulsioJF, Eds. Philadelphia, Saunders, 1993, p.1219–1250

12. Dyck PJ, O’Brien PC, Kosanke JL, GillenDA, Karnes JL: A 4, 2, and 1 steppingalgorithm for quick and accurate estima-tion of cutaneous sensation threshold.Neurology 43:1508–1512, 1993

13. Stålberg E, Falck B: Clinical motor nerveconduction studies. Meth Clin Neuro-physiol 4:61–80, 1993

14. Stålberg E, Falck B, Bischoff C: Sensorynerve conduction studies with surfaceelectrodes. Meth Clin Neurophysiol 5:1–20, 1994

15. Dyck PJ, Kratz KM, Karnes JL, Litchy WJ,Klein R, Pach JM, Wilson DM, O’BrienPC, Melton LJ III, Service FJ: The preva-lence by staged severity of various types ofdiabetic neuropathy, retinopathy, and ne-phropathy in a population-based cohort:the Rochester Diabetic NeuropathyStudy. Neurology 43:817–824, 1993

16. Brash PD, Foster J, Vennart W, AnthonyP, Tooke JE: Magnetic resonance imagingtechniques demonstrate soft tissue dam-age in the diabetic foot. Diabet Med 16:55–61, 1999

17. Dock W, Happak W, Grabenwoger F,Toifl K, Bittner R, Gruber H: Neuromus-

cular diseases: evaluation with high-fre-quency sonography. Radiology 177:825–828, 1990

18. Heckmatt JZ, Pier N, Dubowitz V: Real-time ultrasound imaging of muscles. Mus-cle Nerve 11:56–65, 1988

19. Lamminen A, Jaaskelainen J, Rapola J,Suramo I: High-frequency ultrasonogra-phy of skeletal muscle in children withneuromuscular disease. J Ultrasound Med7:505–509, 1988

20. Wallgren-Pettersson C, Kivisaari L,Jaaskelainen J, Lamminen A, Holmberg C:Ultrasonography, CT, and MRI of musclesin congenital nemaline myopathy. PediatrNeurol 6:20–28, 1990

21. Bargfrede M, Schwennicke A, Tumani H,Reimers CD: Quantitative ultrasonogra-phy in focal neuropathies as compared toclinical and EMG findings. Eur J Ultra-sound 10:21–29, 1999

22. Fischer AQ, Carpenter DW, Hartlage PL,Carroll JE, Stephens S: Muscle imaging inneuromuscular disease using computer-ized real-time sonography. Muscle Nerve11:270–275, 1988

23. Gunreben G, Bogdahn U: Real-timesonography of acute and chronic muscledenervation. Muscle Nerve 14:654–664,1991

24. Schedel H, Reimers CD, Nagele M, WittTN, Pongratz DE, Vogl T: Imaging tech-niques in myotonic dystrophy: a compar-ative study of ultrasound, computedtomography and magnetic resonance im-aging of skeletal muscles. Eur J Radiol 15:230–238, 1992

25. Juul-Kristensen B, Bojsen-Moller F, HolstE, Ekdahl C: Comparison of muscle sizesand moment arms of two rotator cuffmuscles measured by ultrasonographyand magnetic resonance imaging. Eur JUltrasound 11:161–173, 2000

26. Bland JM, Altman DG: Statistical-meth-ods for assessing agreement between 2methods of clinical measurement. Lancet1:307–310, 1986

27. Maurits NM, Bollen AE, Windhausen A,De Jager AE, van der Hoeven JH: Muscleultrasound analysis: normal values anddifferentiation between myopathies andneuropathies. Ultrasound Med Biol 29:215–225, 2003

28. Reimers CD, Fleckenstein JL, Witt TN,Muller-Felber W, Pongratz DE: Muscularultrasound in idiopathic inflammatorymyopathies of adults. J Neurol Sci 116:82–92, 1993

Severinsen and Associates


Neurovascular Factors in Wound Healing inthe Foot Skin of Type 2 Diabetic SubjectsSINGHAN T.M. KRISHNAN, MRCP

1

CRISTIAN QUATTRINI, MD2,3

MARIA JEZIORSKA, PHD3

RAYAZ A. MALIK, MRCP, PHD2

GERRY RAYMAN, FRCP, MD1

OBJECTIVE — Delayed wound healing in diabetic patients without large-vessel disease hasbeen attributed to microvascular dysfunction, neuropathy, and abnormal cellular and inflam-matory responses. The role of these abnormalities has mainly been examined in animal models.Few studies have been undertaken in diabetic patients, and those that have are limited due toanalysis in wounds from chronic ulcers. In this study, we quantified the rate of wound healingin relation to skin neurovascular function and structure following a dorsal foot skin biopsy intype 2 diabetes.

RESEARCH DESIGN AND METHODS — Twelve healthy control subjects and 12 type2 diabetic subjects with neuropathy but without macrovascular disease were studied. We quan-tified rate of wound healing and related it to skin microvascular function (laser Doppler imager[LDI]max), blood vessel density, small nerve fiber function (LDIflare) and nerve fiber density,vascular endothelial growth factor (VEGF) and its receptor (FLK1), and hypoxia-inducible factor(HIF)-1� expression.

RESULTS — The rate of wound closure was identical between control subjects and diabeticpatients despite a significant reduction in maximum hyperemia (LDImax), epidermal and dermalVEGF-A, and epidermal and dermal blood vessel VEGFR-2 expression as well as the neurogenicflare response (LDIflare) and dermal nerve fiber density. There was no significant difference inHIF-1� and dermal blood vessel density between control subjects and diabetic patients.

CONCLUSIONS — In conclusion, the results of this study suggest that wound closure insubjects with type 2 diabetes is not delayed despite significant alterations in neurovascularfunction and structure.

Diabetes Care 30:3058–3062, 2007

Wound healing is impaired in dia-betic patients and has been at-tributed to both macro- and

microvascular disease leading to tissuehypoxia, peripheral neuropathy, and ab-normal cellular and inflammatory path-ways predisposing to infection in footulcers (1– 4). The molecular basis forthese abnormalities has been examinedmainly in animal models, which have alimited translational capacity.

The loss of protective sensation due to

neuropathy and diminished trophic effectby neuropeptide deficiency have beenproposed to lead to trauma and increasedpressure on the foot skin and a dimin-ished hyperemic response to tissue injury,respectively (5). Furthermore, these alter-ations may lead acute wounds to advanceto chronic wounds with impaired healing(6). More recently, small fiber dysfunc-tion has been shown to be an early featurein patients with type 2 diabetes and hasalso been implicated in delayed wound

healing (7,8). Moreover, several micro-vascular abnormalities, including a re-duced response to tissue injury causingunderperfusion, the development of de-pendent edema due to a defective venoar-teriolar reflex, and increased permeabilityof capillaries, have also been proposed todelay wound healing (9,10). Most humanstudies have shown no reduction in skincapillary density, suggesting that micro-vascular function may be sufficiently ab-normal to reduce tissue blood flowwithout an actual reduction in overall vas-cular density in those with diabetes(11,12).

The molecular basis for these alter-ations has not been studied in detail inpatients with diabetes. Few studies onwound healing have been undertaken indiabetic patients, and those to date havebeen limited to chronic ulcers. In diabeticanimals, a reduction in IGF-I, IGF-II, ker-atinocyte growth factor, and platelet-derived growth factor (13) occurs, andapplication of these growth factors nor-malizes wound healing (14). Matrix met-alloprotienases are increased in chroniculcers in diabetic patients and in animalmodels of diabetes (15). Recently, the ex-pression of vascular endothelial growthfactor (VEGF), which promotes angio-genesis, has been shown to be reduced inthe skin wounds of diabetic animals, andtopical VEGF improved wound healing(16,17). Diabetic wounds in animal mod-els also show abnormal angiogenesis anda reduction in the expression of nervegrowth factor and its receptors. Nervegrowth factor, in addition to its neurotro-phic properties, has been shown to beproangiogenic, and nerve growth factorsupplementation improves vascular re-generation via VEGF-A to acceleratewound healing (18,19).

In this study, we quantified the rate ofwound healing in acute ulcers following apunch skin biopsy from the dorsum of thefoot in diabetic patients and control sub-jects. This was examined in relation toskin microvascular function (laser Dopp-ler imager [LDI]max), blood vessel density,and expression of VEGF and its receptor(VEGFR)-2, and hypoxia-inducible factor(HIF)-1�. C-fiber function (LDIflare) anddermal nerve fiber density were alsoquantified.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Diabetes Centre, Ipswich Hospital, Ipswich, U.K.; the 2Division of Cardiovascular Medicine,University of Manchester and Manchester Royal Infirmary, Manchester, U.K.; and the 3Division of Regen-erative Medicine, University of Manchester, Manchester, U.K.

Address correspondence and reprint requests to Dr. G. Rayman, MD, FRCP, The Ipswich Diabetes Centre,Ipswich Hospital, National Health Service Trust, Heath Road, Ipswich, IP4 5PD. E-mail:[email protected].

Received for publication 22 July 2007 and accepted in revised form 24 August 2007.Published ahead of print at http://care.diabetesjournals.org on 26 September 2007. DOI: 10.2337/dc07-

1421.Abbreviations: HIF, hypoxia-inducible factor; LDI, laser Doppler imager; VEGF, vascular endothelial

growth factor; VEGFR, receptor of VEGF.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





RESEARCH DESIGN ANDMETHODS — Twelve healthy controlsubjects (C group) and 12 subjects withtype 2 diabetes and neuropathy (D group)were studied. Subjects were recruited on aconsecutive basis from the diabetes out-patient clinics of the Ipswich DiabetesCentre. All subjects with diabetes selectedfor this study had peripheral neuropathy,as impaired wound healing is typically as-sociated with this complication. Subjectswith clinical features of peripheral vascu-lar disease (ankle brachial pressure index�0.8) were excluded. The study was ap-proved by the local ethical committee,and all subjects gave informed consent totake part in the study.

Assessment of neuropathyNeuropathy was assessed by measure-ment of the vibration perception thresh-old, using the ascending method of limits.A mean of three values was taken for anal-ysis. The results were expressed in volts. Avibration perception threshold of �15 V(i.e., �95th percentile) for this age-groupwas considered abnormal (20).

In addition, sensation was assessedusing the Neuropen (Owen Mumford,Oxford U.K.), which contains a 10-gmonofilament to assess pressure percep-tion and a Neurotip (Owen Mumford) forpinprick sensation. Ten-gram monofila-ments were applied for 2 s on the plantaraspect of the first, third, and fifth metatar-sal heads, and Neurotip was applied at theepinychium of the first toe (i.e., a total offour sites were tested, three for the 10-gmonofilament and one for Neurotip). Atsites where sensation was not felt, the testwas repeated three times to confirm theabnormality. Subjects were assigned tohave impaired sensation if they could notfeel a stimulus on more than one of thetested sites. All diabetic subjects recruitedhad absent ankle reflexes, impaired sen-sation using the Neuropen, and impairedvibration perception threshold.

Assessment of LDIflareSubjects were allowed to acclimatize for30 min in a temperature-controlled room,where the temperature was maintained at25 � 1°C. The foot temperature was mea-sured proximal to the first and secondmetatarsal heads using an infrared ther-mometer (Linear Laboratories, Fremont,CA). Room temperature and relative hu-midity were monitored throughout. Theaxon-reflex–mediated LDIflare was exam-ined using an LDI (Moor Instruments,Devon, U.K.) (8). This uses a stable he-

lium neon gas laser (� � 632.8 nm) beamthat is deflected by a moving mirror tocreate a raster pattern across the surface ofthe skin. The Doppler shifted light frommoving blood, and nonshifted light fromstatic tissue is directed back via the samemirror into two detectors. Fluctuations inthe wavelength are processed to calculatethe flux that is proportional to tissueblood flow. The data were recorded to acomputer using MoorLDI (version 3.11)software, and a flux image was producedusing a palette of 16 equally spaced colorsin which dark blue represented lowestperfusion and red the highest perfusion.

The skin proximal to the first and sec-ond metatarsal heads on the dorsum ofthe foot was heated with a circular skinheater (diameter �0.9 cm) (Moor Instru-ments) to 44°C for 20 min. An area of3.5 � 3.5 cm surrounding the heated skinwas scanned with the LDI aligned to beperpendicular to the dorsum of the foot ata fixed distance of 30 cm, immediatelyafter removing the heater probe. The scanimages were stored in a computer andprocessed offline. On the flux image, theregion of interest demarcated by the edgeof the flare was drawn, and the area of theLDIflare was calculated using MoorLDI(version 3.11) software. The results wereexpressed in centimeters squared.

Assessment of maximum hyperemia(LDImax)The same flux image described above wasalso used to calculate the maximum hy-peremia. A region corresponding exactlyto the size of heater probe was defined,and the mean flux within that region wascalculated using MoorLDI (version 3.11)software. This is the maximum hyperemicresponse that we have termed LDImax.The results are expressed in arbitrary per-fusion units.

Skin biopsySkin biopsies were performed using asterile 3-mm biopsy punch (Stiefel Labo-ratories, Bucks, U.K.) in the same area inwhich the LDIflare had been assessed on aseparate day. No local anesthetic was ap-plied and all subjects tolerated the biopsy.There was no infection or other adverseevent.

Assessment of wound closureWound closure was assessed by digitalmicroscopy at magnification �50 imme-diately after biopsy, day 3, and day 10.Digital photographs were stored in thecomputer, and the wound area was ana-

lyzed offline using Mouseyes software.The computer monitor was calibrated andthe region of interest drawn along the cir-cumference of the wound to enable calcu-lation of wound area expressed inmillimeters squared.

ImmunohistochemistryThe skin biopsy specimen was immedi-ately fixed in 4% paraformaldehyde for18 –24 h, routinely processed (Citadel2000 Processor; ThermoElectron,Waltham, MA), and embedded in paraffinwax. Serial 5-m tissue sections were cutfrom each block (Microtome Leitz Wet-zlas 1512) and mounted onto positivelycharged slides (Fisher Scientific, Lough-borough, U.K.). Sections were dewaxedin xylene and gradually rehydratedthrough decreasing ethanol dilutions.Epidermal melanin was bleached with0.25% KMnO4 followed by 5% oxalicacid. Series selected for blood vessel den-sity assessment by CD31/vWF immuno-localization underwent trypsinization.For VEGF-A and VEGF-R2, sections weremicrowaved to disclose the antigen, and,for HIF-1�, optimal visualization was ob-tained using a tyramide amplification re-agent (CSA I; Dako). Sections wereincubated overnight at 5°C with mousemonoclonal antibodies to CD31 and vWF(diluted 1:100 and mixed) (both fromDako) and to VEGFR-2 (1:50; Santa CruzBiotechnology) and with rabbit poly-clonal antibodies to VEGF-A (1:300;Santa Cruz Biotechnology) and to HIF-1�(1:300; Abcam). For nerve fiber density,sections were incubated overnight with1:1,200 biogenesis polyclonal rabbit anti-human antibody (Serotec, Oxford, U.K.).Biotinylated swine anti-rabbit secondaryantibody (1:300, 1 h) was then applied;sections were quenched with 1% H2O2 in30% MeOH-PBS (30 min) before incuba-tion for 1 h with 1:500 horseradish per-oxidase streptavidin (Vector Laboratories,Peterborough, England). The reactionswere demonstrated using the following,listed sequentially: biotinylated second-ary antibodies, streptavidin horseradishperoxidase, and the chromogenic sub-strate 3-3diaminobenzidine (DAB; Sig-ma-Aldrich, Dorset, U.K.).

Analysis of stainingPatterns of immunostaining were exam-ined by light microscopy. To quantify theamount of VEGF-A, VEGF-R2, andHIF-1� staining, microphotographs weretaken using a Nikon digital cameramounted on a Leitz DM RB microscope.

Krishnan and Associates


Percentages of stained area were quanti-fied separately in the epidermis and in theupper dermis with the computer programLeica QWin Standard, version 2.4 (LeicaMicrosystem Imaging, Cambridge, U.K.),set to detect color intensities within afixed, constant range. Blood vessel andnerve fiber cross-sections in the papillarydermis were counted manually and theirdensity expressed as number per millime-ter squared.

Statistical analysisDescriptive statistics (median and inter-quartile range) were used to describe sub-ject characteristics. Mann-Whitney U testwas used to determine the differences be-tween the groups. Mean � SD for eachvariable is described, and a P value of�0.05 was considered significant. SPSS(version 11.0) software package was usedfor statistical analysis.

RESULTSClinical characteristics of diabetic andcontrol subjects are shown in Table 1. Allsubjects were Caucasian, and there wasno significant difference in age betweengroups C and D.

Wound closureWound closure (Fig. 1) determined byarea change (mean � SD) did not differbetween diabetic patients (in mm2: day 0,6.17 � 0.5; day 3, 4.63 � 0.4; and day10, 2.93 � 0.5) and control subjects (day0, 6.28 � 0.3; day 3, 4.89 � 0.8; and day10, 3.01 � 0.7). There were no compli-cations, and all wounds were fully reepi-thelialized by day 10.

Neurovascular function/structureLDImax expressed as perfusion units (PU)was significantly reduced in the diabeticgroup (C: 577.4 � 125.3 vs. D: 310.33 �97.3; P � 0.0001), whereas dermal blood

vessel density (per mm2) did not differbetween control subjects (116.5 � 21.0mm2) and diabetic patients (116.8 � 27.8mm2) (P � 0.96). LDIflare was signifi-cantly reduced in diabetic patients com-pared with control subjects (in cm2: C:5.2 � 1.8 vs. D: 1.8 � 0.7; P � 0.0001) aswas dermal nerve fiber density (per mm2:C: 456.3 � 160.1 vs. D: 216.0 � 144.0;P � 0.001). LDIflare was significantly as-sociated with nerve fiber density (r � 0.6;P � 0.0001).

Vascular factorsThe expression of HIF-1� in epidermalvessels (C: 6.42 � 6.32 vs. D: 8.68 �11.74; P � 0.63) and dermal vessels (C:16.99 � 15.98 vs. D: 10.22 � 12.55; P �0.14) did not differ significantly betweencontrol subjects and diabetic patients (Ta-

ble 2). However, there was a significantdifference in the expression of epidermalVEGF-A (C: 0.36 � 0.30 vs. D: 0.16 �0.18; P � 0.03) and dermal VEGF-A (C:0.04 � 0.07 vs. D: 0.01 � 0.004; P �0.04). Also, epidermal blood vesselVEGFR-2 (C: 21.58 � 25.99 vs. D:9.66 � 12.91; P � 0.05) and dermalblood vessel VEGFR-2 (C: 7.94 � 6.88 vs.D: 3.45 � 3.14; P � 0.04) expressionwere significantly reduced in diabetic pa-tients compared with control subjects(Table 2, Fig. 2).

CONCLUSIONSThe pathophysiological mechanisms con-tributing to delayed wound healing in di-abetes are complex and may be mediatedby vascular, neuronal, cellular, and im-mune factors. Our study is unique, as wehave quantified the wound-healing re-sponse and related it to neurovascular integ-rity and the expression of vascular factorscentral to the wound-healing response.

Against expectation and in contrastto findings in animal models and theobservation of poor healing in diabeticpatients with foot ulceration, the rate ofwound closure was identical in diabeticand control subjects. It is important tonote that we studied the healing responseof an acute wound on the dorsum of thefoot in an area that is not exposed to con-tinued high pressure that occurs inchronic diabetic plantar foot ulcers.Whether acute wounds on the plantar

Figure 1—Wound area from biopsy to day 10 to assess rate of closure expressed as means � SDin mm2. No significant difference between the control and diabetic groups (day 0: P � 0.78; day3: P � 0.56; day 10: P � 0.95).

Table 1—Subject characteristics

C group D group

n 12 12Age (years) 50.2 (56.0–62.2) 54.0 (55.0–61.5)*Duration (years) — 10.0 (5.8–14.8)BMI (kg/m2) 25.40 (22.9–27.4) 32.3 (30.6–34.8)†A1C (%) — 8.8 (8.4–9.1)Ankle brachial pressure index 1.1 (1.0–1.2) 1.2 (1.0–1.3)*Vibration perception threshold (V) 7.0 (4.3–8.0) 40.7 (23.7–51.0)‡10-g monofilament/pressure perception Normal Abnormal

Data are median (interquartile range). *No significant difference between the groups; †P � 0.01; ‡P �0.0001.

Wound healing in type 2 diabetes


surface behave differently or trigger fac-tors such as infection, which could turnsuch wounds into chronic ulcers, remainsto be determined. None of the wounds inthe present study became infected, andgreat care was taken to ensure that thewounds were well protected.

Impaired hyperemic response to tis-sue injury and iontophoresis of acetylcho-line in the presence of normal vasculardensity has led previous investigators toimplicate functional microvascular de-fects in delayed wound healing in diabeticpatients (21–23). However, this mecha-nism has only been inferred and neverpreviously directly assessed.

While Veves et al. (24) previouslydemonstrated a reduction in endothelialnitric oxide synthase expression, fewstudies have explored in detail other mo-lecular alterations that may be relevant to

the wound-healing response following in-jury in diabetic patients. We believe suchstudies are essential if we are to gain anunderstanding of any perturbation in thewound-healing response following injuryand development of an ulcer. It is knownthat VEGF expression is normally in-creased during the granulation phase ofwound healing, and this response is di-minished in diabetic mice (16). Further-more, topical application of VEGF oroverexpression of VEGF by an adenoviralvector markedly accelerates wound heal-ing in diabetic animals (17,25). While ad-enovirus-mediated gene transfer of asoluble form of VEGFR-2 (Flk-1) reducesangiogenesis, it does not delay woundclosure in db/db mice (26). Although tis-sue hypoxia, a typical feature of healingwounds, is thought to increase the ex-pression of VEGF through HIF-1� (27),

the role of HIF-1� in diabetic wounds hasnot been explored in experimental studiesand in particular in diabetic patients. Inthe present study, we demonstrate no dif-ference in HIF-1� expression per se be-tween diabetic patients and controlsubjects.

Thus, despite an impaired maximalhyperemic response, wound healing wasnormal in our diabetic patients. Bloodvessel density was similar in the controland diabetic groups, consistent with ourprevious findings in those with type 1(28) and type 2 (24) diabetes. The normalvascular density may well have main-tained skin oxygenation, as evidenced bycomparable HIF-1� expression in bothgroups. Despite lower expression ofVEGF and VEGFR-2 in diabetic skin,wound closure did not differ between di-abetic patients and control subjects. Thissuggests that VEGF may play a limitedrole in acute wound healing in diabeticpatients.

With regard to neuropathy, it maycontribute to the development of foot ul-ceration via a loss of protective sensationand reduced axon reflex–mediated vaso-dilatation. Impaired expression and regu-lation of nerve growth factor andreduction in skin nerve density have beenspeculated to delay healing (29). Wedemonstrate a marked reduction in bothdermal nerve fiber density and the axonreflex as assessed by LDIflare. However,despite significant abnormalities in bothparameters there was no impact onwound healing.

One of the perceived limitations ofthis study is that of studying an acute

Figure 2—The picture shows immunostaining for VEGF-A (A and E), VEGFR-2 (B and F), HIF-1� (C and G), and blood vessels (D and H). Thefirst row (A–D) contains normal case subjects. The second row (E–H) contains diabetic case subjects. Note less pronounced epidermal staining forVEGF-A and VEGFR-2 in the diabetic case compared with the corresponding control subject.

Table 2—Neurovascular factors in wound closure

C group D group P value

LDImax (PU) 577.4 � 125.3 310.33 � 97.3 �0.0001LDIflare (cm2) 5.2 � 1.8 1.8 � 0.7 �0.0001Nerve fiber density (per mm2) 456.3 � 160.1 216.0 � 144.0 0.001Blood vessel density (per mm2) 116.5 � 21.0 116.8 � 27.8 0.96VEGF (epidermal) 0.36 � 0.30 0.16 � 0.18 0.03VEGF (dermal) 0.04 � 0.07 0.01 � 0.004 0.04VEGFR-2 (epidermal) 21.58 � 25.99 9.66 � 12.91 0.05VEGFR-2 (blood vessel) 7.94 � 6.88 3.45 � 3.14 0.04HIF-1� (epidermal) 6.42 � 6.32 8.68 � 11.74 0.63HIF-1� (blood vessel) 16.99 � 15.98 10.22 � 12.55 0.14

Data are means � SD. Maximum hyperemia: LDImax (PU), C-fiber function: LDIflare (cm2), and dermal nervefiber density: nerve fiber density (per mm2) was significantly reduced in diabetic patients compared withcontrol subjects. Epidermal and dermal VEGF-A and epidermal and dermal blood vessel VEGFR-2 expres-sion were significantly reduced in diabetic patients. HIF-1� and dermal blood vessel density did not differsignificantly between the diabetic patients and control subjects.

Krishnan and Associates


wound and expression of neuronal andvascular integrity at baseline, with healingby secondary intention compared withchronic wounds in a typical diabetic footulcer. However, this is no different thanall experimental studies where woundsare also acute and yet the wound healingresponse is delayed. Thus, we believe ourstudy has provided important transla-tional insights and questioned establishedconcepts of wound healing mainly de-rived from studies in experimental ani-mals. This study also establishes the safetyof distal skin biopsies in the assessment ofdiabetic neuropathy. Due to the relativelysmall number of study subjects, furtherlarger studies may be needed to confirmthe findings.

In conclusion, this study suggests thatwound closure in subjects with type 2 di-abetes is not delayed despite significantalterations in neurovascular function andstructure. This reiterates the importanceof pressure relief in those with neuro-pathic ulcers, restoration of adequateblood flow in those with ischemic ulcer-ation, and aggressive treatment of woundinfection as the principal strategies to suc-cessfully heal diabetic wounds.

References1. Pham H, Armstrong DG, Harvey C, Har-

kless LB, Giurini JM, Veves A: Screeningtechniques to identify people at high riskfor diabetic foot ulceration: a prospectivemulticenter trial. Diabetes Care 23:606–611, 2000

2. Veves A, Manes C, Murray HJ, Young MJ,Boulton AJ: Painful neuropathy and footulceration in diabetic patients. DiabetesCare 16:1187–1189, 1993

3. Young MJ, Bennett JL, Liderth SA, VevesA, Boulton AJ, Douglas JT: Rheologicaland microvascular parameters in diabeticperipheral neuropathy. Clin Sci (Colch)90:183–187, 1996

4. Flynn MD, Tooke JE: Diabetic neuropa-thy and the microcirculation. Diabet Med12:298–301, 1995

5. Khaodhiar L, Dinh T, Schomacker KT, Pa-nasyuk SV, Freeman JE, Lew R, Vo T, Pa-nasyuk AA, Lima C, Giurini JM, Lyons TE,Veves A: The use of medical hyperspectraltechnology to evaluate microcirculatorychanges in diabetic foot ulcers and to pre-dict clinical outcomes. Diabetes Care 30:903–910, 2007

6. Gibran NS, Jang YC, Isik FF, GreenhalghDG, Muffley LA, Underwood RA, UsuiML, Larsen J, Smith DG, Bunnett N, AnselJC, Olerud JE: Diminished neuropeptide

levels contribute to the impaired cutane-ous healing response associated with dia-betes mellitus. J Surg Res 108:122–128,2002

7. Vinik AI, Erbas T, Stansberry KB, Pit-tenger GL: Small fiber neuropathy andneurovascular disturbances in diabetesmellitus. Exp Clin Endocrinol Diabetes 109:451–473, 2001

8. Krishnan ST, Rayman G: The LDIflare: anovel test of C-fiber function demon-strates early neuropathy in type 2 diabe-tes. Diabetes Care 27:2930–2935, 2004

9. Christopherson K: The impact of diabeteson wound healing: implications of micro-circulatory changes. Br J Community Nurs8:S6–S13, 2003

10. Greenhalgh DG: Wound healing and dia-betes mellitus. Clin Plast Surg 30:37–45,2003

11. Jaap AJ, Shore AC, Stockman AJ, TookeJE: Skin capillary density in subjects withimpaired glucose tolerance and patientswith type 2 diabetes. Diabet Med 13:160–164, 1996

12. Katz MA, McCuskey P, Beggs JL, JohnsonPC, Gaines JA: Relationships between mi-crovascular function and capillary struc-ture in diabetic and nondiabetic humanskin. Diabetes 38:1245–1250, 1989

13. Greenhalgh DG: The role of growth fac-tors in wound healing. J Trauma 41:159–167, 1996

14. Brown DL, Kane CD, Chernausek SD,Greenhalgh DG: Differential expressionand localization of insulin-like growthfactors I and II in cutaneous wounds ofdiabetic and nondiabetic mice. Am JPathol 151:715–724, 1997

15. Trengove NJ, Stacey MC, MacAuley S,Bennett N, Gibson J, Burslem F, MurphyG, Schultz G: Analysis of the acute andchronic wound environments: the role ofproteases and their inhibitors. Wound Re-pair Regen 7:442–452, 1999

16. Frank S, Hubner G, Breier G, LongakerMT, Greenhalgh DG, Werner S: Regula-tion of vascular endothelial growth factorexpression in cultured keratinocytes: im-plications for normal and impairedwound healing. J Biol Chem 270:12607–12613, 1995

17. Galiano RD, Tepper OM, Pelo CR, BhattKA, Callaghan M, Bastidas N, Bunting S,Steinmetz HG, Gurtner GC: Topical vas-cular endothelial growth factor acceler-ates diabetic wound healing throughincreased angiogenesis and by mobilizingand recruiting bone marrow-derivedcells. Am J Pathol 164:1935–1947, 2004

18. Graiani G, Emanueli C, Desortes E, VanLS, Pinna A, Figueroa CD, Manni L, Mad-eddu P: Nerve growth factor promotes re-parative angiogenesis and inhibitsendothelial apoptosis in cutaneous

wounds of type 1 diabetic mice. Diabeto-logia 47:1047–1054, 2004

19. Muangman P, Muffley LA, Anthony JP,Spenny ML, Underwood RA, Olerud JE,Gibran NS: Nerve growth factor acceler-ates wound healing in diabetic mice.Wound Repair Regen 12:44–52, 2004

20. Wiles PG, Pearce SM, Rice PJ, MitchellJM: Vibration perception threshold: influ-ence of age, height, sex, and smoking, andcalculation of accurate centile values. Dia-bet Med 8:157–161, 1991

21. Benarroch EE, Low PA: The acetylcho-line-induced flare response in evaluationof small fiber dysfunction. Ann Neurol 29:590–595, 1991

22. Rayman G, Williams SA, Spencer PD,Smaje LH, Wise PH, Tooke JE, Hassan A:Impaired microvascular hyperaemic re-sponse to minor skin trauma in type I di-abetes. Br Med J (Clin Res Ed) 292:1295–1298, 1986

23. Rayman G, Malik RA, Sharma AK, Day JL:Microvascular response to tissue injuryand capillary ultrastructure in the footskin of type I diabetic patients. Clin Sci(Colch) 89:467–474, 1995

24. Veves A, Akbari CM, Primavera J, Dona-ghue VM, Zacharoulis D, Chrzan JS, De-Girolami U, LoGerfo FW, Freeman R:Endothelial dysfunction and the expres-sion of endothelial nitric oxide synthetasein diabetic neuropathy, vascular disease,and foot ulceration. Diabetes 47:457–463,1998

25. Romano Di PS, Mangoni A, ZambrunoG, Spinetti G, Melillo G, Napolitano M,Capogrossi MC: Adenovirus-mediatedVEGF(165) gene transfer enhanceswound healing by promoting angiogen-esis in CD1 diabetic mice. Gene Ther9:1271–1277, 2002

26. Jacobi J, Tam BY, Sundram U, von DG,Blau HM, Kuo CJ, Cooke JP: Discordanteffects of a soluble VEGF receptor onwound healing and angiogenesis. GeneTher 11:302–309, 2004

27. Yamakawa M, Liu LX, Date T, BelangerAJ, Vincent KA, Akita GY, Kuriyama T,Cheng SH, Gregory RJ, Jiang C: Hypoxia-inducible factor-1 mediates activation ofcultured vascular endothelial cells by in-ducing multiple angiogenic factors. CircRes 93:664–673, 2003

28. Malik RA, Metcalfe J, Sharma AK, Day JL,Rayman G: Skin epidermal thickness andvascular density in type 1 diabetes. DiabetMed 9:263–267, 1992

29. Walmsley D, Wales JK, Wiles PG: Re-duced hyperaemia following skin trauma:evidence for an impaired microvascularresponse to injury in the diabetic foot.Diabetologia 32:736–739, 1989

Wound healing in type 2 diabetes


Evaluation of Polyneuropathy Markers inType 1 Diabetic Kidney Transplant Patientsand Effects of Islet TransplantationNeurophysiological and skin biopsy longitudinal analysis

UBALDO DEL CARRO, MD1

PAOLO FIORINA, MD, PHD2,3

STEFANO AMADIO, MD1

LUISA DE TONI FRANCESCHINI, MD1

ALESSANDRA PETRELLI, MD2

STEFANO MENINI, PHD4

FILIPPO MARTINELLI BONESCHI, MD, PHD1

STEFANIA FERRARI, MD1

GIUSEPPE PUGLIESE, MD, PHD4

PAOLA MAFFI, MD2

GIANCARLO COMI, MD1,5

ANTONIO SECCHI, MD2,5

OBJECTIVE — The purpose of this study was to evaluate whether islet transplantation maystabilize polyneuropathy in uremic type 1 diabetic patients (end-stage renal disease [ESRD] andtype 1 diabetes), who received a successful islet-after-kidney transplantation (KI-s).

RESEARCH DESIGN AND METHODS — Eighteen KI-s patients underwent electro-neurographic tests of sural, peroneal, ulnar, and median nerves: the nerve conduction velocity(NCV) index and amplitudes of both sensory action potentials (SAPs) and compound motoraction potentials (CMAPs) were analyzed longitudinally at 2, 4, and 6 years after islet transplan-tation. Skin content of advanced glycation end products (AGEs) and expression of their specificreceptors (RAGE) were also studied at the 4-year follow-up. Nine patients with ESRD and type1 diabetes who received kidney transplantation alone (KD) served as control subjects.

RESULTS — The NCV score improved in the KI-s group up to the 4-year time point (P � 0.01versus baseline) and stabilized 2 years later, whereas the same parameter did not change signif-icantly in the KD group throughout the follow-up period or when a cross-sectional analysisbetween groups was performed. Either SAP or CMAP amplitudes recovered in the KI-s group,whereas they continued worsening in KD control subjects. AGE and RAGE levels in perineuriumand vasa nervorum of skin biopsies were lower in the KI-s than in the KD group (P � 0.01 forRAGE).

CONCLUSIONS — Islet transplantation seems to prevent long-term worsening of polyneu-ropathy in patients with ESRD and type 1 diabetes who receive islets after kidney transplantation.No statistical differences between the two groups were evident on cross-sectional analysis. Areduction in AGE/RAGE expression in the peripheral nervous system was shown in patientsreceiving islet transplantation.

Diabetes Care 30:3063–3069, 2007

Chronic sensorimotor diabetic poly-neuropathy (DPN) is a commonlong-term complication of type 1 di-

abetes, affecting �50% of patients (1).Electroneurographic studies, based onnerve conduction velocity (NCV) studies,represent an objective method for DPNassessment (2) and may also predict mor-tality (3). Among treatments that mayprevent either DPN onset or progressionby restoring normoglycemia, pancreastransplantation has been widely studiedwith NCV in the recent past (4,5). Untilnow, little has been known of the effect ofislet transplantation on DPN.

Indications for allogenic pancreaticislet transplantation in type 1 diabeteshave been expanding over the past fewyears (6) thanks to recent improvementsin long-term graft survival rates (7) thatdepend on advances in islet isolation andpurification and new immunosuppressiveprotocols (8). Compared with the wholeorgan transplant, which has a 5% mortal-ity rate 1 year after surgery and severesurgical complications (9), islet trans-plantation is a minimally invasive thera-peutic approach, allowing for long-terminsulin independence and metabolic con-trol (7,8).

However, the question of whetherlong-term diabetes complications may behalted or even reversed by transplantationis still under investigation. Immunosup-pressive treatment may interfere with re-nal function (10); yet the protective roleof islet transplantation on both long-termgraft survival and function of the trans-planted kidney in type 1 diabetic patientshas been demonstrated (11,12). In type 1diabetic patients who had received a kid-ney transplant, benefits for either macro-/microangiopathy or cardiovascularfunction (13,14) and also for retinal com-plications (15,16) were reported. Stabili-zation of peripheral neuropathy wasreported after islet transplantation alone;i.e., it was not associated with kidneytransplantation (15,16).

The aim of this study was to evaluatewhether islet transplantation might stop

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Neurology and Clinical Neurophysiology, San Raffaele Scientific Institute, Milan,Italy; the 2Department of Medicine, San Raffaele Scientific Institute, Milan, Italy; the 3TransplantationResearch Center, Brigham and Women’s Hospital/Children’s Hospital/Harvard Medical School, Boston,Massachusetts; the 4Department of Clinical Science, La Sapienza University Rome, Rome, Italy; and the5Universita Vita-Salute San Raffaele, Milan, Italy.

Address correspondence and reprint requests to Paolo Fiorina, MD, Department of Medicine, San RaffaeleScientific Institute, Universita Vita e Salute, Via Olgettina 60, 20132 Milan, Italy. E-mail: [email protected].

Received for publication 31 January 2007 and accepted in revised form 24 August 2007.Published ahead of print at http://care.diabetesjournals.org on 5 September 2007. DOI: 10.2337/dc07-

0206.U.D.C. and P.F. contributed equally to this study.Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/

dc07-0206.Abbreviations: AGE, advanced glycation end product; CMAP, compound motor action potential; CML,

Nε-(carboxymethyl) lysine; DPN, diabetic polyneuropathy; ESRD, end-stage renal disease; NCV, nerveconduction velocity; RAGE, receptor for AGE; SAP, sensory action potential.






the progression of neuropathy in type 1diabetic patients with end-stage renal dis-ease (ESRD) bearing a kidney graft: thus,we performed longitudinal NCV studiesin a group of uremic type 1 diabetic pa-tients who had received kidney trans-plants throughout the 6 years after islettransplantation. Further, at the 4-year fol-low-up, we investigated cross-sectionallythe skin expression of advanced glycationend products (AGEs) and their specificreceptor (RAGE), which were reported toplay a role in the pathogenesis of DPN(17).

RESEARCH DESIGN ANDMETHODS — Among all consecutivepatients with ESRD and type 1 diabeteswho had received a successful islet-after-kidney transplantation from 1991 untilJanuary 2004 (KI-s group), 18 (age38.7 � 5.7 years; male-to-female ratio8:10) with sustained C-peptide secretionof �0.5 ng/ml for �6 months werereferred to our laboratory for electro-neurographic assessment of theirpolyneuropathy, which had been diag-nosed previously on the basis of clinicalfindings (18). Nine patients with ESRDand type 1 diabetes were considered acontrol group (KD [kidney only] group:age 39.1 � 2.2; male-to-female ratio 4:5).The efficacy of islet function in the KI-sgroup was assessed by fasting circulatingC-peptide levels. KI-s patients were fol-lowed for an average of 53.4 � 7.1months after transplantation. Any pa-tients who lost islet function early aftertransplantation (within 6 months) wereenrolled in the KD group. If the trans-planted islets produced �0.5 ng/ml C-peptide, the patient was considered asubject in the KI-s group.

Subanalysis of the KI-s groupWithin the KI-s group, different patientsappeared to have different degrees of met-abolic control. Therefore, a subanalysis ofthe patients who reached full islet func-tion (fasting C-peptide �1 ng/ml) wasperformed.

Diabetes management after kidney-islet transplantationInsulin therapy was used after transplan-tation to maintain strict glycometaboliccontrol in patients. In some patients oralhypoglycemic agents were used to im-prove islet function, particularly when in-sulin resistance was clearly evident on thebasis of a requirement for very high dosesof insulin.

Laboratory assessmentFasting levels of cyclosporine, creatinine,A1C, serum C-peptide, total cholesterol,and triglycerides were assayed at baselineand 2, 4, and 6 years after transplantation(19). Serum C-peptide levels (intra-assayand interassay coefficients of variationboth 3.0%) were assayed by radioimmu-noassay using a commercial kit (MedicalSystem, Genoa, Italy).

Islet transplantationBoth kidneys and islets came from ca-daver donors; transplantation was per-formed according to HLA matching forkidney graft, whereas ABO compatibilitywas used for islet transplantation (13).The cross-match test was negative in allcases. The details of the procedure can befound in the online appendix (available athttp://dx.doi.org/10.2337/dc07-0206).

Nerve conduction studyNerve conduction study was performedafter standard laboratory procedures byan operator who did not know whichgroup the patient belonged to. The detailsof the procedure can be found in the on-line appendix.

NCV indexBesides standard conduction parameters,an NCV index was assessed for each pa-tient, as already reported (20). Briefly,conduction velocities of each nerve seg-ment (motor conduction velocities ofdeep peroneal and ulnar nerves; sensoryconduction velocities of the sural nerveand either wrist-finger or elbow-wristsegments of the median nerve) were firstconsidered to obtain five nerve NCV Zscores [(patient’s NCV value – mean NCVvalue in control healthy subjects)/SD ofthe same nerve in control healthy sub-jects]. The patient’s NCV index was themean of each nerve NCV Z score. Age-related normative values (mean � SD)from our electromyography laboratorywere considered for calculation. The NCVindex estimates to what extent individualNCV values deviate from the mean valueof a reference population in terms of SD,limiting the specific intraindividual vari-ability for each nerve trunk and allowingfor an easier longitudinal evaluation. Be-cause of the method of NCV index calcu-lation, most negative NCV values identifythe most severe polyneuropathies. In sub-jects with complete nerve unexcitability,the last available NCV value was consid-ered for NCV index calculation.

Skin biopsyPatients underwent skin punch biopsy onthe internal surface of the arm 4.5 � 1.2years after transplantation as describedelsewhere (13). The procedure is easy,minimally invasive, and well tolerated bypatients. Samples were taken with the pa-tient’s consent and ethical review boardapproval.

AGEs and RAGE quantificationNε-(carboxymethyl) lysine (CML)-protein adduct content and RAGE expres-sion were assessed in paraffin-embeddedsections by immunohistochemical analy-sis. Immunoreactivity for CML and RAGEin nerves and perineural vessels was eval-uated with a semiquantitative scale forstaining (�, absent; �, mild; ��, mod-erate; ��, strong). The details of theprocedure can be found in the onlineappendix.

Statistical analysisBecause of the limited sample size, statis-tical analysis was performed by means ofnonparametric tests. We performed thefollowing comparisons: 1) NCV indexesat the different follow-up examinations(2, 4, and 6 years) were compared withbasal values by means of the Wilcoxonsigned rank-sum test for paired samplesin the entire sample and in the KI-s andKD subgroups; and 2) changes from base-line of NCV indexes at the different fol-low-up examinations (2, 4, and 6 year)were compared between the KI-s and KDgroups by means of the Wilcoxon rank-sum test. We also performed a repeated-measures ANOVA to simultaneouslyexplore the role of the time of evaluation,defined as a within-subject factor, and ofthe group of treatment (KI-s versus KD),defined as a between-subjects factor, andtheir interaction on the NCV index value.This type of analysis permits taking intoaccount the correlation of measures per-formed on the same subject and exploringthe role of the time of evaluation and thetreatment on the NCV index.

RESULTS

Population and metabolic variablesThe two groups of recipients had similarpretransplant and peritransplant char-acteristics, in particular the pattern ofrejection episodes, cytomegalovirus in-fections, and kidney retransplantation(data not shown). The mean numbers ofHLA matches for the kidney and plasmarenin activity levels were similar in the

DNP after islet transplantation


two groups (data not shown), as wereimmunosuppression, lipid profile, andmedications. After steroid withdrawal,no kidney or islet rejections were evi-dent. At 6 months after islet transplan-tation, when almost all of the patientshad completed steroid withdrawal, asignificant reduction in A1C was evi-dent (data not shown) in the wholegroup of patients.

No significant intergroup differenceswere found for baseline body weight (KI-s61.3 � 9.6 vs. KD 62.0 � 2.2 kg), diabe-tes duration (KI-s 26.3 � 10.4 vs. KD22.2 � 1.4 years), and dialysis duration(KI-s 42.5 � 6.2 vs. KD 27.6 � 4.1 years).Among follow-up data, the main resultwas the mean creatinine value in the KDgroup, which increased significantly overbaseline at 6 years’ follow-up (P � 0.004)(Table 1). These results confirm previousreports of a protective effect of trans-planted islets on kidney grafts. C-peptidesecretion was higher in the KI-s group andabsent in the KD group (Table 1); the in-sulin requirement was lower in the KI-sgroup than in the KD group (data notshown), with better glycometabolic con-trol (Table 1).

NCV indexNo significant differences were observedin pretransplant NCV scores between KI-sand KD groups (P � 0.6). Both KI-s andKD groups were neuropathic at baselineaccording to electroneurographic find-ings, showing NCV index �2 (e.g., withan NCV score exceeding the mean normalvalue by �2 SD). The longitudinal NCVindex study showed that at the 2-year fol-low-up both KI-s and KD groups scored alittle better than at baseline; at the follow-ing time point (4 years after transplant),however, the NCV index continued to im-prove only in the KI-s group, reaching sta-tistical significance in comparison withpretransplant values at 4 years of fol-low-up (P � 0.01), whereas NCV wors-ened toward baseline values in the KDgroup (NS). At the latest follow-up (6years), the NCV score improvement incomparison to baseline values was main-tained in the KI-s group, whereas it wors-ened further in the KD group (Table 1).When we compared the NCV changesfrom baseline across the two groups, de-spite the evidence of a statistically signif-icant difference at the 4-year follow-up inthe KI-s group, results were not statisti-

cally significant. In particular, the medianchange from baseline at each time pointshowed differences between the KI-s andKD groups that did not reach statisticalsignificance (0.41 vs. �0.08, 0.55 vs.0.17, and 0.11 vs. �0.03 at the 2-, 4- and6-year follow-ups, respectively).

Compound motor action potentialIn the KI-s group, the longitudinal trendof compound motor action potential(CMAP) mean amplitudes showed slight,although not significant, improvementsof either peroneal or ulnar nerves at the6-year follow-up compared with baselinevalues. On the contrary, CMAP ampli-tudes of both nerves progressively de-clined in the KD group over time; theworsening was statistically significant forulnar CMAP mean amplitude 6 years afterkidney transplantation (9.7 � 3.0 vs.13.7 � 4.6 mV; P � 0.03) (Table 1).

Sensory action potentialA slightly improving trend of sensory ac-tion potential (SAP) amplitudes throughthe different time points up to the 6-yearfollow-up was also recognized in patientsof the KI-s group even though statistical

Table 1—NCV index and other electroneurographic findings (baseline and longitudinal)

Groups and variables Basal 2 years 4 years 6 years

KI-sn 18 18 18 9Age (years) 41.8 � 6.2A1C (%) 8.0 � 1.1 7.7 � 1.8 7.4 � 1.8 7.5 � 0.4Creatinine (mg/dl) 1.6 � 1.5 1.3 � 0.5 1.4 � 0.7 1.1 � 0.2C-peptide (ng/ml) 0.1 � 0.1 1.8 � 1.0 1.1 � 0.5 1.4 � 1.1NCV index �2.9 � 0.9 �2.8 � 1.1 �2.6 � 1.0* �2.7 � 0.9Sural SAPampl 7.9 � 3.2 9.0 � 8.1 16.4 � 25.3 13.7 � 19.0Median SAPampl 16.7 � 9.0 19.6 � 10.1 19.4 � 11.1 20.0 � 15.6PeronealCMAPampl

3.0 � 4.2 2.5 � 2.6 2.8 � 3.9 4.4 � 4.5

Ulnar CMAPampl 10.2 � 4.0 9.5 � 3.3 10.3 � 3.0 11.6 � 2.2KD

n 9 9 9 9Age (years) 39.1 � 2.2A1C (%) 11.1 � 2.3 8.0 � 0.4 8.6 � 0.4 8.1 � 0.4Creatinine (mg/dl) 1.7 � 0.1 1.9 � 0.2 2.0 � 0.3 2.5 � 0.7*C-peptide (ng/ml) 0.1 � 0.1NCV index �2.7 � 1.2 �2.5 � 0.9 �2.6 � 1.0 �2.8 � 1.1Sural SAPampl 4.4 � 2.8 3.6 � 0.8 6.5 � 0.7 6.3 � 7.3Median SAPampl 13.3 � 6.8 15.7 � 9.0 17.8 � 5.0 11.8 � 6.6PeronealCMAPampl

3.9 � 3.2 3.5 � 2.5 2.0 � 1.8 3.2 � 3.6

Ulnar CMAPampl 13.7 � 4.6 10.6 � 5.1 12.4 � 3.3 9.7 � 3.0*

Data are means � SD. *P � 0.05 versus basal values. SAPampl, sensory action potential amplitude (expressed as microvolts); CMAPampl, compound motor actionpotential amplitude (expressed as millivolts).

Del Carro and Associates


significance was lacking. In general, thiswas true for both median and suralnerves, but only sural SAP amplitudeschanged from neuropathic to normal val-ues, whereas median SAP values startedwithin the normal range at baseline. In theKD group, both median and sural SAPamplitudes fluctuated slightly throughthe different time points, but values al-ways remained in a neuropathic spectrum(Table 1).

Regression modelRepeated-measures ANOVA showed thatthe type of treatment (between-subjectsfactor) did not influence the NCV index(P � 0.5); in addition, the time of evalu-

ation (within-subject factor) and their in-teraction were not statistically significant(P � 0.4 and 0.4). It is also worthwhile tonote that the 2-year response to treatmentwas maintained at the 4-year follow-upbecause all of the patients with an improv-ing 2-year NCV index had a further in-crease 2 years later.

Skin biopsy analysis and AGE/RAGEexpressionAn overall morphological analysis of skinbiopsies dramatically showed the effect ofdiabetes and uremia on skin innervation(Fig. 1). Compared with healthy controlsubjects (Fig. 1A), the skin of uremic type1 diabetic patients appeared to be com-

pletely denervated. Even the sweat glandsshowed no evidence of residual innerva-tion (Fig. 1B, inset).

Paraffin-embedded sections of skinbiopsy specimens were analyzed for CMLand RAGE content by the immunoperox-idase technique. In the KD group (Fig.1C), the perineurium of peripheral nerves(asterisk) and vasa nervorum (arrows) indermis showed strong staining for CML,whereas the KI-s group (Fig. 1D) showedonly mild staining. The score for CML ex-pression (Fig. 1E) differed between thetwo groups, although not significantly(P � 0.07).

Likewise, strong RAGE expressionwas observed in the bundles of axons and

Figure 1— Comparison with skin biopsies of control patients (A) dramatically showed the skin denervation in uremic type 1 diabetic patients(ESRD�T1DM) (B). Even sweat glands appeared to be completely denervated (B, inset). Immunohistochemical analysis for CML. Paraffin-embedded sections of skin biopsy specimens stained with immunoperoxidase technique. CML was detected by binding to anti-CML monoclonalantibody, biotin-conjugated. C: KD group: peripheral nerve (�) with adjoining blood vessels (vasa nervorum, arrows) in dermis. There is strong(��) staining of the perineurium and vasa nervorum. D: KI-s group: another skin biopsy section with only mild (�) staining of the same tissuestructures. Original magnification, �400. E: The score for CML expression showed a difference between the two groups despite no statisticalsignificance (P � 0.07). Immunohistochemical analysis for RAGE. Paraffin-embedded sections of skin biopsy specimens stained with immunoper-oxidase technique. RAGE was detected by binding to anti-human RAGE goat polyclonal antibody followed by a biotinylated anti-goat IgG antibody.F: KD group: peripheral nerve (�) with adjoining blood vessels (vasa nervorum, arrows) in dermis. There is a strong (��) staining of the bundlesof axons and vasa nervorum. G: KI-s group: another skin biopsy section with only mild (�) staining of the same tissue structures. Originalmagnification, �400. H: The score for RAGE expression showed a higher expression of RAGE in the KD group (P � 0.01).



vasa nervorum in the KD group (Fig. 1F),whereas only mild staining of the sametissue structures was evident in the KI-sgroup (Fig. 1G). The score for RAGE ex-pression (Fig. 1H) confirmed the higherexpression of RAGE in the KD than in theKI-s group (P � 0.01).

Subanalysis of the KI-s groupaccording to the degree of metaboliccontrolNine patients in the KI-s group experi-enced full function (average 52.0 � 11.8months) of transplanted islets, and five ofthem did not require insulin treatment for�2 years. Of interest, an improvement inNCV index was evident in the group ofpatients who achieved better glycometa-bolic control and the withdrawal of insu-lin therapy. In particular, the NCV indeximproved in the full function group froma baseline value of �2.9 � 0.3 to �2.0 �0.4 at 4 years after islet transplantation.This tendency was confirmed for suralSAP amplitude too, which improved from16.2 � 13.3 at baseline in the full func-tion group to 20.1 � 17.1 at 4 years afterislet transplantation, but was not evidentin the group that did not experience fullfunction of the transplanted islets (datanot shown).

Limitations of the studyWe acknowledge that a randomized trialwould be the only way to determinewhether islet transplantation can clearlyimprove diabetic neuropathy and be su-perior to intensive insulin treatment. Thepersistence of islet allograft function con-tributed to improved glycemic controland therefore to less severe diabetes com-plications. Similar results most probablywill be obtained with better glycemic con-trol in the control group, as demonstratedby the Diabetes Control and Complica-tion Trial (5).

CONCLUSIONS — This study pro-vided several indications, coming fromboth physiological and pathologicalsources, that islet transplantation may in-duce long-lasting stabilization of DPN. Infurther cross-sectional analysis, however,no statistical differences were evident be-tween the two groups. It is likely that thereason for the lack of significance is thesmall number of patients and the low sta-tistical power for NCV differences overtime.

This evidence is first supported by anobjective method, electroneurography,which is the most sensitive method for

assessing DPN (21). In fact, the NCV in-dex increased in the KI-s group from thefirst control (2 years), reaching maximumimprovement at the 4-year follow-up, be-fore stabilizing at the latest time point (6years). On the other hand, the NCV indexshowed some improvement in the KDgroup as well, but it was short-lived (2years), as experienced previously (20);then, the NCV index declined toward pre-transplant values after the 4-year fol-low-up and further worsened by 6 years.Because pretransplant variables of age,sex, laboratory values, and electroneuro-graphic findings did not differ at baselinebetween groups, we can conclude thatthis result is not affected by a bias in theselection of the groups but is probablydue to the efficiency of islet function.

There are several reasons for this fa-vorable trend of DPN in our KI-s group.First, islet function may prevent the well-known nephropathy of the transplantedkidney (11,12), as suggested by longitu-dinal behavior of creatinine levels (stablein KI-s group and significantly worseningin KD group). Thus, islet transplantationwould eliminate an important neuro-pathic noxa such as chronic renal disease(22). An additional benefit may resultfrom higher levels of C-peptide in the KI-sgroup, which were reported to improvenerve function in both experimental andclinical settings (23). The feature we wantto highlight is that, despite wide and dra-matic skin denervation at baseline, a sig-nificant reduction in vasa nervorumRAGE expression was found in patientswith islet function at the 4-year follow-up, which clearly demonstrates that isletfunction reverses a primary pathogeneticmechanism specifically related to DPN(2,17).

Intensive insulin treatment was in factshown to significantly improve peripheralnerve function, both autonomic and sen-sorimotor, in type 1 diabetic patients,compared with conventional therapy. Inthe secondary intervention cohort pa-tients, i.e., those who had neuropathy atbaseline (as in our population), althoughless severe, intensive therapy reduced theappearance of clinical neuropathy at 5years by 57% (5). The injurious effect ofchronic hyperglycemia on vessels andnerves has been attributed to various bio-chemical consequences of intracellularmetabolism of excess glucose, includingnonenzymatic glycation with formationof AGEs (24). AGEs are heterogeneouscompounds originating from precursorsformed both nonoxidatively and oxida-

tively; the latter group includes the mono-lysyl adduct CML, which has beendetected in peripheral nerves from dia-betic patients (25). In addition to direct,physicochemical effects, such as trappingand cross-linking of macromolecules,AGEs exert indirect, biological effects,mediated by cell surface receptors. RAGE,whose expression is positively regulatedby AGEs (26), is the prototypic AGE re-ceptor mediating AGE-induced tissue in-jury via induction of reactive oxygenspecies formation and activation of redox-sensitive signaling pathways, and CML isa major RAGE ligand (27). The findingthat, despite a comparable, dramatic de-gree of skin denervation, patients in theKI-s groups showed a significant reduc-tion in RAGE expression (associated witha nonsignificant decrease in CML con-tent) in nerves and vasa nervorum com-pared with KD patients, supports a rolefor the downregulation of the AGE-RAGEpathway as a molecular mechanism un-derlying the improvement of neuropathyobserved after successful islet transplan-tation.

Previous works have already reportedbeneficial effects of islet transplantationon DPN (24,25), though there are rele-vant differences from our study. First,both of the above-mentioned studies in-cluded type 1 diabetic patients and nottype 1 diabetic patients with ESRD.Hence, neuropathy was supposed to beeven more severe in our study, and thepositive role of islet transplantation isstrengthened by our study. Second, pe-ripheral nerve function was assessed pre-viously with NCVs only by Lee et al. (16),but the follow-up period was no longerthan 2 years; in the article by Varkonyi etal (14), the follow-up was as long as 9.5years on average, but they evaluated DPNonly with a perceptive test of sensorythreshold. Therefore, this is the first studybased on both nerve conduction and skinbiopsy to demonstrate that islet trans-plantation may induce long-lasting stabi-lization or even improvement of polyneu-ropathy in type 1 diabetic patients whoreceived kidney transplants.

Looking at other electroneurographicvariables such as the amplitude of bothSAPs and CMAPs, which are currentlyrecognized as indicators of axon integrityin peripheral neuropathies (27), wefound that CMAP amplitude remainedstable throughout the follow-up period inthe KI-s group. On the contrary, the am-plitude of both peroneal and ulnar nerveCMAPs declined progressively through



the follow-up period in the KD group(with significant differences for the 6-yearulnar nerve CMAP from baseline) (Table1), as though axonal damage were stillprogressing after isolated kidney trans-plant. The amplitude of both sural andmedian nerve SAPs increased slightlythrough the different time points in theKI-s group, whereas the same parametershowed a fluctuating pattern in KD pa-tients, suggesting that islet transplant pos-itively affected even sensory nerve fibers;however, this conclusion must be madewith caution because of the lack of statis-tical significance and somewhat heteroge-neous baseline values.

Hence, changes of each electroneuro-graphic variable depict distinct aspects ofDPN influenced by islet transplantation.In fact, because NCV changes are relatedto glycemic control (20), improvementsin the NCV index indicate an overall im-provement of patient nerve function. AsNCV mainly reflects pathological pro-cesses of large-diameter axons (27), wealso included, among electroneuro-graphic variables, the longitudinal analy-sis of either SAP or CMAP amplitudes,which were suggested as a means of as-sessing the contribution of smaller slow-conduction nerve fibers (2). These latterdata are in keeping with evidence ofRAGE expression in vasa nervorum andsmall skin nerve terminals in our patients.Taken together, electroneurographic andskin biopsy studies allowed longitudinal,long-term assessment of patients withDPN and served as surrogate markers ofthe restored glycemic control.

In light of the advantages suggested inour article, islet transplantation could be-come an option for improving quality oflife for diabetic patients with brittle dia-betes, for those who are unaware of life-threatening hypoglycemia, and even forthose with severe, often painful forms ofpolyneuropathy. More studies with alarger number of patients are requiredto definitely clarify whether the positivetrend observed in our preliminary studycan ultimately result in a strong positiveassociation.

In our small group of patients therewere no differences with kidney-onlytransplantation. Islet transplantation inaddition to kidney transplantation makesno difference in nerve function comparedwith kidney transplantation only.

Acknowledgments— We thank the Islet Iso-lation Core (particularly Federico Bertuzzi,

Rita Nano, and Barbara Antonioli). We thankMollie Jurewicz for editing of the manuscriptand Alessandra Mello for amazing support.

References1. Dyck PJ, Kratz KM, Karnes JL, Litchy WJ,

Klein R, Pach JM, Wilson DM, O’BrienPC, Melton LJ III: The prevalence bystaged severity of various types of diabeticneuropathy, retinopathy, and nephropa-thy in a population-based cohort: theRochester Diabetic Neuropathy Study.Neurology 43:817–824, 1993

2. Boulton AJM, Malik RA, Arezzo JC, So-senko JM: Diabetic somatic neuropathies.Diabetes Care 27:1458–1486, 2004

3. Carrington AL, Shaw JE, Van Schie CHM;Abbott CA, Vileikite L, Boulton AJM: Canmotor nerve conduction velocity predictfoot problems in diabetic subjects over a6-year outcome period? Diabetes Care 25:2010–2015, 2002

4. Kennedy WR, Navarro X, Goetz FC, Suth-erland DE, Najaran JS: Effects of pancre-atic transplantation on diabetic neuro-pathy. N Engl J Med 322:1031–1037,1990

5. The Diabetes Control and ComplicationTrial Research Group: The effect of inten-sive treatment of diabetes on develop-ment and progression of long termcomplications in insulin-dependent dia-betes mellitus. N Engl J Med 329:977–986,1993

6. Bertuzzi F, Secchi A, Di Carlo V: Islettransplantation in type 1 diabetic pa-tients. Transplant Proc 36:603–604, 2004

7. Ryan EA, Lakey JRT, Paty BW, Imes S,Korbutt GS, Kneteman NM, Bigam D, Ra-jotte RV, Shapiro AMJ: Successful islettransplantation: continued insulin reserveprovides long-term glycemic control. Di-abetes 51:2148–2157, 2002

8. Shapiro AMJ, Lakey JRT, Ryan EA, Kor-butt GS, Toth E, Warnock GL, KnetemanNM, Rajotte RV: Islet transplantation inseven patients with type 1 diabetes melli-tus using a glucocorticoid-free immuno-suppressive regimen. N Engl J Med 343:230–238, 2000

9. Gruessner RW, Sutherland DE, Tropp-mann C, Benedetti E, Hakim N, Dunn DL,Gruessner AC: The surgical risk of pan-creas transplantation in the cyclosporineera: an overview. J Am Coll Surg 185:128–144, 1997

10. Senior PA, Paty BW, Cockfield SM, RyanEA, Shapiro AMJ: Proteinuria developingafter clinical islet transplantation resolveswith sirolimus withdrawal and increasedtacrolimus dosing. Am J Transplant 5:2318–2323, 2005

11. Fiorina P, Folli F, Zerbini G, Maffi P,Gremizzi C, Di Carlo V, Socci C, BertuzziF, Kashgarian M, Secchi A: Islet trans-plantation is associated with improve-ment of renal function among uremicpatients with type I diabetes mellitus and

kidney transplants. J Am Soc Nephrol 14:2150–2158, 2003

12. Fiorina P, Venturini M, Folli F, Losio C,Maffi P, Placidi C, La Rosa S, Orsenigo E,Socci C, Capella C, Del Maschio A, SecchiA: Natural history of kidney graft survival,hypertrophy, and vascular function inend-stage renal disease type 1 diabetickidney-transplanted patients. DiabetesCare 28:1303–1310, 2005

13. Fiorina P, Folli F, Bertuzzi F, Maffi P,Finzi G, Venturini M, Socci C, Davalli A,Orsenigo E, Monti L, Falqui L, Uccella S,La Rosa S, Usellini U, Properzi G, Di CarloV, Del Maschio A, Capella C, Secchi A:Long-term beneficial effect of islet trans-plantation on diabetic macro-/microan-giopathy in type 1 diabetic kidney-transplanted patients. Diabetes Care 26:1129–1136, 2003

14. Fiorina P, Gremizzi C, Maffi P, Caldara R,Tavano D, Monti L, Socci C, Folli F, FazioF, Astorri E, Del Maschio A, Secchi A: Islettransplantation is associated with an im-provement of cardiovascular function intype 1 diabetic kidney transplant patients.Diabetes Care 28:1358–1365, 2005

15. Varkonyi TT, Farkas G, Fulop Z, Voros P,Lengyel CS, Kempler P, Lonovics J: Ben-eficial effect of fetal islet grafting on devel-opment of late diabetic complications.Transplant Proc 30:330–331, 1998

16. Lee TC, Barshes NR, O’Mahony CA,Nguyen L, Brunicardi FC, Ricordi C, Ale-jandro R, Schock AP, Mote A, Goss JA:The effect of pancreatic islet transplanta-tion on progression of diabetic retinopa-thy and neuropathy. Transplant Proc 37:2263–2265, 2005

17. King RH: The role of glycation in thepathogenesis of diabetic polyneuropathy.Mol Pathol 54:400–408, 2001

18. Report and recommendation of the SanAntonio conference on diabetic neuropa-thy: consensus statement. Diabetes 37:1000–1004, 1988

19. Fiorina P, La Rocca E, Venturini M, Mi-nicucci F, Fermo I. Paroni R, D’Angelo A,Sblendido M, Di Carlo V, Cristallo M, DelMaschio A, Pozza G, Secchi A: Effects ofkidney-pancreas transplantation on ath-erosclerotic risk factors and endothelialfunction in patients with uremia and type1 diabetes. Diabetes 50:496–501, 2001

20. Martinenghi S, Comi G, Galardi G, DiCarlo V, Pozza G, Secchi A: Ameliorationof nerve conduction velocity following si-multaneous kidney/pancreas transplanta-tion is due to the glycaemic controlprovided by the pancreas. Diabetologia 40:1110–1112, 1997

21. Krarup C: An update on electrophysiolog-ical studies in neuropathy. Curr Opin Neu-rol 16:603–612, 2003

22. Bolton CF: Peripheral neuropathies asso-ciated with chronic renal failure. CanJ Neurol Sci 7:89–96, 1980

23. Ekberg K, Brismar T, Johansson BL, Jon-



sson B, Lindstrom P, Wahren J: Amelio-ration of sensory nerve dysfunction byC-peptide in patients with type 1 diabe-tes. Diabetes 52:536–541, 2003

24. Di Mario U, Pugliese G: 15th Golgi Lec-ture: from hyperglycaemia to the dysregu-lation of vascular remodelling in diabetes.Diabetologia 44:674–692, 2001

25. Sugimoto K, Nishizawa Y, Horiuchi S,Yagihashi S: Localization in human dia-

betic peripheral nerve of Nε-carboxym-ethyllysine-protein adducts, an advancedglycation endproduct. Diabetologia 40:1380–1387, 1997

26. Tanji N, Markowitz GS, Fu C, KislingerT, Taguchi A, Pischetsrieder M, Stern D,Schmidt AM, D’Agati VD: Expression ofadvanced glycation end products andtheir cellular receptor RAGE in diabeticnephropathy and nondiabetic renal dis-

ease. J Am Soc Nephrol 11:1656 – 66,2000

27. Tankisi H, Pugdahl K, Fuglsang-Frederik-sen A, Johnsen B, de Carvalho M, FawcettPRW, Labarre-Vila A, Liguori R, Nix WA,Schofield IS: Pathophysiology inferredfrom electrodiagnostic nerve tests andclassification of polyneuropathies. Sug-gested guidelines. Clin Neurophysiol 116:1571–1580, 2005



Incidences, Treatments, Outcomes, and SexEffect on Survival in Patients With End-Stage Renal Disease by Diabetes Status inAustralia and New Zealand (1991–2005)EMMANUEL VILLAR, MD, PHD

1,2

SEAN HAW CHANG, MBBS, MRCP1,3

STEPHEN PETER MCDONALD, MBBS, FRACP, PHD1,3

OBJECTIVE — We aimed to update the epidemiology of type 1 and type 2 diabetic patientsamong the incident end-stage renal disease (ESRD) population in Australia and New Zealand(ANZ) and to determine whether outcome is worse for diabetic women, as described in thegeneral population.

RESEARCHDESIGNSANDMETHODS — All resident adults of ANZ who began renalreplacement therapy (RRT) from 1 April 1991 to 31 December 2005 were included using datafrom the ANZ Dialysis and Transplant Registry. Incidence rates, RRT, and survival were ana-lyzed. Risk factors for death were assessed using Cox regression.

RESULTS — The study included 1,284 type 1 diabetic (4.5%), 8,560 type 2 diabetic (30.0%),and 18,704 nondiabetic (65.5%) patients. The incidence rate of ESRD with type 2 diabetesincreased markedly over time (�10.2% annually, P � 0.0001). In patients aged �70 years, ratesof renal transplantation in type 1 diabetic, type 2 diabetic, and nondiabetic patients were 41.8,6.5 (P � 0.0001 vs. other patients), and 40.9% (P � 0.56 vs. type 1 diabetic patients), respec-tively. Compared with nondiabetic patients, the adjusted hazard ratio (HR) for death was 1.64(P � 0.0001) in type 1 diabetes and 1.13 (P � 0.0001) in type 2 diabetes. Survival rates per5-year period improved by 6% in type 1 diabetic patients (P � 0.36), by 9% in type 2 diabeticpatients (P � 0.0001), and by 5% in nondiabetic patients (P � 0.001). In type 2 diabetic patientsaged �60 years, the adjusted HR for death in women versus men was 1.19 (P � 0.0003).

CONCLUSIONS — The incidence of ESRD with type 2 diabetes increased markedly. De-spite high access to renal transplants, type 1 diabetic patients had a poor prognosis after startingRRT. Survival improved significantly in type 2 diabetic patients during the study period. Oldertype 2 diabetic women had a worse prognosis than older type 2 diabetic men.

Diabetes Care 30:3070–3076, 2007

D iabetes is associated with high mor-tality in the general population (1,2).Worse prognosis has also been re-

ported in diabetic women compared withdiabetic men (3,4). End-stage renal disease(ESRD) in patients with type 2 diabetes hasincreased dramatically worldwide duringthe last few decades, and diabetes is associ-ated with worse survival among patientsundergoing dialysis (5–7).

Nevertheless, a study in Denmarkshowed that the survival rate of patientswith ESRD who had type 2 diabetes hasimproved during the 1990–2005 period(8). Available studies on patients withESRD who have type 1 and type 2 diabe-tes have shortcomings because analyseswere limited to patients with diabetic ne-phropathy (6–7), did not differentiate thetwo types of diabetes (9), were short-term(10), or were based on single-center ex-periences (11).

The aim of the present study was toexamine the epidemiology and long-termsurvival of patients with incident ESRD bydiabetes status (type 1 diabetes, type 2diabetes, and no diabetes) in Australiaand New Zealand (ANZ) and to determinewhether outcomes were different betweenthe sexes among patients with diabetes.

RESEARCH DESIGN ANDMETHODS — We performed a pro-spective study including all patients aged�16 years who began chronic renal re-placement therapy (RRT) in ANZ from 1April 1991 to 31 December 2005. Weused data from the Australia and NewZealand Dialysis and Transplant (ANZ-DATA) Registry (5). Patients were fol-lowed until death or 31 December 2005.Data collection consisted of informationon patient demographic characteristics,cause of ESRD, comorbidities at start ofRRT (presence of type 1 diabetes, type 2diabetes, coronary artery disease, periph-eral vascular disease, cerebrovascular dis-ease, or chronic lung disease; BMI; andsmoking status), estimated glomerular fil-tration rate (eGFR) at the first RRT, detailsof RRT modality and of renal transplanta-tion (RTx), and date and cause of death.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Australia and New Zealand Dialysis and Transplant Registry, Woodville, South Australia, Aus-tralia; the 2Department of Nephrology, Hospices Civils de Lyon, Centre Hospitalier Lyon Sud, Pierre Benite,France; and the 3Department of Nephrology, The Queen Elizabeth Hospital, Woodville, South Australia,Australia, and University of Adelaide, Adelaide, South Australia, Australia.

Address correspondence and reprint requests to Emmanuel Villar, MD, PhD, ANZDATA Registry, TheQueen Elizabeth Hospital, 28 Woodville Rd., Woodville South, South Australia 5011, Australia. E-mail:[email protected].


0895.Sponsors have not been involved in any way in the study design, data interpretation, and manuscript

editing. The interpretation of reported data are the responsibility of the authors and in no way should be seenas an official interpretation of the ANZDATA Registry.

Abbreviations: ANZ, Australia and New Zealand; ANZDATA, Australia and New Zealand Dialysis andTransplant Registry; eGFR, estimated glomerular filtration rate; ESRD, end-stage renal disease; RRT, renalreplacement therapy; RTx, renal transplantation.






BMI (ratio of weight in kilograms to thesquare of height in meters at commence-ment of RRT) was analyzed in categories:underweight �18 kg/m2, normal weight18–24.9 kg/m2, overweight 25–29.9 kg/m2, and obese � 30 kg/m2. Smoking sta-tus at the start of RRT was categorized asnever, former, or current smoker. eGFRwas determined by the simplified Modifi-cation of Diet in Renal Disease formula(12) in patients who began RRT after 1April 1998 because data on serum creati-nine before the first RRT were collectedafter this date.

When appropriate, univariate com-parisons were performed using a �2 test orFisher’s exact test for categorical vari-ables, Student’s t test for continuous vari-ables between two groups, and ANOVAfor continuous variables across the threegroups by diabetes status. We calculatedage- and sex-standardized ESRD inci-dence rates by diabetes status among ANZpopulations using direct standardization.For 1991, incidence was projected for theentire year. Data on ANZ populationswere provided by the Australian Bureau ofStatistics and Statistics New Zealand. Thereference populations were the 1991–2005 ANZ populations aged �16 years.Calculation of average annual changes inincidence and comparisons between sub-groups were performed by Poisson regres-sion, and we checked for overdispersion.

Times to RTx or to death were exam-ined with Kaplan-Meier models and Coxregression for multivariate analyses. RTxoutcomes were examined in patients aged�70 years. Cox models to analyze varia-tions in access to RTx by diabetes statusper 5-year periods (1991–1995, 1996–2000, and 2001–2005) were adjusted forage, sex, primary renal disease, comor-bidities at the first RRT, BMI categories,and smoking status and were stratified onracial origin, state where RRT was started(seven states in Australia and one in NewZealand), and initial RRT modality.

Causes of death were classified intosudden death, cardiovascular, infection,malignancy, and other causes. In survivalanalyses, death from any cause was theend point. In multivariate survival analy-sis, diabetes status (type 1 diabetic, type 2diabetic, or nondiabetic) was the variableof interest. We also examined the evolu-tion of all-cause and cause-specific mor-tality over 1991–2005 by using the periodof the first RRT (1991–1995, 1996 –2000, and 2001–2005) as the parameterof interest. Models were adjusted for age,sex, primary renal disease, comorbidities

at the first RRT, BMI categories, andsmoking status. eGFR at the start of RRTwas modeled as a fractional polynomialfunction (analyses restricted to patientswho started RRT from 1 April 1998). Coxregression was stratified on racial origingroup, year of the first RRT (1991–2005)with the exception of analysis by period ofthe first RRT, state where RRT was started,initial RRT modality, and RTx during thestudy period. We checked for interactionsbetween variables by including multipli-cative terms in Cox regression. If signifi-cant interactions were found, weperformed stratified survival analysis asdescribed above. Validity of the Cox pro-portional hazard assumption waschecked by tests based on Shoenfeld’s re-siduals. All statistical analyses were per-formed with S-PLUS 6.0 SoftwareProfessional Release 2 (Insightful).

RESULTS

Baseline patient characteristicsType 1 diabetic patients were theyoungest, and type 2 diabetic patientswere the oldest (P � 0.0001) (Table 1).Rates of cardiovascular disease werehigher in diabetic than in nondiabeticpatients (P � 0.0001). Type 2 diabeticpatients had higher average BMI (P �0.0001). The proportion of currentsmokers was higher in type 1 diabeticpatients (P � 0.0001).

Proportions of type 1 and type 2 dia-betes in Caucasoid, in Australian Aborig-inal, and in Maori/Pacific Islanderpatients were 5.3 and 20.9%, 1.5 and70.9%, and 2.6 and 64.1%, respectively(P � 0.0001). Sex ratios (male to female)in these groups were 1.5, 0.76, and 1.25,respectively (P � 0.0001). Average ages atthe first RRT were 58.8 � 16, 49.9 �11.9, and 53.0 � 12.9 years, respectively(P � 0.0001).

ESRD incidence rates by diabetesstatusStandardized incidence rates of ESRDwith associated type 1 diabetes remainedstable over time at about 5 per millionpopulations. Average annual change was�0.3% per year (�1.6 to �0.9%), with-out significant differences between coun-tries, sex, and age (Fig. 1).

Standardized incidence rates of ESRDwith associated type 2 diabetes rose from10.6 per million populations in 1991 to48.8 per million populations in 2005 inAustralia. In New Zealand, they varied be-tween 23.9 per million populations in

1991 and 68.7 per million populations in2002. Across countries, the average an-nual change was �10.2% per year(�9.6–�10.8%). For incidence of ESRDat age �60 years with associated type 2diabetes, the increase was �8.7% (�7.7–�9.7%) in Australia and �5.3% (�3.9–�6.8%) in New Zealand. For ESRD at age�60 years with associated type 2 diabe-tes, the increase was �11.7% (�10.8–�12.6) and �11.5% (�9.7–�13.4%),respectively (P � 0.001 compared withthose for patients aged �60 years of thesame country).

Standardized incidence rates of ESRDwithout diabetes increased significantly(�1.5% [�1.1–�1.8%] in Australia and�2.9% [�2.1–�3.8%] in New Zealand).

RRT modalities on the 90th day andaccess to RTxType 1 diabetic patients were more likelyto be treated by peritoneal dialysis thantype 2 diabetic and nondiabetic patients(P � 0.0001) (Table 1). Type 2 diabeticpatients were less likely to receive RTx(P � 0.0001). Over time, rates of RTxwere stable in type 1 diabetic patients (ad-justed hazard ratio [HR] 1.02 [95% CI0.91–1.15] per 5-year period, P � 0.72)and in nondiabetic patients (1.00 [0.96–1.03], P � 0.84). Adjusted rates of RTxdecreased in type 2 diabetic patients (0.78[0.68–0.90], P � 0.0005), without a dif-ference between sexes.

Crude survival and causes of deathUnadjusted median (95% CI) survivalsfrom the first RRT in type 1 diabetic, type2 diabetic, and nondiabetic patients were72.5 (66.3–82.1), 40.1 (38.8–41.3), and80.2 (77.7–83.0) months, respectively.Median survivals from birth were 55.7(54.4–56.7), 70.5 (70.2–70.9), and 74.7(74.5–74.9) years, respectively (Fig. 2).

Among type 1 diabetic patients, 627(48.8%) died during the study period.Proportions of sudden death, cardiovas-cular, infection, malignancy, and othercause as cause of death were in men andin women 27.2, 40.4, 11.3, 3.3, and17.8% and 18.7, 34.8, 19.5, 2.0, and25.0%, respectively (P � 0.01). Amongtype 2 diabetic patients, 4,997 (58.4%)died. Proportions were 17.9, 42.2,14.7, 4.6, and 20.6% and 15.7, 41.0,15.9, 3.7, and 23.7%, respectively (P �0.01). Among nondiabetic patients,8,393 (44.9%) died. Proportions were14.9, 35.7, 13.4, 11.5, and 24.5% and12.2, 35.4, 15.4, 8.7, and 28.3%, re-spectively (P � 0.0001). Causes of

Villar, Chang, and McDonald


death were significantly different be-tween patient groups by diabetes status(P � 0.0001).

Over time, there was a decrease inadjusted rates of cardiovascular death(adjusted HR 0.96 [95% CI 0.92– 0.99]per 5-year period, P � 0.04), death

from infectious disease (0.89 [0.83–0.95], P � 0.003), and sudden death(0.88 [0.83– 0.94], P � 0.0001),whereas rates of malignancy death in-creased (1.19 [1.08 –1.30] , P �0.0002). These trends were similar inthe three patient groups.

Multivariate survival analysis in thewhole cohort

Multivariate survival analysis showed thatthe risk for death after the first RRT was64% higher in type 1 diabetic (P �0.0001) and 13% higher in type 2 dia-

Table 1—Baseline characteristics and renal replacement therapy in type 1 diabetic, type 2 diabetic, and nondiabetic patients

Type 1 diabetic Type 2 diabetic Nondiabetic P*

n 1,284 (4.5) 8,560 (30.0) 18,704 (65.5)Male 733 (57.1) 4,943 (57.7) 10,934 (58.5) 0.002Age at first RRT (years) 43.1 � 11.3 61.2 � 11.2 56.5 � 17.0 �0.0001Racial origin �0.0001†

Caucasoid 1,136 (88.5) 4,493 (52.5) 15,882 (84.9)Australian Aboriginal 31 (2.4) 1,444 (16.9) 562 (3.0)Maori/Pacific Islander 71 (5.5) 1,784 (20.8) 930 (5.0)Other people 46 (3.6) 839 (9.8) 1,327 (7.1)

Primary renal disease �0.0001†Diabetes 1,205 (93.8) 6,345 (74.1) 0 (0)Renal vascular disease 15 (1.2) 572 (6.7) 3,114 (16.6)Glomerular nephropathy and

related disease36 (2.8) 775 (9.1) 7,699 (41.2)

Polycystic 2 (0.1) 89 (1.0) 1,842 (9.8)Other 26 (2.1) 779 (9.1) 6,049 (32.4)

Biopsy-proven nephropathy 162 (12.6) 1,421 (16.6) 7,032 (37.6) �0.0001Comorbid conditions at first RRT

Chronic lung disease 84 (6.5) 1,496 (17.5) 2,728 (14.6) �0.0001Coronary artery disease 435 (33.9) 4,802 (56.1) 5,550 (29.7) �0.0001Peripheral vascular disease 555 (43.2) 3,694 (43.2) 2,989 (16.0) �0.0001Cerebrovascular disease 153 (11.9) 1,692 (19.8) 2,134 (11.4) �0.0001

BMI (kg/m2) 25.0 � 4.7 28.6 � 6.4 25.2 � 5.3 �0.0001�18 29 (2.3) 128 (1.5) 844 (4.5) �0.0001b

18–24 727 (56.6) 2,574 (30.1) 9,633 (51.5)25–29 368 (28.7) 2,878 (33.6) 5,495 (29.4)�30 160 (12.5) 2,980 (34.8) 2,832 (15.1)

Cigarette smokingNever 676 (52.6) 3,725 (43.5) 9,135 (48.8) �0.0001†Former 384 (29.9) 3,720 (43.5) 7,131 (38.1)Current 224 (17.5) 1,115 (13.0) 2,438 (13.1)

Serum creatinine at first RRT(�mol/l)‡

686 � 263 735 � 306 795 � 339 �0.0001

eGFR at first RRT (ml/min)‡§ 8.5 � 3.8 7.5 � 4.0 7.0 � 3.6 �0.000190-day RRT modality �0.0001†

Haemodialysis 531 (41.3) 4,971 (58.1) 10,860 (58.1)Peritoneal dialysis 639 (49.8) 3,554 (41.5) 6,992 (37.4)Renal transplantation 114 (8.9) 35 (0.4) 852 (4.5)

Details of RTxn 1,257� 6,551� 10,860�Waiting list registration 522 (41.5) 724 (11.1) 5,069 (36.7) �0.0001Preemptive renal transplantation 85 (6.8)¶ 18 (0.3) 502 (3.6) �0.0001Living donor renal transplantation 89 (7.1)# 111 (1.3) 2,024 (14.6) �0.0001Cadaveric renal transplantation 436 (34.7)** 340 (5.2) 3,638 (26.3) �0.0001Median times to RTx (months) 18.3 (16.7–20.9) 48.8 (45.7–55.9) 26.0 (24.9–26.9) �0.0001

Data are n (%), mean � SE, or median (95% CI). *Comparisons across the three groups. †Comparisons in categorical variables (racial origin, primary renal disease,BMI categories, cigarette smoking status, 90-day RRT modality). ‡Analysis restricted to patients who started RRT after 1 April 1998: n � 17,809; type 1 diabetic, n �694; type 2 diabetic, n � 6,176; nondiabetic, n � 10;939; for conversion to milligrams per deciliter divide by 88.4. §Estimated by the simplified Modification Dietin Renal Disease formula (12). �Analyses restricted to patients aged �70 years. ¶Including 15 living donor renal transplantations, 5 single cadaveric renaltransplantations, and 65 simultaneous kidney-pancreas transplantations. #Including 15 preemptive renal transplantations. **Including 159 single renal transplan-tations and 277 simultaneous kidney-pancreas transplantations.

ESRD in type 1 and type 2 diabetic patients


betic (P � 0.0001) than in nondiabeticpatients (Table 2).

Multivariate survival analysis bydiabetes statusThere was a significant interaction be-tween sex and diabetes status (P �

0.0004). Female sex was significantly as-sociated with higher risk for death in type2 diabetic patients (adjusted HR for deathin women versus men 1.08 [95% CI1.015–1.16], P � 0.02). Sex was not as-sociated with survival in type 1 diabetic(1.12 [0.87–1.46], P � 0.38) and in non-

diabetic patients (0.95 [0.91–1.005], P �0.07).

In type 2 diabetes, there was a signif-icant interaction between sex and age(P � 0.0001). The adjusted HR for deathin women versus men was 0.93 (95% CI0.83–1.04) (P � 0.20) in type 2 diabeticpatients aged �60 years (n � 3,762) and1.19 (1.08–1.30) (P � 0.0003) in type 2diabetic patients aged �60 years (n �4,798). This last adjusted HR was similarfor cardiovascular and noncardiovascularcauses of death.

No other significant interactions werefound with race, cause of ESRD (diabeticnephropathy versus other causes ofESRD), and BMI. Results were unchangedwhen follow-up was censored at the timeof transplant and/or RRT modalityswitches and when analyses were ad-justed for eGFR.

In type 1 diabetic patients, survivaldid not change over time (adjusted HR0.94 [0.83–1.07] per 5-year period, P �0.36), whereas it significantly improvedby 9% per 5-year period in type 2 diabeticpatients (0.91 [0.87–0.95], P � 0.0001)and by 5% in nondiabetic patients (0.95[0.92–0.98], P � 0.001).

CONCLUSIONS — This study inANZ showed a large increase in the inci-dence rate of ESRD with associated type 2diabetes from 1991 to 2005, which wasespecially marked in type 2 diabetic pa-tients aged �60 years (�11.5% per year).The incidence of ESRD with associatedtype 1 diabetes remained stable. After ad-justment for age, sex, and risk factors fordeath, type 1 diabetes had a greater effecton survival in patients with ESRD than intype 2 diabetic patients compared withnondiabetic patients. In each patientgroup, the proportions of cardiovascular,infection, and sudden death decreasedover the study period, whereas rates ofmalignancy death increased. Female sexwas associated with worse outcome thanmale sex in type 2 diabetic patients aged�60 years. This difference did not appearto be explained by the different comorbidconditions, age, race, causes of ESRD,BMI at first RRT, or RRT modalities.

The strength of this analysis is thattype 1 diabetes and type 2 diabetes areseparately reported in a prospective andpopulation-based study. Previous analy-ses may have been biased because theyonly included patients with diabetic ne-phropathy and because nephropathy mayhave been misclassified if it was not bi-opsy proven.

Figure 1—Age- and sex-standardized ESRD incidence per million population (aged �16 years)by diabetes status among the general population in Australia (A, n � 23,417) and in New Zealand(B, n � 5,131). -�-, patients with type 1 diabetes; -f-, patients with type 2 diabetes; -Œ-, patientswithout diabetes.



Despite an increase of about �3% peryear in the incidence of childhood type 1diabetes in ANZ during the last decades(13,14), the incidence of RRT with asso-ciated type 1 diabetes remained stable be-tween 1991 and 2005. The difference intrends between general and ESRD popu-lations may indicate improvements incare of type 1 diabetic patients due totreatment with ACE inhibitors and ag-gressive glycemic control available since1980 (15). High transplant rates, includ-ing simultaneous kidney-pancreas trans-plant, remained stable over time. Thehigher risk for death in type 1 diabeticthan in type 2 diabetic patients was notexplained by risk factors in the multivar-iate analyses. This difference should beaccounted for by differences in diabetesduration and severity or glycemic control.These data were not available for analysis,and the result should be interpreted withthis limitation in mind.

For type 2 diabetes, the overall 10.2%annual increase in ANZ is consistent withstudies in Europe and the U.S. over com-parable periods (6,7). The increase washigher in patients aged �60 years than inyounger patients. Possible explanationsfor this rise are the increasing incidenceand prevalence of overweight, obesity(16), and type 2 diabetes in the generalpopulation (17); improved life expect-ancy in type 2 diabetic patients with ear-lier stage of kidney disease due in part tobetter management of cardiovascular dis-eases (18); and greater access to RRT(5–7).

These results highlighted the specificepidemiology of diabetes and ESRD in theAustralasian population. Two-thirds ofAustralian Aboriginal and Maori/PacificIslander patients with ESRD had type 2diabetes at the start of RRT, which wassignificantly different from the situationin the Caucasoid population (�20% withtype 2 diabetes at the start of RRT). Theincidence and prevalence of type 2 diabe-tes and hypertension are high in Aborigi-nal population (19). This higherincidence of ESRD with associated type 2diabetes may be explained in part by ge-netic susceptibility and higher rates ofkidney disease progression than in theCaucasoid population (19).

After the first RRT, overall survivalwas short in type 2 diabetic patients, withmedian survival times of �3.5 years, sim-ilar to reports from Europe (8,11) and theU.S. (9,12). Less than 10% of type 2 dia-betic patients received RTx, as in France(20) and the U.S. (21). Adjusted rates of

Figure 2—Survival curve for national cohorts after birth by diabetes status, computed for mor-tality rates for the period 1991–2005.

Table 2—Adjusted HR of death of any cause in the whole cohort, patients not censored at renalreplacement modality switches or renal transplantation

HR (95% CI) P

Diabetes statusPatients without diabetes* 1Patients with type 1 diabetes 1.64 (1.47–1.84) �0.0001Patients with type 2 diabetes 1.13 (1.06–1.20) �0.0001Male versus female 1.0 (0.96–1.04) 0.89Age at first RRT (�1 year) 1.024 (1.022–1.026) �0.0001

Primary renal diseaseDiabetes 1.21 (1.12–1.31) �0.0001Renal vascular disease 1.10 (1.04–1.17) 0.002Glomerular nephropathy and related

disease*1

Polycystic 0.76 (0.69–0.83) �0.0001Myeloma, light chain deposit, and amyloid 3.0 (2.72–3.32) �0.0001Renal cancer 1.67 (1.4–2.0) �0.0001Other 1.07 (1.02–1.13) 0.01

Lung disease 1.24 (1.18–1.29) �0.0001Coronaropathy 1.22 (1.17–1.27) �0.0001Peripheral vascular disease 1.21 (1.15–1.26) �0.0001Cerebrovascular disease 1.16 (1.11–1.22) �0.0001BMI (kg/m2)

�18 1.33 (1.21–1.45) �0.000118–24* 125–29 0.89 (0.86–0.93) �0.0001�30 0.91 (0.87–0.96) 0.0005

Cigarette smokingNever* 1Former 0.99 (0.95–1.03) 0.72Current 1.10 (1.04–1.17) 0.001

Whole cohort: n � 28,548. Results were unchanged when patients were censored at time of transplantand/or RRT modality switches, when analyses were adjusted for eGFR, or when analyses were performedonly in patients starting RRT with hemodialysis or in patients starting with peritoneal dialysis. *Referencegroup in categorical variables.



RTx declined over the study periodamong type 2 diabetic patients but re-mained stable in the other two groups.Survival rates improved with a decrease incardiovascular death. We hypothesizethat improvements in dialysis manage-ment and in cardiovascular treatmentsmay explain this improvement over time.

Moreover, female sex was signifi-cantly associated with death in type 2diabetic patients aged �60 years. Interac-tions between female sex, diabetes, andexcess mortality in the ESRD populationcompared with the general populationhave also been noted in France (22). Sev-eral dialysis-specific explanations can beproposed, such as sex differences in theeffects of the dialysis dose (23) and theimportance of glycemic control (12)on survival of patient with ESRD anddiabetes.

Although it remains controversial(24), worse prognosis has also been re-ported in women than in men in the non-ESRD diabetic population who do nothave diabetes (3,4). In diabetic subjectswithout chronic kidney disease, moststudies have found that this differencewas not accounted for by traditional riskfactors (25). Higher risk for death inwomen may be related to interactions be-tween cardiovascular risk factors andmenopause (26), a stronger inverse asso-ciation between coronary disease andcholesterol level in women, and differ-ences in coagulation and in patterns ofobesity and hyperinsulinemia (2–4,25,26).

In summary, this study confirms thatincidences, treatments, and survivals aredifferent between ESRD patients withtype 1 and type 2 diabetes. Future studiesof patients with ESRD and diabetesshould differentiate between these twogroups to provide interpretable results.ESRD remains a dreadful complication inpatients with type 1 diabetes, and greateffort to prevent kidney disease in theseyoung patients is needed. A marked in-crease in the incidence rate of ESRD withassociated type 2 diabetes was seen overthe study period. The study emphasizesthe burden of ESRD with associated type2 diabetes in Australian Aboriginal and inMaori/Pacific Islander populations. Pre-vention of renal impairment (27), ne-phroprotection in patients with overtnephropathy, early referral to nephrolo-gists (28), and access to RTx (29) mayimprove the prognosis of type 2 diabeticpatients. This study also highlights thepoorer prognosis in older type 2 diabetic

women compared with older type 2 dia-betic men. This finding deserves furtherexplanatory studies.

Acknowledgments— We acknowledge allregistry participants, especially the nephrolo-gists and professionals who collected the dataand conducted the quality control studies. TheANZDATA Registry is funded by the Austra-lian Government Department of Health ofAgeing, by the New Zealand Ministry ofHealth, and by Kidney Health Australia. TheRegistry has also received contributions fromvarious pharmaceutical and dialysis compa-nies on an unrestricted basis. E.V. is supportedby research grants from the Hospices Civils deLyon and from Novartis and Roche.

References1. Gu K, Cowie CC, Harris MI: Mortality in

adults with and without diabetes in a na-tional cohort of the U.S. population,1971–1993. Diabetes Care 21:1138–1145, 1998

2. Soedamah-Muthu SS, Fuller JH, MulnierHE, Raleigh VS, Lawrenson RA, ColhounHM: All-cause mortality rates in patientswith type 1 diabetes mellitus comparedwith a non-diabetic population from UKgeneral practice research database, 1992–1999. Diabetologia 49:660–669, 2006

3. Hu G, Jousilahti P, Qiao Q, Katoh S,Tuomilehto J: Sex differences in cardio-vascular and total mortality among dia-betic and non-diabetic individuals with orwithout history of myocardial infarction.Diabetologia 48:856–861, 2005

4. Juutilainen A, Kortelainen S, Lehto S,Ronnemaa T, Pyorala K, Laakso M: Gen-der difference in the impact of type 2 di-abetes on coronary heart disease risk.Diabetes Care 27:2898–2904, 2004

5. ANZDATA Registry Report 2006. Australiaand New Zealand Dialysis and TransplantRegistry. Adelaide, Australia.

6. van Dijk PC, Jager KJ, Stengel B, Gronha-gen-Riska C, Feest TG, Briggs JD: Renalreplacement therapy for diabetic end-stage renal disease: data from 10 registriesin Europe (1991–2000). Kidney Int67:1489–1499, 2005

7. U.S. Renal Data System: USDRS 2006 An-nual Data Report: Atlas of End-Stage RenalDisease in the United States. National Insti-tutes of Health, National Institute of Dia-betes and Digestive and Kidney Diseases,Bethesda, MD, 2006

8. Sorensen VR, Mathiesen ER, Heaf J, Feldt-Rasmussen B: Improved survival rate inpatients with diabetes and end-stage renaldisease in Denmark. Diabetologia 50:922–929, 2007

9. Locatelli F, Pozzoni P, Del Vecchio L: Re-nal replacement therapy in patients withdiabetes and end-stage renal disease. J AmSoc Nephrol 15:S25–S29, 2004

10. Williams ME, Lacson EJ, Teng M, Ofst-hun N, Lazarus JM: Hemodialyzed type Iand type II diabetic patients in the US:characteristics, glycemic control, and sur-vival. Kidney Int 70:1503–1510, 2006

11. Biesenbach G, Hubman R, Grafinger P,Stuby U, Eichbauer-Sturm G, Janko O:5-Year overall survival rates of uremictype 1 and type 2 diabetic patients incomparison with age-matched non-dia-betic patients with end-stage renal dis-ease from a single dialysis center from1991 to 1997. Diabetes Care 12:1860 –1862, 2000

12. Levey AS, Bosch JP, Lewis JB, Greene T,Rogers N, Roth D: A more accuratemethod to estimate glomerular filtrationrate from serum creatinine: a new predic-tion equation. Modification of Diet in Re-nal Disease Study Group. Ann Intern Med130:461–470, 1999

13. Haynes A, Bower C, Bulsara MK, JonesTW, Davis EA: Continued increase in theincidence of childhood type 1 diabetes ina population-based Australian sample(1985–2002). Diabetologia 47:866–870,2004

14. Campbell-Stokes PL, Taylor BJ, the NewZealand Children’s Diabetes WorkingGroup: Prospective incidence of diabetesmellitus in New Zealand children aged 0to 14 years. Diabetologia 48:643–648,2005

15. Astrup AS, Tarnow L, Rossing P, Pi-etraszek L, Hansen PR, Parving HH: Im-proved prognosis in type 1 diabeticpatients with nephropathy. Kidney Int 68:1250–1257, 2005

16. Allman-Farinelli MA, Chey T, BaumanAE, Gill T, James WP: Age, period andbirth cohort effect on prevalence of over-weight and obesity in Australian adultsfrom 1990 to 2000. Eur J Clin Nutr (Epubahead of print)

17. Wild S, Roglic G, Green A, Sicree R, KingH: Global prevalence of diabetes: esti-mates for the year 2000 and projectionsfor 2030. Diabetes Care 27:1047–1053,2004

18. Lipscombe LL, Hux JE: Trends in diabetesprevalence, incidence, and mortality inOntario, Canada 1995–2005: a popula-tion-based study. Lancet 369:750–756,2007

19. Wang Z, Hoy WE: Hypertension, dyslip-idemia, body mass index, diabetes andsmoking status in Aboriginal Australiansin a remote community. Ethn Dis 13:324–330, 2003

20. Villar E, Rabilloud M, Berthoux P, VialtelP, Labeeuw M, Pouteil-Noble C: A multi-centre study of registration on renal trans-plantation waiting list of the elderly andpatients with type 2 diabetes. Nephrol DialTransplant 19:207–214, 2004

21. Wolfe RA, Ashby VB, Milford EL, Bloem-bergen WE, Agodoa LY, Held PJ, Port FK:



Differences in access to cadaveric renaltransplantation in the United States. Am JKidney Dis 36:1025–1033, 2000

22. Villar E, Remontet L, Labeeuw M, Eco-chard R: Effects of age, sex and diabeteson excess death in end-stage renal diseasepopulation. J Am Soc Nephrol 18:2125–2134, 2007

23. Depmer T, Daugirdas J, Greene T, AllonM, Beck G, Chumlea C, Delmez J, GotchF, Kusek J, Levin N, Macin E, Milford E,Owen W, Star R, Toto R, Eknoyan G: Di-alysis dose and the effect of gender andbody size on outcome in HEMO Study.Kidney Int 66:1386–1394, 2004

24. Kanaya AM, Grady D, Barrett-Connor E:Explaining the sex difference in coronary

heart disease mortality among patientswith type 2 diabetes. Arch Intern Med 162:1737–1745, 2002

25. Lee WL, Cheung AM, Cape D, Zinman B:Impact of diabetes on coronary artery dis-ease in women and men: a meta-analysisof prospective studies. Diabetes Care 23:962–968, 2000

26. Kok HS, Van Asselt KM, Van der SchouwYT, Van der Tweel I, Peeters PHM, WilsonPWF, Learson PL, Grobbee DE: Heart dis-ease risk determines menopausal agerather than the reverse. J Am Coll Cardiol47:1976–1983, 2006

27. Gillies CL, Abrams KR, Lambert PC, Coo-per NJ, Sutton AJ, Hsu RT, Khunti K:Pharmacological and lifestyle interven-

tions to prevent or delay type 2 diabetes inpeople with impaired glucose tolerance:systematic review and meta-analysis. BMJ334:299–317, 2007

28. Frimat L, Loos-Ayav C, Panescu V, Cord-ebar N, Briancon S, and Kessler M: Earlyreferral to a nephrologist is associatedwith better outcomes in type 2 diabetespatients with end-stage renal disease. Di-abetes Metab 30:67–74, 2004

29. Wolfe R, Ashby W, Milford E, Ojo AO,Ettenger RE, Agodoa LY, Held PJ, Port FK:Comparison of mortality in all patients ondialysis, patients on dialysis awaitingtransplantation, and recipients of a firstcadaveric transplant. N Engl J Med 341:1725–1730, 1999



Microvascular and C-Fiber Function inDiabetic Charcot Neuroarthropathy andDiabetic Peripheral NeuropathyNEIL BAKER, BSC, DPODM

ALISTAIR GREEN, MRCP

SINGHAN KRISHNAN, MRCP

GERRY RAYMAN, MD, FRCP

OBJECTIVE — Sympathetic denervation and hyperemia are implicated in the pathogenesisof Charcot neuroarthropathy (CN) but are also features of diabetic peripheral neuropathy (DPN).Differences in these physiological parameters were sought by determining C-fiber function (laserDoppler imager [LDI]flare technique) and maximum microvascular hyperemia (MMH) in 13subjects with diabetic CN (DCN), 10 subjects with DPN, and 10 healthy control subjects.Additionally, unaffected limbs of the nine DCN subjects with unilateral CN (UCN) were studiedto determine whether any observed differences precede CN.

RESULTS — LDIflare area was reduced in DPN (mean � SD 1.41 � 0.51 cm2) and DCN(1.42 � 0.37) groups compared with the healthy control group (5.24 � 1.33; P � 0.0001).MMH was higher in DCN (432 � 88 PU [perfusion units]) than in DPN (262 � 71; P � 0.001)subjects but lower than in the control group (564 � 112; P � 0.01). LDIflare area and MMHwere similar in the UCN and DCN groups.

CONCLUSIONS — C-fiber function is equally impaired in neuropathic patients with andwithout CN; however, a higher MMH distinguishes those with CN. Unaffected and affected limbsof those with unilateral CN have the same neurovascular abnormalities, suggesting that theseabnormalities precede CN and are not a result of CN.

Diabetes Care 30:3077–3079, 2007

P eripheral sensory neuropathy andautonomic dysfunction are acceptedprerequisites for the development of

Charcot neuroarthropathy (CN) but arealso features of diabetic peripheral neu-ropathy (DPN) (1,2). CN is rare in com-parison with DPN, suggesting thatadditional factors are involved in itspathogenesis. Small-fiber neuropathy,measured with quantitative sensory test-ing, has been implicated in its develop-ment (3,4). Moreover, a relatively highermaximum microvascular hyperemia(MMH) has been reported (3,4); however,whether this is a result of CN or preexist-ing is unknown. This study examinesthese features in greater detail. Small-fiber

neuropathy was assessed using the laserDoppler imager (LDI)flare technique (5),a more sensitive test of small-fiber func-tion than quantitative sensory testing.MMH was assessed using the LDImaxtechnique (5). The unaffected foot inthose with CN was also studied to deter-mine whether any defects in these mea-sures were preexisting, and thereforeetiological, or consequential of CN.

RESEARCH DESIGN ANDMETHODS — Four matched groupswere studied: the DPN group, 10 subjectswith type 2 diabetes and neuropathy(aged 67.2 � 7.1 years, diabetes duration19 � 8.1 years, vibration perception

threshold [VPT] 30.3 � 6.0 V); the DCNgroup, 13 subjects with type 2 diabetesand quiescent CN (aged 65.5 � 8.7 years,diabetes duration 20 � 11.3 years, VPT36.1 � 9.7 V) (4 with bilateral and 9 withunilateral CN); the unilateral CN (UCN)group, 9 subjects with UCN from theDCN group in whom the unaffected limbwas studied (aged 64.7 � 10.2 years, di-abetes duration 21 � 10.2 years, and VPT33.5 � 8.1 V); and the control group, 10healthy subjects (aged 61.4 � 9.7 years,VPT 8.0 � 2.1 V).

Neuropathy was present if two ormore of four sites on the plantar foot wereinsensate to 10-g monofilaments and ifthe VPT at the hallux was �24 V (Neu-rothesiometer; Horwell Scientific, Not-tingham, U.K.).

CN was determined by clinical andradiological examination. All affectedjoints had been quiescent (�2°C differ-ence between limbs) and ulcer free forover 18 months.

The LDIflare and LDImax were as-sessed using an LDI from Moor Instru-ments (Devon, U.K.). These methodshave been validated and are described indetail elsewhere (5). Briefly, after acclima-tization, a baseline scan was performed ona 7.5 � 4 cm area on the dorsum of thefoot using the LDI. The skin was thenheated to 44°C for 20 min using a 0.64-cm2 circular skin heater and thenrescanned immediately after its removal.Heating induces MMH (LDImax) under-neath the heater but also hyperemia in thesurrounding skin due to axon-reflex–mediated vasodilatation (LDIflare). Fromthe computer-generated flux images, theLDIflare area (centimeters squared) andthe LDImax (PU [perfusion units]) are de-rived. The coefficients of variation for theLDIflare and LDlmax were 6.8 and 6.4%,respectively. Variables from the groupswere compared using one-way ANOVAand Tukey tests.

RESULTS — Al l sub j ec t s werematched for age and sex and those withdiabetes for duration and A1C.

LDIflares were markedly reduced inall diabetic groups compared with thecontrol group (P � 0.0001 for each

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the Ipswich Diabetic Foot Unit and Diabetes Centre, Suffolk, U.K.Address correspondence and reprint requests to Dr. G. Rayman, MD, FRCP, The Ipswich Diabetes Centre,

Ipswich Hospital NHS Trust, Suffolk, U.K. E-mail: [email protected] for publication 5 June 2007 and accepted in revised form 24 August 2007.Published ahead of print at http://care.diabetesjournals.org on 5 September 2007. DOI: 10.2337/dc07-

1063.Abbreviations: CN, Charcot neuroarthropathy; DCN, diabetic CN; DPN, diabetic peripheral neuropa-

thy; LDI, laser Doppler imager; MMH, maximum microvascular hyperemia; NF-�B, nuclear factor-�B;RANKL, receptor activator of the NF-�B ligand; UCN, unilateral CN; VPT, vibration perception threshold.




P a t h o p h y s i o l o g y / C o m p l i c a t i o n sB R I E F R E P O R T


group) (Fig. 1). In the UCN patients,there was no difference in LDIflare areabetween the unaffected (1.14 � 0.51cm2) and the affected (1.42 � 0.37 cm2)limbs. LDImax was also markedly im-paired in the DPN group (262 � 71 PU)compared with the control group (594 �94, P � 0.0001) (Fig. 1). In contrast to theLDIflare findings, LDImax in the DCN(432 � 88) and UCN (417 � 110) groupswas significantly greater than in the DPNgroup (262 � 71; P � 0.001 and P �0.01, respectively) but lower than in thecontrol group (594 � 110, both P �0.01). Finally, there was no difference inthe LDImax between the unaffected(417 � 110) and affected (432 � 88)limbs in the UCN group.

CONCLUSIONS — The principalfindings were as follows: 1) C-fiber func-tion, as assessed by the LDIflare tech-nique, is severely impaired in CN andindistinguishable from DPN alone; 2)MMH is relatively preserved in CN andsignificantly higher than in neuropathyalone; and 3) affected and unaffectedlimbs of patients with CN have similarC-fiber dysfunction and MMH.

The reduced MMH in the DPN groupis not unexpected, having been describedin a variety of diabetic states includingimpaired glucose tolerance (6–8). Whatis surprising is the relative preservation ofMMH in the DCN groups, as this wouldbe expected to be worse or similar to, butnot significantly better than, the relatively

less complicated DPN group. The latterfindings are supported by other studies(13,14).

As first suggested by Charcot (9,10),bone resorption as a result of increasedbone perfusion secondary to the sympa-thetic denervation may be implicated inthe development of CN (11). Preservationof the MMH in CN is consistent with hy-peremia being involved. In contrast, inthose with DPN without CN, the ob-served lower hyperemic responses may beprotective.

Bone dissolution, which is the hall-mark of the condition, is dependent uponosteoclastic activativation by a system ofcytokines, the receptor activator of thenuclear factor-�B (NF-�B) ligand(RANKL)–NF-�B system (RANKL–NF-�B) (10,12,13). This system is activated indiabetes (14–16) and inhibited by neu-ropeptides (13). Thus, diabetic neuropa-thy may favor RANKL–NF-�B systemactivation and may lead to protracted in-flammation in CN (10).

Heat-induced vasodilatation is thoughtto be proportional to the expression ofNO synthase (4,17), and RANKL–NF-�Bactivation increases the production of in-ducible NO synthase (18).

Protracted inflammation in combina-tion with the absence of modifying sym-pathetic tone and neuropeptides may leadto unrestrained and prolonged hyperemicbone perfusion, contributing to the bonedissolution in CN. The relatively highMMH seen in the skin of those with CNwould support the latter suggestion of hy-peremic bone blood flow.

The finding of high MMH in affectedand unaffected limbs of those with CNsuggests that this abnormality is preexist-ing. A high MMH may thus be implicatedin the development of CN rather than sec-ondary to changes in the local microcir-culation as a consequence of CN.

This study supports the suggestionthat preserved MMH is a prerequisite forthe development of CN (3,4). Under-standing why vascular reactivity is retainedmay be important in discovering thecause and identifying treatments for CN.

References1. Jeffcoate W, Lima J, Nobrega L: The Char-

cot foot. Diabet Med 17:253–258, 20002. Rajbhandari SM, Jenkins RC, Davies C,

Tesfaye S: Charcot neuroarthropathy indiabetes mellitus. Diabetologia 45:1085–1096, 2002

3. Shapiro SA, Stansberry KB, Hill MA,Meyer MD, McNitt PM, Bhatt BA, Vinik

Figure 1—A: C-fiber function measured by the LDIflare technique. B: Maximum microvascularhyperemic response measured using LDImax technique. HC, healthy control group.

Nerve and microvascular function in Charcot and DPN


AI: Normal blood flow response and vaso-motion in the diabetic Charcot foot. J Di-abetes Complications 12:147–153, 1998

4. Veves A, Akbari CM, Primavera J, Dona-ghue VM, Zacharoulis D, Chrzan JS, De-Girolami U, LoGerfo FW, Freeman R:Endothelial dysfunction and the expres-sion of endothelial nitric oxide synthetasein diabetic neuropathy, vascular disease,and foot ulceration. Diabetes 47:457–463,1998

5. Krishnan ST, Rayman G: The LDIflare: anovel test of C-fiber function demon-strates early neuropathy in type 2 diabe-tes. Diabetes Care 27:2930–2935, 2004

6. Jaap AJ, Hammersley MS, Shore AC,Tooke JE: Reduced microvascular hype-raemia in subjects at risk of developingtype 2 (non-insulin-dependent) diabetesmellitus. Diabetologia 37:214–216, 1994

7. Sandeman DD, Pym CA, Green EM, Sea-mark C, Shore AC, Tooke JE: Microvascu-lar vasodilatation in feet of newlydiagnosed non-insulin dependent dia-betic patients. BMJ 302:1122–1123, 1991

8. Wilson SB, Jennings PE, Belch JJ: Detec-tion of microvascular impairment in typeI diabetics by laser Doppler flowmetry.Clin Physiol 12:195–208, 1992

9. Charcot JM: Sur Quelques Arthropathies quiParaissent Dependre d’une Lesion duCerveau ou de la Moelle Epiniere. Archivesde Physiologie Normale et Pathologique,Paris, 1868, p. 161–178

10. Jeffcoate WJ, Game F, Cavanagh PR: Therole of proinflammatory cytokines in thecause of neuropathic osteoarthropathy(acute Charcot foot) in diabetes. Lancet366:2058–2061, 2005

11. Duncan CP, Shim SS: The J Edouard Sam-son Address: The autonomic nerve supplyof bone: an experimental study of the in-traosseous adrenergic nervi vasorum inthe rabbit. J Bone Joint Surg Br 59:323–330, 1977

12. Hofbauer LC, Heufelder AE: The role ofreceptor activator of nuclear factor-kap-paB ligand and osteoprotegerin in thepathogenesis and treatment of metabolicbone diseases. J Clin Endocrinol Metab 85:2355–2363, 2000

13. Jeffcoate W: Vascular calcification and os-teolysis in diabetic neuropathy: isRANK-L the missing link? Diabetologia47:1488–1492, 2004

14. Jiang MZ, Tsukahara H, Ohshima Y,Todoroki Y, Hiraoka M, Maeda M,Mayumi M: Effects of antioxidants and ni-

tric oxide on TNF-alpha-induced adhe-sion molecule expression and NF-kappaBactivation in human dermal microvascu-lar endothelial cells. Life Sci 75:1159–1170, 2004

15. Yerneni KK, Bai W, Khan BV, MedfordRM, Natarajan R: Hyperglycemia-in-duced activation of nuclear transcriptionfactor �B in vascular smooth muscle cells.Diabetes 48:855–864, 1999

16. Bierhaus A, Schiekofer S, Schwaninger M,Andrassy M, Humpert PM, Chen J, HongM, Luther T, Henle T, Kloting I, MorcosM, Hofmann M, Tritschler H, Weigle B,Kasper M, Smith M, Perry G, SchmidtAM, Stern DM, Haring HU, Schleicher E,Nawroth PP: Diabetes-associated sus-tained activation of the transcription fac-tor nuclear factor-�B. Diabetes 50:2792–2808, 2001

17. Gooding KM, Hannemann MM, TookeJE, Clough GF, Shore AC: Maximum skinhyperaemia induced by local heating:possible mechanisms. J Vasc Res 43:270–277, 2006

18. Ahn KS, Aggarwal BB: Transcription fac-tor NF-kappaB: a sensor for smoke andstress signals. Ann N Y Acad Sci 1056:218–233, 2005

N. Baker and Associates


C-Reactive Protein in Diabetic andNondiabetic Patients With AcuteMyocardial InfarctionWOLFGANG OTTER, MD

1,2

MICHAEL WINTER, MD3

WITTICH DOERING, MD1

EBERHARD STANDL, MD3

OLIVER SCHNELL, MD3

A therosclerosis has been reported tobe associated with chronic low-grade inflammation of the vasculary

structure and the endothelial cells (1–4).C-reactive-protein (CRP) is a marker forinflammation and is enhanced in bothatherosclerosis and coronary artery dis-ease (5–7). CRP plasma levels above thecutoff of 3 mg/l, as assessed with high-sensitivity immunoassays, have beenshown to indicate an increase in cardio-vascular risk (8). Diabetes is an indepen-dent risk factor of atherosclerosis (9,10).It is considered a state of low-grade in-flammation (11–13). CRP levels havebeen reported to be augmented in dia-betic patients (11–13). The Munich Myo-cardial Infarction Registry analyzes theoutcome of hospital mortality in both di-abetic and nondiabetic subjects (13,14).The present study aimed at determiningthe role of CRP in patients with myocar-dial infarction and comparing the resultsbetween diabetic and nondiabetic patients.

RESEARCH DESIGN ANDMETHODS — All patients of the Mu-nich Myocardial Infarction Registry(2001–2004, n � 1,237) were includedin the analysis. Myocardial infarction wasdefined and treated according to the rec-ommendations of the European Society ofCardiology and the American College ofCardiology (15,16–18).

The presence of diabetes was definedif the patient had been informed of this

diagnosis or was on prescribed anti-diabetes treatment. Patients without diag-nosis but with blood glucose �200 mg/dl(3) were also classified as having diabetes(14). CRP was measured on admis-sion and analyzed by a highly sensitiveCRP-Assay, (Roche Diagnostics, Basel,Switzerland).

StatisticsGroup comparisons were performed byMann-Whitney U test for continuousvariables and �2 test for categorical vari-ables. Crude odds ratios (ORs) and 95%CIs were adjusted for age, sex, renal fail-ure, and diabetes. Multiple logistic regres-sion analysis was performed as binarylogistic regression analysis after transfor-mation of the non-normally distributedCRP levels into quintiles and dichotomi-zation of age.

RESULTS — In the entire group of pa-tients (n � 1,237), mean age was 68 � 13years. Thirty-seven percent presentedwith previously known coronary arterydisease and 25% with a history of previ-ous myocardial infarction. Sixty-four per-cent presented with hypertension, and28% had an impaired kidney function.Total hospital mortality of the entiregroup of patients was 15.9% (n � 210).There were 479 patients (38.7%) whopresented with diabetes.

In the entire group of patients withacute myocardial infarction, the median

(25th–75th percentile) CRP on admissionwas 7 mg/l (3–25). Glucose levels on ad-mission were not significantly differentbetween patients with CRP levels equal toor below the median compared with thosewith CRP levels above the median (166 �68 vs. 175 � 86 mg/dl, respectively). Inpatients with CRP levels above the me-dian, hospital mortality was higher com-pared with that for patients with CRPlevels equal to or below the median (20vs. 9%, respectively; P � 0.001). In pa-tients who died in the hospital (n � 210),median CRP levels were higher (22 mg/l)compared with those in patients who sur-vived (6 mg/l; P � 0.001).

After multiple correction for age,presence of diabetes, impairment of kid-ney function, hypertension, presence ofreinfarction, and known peripheral arte-rial disease, CRP levels remained an inde-pendent predictor of hospital mortality inpatients with acute myocardial infarction(highest vs. lowest quintile of CRP levels:OR [95% CI] 4.66 [2.28 –9.53]; P �0.001).

CRP plasma levels on admission werehigher in diabetic than in nondiabetic pa-tients: median (25th–75th percentile) 8mg/l (3–36) vs. 6 mg/l (3–20) (P �0.001). The prevalence of diabetes rosewith the level of elevation of CRP plasmalevels (1st quintile [2 mg/l], 38%; 2nd[3–4 mg/l], 30%; 3rd [5–9 mg/l], 35%;4th [10–40 mg/l], 43%; and 5th [�40mg/l], 46%; P for trend �0.01).

Diabetic patients who died in the hos-pital presented with higher CRP plasmalevels on admission compared with thosepresented by diabetic patients who sur-vived: median (25th–75th percentile) 23mg/l (6–77) vs. 7 mg/l (2–26), respec-tively; P � 0.001. Nondiabetic patientswho died in the hospital also presentedwith higher CRP levels compared withthose in patients who survived: 16 mg/l(5–98) vs. 5 mg/l (2–17); P � 0.001.

Hospital mortality with regard to CRPquintiles in diabetic and nondiabetic pa-tients are displayed in Fig. 1. In both di-abetic and nondiabetic patients, CRPlevels remained a significant predictor forhospital mortality after correction for age,kidney dysfunction, hypertension, reinfarc-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From 1Cardiology, Academical Teaching Hospital, Schwabing, Munich, Germany; the 2Center of InternalMedicine, Unterschleissheim/Munich, Germany; and the 3Diabetes Research Institute, Munich, Germany.

Address correspondence and reprint requests to Wolfgang Otter, MD, Center of Internal Medicine,Unterschleissheim/Munich, Rathausplatz 2, 85716 Unterschleissheim/Munich, Germany. E-mail:[email protected].


1020.Abbreviations: CRP, C-reactive protein.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





tion, and known peripheral arterial disease:for the highest versus lowest quintiles in di-abetic patients, OR (95% CI) 7.15 (2.02–25.30) (P � 0.01); in nondiabetic patients,3.47 (1.43–8.45) (P � 0.01).

In patients with CRP levels below themedian and absence of both diabetes andimpaired kidney function, hospital mor-tality was 4.8%. In patients with CRP lev-els above the mean and the presence ofboth diabetes and impaired kidney func-tion, hospital mortality increased to35.3% (P � 0.001).

CONCLUSIONS — The study of theMunich Myocardial Infarction Registrydemonstrates that CRP on admission is astrong predictor for hospital mortality inboth diabetic and nondiabetic patients. Acutoff of 7 mg/dl for CRP levels on admis-sion is suggested for patients with acutemyocardial infarction. In patients withCRP levels equal to or below the cutoff,hospital mortality was 9% compared with20% in patients with CRP levels above themedian (P � 0.001).

Diabetic patients presented withhigher CRP levels compared with those innondiabetic subjects. Furthermore, theprevalence of diabetes increased signifi-cantly with quintiles of CRP levels.

The standard cutoff for CRP has beenreported to be 5 mg/l (4). The use of thiscutoff level, however, does not incorpo-rate the overall CRP elevation in acutemyocardial infarction, which is consid-ered to occur as a result of the acute event(19,20). The Munich Myocardial Infarc-tion Registry, which investigates patients

with troponin-positive myocardial infarc-tion with and without ST elevation, sug-gests the use of a 7 mg/dl cutoff for CRPlevels. Previously, Lim et al. (21) sug-gested CRP levels of 10 mg/l and Dibra etal. (22) CRP levels �12 mg/l to detectacute myocardial infarction patients at in-creased risk for short- and long-term mor-tality. In the two studies, however, only147 and 250 patients were included.

In the Munich Myocardial InfarctionRegistry, the combined presence of diabe-tes and CRP levels in the two upper quin-tiles demonstrated that the rate ofmortality was six- to sevenfold higherthan that in diabetic patients who pre-sented with CRP levels in the lowesttertile. The relationship between athero-thrombosis, inflammation, and diabetesis supported in the clinical setting of aregistry (23,24).

The Munich Myocardial InfarctionRegistry emphasizes the importance ofCRP levels on admission with regard tothe hospital outcome of diabetic and non-diabetic patients. A 7 mg/dl cutoff for CRPlevels on admission is suggested for pa-tients with acute myocardial infarction.The excessive risk of mortality in patientswith diabetes and elevated CRP will re-quire an intensification of strategies toovercome the poor prognosis.

References1. Ridker PM, Cushman M, Stampfer MJ,

Tracy RP, Hennekens CH: Inflammation,aspirin, and the risk of cardiovascular dis-ease in apparently healthy men. N Engl

J Med 336:973–979, 19972. Ross R: Atherosclerosis: an inflammatory

disease. N Engl J Med 340:115–126, 19993. Libby P, Ridker PM, Maseri A: Inflamma-

tion and atherosclerosis. Circulation 105:1135–1143, 2002

4. Otter W, Standl E, Schnell O: [Inflamma-tion and atherogenesis in diabetes melli-tus-new therapeutic approaches]. Herz29:524–531, 2004 [in German]

5. Rowe IF, Walker LN, Bowyer DE, SoutarAK, Smith LC, Pepys MB: Immunohisto-chemical studies of C-reactive protein andapolipoprotein B in inflammatory and ar-terial lesions. J Pathol 145:241–249, 1985

6. Blake GJ, Ridker PM: C-reactive proteinand other inflammatory risk markers inacute coronary syndromes. J Am Coll Car-diol. 41:37S–42S, 2003

7. Pai JK, Pischon T, Ma J, Manson JE,Hankinson SE, Joshipura K, Curhan GC,Rifai N, Cannuscio CC, Stampfer MJ,Rimm EB: Inflammatory markers and therisk of coronary heart disease in men andwomen. N Engl J Med 351:2599–2610,2004

8. Blake GJ, Ridker PM: Inflammatory bio-markers and cardiovascular risk predic-tion. J Intern Med 252:283–294, 2002

9. Kannel WB, McGee DL: Diabetes and car-diovascular disease: the Framinghamstudy. JAMA 241:2035–2038, 1979

10. Stern MP: Diabetes and cardiovasculardisease: the “common soil” hypothesis.Diabetes 44:369–374, 1995

11. Festa A, D’Agostino R Jr, Howard G, Myk-kanen L, Tracy RP, Haffner SM: Chronicsubclinical inflammation as part of the in-sulin resistance syndrome: the Insulin Re-sistance Atherosclerosis Study (IRAS).Circulation 102:42–47, 2000

12. Festa A, Hanley AJ, Tracy RP, D’AgostinoR Jr, Haffner SM: Inflammation in the pre-diabetic state is related to increased insu-lin resistance rather than decreasedinsulin secretion. Circulation 108:1822–1830, 2003

13. Otter W, Kleybrink S, Doering W, StandlE, Schnell O: Hospital outcome of acutemyocardial infarction in patients with andwithout diabetes mellitus. Diabet Med 21:183–187, 2004

14. Schnell O, Schafer O, Kleybrink S, Doer-ing W, Standl E, Otter W: Intensificationof therapeutic approaches reduces mor-tality in diabetic patients with acute myo-cardial infarction: the Munich Registry.Diabetes Care 27:455–460, 2004

15. Alpert JS, Thygesen K, Antman E, BassandJP: Myocardial infarction redefined: aconsensus document of The Joint Euro-pean Society of Cardiology/AmericanCollege of Cardiology Committee for theRedefinition of Myocardial Infarction.J Am Coll Cardiol 36:959–969, 2000

16. Bertrand ME, Simoons ML, Fox KA, Wal-lentin LC, Hamm CW, McFadden E, DeFeyter PJ, Specchia G, Ruzyllo W: Man-

Figure 1—CRP quintiles in acute myocardial infarction (1st quintile, �3 mg/dl; 2nd, 3–4 mg/dl;3rd, 5–9 mg/dl; 4th, 10–40 mg/dl; and 5th, �40 mg/dl). P for trend vs. lowest quintile: *P � 0.05;**P � 0.01; and ***P � 0.001.

��

, entire group of patients; f, diabetic patients; �, nondiabeticpatients.

Otter and Associates


agement of acute coronary syndromes inpatients presenting without persistent ST-segment elevation. Eur Heart J 23:1809–1840, 2002

17. Braunwald E, Antman EM, Beasley JW,Califf RM, Cheitlin MD, Hochman JS,Jones RH, Kereiakes D, Kupersmith J,Levin TN, Pepine CJ, Schaeffer JW, SmithEE III, Steward DE, Theroux P, GibbonsRJ, Alpert JS, Faxon DP, Fuster V, Grego-ratos G, Hiratzka LF, Jacobs AK, Smith SCJr: ACC/AHA 2002 guideline update forthe management of patients with unstableangina and non-ST-segment elevationmyocardial infarction: summary article: areport of the American College of Cardi-ology/American Heart Association TaskForce on Practice Guidelines (Committeeon the Management of Patients With Un-stable Angina). J Am Coll Cardiol 40:1366–1374, 2002

18. Van de Werf F, Ardissino D, Betriu A,Cokkinos D, Falk E, Fox KA, Julian D,Lengyel M, Neumann FJ, Ruzyllo W, Thy-

gesen C, Underwood SR, Vahanian A,Verheugt FW, Wijns W, Task Force onthe Management of Acute Myocardial In-farction of the European Society of Cardi-ology: Managment of acute myocardialinfarction in patients presenting with ST-segment elevation. Eur Heart J 24:28–66,2003

19. Sanchis J, Bodi V, Llacer A, Nunez J,Facila L, Ruiz V, Blasco M, Sanjuan R,Chorro FJ: Usefulness of C-reactive pro-tein and left ventricular function for riskassessment in survivors of acute myocar-dial infarction. Am J Cardiol 94:766–769,2004

20. Hoffmann R, Suliman H, Haager P, Chr-istott P, Lepper W, Radke PW, Ortlepp J,Blindt R, Hanrath P, Weber C: Associa-tion of C-reactive protein and myocardialperfusion in patients with ST-elevationacute myocardial infarction. Atherosclero-sis 186:177–183, 2005

21. Lim SY, Jeong MH, Bae EH, Kim W, KimJH, Hong YJ, Park HW, Kang DG, Lee YS,

Kim KH, Lee SH, Yun KH, Hong SN, ChoJG, Ahn YK, Park JC, Ahn BH, Kim SH,Kang JC: Predictive factors of major ad-verse cardiac events in acute myocardialinfarction patients complicated by cardio-genic shock undergoing primary percuta-neous coronary intervention. Circ J69:154–158, 2005

22. Dibra A, Mehilli J, Schwaiger M, SchuhlenH, Bollwein H, Braun S, Neverve J,Schomig A, Kastrati A: Predictive value ofbasal C-reactive protein levels for myocar-dial salvage in patients with acute myo-cardial infarction is dependent on thetype of reperfusion treatment. Eur Heart J24:1128–1133, 2003

23. Biondi-Zoccai GG, Abbate A, Liuzzo G,Biasucci LM: Atherothrombosis, inflam-mation, and diabetes. J Am Coll Cardiol41:1071–1077, 2003

24. Gonzalez MA, Selwyn AP: Endothelialfunction, inflammation, and prognosis incardiovascular disease. Am J Med 115(Suppl. 8A):99S–106S, 2003

CRP and acute myocardial infarction


Pregnancy-Associated Plasma Protein-ALevels Are Related to Glycemic Control butNot to Lipid Profile or HemostaticParameters in Type 2 DiabetesSILVIA PELLITERO, MD

1

JORDI L. REVERTER, MD, PHD1

EDUARDA PIZARRO, MD, PHD2

MARIA CRUZ PASTOR, MD, PHD3

MARIA LUISA GRANADA, MD, PHD3

DOLORS TASSIES, MD, PHD4

JUAN-CARLOS REVERTER, MD, PHD4

ISABEL SALINAS, MD, PHD1

ANNA SANMARTI, MD, PHD1

P regnancy-associated plasma pro-tein-A (PAPP-A) is a zinc-bindingmatrix metalloproteinase that regu-

lates extracellular matrix remodeling.PAPP-A degrades IGFBP-4, increasinglevels of local IGF-1 in response to injury,and could be involved in the pathogenesisof atherosclerosis (1– 6). Inflammatorycytokines tumor necrosis factor (TNF)-�and interleukin (IL)-1�, implicated in in-sulin resistance (7), are potent stimulatorsof PAPP-A (8,9). The association betweenPAPP-A levels and metabolic parameterssuch as cholesterol and high-sensitivityC-reactive protein (hsCRP) is controver-sial (2,10,11). We aimed to study the re-lationship between PAPP-A, glycemiccontrol and other metabolic and hemo-static parameters, inflammatory cyto-kines, and ankle-brachial pressure index(ABI) in diabetic patients.

RESEARCH DESIGN ANDMETHODS — Type 2 diabetic pa-tients (n � 175, 65 of whom werewomen) with stable glycemic control(variation in A1C �10% in the last 5years) and without diagnosis of clinical

macrovascular disease, inflammatory dis-ease, malignancies, or pregnancy werestudied. Fifty-three (20 of whom werewomen) nondiabetic subjects withoutprevious clinical macrovascular diseaseand normal ABI (�0.9) were recruited ascontrol subjects.

Demographic, anthropometric, andclinical data and ABI were recorded in allsubjects. Laboratory data were measuredby commercially available assays, hsCRPby nephelometry, ultrasensitive PAPP-Ausing an enzyme-linked immunosorbentassay, and TNF-� and IL-6 concentra-tions using an enzyme chemilumines-cence immumometric assay.

Continuous variables were expressedas means � SD or median (interquartilerange). Differences between groups wereexamined by Student’s t test or Mann-Whitney and correlation between vari-ables by Pearson’s or Spearman’s tests asrequired. Multiple logistic regressionanalysis was performed.

RESULTS — Clinical and biochemicalcharacteristics of all study subjects areshown in Table 1. PAPP-A levels were sig-

nificantly higher in male than in femalesubjects in both groups (median [inter-quartile range] 1.04 [0.6–1.47] vs. 0.52mIU/l [0.43–0.94], P � 0.025 in controlsubjects and 0.49 [0.23–0.93] vs. 0.35mIU/l [0.13–0.63], P � 0.01 in diabeticpatients).

Serum PAPP-A concentrations weresignificantly higher in control than in di-abetic subjects (median [interquartilerange] 0.73 [0.48–1.33] vs. 0.45 mIU/l[0.19–0.82], respectively, P � 0.0001)and correlated negatively with A1C (r ��0.2, P � 0.03). Diabetic patients werestratified according to mean � SD valuesof A1C (�5.9, 5.9 – 8.2, and �8.2%).PAPP-A concentration was significantlylower in patients with A1C �8.2% (0.35mUI/l [0.07–0.43]) compared with thatin patients with A1C �5.9% (0.72 mUI/l[0.2–0.92], P � 0.03) and between 5.9and 8.2% (0.56 mUI/l [0.15–0.83], P �0.02) and control subjects (0.73 mUI/l[0.48–1.33], P � 0.001).

No differences in PAPP-A levels wereobserved when subjects with normocho-lesterolemia were compared with thosewith hypercholesterolemia (median [in-terquartile range] 0.6 [0.45–1.14] vs. 0.8mUI/l [0.48 –1.38], respectively) andwith diabetic patients (0.6 [0.45–1.14]vs. 0.8 mUI/l [0.48–1.38]). On the otherhand, when control subjects and diabeticpatients with normocholesterolemia werecompared, PAPP-A levels remained sig-nificantly higher in control subjects thanin diabetic patients (0.6 [0.45–1.14] vs.0.33 mUI/l [0.13– 0.83], respectively,P � 0.04). We obtained the same resultswhen control subjects and diabetic pa-tients with hypercholesterolemia werecompared (0.8 [0.48 –1.38] vs. 0.44mUI/l [0.22–0.78], P � 0.0001). More-over, PAPP-A levels were similar in sub-jects treated with statins compared withthose in untreated subjects in bothgroups.

No differences were observed inPAPP-A levels between diabetic patientswith or without a history of diabetic vas-culopathy (n � 25), abnormal ABI (n �54), nephropathy (n � 42), or retinopa-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Endocrinology and Nutrition Service, Department of Medicine, Germans Trias i Pujol UniversityHospital, Universitat Autonoma de Barcelona, Badalona, Spain; the 2Endocrinology Unit, Mataro Hospital,Barcelona, Spain; 3Clinical Biochemistry Service, Germans Trias i Pujol University Hospital UniversitatAutonoma de Barcelona, Badalona, Spain; and the 4Hemotherapy and Hemostasis Service, Hospital Clinic,Barcelona, Spain.

Address correspondence and reprint requests to Jordi L. Reverter, MD, PHD, Endocrinology and Nutri-tion Service, Germans Trias i Pujol University Hospital, Via Canyet s/n, 08916, Badalona, Spain. E-mail:[email protected].

Received for publication 8 June 2007 and accepted in revised form 21 August 2007.Published ahead of print at http://care.diabetesjournals.org on 28 August 2007. DOI: 10.2337/dc07-

1092.Abbreviations: ABI, ankle-brachial pressure index; hsCRP, high-sensitivity C-reactive protein; IL, inter-

leukin; PAPP-A, pregnancy-associated plasma protein-A; TNF, tumor necrosis factor.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





thy (n � 59), and no relationship wasfound between plasma levels of PAPP-Aand those of hemostasis parameters andinflammatory cytokines.

Multiple logistic regression analysesusing PAPP-A as the dependent variableand age, BMI, and biochemical parame-ters as independent variables showed noassociated factors other than A1C (P �0.02) and glycemia (P � 0.04).

CONCLUSIONS — In the presentstudy, PAPP-A levels were significantlylower in diabetic patients compared withthose in age- and sex-matched controlsubjects without clinical macrovasculardiseases, and, for the first time, a signifi-cant inverse correlation was found be-tween PAPP-A and A1C, independent ofother clinical and metabolic factors. Therelationship between PAPP-A levels andA1C could reflect the influence of glyce-mic control on the regulation of PAPP-Aexpression. A possible hypothesis to con-sider is that PAPP-A may not be a goodmarker of vascular risk in chronic diseases

such as diabetes. In fact, PAPP-A wouldbe a modulator of proliferative local ac-tion of IGF-1 in atherosclerotic plaques(1–6). IGF-1 acts as a promoter of repairat damaged tissues (4,5,12), and in-creased PAPP-A levels may reflect a re-paired mechanism against vasculardamage (10).

In previous articles (13), diabetic pa-tients with hypercholesterolemia showedhigher PAPP-A levels than control sub-jects. We do not have an explanation forthis disparity in our results; however, inthat previous study, PAPP-A levels weremeasured in diabetic patients with a widerange of A1C, without a hypercholester-olemic control group.

No correlations were observed be-tween PAPP-A and clinical and other bio-chemical data. In previous studies on thistopic, there were discrepancies, and someauthors have reported a relationship(2,10,14) between PAPP-A and hsCRP,while others have not (11).

The lack of an observed correlationbetween PAPP-A levels and IL-6 or

TNF-� could reflect the complex interac-tion of multiple cytokines, and it is evenpossible that their exact role in PAPP-Aexpression is unknown. We found noprevious reports on the relationship be-tween plasma PAPP-A levels and the he-mostatic parameters evaluated that wereselected to globally identify coagulationor fibrinolysis activation in diabetic pa-tients. The absolute PAPP-A serum con-centrations found in our study couldnot be compared with those reported inother studies (2,10,11,15) owing to thedifferent methods used for PAPP-Ameasurement (16).

Acknowledgments— This study was sup-ported by a grant “Ajut a la recerca en diabetisGoncal Lloveras i Valles” from the Catalan Di-abetes Association.

References1. Bayes-Genis A, Conover CA, Schwartz RS:

The Insulin-like growth factor axis: a re-view of atherosclerosis and restenosis.Circ Res 86:125–130, 2000

2. Bayes-Genis A, Conover CA, OvergaardMT, Bailey KR, Christiansen M, HolmesDR, Virmani R, Oxvig C, Schwartz RS:Pregnancy-associated plasma protein Aas a marker of acute coronary syn-dromes. N Engl J Med 345:1022–1029,2001

3. Cosin-Sales J, Kaski JC, Christiansen M,Kaminski P, Oxvig C, Overgaard MT,Cole D, Holt DW: Relationship amongpregnancy associated plasma protein-Alevels, clinical characteristics, and coro-nary artery disease extent in patients withchronic stable angina pectoris. Eur Heart J26:2093–2098, 2005

4. Conti E, Andreotti F, Zuppi C: Pregnan-cy-associated plasma protein-A as predic-tor of outcome in patients with suspectedacute coronary syndromes. Circulation109:e211–e212, 2004

5. Crea F, Andreotti F: Pregnancy associatedplasma protein-A and coronary athero-sclerosis: marker, friend, or foe? EurHeart J 26:2075–2076, 2005

6. Pinon P, Kaski JC: Inflammation, athero-sclerosis and cardiovascular disease risk:PAPP-A, Lp-PLA2 and cystatin C: new in-sights or redundant information? Rev EspCardiol 59:247–258, 2006

7. Reaven GM: The metabolic syndrome: isthis diagnosis necessary? Am J Clin Nutr83:1237–1247, 2006

8. Resch ZT, Oxvig C, Bale LK, Conover CA:Stress-activated signaling pathways medi-ate the stimulation of pregnancy-associ-ated plasma protein-A expression incultured human fibroblasts. Endocrinology147:885–890, 2006

Table 1—Clinical, biochemical, and hemostatic characteristics of control and diabetic subjects

Controlsubjects

Diabeticpatients P

n 53 175 —Male/female 33/20 110/65 NS

Age (years) 63.3 � 7.5 62 � 7.9 NSTreatment (%)

Antidiabetes drugs — 39.3 —Insulin — 27.7 —Combined — 34.0 —Statins 27 70 0.001

ABI �0.9 — 30.8 —Hypercholesterolemia 69 84 0.02Current smoker 19 17 NSBMI (kg/m2) 29.5 � 2.6 31.1 � 4.7 0.002Waist circumference (cm) 98.7 � 7.9 104.3 � 11.8 0.002SBP (mmHg) 149 � 19.5 146.7 � 20.7 NSDBP (mmHg) 87.6 � 9.3 79.1 � 11.4 �0.0001Fasting plasma glucose (mg/dl) 81.3 � 11.9 156.2 � 46.9 �0.0001A1C (%) 5.2 � 0.9 7.1 � 1.1 �0.0001Cholesterol (mg/dl)

Total 210.6 � 39 182.3 � 39.7 �0.0001HDL 48.3 � 13.6 44.8 � 13.6 NSNon-HDL 162.2 � 34.5 136.5 � 36.8 �0.0001

Triglycerides (mg/dl) 97 (73–153) 131 (88–193) 0.01Uric acid (mg/dl) 5.1 � 1.5 5.8 � 1.6 0.03hsCRP (mg/l) 3.47 (1.13–5.86) 3.15 (1.55–5.78) NSFibrinogen (g/l) 3.2 � 1.0 4.5 � 1.1 �0.0001F1 � 2 (nmol/l) 0.81 � 0.32 2.03 � 0.65 �0.0001PAP (g/l) 198.9 � 90.2 207.1 � 87.3 NS

Data are means � SD, percentages, or median (interquartile range). DBP, diastolic blood pressure; F1 � 2,prothrombin fragment 1 � 2; NS, nonsignificant; PAP, plasmin-antiplasmin complexes; SBP, systolic bloodpressure.

Pregnancy-associated plasma protein-A and diabetes


9. Resch ZT, Chen BK, Bale LK, Oxvig C,Overgaard MT, Conover CA: Pregnan-cy-associated plasma protein A gene ex-pression as a target of inflammatorycytokines. Endocrinology 145:1124 –1129, 2004

10. Beaudeux JL, Burc L, Imbert-Bismut F,Giral P, Bernard M, Bruckert E, ChapmanMJ: Serum plasma pregnancy-associatedprotein A: a potential marker of echogeniccarotid atherosclerotic plaques in asymp-tomatic hyperlipidemic subjects at highcardiovascular risk. Arterioscler ThrombVasc Biol 23:e7–e10, 2003

11. Stulc T, Malbohan I, Malik J, Fialova L,Soukupova J, Ceska R: Increased levels ofpregnancy-associated plasma protein-Ain patients with hypercholesterolemia:the effect of atorvastatin treatment. Am

Heart J 146:1060–1063, 200312. Conti E, Carrozza C, Capoluongo E,

Volpe M, Crea F, Zuppi C, Andreotti F:Insuline-like growth factor-1 as a vascularprotective factor. Circulation 110:2260–2265, 2004

13. Aso Y, Okumura K, Wakabayashi S, Take-bayashi K, Taki S, Inukai T: Elevatedpregnancy-associated plasma protein-Ain sera from type 2 diabetic patients withhypercholesterolemia: associations withcarotid atherosclerosis and toe-brachialindex. J Clin Endocrinol Metab 89:5713–5717, 2004

14. Heeschen C, Dimmeler S, Hamm CW,Fichtlscherer S, Simoons ML, Zeiher AM:Pregnancy-associated plasma protein-Alevels in patients with acute coronary syn-dromes: comparison with markers of sys-

temic inflammation, platelet activation,and myocardial necrosis. J Am Coll Cardiol45:229–237, 2005

15. Cosin-Sales J, Christiansen M, KaminskiP, Oxvig C, Overgaard MT, Cole D, HoltDW, Kaski JC: Pregnancy-associatedplasma protein A and its endogenous in-hibitor, the proform of eosinophil majorbasic protein (proMBP), are related tocomplex stenosis morphology in patientswith stable angina pectoris. Circulation109:1724–1728, 2004

16. Fredericks S, Bertomeu-Gonzalez V,Petrovic I, Holt DW, Kaski JC: Commenton immunoassays developed for pregnan-cy-associated plasma protein-A (PAPP-A)in pregnancy may not recognize PAPP-Ain acute coronary syndromes. Clin Chem52:1619–1620, 2006

Pellitero and Associates


Low Serum Angiogenin Concentrations inPatients With Type 2 DiabetesJANUSZ SIEBERT, MD, PHD

1

MAGDALENA REIWER-GOSTOMSKA, MD1

ZOFIA BABINSKA, MD1

JOLANTA MYSLIWSKA, MD, PHD2

ANDRZEJ MYSLIWSKI, MD, PHD3

EWA SKOPINSKA-ROZEWSKA, MD, PHD4

EWA SOMMER, MD4

PIOTR SKOPINSKI, MD5,6

P remature development of microvas-cular and macrovascular disease isthe most frequent complication of

diabetes. It is responsible for diabetic ret-inopathy, nephropathy, and neuropathy(1). Moreover, diabetes leads to reducedcollateralization in ischemic tissues,which causes impaired wound healing,exacerbation of peripheral limb ischemia,and a three- to fourfold increase in cardiacmortality in comparison with nondiabeticpatients (2,3).

The pathophysiological mechanismsresponsible for impaired angiogenic ac-tivity in diabetes remain unknown. Al-though angiogenesis impairment hasbeen attributed to alterations in the generegulatory network, which can be in-volved in the physiological revasculariza-tion process, the role of angiogenin in thisprocess has not been clarified (2,4).

The aim of this study therefore was tocompare angiogenin serum levels inhealthy and type 2 diabetic age-matchedindividuals. The cutaneous angiogenesisin vivo (serum-induced angiogenesis[SIA]) test in mice was performed in par-allel to determine the functional differ-ences between the groups examined.

RESEARCH DESIGN ANDMETHODS — A total of 43 patientswith type 2 diabetes and 43 age-matched

healthy control subjects volunteered forthe study. Patients suffering from acuteand chronic infections and neoplastic dis-eases were excluded. Diabetes was de-fined according to American DiabetesAssociation criteria (5). All patients hadblood glucose levels �7 mmol/l. Serumangiogenin levels were measured in du-plicate with the ELISA Quantikine kit(R&D System, Minneapolis, MN). TheSIA test was performed on Balb/c miceaccording to the method of Sidky andAuerbach (6) with own modifications (7–9).

The results were analyzed using Sta-tistica software, Version 7 (StatSoft Pol-ska). The level of significance was set atP � 0.05, and two-sided tests were per-formed as the standard.

RESULTS — The diabetic patientswere characterized by significantly higherlevels of triglycerides and serum creati-nine and lower levels of LDL cholesterol,while BMI and total as well as HDL cho-lesterol concentrations were not signifi-cantly different from those of healthyindividuals (Table 1).

Serum angiogenin levels were signifi-cantly lower in type 2 diabetic patients inrelation to control subjects (P �0.000002). At the same time, sera fromdiabetic patients induced significantly

fewer newly formed blood vessels on theinner surface of skin from Balb/c mice(P � 0.03).

Serum angiogenin levels as well asSIA values were significantly lower in thepatients with late complications (retinop-athy and nephropathy) in relation tothose without complicated disease (P �0.04 for angiogenin and P � 0.02 forSIA).

There were no differences, however,in angiogenin and SIA values between thepatients receiving statins (P � 0.93 forangiogenin and P � 0.74 for SIA) or ACEinhibitors (P � 0.70 for angiogenin andP � 0.51 for SIA) and those not treatedwith either. The multivariate linear step-wise regression analysis confirmed theprevious results revealing that none of theparameters examined, including medi-cines, exerted an effect on angiogenin orSIA values (P � 0.70). The angiogeninand SIA values were not correlated withthe A1C concentrations (P � 0.4).

CONCLUSIONS — Angiogenin, a14-kDa protein, is implicated in immuno-logical and inflammatory angiogenesis(10). The level of angiogenin in humanplasma is strictly regulated (11); the pro-tein is a normal constituent of blood, andits level usually remains unchanged.However, in some pathological condi-tions such as peripheral vascular disease,inflammatory bowel disease, rheumatoidarthritis, obesity, proliferative diabeticretinopathy, and proliferative vitreopa-thy, it can intensify the induction of newblood vessel formation (10).

We discovered that the serum angio-genin level is markedly decreased in type2 diabetic patients in comparison withage-matched healthy subjects. The lowerangiogenic potential of sera from type 2diabetic patients was confirmed by the invivo SIA test. The more severe disease,i.e., complicated with retinopathy or ne-phropathy, was associated with lower val-ues of both angiogenin and SIA. On theother hand, the differences in values ofboth indicators of angiogenesis were notrelated, in our analysis, to any of the otherclinical parameters examined in the dia-betic patients. Since the number of dia-betic patients was limited in our studyand almost all patients had poorly con-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1University Centre for Cardiology, Department of Family Medicine, Medical University of Gdansk,Gdansk, Poland; the 2Department of Immunology, Medical University of Gdansk, Gdansk, Poland; the3Department of Histology, Medical University of Gdansk, Gdansk, Poland; the 4Department of Pathology,Biostructure Centre, Medical University of Warsaw, Warsaw, Poland; the 5Department of Histology, Emb-riology, Biostructure Centre, Medical University of Warsaw, Warsaw, Poland; and the 62nd Department ofOphthalmology, Medical University of Warsaw, Warsaw, Poland.

Address correspondence and reprint requests to Prof. Janusz Siebert, Medical University of Gdansk, 2Debinki St., Gdansk 80952, Poland. E-mail: [email protected].

Received for publication 30 March 2007 and accepted in revised form 12 September 2007.Published ahead of print at http://care.diabetesjournals.org on 18 September 2007. DOI: 10.2337/dc07-


dc07-0629.Abbreviations: SIA, serum-induced angiogenesis.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





trolled A1C levels, we can assume thatthey constituted a very homogenousgroup. The angiogenin and SIA valueswere in a rather narrow range. Therefore,they most likely reached a point when anytreatment ceased to play a role. The mostimportant finding was that neither statinsnor angiotensin inhibitors influenced thevalues of both angiogenesis indicators.

Our data are compatible with theknown phenomenon of limited collateralvessel development in coronary heart dis-ease, which is associated with a pro-nounced myocardial ischemia in diabeticpatients (2). In experimental in vitro andin vivo models of retinal angiogenesis as-says, Stitt et al. (1) discovered that sera oftype 2 diabetic patients have strong anti-angiogenic effects. The poorer the glyce-mic control of patients, the more evidentthe inhibition of angiogenesis. The au-thors documented that advanced glyca-tion end products and their receptors maybe mediators in retinal angiogenesis inhi-bition. Adding to that, Chou et al. (12)suggest a possible explanation for im-paired collateral formation in cardiac tis-sue by reduction in vascular endothelialgrowth factor in ventricles from diabeticpatients compared with those in controlsubjects. Similar results were obtained by

Chung et al. (13). Their report indicates a48% reduction of vascular endothelialgrowth factor in internal mammary arteryof diabetic patients undergoing coronaryartery bypass graft surgery. Our resultsimply, additionally, that angiogenesis in-hibition may be realized through a reduc-tion in angiogenin level. Whether such arelation really exists is a matter for furtherinvestigation.

References1. Stitt AW, McGoldrick C, Rice- McCaldin

A, McCance DR, Glenn JV, Hsu DK, Fu-Tong Liu, Thorpe SR, Gardiner TA: Im-paired retinal angiogenesis in diabetes:role of advanced glycation end productsand galectin-3. Diabetes 53:857–794,2005

2. Abaci A, Oguzhan A, Kahraman S, EryolNK, UnalS, Arinc H, Egrin A: Effect ofdiabetes mellitus on formation of coro-nary collateral vessels. Circulation 99:2239–2242, 1999

3. Schiekofer S, Gallasso G, Sato K, Kraus BJ,Walsh K: Impaired revascularization in amouse model of type 2 diabetes is associ-ated with dysregulation of complex angio-genic-regulatory network. ArterisclerThromb Vasc Biol 25:1603–1609, 2005

4. Taniyama Y, Morishita R, Hiraoka K, AokiM, Nakagami H, Yamasaki K, Matsumoto

K, Nakamura T, Kaneda Y, Ogihara T:Therapeutic angiogenesis induced by hu-man hepatocyte growth factor gene in ratdiabetic hind limb ischemia model: mo-lecular mechanisms of delayed angiogen-esis in diabetes. Circulation 104:2344–2350, 2001

5. American Diabetes Association: Diagnosisand classification of diabetes mellitus. Di-abetes Care 28 (Suppl. 1): S37–S47, 2005

6. Sidky YA, Auerbach R: Lymphocyte-in-duced angiogenesis: a quantitative andsensitive assay of the g-v-h reaction. J ExpMed 141:1084–1100, 1975

7. Skopinski P, Rogala E, Duda-Krol B, Lip-inska A, Sommer E, Chorostowska-Wyn-imko J, Szaflik J, Partyka I, Skopinska-Rozewska E: Increased interleukin 18content and angiogenic activity of serafrom diabetic (type 2) patients with back-ground retinopathy. J Diabetes Complica-tions 19:335–338, 2005

8. Skopinski P, Szaflik J, Partyka I, Choros-towska- Wynimko J, Duda-Krol B, Lipin-ska A, Sommer E, Skopinska-Rozewska E:Serum in vivo angiogenic activity andsome pro-angiogenic cytokine levels indiabetes mellitus type 2 (DM2) patientswith or without background retinopathy.Centr Eur J Immunol 32:48–52, 2007

9. Skopinski P, Szaflik J, Duda-Krol B, Nar-towska J, Sommer E, Chorostowska-Wynimko J, Demkow U, Skopinska-Rozewska E: Suppression of angiogenicactivity of sera from diabetic patients withnon-proliferative retinopathy by com-pounds of herbal origin and sulindac sul-fone. Int J Mol Med 14:707–711, 2004

10. Hu G, Riordan JF., Vallee BL: Angiogeninpromotes invasiveness of cultured endo-thelial cells by stimulation of cell-associ-ated proteolytic activities. Proc Natl AcadSci U S A 91:12096–12100, 1994

11. Hu G: Neomycin inhibits angiogenin-in-duced angiogenesis. Proc Natl Acad Sci U SA 95:9791–9795, 1998

12. Chou E, Suzuma I, Way KJ, Opland D,Clemont AC, Naruse K, Suzuma K, Bowl-ing NL, Vlahos CJ, Aiello LP, King GL:Decreased cardiac expression of vascularendothelial growth factor and its recep-tors in insulin-resistant and diabeticstates. Circulation 105:373–379, 2002

13. Chung AWY, Hsiang YN, Matzke LA, Mc-Manus BM, van Breemen C, Okon EB: Re-duced expression of vascular endothelialgrowth factor paralleled with the in-creased angiostatin expression resultingfrom the upregulatd activities of matrixmetalloproteinase-2 and -9 in humantype 2 diabetic arterial vasculature. CircRes 99:140–148, 2006

Table 1—Basic parameters in diabetic patients and healthy control subjects

ParameterType 2 diabetic

patientsHealthy control

subjects P value

n 43 43Age (years) 64.73 � 11.47 65.16 � 11.91 0.78BMI (kg/m2) 32.09 � 5.91 30.11 � 4.51 0.13Systolic blood pressure (mmHg) 140.81 � 15.80 137.83 � 11.91 0.4Diastolic blood pressure (mmHg) 78.10 � 7.71 81.01 � 4,80 0.1Diabetes duration (years) 9.63 � 5.60A1C 8.23 � 1.72Total cholesterol (mmol/l) 224.36 � 45.20 240.45 � 39.79 0.1LDL cholesterol (mmol/l) 135.59 � 42.61 156.52 � 32.96 0.02HDL cholesterol (mmol/l) 51.42 � 17.74 55.12 � 12.10 0.3Triglycerides (mmol/l) 187.25 � 89.69 135.6 � 54.0 0.004Serum creatinine (mg/100 ml) 1.01 � 0.185 0.86 � 0.23 0.04Retinopathy (%) 71.87Nephropathy (%) 35.71ACE inhibitor treatment (%) 86.49Statin treatment (%) 48.65Angiogenin serum level (pg/ml) 319.72 � 107.04 550.54 � 187.99 0.0000015Cutaneous angiogenesis in vivo in

serum-induced angiogenesis test*38.17 � 11.80 52.51 � 20.22 0.03

Data are arithmetic means � SD or percentage of patients, unless otherwise indicated.*Mean � SD newly formed blood vessels on inner skin surface of Balb/c mice.

J. Siebert and Associates


Latent Autoimmune Diabetes in Adults in aSouth Asian Population of the U.K.ABIGAIL C. BRITTEN, BSC

1

KAREN JONES, BSC1

CARINA TORN, PHD2

MAGNUS HILLMAN, PHD2

BIRGITTE EKHOLM2

SUDHESH KUMAR, MD, FRCP3

ANTHONY H. BARNETT, MD, FRCP1,4

MARILYN ANN KELLY, PHD1

T ype 2 diabetes is four- to sixfoldmore common in the South Asianpopulation of the U.K. than in the

indigenous white Caucasian population.A subset of all patients initially diagnosedwith type 2 diabetes shows evidence ofslowly evolving islet autoimmunity,termed latent autoimmune diabetes inadults (LADA). LADA is characterized bythe presence of circulating autoantibodiesspecific for islet proteins and by insulinindependence for at least 6 months post-diagnosis (1).

A recent pilot study in Birmingham,U.K., suggested that 27% of South Asiansinitially presenting with type 2 diabeteswere positive for autoantibodies toGAD65 and/or insulinoma-associatedprotein (IA)-2 (2). This is significantlyhigher than the islet autoimmunity fre-quency of 10% observed in white Cauca-sians diagnosed with type 2 diabetes(3,4). The study in South Asians was car-ried out in a very small cohort, however,and the findings require confirmation in amuch larger study group. The aim of thisstudy was to determine the prevalence ofLADA in a larger U.K.-resident SouthAsian population and to characterize thephenotypic features and genetic basis ofthe disease in this ethnic group.

RESEARCH DESIGN ANDMETHODS — A total of 500 SouthAsian subjects with type 2 diabetes (mean[range] age 55 years [31-89] and diseaseduration 7 years [0-29]) (Table 1) were

consecutively recruited in Birmingham,U.K., and Coventry, U.K., as part of theU.K. Asian Diabetes Study. A total of 206normoglycemic control subjects (age 49years [30-83]) were recruited in Birming-ham. All subjects were of Punjabi ances-try. Type 2 diabetes was definedaccording to World Health Organizationcriteria (5). The study was approved bythe local ethics committee, and writteninformed consent was obtained from allparticipants. Venous blood samples werecollected from each subject, plasma wasremoved for autoantibody analysis, andDNA was extracted from the remainingblood using an adaptation of the Nucleonprotocol (Nucleon Biosciences, Coat-bridge, U.K.). LADA was defined as de-scribed above (1).

Antibody analysisPlasma samples were incubated with anexcess of calcium ions overnight, fol-lowed by centrifugation. The superna-tants were analyzed for autoantibodies toGAD65 and IA-2 using commerciallyavailable enzyme-linked immunosorbentassay kits according to the manufacturer’sinstructions (RSR, Cardiff, U.K.). The ref-erence value was 10 units/ml for GAD65antibodies and 15 units/ml for IA-2 auto-antibodies (based on the World HealthOrganization standard).

Genetic analysisDNA samples were typed for alleles ofHLA-DRB1, -DQA1, and -DQB1 using

the phototyping method (6,7). The insu-lin gene variable number tandem repeat(INS-VNTR) type was determined usingrestriction fragment–length polymor-phism analysis with HphI (8). Alleles ofthe GCT microsatellite in the major histo-compatibility complex class I chain-related gene-A (MIC-A) gene were typedusing the method described by Gambel-unghe et al. (9).

Statistical analysisAssociations between genotype and auto-antibody status were analyzed using the�2 test or Fisher’s exact test. Differences incontinuous variables were investigatedusing the Mann-Whitney U test. All sta-tistical analyses were performed usingSPSS (version 13.0; SPSS, Chicago, IL).

RESULTS — Autoantibodies were de-tected in 13 of 500 (2.6%) individualswith type 2 diabetes (of whom 8 wereGAD65 positive [1.6%] and 6 were IA-2positive [1.2%], including 1 subject whowas positive for both) and 8 of 206 (3.9%)control subjects (of whom 3 were GAD65positive [1.5%] and 5 were IA-2 positive[2.4%]). There was no significant differ-ence in antibody titers between diabeticand control subjects.

The small number of autoantibody-positive subjects found in this cohort lim-ited investigation of associations betweengenotype and antibody status in the SouthAsian population, but some trends wereobserved. The DRB1*04 and DQB1*0302alleles were increased in frequency amongthe IA-2 autoantibody–positive diabetic(P � 0.020 and P � 0.015, respectively)and control (P � not significant) subjectscompared with those in subjects lackingthese markers. The distribution of the INS-VNTR genotypes did not differ significantlybetween the autoantibody-positive and au-toantibody-negative subjects in either thediabetic or control groups. The MIC-A6 al-lele was significantly less frequent amongIA-2 autoantibody–positive diabetic thanautoantibody-negative diabetic subjects(P � 0.044).

Clinical, biochemical, and anthropo-metric measurements were compared be-tween the autoantibody-positive andautoantibody-negative diabetic subjects

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Medicine, University of Birmingham, Birmingham, U.K.; 2Clinical Sciences, LundUniversity, Lund, Sweden; 3Diabetes and Metabolism, University of Warwick, Coventry, U.K.; and the4Heart of England NHS Foundation Trust, Birmingham, U.K.

Address correspondence and reprint requests to Abigail Britten, Diabetes Research Group, ELG54, TheMedical School, Vincent Drive, Edgbaston, Birmingham, B15 2TT, U.K. E-mail: [email protected].


0896.Abbreviations: IA, insulinoma-associated protein; LADA, latent autoimmune diabetes in adults.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





(Table 1). Mean weight and BMI were sig-nificantly lower in the autoantibody-positive subjects (P � 0.029 and P �0.032, respectively). A longer mean dura-tion of diabetes was observed amongindividuals positive for GAD65 autoanti-bodies compared with that in autoanti-body-negative and IA-2 autoantibody–positive subjects (P � 0.019 and P �0.009, respectively). A higher percentageof the autoantibody-positive than autoan-tibody-negative diabetic subjects wastreated with insulin (53.8 and 18.8%, re-spectively) (Table 1).

CONCLUSIONS — Our study showsthat islet autoimmunity is considerablyless common among type 2 diabetic indi-viduals of Punjabi ancestry in Birming-ham, U.K., than in those of whiteCaucasian origin. The differences ob-served between the two ethnic groupsmay reflect both the higher prevalence ofclassical type 2 diabetes in South Asiansand their lower susceptibility to autoim-mune disease. The overall prevalence ofislet autoantibodies among the individu-als diagnosed with type 2 diabetes in ourstudy cohort (2.6%) was significantlylower than that observed in the pilotstudy (27%), as was the frequency in thecontrol group (3.9 compared with 9%,respectively) (2). The reasons for these

differences are unclear, as both methodswere approved by the Diabetes AntibodyStandardization Programme (10). Themost likely explanation is that the highprevalence of autoimmunity observed inthe pilot study is a spurious result due tothe small number of individuals investi-gated (33 type 2 diabetic and 98 controlsubjects). A higher percentage of patientspositive for diabetes autoantibodies haspreviously been reported in South Indian(GAD65 and ICA512) (11) and EasternIndian (GAD65 and IA-2) (12) popula-tions resident in India compared with thatin the present study. It remains to be de-termined whether these differences aredue to genetic, environmental, or popu-lation-selection influences.

The low frequency of islet autoanti-bodies in the current study made it diffi-cult to detect statistically significantassociations with the genetic loci studiedand the clinical, biochemical, and anthro-pometric measurements. The trends thatwere observed, however, are generallyconsistent with previous associationswith islet autoimmunity seen in other eth-nic groups. Based on the findings of ourstudy, screening for LADA in the U.K.Punjabi population would offer little clin-ical benefit and is not routinely indicated.

Acknowledgments— This study was spon-sored by Diabetes UK Grant BDA:RD03/0002693). The recruitment of samplesthrough the U.K. Asian Diabetes Study wassupported by Pfizer, Sanofi-Aventis, Servier,MSD/SP, Takeda UK, Merck, Roche, Boehr-inger Ingelheim, Eli Lilly, Novo Nordisk, BMS,and Daichi Sankyo.

We thank Dr. Anthony Dixon, Dr. SrikanthBellary, Shanaz Mughal, and Kam Johal fortheir roles in the collection of samples.

References1. Pozzilli P, DiMario U: Autoimmune diabe-

tes not requiring insulin at diagnosis (latentautoimmune diabetes of the adult): defini-tion, characterization, and potential preven-tion. Diabetes Care 24:1460–1467,2001

2. Tica V, Hanif MW, Andersson A, Valsa-makis G, Barnett AH, Kumar S, SanjeeviCB: Frequency of latent autoimmune dia-betes in adults in Asian patients diagnosedas type 2 diabetes in Birmingham, UnitedKingdom. Ann N Y Acad Sci 1005:356–358, 2003

3. Romkens TE, Kusters GC, Netea MG,Netten PM: Prevalence and clinical char-acteristics of insulin-treated, anti-GAD-positive, type 2 diabetic subjects in anoutpatient clinical department of a Dutchteaching hospital. Neth J Med 64:114–118, 2006

4. Tuomi T, Carlsson A, Li H, Isomaa B, Mi-ettinen A, Nilsson A, Nissen M, Ehrn-

Table 1—Clinical parameters recorded for autoantibody-positive and autoantibody-negative diabetic subjects

Clinical parameterAutoantibody

positiveAutoantibody

negative

GAD65 autoantibody IA-2 autoantibody

Positive Negative Positive Negative

n 13 479 8 484 6 486Age (years) 58 (40–78) 55 (31–89) 55 (46–73) 55 (31–89) 65 (40–78) 55 (31–89)Disease duration (years) 9 (2–25) 7 (0–29) 13 (6–25)*† 7 (0–29) 7 (2–25) 8 (0–29)Height (cm) 165 (146–177) 162 (152–173) 166 (157–177) 162 (146–173) 164 (146–174) 162 (152–177)Weight (kg) 70 (61–78)‡ 77 (43–139) 71 (64–78) 77 (43–139) 69 (61–76) 77 (43–139)BMI (kg/m2) 26 (23–32)§ 29 (16–49) 26 (23–30) 29 (16–49) 26 (23–32) 29 (16–49)Waist circumference (cm) 98 (93–109) 102 (60–139) 99 (93–108) 102 (60–139) 95 (93–109) 102 (60–139)Diastolic blood pressure (mmHg) 80 (54–96) 83 (53–124) 81 (66–96) 83 (53–124) 75 (54–87) 83 (53–124)Systolic blood pressure (mmHg) 130 (105–170) 137 (80–203) 126 (107–146) 137 (80–203) 134 (105–170) 137 (80–203)A1C (%) 6.3 (4.4–9.5) 7.0 (2.0–15.7) 6.8 (5.4–9.5) 7.0 (2.0–15.7) 5.6 (4.4–6.5)� 7.0 (2.0–15.7)Cholesterol (mmol/l)

Total 4.9 (3.2–6.1) 4.8 (2.2–11.8) 4.7 (3.2–6.1) 4.8 (2.2–11.8) 4.7 (3.2–5.4) 4.8 (2.2–11.8)HDL 1.3 (0.8–2.1) 1.2 (0.6–3.1) 1.4 (1.1–2.1) 1.2 (0.6–3.1) 1.1 (0.8–1.2) 1.2 (0.6–3.1)LDL 2.3 (1.6–3.4) 2.4 (0.49–6.6) 2.1 (1.6–3.0) 2.5 (0.49–6.6) 2.5 (1.6–3.4) 2.4 (0.49–6.6)

Triglycerides (mmol/l) 2.6 (0.9–3.9) 2.9 (0.3–11.6) 2.4 (0.9–3.7) 2.9 (0.3–11.6) 2.5 (0.9–3.9) 2.9 (0.3–11.6)Treatment

Insulin 53.8 18.8 75.0 19.0 33.0 19.9Oral hypoglycemic agents 38.5 78.2 25.0 78.9 50.0 78.7Diet 7.7 18.6 0.0 18.8 16.7 18.6

Data are means (range) or percentages unless otherwise indicated. *P � 0.019, GAD65 autoantibody positive vs. GAD65 autoantibody negative; †P � 0.009, GAD65autoantibody positive vs. IA-2 autoantibody positive; ‡P � 0.029, autoantibody positive vs. autoantibody negative; §P � 0.032, autoantibody positive vs.autoantibody negative; �P � 0.046, IA-2 autoantibody positive vs. IA-2 autoantibody negative.

Britten and Associates


strom BO, Forsen B, Snickars B, Lahti K,Forsblom C, Saloranta C, Taskinen MR,Groop LC: Clinical and genetic character-istics of type 2 diabetes with and withoutGAD antibodies. Diabetes 48:150–157,1999

5. World Health Organization: Definition,Diagnosis and Classification of Diabetes Mel-litus and its Complications, Part 1: Diagnosisand Classification of Diabetes Mellitus: Re-port of the WHO Consultation. Geneva,World Health Org., 1999 (Tech. Rep.Ser., no. WHO/NCD/NCS/99.2)

6. Bunce M, O’Neill CM, Barnardo MC,Krausa P, Browning MJ, Morris PJ, WelshKI: Phototyping: comprehensive DNAtyping for HLA-A, B, C, DRB1, DRB3,DRB4, DRB5 & DQB1 by PCR with 144primer mixes utilizing sequence-specific

primers (PCR-SSP). Tissue Antigens 46:355–367, 1995

7. Heward JM, Mijovic CH, Kelly MA, Mor-rison E, Barnett AH: HLA-DQ and DRB1polymorphism and susceptibility to type1 diabetes in Jamaica. Eur J Immunogenet29:47–52, 2002

8. Julier C, Hyer RN, Davies J, Merlin F, Sou-larue P, Briant L, Cathelineau G, De-schamps I, Rotter JI, Froguel P, Boitard C,Bell J, Lathrop G: Insulin-IGF2 region onchromosome 11p encodes a gene impli-cated in HLA-DR4-dependent diabetessusceptibility. Nature 354:155–159, 1991

9. Gambelunghe G, Ghaderi M, CosentinoA, Falorni A, Brunetti P, Falorni A, San-jeevi CB: Association of MHC class Ichain-related A (MIC-A) gene polymor-phism with type I diabetes. Diabetologia

43:507–514, 200010. Bingley PJ, Bonifacio E, Mueller PW: Dia-

betes Antibody Standardization Program:first assay proficiency evaluation. Diabetes52:1128–1136, 2003

11. Das AK, Shtauvere-Brameus A, SanjeeviCB: GAD65 and ICA512 antibodies in un-dernourished and normally nourishedsouth Indian patients with diabetes. AnnN Y Acad Sci 958:247–250, 2002

12. Sanjeevi CB, Kanungo A, Berzina L,Shtauvere-Brameus A, Ghaderi M, SamalKC: MHC class I chain-related gene a al-leles distinguish malnutrition-modulateddiabetes, insulin-dependent diabetes, andnon-insulin- dependent diabetes mellituspatients from eastern India. Ann N Y AcadSci 958:341–344, 2002

LADA in South Asians


GAD Antibody in Multiplex DiabeticPedigrees of ChineseTAO CHEN, MD

1

YAN REN, PHD1

YANG LONG, MD2

XIANGXUN ZHANG2

HONGLIN YU2

HAOMING TIAN, MD1

Multiplex diabetic families may becaused by mutations in genes ofhepatocyte nuclear factors (matu-

r i ty-onset diabetes of the young[MODY]1, -3, and -5), glucokinase(MODY2), insulin promoter factor-1(MODY4), NeuroD1 (MODY6), mito-chondrion ND1 and tRNALeu(UUR), andsome other unknown genetic mutations.Recent studies showed that mutations ofMODY are not common causes of diabe-tes in Chinese diabetic families (1–3), in-dicating that there may be other causesand mechanisms involved. Latent auto-immune diabetes in adults (LADA) sharessome similar clinical phenotypes withMODY and is common in phenotypictype 2 diabetes (10–25%) (4). This studywas conducted to investigate the distribu-tion of GAD antibody (GADA) in multi-plex diabetic pedigrees from theChengDu area of China.

RESEARCH DESIGN ANDMETHODS — A total of 140 familymembers of 18 families were recruited. Inour previous study, most members hadbeen screened for variants in mitochon-drion ND1 and tRNALeu(UUR) and hadbeen proved to be without deficiencies inthese genes (5). Among them, 42 subjectsin four families had been further screenedin the hepatic nuclear factor-1� gene andhad been found to have no suspected vari-ant (T.C., Y.R., H.Y., X.Z., H.T., X. Cao,unpublished observations). Each familyhad three or more diabetic individuals in

at least two generations, and most pro-bands were diagnosed as diabetic be-tween ages 23 and 40 years. The probandshad a mean � SD BMI of 24.4 � 3.1kg/m2. Only one (5.6%) had hyperten-sion, and seven (38.9%) met the Interna-tional Diabetes Federation diagnosiscriteria of metabolic syndrome (6). Themean (range) age of whole family mem-bers was 42.8 � 18.1 years (2–83), andmean BMI was 23.6 � 3.4 kg/m2. Eigh-teen subjects (12.9%) had hypertension,and 30 (21.4%) met diagnosis criteria ofmetabolic syndrome. Another 50 unre-lated healthy subjects were recruited asnormal controls. All subjects underwentan oral glucose tolerance test and insulinrelease tests after 8 h of fasting. GADAswere measured with a radioimmunoassaykit (Beijing North Institute of BiologicalTechnology). Categorical variables werecompared by �2 tests. Continuous vari-ables were evaluated by Student’s t tests.

RESULTS — There were 59 diabeticpatients, 20 pre-diabetic subjects (includ-ing 10 with impaired glucose tolerance, 6with impaired fasting glucose, and 4 withboth impaired glucose tolerance and im-paired fasting glucose), and 61 subjectswith normal glucose tolerance (NGT) inthe 18 families. GADA presented in 38 of140 family members (27.1%), a percent-age higher than that of the normal controlgroup (6%); P � 0.002. Further analysisshowed that 21 of 59 (35.6%) diabeticpatients were positive for GADA, as were

4 of 20 (20%) pre-diabetic patients, 13 of61 NGT subjects (21.3%), and 3 of 50normal control subjects (6.0%). The dif-ference was significant (P � 0.002).When compared between groups, dia-betic patients had a higher prevalence ofGADA than found in the normal controlgroup (P � 0.000) and showed a ten-dency of higher prevalence of GADAcompared with pre-diabetic and NGTsubjects, though the difference was notsignificant. Pre-diabetic and NGT pa-tients also showed a high GADA-positiverate compared with the normal controlgroup, while no significant difference wasfound (Table 1).

There were 17 newly diagnosed and42 previously diagnosed diabetic pa-tients. The prevalence of GADA was 5 of17 (29.4%) in newly diagnosed and 16 of42 (38.1%) in previously diagnosed pa-tients. No significant difference wasfound (P � 0.05).

Of the 59 diabetic patients, 21 wereGADA positive and 38 GADA negative.No difference was found between thesetwo subgroups in age of onset (50.9 �13.0 vs. 50.0 � 12.1 years), BMI (23.4 �2.5 vs. 23.8 � 2.9 kg/m2), rate of meta-bolic syndrome (38.1 vs. 32.4%), orcourse of diabetes (4.1 � 5.5 vs. 5.8 � 7.3years).

CONCLUSIONS — Published stud-ies showed that GADA-positive fre-quencies in newly diagnosed pheno-typic type 2 diabetic patients were 10%in UK Prospective Diabetes Study 25 (7)and 7.1% in a study of a Chinese pop-ulation (8). Our study found that GADAwas present in 36.1% of diabetic pa-tients in our nonmitochondrion ND1and tRNAL e u ( U U R ) gene var iant–predisposed pedigrees, a rate higherthan that in the above studies (P �0.000). GADA was one established se-rological marker indicating autoim-mune damage to islet cells. GADAcombined with type 2 diabetic pheno-type could lead to the diagnosis of LADA(4). Thus, more than one-third of diabeticpatients in our diabetic families could beLADA patients, which is more commonthan the rate observed in sporadic pheno-typic type 2 diabetic patients. Whethergenetic or environmental factors caused

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Endocrinology, West China Hospital of Sichuan University, Sichuan, China; andthe 2Laboratory of Endocrinology and Metabolism, West China Hospital of Sichuan University, Sichuan,China.

Address correspondence and reprint requests to Haoming Tian, MD, Department of Endocrinology, WestChina Hospital, Sichuan University, 37 GuoXue Street, Chengdu, Sichuan 610041, China. E-mail:[email protected].


0954.Abbreviations: GADA, GAD antibody; LADA, latent autoimmune diabetes in adults; MODY, maturity-

onset diabetes of the young; NGT, normal glucose tolerance.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





this gathering phenomenon remains to beestablished.

In addition, our study found thatGADA-positive and -negative subjects indiabetic families did not show a differencein age of onset, diabetes course, BMI, ormetabolic syndrome rate. This may bedue to our limited case number or, as de-scribed by Juneja et al., because only an-tibodies, mainly islet cell antibody andGADA, identify LADA, while age and BMIdo not (9).

In summary, we have shown a highfrequency of GADA in 18 non–insulin de-pendent diabetic pedigrees, which sug-gests that LADA is common in thesespecial populations and that auto-inmmune �-cell destruction might be an-other etiology in apparently genedeficient–predisposed diabetic pedigrees.It is necessary to detect GADA in patientswith a multiplex diabetes family history toprovide better diagnosis and treatment.

Acknowledgments— This study was sup-ported by a grant for Scientific Research from

the Ministry of Education, Beijing, China(20030610073).

References1. Fang QC, Zhang R, Wang CR, Lin X,

Xiang KS: Scanning HNF-1� gene muta-tion in Chinese early-onset and/or multi-plex diabetes pedigrees. Zhonghua Yi XueYi Chuan Xue Za Zhi 21:329–334, 2004

2. Zhang R, Hu C, Wang CR, Fang QC, MaXJ, Jia WP, Xiang KS: Scanning theHNF4� gene mutation from Chinese ped-igrees with early-and/or multiple-onsetdiabetes. Zhonghua Yi Xue Yi Chuan Xue ZaZhi 23:406–409, 2006

3. Zheng TS, Wu SH, Yang Z, Lu HJ, XiangKS: Mutation screening of GCK gene inChinese early-onset diabetes population.Zhonghua Yi Xue Yi Chuan Xue Za Zhi 22:671–674, 2005

4. Stenstrom G, Gottsater A, Bakhtadze E,Berger B, Sundkvist G: Latent autoim-mune diabetes in adults: definition, prev-alence, �-cell function, and treatment.Diabetes 54 (Suppl. 2):S68–S72, 2005

5. Ren Y, Li XJ, Tian HM, Liang JZ, Han LC,Zhang XX, Yu HL, Yu YR, Liu R, Zhao GZ,Wang JN: Molecular scanning of candi-

date mtDNA gene fragment in diabeticpedigrees. ZhongHua Yi Xue Yi Chuan XueZa Zhi 20:181–185, 2003

6. Alberti KG, Zimmet P, Shaw J; IDF Epide-miology Task Force Consensus Group:The metabolic syndrome: a new world-wide definition. Lancet 366:1059–1062,2005

7. Turner R, Stratton I, Horton V, Manley S,Zimmet P, Mackay IR, Shattock M, Bot-tazzo GF, Holman R: UKPDS 25: Autoan-tibodies to islet-cell cytoplasm andglutamic acid decarboxylase for predic-tion of insulin requirement in type 2 dia-betes. Lancet 350:1288–1293, 1997

8. Li X, Zhou ZG, Huang G, Peng J, Yan X,Yang L, Wang JP, Deng ZM: Study on thepositive frequency and distribution ofglutamic acid decarboxylase antibody inphenotypic type 2 diabetic patients. Chin JEpidemiol 26:800–803, 2005

9. Juneja R, Hirsch IB, Naik RG, Brooks-Worrell BM, Greenbaum CJ, Palmer JP:Islet cell antibodies and glutamic aciddecarboxylase antibodies, but not theclinical phenotype, help to identify type1(1/2) diabetes in patients presentingwith type 2 diabetes. Metabolism 50:1008 –1013, 2001

Table 1—Frequencies of GADA in subgroups of diabetic pedigrees and normal control subjects

Diabetic pedigrees Normal controlsubjects P*Diabetes Pre-diabetes NGT

n 59 20 61 50GADA 35.6 (21)† 20.0 (4) 21.3 (13) 6.0 (3) 0.003

Data are % (n) unless otherwise indicated. *P � 0.003, compared among the four groups; †P � 0.000, compared with the normal control group.

Detecting GADA in diabetic pedigrees


Intrahepatic Fat Accumulation andAlterations in Lipoprotein Composition inObese AdolescentsA perfect proatherogenic state

ANNA M.G. CALI, MD1

TOSCA L. ZERN, PHD1

SARA E. TAKSALI, MPH1

ANA MAYRA DE OLIVEIRA, MD2

SYLVIE DUFOUR, PHD3

JAMES D. OTVOS, PHD4

SONIA CAPRIO, MD1

OBJECTIVE — Among other metabolic consequences, a dyslipidemic profile often accom-panies childhood obesity. In adults, type 2 diabetes and hepatic steatosis have been shown toalter lipoprotein subclass distribution and size; however, these alterations have not yet beenshown in children or adolescents. Therefore, our objective was to determine the effect of hepaticsteatosis on lipoprotein concentration and size in obese adolescents.

RESEARCH DESIGN AND METHODS — Using fast magnetic resonance imaging, wemeasured intrahepatic fat content in 49 obese adolescents with normal glucose tolerance. Thepresence or absence of hepatic steatosis was determined by a threshold value for hepatic fatfraction (HFF) of 5.5%; therefore, the cohort was divided into two groups (HFF � or �5.5%).Fasting lipoprotein subclasses were determined using nuclear magnetic resonance spectroscopy.

RESULTS — Overall, the high-HFF group had 88% higher concentrations of large VLDLcompared with the low-HFF group (P � 0.001). Likewise, the high-HFF group had significantlyhigher concentrations of small dense LDL (P � 0.007); however, the low-HFF group hadsignificantly higher concentrations of large HDL (P � 0.001). Stepwise multiple regressionanalysis revealed that high HFF was the strongest single correlate, accounting for 32.6% of thevariance in large VLDL concentrations (P � 0.002).

CONCLUSIONS — The presence of fatty liver was associated with a pronounced dyslipide-mic profile characterized by large VLDL, small dense LDL, and decreased large HDL concentra-tions. This proatherogenic phenotype was strongly related to the intrahepatic lipid content.

Diabetes Care 30:3093–3098, 2007

S tudies from autopsies on 742 chil-dren (aged 2–19 years) reportedfatty liver prevalence at 9.6%, and in

obese children this rate increased to analarming 38% (1). An imbalance betweenfatty acid flux and utilization and VLDL

secretion leads to an accumulation of trig-lycerides within the hepatocytes and ulti-mately to hepatic steatosis (2). It isbecoming increasingly clear that fat accu-mulation in the liver, per se, is not a be-nign condition (3). Indeed, it is frequently

associated with type 2 diabetes in bothadults and children (4,5) and has beenlabeled as the hepatic component of themetabolic syndrome (2,3).

Worsening of the dyslipidemic pro-file has been described in adults in asso-ciation with insulin resistance and type 2diabetes (6–8). Garvey et al. (7) haveshown that subjects with type 2 diabeteshave larger VLDL and smaller LDL andHDL particles compared with insulin-sensitive subjects. The insulin-resistantand type 2 diabetic groups also hadgreater concentrations of these athero-genic particles. Further studies by Toledoet al. (8) reported that the presence of he-patic steatosis in obese subjects with type2 diabetes further altered lipoproteincomposition compared with type 2 dia-betic subjects without fatty liver. Type 2diabetic subjects with fatty liver hadlarger triglyceride-rich VLDL particles,smaller LDL and HDL particles, and re-duced concentrations of large LDL com-pared with type 2 diabetic subjectswithout fatty liver (8).

Obese children and adolescents areoften diagnosed with dyslipidemia char-acterized by high triglycerides and lowHDL cholesterol concentrations. In addi-tion, the presence of small dense LDL par-ticles has been shown in obese children(9,10). Recent studies from our group re-ported dyslipidemia and a deteriorationin glucose metabolism in obese nondia-betic adolescents with excessive intrahe-patic fat accumulation. In particular, wefound rising levels of triglycerides and de-creasing levels of HDL cholesterol withincreasing accumulation of fat in the liver(11). Although studies in adults haveshown insulin resistance, obesity, andfatty liver playing a role in the composi-tion of lipoproteins, there are no currentstudies for this comprehensive phenotypein children. Therefore, our objectiveswere 1) to determine whether obese nor-mal glucose-tolerant adolescents withfatty liver had alterations in lipoproteincomposition and size compared withobese normal glucose-tolerant adoles-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Pediatrics, Yale University School of Medicine, New Haven, Connecticut; the2Department of Health, State University of Feira de Santana, Feira De Santana, Brazil; the 3Department ofInternal Medicine and the Howard Hughes Institute, Yale University School of Medicine, New Haven,Connecticut; and 4LipoScience, Raleigh, North Carolina.

Address correspondence and reprint requests to Sonia Caprio, MD, Yale University School of Medicine,Department of Pediatrics, 330 Cedar St., P.O. Box 208064, New Haven, CT 06520. E-mail: [email protected].

Received for publication 7 June 2007 and accepted in revised form 17 August 2007.Published ahead of print at http://care.diabetesjournals.org on 23 August 2007. DOI: 10.2337/dc07-

1088.Abbreviations: EMCL, extramyocellular triglyceride content; HFF, hepatic fat fraction; IMCL, intramyo-

cellular triglyceride content; MRI, magnetic resonance imaging; NMR, nuclear magnetic resonance; WBISI,whole-body insulin sensitivity index.




C a r d i o v a s c u l a r a n d M e t a b o l i c R i s kO R I G I N A L A R T I C L E


cents without fatty liver, matched for thedegree of overall obesity and age; and 2)to examine to what degree differences, ifany, in the lipoprotein composition maybe accounted for by the level of intrahe-patic fat accumulation. Nuclear magneticresonance (NMR) spectroscopy was uti-lized to determine lipoprotein subclasscomposition in fasting plasma samples,while fast magnetic resonance imaging(MRI) was used to determine the hepaticfat fraction (fatty liver).

RESEARCH DESIGN ANDMETHODS — We recruited 49 obeseadolescents from our Pediatric ObesityClinic. Some subjects are part of a largerstudy on the prevalence of fatty liver dis-ease in youth and thus were reported onpreviously (11). All subjects had a BMIgreater than the 95th percentile, were tak-ing no medications known to affect liverfunction or alter glucose or lipid metabo-lism, and all denied the use of alcohol.The Yale University School of MedicineHuman Investigation Committee ap-proved the study, and written informedconsent and assent were obtained.

Metabolic studiesOral glucose tolerance test. All subjectswere invited to the Yale Center for Clini-cal Investigation for an oral glucose tol-erance test at 8:00 A.M., following anovernight fast, as previously reported(12). The subjects were instructed to con-sume a diet containing at least 250 g car-bohydrate the day before the study and torefrain from vigorous physical activity.Baseline blood samples were obtainedfrom subjects while they were fasting,with the use of an indwelling venous linefor measurement of glucose, insulin, lipidprofile, lipoprotein concentration andsize, free fatty acids, adiponectin, and lep-tin. A standard 3-h oral glucose tolerancetest was then performed with the admin-istration of 1.75 g glucose/kg body wt(maximum dose: 75 g), and blood sam-ples were obtained every 30 min forplasma glucose, insulin, and C-peptidemeasurements.

Lipoprotein analysisFasting plasma samples were obtained todetermine lipoprotein particle concentra-tion and size. The analysis utilized a 400MHz proton NMR analyzer at Liposcience(Raleigh, NC). The methodology has beendescribed in detail (13,14). In brief, eachlipoprotein subclass concentration wasdetermined by the measured amplitudes

of the characteristic lipid methyl groupNMR signals they emit (15). The intensityof each signal is proportional to the quan-tity of the subclass, which is reported inparticle concentration units (nanomolesof particles per liter for VLDL and LDLand micromoles per liter for HDL). VLDL,LDL, and HDL were separated into 10subclass categories: large VLDL (includ-ing chylomicrons) (�60 nm), mediumVLDL (35–60 nm), small VLDL (27–35nm), intermediate-density lipoprotein(23–27 nm), large LDL (21.2–23 nm),medium-small LDL (19.8–21.2 nm), verysmall LDL (18 –19.8 nm), large HDL(8.8 –13 nm), medium HDL (8.2– 8.8nm), and small HDL (7.3–8.2 nm). Aver-age particle sizes were computed as thesum of the diameter of each subclass mul-tiplied by its relative mass percentage asestimated from the amplitude of itsmethyl NMR signal.

Imaging studiesFast MRI: liver fat content. Measure-ment of hepatic fat accumulation was per-formed using MRI along with the Dixonmethod as modified by Fishbein et al.(16). The description of the method hasbeen reported recently by our group (11).Following the analysis of hepatic fat frac-tion (HFF), subjects were stratified intotwo groups: HFF �5.5% (n � 37) and�5.5% (n � 12). This cutoff has beenused previously by our research groupand is a threshold to denote steatosis inthis population (11,17). Secondary causesof fatty liver, such as autoimmune hepa-titis, Wilson disease, �1-antitripsin defi-ciency, and hepatitis B and C, wereexcluded with appropriate tests.

Abdominal fat distribution: MRIAbdominal MRI studies were performedon a Siemens Sonata 1.5 Tesla system, aspreviously described (12).

1H-NMR spectroscopy:intramyocellular triglyceride contentMuscle triglyceride content was measuredusing a 4.0T Biospec system (Bruker Bio-spin MRI, Ettlingen, Germany), as previ-ously described (18).

Limitations of the imagingtechniquesLiver MRI. This method is limited to pix-els with an HFF of �50%; however, otherstudies (17) using various methods haveshown that values �50% are rare. In ad-dition, we measured only a single slice ofthe liver, which in some cases may not

reflect the fat content of the liver as awhole. Despite these limitations, the two-point Dixon method is the most widelyused technique in clinical MRI studies. Inour group, we validated the modifiedDixon method against hepatic fat mea-sured by 1H-NMR in 34 subjects (leanand obese) and found a very strong cor-relation between the two methods (r �0.93, P � 0.001) (T. Constable and S.C.,personal data). Furthermore, at our insti-tution, four obese adolescents underwenta liver biopsy—the gold standard for di-agnosing nonalcoholic fatty liver dis-ease—which confirmed fatty livermeasured by fast MRI.Abdominal MRI. This technique givesan estimate of abdominal fat content,which is dependent on the sensitivity ofthe images, the intensity threshold, andthe reader’s ability; this method has beenvalidated against dissection in human ca-davers (19). As with the liver MRI, weonly measure one slice (at the level of theL4/L5 disc space); however, single-slicemethods have been shown to correlatewell with multislice methods (20).1H-NMR spectroscopy. Single-voxelmagnetic resonance spectroscopy is themost accurate way to determine musclelipids, and it can distinguish intra- versusextramyocellular lipids. The disadvantageis that it only measures a single site andtherefore cannot give information on heter-ogeneity across a tissue. Moreover, toensure good separation of the extramyo-cellular triglyceride content (EMCL)intramyocellular triglyceride content(IMCL) resonances, the muscle fibersmust be aligned along the Z direction ofthe magnet to avoid contamination of theIMCL peak with the EMCL signal. Thisrequires the use of a smaller voxel to ob-tain a better fiber alignment. Overall, theimaging techniques have the advantage ofbeing noninvasive without the use of ion-izing radiation; however, they are rela-tively expensive and therefore notappropriate for routine screening.

Analytical methodsPlasma glucose levels were measured us-ing the YSI 2700 STAT Analyzer (YellowSprings Instruments) and lipid levels us-ing an autoanalyzer (model 747-200;Roche-Hitachi). Plasma insulin, leptin,and adiponectin levels were measured us-ing an radioimmunoassay from Millipore(St. Charles, MO). Free fatty acids weremeasured using a Wako Diagnostics as-say. Estimated insulin sensitivity was cal-culated using the Matsuda index (whole-

Lipoprotein alterations in hepatic steatosis


body insulin sensitivity index [WBISI]),which has been validated by comparisonwith hyperinsulinemic-euglycemic clampstudies in obese children and adolescents(21).

StatisticsData are represented as means � SE. Pa-rameters that were not normally distrib-uted were log transformed for analysis.Independent t tests were used to analyzedifferences between groups. Univariateanalysis was utilized to adjust for poten-tial confounders (race and sex). To furtherexamine the independent association be-tween large VLDL particle concentrationsand HFF, we used a stepwise forwardmultiple regression analysis in the totalcohort. In step 1, we entered sex and race;in step 2, visceral fat was added; in step 3,insulin sensitivity (WBISI) was added;and finally in step 4, HFF was added tothe model. A P value of �0.05 was con-sidered statistically significant. All analy-ses were performed using SPSS 14.0 forWindows.

RESULTS

Demographic and anthropometriccharacteristicsAs shown in Table 1, a total of 49 maleand female obese adolescents were in-cluded in the study. It is evident that therewere large discrepancies in race and sex inthe cohort, and HFF varied widely, from

undetectable to 37.3%. The wide range inHFF allowed for stratification into twogroups: HFF �5.5% (n � 37) and HFF�5.5% (n � 12). There were no differ-ences in age, BMI, or BMI z scores be-tween the two groups. Although malesubjects were equally represented, �85%of the female subjects were included inthe group with the low HFF (P � 0.001).Likewise, there were significantly moreCaucasians and no African Americans inthe high-HFF group (P � 0.033). Due tothese significant differences, all analyseswere statistically adjusted for race andsex.

Although there were no differences inpercent body fat, visceral adiposity wassignificantly higher in the high-HFFgroup compared with the low-HFF group(87.3 � 6.6 vs. 56.0 � 4.3 cm2; P �0.001). There were no differences in sub-cutaneous fat between groups; however,due to the differences in visceral adipos-ity, there were also significant differencesin the ratio of visceral to subcutaneous fatdepots. In addition, the high-HFF grouphad elevated IMCL concentrations com-pared with the low-HFF group (1.8 �0.23 vs. 1.1 � 0.15; P � 0.01).

Metabolic characteristicsMetabolic characteristics are shown in Ta-ble 2. All subjects included in the studyhad normal glucose tolerance (2 h glucose�7.7 mmol/l [140 mg/dl]). Despite com-parable fasting glucose concentrations,

the high-HFF group had significantlyhigher concentrations of fasting insulin(41.4 � 1.13 vs. 26.3 � 1.07; P � 0.002)compared with the low-HFF group. Bothgroups were insulin resistant, as illus-trated by both homeostasis model assess-ment of insulin resistance and WBISIcalculations; however, the high-HFFgroup had significantly higher homeosta-sis model assessment of insulin resistancelevels and lower WBISI levels comparedwith the low-HFF group. Leptin concen-trations were similar among the groups,whereas the low-HFF group had higherconcentrations of adiponectin comparedwith the high-HFF group. This differenceremained after controlling for age and sexdifferences.

Plasma lipids and lipoproteincomposition and sizeA standard lipid panel revealed plasmalipid alterations in the high-HFF group(Table 2). There were no differences be-tween groups with regards to total choles-terol and LDL cholesterol concentrations.As expected, the high-HFF group had sig-nificantly higher triglyceride concentra-tions than the low-HFF group (P �0.001). HDL cholesterol concentrationswere also significantly lower in the high-HFF group (P � 0.006). Free fatty acidconcentrations were not different be-tween groups.

A more complete lipoprotein analysisrevealed significant alterations in both li-

Table 1—Demographic and anthropometric characteristics

Total cohort HFF �5.5% HFF �5.5% Unadjusted Adjusted*

n 49 37 12Age (years) 15.3 � 0.33 15.32 � 0.4 15.1 � 0.7 0.738 —Sex

Male 17 10 7 0.467 —Female 32 27 5 �0.001 —

RaceWhite 22 16 6 0.033 —Hispanic 18 12 6 0.439 —Black 9 9 0 —

BMI (kg/m2) 35.7 � 0.9 35.63 � 1.12 36.0 � 1.17 0.862 —BMI Z score 2.24 � 0.05 2.21 � 0.06 2.35 � 0.06 0.191 —% fat 43.0 � 1.0 42.0 � 1.3 43.3 � 2.0 0.510 —% HFF 6.83 � 1.61 1.15 � 0.23 24.31 � 3.0 �0.001 —Visceral fat (cm2) 64.32 � 3.5 56.0 � 4.3 87.3 � 6.6 �0.001 �0.001Subcutaneous fat (cm2) 558.0 � 28.0 568.0 � 39.6 531.1 � 61.0 0.415 0.616Visceral-to-subcutaneous fat ratio 0.124 � 0.01 0.106 � 0.01 0.171 � 0.014 �0.001 �0.001IMCL (%) 1.33 � 0.12 1.1 � 0.15 1.8 � 0.23 0.01 0.01EMCL (%) 1.6 � 0.14 1.42 � 0.18 1.93 � 0.27 0.118 0.126

Data are means � SE. *Adjusted for sex and race; P � 0.05.

Cali and Associates


poprotein subclass particle concentrationand size in the high-HFF group. With re-gard to VLDL, the high-HFF group had�25% more VLDL particles and 88%higher concentrations of large VLDL com-pared with the low-HFF group (P �0.021 and P � 0.001, respectively). Themedium VLDL particle concentration was�50% higher in the high-HFF group(P � 0.001). No differences were ob-served in the small VLDL particles. Al-though there was no difference in the totalnumber of LDL particles between groups,the high-HFF group had significantlyhigher concentrations of small dense LDL(P � 0.007). Likewise, total HDL particleconcentration was not different betweengroups, but the low-HFF group had sig-nificantly higher concentrations of largeHDL and lower concentrations of me-dium HDL particles compared with thehigh-HFF group (P � 0.001 and P �0.005, respectively). Overall size of thelipoproteins followed a similar trend. Asexpected, the high-HFF group had a

larger VLDL particle size and smaller LDLand HDL particle sizes compared with thelow-HFF group.

Relationship between HFF, insulinsensitivity, body fat distribution,and large VLDL concentrationsTo assess the relationships between largeVLDL and HFF, insulin sensitivity, andbody fat distribution, stepwise multipleregression analysis was used. From ourmodel, it is evident that large VLDL con-centrations are independently associatedwith fatty liver. Furthermore, HFF ex-plains �32.6% of the variance in largeVLDL concentrations (P � 0.002). Noother parameter was significantly associ-ated with large VLDL concentrations.

CONCLUSIONS — By combiningNMR spectroscopy to assess lipoproteincomposition with fast MRI to quantifyliver fat content, we demonstrate, in thepresent study, that obese adolescents withnormal glucose tolerance and fatty liver

have the prototypic proatherogenic li-poprotein phenotype. In particular, wefound that the presence of hepatic steato-sis was associated with 1) an increase inVLDL particle size and number, 2) an in-crease in small dense LDL concentrations,and 3) a decrease in the number of largeHDL particles. These alterations werereflected by an increase in triglycerideconcentrations and decreased HDL cho-lesterol. Of note, hepatic steatosis wasfound to predict the concentration of thelarge VLDL particles, independent ofoverall adiposity, insulin sensitivity, andvisceral adiposity, thereby suggesting thatliver steatosis is important in the earlypathogenesis of insulin resistance andtype 2 diabetes in youth. Hence, theatherogenic profile is already fully estab-lished at this very young age.

It is widely appreciated that hepaticoverproduction of VLDL constitutes themetabolic basis of various hyperlipidemicstates in humans, such as the familialcombined hyperlipidemia and the dyslip-

Table 2—Metabolic, plasma lipids, and lipoprotein profile in the entire cohort stratified by HFF (%)

HFF �5.5% HFF �5.5% Unadjusted Adjusted*

n 37 12Fasting glucose (mmol/l) 5.12 � 0.07 5.01 � 0.12 0.946 0.449Fasting insulin (pmol/l) 150.0 � 6.6 246.0 � 6.78 0.002 0.002Homeostasis model assessment of insulin resistance 5.65 � 1.1 9.02 � 1.13 0.003 0.003WBISI 2.05 � 1.1 1.19 � 1.15 0.002 0.002Adiponectin (�g/ml) 9.94 � 0.67 6.5 � 1.1 0.004 0.01Leptin (ng/ml) 27.9 � 2.5 27.4 0.276 0.937Free fatty acids (mmol/l) 0.463 � 0.028 0.517 � 0.055 0.337 0.397Plasma lipids (mmol/l)

Total cholesterol 3.76 � 0.026 3.84 � 0.03 0.788 0.909LDL cholesterol 2.24 � 0.16 2.00 � 0.24 0.438 0.425HDL cholesterol 1.19 � 0.04 0.95 � 0.07 0.006 0.006Triglycerides 0.73 � 0.021 1.62 � 0.015 �0.001 �0.001

VLDL particles (nmol/l)Total VLDL and chylomicrons 41.1 � 3.2 55.03 � 4.9 0.012 0.021Large VLDL and chylomicrons 0.77 � 0.749 6.25 � 1.15 �0.001 �0.001Medium VLDL 8.7 � 1.3 18.7 � 2.0 �0.001 �0.001Small VLDL 32.0 � 2.6 30.1 � 4.0 0.821 0.754

IDL particles (nmol/l) 24.0 � 5.43 39.4 � 8.34 0.058 0.126LDL particles (nmol/l)

Total LDL 938.4 � 51.0 1,076.46 � 78.0 0.084 0.146Large LDL 344.5 � 31.0 255.4 � 47.3 0.062 0.123Total small LDL 570.2 � 41.0 781.4 � 63.0 0.002 0.007Medium-small LDL 119.8 � 9.52 150.4 � 14.6 0.032 0.087Very small LDL 450.2 � 32.65 631.1 � 50.15 0.001 0.004

HDL particles (�mol/l)Total HDL 27.0 � 0.7 25.4 � 1.1 0.642 0.271Large HDL 5.2 � 0.34 2.8 � 0.52 �0.001 �0.001Medium HDL 2.2 � 0.54 5.2 � 0.84 0.004 0.005Small HDL 19.4 � 0.6 17.3 � 0.851 0.220 0.052

Data are means � SE. *Adjusted for sex and race; P � 0.05.



idemia of type 2 diabetes (22). AlthoughLDL subclasses have received the most at-tention, subclasses of VLDL may also dif-fer in atherogenicity. It is possible thatlarge VLDL particles may be selectivelyretained in the intima of the arterial wallor may be a marker of delayed chylomi-cron clearance, a metabolic condition thathas been related to disease severity (23).In the high-HFF group, we found amarked increase in large VLDL and, to alesser extent, medium VLDL and no dif-ferences regarding small VLDL comparedwith the low-HFF group. Large VLDLparticles are triglyceride rich and are ex-cellent substrates for cholesterol estertransfer protein. Cholesterol ester transferprotein is a key enzyme in the reversecholesterol transport system, whose activ-ity is mediated by substrate availability(24). In these instances of hypertriglycer-idemia, there is an increase in the exchangeof cholesterol ester and triglycerides viacholesterol ester transfer protein betweentriglyceride-rich lipoproteins and HDL orLDL. This interaction yields triglyceride-rich LDL and HDL particles that can behydrolyzed by hepatic lipase, thus pro-moting the formation of small dense LDLand decreased large HDL (22). The im-portance of triglyceride levels likely re-flects the exchange of triglyceride andcholesterol esters between VLDL and LDLparticles, with the subsequent hydrolysisof triglycerides. Triglyceride levels arestrongly related to the size of VLDL parti-cles and to the relative amount of eachVLDL subclass, and the relative propor-tion of large VLDL increases rapidly athigher triglyceride levels (24). It is wellknown that the LDL receptor has a de-creased affinity for smaller particles, and,therefore, particles are left in circulation(25). However, triglyceride-enrichedHDL has been shown to be cleared morerapidly from circulation (26). This may bethe case in the high-HFF group, wherethere was an �50% reduction in largeHDL subclasses and a 27% increase in to-tal small LDL particles compared with thelow-HFF group.

It is noteworthy that had the lipopro-tein subclasses not been measured by theNMR technique, we would have missedthe important finding regarding the pat-tern of changes in the LDL subclassespresent in these youngsters with fattyliver. Indeed, the traditional fasting lipidprofile revealed normal LDL cholesterolconcentrations in both groups. In con-trast, we found a significant increase insmall LDL particles with increasing liver

fat content. Small dense LDL is known tobe proatherogenic; they are more suscep-tible to oxidation and may be taken up bymacrophages, which eventually leads tothe development of atheroscleroticplaque formation in the arterial wall.

The obese adolescents with fatty liveralso had a greater visceral fat depot andhigher IMCL and were more insulin resis-tant than their matched controls. Inadults, a strong association between fattyliver and visceral adiposity has been re-ported, but no associations have been re-ported with IMCL content (27–29).Petersen et al. (30) reported both intrahe-patic triglyceride and IMCL content to beincreased in Asian-Indian men comparedwith Caucasian men. However, after ad-justing for insulin sensitivity, the Asian-Indian men had more than a twofoldincrease in hepatic triglyceride contentcompared with the Caucasian men,whereas the differences in the amount ofIMCL between the groups did not persist.In the present study, no significant rela-tion between HFF (fatty liver) and visceralfat (r2 � 0.232, P � 0.108 data notshown) was found. This, however, may bedue to the small sample size and ratherhomogeneous group of obese adoles-cents. In an attempt to discern the rela-t i onsh ips be tween la rge VLDLconcentrations and fatty liver, we per-formed a stepwise regression analysis.HFF was the strongest single correlate, ac-counting for 32.6% of the variance in theVLDL concentration.

The disturbances in triglyceride me-tabolism may, in part, explain the risk offuture cardiovascular disease. Recently,Godsland et al. (18) showed that inadults, triglyceride concentrations are astrong correlate of ethnic differences inischemic heart disease risk. In particular,they showed that both medium and largeVLDL levels were significantly higher inCaucasian compared with African-American men and women. Triglycerideconcentrations predict ischemic heartdisease, even though triglycerides per sedo not seem to be directly involved in theatherogenic process (30). Moreover, Herdet al. (31) found that African-Americanchildren have lower triglyceride concen-trations than Caucasian children, but dif-ferences in visceral fat did not explain thisresult, and VLDL concentrations rosemore slowly with increasing waist cir-cumference in African-American com-pared with Caucasian children.

The marked differences in the li-poprotein composition between the

groups with and without steatosis cannotbe accounted for by unequal sex and eth-nic distribution, since we have adjustedfor these variables during the analysis.

Triglyceride levels can quantitativelyand qualitatively affect circulation. Hy-pertriglyceridemia results in the accumu-lat ion of excess tr iglyceride-r ichlipoproteins including chylomicrons,VLDL, and their remnants. Interestingly,the use of thiazolidinediones has been as-sociated with changes in the lipoproteinsubclass particles, which to some extentmay be related to their increased risk ofcoronary artery disease (32).

In summary, among a small group ofobese adolescents with normal glucosetolerance, the presence of fatty liver wasassociated with a pronounced dyslipide-mic profile characterized by large VLDL,small LDL, and decreased large HDLconcentrations. This proatherogenic phe-notype was strongly related to the intra-hepatic lipid content. The coexistence offatty liver with severe insulin resistanceand dyslipidemia may represent the un-derlying metabolic defects that could pre-cede the onset of type 2 diabetes in theseyoungsters.

Acknowledgments— This study was sup-ported by National Institutes of Health GrantsR01-HD40787, R01-HD28016, and K24-HD01464 (to S.C.); M01-RR00125 (to theYale General Clinical Research Center); andR01-EB006494 (Bioimage Suite).

We are grateful to all of the adolescents whoparticipated in the study, to the researchnurses for the excellent care given to our sub-jects, and to Aida Groszmann, Andrea Belous,and Codruta Todeasa for their superb techni-cal assistance.

The authors had full access to the data andtake responsibility for its integrity. All authorshave read and agree to the manuscript as written.

References1. Schwimmer JB, Deutsch R, Kahen T, La-

vine JE, Stanley C, Behling C: Prevalenceof fatty liver in children and adolescents.Pediatrics 118:1388–1393, 2006

2. Sanyal AJ: Mechanisms of disease: patho-genesis of nonalcoholic fatty liver disease.Nat Clin Prac Gastro Hep 2:46–53, 2005

3. Angulo P: Nonalcoholic fatty liver disease.N Engl J Med 346:1221–1231, 2002

4. Vozarova B, Stefan N, Lindsay RS, SaremiA, Pratley RE, Bogardus C, Tataranni PA:High alanine aminotransferase is associ-ated with decreased hepatic insulin sensi-tivity and predicts the development oftype 2 diabetes. Diabetes 51:1889–1895,2002

Cali and Associates


5. Nadeau KJ, Klingensmith G, Zeitler P:Type 2 Diabetes in children is frequentlyassociated with elevated alanine amino-transferase. J Pediatr Gastroenterol Nutr41:94–98, 2005

6. Festa A, Williams K, Hanley AJG, OtvosJD, Goff DC, Wagenknecht LE, HaffnerSM: Nuclear magnetic resonance lipopro-tein abnormalities in prediabetic subjectsin the Insulin Resistance AtherosclerosisStudy. Circulation 111:3465–3472, 2005

7. Garvey WT, Kwon S, Zheng D, Shaugh-nessy S, Wallace P, Hutto A, Pugh K, Jen-kins AJ, Klein RL, Liao Y: Effects of insulinresistance and type 2 diabetes on lipopro-tein subclass particle size and concentra-tion determined by nuclear magneticresonance. Diabetes 52:453–462, 2003

8. Toledo FGS, Sniderman AD, Kelley DE:Influence of hepatic steatosis (fatty liver)on severity and composition of dyslipide-mia in type 2 diabetes. Diabetes Care 29:1845–1850, 2006

9. Kang H-S, Gutin B, Parbeau P, LItakerMS, Allison J, Lee N-A: Low-density li-poprotein particle size, central obesity,cardiovascular fitness, and insulin resis-tance syndrome markers in obese youths.Int J Obes 26:1030–1035, 2002

10. Miyashita M, Okada T, Kuromori Y,Harada K: LDL particle size, fat distribu-tion, and insulin resistance in obese chil-dren. Eur J Clin Nut 60:416–420, 2006

11. Burgert TS, Taksali SE, Dziura J, Good-man TR, Yeckel CW, Papademetris X,Constable RT, Weiss R, Tamborlane WV,Savoye M, Seyal AA, Caprio S: Alanineaminotransferase levels and fatty liver inchildhood obesity: associations with insu-lin resistance, adiponectin, and visceralfat. J Clin Endocrinol Metab 91:4287–4294, 2006

12. Weiss J, Dziura J, Burgert TS, TamborlaneWV, Taksali SE, Yeckel CW, Allen K,Lopes M, Savoye M, Morrison J, SherwinRS, Caprio S: Obesity and metabolic syn-drome in children and adolescents. N EnglJ Med 350:3262–3274, 2004

13. Otvos JD, Jeyarajah EJ, Bennett DW,Krauss RM: Development of a proton nu-clear magnetic resonance spectroscopicmethod for determining plasma lipo-protein concentrations and subspeciesdistributions from a single, rapid mea-surement. Clin Chem 38:1632–1638, 1992

14. Otovs JD: Measurement of lipoproteinsubclass profiles by nuclear magnetic res-onance spectroscopy. In Handbook of Li-poprotein Testing. Rifai N, Wainick R,Cominazak M, Eds. Washington, DC,AACC Press, 2000, p. 609–623

15. Jeyarajah EJ, Cromwell WC, Otvos JD: Li-poprotein particle analysis by nuclearmagnetic resonance spectroscopy. ClinLab Med 26:847–870, 2006

16. Fishbein MH, Gardner KG, Potter CJ,Schmalbrock P, Smith MA: Introductionof fast MR imaging in the assessment ofhepatic steatosis. Magn Reson Imaging 15:287–293, 1997

17. Szczepaniak LS, Nurenberg P, Leonard D,Browning JD, Reingold JS, Grundy S,Hobbs HH, Dobbins RL: Magnetic reso-nance spectroscopy to measure hepatictriglyceride content: prevalence of hepaticsteatosis in the general population. Am JPhysiol Endocrinol Metab 288:E462–E468,2005

18. Abate N, Burns D, Peshock RM, Garg A,Grundy SM: Estimation of adipose tissuemass by magnetic resonance imaging: val-idation against dissection in human ca-davers. J Lipid Res 35:1490–1496, 1994

19. Ross R, Leger L, Morris D, de Guise J,Guardo R: Quantification of adipose tis-sue by MRI: relationship with anthropo-metric variables. J Appl Physiol 72:787–795, 1992

20. Yeckel CW, Weiss R, Dziura J, Taksali SE,Dufour S, Burgert TS, Tamborlane WV,Caprio S: Validation of insulin sensitivityindices from oral glucose tolerance testparameters in obese children and adoles-cents. J Clin Endocrinol Metab 89:1096–1101, 2004

21. Ayyobi A, Brunzell JD: Lipoprotein distri-bution in the metabolic syndrome, type IIdiabetes mellitus, and familial combinedhyperlipidemia. Am J Cardiol 92 (Suppl.4A):27J–33J, 2003

22. Morton RE, Zilversmit DB: Inter-relation-ship of lipids transferred by the lipidtransfer protein isolated from human li-poprotein-deficient plasma. J Biol Chem258:11751–11757, 1983

23. Freedman DS, Bowman BA, Otvos JD,Srinivasan SR, Berenson GS: Levels andcorrelates of LDL and VLDL particle sizesamong children: the Bogalusaheart Study.Atherosclerosis 152:441–449, 2000

24. Nigon F, Lesnik P, Rouis M, ChapmanMJ: Discrete subspecies of human lowdensity lipoproteins are heterogeneous intheir activation with the cellular LDL re-ceptor. J Lipid Res 32:1741–1753, 1991

25. Lamarche B, Uffelman KD, Carpentier A,Cohn JS, Steiner G, Barrett PH, Lewis GF:Triglyceride enrichment of HDL enhancesin vivo metabolic clearance of HDL apoA-1 in healthy mean. J Cli Invest 103:1191–1199, 1999

26. Fencki S, Rota S, Sabir N, Akdaq B: Ultra-sonographic and biochemical evaluationof visceral obesity in obese women withNon-alcoholic fatty liver disease. EurJ Med Res 12:68–73, 2007

27. Sabir N, Sermen Y, Kazil S, Zencir M: Cor-relation of abdominal fat accumulationand liver steatosis: importance of ultra-sonoagraphic and anthropometric mea-surements. Eur J Ultrasound 14:121–128,2001

28. Busetto L, Tregnaghi A, De Marchi F, Se-gato G, Foletto M, Sergi G, Favretti F, LiseM, Enzi G: Liver volume and visceral obe-sity in women with hepatic steatosis un-dergoing gastric banding. Obes Res10:408–411, 2002

29. Petersen KF, Dufour S, Feng J, Befroy D,Dziura J, Dalla Man C, Cobelli C, Shul-man GI: Increased prevalence of insulinresistance and nonalcoholic fatty liver dis-easein Asian-Indian men. Proc Natl AcadSci U S A 103:18273–18277, 2006

30. Godsland IF, Johnston DG, ChaturvediN: Mechanisms of disease: lessons fromethnicity in the role of triglyceride metab-olism in ischemic heart disease. Nat ClinPract Endocrinol Metab 3:530–538, 2007

31. Herd SL, Gower BA, Dashti N, Goran MI:Body fat, fat distribution and serum lip-ids, lipoproteins and apolipoproteins inAfrican-American and Caucasian-Ameri-can prepubertal children. Int J Obes RelatMetab Disord 25:198–204, 2001

32. Deeg MA, Buse JB, Goldberg RB, KendallDM, Zagar AJ, Jacober SJ, Khan MA, PerezAT, Tan MH, the GLAI Study Investiga-tors: Pioglitazone and rosiglitazone havedifferent effects on serum lipoprotein par-ticle concentrations and sizes in patientswith type 2 diabetes and dyslipidemia. Di-abetes Care 30:2458–2464, 2007



Metabolic Syndrome and Incident End-Stage Peripheral Vascular DiseaseA 14-year follow-up study in elderly Finns

JIANJUN WANG, MD

SANNA RUOTSALAINEN, MD

LEENA MOILANEN, MD

PAIVI LEPISTO, MD

MARKKU LAAKSO, MD

JOHANNA KUUSISTO, MD

OBJECTIVE — We investigated the relationship of the metabolic syndrome and its singlecomponents, defined by four different criteria, with peripheral vascular disease (PVD) in aprospective population-based study.

RESEARCH DESIGN AND METHODS — The metabolic syndrome was defined ac-cording to the World Health Organization (WHO), the National Cholesterol Education Program(NCEP), the International Diabetes Federation (IDF), and the American Heart Association (up-dated NCEP) criteria. We investigated the relationship of the metabolic syndrome defined by theaforementioned four criteria with PVD (revacularization and amputation) by Cox regressionanalyses in a Finnish population of 1,212 subjects, aged 65–74 years, with and without diabetesduring a 14-year follow-up.

RESULTS — The metabolic syndrome defined by the WHO, NCEP, and updated NCEPcriteria was associated with a statistically significant risk for incident PVD (n � 57) with adjust-ment for all confounding variables except for prevalent diabetes (hazard ratios [HRs] from 1.91to 2.62). After adjustment for prevalent diabetes or after the exclusion of subjects with prevalentdiabetes, there was no association between the metabolic syndrome by any criteria and incidentPVD. Of the single components of the metabolic syndrome, elevated fasting glucose by the WHOand NCEP criteria (HR 2.35) and microalbuminuria by the WHO definition (2.56) predictedPVD in multivariable models (prevalent diabetes included).

CONCLUSIONS — The metabolic syndrome defined by the WHO, NCEP, and updatedNCEP criteria predicted incident end-stage PVD in elderly Finns but only when not adjusted fordiabetes status. Two of the single components of the metabolic syndrome, elevated fastingplasma glucose and microalbuminuria, predicted PVD. We conclude that the metabolic syn-drome predicts PVD but not above and beyond the risk associated with diabetes andmicroalbuminuria.

Diabetes Care 30:3099–3104, 2007

P eripheral vascular disease (PVD) re-fers to the atherosclerotic disease ofperipheral arteries, most commonly

in the lower extremities. Smoking and di-abetes are considered to be main risks oflower extremity PVD (1), but it is unclearwhether other risk factors for coronary

heart disease (CHD) are also risk factorsfor PVD. The metabolic syndrome, a clus-tering of cardiovascular risk factors thatconfers an increased risk of cardiovascu-lar disease (CVD), has been defined by avariety of groups, including the WorldHealth Organization (WHO) in 1999 (2),

the European Group for the Study of In-sulin Resistance (EGIR) in 1999 (3), theNational Cholesterol Education Program(NCEP) Expert Panel in 2001 (4), theAmerican College of Endocrinology(ACE) in 2003 (5), the International Dia-betes Federation (IDF) in 2005 (6), andthe American Heart Association and theNational Heart, Lung, and Blood Institute(updated the NCEP criteria) in 2005 (7).Since these different definitions werepublished, various prospective studieshave reported that the metabolic syn-drome defined by the criteria is associatedwith incidence or mortality of CHD andCVD and stroke (8–17). However, thereare limited data on the effect of the meta-bolic syndrome on PVD (18). In particu-lar, it is unknown whether the metabolicsyndrome predicts PVD above and be-yond diabetes. Therefore, the aim of thepresent study was to investigate the rela-tionship of the metabolic syndrome andits single components, defined by theWHO, NCEP, IDF, and updated NCEPcriteria, with the risk of end-stage lowerextremity PVD in an elderly cohort ofFinnish subjects during a 14-year follow-up.


Baseline studyThe formation (19) and representative-ness (20) of the study population havebeen described in detail previously.Briefly, the study was conducted in Kuo-pio, east Finland, between 1986 and1988. Altogether, 1,910 subjects born be-tween 1912 and 1921 were randomly se-lected from the population registerincluding all inhabitants of Kuopio. Thisrandom sample covered 35% of all resi-dents in the age-group of 65–74 years.The overall participation rate was 71%.All subjects with intermittent claudica-tion and gangrene diagnosed by physi-cians at the baseline examination or with aprevious history of leg amputation andperipheral revascularization surgery wereexcluded from the statistical analyses forincident PVD. The WHO criteria (2) for

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the Department of Medicine, Kuopio University Hospital, Kuopio, Finland.Address correspondence and reprint requests to Johanna Kuusisto, MD, Medicine, Cardiology Unit,

Kuopio University Hospital, P.O. Box 1777, 70211 Kuopio, Finland. E-mail: [email protected] for publication 22 May 2007 and accepted in revised form 1 September 2007.Published ahead of print at http://care.diabetesjournals.org on 11 September 2007. DOI: 10.2337/dc07-

0985.Abbreviations: ACR, ratio of urinary albumin to urinary creatinine; CHD, coronary heart disease; CVD,

cardiovascular disease; FPG, fasting plasma glucose; IDF, International Diabetes Federation; IGT, impairedglucose tolerance; NCEP, National Cholesterol Education Program; PVD, peripheral vascular disease; WHO,World Health Organization.






impaired glucose tolerance (IGT) and di-abetes were used in the classification ofsubjects without previously known dia-betes based on fasting plasma glucose(FPG) and 2-h postload glucose values atbaseline. The diagnosis of previouslyknown diabetes was based on drug treat-ment for diabetes or a history of a diagno-sis of diabetes made by a physician. A totalof 1,212 subjects aged 65–74 years wereincluded in the current study. Amongthem, 962 were nondiabetic, 122 hadknown diabetes, and 128 had newly diag-nosed diabetes at baseline.

Previous verified definite and possi-ble myocardial infarctions before thebaseline study were defined according tothe WHO Monitoring of Trends and De-terminants in Cardiovascular Disease(MONICA) project criteria (21) as modi-fied by the FINMONICA Acute Myocar-dial Infarction (AMI) Register StudyGroup (22).

Weight, height, waist and hip cir-cumference, and blood pressure weremeasured. Waist-to-hip ratio was de-fined as the ratio of waist circumferenceto hip circumference. BMI was calcu-lated as weight in kilograms divided bythe square of height in meters. Smokingstatus was defined as current smoking.With respect to alcohol consumption,subjects were classified as alcohol usersor nonusers. Physical activity duringleisure time was classified as physicallyinactive (little and occasional activity)and physically active (regular exerciseat least once a week and at least 30 minper time). Physical activity at work wasclassified as light physical work (seden-tary; standing, and walking a little) andheavy physical work (exhausting work-load or heavy manual work).

Blood samples were taken in themorning after a 12-h overnight fast. Allsubjects, except for those receiving insu-lin, underwent an oral glucose tolerancetest (75 g glucose). Plasma glucose andinsulin, serum lipids and lipoproteins,and urinary albumin were determined asdescribed previously (19,23). The ratio ofurinary albumin (milligrams per liter) tourinary creatinine (millimoles per liter)(ACR) was used as a measure of albuminexcretion.

The study complies with the Declara-tion of Helsinki and was approved by theEthics Committee of Kuopio UniversityHospital. All study subjects gave in-formed consent.T

able

1—D

efini

tion

sof

the

met

abol

icsy

ndro

me

byth

eW

HO

,NC

EP,

IDF

,and

upda

ted

NC

EP

crit

eria

WH

ON

CE

PID

FU

pdat

edN

CE

P

Req

uire

dFa

stin

gin

sulin

into

p25

%;F

PG�

6.1

mm

ol/l;

2-h

gluc

ose

�7.

8,ty

pe2

diab

etes

(FPG

�7.

0m

mol

/lan

d/or

2-h

gluc

ose

�11

.1m

mol

/l)

—Fo

rE

urop

eans

:wai

st�

94cm

inm

enor

�80

cmin

wom

en—

No.

ofab

norm

alit

ies

Plus

atle

ast

two

ofth

efo

llow

ing:

At

leas

tth

ree

ofth

efo

llow

ing:

Plus

atle

ast

two

ofth

efo

llow

ing:

At

leas

tth

ree

ofth

efo

llow

ing:

FPG

�6.

1m

mol

/l�

6.1

mm

ol/l;

prev

ious

lydi

agno

sed

type

2di

abet

es�

5.6

mm

ol/l

HD

Lch

oles

tero

l�

0.9

mm

ol/l

inm

enor

�1.

0m

mol

/lin

wom

en�

1.03

mm

ol/l

inm

enor

�1.

29m

mol

/lin

wom

en�

1.03

mm

ol/l

(men

)or

�1.

29m

mol

/l(w

omen

)�

1.03

mm

ol/l

(men

)or

�1.

29m

mol

/l(w

omen

)or

Tri

glyc

erid

es�

1.7

mm

ol/l

�1.

7m

mol

/l�

1.7

mm

ol/l

�1.

7m

mol

/lO

besi

tyW

aist

-to-

hip

rati

o�

0.90

inm

enor

�0.

85in

wom

en;B

MI

�30

kg/m

2W

aist

�10

2cm

inm

enor

�88

cmin

wom

enW

aist

�10

2cm

inm

enor

�88

cmin

wom

enH

yper

tens

ion

�14

0/90

mm

Hg

oran

tihy

pert

ensi

vedr

ugs

�13

0/85

mm

Hg

oran

tihy

pert

ensi

vedr

ugs

�13

0/85

mm

Hg

oran

tihy

pert

ensi

vedr

ugs

�13

0/85

mm

Hg

oran

tihy

pert

ensi

vedr

ugs

Mic

roal

bum

inur

iaA

CR

�3.

39m

g/m

mol

(30

mg/

g)

Metabolic syndrome and end-stage PVD


Follow-up studyMedical records of all study subjects whoparticipated in the baseline study in1986–1988 were reviewed by two of theauthors (S.R. and J.K.). Clinical recordsfor individuals who developed end-stagePVD during the 14-year follow-up wereobtained from the medical records of theKuopio University Hospital where all pa-tients with PVD are treated (S.R. and J.K.).PVD was defined as lower extremity am-putation (n � 26) due to ischemic vascu-lar disease or peripheral revascularization(angioplasty or surgery) (n � 31), re-corded until the end of June 2001.

Definitions of the metabolicsyndromeThe WHO, NCEP, IDF, and updatedNCEP definitions include diabetic indi-viduals, and, therefore, the present studywas based on these definitions. Each com-ponent of the four definitions was definedaccording to the original criteria. Criteriafor the four definitions of the metabolicsyndrome are shown in Table 1.

Statistical analysesAll statistical analyses were performedwith SPSS 14.0 statistical programs. Be-cause of the skewed distribution of fast-ing insulin, triglyceride concentrations,and ACR, these variables were log trans-

formed for statistical analyses. Differ-ences in basel ine character is t icsbetween subjects with and without in-cident PVD were tested by a �2 test andunivariate ANOVA adjusted for age andsex. The baseline variables not includedin the definitions of metabolic syn-drome but showing a statistically signif-icant association with incident PVDwere added into the multivariable Coxregression models as covariates. Themultivariable Cox regression analyseswere applied to investigate the associa-tion of the metabolic syndrome definedby the four criteria with incident PVD inadjusted models (model 1: adjusted forage and sex; model 2: adjusted for age,sex, history of myocardial infarction,and physical activity of work; andmodel 3: adjusted for age, sex, history ofmyocardial infarction, physical activityof work, and prevalent diabetes). Aproduct term of sex � each of four def-initions was added to the model to rep-resent interaction. The null hypothesisof no interaction was tested using thechange in �2 log likelihoods betweenCox models with and without the prod-uct term. The effect of the single com-ponents of the metabolic syndrome onincident PVD was tested by the multi-variable Cox regression models ad-justed for other risk factors. P � 0.05

(two sided) was considered to be statis-tically significant. Exact P values and95% CIs are given in the tables.

RESULTS — The median follow-upfor incident PVD (nondiabetic subjects:24 revascularizations and 2 amputations;diabetic subjects: 9 revascularizations and22 amputations) was 14.0 years (the 25thand the 75th quartiles were 13.6 and 14.7years, respectively). Of the 57 subjectswith PVD during the follow-up, 31 haddiabetes. Compared with subjects with-out incident PVD, more subjects with in-cident PVD had previous myocardialinfarctions and diabetes and were physi-cally active at work. Subjects with PVDhad also higher levels of systolic bloodpressure, ACR, triglycerides, FPG, and2-h postload glucose and lower levels ofHDL cholesterol (Table 2). Althoughthere was a trend that more subjects withincident PVD were current smokers com-pared with those without PVD (12.3 vs.7.9%), no statistically significant differ-ence in smoking was found.

Table 3 shows hazard ratios (HRs) ofthe metabolic syndrome defined by thefour different criteria to predict PVD dur-ing the 14-year follow-up among all sub-jects. The prevalence of the metabolicsyndrome at baseline varied from 51.1%(WHO criteria) to 61.1% (IDF criteria),depending on the metabolic syndromecriteria. The metabolic syndrome by theWHO, NCEP, IDF, and updated NCEPcriteria was associated with a 1.84- to2.74-fold risk for incident PVD when ad-justed for age and sex (model 1). Afterfurther adjustment for history of myocar-dial infarction and physical activity atwork (model 2), the metabolic syndromedefined by the WHO, NCEP, and the up-dated NCEP criteria was associated with astatistically significant 1.91- to 2.62-foldrisk for future PVD. However, the meta-bolic syndrome by the IDF criteria did notpredict PVD when adjusted for all ofaforementioned factors in model 2. Whenprevalent diabetes was added into model2, none of the four definitions predictedfuture PVD (model 3). Interaction termsbetween sex and the metabolic syndromeby the four definitions were not signifi-cant for PVD (P � 0.50).

We also repeated statistical analysesby excluding subjects with prevalent dia-betes (n � 250), with previous myocar-dial infarction (n � 107), and withmicroalbuminuria (n � 277), respec-tively. None of the four definitions pre-dicted incident PVD in nondiabetic

Table 2—Baseline characteristics of subjects with and without incident PVD during the 14-year follow-up in 1,212 elderly subjects

IncidentPVD Non-PVD P

n 57 1,155Male/female 24/33 402/753 0.266Age (years) 69.6 � 2.9 69.0 � 2.9 0.087Previous diabetes 24 (42.1) 98 (8.5) �0.001Prevalent diabetes 31 (54.4) 219 (19.0) �0.001Previous myocardial infarction 11 (19.3) 96 (8.3) 0.013Previous stroke 1 (1.8) 35 (3.6) 1.000Current smokers 7 (12.3) 91 (7.9) 0.216Alcohol user 15 (26.3) 327 (28.3) 0.880Physically inactive at leisure time 10 (17.5) 300 (26.0) 0.139Physically active at work 43 (75.4) 696 (60.3) 0.018BMI (kg/m2) 27.1 � 4.5 27.5 � 4.2 0.476Waist circumference (cm) 93.3 � 13.0 91.7 � 11.1 0.274Systolic blood pressure (mmHg) 166 � 26 157 � 24 0.004ACR (mg/mmol) 12.6 � 25.2 3.9 � 13.8 �0.001Total cholesterol (mmol/l) 6.48 � 1.15 6.52 � 1.29 0.816Triglycerides (mmol/l) 2.13 � 1.15 1.79 � 0.95 0.009HDL cholesterol (mmol/l) 1.16 � 0.28 1.28 � 0.35 0.007Fasting plasma glucose (mmol/l) 9.2 � 4.2 6.3 � 2.1 �0.0012-h postload glucose (mmol/l) 13.7 � 8.8 8.2 � 4.7 �0.001

Data are n (%) or means � SD.

Wang and Associates


subjects and in subjects without mi-croalbuminuria (data not shown). When107 subjects with previous myocardial in-farction were excluded, results similar tothose given in Table 3 were obtained.When analyses were done separately indiabetic and nondiabetic subjects, themetabolic syndrome was not a predictorof incident PVD because of a small num-ber of events.

Table 4 shows HRs for the single com-ponents of the metabolic syndrome defi-nitions for risk of PVD in multivariableCox regression models after adjustmentfor other risk factors in all subjects. In allsubjects, the following single componentsof the metabolic syndrome predicted PVDafter the adjustment for age, sex, historyof myocardial infarction, and physical ac-tivity at work (model 1): elevated FPG(FPG �6.1 mmol, HR 4.03) according tothe WHO and NCEP criteria and elevatedFPG (FPG �5.6 mmol, HR 2.55) accord-ing to the updated NCEP criteria, lowHDL cholesterol (HDL cholesterol �1.03mmol/l in men or �1.29 mmol/l inwomen, HR 1.90) according to the NCEPcriteria, and microalbuminuria (ACR�3.39 mg/mmol, HR 3.23) according tothe WHO definition. After further adjust-ment for prevalent diabetes (model 2), thefollowing single components of the meta-bolic syndrome still predicted PVD: ele-vated FPG (FPG �6.1 mmol, HR 2.35)and microalbuminuria (ACR �3.39 mg/mmol, HR 2.56). Of the single compo-nents, only low HDL cholesterol (HDLcholesterol �1.03 mmol/l in men or�1.29 mmol/l in women) according tothe NCEP criteria predicted PVD risk inmodel 2 (HR 3.02) among nondiabeticsubjects.

CONCLUSIONS — To our knowl-edge, this is the first study investigatingthe role of the metabolic syndrome, de-fined by the WHO, NCEP, IDF, and up-dated NCEP criteria, to predict incidentend-stage PVD. In the present study, themetabolic syndrome defined by theWHO, NCEP, and updated NCEP criteriapredicted PVD in the elderly population.After the diabetes status was taken intoaccount, none of the metabolic syndromedefinitions predicted incident PVD.

Only one previous prospective study(Dutch) has reported that the modifiedNCEP definition predicted PVD in sub-jects with familial hypercholesterolemia(18). However, the original definition ofthe metabolic syndrome was not used inthis study because of the lack of waist cir-T

able

3—H

Rs

(95%

CIs

)of

the

met

abol

icsy

ndro

me

defi

ned

byth

eW

HO

,NC

EP,

IDF

,and

upda

ted

NC

EP

crit

eria

for

inci

dent

PVD

duri

ngth

e14

-yea

rfo

llow

-up

in1,

212

nond

iabe

tic

and

diab

etic

subj

ects

All

subj

ects

Non

diab

etic

subj

ects

Dia

beti

csu

bjec

ts

Mod

els*

HR

(95%

CI)

PM

odel

s*H

R(9

5%C

I)P

Mod

els*

HR

(95%

CI)

P

n1,

212

962

250

PVD

even

ts†

5726

31W

HO

defin

itio

n(5

1.1%

)‡1

2.74

(1.5

2–4.

94)

0.00

11

1.59

(0.7

3–3.

43)

0.24

21

2.07

(0.6

3–6.

85)

0.23

32

2.62

(1.4

5–4.

72)

0.00

12

1.55

(0.7

2–3.

36)

0.26

62

2.15

(0.6

5–7.

12)

0.21

23

1.68

(0.8

9–3.

17)

0.11

0N

CE

Pde

finit

ion

(51.

1%)

12.

28(1

.30–

4.00

)0.

004

11.

43(0

.66–

3.11

)0.

366

11.

23(0

.46–

3.29

)0.

681

22.

14(1

.21–

3.77

)0.

009

21.

36(0

.62–

2.96

)0.

445

21.

37(0

.50–

3.74

)0.

538

31.

36(0

.73–

2.51

)0.

331

IDF

defin

itio

n(6

1.1%

)1

1.84

(1.0

2–3.

32)

0.04

21

1.15

(0.5

2–2.

54)

0.72

91

1.60

(0.5

9–4.

35)

0.35

72

1.74

(0.9

6–3.

15)

0.06

72

1.09

(0.4

9–2.

42)

0.82

82

1.76

(0.6

4–4.

84)

0.27

33

1.32

(0.7

2–2.

43)

0.37

3U

pdat

edN

CE

Pde

finit

ion

(58.

0%)

12.

04(1

.14–

3.65

)0.

017

11.

39(0

.64–

3.05

)0.

409

11.

14(0

.43–

3.04

)0.

798

21.

91(1

.06–

3.44

)0.

031

21.

85(0

.62–

5.48

)0.

271

21.

27(0

.47–

3.47

)0.

642

31.

29(0

.69–

2.40

)0.

420

*Mod

el1:

adju

sted

for

age

and

sex;

mod

el2:

adju

sted

for

age,

sex,

hist

ory

ofm

yoca

rdia

linf

arct

ion,

and

phys

ical

acti

vity

ofw

ork;

mod

el3:

adju

sted

for

age,

sex,

hist

ory

ofm

yoca

rdia

linf

arct

ion,

phys

ical

acti

vity

ofw

ork,

and

prev

alen

tdi

abet

es.†

Inci

dent

PVD

:n�

57.‡

Perc

enta

ges

inpa

rent

hese

sar

eth

epr

eval

ence

sof

the

met

abol

icsy

ndro

me

acco

rdin

gto

diff

eren

tcr

iter

iain

alls

ubje

cts.



cumference measurement. Moreover, theauthors did not investigate whether themetabolic syndrome predicted PVD whendiabetes status and CHD were taken intoaccount or whether all risk factors in-cluded in the definition of the metabolicsyndrome are equally important in pre-dicting PVD. In our study, we investi-gated the relationship between PVD andthe metabolic syndrome defined by thefour originally proposed criteria and in-cluded all components of each definition.Given the fact that the risk of PVD is in-creased by the presence of CHD and dia-betes (24,25), we also controlled fordiabetes status and previous myocardialinfarction in statistical analyses.

We also investigated whether all sin-gle components of the metabolic syn-drome were equally important inpredicting PVD and whether the singlecomponents were better predictors ofPVD than was the metabolic syndrome.We found that of the single componentsof the metabolic syndrome, elevated FPG(FPG �6.1mmol/l) and microalbumin-uria, were predictive of PVD with higherHRs compared with the metabolic syn-drome definitions. Although epidemio-logical and experimental data show thatmicroalbuminuria is associated with an

increased risk for all-cause and CVD mor-tality, hypertension, and diabetes, there isa little information on the relationship be-tween microalbumin and PVD (26,27). Inthis study, we found that the metabolicsyndrome did not predict PVD when sub-jects with microalbuminuria were ex-cluded, supporting the findings ofprevious studies (26 –28), which haveshown that microalbuminuria is associ-ated with PVD and diabetes. Accordingly,the WHO and NCEP criteria for the met-abolic syndrome, which include FPG�6.1mmol/l and microalbuminuria intheir definitions, had the highest HRs indifferent statistical models. However,with adjustment for diabetes status, thepredictive powers of elevated FPG andmicroalbuminuria were significantly at-tenuated. Furthermore, low HDL choles-terol predicted PVD risk only withoutadjustment for diabetes status. IGT wasnot a predictor of incident PVD in anymodel.

In the present study, the metabolicsyndrome was not a statistically signifi-cant risk factor for PVD in nondiabeticand diabetic subjects when it was ana-lyzed separately. In contrast, our previousstudy showed that the metabolic syn-drome was a predictor of CVD mortality

in subjects without diabetes, although notabove and beyond the risk associated withits individual components, such as im-paired fasting glucose, IGT, low HDL cho-lesterol, and microalbuminuria (17). Therelationship between the metabolic syn-drome and PVD may be different fromthat of the metabolic syndrome and CVD(17), but it is also possible that the num-ber of cases of PVD in subgroups was toosmall (26 in nondiabetic subjects and 31in diabetic subjects) to demonstrate a sta-tistically significant effect of the metabolicsyndrome on the risk of PVD.

Although smoking is probably themost important risk factor for the devel-opment of PVD in middle-aged men (29),we did not find this association in ourstudy. A low percentage of smokers and alow number of PVD events (n � 57) aswell as the elderly population may explainthe results.

A major limitation of our study is arelatively low number of PVD events,even though the follow-up time waslong. We restricted our analysis to end-stage PVD, and, thus, milder cases ofPVD (claudication and limb ischemia)were not included. Therefore, our find-ings may apply only to severe cases ofPVD. The diagnosis of PVD was deter-mined by strict clinically relevant crite-ria at both baseline and follow-upexaminations similarly in diabetic andnondiabetic subjects, and the propor-tion of revascularizations of all cases ofPVD was �50%. However, amputationsdo not necessarily result from athero-sclerosis alone, because in diabetic sub-jects, neuropathy and concurrentinfections may contribute to gangrene.Furthermore, the absence of middle-aged individuals in the cohort may leadto bias in the incidence of PVD. Finally,because of several definitions of themetabolic syndrome, multiple testingincreases the likelihood of false-positiveP values.

In summary, the metabolic syndromedefined by the WHO, NCEP, and updatedNCEP criteria predicts incident end-stagePVD in elderly subjects but only when notadjusted for diabetes status. Two singlecomponents of the metabolic syndrome,namely elevated FPG and microalbumin-uria, predicted PVD with higher HRs thanthose for the metabolic syndrome. There-fore, the metabolic syndrome is a risk fac-tor for PVD, but not above and beyondthe risk associated with diabetes and mi-croalbuminuria.

Table 4—HRs (95% CIs) of individual components of the metabolic syndrome based on theWHO, NCEP, IDF and updated NCEP definitions for incident PVD in 1,212 subjects

HR (95% CI)

Model 1 Model 2

FPG �6.1 mmol/l 4.03 (2.26–7.20)* 2.35 (1.15–4.79)†FPG �5.6 mmol/l 2.55 (1.29–5.07)‡ 1.74 (0.80–3.75)2-h postload glucose 7.8–11.0 mmol/l 0.64 (0.28–1.51) 1.18 (0.48–2.91)Blood pressure �130/85 mmHg or

drug treatment0.98 (0.39–2.47) 0.82 (0.32–2.09)

Blood pressure �140/90 mmHg ordrug treatment

1.39 (0.65–2.97) 1.21 (0.56–2.60)

Waist circumference �94 cm(women: �80 cm)

1.01 (0.55–1.84) 1.09 (0.59–2.02)

Waist circumference �102 cm(women: �88 cm)

0.94 (0.54–1.65) 0.83 (0.46–1.50)

Waist-to-hip ratio �0.90 (women:�0.85)

1.11 (0.56–2.19) 0.95 (0.47–1.91)

BMI �30 kg/m2 0.93 (0.48–1.78) 0.65 (0.33–1.27)Triglycerides �1.7 mmol/l 1.64 (0.96–2.80) 1.30 (0.75–2.25)HDL cholesterol �0.9 mmol/l

(women: �1.0 mmol/l)1.70 (0.94–3.07) 1.63 (0.91–2.93)

HDL cholesterol �1.03 mmol/l(women: �1.29 mmol/l)

1.85 (1.09–3.17)† 1.55 (0.90–2.66)

ACR �3.39 mg/mmol 3.22 (1.90–5.46)* 2.56 (1.49–4.40)‡

Model 1: adjusted for age, sex, history of myocardial infarction, and physical activity of work; model 2:adjusted for age, sex, history of myocardial infarction, physical activity of work, and prevalent diabetes. *P �0.001; †P � 0.05; ‡P � 0.01.

Wang and Associates


Acknowledgments— This work was sup-ported by a grant from the Academy of Finlandto M.L.

References1. Hirsch AT, Haskal ZJ, Hertzer NR, Bakal

CW, Creager MA, Halperin JL, HiratzkaLF, Murphy WR, Olin JW, Puschett JB,Rosenfield KA, Sacks D, Stanley JC, Tay-lor LM Jr, White CJ, White J, White RA,Antman EM, Smith SC Jr, Adams CD,Anderson JL, Faxon DP, Fuster V, Gib-bons RJ, Hunt SA, Jacobs AK, NishimuraR, Ornato JP, Page RL, Riegel B: ACC/AHA2005 guidelines for the management ofpatients with peripheral arterial disease(lower extremity, renal, mesenteric, andabdominal aortic): executive summary acollaborative report from the AmericanAssociation for Vascular Surgery/Societyfor Vascular Surgery, Society for Cardio-vascular Angiography and Interventions,Society for Vascular Medicine and Biol-ogy, Society of Interventional Radiology,and the ACC/AHA Task Force on PracticeGuidelines (Writing Committee to De-velop Guidelines for the Management ofPatients With Peripheral Arterial Disease)endorsed by the American Association ofCardiovascular and Pulmonary Rehabili-tation; National Heart, Lung, and BloodInstitute; Society for Vascular Nursing;TransAtlantic Inter-Society Consensus;and Vascular Disease Foundation. J AmColl Cardiol 47:1239–1312, 2006

2. World Health Organization: Definition,Diagnosis and Classification of Diabetes Mel-litus and Its Complications: Report of a WHOConsultation. Part 1: Diagnosis and Classifi-cation of Diabetes and Mellitus. Geneva,World Health Organization, 1999 (publ.no. WHO/NCD/NCS/99.2). Availablefrom http://whqlibdoc.who.int/hq/1999/who_ncd_ncs_99.2.pdf. Accessed Febru-ary 1999

3. Balkau B, Charles MA: Comment on theprovisional report from the WHO consul-tation: European Group for the Study ofInsulin Resistance (EGIR). Diabet Med 16:442–443, 1999

4. Executive summary of the Third Report ofthe National Cholesterol Education Pro-gram (NCEP) Expert Panel on Detection,Evaluation, and Treatment of High BloodCholesterol in Adults (Adult TreatmentPanel III). JAMA 285:2486–2497, 2001

5. Einhorn D, Reaven GM, Cobin RH, FordE, Ganda OP, Handelsman Y, Hellman R,Jellinger PS, Kendall D, Krauss RM,Neufeld ND, Petak SM, Rodbard HW, Sei-bel JA, Smith DA, Wilson PW: AmericanCollege of Endocrinology position state-ment on the insulin resistance syndrome.Endocr Pract 9:237–252, 2003

6. Alberti KG, Zimmet P, Shaw J: The meta-bolic syndrome—a new worldwide defi-

nition. Lancet 366:1059–1062, 20057. Grundy SM, Cleeman JI, Daniels SR, Do-

nato KA, Eckel RH, Franklin BA, GordonDJ, Krauss RM, Savage PJ, Smith SC Jr,Spertus JA, Costa F: Diagnosis and man-agement of the metabolic syndrome: anAmerican Heart Association/NationalHeart, Lung, and Blood Institute ScientificStatement. Circulation 112:2735–2752,2005

8. Wang JJ, Li HB, Kinnunen L, Hu G, Jarvi-nen TM, Miettinen ME, Yuan S, Tuom-ilehto J: How well does the metabolicsyndrome defined by five definitions pre-dict incident diabetes and incident coro-nary heart disease in a Chinesepopulation? Atherosclerosis 192:161–168,2007

9. Lakka HM, Laaksonen DE, Lakka TA, Ni-skanen LK, Kumpusalo E, Tuomilehto J,Salonen JT: The metabolic syndrome andtotal and cardiovascular disease mortalityin middle-aged men. JAMA 288:2709–2716, 2002

10. Hunt KJ, Resendez RG, Williams K,Haffner SM, Stern MP: National Choles-terol Education Program versus WorldHealth Organization metabolic syndromein relation to all-cause and cardiovascularmortality in the San Antonio Heart Study.Circulation 110:1251–1257, 2004

11. Dekker JM, Girman C, Rhodes T, NijpelsG, Stehouwer CD, Bouter LM, Heine RJ:Metabolic syndrome and 10-year cardio-vascular disease risk in the Hoorn Study.Circulation 112:666–673, 2005

12. McNeill AM, Rosamond WD, Girman CJ,Golden SH, Schmidt MI, East HE, Ballan-tyne CM, Heiss G: The metabolic syn-drome and 11-year risk of incidentcardiovascular disease in the atheroscle-rosis risk in communities study. DiabetesCare 28:385–390, 2005

13. Bonora E, Kiechl S, Willeit J, Oberhollen-zer F, Egger G, Bonadonna RC, MuggeoM: Carotid atherosclerosis and coronaryheart disease in the metabolic syndrome:prospective data from the Bruneck study.Diabetes Care 26:1251–1257, 2003

14. Wilson PW, D’Agostino RB, Parise H, Sul-livan L, Meigs JB: Metabolic syndrome asa precursor of cardiovascular disease andtype 2 diabetes mellitus. Circulation 112:3066–3072, 2005

15. Koren-Morag N, Goldbourt U, Tanne D:Relation between the metabolic syndromeand ischemic stroke or transient ischemicattack: a prospective cohort study in pa-tients with atherosclerotic cardiovasculardisease. Stroke 36:1366–1371, 2005

16. Kurl S, Laukkanen JA, Niskanen L, Laak-sonen D, Sivenius J, Nyyssonen K, Sa-lonen JT: Metabolic syndrome and therisk of stroke in middle-aged men. Stroke37:806–811, 2006

17. Wang J, Ruotsalainen S, Moilanen L, Lepi-sto P, Laakso M, Kuusisto J: The meta-bolic syndrome predicts cardiovascular

mortality: a 13-year follow-up study in el-derly non-diabetic Finns. Eur Heart J 28:857–864, 2007

18. Rana JS, Jansen AC, Zwinderman AH,Nieuwdorp M, van Aalst-Cohen ES,Jukema JW, Trip MD, Kastelein JJ: Meta-bolic syndrome and risk of coronary, ce-rebral, and peripheral vascular disease ina large Dutch population with familial hy-percholesterolemia. Diabetes Care 29:1125–1127, 2006

19. Mykkanen L, Laakso M, Uusitupa M,Pyorala K: Prevalence of diabetes and im-paired glucose tolerance in elderly sub-jects and their association with obesityand family history of diabetes. DiabetesCare 13:1099–1105, 1990

20. Mykkanen L, Laakso M, Penttila I, PyoralaK: Asymptomatic hyperglycemia and car-diovascular risk factors in the elderly. Ath-erosclerosis 88:153–161, 1991

21. World Health Organization: MONICAManual: CVD/MNC. Geneva, WorldHealth Organization, 1990

22. Tuomilehto J, Arstila M, Kaarsalo E, Kan-kaanpaa J, Ketonen M, Kuulasmaa K,Lehto S, Miettinen H, Mustaniemi H,Palomaki P, et al: Acute myocardial in-farction (AMI) in Finland—baseline datafrom the FINMONICA AMI register in1983–1985. Eur Heart J 13:577–587,1992

23. Kuusisto J, Mykkanen L, Pyorala K,Laakso M: Hyperinsulinemic microalbu-minuria: a new risk indicator for coronaryheart disease. Circulation 91:831–837,1995

24. Murabito JM, D’Agostino RB, SilbershatzH, Wilson WF: Intermittent claudication:a risk profile from the Framingham HeartStudy. Circulation 96:44–49, 1997

25. Laakso M, Lehto S: Epidemiology of riskfactors for cardiovascular disease in dia-betes and impaired glucose tolerance.Atherosclerosis 137 (Suppl.):S65–S73,1998

26. Luft FC, Agrawal B: Microalbuminuria asa predictive factor for cardiovascularevents. J Cardiovasc Pharmacol 33 (Suppl.1):S11–S15; discussion S41–S43, 1999

27. Zander E, Heinke P, Reindel J, KohnertKD, Kairies U, Braun J, Eckel L, Kerner W:Peripheral arterial disease in diabetes mel-litus type 1 and type 2: are there differentrisk factors? Vasa 31:249–254, 2002

28. Wang JJ, Qiao Q, Miettinen ME, Lappa-lainen J, Hu G, Tuomilehto J: The meta-bolic syndrome defined by factor analysisand incident type 2 diabetes in a Chinesepopulation with high postprandial glu-cose. Diabetes Care 27:2429–2437, 2004

29. Bowlin SJ, Medalie JH, Flocke SA, Zyzan-ski SJ, Goldbourt U: Epidemiology of in-termittent claudication in middle-agedmen. Am J Epidemiol 140:418–430,1994



Does Waist Circumference Predict Diabetesand Cardiovascular Disease BeyondCommonly Evaluated Cardiometabolic RiskFactors?PETER M. JANISZEWSKI, MSC

1

IAN JANSSEN, PHD1,2

ROBERT ROSS, PHD1,3

OBJECTIVE — While the measurement of waist circumference (WC) is recommended incurrent clinical guidelines, its clinical utility was questioned in a recent consensus statement. Inresponse, we sought to determine whether WC predicts diabetes and cardiovascular disease(CVD) beyond that explained by BMI and commonly obtained cardiometabolic risk factorsincluding blood pressure, lipoproteins, and glucose.

RESEARCH DESIGN AND METHODS — Subjects consisted of 5,882 adults from the1999–2004 National Health and Nutrition Examination Survey, which is nationally represen-tative and cross-sectional. Subjects were grouped into sex-specific WC and BMI tertiles. Bloodpressure, triglycerides, LDL and HDL cholesterol, and glucose were categorized using standardclinical thresholds. Logistic regression analyses were used to calculate the odds for diabetes andCVD according to WC tertiles.

RESULTS — After controlling for basic confounders, the medium and high WC tertiles weremore likely to have diabetes and CVD compared with the low WC tertile (P � 0.05). Afterinclusion of BMI and cardiometabolic risk factors in the regression models, the magnitude of theodds ratios were attenuated (i.e., for diabetes the magnitude decreased from 6.54 to 5.03 for thehigh WC group) but remained significant in the medium and high WC tertiles for the predictionof diabetes, though not for CVD.

CONCLUSIONS — WC predicted diabetes, but not CVD, beyond that explained by tradi-tional cardiometabolic risk factors and BMI. The findings lend critical support for the recom-mendation that WC be a routine measure for identification of the high-risk, abdominally obesepatient.

Diabetes Care 30:3105–3109, 2007

I t is established that waist circumfer-ence (WC) predicts increased risk ofmorbidity (1–4) and mortality (5) be-

yond that explained by BMI alone. Severalorganizations, including the National In-stitutes of Health (6), currently advocatefor the measurement of WC in clinical

practice. However, a recent consensusstatement from the American DiabetesAssociation (ADA), the Obesity Society,and the American Society for Nutritionquestioned the clinical utility of WC mea-surement (7). Opposition to the inclusionof WC measurement in clinical practice is

hinged on the observation that it is un-clear whether WC predicts health risk be-yond that explained by BMI andcommonly evaluated cardiometabolicrisk factors (7). It is reasoned that clini-cians would be unnecessarily burdenedby the measurement of WC if this mea-sure failed to explain health risk beyondthe risk factors routinely obtained in clin-ical practice.

Limited evidence suggests that WCpredicts risk of cardiovascular disease(CVD) after control for hypertension(1,2), hypercholesterolemia (2), and theapolipoprotein B–to–A ratio (1). Absentfrom the literature is a clear demonstra-tion that WC predicts the risk of diabetesand CVD in men and women beyond thatexplained by the commonly evaluatedcardiometabolic risk factors (blood pres-sure, triglyceride, LDL and HDL choles-terol, and glucose levels) and BMI. Weaddressed this issue using data from themost recent National Health and Nutri-tion Survey (NHANES).

RESEARCH DESIGN ANDMETHODS — The study sample wasobtained from the 1999 –2000, 2001–2002, and 2003–2004 NHANES.NHANES was designed to be a nationallyrepresentative cross-sectional survey,which allows for two or three surveyrounds to be combined, as done here.NHANES was conducted by the U.S. Na-tional Center for Health Statistics to esti-mate the prevalence of major diseases,nutritional disorders, and risk factors forthese diseases. The sampling plan used astratified, multistage, probability clusterdesign. Full details of the study designand procedures are available elsewhere(8). Informed consent was obtainedfrom all participants and the protocol ap-proved by the National Center for HealthStatistics.

Participants who were aged �18years, pregnant women, or missing waistcircumference, BMI, outcome measures,or covariates required for the analyseswere excluded from this study. This left a

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1School of Kinesiology and Health Studies, Queen’s University, Kingston, Ontario, Canada; the2Department of Community Health and Epidemiology, Queen’s University, Kingston, Ontario, Canada; andthe 3Division of Endocrinology and Metabolism, Department of Medicine, Queen’s University, Kingston,Ontario, Canada.

Address correspondence and reprint requests to Robert Ross, PhD, School of Kinesiology and HealthStudies, Queen’s University, Kingston, Ontario, Canada, K7L 3N6. E-mail: [email protected].


0945.Abbreviations: ADA, American Diabetes Association; CVD, cardiovascular disease; NHANES, National

Health and Nutrition Examination Survey; WC, waist circumference.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





total of 5,882 subjects (3,001 men and2,881 women).

Measurement and classification ofanthropometric variablesWC was measured during minimal respi-ration to the nearest 0.1 cm at the level ofthe iliac crest (8). Height was measured tothe nearest 0.1 cm and body mass to thenearest 0.1 kg (8). BMI was calculated asweight in kilograms divided by the squareof height in meters. Subjects were dividedinto sex-specific tertiles for WC and BMI.We divided the subjects into WC and BMItertiles instead of using commonly em-ployed clinical thresholds to match thegroups for size both within (i.e., threeequally sized WC groups) and across (i.e.,with high WC group the same size as highBMI group) anthropometric measures. Inmen, WC tertiles were defined by the fol-lowing thresholds: �90.9, 90.9–102.9,and �102.9 cm. The corresponding val-ues in women were �85.5, 85.5–98.7,and �98.7 cm. In men, BMI tertiles weredefined by the following thresholds:�24.8, 24.8 –28.8, and �28.8 kg/m2.The corresponding values in women were�24.6, 24.6–29.9, and �29.9 kg/m2.

Measurement and classification ofcardiometabolic risk factorsBlood pressure. Three blood pressuremeasurements were obtained with thesubject in a seated position using a stan-dard manual mercury sphygmomanome-ter (8). The average of the three readingswas utilized. Blood pressure was classifiedaccording to established guidelines (9):normal (systolic �120 and diastolic �80mmHg), prehypertension (systolic 120–139 or diastolic 80–89 mmHg), or hyper-tension (systolic �140 or diastolic �90mmHg). When systolic and diastolic bloodpressures fell into different categories, thehigher category was selected for classifica-tion. Participants who reported takingblood pressure medication were consideredto have hypertension regardless of theirblood pressure measurements.Lipids and lipoproteins. Blood sampleswere obtained after an overnight fast forthe measurement of serum LDL choles-terol, HDL cholesterol, triglycerides, andglucose as described in detail elsewhere(8,10). Briefly, cholesterol and triglycer-ide levels were measured enzymatically ina series of coupled reactions hydrolyzingcholesterol ester and triglyceride to cho-lesterol and glycerol, respectively. LDLcholesterol, HDL cholesterol, and triglyc-eride levels were classified according to

the National Cholesterol Education Pro-gram guidelines (11). LDL cholesterolwas categorized as optimal (�100 mg/dl),near optimal (100–129 mg/dl), border-line high (130 –159 mg/dl), or high(�160 mg/dl). Participants who reportedtaking a cholesterol-lowering medicationwere placed into the high LDL cholesterolcategory regardless of their LDL choles-terol level. HDL cholesterol was catego-rized as low (�40 mg/dl), normal (40–59mg/dl), or high (�60 mg/dl). Triglycer-ides were categorized as normal (�150mg/dl), borderline high (150–199 mg/dl), or high (�200 mg/dl).Glucose and diabetes. Fasting plasmaglucose samples were assayed using a hex-okinase enzymatic method (8,12). Subjectswere classified as having normal glucose(�100 mg/dl), impaired fasting glucose(100–125 mg/dl), or diabetes (�126 mg/dl) in accordance with ADA guidelines(13). All participants with physician-diagnosed diabetes (outside of preg-nancy) were coded positive for diabetes,as were those who reported using insulinor blood glucose–lowering medications.CVD. Participants who reported that aphysician had ever told them they had aheart attack, stroke, angina, congestiveheart failure, or coronary heart diseasewere coded positive for CVD. All otherparticipants were coded negative for CVD.

Confounding variablesConfounding variables included age,race/ethnicity, sex, and smoking status.Age was included in the analysis as a con-tinuous variable. Race was categorized asnon-Hispanic white, non-Hispanic black,Hispanic, and other. Subjects were con-sidered current smokers if they smokedcigarettes at the time of the interview, pre-vious smokers if they were not current

smokers but had smoked 100 cigarettes intheir entire life, and nonsmokers if theysmoked less than this amount.

Statistical analysisThe Intercooled Stata program (version 7;Stata, College Station, TX) was used toproperly weight the sample to be repre-sentative of the U.S. population and totake into account the complex samplingstrategy of the NHANES design. Initially,logistic regression tests were used to ex-amine associations among WC or BMIcategories, CVD, and diabetes. Threemodels were run for each disease out-come. The first model controlled for thebasic confounding variables (age, sex,race, and smoking). The second modelcontrolled for the basic confounding vari-ables and the risk factor categories for thecardiometabolic variables (glucose cate-gories were not controlled for in the dia-betes analysis). The third model con-trolled for the basic confounding vari-ables, the cardiometabolic risk factor cat-egories, and BMI (or WC) categories.Next, subjects were cross-classified ac-cording to WC (low, moderate, or high)and the number of metabolic risk factors(0, 1, 2, or �3), creating 12 different cat-egories. Odds ratios (ORs) for CVD anddiabetes were then computed for these 12groups. P for trend values were calculatedto determine whether the WC and meta-bolic risk factor groups had independenteffects on CVD and diabetes.

To further explore the added value ofWC, we determined the discriminatoryability of the diabetes and CVD models(e.g., ability to correctly separate thosewho did and did not have disease) usingthe c statistic. For each disease outcome,the c statistic was calculated for three sep-arate models that included the following

Table 1—Descriptive characteristics of study participants

Variable Total Men Women

n 5,882 3,001 2,881Age (years) 44.2 � 0.5 43.3 � 0.5 45.1 � 0.5Waist circumference (cm) 95.3 � 0.4 98.5 � 0.4 92.1 � 0.5BMI (kg/m2) 27.7 � 0.1 27.6 � 0.1 27.8 � 0.2Impaired fasting glucose 24.6 (1.1) 30.3 (1.4) 19.0 (1.1)Diabetes 8.1 (0.5) 9.2 (0.7) 7.1 (0.5)Cardiovascular disease 7.0 (0.5) 8.0 (0.7) 6.0 (0.6)Hypertension 27.3 (0.9) 26.6 (1.2) 28.1 (1.0)High LDL cholesterol 21.6 (0.8) 22.8 (0.9) 20.4 (0.1)Low HDL cholesterol 19.9 (0.8) 27.8 (1.0) 12.0 (0.9)High triglycerides 14.7 (0.6) 17.0 (1.0) 12.4 (0.6)

Data are means � SE for continuous variables or prevalence �SE� (%) for dichotomous variables.

Waist circumference and morbidity


variables: 1) demographics (age, race, sex,and smoking), 2) demographics plus tra-ditional risk factors (blood pressure, LDLand HDL cholesterol, and triglyceride cat-egories), and 3) demographics, traditionalrisk factors, and WC categories. The c sta-tistic is identical to the area under the re-ceiver operating characteristic curve, withvalues ranging from 0.5 (no better thanchance alone) to 1.0 (perfect).

RESULTS — The descriptive charac-teristics of the study sample are containedwithin Table 1 . Table 2 presents the re-sults of the logistic regression models inwhich WC groups were used to predictthe likelihood of having diabetes andCVD. After controlling for age, sex, race,and smoking, participants in the mediumand high WC groups were more likely tohave diabetes and CVD compared withparticipants in the low WC group (P �

0.05). After inclusion of the cardiometa-bolic risk factor categories in the logisticregression models, the magnitude of theORs were attenuated but remained signif-icant in the medium and high WC groupsfor the prediction of diabetes (OR 1.98[95% CI 1.26 –3.11] and 4.62 [3.16 –6.75], respectively) but not for CVD. Afinal set of logistic regression models in-cluded BMI categories among the covari-ates. After controlling for demographiccharacteristics, smoking, cardiometabolicrisk factors categories, and BMI catego-ries, the moderate and high WC catego-ries remained predictive of a higherlikelihood of diabetes (2.32 [1.30–4.12]and 5.03 [2.87–8.83], respectively) butnot CVD (P � 0.1) (Table 2).

Table 3 presents the results of the lo-gistic regression models in which BMIgroups were used to predict the likeli-hood of having diabetes and CVD. After

controlling for age, sex, race, and smok-ing, participants in the medium and highBMI groups were more likely to have dia-betes and CVD compared with partici-pants in the low BMI group (P � 0.05).After the inclusion of cardiometabolicrisk factor categories in the logistic regres-sion models, ORs for the medium andhigh BMI categories were attenuated forboth diabetes and CVD, with only thehigh BMI category remaining associatedwith a greater risk of diabetes (OR 2.92[95% CI 1.95–4.37]). Lastly, after inclu-sion of WC in addition to demographiccharacteristics, smoking, and cardiometa-bolic risk factors, neither the moderatenor the high BMI categories remainedpredictive of a higher likelihood of diabe-tes or CVD (P � 0.1) (Table 3).

To further illustrate the effect of WC,we divided the study participants intogroups based on their number of high-risk metabolic variables. We then cross-tabulated the WC and cardiometabolicrisk factor groups to form 12 WC � met-abolic risk factor groups. As illustrated inFig. 1A, both WC and cardiometabolicrisk factor groups were independent pre-dictors of diabetes (Ptrend � 0.001); i.e.,for a given number of cardiometabolicrisk factors, the likelihood of having dia-betes increased when moving from thelow to high WC groups. Conversely,within a given WC group, the likelihoodof having diabetes increased when mov-ing from the group with no cardiometa-bolic risk factors to the group with threeor more. As illustrated in Fig. 1B, cardio-metabolic risk factor groups, but not WCgroups, significantly predicted CVD.Thus, for a given number of cardiometa-bolic risk factors, the likelihood of havingCVD was not different across WC groups(Ptrend � 0.415).

Table 2—ORs for diabetes and CVD according to WC

Covariates included in regressionmodel

Waist circumference tertile

Low Medium High

DiabetesAge, sex, race, and smoking 1.00 2.44 (1.53–3.89)* 6.54 (4.43–9.67)*Age, sex, race, smoking, and

metabolic risk factors†1.00 1.98 (1.26–3.11)* 4.62 (3.16–6.75)*

Age, sex, race, smoking, metabolicrisk factors, and BMI‡

1.00 2.32 (1.30–4.12)* 5.03 (2.87–8.83)*

Cardiovascular diseaseAge, sex, race, and smoking 1.00 1.41 (1.01–1.98)* 1.73 (1.22–2.44)*Age, sex, race, smoking, and

metabolic risk factors‡1.00 1.14 (0.79–1.64) 1.16 (0.79–1.70)

Age, sex, race, smoking, metabolicrisk factors, and BMI‡

1.00 0.97 (0.62–1.50) 0.80 (0.43–1.52)

Data are OR (95% CI). The low waist circumference group was used as the referent. *Significantly greaterthan low waist circumference group (P � 0.05). †Metabolic risk factors include blood pressure, LDLcholesterol, HLD cholesterol, and triglyceride risk factor categories. ‡Metabolic risk factors include bloodpressure, LDL cholesterol, HLD cholesterol, triglyceride, and fasting glucose risk factor categories.

Table 3—ORs for diabetes and CVD according to BMI

Covariates included in regression model

BMI tertile

Low Medium High

DiabetesAge, sex, race, and smoking 1.00 1.71 (1.15–2.55)* 4.12 (2.72–6.24)*Age, sex, race, smoking, and metabolic risk factors† 1.00 1.38 (0.93–2.05) 2.92 (1.95–4.37)*Age, sex, race, smoking, metabolic risk factors, and WC† 1.00 0.73 (0.42–1.25) 0.91 (0.49–1.68)

Cardiovascular diseaseAge, sex, race, and smoking 1.00 1.41 (1.01–1.97)* 1.84 (1.38–2.44)*Age, sex, race, smoking, and metabolic risk factors‡ 1.00 1.19 (0.82–1.72) 1.32 (0.95–1.85)Age, sex, race, smoking, metabolic risk factors, and WC‡ 1.00 1.26 (0.81–1.99) 1.58 (0.88–2.23)

Data are OR (95% CI). The normal-weight BMI group was used as the referent. *Significantly greater than low BMI (P � 0.05). †Metabolic risk factors include bloodpressure, LDL cholesterol, HLD cholesterol, and triglyceride risk factor categories. ‡Metabolic risk factors include blood pressure, LDL cholesterol, HLD cholesterol,triglyceride, and fasting glucose risk factor categories.

Janiszewski, Janssen, and Ross


Finally, the c statistic was calculatedto determine the discriminatory ability ofdiabetes and CVD models. For diabetes,the c statistic increased from 0.77 to 0.80to 0.82 across modes that included basicdemographic characteristics; demograph-ics plus traditional risk factor categories;and demographics, traditional risk fac-tors, and waist circumference categories,respectively. The corresponding c statisticvalues for the CVD models were 0.83,0.85, and 0.85.

CONCLUSIONS — The pr ima ryfinding of this study is that WC predictsthe likelihood of diabetes beyond that ex-plained by commonly evaluated cardiom-etabolic risk factors and BMI. Conversely,BMI did not predict diabetes after consid-eration of common cardiometabolic riskfactors and WC. Although both elevatedWC and BMI were associated with greaterCVD risk, these effects were eliminated

after control for cardiometabolic riskfactors.

Clinical guidelines for the assessmentand/or management of obesity in the U.S.(14) and Canada (15) recommend thatmeasurement of WC be used to identifythe need for further assessment includingmeasurement of cardiometabolic risk fac-tors. The recent consensus statement ofthe ADA, the Obesity Society, and theAmerican Society for Nutrition questionsthe sequence of these clinical measuresand, more importantly, the relevance ofWC measurement in clinical practice (7).Our finding that WC predicts the risk ofdiabetes beyond that explained by cardio-metabolic risk factors and BMI extendsprevious observations that document anapproximately fivefold greater risk of di-abetes in the highest relative to the lowestcategory of WC in multivariate analysiscontrolling for lifestyle factors and BMI(3,4). Combined with the fact that the

sex-specific WC cut points used in thecurrent study approximate those advo-cated in the guidelines (�102 and 88 cmin men and women, respectively), theseobservations reinforce the utility of WC asa first step in the identification of thehigh-risk, abdominally obese patient. In-deed, although an elevated WC per sealerts the clinician to the need for furtherclinical assessment, we (16) and others(17) have shown that only patients withan elevated WC in combination with ele-vations in one or more cardiometabolicrisk factors represent those who are atsubstantially increased health risk andthus require aggressive treatment.

The mechanistic link that explains theassociation between WC and diabetes riskindependent of cardiometabolic risk fac-tors and BMI is unclear and remains thefocus of ongoing investigation (18). Al-though the portal theory originally pro-posed a substrate-driven mechanism(19), recent evidence suggests that thepathophysiology of abdominal adipositymay result from the augmented secretionof various prothrombotic and proinflam-matory cytokines from an expanded ab-dominal fat depot (20).

Although WC was associated withCVD, such that individuals with a highWC were 73% more likely to have CVDthan those with a low WC, the associationdid not remain significant after control forthe cardiometabolic risk factors. Thisfinding was not unexpected given thatWC is a strong correlate of dyslipidemia,hypertension, and the metabolic syn-drome (21), themselves established ante-cedents for CVD. Accordingly, thisfinding does not indicate that a high WCis not a risk factor for CVD but, rather,that WC predicts CVD via its influence oncardiometabolic risk factors. Indeed, theutility of WC to predict CVD risk will al-ways be attenuated when metabolic riskfactors that lie in the causal pathway be-tween WC and risk of CVD are includedin the prediction model. This observationagrees with the findings of the INTER-HEART study, wherein the strong associ-ation between WC and myocardialinfarction was substantially attenuated af-ter control for hypertension and the apo-lipoprotein B–to–A ratio (1).

From a clinical perspective, it is note-worthy that in addition to the utility ofWC measurement to identify the high-risk, abdominally obese patient, WC isthe single best anthropometric measurefor detecting changes in abdominal obe-sity in response to treatment. It has re-

Figure 1—ORs for diabetes (A) and CVD (B) according to WC � metabolic risk factor groups.Both WC and metabolic risk factor groups were independent predictors of diabetes (Ptrend �0.001). The metabolic risk factor groups were independent predictors of CVD (Ptrend � 0.001),whereas the WC groups were not (Ptrend � 0.415). Mod, moderate.

Waist circumference and morbidity


peatedly been demonstrated thatalthough WC is reduced consequent toweight loss, WC can also be reduced inresponse to treatment in obese individu-als who are resistant to weight loss orchanges in BMI (22). The implication isthat when considering the efficacy oftreatment strategies designed to manageabdominal obesity, practitioners are en-couraged to look beyond body weight asthe measure of benefit and measure WC.

The analyses presented here are basedon a large and representative dataset andare therefore generalizable to the U.S.adult population. However, the cross-sectional nature of this study precludesdefinitive causal inferences about the as-sociation between WC and BMI with dia-betes and CVD. Numerous studies,however, have shown that high WC andBMI precede the onset of morbidity (1–4)and mortality (5). The assessment of CVDpresence in the current study relied onparticipant recall of previous diagnosisand thus may have been a source of error.Additionally, as our assessment of diabe-tes was based on fasting plasma glucosevalues, a limited number of new diabetescases may have been misclassified as non-diabetes. Lastly, due to the limited samplesize, we were not able to perform ethnic-ity- and/or sex-specific analyses.

The demonstration that WC predictsrisk of diabetes beyond that explained bycardiometabolic risk factors routinely ac-quired in clinical practice responds toprior criticism (7) and lends critical sup-port for the recommendation that WC bea routine measure for identification andmanagement of the high-risk, abdomi-nally obese patient (14,15). Indeed, com-bined with the observation that WC isassociated with changes in abdominalobesity in response to treatment with orwithout weight loss (22), it is difficult toimagine a cogent argument against inclu-sion of WC in clinical practice.

Acknowledgments— This study was sup-ported by Canadian Institutes of Health Re-search Grant MT13448 (to R.R.).

References1. Yusuf S, Hawken S, Ounpuu S, Bautista L,

Franzosi MG, Commerford P, Lang CC,Rumboldt Z, Onen CL, Lisheng L, Tan-

omsup S, Wangai P Jr, Razak F, SharmaAM, Anand SS: Obesity and the risk ofmyocardial infarction in 27,000 partici-pants from 52 countries: a case-controlstudy. Lancet 366:1640–1649, 2005

2. Rexrode KM, Carey VJ, Hennekens CH,Walters EE, Colditz GA, Stampfer MJ,Willett WC, Manson JE: Abdominal adi-posity and coronary heart disease inwomen. JAMA 280:1843–1848, 1998

3. Wang Y, Rimm EB, Stampfer MJ, WillettWC, Hu FB: Comparison of abdominaladiposity and overall obesity in predictingrisk of type 2 diabetes among men. Am JClin Nutr 81:555–563, 2005

4. Carey VJ, Walters EE, Colditz GA, So-lomon CG, Willett WC, Rosner BA,Speizer FE, Manson JE: Body fat distribu-tion and risk of non-insulin-dependentdiabetes mellitus in women: the Nurses’Health Study. Am J Epidemiol 145:614–619, 1997

5. Bigaard J, Tjonneland A, Thomsen BL,Overvad K, Heitmann BL, Sorensen TI:Waist circumference, BMI, smoking, andmortality in middle-aged men andwomen. Obes Res 11:895–903, 2003

6. National Institutes of Health: Clinicalguidelines on the identification, evalua-tion, and treatment of overweight andobesity in adults: the evidence report.Obes Res 6 (Suppl. 2):51S–209S, 1998

7. Klein S, Allison DB, Heymsfield SB, KelleyDE, Leibel RL, Nonas C, Kahn R: WaistCircumference and cardiometabolic risk:a consensus statement from ShapingAmerica’s Health: Association for WeightManagement and Obesity Prevention;NAASO, The Obesity Society; the Ameri-can Society for Nutrition; and the Ameri-can Diabetes Association. Diabetes Care30:1647–1652, 2007

8. National Health and Nutrition Examina-tion Survey Data [article online], 2006.Hyattsville, MD, National Center forHealth Statistics. Available from http://www.cdc.gov/nchs/about/major/nhanes/datalink.htm. Accessed 6 April 2007

9. Joint National Committee on DetectionEvaluation and Treatment of High BloodCholesterol in Adults: The fifth report ofthe Joint National Committee on Detec-tion Evaluation and Treatment of HighBlood Pressure (JNC V). Arch Intern Med153:154–183, 1993

10. Johnson CL, Rifkind BM, Sempos CT,Carroll MD, Bachorik PS, Briefel RR, Gor-don DJ, Burt VL, Brown CD, Lippel K, etal.: Declining serum total cholesterol lev-els among US adults: the National Healthand Nutrition Examination Surveys.JAMA 269:3002–3008, 1993

11. Expert Panel on Detection Evaluation and

Treatment of High Blood Cholesterol inAdults: Executive summary of the thirdreport of the National Cholesterol Educa-tion Program (NCEP) Expert Panel on De-tection Evaluation and Treatment of HighBlood Cholesterol in Adults (Adult Treat-ment Panel III). JAMA 285:2486–2497,2001

12. Harris MI, Flegal KM, Cowie CC, Eber-hardt MS, Goldstein DE, Little RR, Wied-meyer HM, Byrd-Holt DD: Prevalence ofdiabetes, impaired fasting glucose, andimpaired glucose tolerance in U.S. adults:the Third National Health and NutritionExamination Survey, 1988–1994. Diabe-tes Care 21:518–524, 1998

13. American Diabetes Association: Diagnosisand classification of diabetes mellitus (Po-sition Statement). Diabetes Care 27(Suppl. 1):S5–S10, 2004

14. Aronne LJ: Classification of obesity andassessment of obesity-related health risks.Obes Res 10 (Suppl. 2):105S–115S, 2002

15. Lau DC, Douketis JD, Morrison KM,Hramiak IM, Sharma AM, Ur E: 2006 Ca-nadian clinical practice guidelines on themanagement and prevention of obesity inadults and children [summary]. Cmaj176:S1–S13, 2007

16. Katzmarzyk PT, Janssen I, Ross R, ChurchTS, Blair SN: The importance of waist cir-cumference in the definition of metabolicsyndrome: prospective analyses of mor-tality in men. Diabetes Care 29:404–409,2006

17. Despres JP, Lemieux I, Prud’homme D:Treatment of obesity: need to focus onhigh risk abdominally obese patients. BMJ322:716–720, 2001

18. Snijder MB, van Dam RM, Visser M, Sei-dell JC: What aspects of body fat are par-ticularly hazardous and how do wemeasure them? Int J Epidemiol 35:83–92,2006

19. Bjorntorp P: “Portal” adipose tissue as agenerator of risk factors for cardiovasculardisease and diabetes. Arteriosclerosis 10:493–496, 1990

20. Wajchenberg BL: Subcutaneous and vis-ceral adipose tissue: their relation to themetabolic syndrome. Endocr Rev 21:697–738, 2000

21. Janssen I, Katzmarzyk PT, Ross R: Waistcircumference and not body mass indexexplains obesity-related health risk. Am JClin Nutr 79:379–384, 2004

22. Ross R, Dagnone D, Jones PJ, Smith H,Paddags A, Hudson R, Janssen I: Reduc-tion in obesity and related comorbid con-ditions after diet-induced weight loss orexercise-induced weight loss in men: arandomized, controlled trial. Ann InternMed 133:92–103, 2000

Janiszewski, Janssen, and Ross


Gene Expression of Adiponectin Receptorsin Human Visceral and SubcutaneousAdipose Tissue Is Related to InsulinResistance and Metabolic Parameters andIs Altered in Response to Physical TrainingMATTHIAS BLUHER, MD

1

CATHERINE J. WILLIAMS, BS2

NORA KLOTING, PHD1

ALEX HSI, BS2

KAREN RUSCHKE, PHD1

ANDREAS OBERBACH, PHD1

MATHIAS FASSHAUER, MD1

JANIN BERNDT1

MICHAEL R. SCHON, MD3

ALICJA WOLK, DMSC4

MICHAEL STUMVOLL, MD1

CHRISTOS S. MANTZOROS, MD2

OBJECTIVE — Adiponectin receptors 1 and 2 (AdipoR1 and AdipoR2, respectively) mediatethe effects of adiponectin on glucose and lipid metabolism in vivo. We examined whetherAdipoR1 and/or AdipoR2 mRNA expression in human adipose tissue is fat-depot specific. Wealso studied whether their expression in visceral and subcutaneous fat depots is associated withmetabolic parameters and whether their expression is regulated by intensive physical exercise.

RESEARCH DESIGN AND METHODS — We determined metabolic parameters andassessed AdipoR1 and AdipoR2 mRNA expression using quantitative real-time PCR in adiposetissue in an observational study of 153 subjects and an interventional study of 60 subjects (20each with normal glucose tolerance, impaired glucose tolerance, and type 2 diabetes) before andafter intensive physical training for 4 weeks.

RESULTS — AdipoR1 and AdipoR2 mRNA expression is not significantly different betweenomental and subcutaneous fat, but their expression is several-fold lower in adipose tissue than inmuscle. AdipoR2 mRNA expression in visceral fat is highly correlated with its expression insubcutaneous fat. AdipoR2 mRNA expression in both visceral and subcutaneous fat is positivelyassociated with circulating adiponectin and HDL levels but negatively associated with obesity aswell as parameters of insulin resistance, glycemia, and other lipid levels before and after adjust-ment for fat mass. Physical training for 4 weeks resulted in increased AdipoR1 and AdipoR2mRNA expression in subcutaneous fat.

CONCLUSIONS — AdipoR2 mRNA expression in fat is negatively associated with insulinresistance and metabolic parameters independently of obesity and may mediate the improve-ment of insulin resistance in response to exercise.

Diabetes Care 30:3110–3115, 2007

A diponectin is an adipose tissue–secreted cytokine that acts as a keymodulator of insulin sensitivity

(1,2) and glucose and lipid metabolism(3) and has pronounced antiatheroscle-rotic effects (4,5). The beneficial effects ofthis highly abundant 244 –amino acidprotein hormone (circulating at �10�g/ml concentration in human serumand accounting for �0.01% of totalplasma protein) are predominantly medi-ated by two cell membrane receptors, adi-ponectin receptors 1 (AdipoR1) and 2(AdipoR2) (6).

AdipoR1 is a high-affinity receptor forglobular adiponectin, and studies in micehave shown that it is ubiquitously ex-pressed (6–10) but most abundantly inskeletal muscle. AdipoR2 is predomi-nantly expressed in liver and has interme-diate affinity for both full-length andglobular adiponectin (6,8). Simultaneousdisruption of both AdipoR1 and AdipoR2abolished adiponectin binding andactions, resulting in increased tissue tri-glyceride content, inflammation, and ox-idative stress, leading to insulin resistanceand glucose intolerance in mice (9). Ele-vated expression of AdipoR1 and Adi-poR2 has been associated with decreasedplasma insulin levels in mice in eitherphysiological (i.e., fasting) or pathologi-cal conditions (11). We have previouslyreported that prolonged exposure tohigh-fat feeding decreases adiponectinand upregulates expression of adiponec-tin receptors in mice (12).

Both adiponectin receptors are ex-pressed in human adipocytes (13–15)and muscle cells (16). Moreover, we havedemonstrated that adiponectin receptorexpression in skeletal muscle is increasedin conditions of insulin resistance andtype 2 diabetes, and an exercise inter-vention for 4 weeks, which improvesmetabolic parameters, also increases cir-culating adiponectin levels and upregu-lates expression of both adiponectinreceptors in skeletal muscle (16). It wasrecently shown that improvement of in-

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Internal Medicine III, University of Leipzig, Leipzig, Germany; the 2Division ofEndocrinology, Diabetes, and Metabolism, Beth Israel Deaconess Medical Center, Harvard Medical School,Boston, Massachusetts; the 3Department of Surgery II, University of Leipzig, Leipzig, Germany; and the4Division of Nutritional Epidemiology, The National Institute of Environmental Medicine, Karolinska Insti-tutet, Stockholm, Sweden.

Address correspondence and reprint requests to Christos S. Mantzoros, MD, 330 Brookline Ave., ST 816,Boston, MA 02215. E-mail: [email protected].

Received for publication 2 July 2007 and accepted in revised form 12 September 2007.Published ahead of print at http://care.diabetesjournals.org on 18 September 2007. DOI: 10.2337/dc07-

1257.M.B. and C.J.W. contributed equally to this work.Abbreviations: AdipoR1, adiponectin receptor 1; AdipoR2, adiponectin receptor 2; IGT, impaired glu-

cose tolerance; NGT, normal glucose tolerance; OGTT, oral glucose tolerance test; WHR, waist-to-hip ratio.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





sulin sensitivity by thiazolidinediones isnot related to AdipoR1/AdipoR2 expres-sion changes (15).

Here, we first explored associations ofAdipoR1 and/or AdipoR2 expression invisceral and subcutaneous fat with meta-bolic parameters and insulin sensitivity inpaired samples of omental and subcuta-neous adipose tissue from 153 subjectswith a wide range of obesity, body fat dis-tribution, insulin sensitivity, and glucosetolerance in the context of a cross-sectional study. We further tested the hy-pothesis that AdipoR1 and/or AdipoR2mRNA expression in subcutaneous adi-pose tissue is upregulated by the previ-ously described intensive exerciseregimen for 4 weeks in 60 subjects withvarious degrees of insulin resistance (16).


Cross-sectional studyPaired samples of visceral and subcutane-ous adipose tissue were obtained from153 consecutively enrolled Caucasianmen (n � 75) and women (n � 78) whounderwent open abdominal surgery forgastric banding, cholecystectomy, appen-dectomy, weight reduction, abdominalinjuries, or explorative laparotomy. Per-cent body fat was measured by dual-energy X-ray absorptiometry. In addition,abdominal visceral and subcutaneous fatarea was calculated using computed to-mography scans at the level of L4–L5 asdescribed previously (17,18). Areas ofsubcutaneous and intra-abdominal adi-pose tissue (attenuation range of �30 to�190 Hounsfield units) were counted us-ing ImageAccess software (Imagic, Glatt-brugg, Switzerland). In obese subjectsonly (BMI �30 kg/m2), the ratio of intra-abdominal visceral divided by abdominalsubcutaneous fat area was calculated asdescribed previously (17–19)

Using oral glucose tolerance tests(OGTTs), we identified 67 individualswith either type 2 diabetes (n � 36) orimpaired glucose tolerance (IGT, n � 31).All subjects had stable weight with nofluctuations of body weight �2% for atleast 3 months before surgery. Patientswith malignant diseases or any acute orchronic inflammatory disease as deter-mined by a leukocyte count �7.0 �109/l, C-reactive protein �5.0 mg/dl, orclinical signs of infection were excludedfrom the study. Samples of visceral andsubcutaneous adipose tissue were imme-diately frozen in liquid nitrogen after ex-

plantation. The study was approved bythe ethics committee of the University ofLeipzig. All subjects gave written in-formed consent before taking part in thestudy. Insulin sensitivity was assessedwith the euglycemic-hyperinsulinemicclamp method as described previously(16,20). Basal blood samples were takenafter an overnight fast. Plasma insulin wasmeasured with an enzyme immunometricassay for the IMMULITE automated ana-lyzer (Diagnostic Products, Los Angeles,CA). Plasma adiponectin levels were as-sessed by radioimmunoassay (Linco Re-search, St. Charles, MO).

Exercise interventional studyWe studied 60 Caucasian men andwomen with no acute or chronic inflam-matory disease, alcohol or drug abuse, ordiabetic retinopathy or nephropathy.These subjects, who have not been in-cluded in the cross-sectional study, werecategorized into groups of normal glucosetolerance (NGT) (n � 20, 9 men and 11women), IGT (n � 20, 9 men and 11women), and type 2 diabetes (n � 20, 11men and 9 women). All subjects were en-rolled in 60 min of supervised physicaltraining sessions 3 days per week as de-scribed previously (16). At baseline andafter 4 weeks of training (48 h after thelast training session), subcutaneous adi-pose tissue and blood samples were ob-tained in the fasting state, and dual-energy X-ray absorptiometry analyses andmeasurements of anthropometric param-eters were performed. All baseline bloodsamples and adipose tissue samples werecollected between 8 and 10 A.M. after anovernight fast. Subcutaneous adipose tis-sue samples were immediately frozen inliquid nitrogen after explantation. Thestudy was approved by the ethics com-mittee, and all subjects gave written in-formed consent.

Analysis of AdipoR1/R2 mRNAexpression in adipose tissueHuman AdipoR1 and AdipoR2 gene ex-pression was measured by quantitativereal-time PCR in a fluorescent tempera-ture cycler using the TaqMan assay, andfluorescence was detected on an ABIPRISM 7000 sequence detector (AppliedBiosystems, Darmstadt, Germany) as de-scribed previously (16). The followingprimers were used: human AdipoR1, TTCTTC CTC ATG GCT GTG ATG T (sense)and AAG AAG CGC TCA GGA ATT CG(antisense); human AdipoR2, ATA GGGCAG ATA GGC TGG TTG A (sense) and

GGA TCC GGG CAG CAT ACA (anti-sense); and human 18s rRNA, TGC CATGTC TAA GTA CGC ACG (sense) andTTG ATA GGG CAG ACG TTC GA (anti-sense).

Statistical analysesIn both cross-sectional and interventionalstudies, comparisons of descriptive char-acteristics, expressed as means � SE ormeans with 95% CIs, were conducted us-ing one-way ANOVA with Bonferronicorrected post hoc tests and were re-peated using a nonparametric Kruskal-Wallis test Nonparametric Spearmancorrelation coefficients were calculated toexamine the cross-sectional associationsof adiponectin and its receptors withanthropometric and insulin resistance–related parameters. Analyses were re-peated after adjustments for age, body fat,and sex. For the interventional study, posthoc comparisons of baseline and after-training measures, expressed as means �SE, were conducted using paired t testswithin groups of glucose tolerance (NGT,IGT, and type 2 diabetes). Differences inchange between groups in measurementswere compared by one-way ANOVA withBonferroni corrected post hoc tests.

RESULTS — We present descriptivecharacteristics of participants in the cross-sectional study (n � 153) and exerciseintervention study (n � 60) in Table 1. Inlean and healthy subjects, AdipoR2 geneexpression in adipose tissue was higherthan that of AdipoR1 (P � 0.05), and ex-pressions of both AdipoR1 and AdipoR2were substantially lower in subcutaneousadipose tissue than in muscle (P � 0.05for both) (Fig. 1A). In addition, visceraland subcutaneous AdipoR2 expressionswere highly interrelated (Table 2) andboth significantly higher in lean than inobese subjects and in subjects with NGTcompared with subjects with IGT or type2 diabetes (Fig. 1B). Expression of Adi-poR1 did not differ between lean,subcutaneous obese, or visceral obesesubjects in either depot (Fig. 1A). Subcu-taneous AdipoR1 mRNA expression waslower in subjects with NGT comparedwith subjects with IGT or type 2 diabetes(Fig. 1B). Age was negatively correlatedwith circulating adiponectin levels (r ��0.18, P � 0.033) as well as AdipoR2expression in both subcutaneous (r ��0.30, P � 0.001) and visceral adiposetissue (r � �0.32, P � 0.001). The asso-ciations of adiponectin receptor gene ex-pression with insulin resistance and

Bluher and Associates


obesity were similar, however, when sub-jects were stratified by age in two catego-ries, i.e., � or �60 years (data not shown).

We then examined age-adjusted asso-ciations of adiponectin receptor expres-sion in fat with anthropometric andmetabolic variables among subjects en-rolled in the cross-sectional study (Table2). Serum adiponectin was significantlyand negatively correlated with adipositymeasures, IGT, and dyslipidemia. Afteradjustments for age, sex, and percentbody fat, adiponectin remained nega-tively associated with waist-to-hip ratio(WHR) and free fatty acids and positivelyassociated with HDL cholesterol and in-sulin sensitivity (clamp). Expression ofAdipoR1 in visceral fat was only corre-lated with subcutaneous AdipoR1 mRNAexpression and free fatty acid serum con-centrations beyond the effects of age, sex,and percent body fat (Table 2). In con-trast, AdipoR2 expression in both fat de-pots showed strong correlations withadiponectin, HDL cholesterol, measuresof adiposity, IGT, and dyslipidemia (Ta-ble 2). Among subjects in the interven-tional study, moderate correlationsbetween AdipoR1 expression in subcuta-

neous fat and adiposity/IGT (r � 0.5 forBMI, WHR, percent body fat, and 2-hOGTT glucose) disappeared after adjust-ing for sex and percent body fat. Associa-t ions o f AdipoR2 expres s ion insubcutaneous fat with the metabolic vari-ables studied were very similar in bothdirection and magnitude to those re-ported for the cross-sectional study (datanot shown).

Associations between AdipoR2 ex-pression and metabolic variables also re-mained significant upon adjustments forage, sex, and body fat. To eliminate a po-tential effect of type 2 diabetes (n � 26 of153, 17%), we performed additionalanalyses excluding type 2 diabetic sub-jects. Neither mean AdipoR2 mRNA ex-pression differences nor statisticallysignificant associations were altered byexcluding subjects with type 2 diabetesfrom the analyses (data not shown).

Four weeks of intensive physicaltraining resulted in significant improve-ments of body weight, percent body fat,insulin sensitivity, and circulating adi-ponectin, leading to increases in skeletalmuscle AdipoR1/R2 mRNA expression(16). We report here that the same inter-

vention led to an increased expression ofAdipoR1/R2 in subcutaneous fat (Fig.1B). Subcutaneous AdipoR1 expressionwas elevated by 63% in subjects withNGT, by 48% in participants with IGT,and by 89% in type 2 diabetic patients.Similarly, AdipoR2 expression in subcu-taneous fat was significantly higher, withincreases of 48, 58, and 57% for the samesubgroups, respectively. These changes inreceptor expression were independent ofdecreases in body weight and fat mass(data not shown, adjusted P � 0.05 forboth in all groups). Finally, the increase ofAdipoR2 in subcutaneous fat is signifi-cantly and positively correlated with theincreases of AdipoR2 in skeletal muscle(R2 � 0.34, P � 0.01).

CONCLUSIONS — Adiponectin hasbecome widely accepted as a key regula-tor of insulin sensitivity and metabolism(1–5), but the physiological regulationand role of adiponectin receptors, Adi-poR1 and AdipoR2, in mediating the ben-eficial effects of adiponectin in humansremain to be fully elucidated. We confirmage-, sex-, and BMI-independent relation-ships between circulating adiponectin

Table 1—Descriptive and metabolic characteristics, along with adiponectin and receptor gene expression, from a cross-sectional study ofsubjects categorized as lean, subcutaneous obese, or visceral obese and from a separate exercise intervention trial of 60 subjects categorizedin groups of NGT, IGT, and type 2 diabetes

Variable

Cross-sectional study Exercise intervention study

LeanSubcutaneous

obeseVisceralobese NGT IGT

Type 2diabetes

n 58 58 37 20 20 20Male/female sex 28/30 28/30 19/18 9/11 9/11 11/9Age (years) 50.2 � 2.1 55.3 � 1.7 64.4 � 1.9‡§ 32.8 � 2.5 56.0 � 3.6‡ 53.1 � 1.5‡Anthropometric

BMI (kg/m2) 23.9 � 0.2 35.9 � 0.9‡ 33.6 � 1.0‡ 24.3 � 0.3 29.8 � 0.9‡ 31.4 � 0.7‡WHR 0.85 � 0.02 1.05 � 0.02‡ 1.13 � 0.02‡§ 0.84 � 0.02 1.21 � 0.04‡ 1.28 � 0.03‡Body fat (%) 21.6 � 0.4 41.1 � 1.2‡ 32.9 � 1.1‡� 24.5 � 0.7 34.9 � 1.9‡ 38.2 � 1.8‡Visceral fat area (cm2) 60.9 � 2.5 146.6 � 4.1‡ 272.7 � 10.2‡� — — —Subcutaneous fat area (cm2) 77.1 � 3.9 620.5 � 39.7‡ 381.4 � 28.7‡� — — —Computed tomography ratio — 0.48 � 0.04 0.84 � 0.03� — — —

MetabolicFasting plasma insulin (pmol/l) 28.0 � 1.5 170.4 � 17.2‡ 198.0 � 18.5‡ 66.0 � 8.0 695.0 � 110‡ 319.0 � 48*§Fasting plasma glucose (mmol/l) 5.4 � 0.1 5.9 � 0.19† 5.8 � 0.2 5.2 � 0.1 5.7 � 0.12* 6.2 � 0.1‡§2-h OGTT glucose (mmol/l) 5.9 � 0.1 6.8 � 0.30* 7.4 � 0.39† 6.0 � 0.2 9.4 � 0.20‡ 13.1 � 0.3‡�

WBGU (�mol � kg�1 � min�1) 97.2 � 1.0 57.4 � 4.0‡ 38.6 � 4.7‡§ 75.9 � 3.8 18.7 � 2.0‡ 21.5 � 2.1‡Lipids

Total cholesterol (mg/dl) 201.0 � 4.8 202.2 � 3.9 224.2 � 4.8†§ 178.7 � 4.3 206.5 � 4.6† 216.6 � 6.2‡HDL cholesterol (mg/dl) 62.0 � 2.5 48.6 � 1.9‡ 41.6 � 3.0‡ 46.4 � 1.9 63.4 � 2.7‡ 56.8 � 2.7*LDL cholesterol (mg/dl) 107.1 � 3.9 110.6 � 3.7 136.5 � 4.4‡� 90.5 � 3.9 124.5 � 4.6‡ 127.6 � 7.3‡Free fatty acids (mmol) 0.30 � 0.02 0.66 � 0.05‡ 0.78 � 0.05‡ 0.41 � 0.04 0.54 � 0.06 0.56 � 0.06

Data are means � SE and were compared using ANOVA with Bonferroni corrections for post hoc tests. *P � 0.05, †P � 0.01, ‡P � 0.001 versus lean group(cross-sectional study) or NGT group (intervention study). §P � 0.01, �P � 0.001 versus subcutaneous obese group (cross-sectional study) or IGT group(intervention study). WBGU, whole-body glucose uptake during the steady state of euglycemic-hyperinsulinemic clamp.

Regulation of adiponectin receptors


and measures of obesity and insulin sen-sitivity, supporting previous reports thatadiponectin is related to central adiposityand improved insulin sensitivity beyondassociations with total body fat (21). Inthis study, we confirm using quantitativereal-time PCR that both receptors are ex-pressed in subcutaneous and visceral ad-ipose tissue but at levels significantlylower than those in skeletal muscle. Wealso demonstrate that expression of Adi-poR2 is several-fold higher than that ofAdipoR1 in fat. It correlates negativelywith obesity, lipid level, glycemia, and in-sulin resistance and is significantly de-creased in states of insulin resistanceincluding obesity and type 2 diabetes.

It was recently shown that in elderlymen, high adiponectin levels are associ-ated with increased all-cause and cardio-vascular disease mortality (22). However,subgroup analyses in individuals aged�60 years revealed similar associations ofadiponectin receptor gene expressionwith insulin resistance and obesity com-pared with younger subjects. In parallel toour previous analysis of skeletal muscleAdipoR1/R2 mRNA expression (16), 4weeks of intensive exercise training signif-icantly elevated gene expression of bothAdipoR1 and AdipoR2 in subcutaneousadipose tissue in subjects with NGT, IGT,and type 2 diabetes, with comparable rel-ative increases across groups. However,

because of the lack of association betweenbaseline AdipoR1 expression and mostparameters in the cross-sectional study,the potential mechanisms causing in-creased AdipoR1 expression in responseto exercise remain elusive. Therefore,more sophisticated study designs are nec-essary to elucidate the causal factors fortraining-induced AdipoR1 expressionchanges.

Only a limited number of studieshave examined adiponectin receptor ex-pression in adipose tissue in a small num-ber of humans and have yieldedinconclusive results (13,15,23). A recentcross-sectional study reported decreasedAdipoR2 expression, similar to our study,

Figure 1—A: Cross-sectional study. Adiponectin and adiponectin receptor gene expression in visceral and subcutaneous (SC) adipose tissue atbaseline in lean subjects and subjects who are classified as having subcutaneous and visceral obesity. *P � 0.05, **P � 0.01, ***P � 0.001 versuslean. B: Interventional study. Adiponectin receptor gene expression in skeletal muscle and subcutaneous adipose tissue at baseline (�) and afterintensive physical training (f) in subjects with NGT, IGT, and type 2 diabetes (T2D). *P � 0.01, **P � 0.001 versus baseline; P � 0.01, P �0.001 versus NGT; #P � 0.05 versus IGT.



but no differences in AdipoR1 expressionin intra-abdominal adipose tissue amongobese subjects with and without diabetescompared with lean control subjects (23).Our study confirms these findings in vis-ceral adipose samples from a larger studysample (n � 153) and extends previousresults by demonstrating similar associa-tions in subcutaneous fat.

Similar to our study, a recent weightloss intervention study of lean and obesewomen consistently reported reducedAdipoR2 gene expression in subcutane-ous adipose tissue from obese versus leanwomen (13) and significant decreases inAdipoR1 with obesity. In contrast to ourand other previous (23) findings, how-ever, this study did not show decreasedAdipoR2 mRNA expression in visceral fat.These differences may be due to the ex-clusion of males and a lower mean age of

this study population (13). We confirmthe data for to AdipoR2 expression in sub-cutaneous adipose tissue using a longerstudy, and consistent with the previousstudy (13), we show that serum adiponec-tin and adiponectin receptor expressionsincrease with exercise in both lean andobese subjects independent of glycemiccontrol. Moreover, in the exercise inter-vention study, we also show that AdipoR1expression in subcutaneous adipose tis-sue is �20-fold lower and AdipoR2 ex-pression is �1.5-fold lower than inmuscle in humans.

Although AdipoR1/R2 gene expres-sion increases in both skeletal muscle andsubcutaneous fat with intensive exercise,mRNA expression of both receptors ishigher and shows greater improvementswith exercise training in skeletal musclethan in subcutaneous adipose tissue. It

was recently shown that the insulin-sensitizing effects of thiazolidinedionesare not linked to AdipoR1/R2 expressionchanges (15), suggesting that mecha-nisms other than improved insulin sensi-tivity cause increased AdipoR1/R2 mRNAlevels in response to training. Additionalparacrine or autocrine adiponectin effectsin adipose tissue may explain the differentaspects of AdipoR1/R2 mRNA expressionregulation between fat and muscle. Wefurther investigated whether adiponectinreceptor mRNA expression varies in rela-tion to differences in fat distribution andglycemic control. Based on computedtomography scanning measurements(L4–L5) of abdominal visceral and sub-cutaneous fat areas, obese subjects werefurther categorized as predominantly vis-ceral or subcutaneous obese, with pre-dominantly visceral obesity defined as aratio of visceral-to-subcutaneous fat area�0.5, as described previously (17,18).Independent of fat distribution, AdipoR2mRNA was reduced in subcutaneous andvisceral fat among obese subjects com-pared with lean individuals. No such dif-ferences were found for AdipoR1 mRNAexpression among all groups. Similarly,AdipoR2 mRNA expression in subcutane-ous fat was reduced in patients with IGTand type 2 diabetes, whereas AdipoR1mRNA was higher in IGT and type 2 dia-betic patients. It has been suggested pre-viously that AdipoR2 may play a moresubstantial role in the pathogenesis oftype 2 diabetes (23–25). This hypothesisis supported by the significant associa-tions of AdipoR2 mRNA expression in ad-ipose tissue with measures of obesity anddiabetes in this study. Dysregulation ofAdipoR2 expression (6,23) in adipose tis-sue may promote accumulation of lipidsdue to reduced adiponectin action. Fur-ther studies are needed to fully clarify thepotentially distinct roles of AdipoR1 andAdipoR2 in regulating the action of adi-ponectin in various tissues and metabolicconditions.

In summary, we show that AdipoR2mRNA expression in subcutaneous andvisceral adipose tissue is reduced in statesof obesity and type 2 diabetes. In contrastto AdipoR1 mRNA expression, AdipoR2mRNA expression in fat correlates withcirculating adiponectin levels, lipid levels,parameters of insulin sensitivity, and gly-cemic control. An exercise interventionfor 4 weeks resulted in increased expres-sion of both adiponectin and adiponectinreceptors, which may thus mediate the

Table 2—Spearman correlation matrix of gene expression of AdipoR1 and AdipoR2 in visceraland subcutaneous adipose tissue with study variables for cross-sectional study subjects

Variable Adiponectin

Subcutaneous Visceral

AdipoR1 AdipoR2 AdipoR1 AdipoR2

Age adjustedBMI �0.34* �0.10 �0.63* 0.12 �0.68*WHR �0.39* �0.23† �0.49* 0.06 �0.46*Body fat (%) �0.36* �0.16 �0.50* 0.12 �0.65*Fasting plasma glucose �0.23† �0.01 �0.30* �0.06 �0.32*Fasting plasma insulin �0.36* �0.11 �0.60* 0.11 �0.67*2-h OGTT glucose �0.20‡ �0.13 �0.22‡ �0.10 �0.19‡WGBU 0.42* 0.25† 0.63* 0.03 0.69*Total cholesterol �0.15 �0.02 �0.18‡ 0.05 �0.21†HDL cholesterol 0.29† 0.15 0.32* �0.11 0.35*LDL cholesterol �0.19‡ �0.05 �0.21† 0.03 �0.31*Free fatty acids �0.33* �0.06 �0.41* 0.28‡ �0.50*Adiponectin 0.13 0.29† �0.02 0.36*AdipoR1 subcutaneous fat 0.08 0.22† 0.15AdipoR2 subcutaneous fat �0.04 0.80*AdipoR1 visceral fat �0.04

Age, sex, and body fat adjustedBMI �0.04 0.09 �0.45* 0.01 �0.33*WHR �0.24† �0.15 �0.35* �0.08 �0.27†Fasting plasma glucose �0.16 0.04 �0.21‡ �0.10 �0.22†Fasting plasma insulin �0.19‡ �0.01 �0.42* 0.05 �0.45*2-h OGTT glucose �0.14 �0.11 �0.15 �0.15 �0.11WGBU 0.29† 0.17 0.48* 0.08 0.49*Total cholesterol �0.10 0.01 �0.12 0.03 �0.14HDL cholesterol 0.18‡ 0.11 0.20‡ �0.07 0.22†LDL cholesterol �0.13 �0.03 �0.14 0.02 �0.24†Free fatty acids �0.19‡ 0.02 �0.21‡ 0.27† �0.24†Adiponectin 0.06 0.14 0.03 0.19‡AdipoR1 subcutaneous fat 0.00 0.26† 0.07AdipoR2 subcutaneous fat 0.03 0.72*AdipoR1 visceral fat 0.04

n � 153. *P � 0.001; †P � 0.01; ‡P � 0.05. WBGU, whole-body glucose uptake during the steady state ofthe euglycemic-hyperinsulinemic clamp.

Regulation of adiponectin receptors


beneficial effects of exercise on insulin re-sistance, glycemia, and lipidemia.

Acknowledgments— This work was sup-ported by grants from Deutsche Forschungs-gemeinschaft (BL 580/3-1 to M.B.), theClinical Research group “Atherobesity” KFO152 (project BL 833/1-1 to M.B.), and the In-terdisciplinary Center of Clinical ResearchLeipzig at the Faculty of Medicine of the Uni-versity of Leipzig (project B24 to M.B.), byNational Institute of Diabetes and Digestiveand Kidney Diseases (NIDDK) Grant R0158785 (to C.S.M.), and, in part, by NationalInstitutes of Health NIDDK Grant P30 DK57521 (“The Metabolic Physiology Core”), bythe Humboldt Foundation, and by a discre-tionary grant from Beth Israel Deaconess Med-ical Center (to C.S.M.).

References1. Cnop M, Havel PJ, Utzschneider KM, Carr

DB, Sinha MK, Boyko EJ, Retzlaff BM,Knopp RH, Brunzell JD, Kahn SE: Rela-tionship of adiponectin to body fat distri-bution, insulin sensitivity and plasmalipoproteins: evidence for independentroles of age and sex. Diabetologia 46:459–469, 2003

2. Pajvani UB, Scherer PE: Adiponectin: sys-temic contributor to insulin sensitivity.Curr Diab Rep 3:207–213, 2003

3. Berg AH, Combs TP, Scherer PE:ACRP30/adiponectin: an adipokine regu-lating glucose and lipid metabolism.Trends Endocrinol Metab 13:84–89, 2002

4. Wolf AM, Wolf D, Rumpold H, Enrich B,Tilg H: Adiponectin induces the anti-in-flammatory cytokines IL-10 and IL-1RAin human leukocytes. Biochem Biophys ResCommun 323:630–635, 2004

5. Kobashi C, Urakaze M, Kishida M, Kiba-yashi E, Kobayashi H, Kihara S, Fu-nahashi T, Takata M, Temaru R, Sato A,Yamazaki K, Nakamura N, Kobayashi M:Adiponectin inhibits endothelial synthe-sis of interleukin-8. Circ Res 97:1245–1252, 2005

6. Kadowaki T, Yamauchi T: Adiponectinand adiponectin receptors. Endocr Rev 26:439–451, 2005

7. Yamauchi T, Kamon J, Ito Y, Tsuchida A,Yokomizo T, Kita S, Sugiyama T, Miy-agishi M, Hara K, Tsunoda M, MurakamiK, Ohteki T, Uchida S, Takekawa S, WakiH, Tsuno NH, Shibata Y, Terauchi Y,Froguel P, Tobe K, Koyasu S, Taira K,Kitamura T, Shimizu T, Nagai R, Kad-

owaki T: Cloning of adiponectin recep-tors that mediate antidiabetic metaboliceffects. Nature 423:762–769, 2003

8. Goldstein BJ, Scalia R: Adiponectin: anovel adipokine linking adipocytes andvascular function. J Clin Endocrinol Metab89:2563–2568, 2004

9. Yamauchi T, Nio Y, Maki T, Kobayashi M,Takazawa T, Iwabu M, Okada-Iwabu M,Kawamoto S, Kubota N, Kubota T, Ito Y,Kamon J, Tsuchida A, Kumagai K, Ko-zono H, Hada Y, Ogata H, Tokuyama K,Tsunoda M, Ide T, Murakami K, AwazawaM, Takamoto I, Froguel P, Hara K, TobeK, Nagai R, Ueki K, Kadowaki T: Targeteddisruption of AdipoR1 and AdipoR2causes abrogation of adiponectin bindingand metabolic actions. Nat Med 13:332–339, 2007

10. Kharroubi I, Rasschaert J, Eizirik DL,Cnop M: Expression of adiponectin re-ceptors in pancreatic cells. BiochemBiophys Res Commun 312:1118 –1122,2003

11. Tsuchida A, Yamauchi T, Ito Y, Hada Y,Maki T, Takekawa S, Kamon J, KobayashiM, Suzuki R, Hara K, Kubota N, TerauchiY, Froguel P, Nakae J, Kasuga M, Accili D,Tobe K, Ueki K, Nagai R, Kadowaki T:Insulin/Foxo1 pathway regulates expres-sion levels of adiponectin receptors andadiponectin sensitivity. J Biol Chem 279:30817–30822, 2004

12. Bullen JW Jr, Bluher S, Kelesidis T, Mant-zoros CS: Regulation of adiponectin andits receptors in response to developmentof diet-induced obesity in mice. Am JPhysiol 292:E1079–E1086, 2007

13. Rasmussen MS, Lihn AS, Pedersen SB,Bruun JM, Rasmussen M, Richelsen B:Adiponectin receptors in human adiposetissue: effects of obesity, weight loss, andfat depots. Obesity (Silver Spring) 14:28–35, 2006

14. Fasshauer M, Klein J, Kralisch S, Klier M,Lossner U, Bluher M, Paschke R: Growthhormone is a positive regulator of adi-ponectin receptor 2 in 3T3–L1 adipo-cytes. FEBS Lett 558:27–32, 2004

15. Li W, Tonelli J, Kishore P, Owen R, Good-man E, Scherer PE, Hawkins M: Insulin-sensitizing effects of thiazolidinedionesare not linked to adiponectin receptor ex-pression in human fat or muscle. Am JPhysiol 292:E1301–E1307, 2007

16. Bluher M, Bullen JW, Jr, Lee JH, KralischS, Fasshauer M, Kloting N, Niebauer J,Schon MR, Williams CJ, Mantzoros CS:Circulating adiponectin and expressionof adiponectin receptors in human skel-

etal muscle: associations with metabolicparameters and insulin resistance andregulation by physical training. J ClinEndocrinol Metab 91:2310 –2316, 2006

17. Berndt J, Kloting N, Kralisch S, Kovacs P,Fasshauer M, Schon MR, Stumvoll M,Bluher M: Plasma visfatin concentrationsand fat depot-specific mRNA expressionin humans. Diabetes 54:2911–2916, 2005

18. Bluher M, Engeli S, Kloting N, Berndt J,Fasshauer M, Batkai S, Pacher P, SchonMR, Jordan J, Stumvoll M: Dysregulationof the peripheral and adipose tissue endo-cannabinoid system in human abdominalobesity. Diabetes 55:3053–3060, 2006

19. Fujioka S, Matsuzawa Y, Tokunaga K, Ta-rui S: Contribution of intra-abdominal fataccumulation to the impairment of glu-cose and lipid metabolism in human obe-sity. Metabolism 36:54–59, 1987

20. DeFronzo RA, Tobin JD, Andres R: Glu-cose clamp technique: a method for quan-tifying insulin secretion and resistance.Am J Physiol 237:E214–E223, 1979

21. Mantzoros CS, Li T, Manson JE, Meigs JB,Hu FB: Circulating adiponectin levels areassociated with better glycemic control,more favorable lipid profile, and reducedinflammation in women with type 2 dia-betes. J Clin Endocrinol Metab 90:4542–4548, 2005

22. Wannamethee SG, Whincup PH, LennonL, Sattar N: Circulating adiponectin levelsand mortality in elderly men with andwithout cardiovascular disease and heartfailure. Arch Intern Med 167:1510–1517,2007

23. Morinigo R, Musri M, Vidal J, Casamit-jana R, Delgado S, Lacy AM, Ayuso C,Gomis R, Corominola H: Intra-abdominalfat adiponectin receptors expression andcardiovascular metabolic risk factors inobesity and diabetes. Obes Surg 16:745–751, 2006

24. Bauche IB, it El MS, Rezsohazy R, Fu-nahashi T, Maeda N, Miranda LM, Bri-chard SM: Adiponectin downregulates itsown production and the expression of itsAdipoR2 receptor in transgenic mice. Bio-chem Biophys Res Commun 345:1414–1424, 2006

25. Damcott CM, Ott SH, Pollin TI, ReinhartLJ, Wang J, O’Connell JR, Mitchell BD,Shuldiner AR: Genetic variation in adi-ponectin receptor 1 and adiponectin re-ceptor 2 is associated with type 2 diabetesin the Old Order Amish. Diabetes 54:2245–2250, 2005



Waist Circumference Thresholds Provide anAccurate and Widely Applicable Methodfor the Discrimination of DiabetesOBESITY IN ASIA COLLABORATION*

E xcess weight, particularly centralobesity, is recognized to be a majordeterminant of diabetes risk in all

populations, with the magnitude of theassociation reported as being stronger inAsians than whites (1–3). Consequently,indicators of overweight have been incor-porated into several guidelines for theearly identification of individuals withtype 2 diabetes (4). However, the anthro-pometric cut points for different ethnicgroups have been determined in variousways, leading to uncertainty about theirapplicability to diabetes screening. Here,we clarify current uncertainty regardingethnic differences in the relationship be-tween overweight and diabetes andwhether there is a single measure of over-weight that can be determined routinelyand applied universally in clinical prac-tice to facilitate earlier detection of diabe-tes in the general population.


Eligibility criteria for studies in theObesity in Asia CollaborationStudies were eligible if they contained thefollowing information: age, sex, weight,height, waist circumference, hip circum-ference, fasting plasma glucose, bloodpressure, and smoking status (5). Diabe-tes was defined as fasting plasma glucose�7 mmol/l. Individuals with a history ofdiabetes or on diabetic medication wereexcluded.

Logistic regression models, stratified

by sex and study and adjusted for age(and subsequently by smoking and bloodpressure), were used to estimate the oddsratio (OR) and 95% CI for prevalent dia-betes associated with a 0.5-SD incrementin each of the measures. To assess the abil-ity of each anthropometric variable to dis-criminate between those with andwithout diabetes, areas under the receiveroperating characteristic curves were com-puted. The areas under the curve (AUCs)was subsequently pooled to find region-specific AUCs using a random-effectsmeta-analysis.

The dataset was randomly dividedinto two equal-sized samples, one ofwhich was used to derive the anthropo-metric threshold values for diabetes. Cutpoints from the receiver operating char-acteristic that maximized the sum of sen-sitivity and specificity were identified foreach sex and study. These thresholdswere subsequently validated on the sec-ond random half-sample.

RESULTS — Data were available for155,122 individuals (86% Asian; 52% fe-male) from 18 study populations from 10countries in the Asia-Pacific region (sup-plemental Tables 1 and 2 [available in anonline appendix at http://dx.doi.org/10.2337/dc07-1455]).

In women, the association betweenBMI, waist circumference, and waist-to-hip ratio (WHR) with diabetes was �10–20% stronger in whites compared withAsians (Fig. 1). In men, this was true onlyfor BMI. This remained unchanged after

adjustment for smoking and blood pres-sure. In all groups (with the exception ofwhite men), measures of central obesitywere more strongly associated with prev-alent diabetes than BMI (Fig. 1).

At any given level of BMI, waist cir-cumference, or WHR, the prevalence ofdiabetes was consistently higher in Asiansthan in whites. At a BMI of 24 kg/m2, theproportion of men with diabetes was 5%in Asians compared with 2% in whites; inwomen, the corresponding values were 5and 1%, respectively. At a waist circum-ference of 90 cm or WHR of 0.9, the pro-portion of Asian men with diabetes was6% compared with 2% in white men; inwomen, at a waist circumference of 80 cmand a WHR of 0.8, the estimates were �5and 1%, respectively.

The ability of each of the measures todiscriminate diabetes ranged from 0.63 to0.71 in men and from 0.66 to 0.80 inwomen (supplemental tables 3 and 4).The AUCs tended to be slightly (but inmost instances nonsignificantly) higherfor waist circumference than for BMI orWHR across the groups. Measuring BMIand/or waist circumference or BMI and/orWHR did not improve the discriminatorycapabilities of any single measure.

Anthropometric cut points for the op-timal discrimination of diabetes werelower in Asians compared with whites.The optimal cut points for Asian menwere BMI 23.7 kg/m2, waist circumfer-ence 85 cm, and WHR 0.90 versus 27.7kg/m2, 99 cm, and 0.94, respectively, inwhite men. For Asian women, the corre-sponding values were 24.5 kg/m2, 80 cm,and 0.80, and in white women the valueswere 27.9 kg/m2, 85 cm, and 0.85, re-spectively. These cut points optimizedsensitivity and specificity such that bothvalues nearly always exceeded 60% in allgroups.

CONCLUSIONS — Irrespective ofwhich measure of excess weight is used,the prevalence of diabetes is consistentlyhigher among Asians than whites at anygiven level, in agreement with earlier find-ings (6). The mechanisms that might un-derlie this apparent greater susceptibilityamong Asians are unknown, but the data

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Address correspondence and reprint requests to Rachel Huxley, DPhil, The George Institute for InternationalHealth, P.O. Box M201, Missenden Road, Sydney, NSW 2050, Australia. E-mail: [email protected].

Received for publication 27 July 2007 and accepted in revised form 29 August 2007.Published ahead of print at http://care.diabetesjouranals.org on 5 September 2007. DOI: 10.2337/dc07-

1455.*A complete list of the members of the Obesity in Asia Collaboration can be found in the APPENDIX.Additional information for this article can be found in an online appendix at http://dx.doi.org/10.2337/

dc07-1455.Abbreviations: AUC, area under the curve; WHR, waist-to-hip ratio.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion



C a r d i o v a s c u l a r a n d M e t a b o l i c R i s kB R I E F R E P O R T


do not consistently support the hypothe-sis that this is due to a stronger associationbetween diabetes and body size amongAsians compared with whites.

There were marginal differences inthe capability of BMI, waist circumfer-ence, and WHR in discriminating diabe-tes, although measures of central obesity(waist circumference in particular)tended to perform slightly better thanBMI. Compared with either BMI or WHR,waist circumference is a more readily un-derstood measure that, with adequatetraining, is easily determined with a tapemeasure. Our findings suggest that cur-rent recommended waist circumferencecut points should be modified to 80 cm inAsian women, 85 cm in white women, 85cm in Asian men, and 99 cm in white mento optimize the discrimination of diabetesin these populations.

Acknowledgments—This study was supported by the National

Health and Medical Research Council of Aus-tralia, the National Heart Foundation of Aus-tralia, and sanofi-aventis.

APPENDIX

Members of the Obesity in AsiaCollaborationWriting committee. Rachel Huxley, Fe-derica Barzi, Crystal M.Y. Lee, Scott Lear,Jonathan Shaw, Tai Hing Lam, Ian Cater-son, Fereidoun Azizi, Jeetesh Patel, PaibulSuriyawongpaisal, Sang Woo Oh, Jae-Heon Kang, Tim Gill, Paul Zimmet, PhilipT. James, and Mark Woodward.Statistical analyses team. FedericaBarzi and Mark Woodward.Principal collaborators. John Adam,Fereidoun Azizi, Corazon Barba, Zhou

Beifan, Chen Chunming, Stephen Colagi-uri, Jeffery Cutter, Chee Weng Fong, Gra-ham Giles, Kuo-Chin Huang, EdwardJanus, Jae-Heon Kang, Gary Ko, ShinichiKuriyama, Tai Hing Lam, Scott Lear,Viswanathan Mohan, Sang Woo Oh,Jeetesh Patel, Dorairaj Prabhakaran, Sri-nath Reddy, Jonathan Shaw, PiyamitrSritara, Paibul Suriyawongpaisal, TimWelborn, and Paul Zimmet.

References1. McKeigue PM, Shah B, Marmot MG: Re-

lation of central obesity and insulin resis-tance with high diabetes prevalence andcardiovascular risk in South Asians. Lan-cet 337:382–386, 1991

2. James WPT, Jackson-Leach R, NiMhurchu C, Kalmara E, Shayeghi M,Rigby N: Overweight and obesity (high

Figure 1—Age-adjusted ORs and 95% CIs for prevalent type 2 diabetes associated with a 0.5-SD increment in each anthropometric measure: BMI,waist circumference (Waist), and WHR (Waist:Hip). Results are shown separately by sex and ethnic group (Asian and white). The strength of theassociation between waist circumference and diabetes and between WHR and diabetes are compared against the strength of the association betweenBMI and diabetes. For each variable the strength of the association with diabetes is compared between Asian and white individuals. P values for thedifferences are shown.

Obesity in Asia Collaboration


body mass index). In Comparative Quan-tification of Health Risks. Global and Re-gional Burden of Disease Attributable toSelected Major Risk Factors. Chpt. 8, vol. 1.Ezzati M, Lopez AD, Rodgers A, MurrayCJL, Eds. Geneva, World Health Org.,2004

3. Whincup PH, Gilg JA, Papacosta O, Sey-mour C, Miller GJ, Alberti KG, Cook DG:Early evidence of ethnic differences in car-

diovascular risk: cross sectional compari-son of British South Asian and whitechildren. BMJ 324:635, 2002

4. Griffin SJ, Little PS, Hales CN, KinmonthAL, Wareham NJ: Diabetes risk score: to-wards earlier detection of type 2 diabetesin general practice. Diabetes Metab Res Rev16:164–171, 2000

5. Obesity in Asia Collaboration: Ethniccomparisons of obesity in the Asia-Pacific

region: protocol for a collaborative over-view of cross-sectional studies. Obes Rev6:193–198, 2005

6. Razak F, Anand S, Vuksan V, Davis B,Jacobs R, Teo KK, Yusuf S, the SHAREInvestigators: Ethnic differences in the re-lationships between obesity and glucose-metabolic abnormalities: a cross-sectionalpopulation-based study. Int J Obes (Lond)29:656–667, 2005

Waist circumference threshold and diabetes


Change in Albuminuria Is Predictive ofCardiovascular Outcome in NormotensivePatients With Type 2 Diabetes andMicroalbuminuriaADRIENNE A.M. ZANDBERGEN, MD, PHD

1

LIFFERT VOGT, MD2

DICK DE ZEEUW, MD, PHD2

STEVEN W.J. LAMBERTS, MD, PHD1

ROB J.T.H. OUWENDIJK, MD, PHD3

MARINUS G.A. BAGGEN, MD, PHD3

AART H. BOOTSMA, MD, PHD1

M icroalbuminuria is associatedwith cardiovascular complica-tions and all-cause mortality in

patients with diabetes (1–3). Inhibitors ofthe renin-angiotensin system (RAS) pro-tect renal and cardiac function in thesepatients, at least partly independent of theassociated blood pressure reduction (4–9). Recently, a few studies showed thatreduction in albuminuria in hypertensivediabetic patients reduces the risk of sub-sequent cardiovascular events (10–12).The question remains whether this riskreduction is explained by the reduction ofhigh blood pressure. No data are availablefor normotensive diabetic patients. There-fore, we investigated whether sustainedchange in albuminuria independentlypredicts cardiovascular outcome in pa-tients with type 2 diabetes and microalbu-minuria but without hypertension.

RESEARCH DESIGN ANDMETHODS — The present study is aprospective follow-up study of 67 normo-tensive patients (baseline blood pressure�140/90 mmHg without antihypertensivetreatment) with type 2 diabetes and mi-croalbuminuria (urinary albumin excretion20–200 mg/l), who participated in a previ-ously published, larger, randomized, dou-

ble-blind, placebo-controlled, multicentertrial investigating the short-term effects ofthe angiotensin-receptor antagonist losar-tan on microalbuminuria (7). Exclusioncriteria included a history of macrovascu-lar complications and a baseline serumcreatinine level �150 �mol/l.

After the original 20-week study pe-riod, a cohort of 67 patients from thatstudy was prospectively followed duringmean � SEM 4.7 � 0.1 years. They wererecruited on the basis of their address andreceived standard medical care. Data werecollected on current and past health,medication use, blood pressure, renalfunction, and albuminuria, which was an-nually assessed in morning spot urines.The end point was a composite of death,cardiovascular disease, cerebrovascularevents, and peripheral artery disease.

Paired Student’s t test was used forcomparisons within similar variables. Re-garding the rate of change in albuminuriafrom baseline over each year, three groupswere discerned: one with reduction of al-buminuria of �30%, one with stable al-buminuria (change �30%), and one withrapid progression of albuminuria of atleast 30%. The correlation between rate ofchange in albuminuria at 1 year and cu-mulative event-free survival was analyzed

with the Kaplan-Meier method, the log-rank test, and multivariate Cox regres-sion. The 95% CI of the hazard ratio (HR)was calculated as the exponent of the re-gression coefficient. P values �0.05 de-fined statistical significance. We usedSPSS for Windows (version 12.0; SPSS,Chicago, IL) for all analyses.

RESULTS — Baseline characteristics ofthe three groups, including blood pres-sure and albuminuria, did not differ sig-nificantly, with the exception of age, forwhich we corrected. Albuminuria re-duced from 69.1 mg/l at baseline to 39.4mg/l after 1 year (mean difference�29.7 mg/l [95% CI �39.7 to �19.8],P � 0.0001) and returned to 62.0 mg/lat the end of follow-up in the groupwith albuminuria reduction. In patientswith rapid progression of albuminuria,mean levels were 84.3 mg/l at baseline,223.3 mg/l after 1 year (139.0 mg/l[46.3–209.7], P � 0.01), and 354.1mg/l at the end of follow-up. Albumin-uria levels did not change significantlyin the group with stable albuminuria.Importantly, the course of blood pres-sure was similar in the three groups,without significant changes in systolicor diastolic blood pressure (Fig. 1).

During follow-up, 14 patients (21%)reached the composite end point. A sig-nificant difference in event-free survivalwas observed between the three groups(P � 0.02). Patients with rapid progres-sion of albuminuria were at highest risk toreach the end point, whereas patientswith reduction in albuminuria of �30%were at lowest risk. After adjustment forsex, age, systolic blood pressure, total–to–HDL cholesterol ratio, and current smok-ing in a multivariate Cox regressionmodel, change of albuminuria remainedan independent, significant predictor (HR5.1 [95% CI 1.5–18.1], P � 0.01).

Relevant medication during fol-low-up was used similarly in the threegroups. Following the original study pro-tocol, 63 patients (95%) received an RASinhibitor because of microalbuminuria.At the end of follow-up, 62 patients

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Internal Medicine, Erasmus University Medical Centre, Erasmus University Rot-terdam, Rotterdam, the Netherlands; the 2Department of Internal Medicine and Clinical Pharmacology,University Medical Centre Groningen, Groningen, the Netherlands; and the 3Department of Internal Med-icine, Ikazia Hospital Rotterdam, Rotterdam, the Netherlands.

Address correspondence and reprint requests to Adrienne A.M. Zandbergen, MD, PHD, Erasmus Uni-versity Medical Centre, Internal Medicine, ’s Gravendijkwal 230, 3015 CE, Rotterdam, Netherlands. E-mail:[email protected].


0960.Abbreviations: RAS, renin-angiotensin system.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





(93%) still used RAS inhibition. More-over, neither statin nor aspirin use dif-fered between groups.

CONCLUSIONS — This study dem-onstrates that normotensive patients withtype 2 diabetes and miroalbuminuria runa marked risk for cardiovascular compli-cations. The risk depends on the rate of 1-year change in urinary albumin excretion.Patients with rapid progression of albu-minuria were at highest risk, whereas pa-tients with regression of albuminuria hadthe lowest risk. This association persistedafter adjustment for classic cardiovascularrisk factors.

Besides reducing blood pressure, RASinhibitors are effective in preserving renaland cardiac function in diabetic patients(4–9). Moreover, they reduce albumin-uria up to 40%, significantly more thanother classes of antihypertensive drugs (4).Since albuminuria is strongly associatedwith cardiovascular outcome, changes inalbuminuria during treatment might re-flect changes in cardiovascular diseaserisk (13,14). A few studies recentlyshowed that reduction of albuminuria inhypertensive diabetic patients reduces therisk of subsequent cardiovascular events(10–12,15). However, these studies in-vestigated hypertensive patients, thereby

leaving open the possibility that bloodpressure lowering explains the cardiovas-cular risk reduction, with albuminuriachange just an innocent bystander. To ourknowledge, our study is the first demon-strating that, even with no appreciablechanges or even rises in blood pressure,change in albuminuria differentiates thecardiovascular outcome in type 2 diabeticpatients without hypertension.

An important limitation of this studyis the small sample size. Nonetheless, theassociation that we observed betweenchanges in albuminuria and cardiovascu-lar outcome was statistically significant inmultivariate analysis. The strength of ourstudy lies in the fact that we studied type2 diabetic patients with microalbumin-uria but without hypertension at baselinein a prospective design. Clearly, this studyneeds further follow-up in larger cohorts.

In summary, sustained reduction inalbuminuria reflected cardiovascular riskreduction in type 2 diabetic patients with-out hypertension. Hence, albuminuriachange during treatment seems to revealtherapeutic responsiveness independentof blood pressure changes and is thereforeuseful as a modifiable treatment goal. Theseobservations advocate a more aggressive ap-proach to treating albuminuria in addi-tion to more aggressive cardioprotective

treatment in normotensive diabetic pa-tients with elevated levels of albuminuria.

Acknowledgments— The authors thank thefollowing investigators: Drs. T.L.J.M. van derLoos and F.J.M. Klessens-Godfroy (RotterdamEye Hospital, Rotterdam, the Netherlands),Dr. J.W.F. Elte (Sint Franciscus Hospital, Rot-terdam, the Netherlands), Dr. R.J.M. vanLeendert (Albert Schweitzer Hospital, Zwyn-drecht,theNetherlands),Dr.S.G.THulst(Vliet-land Hospital, Schiedam, the Netherlands),and Dr. J.W. van der Beek-Boter (HofpoortHospital, Woerden, the Netherlands).

These data were presented in abstract formas a poster at the scientific meeting of the In-ternational Diabetes Federation at the CapeTown International Convention Centre, CapeTown, South Africa, on 4 December 2006.

References1. Gerstein HC, Mann JF, Yi Q, Zinman B,

Dinneen SF, Hoogwerf B, Halle JP, Young J,Rashkow A, Joyce C, Nawaz S, Yususf S;HOPE Study Investigators: Albuminuriaand risk of cardiovascular events, death,and heart failure in diabetic and nondiabeticindividuals. JAMA 286:421–426, 2001

2. Arnlov J, Evans JC, Meigs JB, Wang TJ,Fox CS, Levy D, Benjamin EJ, D’AgustinoRB, Vasan RS: Low-grade albuminuriaand incidence of cardiovascular disease

Figure 1—Course of urinary albumin excretion and blood pressureduring follow-up in the three groups of urinary albumin excretionchange at 1 year (F, urinary albumin excretion difference �30% [notsignificant]; f, urinary albumin excretion reduction �30% [P �0.001]; �, urinary albumin excretion progression �30% [P �0.01]). The course of blood pressure was similar in the three groups;changes during follow-up: not significant.

Change in albuminuria and cardiovascular outcome


events in nonhypertensive and nondia-betic individuals: the Framingham HeartStudy. Circulation 112:969–975, 2005

3. Yuyun MF, Adler AI, Wareham NJ: Whatis the evidence that microalbuminuria isa predictor of cardiovascular diseaseevents? Curr Opin Nephron Hypertens14:271–276, 2005

4. Ruilope LM, Segura J: Losartan and otherangiotensin II antagonists for nephropa-thy in type 2 diabetes mellitus: a review ofthe clinical trial evidence. Clin Ther 25:3044–3064, 2003

5. Brenner BM, Cooper ME, de Zeeuw D,Keane WF, Mitch WE, Parving HH, Re-muzzi G, Snapinn SM, Zhang Z, ShahinfarS; RENAAL Study Investigators: Effects oflosartan on renal and cardiovascular out-comes in patients with type 2 diabetes andnephropathy. N Engl J Med 345:861–869,2001

6. Parving HH, Lehnert H, Brochner-Mortensen J, Gomis R, Andersen S, ArnerP; Irbesartan in Patients with Type 2 Dia-betes and Microalbuminuria StudyGroup: The effect of irbesartan on the de-velopment of diabetic nephropathy in pa-tients with type 2 diabetes. N Engl J Med345:870–878, 2001

7. Zandbergen AA, Baggen MG, LambertsSW, Bootsma AH, de Zeeuw D, Ouwen-dijk RJ: Effect of losartan on microalbu-

minuria in normotensive patients withtype 2 diabetes mellitus: a randomizedclinical trial. Ann Intern Med 139:90–96,2003

8. Dahlof B, Devereux RB, Kjeldsen SE,Julius S, Beevers G, de Faire U, FyhrquistF, Ibsen H, Kristiansson K, Lederballe-Pedersen O, Lindholm LH, Nieminen MS,Omvik P, Oparil S, Wedel H; LIFE StudyGroup: Cardiovascular morbidity andmortality in the losartan intervention forendpoint reduction in hypertensive study(LIFE): a randomised trial against ateno-lol. Lancet 359:995–1003, 2002

9. Heart Outcomes Prevention EvaluationStudy Investigators: Effects of an angio-tensin-converting-enzyme inhibitor,ramipril, on cardiovascular events inhigh-risk patients. N Engl J Med 342:145–153, 2000

10. Ibsen H, Olsen MH, Wachtell K, Borch-Johnsen K, Lindholm LH, Mogensen CE,Dahlof B, Devereux RB, de Faire U, Fyhr-quist F, Julius S, Kjeldsen SE, Lederballe-Pedersen O, Nieminen MS, Omvik P,Oparil S, Wan Y: Reduction in albumin-uria translates to reduction in cardiovas-cular events in hypertensive patients:losartan intervention for endpoint reduc-tion in hypertension study. Hypertension45:198–202, 2005

11. De Zeeuw D, Remuzzi G, Parving HH,

Keane WF, Zhang Z, Shahinfar S, SnapinnS, Cooper ME, Mitch WE, Brenner BM:Albuminuria, a therapeutic target for car-diovascular protection in type 2 diabeticpatients with nephropathy. Circulation110:921–927, 2004

12. Yuyun MF, Dinneen SF, Edwards OM,Wood E, Wareham NJ: Absolute level andrate of change of albuminuria over 1 yearindependently predict mortality and car-diovascular events in patients with dia-betic nephropathy. Diabet Med 20:277–282, 2003

13. De Zeeuw D: Albuminuria, not only a car-diovascular/renal risk marker, but also atarget for treatment? Kidney Int 68:1899–1901, 2005

14. Burnier M, Zanchi A: Blockade of the re-nin-angiotensin-aldosterone system: akey therapeutic strategy to reduce renaland cardiovascular events in patients withdiabetes. J Hypertens 24:11–25, 2006

15. Ibsen H, Olsen MH, Wachtell K, Borch-Johnsen K, Lindholm LH, Mogensen CE,Dahlof B, Snapinn SM, Wan Y, Lyle PA:Does albuminuria predict cardiovascu-lar outcomes on treatment with losartanversus atenolol in patients with diabe-tes, hypertension, and left ventricularhypertrophy? The LIFE study. DiabetesCare 29:595– 600, 2006

Zandbergen and Associates


Simvastatin Reduces PlasmaOsteoprotegerin in Type 2 DiabeticPatients With MicroalbuminuriaBIRGITTE NELLEMANN, RF

1

LARS C. GORMSEN, MD1

JENS DOLLERUP, MSC2

OLE SCHMITZ, MD3

CARL E. MOGENSEN, MD1

LARS M. RASMUSSEN, MD4,5

SØREN NIELSEN, MD1

O steoprotegerin (OPG), a secretedglycoprotein and member of the tu-mor necrosis factor (TNF) receptor

superfamily, is a soluble receptor activa-tor of nuclear factor-�B (RANK) ligand(RANKL) and TNF-related apoptosis-inducing ligand (1). OPG works as a de-coy receptor preventing RANK/RANKL-induced osteoclast differentiation andactivation (2). Moreover, the RANK/RANKL system has potential cardiovascu-lar effects; the system induces vascularcell adhesion molecule (VCAM)-1 synthe-sis, prolongs endothelial cell survival, andpromotes angiogenesis (3,4). Furthermore,OPG may be involved in cardiovasculardisease (CVD). An epidemiological studyidentified OPG as an independent riskfactor for CVD (5) and OPG is present inhigh concentrations in the arterial wall(6,7). Of note, diabetic patients are char-acterized by elevated OPG (3), which isassociated with subclinical atherosclero-sis in both type 1 (8) and type 2 (9) dia-betes. Conversely, OPG may inhibitcalcification in mice (10). Hence, it is pos-sible that vascular calcification increasesOPG, which then, in turn, is involved incalcification inhibition (4).

Development of atherosclerosis in-volves expression of adhesion molecules(e.g., VCAM-1 and intercellular adhesionmolecule [ICAM]), allowing cellular at-

tachment and migration of monocytesand macrophages into the vascular wall(11). Recent in vitro studies suggest thatstatins may suppress both OPG (12) andadhesion molecule (13) production.

Statin treatment reduces cardiovascu-lar disease in type 2 diabetes (14). More-over, additional so-called pleiotropiceffects have also been proposed (15).Since both OPG and adhesion moleculesare associated with CVD and potentiallymodifiable by statins, we examined theeffect of simvastatin on OPG and adhe-sion molecules in type 2 diabetic patientsat increased risk for CVD due to persistentmicroalbuminuria.

RESEARCH DESIGN ANDMETHODS — Informed consent wasobtained from all participants, and thestudy received ethics committee ap-proval. Eighteen type 2 diabetic patientswere randomly recruited from the outpa-tient clinic (16). Inclusion criteria weremicroalbuminuria (overnight urinary al-bumin excretion 15–200 �g/min),plasma cholesterol �5.5 mmol/l, plasmatriglyceride �4.5 mmol/l, A1C �10%,serum C-peptide �0.49 nmol/l, andblood pressure �160/95 mmHg.

The study design has previously beendescribed (16). In brief, in a randomized,double-blind design, patients were allo-

cated to treatment with 10 mg/day simva-statin or the placebo group for 18 weeks.If plasma cholesterol was �5.2 mmol/l at6 weeks, the dose was doubled. Bloodsamples for OPG, VCAM-1, and ICAMwere collected after an overnight fast atbaseline and week 18. Sampling for OPGwas insufficient for one patient in the pla-cebo group. Sample size was based onpreviously decided main outcome mea-surements of renal function (16). Since nointerventional studies describing changesin OPG could be identified from the liter-ature, we were unable to perform a validsample size calculation. We therefore in-cluded all patients from that study.

OPG was measured by a sandwich en-zyme-linked immunosorbent assay (R&DSystems, Minneapolis, MN) using amouse anti-human OPG as capture anti-body and a biotinylated goat anti-humanOPG for detection. Recombinant humanOPG was used for calibration. Sampleswere diluted and measured in duplicate(8). Serum ICAM and VCAM-1 were mea-sured by monoclonal antibody–based en-zyme-linked immunosorbent assays asdescribed by the manufacturer (catalognos. BBE1B, BBE3, and DY809, respec-tively; R&D Systems).

Data are presented as means � SEM.Between-group differences were analyzedusing Student’s t test or the Mann-Whitney two-sample test. Changes werealso evaluated as the 18 weeks–to–baseline ratio. Correlations were evalu-ated by Pearson’s r.

RESULTS — The treatment groupswere similar with respect to age, sex, dia-betes duration, BMI, A1C, and serum C-peptide. The average simvastatin dosewas 12.5 mg/day. Cholesterol was signif-icantly reduced by simvastatin. A1C re-mained unchanged.

OPG levels were comparable at base-line (1,660 � 161 vs. 1,961 � 131 pg/mlfor the placebo vs. simvastatin groups, re-spectively) and after 18 weeks (1,684 �154 vs. 1,816 � 95 pg/ml), and within-group changes were not statistically sig-nificant. However, simvastatin treatmentwas associated with a significant reduc-tion of baseline–to–18 weeks OPG ratio

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From 1Medical Department M, Aarhus University Hospital, Aarhus, Denmark; 2Merck Research Laborato-ries, Copenhagen, Denmark; 3Pharmacology, Aarhus University Hospital, Aarhus, Denmark; the 4Labora-tory for Biochemical Pathology, Aarhus University Hospital, Aarhus, Denmark; and the 5Department ofClinical Biochemistry, Odense University Hospital, Odense, Denmark.

Address correspondence and reprint requests to Søren Nielsen, MD, Aarhus University Hospital, Nørre-brogade 44, 8000 Aarhus C, Denmark. E-mail: [email protected].


0919. Clinical trial reg. no. NCT00471549, clinicaltrials.gov.Abbreviations: CVD, cardiovascular disease; ICAM, intercellular adhesion molecule; OPG, osteoprote-

gerin; RANK, receptor activator of nuclear factor-�B; RANKL, RANK ligand; TNF, tumor necrosis factor;VCAM, vascular cell adhesion molecule.






compared with that in the placebo group(Fig. 1).

No significant differences were ob-served at baseline in VCAM-1 (755 � 64vs. 690 � 51 ng/ml for the placebo vs.simvastatin groups, respectively) orICAM (307 � 83 vs. 336 � 33 ng/ml).Moreover, the 18 weeks–to–baseline ra-tios were comparable between groups.

There was no significant correlationbetween the change in cholesterol andOPG ratio in the simvastatin group or be-tween total cholesterol and OPG,VCAM-1, or ICAM at baseline in the com-bined group. OPG and A1C tended tocorrelate at baseline (r2 � 0.46, P � 0.06)and week 18 (r2 � 0.48, P � 0.07).Plasma insulin (picomoles per liter) cor-related significantly with OPG at week 18(r2 � 0.56, P � 0.05).

CONCLUSIONS — In this study,low-dose simvastatin treatment for 18weeks reduced OPG levels in type 2 dia-betic patients with microalbuminuria andmild hypercholesterolemia. To ourknowledge, this has not previously beendemonstrated in humans in vivo.

The cellular mechanism whereby stat-ins affect OPG is unclear. Statins modu-late inflammatory mediators on OPGsecretion. Thus, simvastatin reducesTNF-�–induced OPG levels in vitro (12)and directly inhibits the nuclear factor-�Bsystem (12,17). Moreover, OPG mRNAand RANKL mRNA were increased inmouse bone-cell cultures incubated with

simvastatin for 7–16 days (18). These re-sults seem contrary to ours, since wefound decreased OPG levels. However,the different study design (in vitro vs. invivo) and treatment duration (7 and 16days vs. 18 weeks) may partly explain thedifferences. Moreover, since OPG is in-volved in calcification inhibition, at leastin mice (10), our findings may reflect asimvastatin-mediated calcification reduc-tion, which, in turn, might downregulateOPG. In type 2 diabetes, microalbumin-uria is not known to be related to alteredbone mineral metabolism.

Contrary to previous reports (19), wefound no change in VCAM-1 or ICAM.These adhesion molecules are inducedby inflammatory cytokines (i.e., TNF-�and interleukin-1), oxidative stress, andoxidized LDL (20). In theory, the anti-inflammatory and LDL-lowering effectsof simvastatin should, therefore, exertan inhibitory effect on adhesion mole-cule expression (17,19). To our knowl-edge, this has never been shown inhuman in vivo studies. Mulder et al.(21) recently reported that switching tomore aggressive lipid lowering in patientsalready on statin treatment may not havefurther effects on adhesion molecule ex-pression. We included statin-naive pa-tients in the present study. Although theinclusion criteria do not adhere to today’sstandards of clinical care, we think thatthe interpretations of the results are nothampered in a major way. Our inability toshow changes in adhesion molecules may

be due to the relatively low statin dose,the study duration, or the number of pa-tients included.

In summary, 18 weeks of low-dosesimvastatin treatment reduced circulat-ing OPG levels in type 2 diabetic pa-tients with microalbuminuria but hadno effect on VCAM-1 or ICAM. The re-duction of OPG was independent ofcholesterol and suggests a pleiotropiceffect of simvastatin per se. The OPG-lowering effect of simvastatin may sig-nal diminished vascular calcification.

Acknowledgments— This study was sup-ported by Novo Nordisk Research Founda-tion, the Danish Medical Research Council,and Merck Sharp & Dohme.

We thank L. Larsen, A. Mengel, and M.Møller for technical assistance.

References1. Emery JG, McDonnell P, Burke MB, Deen

KC, Lyn S, Silverman C, Dul E, Appel-baum ER, Eichman C, DiPrinzio R, DoddsRA, James IE, Rosenberg M, Lee JC,Young PR: Osteoprotegerin is a receptorfor the cytotoxic ligand TRAIL. J Biol Chem273:14363–14367, 1998

2. Lacey DL, Timms E, Tan HL, Kelley MJ,Dunstan CR, Burgess T, Elliott R, Co-lombero A, Elliott G, Scully S, Hsu H, Sul-livan J, Hawkins N, Davy E, Capparelli C,Eli A, Qian YX, Kaufman S, Sarosi I, Shal-houb V, Senaldi G, Guo J, Delaney J, BoyleWJ: Osteoprotegerin ligand is a cytokinethat regulates osteoclast differentiationand activation. Cell 93:165–176, 1998

3. Secchiero P, Corallini F, Pandolfi A, Con-soli A, Candido R, Fabris B, Celeghini C,Capitani S, Zauli G: An increased osteo-protegerin serum release characterizes theearly onset of diabetes mellitus and maycontribute to endothelial cell dysfunction.Am J Pathol 169:2236–2244, 2006

4. Schoppet M, Sattler AM, Schaefer JR, Her-zum M, Maisch B, Hofbauer LC: Increasedosteoprotegerin serum levels in men withcoronary artery disease. J Clin EndocrinolMetab 88:1024–1028, 2003

5. Kiechl S, Schett G, Wenning G, Redlich K,Oberhollenzer M, Mayr A, Santer P,Smolen J, Poewe W, Willeit J: Osteopro-tegerin is a risk factor for progressive ath-erosclerosis and cardiovascular disease.Circulation 109:2175–2180, 2004

6. Dhore CR, Cleutjens JP, Lutgens E, Cleu-tjens KB, Geusens PP, Kitslaar PJ, TordoirJH, Spronk HM, Vermeer C, Daemen MJ:Differential expression of bone matrixregulatory proteins in human atheroscle-rotic plaques. Arterioscler Thromb VascBiol 21:1998–2003, 2001

7. Olesen P, Ledet T, Rasmussen LM: Arte-rial osteoprotegerin: increased amounts

Figure 1—Baseline–to–18 weeks OPG ratio: mean � SEM 1.021 � 0.035 vs. 0.932 � 0.024 forthe placebo (F) vs. simvastatin (�) groups; P � 0.05. A relative reduction of 7% was observed inthe simvastatin group (P � 0.05). n � 17 (9 in the placebo group and 8 receiving simvastatintreatment). Medians are represented by solid lines.

Nellemann and Associates


in diabetes and modifiable synthesis fromvascular smooth muscle cells by insulinand TNF-alpha. Diabetologia 48:561–568, 2005

8. Rasmussen LM, Tarnow L, Hansen TK,Parving HH, Flyvbjerg A: Plasma osteo-protegerin levels are associated with gly-caemic status, systolic blood pressure,kidney function and cardiovascular mor-bidity in type 1 diabetic patients. Eur JEndocrinol 154:75–81, 2006

9. Anand DV, Lahiri A, Lim E, Hopkins D,Corder R: The relationship betweenplasma osteoprotegerin levels and coro-nary artery calcification in uncomplicatedtype 2 diabetic subjects. J Am Coll Cardiol47:1850–1857, 2006

10. Bucay N, Sarosi I, Dunstan CR, Morony S,Tarpley J, Capparelli C, Scully S, Tan HL,Xu W, Lacey DL, Boyle WJ, Simonet WS:Osteoprotegerin-deficient mice developearly onset osteoporosis and arterial calci-fication. Genes Dev 12:1260–1268, 1998

11. Ross R: Atherosclerosis is an inflamma-tory disease. Am Heart J 138:S419–S420,1999

12. Ben Tal CE, Hohensinner PJ, Kaun C,

Maurer G, Huber K, Wojta J: Statins de-crease TNF-alpha-induced osteoprote-gerin production by endothelial cells andsmooth muscle cells in vitro. BiochemPharmacol 73:77–83, 2007

13. Rasmussen LM, Hansen PR, NabipourMT, Olesen P, Kristiansen MT, Ledet T:Diverse effects of inhibition of 3-hydroxy-3-methylglutaryl-CoA reductase on theexpression of VCAM-1 and E-selectin inendothelial cells. Biochem J 360:363–370,2001

14. Kempler P: Learning from large cardio-vascular clinical trials: classical cardiovas-cular risk factors. Diabetes Res Clin Pract68 (Suppl. 1):S43–S47, 2005

15. Skaletz-Rorowski A, Walsh K: Statin ther-apy and angiogenesis. Curr Opin Lipidol14:599–603, 2003

16. Nielsen S, Schmitz O, Moller N, PorksenN, Klausen IC, Alberti KG, Mogensen CE:Renal function and insulin sensitivityduring simvastatin treatment in type 2(non-insulin-dependent) diabetic pa-tients with microalbuminuria. Diabeto-logia 36:1079 –1086, 1993

17. Devaraj S, Chan E, Jialal I: Direct demon-

stration of an antiinflammatory effect ofsimvastatin in subjects with the metabolicsyndrome. J Clin Endocrinol Metab 91:4489–4496, 2006

18. Kaji H, Kanatani M, Sugimoto T, ChiharaK: Statins modulate the levels of osteopro-tegerin/receptor activator of NFkappaB li-gand mRNA in mouse bone-cell cultures.Horm Metab Res 37:589–592, 2005

19. Zapolska-Downar D, Siennicka A, Kacz-marczyk M, Kolodziej B, Naruszewicz M:Simvastatin modulates TNFalpha-in-duced adhesion molecules expression inhuman endothelial cells. Life Sci 75:1287–1302, 2004

20. Springer TA: Traffic signals for lympho-cyte recirculation and leukocyte emigra-tion: the multistep paradigm. Cell 76:301–314, 1994

21. Mulder DJ, van Haelst PL, Wobbes MH,Gans RO, Zijlstra F, May JF, Smit AJ, Ter-vaert JW, van Doormaal JJ: The effect ofaggressive versus conventional lipid-low-ering therapy on markers of inflammatoryand oxidative stress. Cardiovasc DrugsTher 21:91–97, 2007

Simvastatin reduces OPG in type 2 diabetes


Lifestyle Intervention and AdipokineLevels in Subjects at High Risk for Type 2DiabetesThe Study on Lifestyle intervention and Impaired glucose toleranceMaastricht (SLIM)

EVA CORPELEIJN, PHD1

EDITH J.M. FESKENS, PHD2,3

EUGENE H.J.M. JANSEN, PHD4

MARCO MENSINK, MD, PHD1

WIM H.M. SARIS, MD, PHD1

ELLEN E. BLAAK, PHD1

OBJECTIVE — We investigated whether circulating adipokine concentrations can be alteredby lifestyle intervention according to general recommendations in subjects at risk for diabetes aswell as the potential of leptin, adiponectin, and resistin as biomarkers for lifestyle-inducedimprovements in glucose metabolism and insulin resistance.

RESEARCH DESIGN AND METHODS — In the Study on Lifestyle intervention andImpaired glucose tolerance Maastricht, 147 men and women with impaired glucose tolerance(IGT) were randomized to either a combined diet-and-exercise intervention or a control pro-gram. At baseline and after 1 year, an oral glucose tolerance test, an exercise test, and anthro-pometric measurements were performed. After 1 year, complete data of 103 subjects (50intervention and 53 control subjects) were obtained.

RESULTS — Lifestyle intervention reduced plasma leptin concentrations (�14.2%) in IGTsubjects but did not alter plasma adiponectin (�0.3%) or resistin (�6.5%) concentrationsdespite marked improvements in glucose tolerance and insulin resistance.

CONCLUSIONS — Changes in leptin concentration were related to improvements in insu-lin sensitivity independent of changes in body composition.

Diabetes Care 30:3125–3127, 2007

Adipokines produced by adipose tis-sue, such as adiponectin, resistin,and leptin, may link obesity to in-

sulin resistance, impaired glucose metab-olism, and type 2 diabetes (1). Cross-sectional evidence for an associationbetween insulin resistance and inflamma-tion profile is ample (2). Nevertheless, thereported effects of lifestyle interventionon adipokines are limited and inconclu-

sive (3,4). The first aim of the presentstudy was to investigate whether circulat-ing adipokine concentrations can be al-tered by lifestyle intervention accordingto general recommendations in subjects atrisk for diabetes. Second, we investigatedthe potential of leptin, adiponectin, andresistin as biomarkers for lifestyle-induced improvements in glucose metabo-lism and insulin resistance. We addressed

these aims in the Study on Lifestyle inter-vention and Impaired glucose toleranceMaastricht (SLIM).

RESEARCH DESIGN ANDMETHODS — Study design, inclusionand exclusion criteria, and the diet andexercise program of SLIM, an ongoing,randomized, controlled trial, have previ-ously been described in detail (5,6). Eachyear, anthropometry, body fat percentage(7), physical fitness (VO2max), and plasmametabolites during fasting and 2 h after a75-g oral glucose load are determined.The study protocol was approved by thelocal medical ethical committee of theMaastricht University. All subjects gavewritten informed consent.

Fasting plasma adiponectin (fulllength and globular, coefficient of varia-tion [CV] 6.2%), resistin (homodimeric,CV 4.0%), and leptin (CV 5.8%) weresimultaneously analyzed with enzyme-linked immunosorbent assays (Bioven-dor, Heidelberg, Germany). Of the 147subjects enrolled, 131 completed the firstyear. For regression analysis, 28 subjectswere excluded because of missing valuesfor dietary intake (n � 3), VO2max (n �23), or both (n � 2). No differences wereobserved between the included and ex-cluded subjects.

Repeated-measures ANOVA wereused for differences between groups overtime. A two-tailed P value �0.05 was con-sidered statistically significant. Data arepresented as means � SEM or, if not nor-mally distributed (insulin, leptin, andadiponectin), as median (25th–75thpercentile).

RESULTS — Lifestyle interventionwas effective to improve glucose toleranceand insulin sensitivity, as shown by animproved 2-h glucose concentration(control group 0.36 � 0.3 mmol/l andintervention group �0.78 � 0.2 mmol/l;P � 0.001) and a reduced 2-h insulin re-sponse (1-year change 9.9 mU/l [�22.8to 33.1] for the control group and �15.8

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From 1Human Biology, Nutrition and Toxicology Research Institute, Maastricht, the Netherlands; 2HumanNutrition, Wageningen University and Research Center, Wageningen, the Netherlands; the 3Centre forNutrition and Health, National Institute for Public Health and Environment, Bilthoven, the Netherlands; andthe 4Laboratory for Toxicology, Pathology and Genetics, National Institute for Public Health and Environ-ment, Bilthoven, the Netherlands.

Address correspondence and reprint requests to Eva Corpeleijn, PHD, Human Biology, Maastricht Uni-versity, P.O. Box 616, 6200 MD Maastricht, Netherlands. E-mail: [email protected].

Received for publication 7 March 2007 and accepted in revised form 14 September 2007.Published ahead of print at http://care.diabetesjournals.org on 21 September 2007. DOI: 10.2337/dc07-

0457.Abbreviations: HOMA-IR, homeostasis model assessment of insulin resistance; SLIM, Study on Lifestyle

intervention and Impaired glucose tolerance Maastricht.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





mU/l [�33.2 to 10.0] for the interventiongroup; P � 0.01). Adherence to the inter-vention was demonstrated by reducedbody weight (�3%; P � 0.003) and BMI(�3%; P � 0.001) and increased physicalfitness (VO2max � 5.6%; P � 0.001) after1 year, as previously described (8). Leptinconcentrations were reduced by �14.2%in the intervention group (baseline 16.7ng/ml [6.4 –27.2] and 1-year change�2.37 ng/ml [�5.5 to �0.3]) comparedwith the 0.5% increase in control subjects(baseline 11.6 ng/ml [5.9 –24.8] and1-year change 0.06 ng/ml [�2.2 to 1.5];P � 0.01). No effects were observed onadiponectin for the intervention group(baseline 12.6 �g/ml [8.6 –19.6] and1-year change �0.06 �g/ml [�1.4 to1.1]; �0.3%) or control group (baseline14.1 �g/ml [9.3–19.3] and 1-year change�0.03 �g/ml [�1.0 to 1.2]; �0.2%).Also, in subjects with the highest bodyweight loss (�5%, n � 18), adiponectinconcentrations did not change (�adi-ponectin �0.1 �g/ml [�0.7 to 3.8]; P �0.43). Observations were similar for resis-tin in intervention (baseline 3.69 � 0.21�g/ml and 1-year change �0.24 � 0.11�g/ml; �6.5%) and control (baseline3.67 � 0.21 �g/ml and 1-year change�0.06 � �0.07 �g/ml; �1.6%) subjects.

In the intervention group, leptinchanges were positively associated withchanges in body weight, BMI, body fatpercentage, waist circumference, fastingglucose, 2-h glucose, fasting insulin, ho-meostasis model assessment of insulin re-sistance (HOMA-IR), total cholesterol,and plasma triglycerides (all Pearson cor-relation coefficients �0.48; all P values�0.001). Regression analyses in the inter-vention group revealed that the associa-tions with fasting insulin, HOMA-IR, and2-h glucose were less strong after adjust-ment for age, sex, lifestyle factors, andbody composition but remained highly

significant (Table 1). After full adjustment(Table 1, model 3), a decrease inHOMA-IR of 10% corresponds with a de-crease in leptin concentrations of 3.9%.Changes in adiponectin or resistin werenot significantly associated with changesin parameters under investigation, al-though after full adjustment in the regres-sion analysis (model 3), a 1-unit (1%)decrease in A1C was related to a 15.6%increase in adiponectin concentration.

CONCLUSIONS — The lifestyle in-tervention–induced decrease in leptinwas strongly associated with a decrease ininsulin resistance, and this associationwith insulin resistance was only partiallyexplained by a reduction in body fat per-centage. This is consistent with evidencethat exercise may decrease circulating lep-tin concentration, independent of bodycomposition (9). The decrease in circulat-ing leptin may be explained by increasedleptin sensitivity (10) with an effect onleptin production and clearance by feed-back mechanisms. Leptin may also haveperipheral effects on insulin signaling(12). Since most body weight was lost inthe first 3 months and since weight loss inthe last one-half year was only minor(�0.68 kg), it does not seem likely thatleptin concentrations were reduced as aresult of a catabolic state.

In the present study, adiponectinconcentrations were not clearly altered bylifestyle changes. This was unexpected,since (extreme) weight reduction in obeseindividuals (10–57 kg) was convinc-ingly associated with an increase inplasma adiponectin concentrations rang-ing from 2.1 to 9.2 �g/ml (13–16). Ourfindings are supported, however, by otherstudies using lifestyle intervention pro-grams according to general guidelinesthat failed to show an effect on adiponec-tin in diabetic (17) and obese (18) sub-

jects. Although the molecular form ofadiponectin may be of importance, pe-ripheral insulin resistance has not beenassociated with a specific form of adi-ponectin thus far (4,18–20). The effect ofweight loss on adiponectin concentra-tions seems to depend on the amount ofweight loss and on the way in whichweight loss was achieved.

This study shows that lifestyle inter-vention reduces plasma leptin concentra-tions in subjects with IGT but does notseem to alter plasma adiponectin or resis-tin concentrations despite marked im-provements in glucose tolerance andinsulin resistance. Leptin can be a biomar-ker for improvements in insulin sensitiv-ity and glucose tolerance after lifestyleintervention, independent of changes inbody composition.

Acknowledgments— This study was sup-ported by grants from the Dutch Diabetes Re-search Foundation (DFN 98.901 and DFN2000.00.020) and the Netherlands Organiza-tion for Scientific Research (ZonMW 940-35-034, 2,200.0139).

We thank Jos Stegen, Hans Cremers, TanjaHermans-Limpens, Ilse Nijs, and Marja Ock-eloen.

References1. Arner P: Insulin resistance in type 2 dia-

betes: role of the adipokines. Curr MolMed 5:333–339, 2005

2. Duncan BB, Schmidt MI: The epidemiol-ogy of low-grade chronic systemic inflam-mation and type 2 diabetes. DiabetesTechnol Ther 8:7–17, 2006

3. Monzillo LU, Hamdy O, Horton ES, Led-bury S, Mullooly C, Jarema C, Porter S,Ovalle K, Moussa A, Mantzoros CS: Effectof lifestyle modification on adipokine lev-els in obese subjects with insulin resis-tance. Obes Res 11:1048–1054, 2003

4. Bluher M, Brennan AM, Kelesidis T,

Table 1—�-Coefficients of changes in adiponectin and leptin concentrations after 1 year related to changes in metabolic parameters adjustedfor age, sex, changes in lifestyle factors, and body composition in intervention subjects

�Leptin ln (ng/ml) �Adiponectin ln (�g/ml)

Model 1 Model 2 Model 3 Model 1 Model 2 Model 3

�A1C (%) 0.13 0.12 �0.01 �0.11 �0.14 �0.17*�Fasting insulin ln (mU/l) 0.60† 0.63† 0.41* 0.11 0.07 0.08�HOMA-IR (ln) 0.53† 0.56† 0.40* 0.06 0.03 0.02�2-h glucose (mmol/l) 0.09† 0.09‡ 0.07‡ 0.01 0.01 0.00

Data are unstandardized -coefficients. Multiple regression analysis was performed to identify the contribution of changes in adipokines to changes in metabolicparameters independent of other factors, with the adipokine as the dependent variable and, as independent variables, the means of the dependent (�leptin �ln� or�adiponectin �ln�), the main independent variable (�A1C, �fasting insulin, �HOMA-IR, or �2-h glucose), age (years), and sex in model 1, plus lifestyle factors(�total fat intake in percentage of energy, �Vo2max per kg fat-free mass in ml O2 � min�1 � kg fat-free mass�1) in model 2, plus body composition (�waistcircumference �cm� and �body fat percentage estimated with skinfolds) in model 3. n � 49. *P � 0.05; †P � 0.001; ‡P � 0.01.

Lifestyle intervention and adipokine levels


Kratzsch J, Fasshauer M, Kralisch S, Wil-liams CJ, Mantzoros CS: Total and high–molecular weight adiponectin in relationto metabolic variables at baseline and inresponse to an exercise treatment pro-gram: comparative evaluation of three as-says. Diabetes Care 30:280–285, 2007

5. Mensink M, Blaak EE, Corpeleijn E, SarisWH, de Bruin TW, Feskens EJ: Lifestyleintervention according to general recom-mendations improves glucose tolerance.Obes Res 11:1588–1596, 2003

6. Mensink M, Corpeleijn E, Feskens EJ,Kruijshoop M, Saris WH, de Bruin TW,Blaak EE: Study on lifestyle-interventionand impaired glucose tolerance Maas-tricht (SLIM): design and screening re-sults. Diabetes Res Clin Pract 61:49–58,2003

7. Durnin JV, Womersley J: Body fat as-sessed from total body density and itsestimation from skinfold thickness: mea-surements on 481 men and women agedfrom 16 to 72 years. Br J Nutr 32:77–97,1974

8. Corpeleijn E, Feskens EJ, Jansen EH,Mensink M, Saris WH, de Bruin TW,Blaak EE: Improvements in glucose toler-ance and insulin sensitivity after lifestyleintervention are related to changes in se-rum fatty acid profile and desaturase ac-tivities: the SLIM study. Diabetologia 49:2392–2401, 2006

9. Pasman WJ, Westerterp-Plantenga MS,Saris WH: The effect of exercise trainingon leptin levels in obese males. Am JPhysiol 274:E280–E286, 1998

10. Montez JM, Soukas A, Asilmaz E, Fayzik-hodjaeva G, Fantuzzi G, Friedman JM:Acute leptin deficiency, leptin resistance,and the physiologic response to leptinwithdrawal. Proc Natl Acad Sci U S A 102:2537–2542, 2005

11. Steinberg GR, Parolin ML, HeigenhauserGJ, Dyck DJ: Leptin increases FA oxida-tion in lean but not obese human skeletalmuscle: evidence of peripheral leptin re-sistance. Am J Physiol Endocrinol Metab283:E187–E192, 2002

12. Fruhbeck G: Intracellular signalling path-ways activated by leptin. Biochem J 393:7–20, 2006

13. Lazzer S, Vermorel M, Montaurier C,Meyer M, Boirie Y: Changes in adipocytehormones and lipid oxidation associatedwith weight loss and regain in severelyobese adolescents. Int J Obes (Lond) 29:1184–1191, 2005

14. Yang WS, Lee WJ, Funahashi T, Tanaka S,Matsuzawa Y, Chao CL, Chen CL, Tai TY,Chuang LM: Weight reduction increasesplasma levels of an adipose-derived anti-inflammatory protein, adiponectin. J ClinEndocrinol Metab 86:3815–3819, 2001

15. Hulver MW, Zheng D, Tanner CJ,Houmard JA, Kraus WE, Slentz CA, SinhaMK, Pories WJ, MacDonald KG, DohmGL: Adiponectin is not altered with exer-cise training despite enhanced insulin ac-tion. Am J Physiol Endocrinol Metab 283:E861–E865, 2002

16. Esposito K, Pontillo A, Di Palo C, Giugli-ano G, Masella M, Marfella R, GiuglianoD: Effect of weight loss and lifestylechanges on vascular inflammatory mark-ers in obese women: a randomized trial.JAMA 289:1799–1804, 2003

17. Aas AM, Seljeflot I, Torjesen PA, Diep LM,Thorsby PM, Birkeland KI: Blood glucoselowering by means of lifestyle interven-tion has different effects on adipokines ascompared with insulin treatment in sub-jects with type 2 diabetes. Diabetologia 49:872–880, 2006

18. Bobbert T, Rochlitz H, Wegewitz U, Ak-pulat S, Mai K, Weickert MO, Mohlig M,

Pfeiffer AF, Spranger J: Changes of adi-ponectin oligomer composition by mod-erate weight reduction. Diabetes 54:2712–2719, 2005

19. Abbasi F, Chang SA, Chu JW, Ciaraldi TP,Lamendola C, McLaughlin T, Reaven GM,Reaven PD: Improvements in insulin re-sistance with weight loss, in contrast torosiglitazone, are not associated withchanges in plasma adiponectin or adi-ponectin multimeric complexes. Am JPhysiol Regul Integr Comp Physiol 290:R139–R144, 2006

20. Waki H, Yamauchi T, Kamon J, Ito Y,Uchida S, Kita S, Hara K, Hada Y, VasseurF, Froguel P, Kimura S, Nagai R, Kad-owaki T: Impaired multimerization of hu-man adiponectin mutants associated withdiabetes: molecular structure and multi-mer formation of adiponectin. J Biol Chem278:40352–40363, 2003

21. Page ST, Herbst KL, Amory JK, CovielloAD, Anawalt BD, Matsumoto AM, Brem-ner WJ: Testosterone administration sup-presses adiponectin levels in men. JAndrol 26:85–92, 2005

22. Chevillotte E, Giralt M, Miroux B, Ric-quier D, Villarroya F: Uncouplingprotein-2 controls adiponectin gene ex-pression in adipose tissue through themodulation of reactive oxygen speciesproduction. Diabetes 56:1042–1050,2007

23. Rubin D, Helwig U, Nothnagel M, LemkeN, Schreiber S, Folsch UR, Doring F,Schrezenmeir J: Postprandial plasma adi-ponectin decreases after glucose and highfat meal and is independently associatedwith postprandial triacylglycerols but notwith - 11388 promoter polymorphism.Br J Nutr. 30 July 2007 [Epub ahead ofprint]

Corpeleijn and Associates


Impaired Postprandial Blood Flow inAdipose Tissue May Be an Early Marker ofInsulin Resistance in Type 2 DiabetesGEORGE DIMITRIADIS, MD, DPHIL

1

VAIA LAMBADIARI, MD1

PANAYOTA MITROU, MD1

EIRINI MARATOU, PHD2

ELENI BOUTATI, MD1

DEMOSTHENES B. PANAGIOTAKOS, PHD3

THEOFANIS ECONOMOPOULOS, MD1

SOTIRIOS A. RAPTIS, MD1,2

OBJECTIVE — We investigated the changes in subcutaneous adipose tissue blood flow(ATBF) after a meal in the various stages of type 2 diabetes.

RESEARCH DESIGN AND METHODS — Five groups were examined: healthy controlsubjects, first-degree relatives of subjects with type 2 diabetes, subjects with impaired glucosetolerance (IGT), subjects with type 2 diabetes and postprandial hyperglycemia but normalfasting plasma glucose levels (diabetes group A [DMA]), and subjects with type 2 diabetes withboth postprandial and fasting hyperglycemia (diabetes group B [DMB]). ATBF was measuredwith 133Xe.

RESULTS — ATBF was higher in control subjects (1,507 � 103 ml/100 cm3 tissue � min)versus relatives and IGT, DMA, and DMB subjects (845 � 123, 679 � 69, 765 � 60, and 757 �69 ml/100 cm3 tissue � min, respectively; P � 0.001). Insulin sensitivity index (ISI) in controlsubjects (82 � 3 mg � l2/mmol � mU � min) was higher versus that for relatives and IGT, DMA,and DMB subjects (60 � 3, 45 � 1, 40 � 6, and 29 � 4 mg � l2/mmol � mU � min,respectively; P � 0.0001). ISI was positively associated with peak-baseline ATBF (� coefficient0.029 � 0.013, P � 0.03).

CONCLUSIONS — After meal ingestion, insulin-stimulated ATBF was decreased in relativesand and IGT, DMA, and DMB subjects. This defect could be an early marker of insulin resistancethat precedes the development of type 2 diabetes.

Diabetes Care 30:3128–3130, 2007

B lood flow plays an important role inthe metabolic function of adiposetissue and normally increases after

meal ingestion (1). In insulin-resistantstates such as obesity or type 2 diabetes, thisresponse is blunted (2–4). Whether this de-fect, which may be another facet of the in-sulin resistance syndrome (5), occurs early

in the development of type 2 is unknown.Our study was undertaken to examine

adipose tissue blood flow (ATBF) at allstages of type 2 diabetes. In addition,changes in plasma levels of adiponectin andapelin were also examined, since these adi-pokines correlate positively with endotheli-um-dependent vasodilatation (6–8).

RESEARCH DESIGN ANDMETHODS — A meal (730 kcal, 50%carbohydrate, 38% starch, 40% fat, and10% protein, consisting of bread, cheese,tomato, cucumber, olive oil, orange juice,and apple) was given to five groups: 1)healthy control subjects (aged 40 � 3years, with BMI 24 � 1 kg/m2; n � 10), 2)relatives of subjects with type 2 diabetes(two first-degree relatives [parents andsiblings] aged 41 � 3 years, with BMI25 � 1 kg/m2; n � 11), 3) subjects withimpaired glucose tolerance (IGT) (aged43 � 3 years, with BMI 26 � 1 kg/m2; n �6), 4) subjects with type 2 diabetes andpostprandial hyperglycemia but normalfasting plasma glucose (diabetes group A[DMA]) (aged 53 � 4 years, with BMI25 � 1 kg/m2), and 5) subjects with type2 diabetes and both fasting and postpran-dial hyperglycemia (diabetes group B[DMB]) (aged 56 � 2 years, with BMI26 � 1 kg/m2; n � 13).

Blood samples were withdrawn fromradial artery for measurements of insulin(Linco Research, St. Charles, MO), glu-cose (Yellow Springs Instruments, YellowSprings, OH), triglycerides, and non-esterified fatty acids (NEFAs) (Roche Diag-nostics, Penzberg, Germany), adiponectin(DRG Diagnostics, Marbourg, Germany),and apelin (Phoenix Pharmaceuticals,Phoeniz, AZ).

ATBF was measured immediately be-fore each blood sample (9,10). Insulinsensitivity in fasting state was measuredby homeostasis model assessement (11)and in postprandial state by Gutt index(insulin sensitivity index [ISI] [12]). Thestudy was approved by a hospital ethicscommittee, and subjects gave informedconsent.

Statistical analysisComparisons between groups were per-formed with repeated-measures ANOVA.Multiple linear regression analysis evalu-ated the association between ISI, triglyc-erides, and NEFAs with peak-baselineATBF after correcting for potential con-founders.

RESULTS — At 120 min, plasma glu-cose and insulin in control subjects were

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 12nd Department of Internal Medicine—Propaedeutic and Research Institute, Athens UniversityMedical School, “Attikon” University Hospital, Athens, Greece; the 2Hellenic National Center for Research,Prevention and Treatment of Diabetes Mellitus and Its Complications (HNDC), Athens, Greece; and 3Nu-trition Science-Dietetics, Harokopio University, Athens, Greece.

Address correspondence and reprint requests to George Dimitriadis, MD, Internal Medicine, AthensUniversity, “Attikon” University Hospital, 1 Rimini St., GR-12462 Haidari, Greece. E-mail:[email protected] and [email protected].

G.D. and V.L. contributed equally to the work presented in this article.Received for publication 10 April 2007 and accepted in revised form 14 September 2007.Published ahead of print at http://care.diabetesjournals.org on 21 September 2007. DOI: 10.2337/dc07-

0699.Abbreviations: ATBF, adipose tissue blood flow; IGT, impaired glucose tolerance; ISI, insulin sensitivity

index; NEFA, nonesterified fatty acids.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion





lower than in relatives and IGT, DMA,and DMB subjects (P � 0.05) (Fig. 1).ATBF after meal ingestion was suppressedin relatives and subjects with IGT, DMA,and DMB versus control subjects (Poverall� 0.001) (Fig. 1).

Fasting triglycerides were lower incontrol subjects (463 � 52 �mol/l) ver-sus relatives and IGT, DMA, and DMBsubjects (671 � 52, 821 � 162, 888 �94, and 928 � 133 �mol/l, respectively;Poverall � 0.014). Postprandial triglycer-

ides were lower in control subjects(264 � 32 mmol/l � 360 min) versusrelatives and IGT, DMA, and DMB sub-jects (336 � 39, 505 � 84, 470 � 90, and498 � 84 mmol/l � 360 min, respec-tively; Poverall � 0.04). Fasting and post-prandial triglycerides were negativelyassociated with peak-baseline ATBF (�coefficient �6.2 � 2.8, P � 0.03 and�2.7 � 1.8, P � 0.09, respectively).

Preprandial NEFAs were similar incontrol subjects (461 � 53 �mol/l), rela-

tives, and IGT, DMA, and DMB subjects(434 � 44, 453 � 129, 456 � 50, and714 � 145 �mol/l, respectively; Poverall �0.218). Postprandial NEFAs (areas undercurve0 –360 min of the postprandial de-creases) were lower in control subjects(98 � 11 mmol/l � 360 min) versus rel-atives and IGT, DMA, and DMB subjects(128 � 22, 156 � 19, 140 � 31, and182 � 16 mmol/l � 360 min, respec-tively; Poverall � 0.02). Fasting NEFAswere not associated with peak-baselineATBF (P � 0.256); postprandial NEFAswere negatively associated with peak-baseline ATBF (� coefficient �9.6 �10

�6

� 0.01; P � 0.001).Homeostasis model assessment in

control subjects (0.9 � 0.1) was lowerversus that in relatives and IGT, DMA,and DMB subjects (1.43 � 0.1, 1.7 � 0.1,1.8 � 0.2, and 2.2 � 0.2, respectively;Poverall � 0.003). ISI in control subjects(82 � 3 mg � l2/mmol � mU � min) washigher versus that in relatives and IGT,DMA, and DMB subjects (60 � 3, 45 � 1,40 � 6, and 29 � 4 mg � l2/mmol �mU � min, respectively; Poverall �0.0001). ISI was positively associatedwith peak-baseline ATBF (� coefficient0.029 � 0.013, P � 0.03).

Adiponectin was higher in controlsubjects (21 � 3 ng/ml) and relatives(23 � 3 ng/ml) versus IGT, DMA, andDMB subjects (11 � 2, 13 � 4, and 12 �3, respectively; Poverall � 0.007).

Apelin was similar in control subjects(1.13 � 0.21 ng/ml), relatives, and IGT,DMA, and DMB subjects (1.02 � 0.2,1.51 � 0.3, 1.5 � 0.4, and 1.3 � 0.3ng/ml, respectively).

CONCLUSIONS — ATBF is bluntedafter meal ingestion at all stages of type 2diabetes. Since insulin is a mediator of thepostprandial increases in ATBF (13),these results suggest that suppressedATBF may be a marker of insulin resis-tance. Indeed, insulin sensitivity in oursubjects was positively associated withthe increases in ATBF after the meal.However, it should be pointed out thatthis is a cross-sectional analysis; althoughfindings of impaired ATBF in people athigh risk for diabetes imply that this ab-normality might precede the develop-ment of clinical diabetes, the analysis doesnot actually prove this, and the suggestionremains speculative.

Our results confirm previous findingsin obese (2,3) or lean (4) subjects withovert type 2 diabetes in whom ATBF rateswere decreased after the consumption of a

Figure 1— Plasma glucose, plasma insulin, and adipose tissue blood flow in healthy subjects(control), first-degree relatives of subjects with type 2 diabetes (relatives), and IGT, DMA, andDMB subjects. P values represent overall comparison (repeated-measures ANOVA) between con-trol and patient groups. At t � 0, a mixed meal was given.

Dimitriadis and Associates


mixed meal or glucose. Jansson et al. (2)showed that ATBF is lower in insulin-resistant subjects with obesity and/or type2 diabetes and that this correlates nega-tively with the blood pressure.

Our results do not agree with a report(14) in first-degree relatives of diabeticsubjects in whom ATBF was measuredduring a hyperinsulinemic-euglycemicclamp: in the presence of insulin, theserates were decreased by 46% comparedwith those in healthy control subjects, butthe differences were not significant. Thedifferences with our study can be ex-plained by the findings of Karpe et al.(13): 1) the increases in ATBF after oraladministration of glucose were signifi-cantly greater than those after intravenousinfusion of insulin, and 2) locally infusedinsulin at the abdominal subcutaneousadipose tissue had no demonstrable ef-fects on blood flow, suggesting that insu-lin does not have a direct effect on ATBFbut, rather, is a mediator acting via sym-pathetic activation.

The physiological significance of thenutrient-related decreases in ATBF in thepatient groups of our study is unclear.However, changes in postprandial plasmatriglyceride and NEFA responses werenegatively associated with ATBF. Theseresults agree with the study of Samra et al.(15), in which triglyceride clearance byadipose tissue was closely related to ATBFwhen increased by epinephrine infusion.Moreover, Karpe at al (5) showed that, inhealthy subjects, the postmeal ATBF re-sponse is related to insulin sensitivity; ofall of the indexes of insulin sensitivityused, the estimation based on NEFA sup-pression to insulin was most strongly re-lated to the ATBF response, suggestingthat ATBF may be a major determinant ofthe insulin-related changes in plasmaNEFAs after the meal.

Plasma adiponectin was decreased inthe subjects with IGT and type 2 diabetes.

However, it is unlikely that adiponectinmay mediate the changes seen in ATBF,since in the relatives, plasma adiponectinlevels were normal but ATBF decreased.

In conclusion, we have shown thatATBF is suppressed after meal ingestion atall stages of type 2 diabetes. These find-ings may provide a marker of insulin re-sistance that occurs early in thedevelopment of type 2 diabetes.

References1. Frayn KN: Adipose tissue as a buffer for

daily lipid flux. Diabetologia 45:1201–1210, 2002

2. Jansson PAE, Larsson A, Lonnroth PN:Relationship between blood pressure,metabolic variables, and blood flow inobese subjects with or without non-insu-lin dependent diabetes mellitus. Eur J ClinInvest 28:813–818, 1998

3. Coppack S, Fisher R, Humphreys S, ClarkM, Pointon J, Frayn K: Carbohydrate me-tabolism in insulin resistance: glucose up-take and lactate production by adiposetissue and forearm tissues in vivo beforeand after a mixed meal. Clin Sci 90:409–15, 1996

4. Dimitriadis G, Boutati E, Lambadiari V,Mitrou P, Maratou E, Brunel P, Raptis SA:Restoration of early insulin secretion aftera meal in type 2 diabetes: effects on lipidand glucose metabolism. Eur J Clin Invest34:490–497, 2004

5. Karpe F, Fielding BA, Ilic V, MacdonaldIA, Summers LKM, Frayn KN: Impairedpostprandial adipose tissue blood flow re-sponse is related to aspects of insulin sen-sitivity. Diabetes 51:2467–2473, 2002

6. Shimabukuro M, Higa N, Asahi T, OshiroY, Takasu N, Tagawa T, Ueda S, Shimo-mura I, Funahashi T, Matsuzawa Y: Hy-poadiponectinemia is closely linked toendothelial dysfunction in man. J Clin En-docrinol Metab 88:3236–3240, 2003

7. Fernandez-Real JM, Castro A, Vazquez G,Casamitjana R, Lopez-Bermejo A, Penar-roja G, Ricart W: Adiponectin is associ-ated with vascular function independentof insulin sensitivity. Diabetes Care

27:739–745, 20048. Foldes G, Horkay F, Szokodi I, Vuol-

teenaho O, Ilves M, Lindstedt K, Mayran-paa M, Sarman B, Seres L, Skoumal R,Lako-Futo Z, deChatel R, Ruskoaho H,Toth M: Circulating and cardiac levels ofapelin, the novel ligand of the orphan re-ceptor APJ, in patients with heart failure.Biochem Biophys Res Commun 308:480–485, 2003

9. Coppack S, Fisher R, Gibbons G, Frayn K:Postprandial substrate deposition in hu-man forearm and adipose tissue in vivo.Clin Sci 79:339–348, 1990

10. Dimitriadis G, Mitrou P, Lanbadiari V,Boutati E, Maratou E, Panagiotakos D,Koukkou E, Tzanella M, Thalassinos N,Raptis SA: Insulin action in adipose tissueand muscle in hypothyroidism. J Clin En-docrinol Metab 91:4930–4937, 2006

11. Matthews D, Hosker J, Rudenski A, Nay-lor B, Treacher D, Turner R: Homeostasismodel assessment: insulin resistance andb-cell function from fasting plasma glu-cose and insulin concentrations. Diabeto-logia 28:412–419, 1985

12. Gutt M, Davis CL, Spitzer SB, Llabre MM,Kumar M, Czarnecki EM, SchneidermanN, Skyler JS, Marks JB: Validation of theinsulin sensitivity index (ISI1,120): com-parison with other measures. Diabetes ResClin Pract 47:177–184, 2000

13. Karpe F, Fielding BA, Ardilouze JL, Ilic V,Macdonald IA, Frayn KN: Effects of insu-lin on adipose tissue blood flow in man.J Physiol 540:1087–1093, 2002

14. Eriksson JW, Smith U, Waagstein F,Wysocki M, Janssson PA: Glucose turn-over and adipose tissue lipolysis are insu-lin resistant in healthy relatives of type 2diabetes patients: is cellular insulin resis-tance a secondary phenomenon? Diabetes48:1572–1578, 1999

15. Samra JS, Simpson EJ, Clark ML, ForsterCD, Humphreys SM, MacDonald IA,Frayn KN: Effects of epinephrine infusionon adipose tissue: interactions betweenblood flow and lipid metabolism. Am JPhysiol (Endocrinol Metab) 271:E834–E839, 1996

Postprandial blood flow in adipose tissue


Diabetes, the Metabolic Syndrome, andIschemic StrokeEpidemiology and possible mechanisms

ELLEN L. AIR, MD, PHD1

BRETT M. KISSELA, MD2

S troke affects more than 700,000 in-dividuals each year; it is the thirdlargest cause of death and the largest

cause of adult disability in the U.S. Diabe-tes is a major risk factor for the develop-ment of stroke, yet this risk is not realizedor understood by patients with diabetes.This likely reflects a lack of understandingwithin the medical community of how di-abetes confers this risk. We will explorethe potential underlying mechanisms thatlead to increased incidence of strokeamong diabetic patients. Beyond diabetesitself, the metabolic syndrome and itscomponents will also be discussed. Theimpact of diabetes and hyperglycemia onstroke outcomes and a discussion of cur-rent approaches to reduce stroke in thishigh-risk population are included. Becausetype 2 diabetes affects the vast majority ofthose diagnosed with diabetes, it will be theprimary focus of this discussion.

DEFINING THE PROBLEM — Ithas been well documented that diabetesconfers a significantly increased risk ofstroke, as well as increased mortality fol-lowing stroke (1–7). Stroke is a prevent-able disease with high personal andsocietal cost. While great progress hasbeen made in understanding the link be-tween diabetes and coronary heart disease(CHD), the literature on diabetes andstroke has been less enlightening. CHD isa larger problem that accounts for 40–

50% of mortality in diabetes. Because ofthe overwhelming impact of CHD, theimpact of stroke has been relatively under-appreciated. Thus, physicians, diabetes ed-ucators, and nurses are less equipped toeducate patients. We therefore review therelationship between diabetes and stroke.

Given that more than one millionpeople are diagnosed with diabetesyearly, a figure that is expected to rise, theimpact of diabetes on the incidence ofstroke is of increasing importance. Dia-betic patients compose roughly 6.3% ofthe U.S. population but account for 15–27% of all incident strokes, based on2002 estimates (4,7–12). This is certainlyan underestimation, as most studies clas-sify patients as having diabetes only if di-agnosed before stroke. When consideringage-adjusted incidence rates, diabetic pa-tients are 2.9 times as likely to have astroke compared with nondiabetic pa-tients, a disparity that is seen in multipleracial/geographic groups (4,7,9,13–15).This is due specifically to an increase inthe rate of ischemic stroke rather thanhemorrhagic stroke (7,16–18).

The heaviest burden of stroke for thegeneral population lies with older and mi-nority groups (4,12,19–22). Diabetes ap-pears to amplify these nonmodifiablerisks, in part due to the increased preva-lence of diabetes in these groups(7,23,24). Diabetes also confers an in-creased risk for neurovascular disease at

younger ages (25). The Greater Cincin-nati–Northern Kentucky Stroke Study(GCNKSS) found that the risk for isch-emic stroke in white diabetic patients ishigher at every age-group compared withnondiabetic patients, with highest relativerisk (RR) of 5.3 found in the 45- to 54-year age-group. Among African Ameri-cans, the highest risk was even greater (RR9.9) and was found in the 35- to 44-yearage-group. A substantial peak in strokerisk is seen in the 45- to 64-year age-group in whites and in the 35- to 54-yearage-group in African Americans (7).

Although stroke is more commonamong diabetic patients, most studies re-port a significantly reduced rate of tran-sient ischemic attacks (TIAs) in diabeticpatients compared with nondiabetic pa-tients. Diabetic patients are more likely topresent with cerebral infarct, indicatingthat ischemia in diabetic patients is lesslikely to be reversible (7,26–28). Thispresents a unique problem for preventingstroke in this population. TIAs can serveas a warning sign, providing a window ofopportunity for medical intervention toprevent a completed stroke. The relativelack of warning in diabetic patients re-quires that physicians, nurses, and educa-tors be aggressive about risk factorintervention, as comprehensive programsto reduce risk can be highly successful(29). For those who do present with aTIA, aggressive treatment is equally im-portant since diabetes has been shown toincrease the risk of subsequent completedstroke (30).

CAUSE AND EFFECT? — Many at-tempts have been made to discern the un-derlying mechanisms through whichdiabetes increases stroke risk. Such stud-ies have largely taken cues from the car-diovascular literature in which diabetesand the associated components of themetabolic syndrome (i.e., hypertensionand hyperlipidemia) have been found tocontribute to cardiovascular disease de-velopment (31–33). This approach hasbeen informative, yet the relationshipsbetween diabetes, the components of themetabolic syndrome, and stroke are

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the 1Department of Neurosurgery, University of Cincinnati College of Medicine, Cincinnati, Ohio; andthe 2Department of Neurology, University of Cincinnati College of Medicine, Cincinnati, Ohio.

Address correspondence and reprint requests to Brett M. Kissela, Department of Neurology, University ofCincinnati College of Medicine, 231 Albert Sabin Way, ML 0525, Cincinnati, OH 45267-0525. E-mail:[email protected].

Received for publication 21 July 2006 and accepted in revised form 8 September 2007.Published ahead of print at http://care.diabetesjournals.org on 11 September 2007. DOI: 10.2337/dc06-

1537.Abbreviations: ARIC, Atherosclerosis Risk in Communities; CAD, coronary artery disease; CARDS,

Collaborative AtoRvastatin Diabetes Study; CHD, coronary heart disease; CIMT, carotid intima-media thick-ness; EPIC, European Prospective Investigation Into Cancer; GCNKSS, Greater Cincinnati–Northern Ken-tucky Stroke Study; NHANES III, Third National Health and Nutrition Survey; TIA, transient ischemicattack; UKPDS, UK Prevention in Diabetes Study; WHR, waist-to-hip ratio.


© 2007 by the American Diabetes Association.

R e v i e w s / C o m m e n t a r i e s / A D A S t a t e m e n t sR E V I E W A R T I C L E


clearly unique. Here, we discuss these in-dividual relationships, highlighting thedifferences between stroke and cardiacrisk.

DIABETES VERSUSHYPERGLYCEMIA — As in any dis-cussion of diabetes and its sequelae, thefundamental question arises as to whetherstroke risk is increased due to chronic hy-perglycemia. Published studies provideconflicting evidence. Lehto et al. (34)studied 1,059 diabetic patients and corre-lated their baseline fasting glucose levels,A1C, and duration of diabetes with strokeover 7 years of follow-up. All three factorscontributed significantly to increased riskof stroke, while fasting hyperglycemia(�13.4 mmol/l) remained significant af-ter accounting for other cardiovascularrisk factors (odds ratio [OR] 2.6 [95% CI1.5–3.8] compared with normoglycemia)(34). The Honolulu Heart Program re-ported similar results in nondiabetic pa-tients when comparing the extremes(80th and 20th percentiles) of serum glu-cose levels (RR for thromboembolicstroke 1.4 [95% CI 1.1–1.8]) (16). AFinnish cohort study measured A1C andfasting glucose in diabetic and nondia-betic patients. In both groups, they founda significant association between eachmeasure of glucose control and stroke riskusing multivariate analysis (35). More re-cent data from the Atherosclerosis Risk inCommunities (ARIC) Study reiteratedthis relationship, finding an increased RRof stroke with increasing levels of A1C inboth diabetic and nondiabetic patients(36). In contrast, the European Prospec-tive Investigation Into Cancer (EPIC)-Norfolk Study did not find a significantrelationship between A1C and stroke riskuntil a threshold level was reached (37).

The only clinical trial to date that hasdirectly evaluated the effect of tight glu-cose control on stroke is the UK Preven-tion in Diabetes Study (UKPDS). Type 2diabetic patients in the intensive treat-ment group (average A1C 7.0%) had nosignificant reduction in stroke incidence(P � 0.52) compared with those receivingtraditional medical therapy (average A1C7.9%), indicating that tight glucose con-trol is not sufficient to prevent excessstrokes (38,39), though the study maynot have been sufficiently powered to de-tect a stroke-specific relationship and/orthe intensive control may not have been“intensive enough” to substantially im-pact stroke incidence.

To summarize, there is no clear rela-

tionship between hyperglycemia andstroke incidence. Rather, it is apparentthat diabetic patients have an increasedrisk of stroke regardless of their level ofmetabolic control.

INSULIN RESISTANCE, THEMETABOLIC SYNDROME,AND STROKE — Without substan-tive evidence that intensive glucose con-trol reduces stroke risk, the focus hasshifted to insulin resistance and its asso-ciated metabolic syndrome. Type 2 diabe-tes, characterized by an inability to produceenough insulin to overcome insulin resis-tance, frequently coexists with a constella-tion of cardiovascular risk factors includinghypertension, obesity, and hyperlipidemia.Together, these have been termed the met-abolic syndrome (also known as syndromeX or insulin-resistance syndrome). The rolethat these factors have played individually,as well as together, in the development ofcardiovascular disease (40) has made themthe target of studies regarding stroke as well.

INSULIN RESISTANCE — Insulinresistance, as measured by basal hyperin-sulinemia (or impaired glucose tolerance,which is equated to a state of insulin re-sistance) has been associated with coro-nary artery disease (CAD) and subsequentcardiovascular events (41– 44). Severalstudies have evaluated whether an analo-gous relationship exists between insulinresistance and stroke. In a retrospectivestudy, impaired glucose tolerance was notassociated with stroke (45). A prospectivestudy of Japanese men found no relation-ship between insulin resistance andstroke incidence (46). In contrast, theARIC Study found an increase in RR forischemic stroke of 1.19 for every 50pmol/l increase in basal insulin amongnondiabetic patients, supporting a rolefor insulin resistance (2). This was similarto results from the elderly patient popu-lation of the Finnish cohort study that in-cluded both diabetic patients andnondiabetic patients (35). As with studiesof insulin resistance and cardiovasculardisease, the association of insulin resis-tance with stroke is attenuated by the ad-justment for other cardiovascular riskfactors (2,35,43,44). However, data fromthe Third National Health and NutritionSurvey (NHANES III) revealed a small,but significant, independent associationbetween insulin resistance and strokewhen other risk factors such as hyperten-sion and level of glycemic control weretaken into account (OR 1.06) (47). To

summarize, a significant association be-tween insulin resistance and stroke riskhas been found, but the magnitude of thisassociation is less than the associationseen with cardiovascular disease.

HYPERTENSION — Among thecomponents of the metabolic syndrome,hypertension is the single most importantrisk factor for the development of stroke.In this respect, stroke varies significantlyfrom cardiac disease, where hypertensionis a lesser risk factor.

Evidence suggests that some of the in-creased risk of stroke among diabetic pa-tients is attributable to the increasedprevalence of hypertension. The GCNKSSfound that the prevalence of hypertensionwas 79% among diabetic patients and57% among nondiabetic patients (P �0.0001) (7). A significant, thoughsmaller, difference was found in theCopenhagen Stroke Study (48 vs. 30%,respectively, P � 0.0001) (10). Prospec-tively, follow-up of diabetic patients inthe UKPDS found that the occurrence ofvascular complications, including stroke,were significantly associated with hyper-tension (48). The converse relationshiphas also been seen. Among hypertensivepatients, diabetes is a significant predictorof ischemic stroke (OR 3.76 [95% CI1.67–8.46]) (49). Data from the ARICStudy suggest a similar increased riskamong diabetic patients with prehyper-tension, as compared with nondiabeticpatients, although the number of strokeswas insufficient to calculate an RR forstroke alone (50). No study has includedstatistical modeling to specifically addresswhether hypertension fully accounts forthe increased risk of stoke in diabetic pa-tients. It appears that the two are syner-gistic in increasing stroke risk andaccount for up to 40% of the population-attributable risk for all ischemic strokes(7). A number of studies have found anti-hypertensive treatment to reduce the inci-dence of cardiovascular events, includingstroke, in those with diabetes (51–57), butfewer studies have focused on stroke specif-ically. The Systolic Hypertension in EuropeTrial specifically noted a 73% decrease instroke incidence in diabetic patients treatedwith antihypertensive medication. Strokeincidence was decreased in nondiabetic pa-tients by 38% (58). Thus, diabetic patientsappear to benefit preferentially from antihy-pertensive treatment.

HYPERLIPIDEMIA — Hyperlipid-emia is one of the most important risk

Diabetes and ischemic stroke


factors for CHD and CAD but a less im-portant risk factor for stroke. As with hy-pertension, diabetic patients who havesuffered a stroke are more likely to havehyperlipidemia than those without diabe-tes (16 vs. 8%, respectively, P � 0.0001 inthe GCNKSS) (7,10). It is currently notclear to what degree the increased preva-lence of hyperlipidemia accounts for theincreased risk of stroke, especially as thecontribution of hyperlipidemia alone tostroke incidence is controversial (59 –64). Subset analysis from large placebo-controlled trials, such as the HelsinkiHeart Study and Scandinavian SimvastatinSurvival Study, which evaluated choles-terol reduction as primary or secondaryprevention of cardiovascular disease, in-dicate that diabetic patients may benefitpreferentially from treatment in stroke re-duction. Recently reported results fromthe Heart Protection Study, in contrast,did not support this difference, findingthat risk reduction did not vary with dia-betic status (65). The Collaborative AtoR-vastatin Diabetes Study (CARDS), whichexpressly evaluated the contribution ofhyperlipidemia to stroke risk in the dia-betic population without known CAD,was halted early due to a significant 48%reduction in the incidence of strokeamong the treatment group (66). CARDS,taken together with the Stroke Preventionby Aggressive Reduction in CholesterolLevels Study (67), is highly significant inthat statin treatment can now be recom-mended for stroke prevention even in pa-tients who do not have cardiovasculardisease, regardless of diabetes status.However, based on the CARDS results, itseems that patients with diabetes may sig-nificantly benefit from statins, making iteven more important that those with dia-betes be considered for statin treatment aspart of their stroke prevention regimen.

OBESITY — Obesity contributes tomore than 300,000 deaths per year andnearly doubles the risk of death from allcauses (68–70). Given its particular asso-ciation with CAD, hypertension, and dia-betes (71), investigators have attemptedto discern the contribution that obesitymakes to stroke incidence with variableresults. Many studies utilize BMI (mea-sured as weight divided by the square ofheight in meters), which provides a broadthough nonspecific estimate of obesity, iseasily obtained from patient self-report ormedical charts and is commonly used inclinical practice. Both the ARIC andNorthern Manhattan Stroke Study failed

to find a convincing association betweenBMI and risk for stroke (2,72). An associ-ation has been noted in studies of specificsubpopulations, such as middle-aged,Korean, or nonsmoking Japanese men(73–75). The Nurses’ Health Study re-ported a significant association with BMI,such that subjects with BMI 27–28.9kg/m2 had an RR of 1.8 (95% CI 1.2–2.6),subjects with BMI 29–31.9 kg/m2 had anRR of 1.9 (1.3–2.8), and subjects withBMI �32 kg/m2 had an RR of 2.4 (1.6–3.5) compared with those with BMI �25kg/m2 (76). A less robust, but still signif-icant, association was found in the Wom-en’s Health Study (77). The Physician’sHealth Study found an RR of 1.95 (1.39–2.72) for ischemic stroke for those withBMI �30 kg/m2 compared with thosewith BMI �23 kg/m2. The risk increasedby 6% for each unit increase in BMI, al-though it was attenuated when other car-diovascular risk factors were taken intoaccount (78).

While BMI has been commonly usedin the literature as an obesity measure,many studies have shown it to poorly re-flect the health impact of obesity. Rather,abdominal obesity has been more specif-ically associated with vascular disease andotherhealthcomplications (79).Waist-to-hip ratio (WHR), while highly correlatedwith BMI, better represents abdominalobesity and therefore may provide addi-tional information on stroke risk. Despitethe lack of a relationship between strokeand BMI, the Northern Manhattan StrokeStudy did find a significant relationshipbetween WHR and risk of stroke. Analysisincluded 576 ischemic stroke patientsand 1,142 age-, sex-, and race/ethnicity-matched control subjects. Comparedwith the first quartile, the third and fourthquartiles of WHR had an increased risk ofstroke (third quartile: OR 2.4 [95% CI1.5–3.9]; fourth quartile: 3.0 [1.8–4.8])after adjustment for other risk factors.These findings were consistent acrossboth sexes and all race/ethnic groups, al-though the effect of WHR was strongeramong younger subjects (72). Directcomparison of BMI versus WHR andstroke risk in 28,643 male health careprofessionals without previous cardiovas-cular or cerebrovascular disease yieldedsimilar results. RR for the first and fifthquintiles of WHR was 2.33 (95% CI 1.25–4.37), whereas that for the first and fifthquintiles of BMI was 1.29 (0.73–2.27)(80).

Taken together, these studies suggestthat obesity—in particular, abdominal

obesity—is a significant risk factor forischemic stroke (81). Regardless, the im-pact that obesity has on the risk ofdiabetes, CAD, hypertension, andhyperlipidemia will confound studiesthat address the risk of stroke (71). It hasbeen estimated that the reductions in di-abetes, hypertension, and hyperlipidemiaassociated with a 10% weight loss couldlead to reduction of stroke of up to 13 per1,000 people (82).

MICROALBUMINURIA — TheWorld Health Organization definition ofthe metabolic syndrome also includes mi-croalbuminuria (30–300 mg/24 h) as afinal component. Microalbuminuria is asignificant marker of cardiovascular dis-ease and is highly associated with hyper-tension (83,84). It is encountered indiabetic patients more than twice as oftenas in nondiabetic patients (84) and mayalso contribute to the increased risk ofstroke. The largest population-based pro-spective study to evaluate microalbumin-uria and stroke risk is the EPIC-NorfolkStudy. Among 23,630 individuals aged40–79 years over 7.2 years of follow-up,microalbuminuria conferred a signifi-cantly increased risk of total and ischemicstroke in multivariate modeling (hazardratio [HR] 1.49 [95% CI 1.13–2.14] and2.01 [1.29 –3.31], respectively) (85).Data from the Heart Outcomes Preven-tion Evaluation Study implicate mi-croalbuminuria as a factor in strokeincidence among those with diabetes(57). Treatment of nonhypertensive dia-betic patients with an ACE inhibitor, aclass of medications known to reduce mi-croalbuminuria (86–88), reduced strokeincidence by 32% despite a minimal de-crease in blood pressure (57). These datasupport a role for microalbuminuria inincreasing the risk of ischemic stroke,which may not be entirely dependent onits direct relationship with hypertensionand other well-known stroke risk factors.

THE METABOLICSYNDROME — Each of the compo-nents of the metabolic syndrome is asso-ciated with higher stroke risk to variousdegrees, as described above. As has beenmentioned, analysis of individual factorscauses substantial adjustment of observedrisk because of the interrelationship ofthese factors. Therefore, studying themetabolic syndrome as a whole may pro-vide a better estimation of the true risk forischemic stroke.

The Botnia Study examined risk for

Air and Kissela


cardiovascular events and stroke con-ferred by the metabolic syndrome in4,483 subjects. In a multiple logistic re-gression analysis, the metabolic syndromewas a significant independent risk forstroke (RR 2.3, P � 0.001 compared withthose without the metabolic syndrome).None of the individual components of themetabolic syndrome contributed signifi-cantly to stroke risk (89). Similar resultswere obtained from examination of morethan 10,000 subjects in the NHANES III.In logistic regression modeling, the meta-bolic syndrome was associated with in-creased odds of stroke (OR 2.2 [95% CI1.5–3.2] compared with those withoutthe metabolic syndrome). After the meta-bolic syndrome was in the model, eachindividual component was also tested.Only hypertriglyceridemia entered as anadditional factor with independent sig-nificance, while hypertension and insu-lin resistance/diabetes trended towardsignificance (90). A few studies haveevaluated the risk of stroke associatedwith the metabolic syndrome in the ab-sence of diabetes, revealing similar two-fold increases (91,92). In the ARICStudy, both hypertension and low HDLcholesterol independently and signifi-cantly increased risk (92).

The data presented above provide ev-idence that the individual components ofthe metabolic syndrome significantlycontribute to the incidence of ischemicstroke. These components are more prev-alent among diabetic patients and may actsynergistically to promote increased riskof stroke. In addition, several studies sup-port a significant relationship between thecollective metabolic syndrome and isch-emic stroke.

The metabolic syndrome and diabe-tes have their association with insulin re-sistance in common. At a cellular andmolecular level, insulin resistance conferschanges that are becoming recognized asincreasingly important in the pathophys-iology of vascular disease, includingstroke.

ENDOTHELIALDYSFUNCTION AND NITRICOXIDE — Both diabetic patients andthose with impaired glucose tolerancehave decreased endothelium-dependentvasodilation (93–95), due to either de-creased nitric oxide (NO) production orimpaired NO metabolism (95). Normally,NO exerts a protective effect against plate-let aggregation and plays an important

role in the response to ischemic challenge(96,97).

Only indirect evidence is available atthe present time linking NO dysregula-tion and stroke. A recent study found adecreased response of cerebrovascularblood flow to NO synthase inhibition indiabetic patients compared with nondia-betic patients, although not enough pa-tients were enrolled to determinesignificance (98). In addition, parasym-pathetic neurons that secrete NO into theperivascular space have been docu-mented to degenerate and eventually diein the absence of insulin signaling (99).Numerous studies have found that HMG-CoA reductase inhibitors (statins), whichupregulate NO synthesis in addition totheir activity in stabilizing atheroscleroticplaques (100), significantly reduce therisk of stroke (56,67,101–104). The dualactions of statins make it difficult to dis-tinguish which action exerts the greatesteffect. However, the growing body of ev-idence indicates that statins exert protec-tive effects against stroke independent ofchanges in cholesterol levels.

HYPERCOAGUABILITYCONFERRED BY DIABETES —Defects in endothelial function may befurther confounded by the hypercoagula-ble state of diabetic patients. Plasminogenactivator inhibitor-1 and antithrombinIII, which inhibit fibrinolysis, as well astissue plasminogen activator antigen, amarker of impaired fibrinolysis, consis-tently have been found to be elevated indiabetic patients and in those with insulinresistance (105–107). Some studies havefurther suggested that coagulation fac-tors, such as factor VII, factor VIII, andvon Willebrand factor, also rise with de-gree of insulin resistance (108,109). Thisupregulation is likely secondary to achronic inflammatory state induced bydiabetes, as several inflammatory markers(e.g., C-reactive protein, lipoprotein-associated phospholipase A2) have beencorrelated with increased thrombotic fac-tors and stroke incidence (108,110 –112). The promotion of thrombusformation likely occurs via platelet hyper-reactivity. Studies of platelets from dia-betic patients have found increasedaggregation in response to ADP (113), aresponse that may be mediated by the up-regulation of GPIIb-IIIa receptors that oc-curs in diabetic patients (114). Insulinnormally acts to inhibit platelet aggrega-tion in response to ADP; however, thisaction is attenuated in diabetic patients

(115). Thromboxane A2 is also elevated indiabetic patients, possibly contributing tohyperaggregation as well (116).

The relative contribution of thesemechanisms to increased ischemic strokerisk in those with diabetes has not beenspecifically evaluated, although severalstudies have implicated these pathways inthe general population. In both cross-sectional and prospective studies, in-creased tissue plasminogen activatorantigen and plasminogen activator inhib-itor-1 levels have been significantly asso-ciated with ischemic stroke (117–119).Treatment with aspirin or clopidogrel tar-gets platelet aggregation by inhibitingthromboxane A2 and ADP, respectively,and are now widely used in the secondaryprevention of stroke, as they significantlyreduce the risk of recurrent stroke (120–125). Several trials, such as the Clopi-dogrel for High Atherothrombotic Riskand Ischemic Stabilization, Management,and Avoidance (CHARISMA), Clopi-dogrel versus Aspirin in Patients at Risk ofIschemic Events (CAPRIE), and Manage-ment of ATherothrombosis with Clopi-dogrel in High-risk patients (MATCH)studies, evaluated whether diabetic pa-tients derived more or less benefit fromantiplatelet therapy in preventing recur-rent ischemic events with mixed results.As the reported end point in these studieswas a composite of all ischemic eventsand mortality, the specific impact of anti-platelet therapy in diabetic patients onstroke is unclear (126–128). Further in-vestigation is required to determine therelative importance of these mechanismsin diabetic patients.

CAROTID INTIMA-MEDIATHICKNESS — Consideration hasalso been given to the impact of the in-creased incidence of atherosclerosisamong those with diabetes and stroke in-cidence. Carotid intima-media thickness(CIMT) has been found in a number ofstudies to be increased with diabetes. TheInsulin Resistance Atherosclerosis Studyfound a significant increase in commoncarotid thickness in the setting of estab-lished diabetes as compared with thosewith newly diagnosed diabetes (129). Al-though not to the same degree, impairedglucose tolerance is also associated withincreased CIMT (130). Diabetic patientsthat have suffered a stroke have signifi-cantly greater CIMT than both those with-out stroke and nondiabetic patients(131,132). As hyperglycemia, regardlessof diabetes duration, was directly related



to CIMT, tight glucose control may yieldbenefits on carotid disease (129).

SURVIVING STROKE — Despitethe uncertainties of the pathogenesis ofstroke in those with diabetes, the im-pact of hyperglycemia and diabetes onoutcomes has been more consistently de-fined. Hyperglycemia during the post-stroke period, regardless of diabeticstatus, is associated with increased mor-bidity and mortality. Studies have gener-ally found increased 30-day and 1-yearmortality rates among hyperglycemic pa-tients (133–137), although increasedmortality was not seen in other studies(7,138). Morbidity, as defined by func-tional outcome and neurologic recov-ery, is also worsened in the setting ofhyperglycemia and diabetes (134,139 –142). This holds true among those withonly transient hyperglycemia, althoughsuch individuals fare better than thosewith chronically elevated glucose levelswhether diagnosed pre- or poststrokewith diabetes (143,144). In imagingstudies, the initial infarct size and in-farct progression are greater in hyper-glycemic patients (142,145–147). Onerecent study has found a decreased re-canalization rate following recombinanttissue plasminogen activator (rt-PA) ad-ministration in the presence of hyper-glycemia, although this was not seen inthe pivotal National Institute of Neuro-logical Diseases and Stroke rt-PA trial(139,148). Normalization of glucose lev-els was associated with 4.6 timesdecreased risk in mortality in oneretrospective study, indicating the poten-tially large impact that can be made withaggressive medical management in thesepatients (149).

Diabetes is also one of the most con-sistent predictors of recurrent stroke orstroke after TIA (150 –162). The in-creased risk of recurrent stroke due to di-abetes ranges from 2.1 to 5.6 times that ofnondiabetic patients (154,156) and is in-dependent of glucose control during theinterstroke period (163). The significanceof these findings is underscored by theincreased morbidity and mortality associ-ated with recurrent stroke (164).

CHALLENGES AHEAD — Diabetessignificantly increases the risk of incidentstroke and stroke recurrence. The magni-tude of this problem will continue to ex-pand as the prevalence of diabetes increasesin the U.S., thus presenting numerouschallenges for the future. Foremost

among these is educating those with di-abetes as to their true risk of stroke. Asignificant barrier appears to be the in-congruence between the informationthe medical community believes it isimparting to patients and the actuallevel of knowledge demonstrated by pa-tients. Ninety percent of physicians reportdiscussing the risk of cardiovascular dis-ease and the importance of prevention,although only one-half of patients reporttheir physician had discussed risk factormodification (165). Recent data from theREduction of Atherothrombosis for Con-tinued Health (REACH) registry corrobo-rates the continued undertreatment ofcardiovascular risk factors (166). Frequentand repeated patient advising regardingcardiovascular and cerebrovascularcomplications of diabetes and warningsigns is necessary to improve utilizationof primary and secondary preventionmeasures.

The potential benefit of aggressivemultiple risk-reduction measures in thosewith diabetes has been highlighted by theSteno-2 Study. Intensive standardizedrisk factor reduction, including 1) treat-ment of hyperglycemia, hypertension,dyslipidemia, and microalbuminuria; 2)secondary prevention of cardiovasculardisease with aspirin; and 3) behavioralmodification, resulted in significant re-ductions in cardiovascular disease, in-cluding stroke (HR 0.47 [95% CI 0.24–0.73]). This effect was larger than thatseen in studies that targeted treatment toindividual risk factors (29). Although thespecific mechanisms that underlie the re-lationship between diabetes, the meta-bolic syndrome, and stroke requireongoing investigation to provide newmethods for prevention and treatment,these data underscore the strides that canbe made with the tools at hand.

References1. Kothari V, Stevens RJ, Adler AI, Stratton

IM, Manley SE, Neil HA, Holman RR:UKPDS 60: risk of stroke in type 2 dia-betes estimated by the UK ProspectiveDiabetes Study risk engine. Stroke 33:1776–1781, 2002

2. Folsom AR, Rasmussen ML, ChamblessLE, Howard G, Cooper LS, Schmidt MI,Heiss G: Prospective associations of fast-ing insulin, body fat distribution, and di-abetes with risk of ischemic stroke: theAtherosclerosis Risk in Communities(ARIC) Study Investigators. DiabetesCare 1077–1083, 1999

3. Tuomilehto J, Rastenyte D, Jousilahti P,

Sarti C, Vartiainen E: Diabetes mellitusas a risk factor for death from stroke:prospective study of the middle-agedFinnish population. Stroke 27:210–215,1996

4. Wolf PA, D’Agostino RB, Belanger AJ,Kannel WB: Probability of stroke: a riskprofile from the Framingham Study.Stroke 22:312–318, 1991

5. Ho JE, Paultre F, Mosca L: Is diabetesmellitus a cardiovascular disease riskequivalent for fatal stroke in women?Data from the Women’s Pooling Project.Stroke 34:2812–2816, 2003

6. Davis BR, Vogt T, Frost PH, Burlando A,Cohen J, Wilson A, Brass LM, FrishmanW, Price T, Stamler J: Risk factors forstroke and type of stroke in persons withisolated systolic hypertension: SystolicHypertension in the Elderly ProgramCooperative Research Group. Stroke 29:1333–1340, 1998

7. Kissela BM, Khoury J, Kleindorfer D,Woo D, Schneider A, Alwell K, Miller R,Ewing I, Moomaw CJ, Szaflarski JP,Gebel J, Shukla R, Broderick JP: Epide-miology of ischemic stroke in patientswith diabetes: the Greater Cincinnati/Northern Kentucky Stroke Study. Diabe-tes Care 28:355–359, 2005

8. Centers for Disease Control and Preven-tion: National Diabetes Fact Sheet: Gen-eral Information and National Estimates onDiabetes in the United States, 2002. At-lanta, GA, Department of Health andHuman Services, Centers for DiseaseControl and Prevention Atlanta, 2003

9. Grau AJ, Weimar C, Buggle F, HeinrichA, Goertler M, Neumaier S, Glahn J,Brandt T, Hacke W, Diener HC: Risk fac-tors, outcome, and treatment in sub-types of ischemic stroke: the Germanstroke data bank. Stroke 32:2559–2566,2001

10. Jorgensen H, Nakayama H, RaaschouHO, Olsen TS: Stroke in patients withdiabetes: the Copenhagen Stroke Study.Stroke 25:1977–1984, 1994

11. Lai SM, Alter M, Friday G, Sobel E: Prog-nosis for survival after an initial stroke.Stroke 26:2011–2015, 1995

12. Kissela B, Schneider A, Kleindorfer D,Khoury J, Miller R, Alwell K, Woo D,Szaflarski J, Gebel J, Moomaw C, Pan-cioli A, Jauch E, Shukla R, Broderick J:Stroke in a biracial population: the ex-cess burden of stroke among blacks.Stroke 35:426–431, 2004

13. Self-reported heart disease and strokeamong adults with and without diabete-s–United States, 1999–2001. MMWRMorb Mortal Wkly Rep 52:1065–1070,2003

14. Almdal T, Scharling H, Jensen JS, Vest-ergaard H: The independent effect oftype 2 diabetes mellitus on ischemicheart disease, stroke, and death: a pop-ulation-based study of 13,000 men and

Air and Kissela


women with 20 years of follow-up. ArchIntern Med 164:1422–1426, 2004

15. Ottenbacher KJ, Ostir GV, Peek MK,Markides KS: Diabetes mellitus as a riskfactor for stroke incidence and mortalityin Mexican American older adults. J Ger-ontol A Biol Sci Med Sci 59:M640–M645,2004

16. Abbott RD, Donahue RP, MacMahonSW, Reed DM, Yano K: Diabetes and therisk of stroke: the Honolulu Heart Pro-gram. JAMA 257:949–952, 1987

17. Adams HP Jr, Putman SF, Kassell NF,Torner JC: Prevalence of diabetes melli-tus among patients with subarachnoidhemorrhage. Arch Neurol 41:1033–1035, 1984

18. Karapanayiotides T, Piechowski-JozwiakB, van Melle G, Bogousslavsky J, DevuystG: Stroke patterns, etiology, and progno-sis in patients with diabetes mellitus. Neu-rology 62:1558–1562, 2004

19. Brown RD, Whisnant JP, Sicks JD,O’Fallon WM, Wiebers DO: Stroke inci-dence, prevalence, and survival: seculartrends in Rochester, Minnesota, through1989. Stroke 27:373–380, 1996

20. Sacco RL, Boden-Albala B, Gan R, ChenX, Kargman DE, Shea S, Paik MC,Hauser WA: Stroke incidence amongwhite, black, and Hispanic residents ofan urban community: the NorthernManhattan Stroke Study. Am J Epidemiol147:259–268, 1998

21. Broderick J, Brott T, Kothari R, Miller R,Khoury J, Pancioli A, Gebel J, Mills D,Minneci L, Shukla R: The Greater Cin-cinnati/Northern Kentucky StrokeStudy: preliminary first-ever and totalincidence rates of stroke among blacks.Stroke 29:415–421, 1998

22. Morgenstern LB, Smith MA, LisabethLD, Risser JM, Uchino K, Garcia N,Longwell PJ, McFarling DA, AkuwumiO, Al-Wabil A, Al-Senani F, Brown DL,Moye LA: Excess stroke in MexicanAmericans compared with non-His-panic whites: the Brain Attack Surveil-lance in Corpus Christi Project. Am JEpidemiol 160:376–383, 2004

23. Harris MI: Diabetes in America: epide-miology and scope of the problem. Dia-betes Care 21 (Suppl. 3):C11–C14, 1998

24. Kannel WB, McGee DL: Diabetes andcardiovascular disease: the Framinghamstudy. JAMA 241:2035–2038, 1979

25. Booth GL, Kapral MK, Fung K, Tu JV:Relation between age and cardiovasculardisease in men and women with diabetescompared with non-diabetic people: apopulation-based retrospective cohortstudy. Lancet 368:29–36, 2006

26. Fritz VU, Bilchik T, Levien LJ: Diabetesas risk factor for transient ischaemic at-tacks as opposed to strokes. Eur J VascSurg 1:259–262, 1987

27. Weinberger J, Biscarra V, Weisberg MK,Jacobson JH: Factors contributing to

stroke in patients with atheroscleroticdisease of the great vessels: the role ofdiabetes. Stroke 14:709–712, 1983

28. Lithner F, Asplund K, Eriksson S, HaggE, Strand T, Wester PO: Clinical charac-teristics in diabetic stroke patients. Dia-betes Metab 14:15–19, 1988

29. Gaede P, Vedel P, Larsen N, Jensen GV,Parving HH, Pedersen O: Multifactorialintervention and cardiovascular diseasein patients with type 2 diabetes. N EnglJ Med 348:383–393, 2003

30. Johnston SC, Rothwell PM, Nguyen-Huynh MN, Giles MF, Elkins JS, Bern-stein AL, Sidney S: Validation andrefinement of scores to predict very earlystroke risk after transient ischaemic at-tack. Lancet 369:283–292, 2007

31. Mooradian AD: Cardiovascular diseasein type 2 diabetes mellitus: current man-agement guidelines. Arch Intern Med 163:33–40, 2003

32. Bloomgarden ZT: Cardiovascular dis-ease in type 2 diabetes. Diabetes Care 22:1739–1744, 1999

33. Grundy SM, Benjamin IJ, Burke GL,Chait A, Eckel RH, Howard BV, MitchW, Smith SC Jr, Sowers JR: Diabetes andcardiovascular disease: a statement forhealthcare professionals from the Amer-ican Heart Association. Circulation 100:1134–1146, 1999

34. Lehto S, Ronnemaa T, Pyorala K, LaaksoM: Predictors of stroke in middle-agedpatients with non-insulin-dependent di-abetes. Stroke 27:63–68, 1996

35. Kuusisto J, Mykkanen L, Pyorala K,Laakso M: Non-insulin-dependent dia-betes and its metabolic control are im-portant predictors of stroke in elderlysubjects. Stroke 25:1157–1164, 1994

36. Selvin E, Coresh J, Shahar E, Zhang L,Steffes M, Sharrett AR: Glycaemia (hae-moglobin A1c) and incident ischaemicstroke: the Atherosclerosis Risk in Com-munities (ARIC) Study. Lancet Neurol4:821–826, 2005

37. Myint PK, Sinha S, Wareham NJ, Bing-ham SA, Luben RN, Welch AA, KhawKT: Glycated hemoglobin and risk ofstroke in people without known diabe-tes in the European Prospective Investi-gation into Cancer (EPIC)-Norfolkprospective population study: a thresh-old relationship? Stroke 38:271–275,2007

38. Effect of intensive blood-glucose controlwith metformin on complications inoverweight patients with type 2 diabetes(UKPDS 34): UK Prospective DiabetesStudy (UKPDS) Group. Lancet 352:854–865, 1998

39. Intensive blood-glucose control withsulphonylureas or insulin comparedwith conventional treatment and risk ofcomplications in patients with type 2 di-abetes (UKPDS 33): UK Prospective Di-abetes Study (UKPDS) Group. Lancet

352:837–853, 199840. Uwaifo GI, Ratner RE: The roles of insu-

lin resistance, hyperinsulinemia, andthiazolidinediones in cardiovascular dis-ease. Am J Med 115 (Suppl. 8A):12S–19S, 2003

41. Despres JP, Lamarche B, Mauriege P,Cantin B, Dagenais GR, Moorjani S, Lu-pien PJ: Hyperinsulinemia as an inde-pendent risk factor for ischemic heartdisease. N Engl J Med 334:952–957,1996

42. Pyorala M, Miettinen H, Halonen P,Laakso M, Pyorala K: Insulin resistancesyndrome predicts the risk of coronaryheart disease and stroke in healthy mid-dle-aged men: the 22-year follow-up re-sults of the Helsinki Policemen Study.Arterioscler Thromb Vasc Biol 20:538–544, 2000

43. Lakka HM, Lakka TA, Tuomilehto J, Siv-enius J, Salonen JT: Hyperinsulinemiaand the risk of cardiovascular death andacute coronary and cerebrovascularevents in men: the Kuopio IschaemicHeart Disease Risk Factor Study. ArchIntern Med 160:1160–1168, 2000

44. Bonora E, Willeit J, Kiechl S, Oberhollen-zer F, Egger G, Bonadonna R, Muggeo M:U-shaped and J-shaped relationships be-tween serum insulin and coronary heartdisease in the general population: theBruneck Study. Diabetes Care 21:221–230, 1998

45. Qureshi AI, Giles WH, Croft JB: Im-paired glucose tolerance and the likeli-hood of nonfatal stroke and myocardialinfarction: the Third National Healthand Nutrition Examination Survey.Stroke 29:1329–1332, 1998

46. Adachi H, Hirai Y, Tsuruta M, Fujiura Y,Imaizuml T: Is insulin resistance or dia-betes mellitus associated with stroke? An18-year follow-up study. Diabetes ResClin Pract 51:215–223, 2001

47. Bravata DM, Wells CK, Kernan WN,Concato J, Brass LM, Gulanski BI: Asso-ciation between impaired insulin sensi-tivity and stroke. Neuroepidemiology 25:69–74, 2005

48. Adler AI, Stratton IM, Neil HA, YudkinJS, Matthews DR, Cull CA, Wright AD,Turner RC, Holman RR: Association ofsystolic blood pressure with macrovas-cular and microvascular complicationsof type 2 diabetes (UKPDS 36): prospec-tive observational study. BMJ 321:412–419, 2000

49. Khealani BA, Syed NA, Maken S, MapariUU, Hameed B, Ali S, Qureshi R, AkhterN, Hassan A, Sonawalla AB, Baig SM,Wasay M: Predictors of ischemic versushemorrhagic strokes in hypertensive pa-tients. J Coll Physicians Surg Pak 15:22–25, 2005

50. Kshirsagar AV, Carpenter M, Bang H,Wyatt SB, Colindres RE: Blood pressureusually considered normal is associated



with an elevated risk of cardiovasculardisease. Am J Med 119:133–141, 2006

51. Efficacy of atenolol and captopril in re-ducing risk of macrovascular and mi-crovascular complications in type 2diabetes: UKPDS 39: UK ProspectiveDiabetes Study Group. BMJ 317:713–720, 1998

52. Pahor M, Psaty BM, Alderman MH,Applegate WB, Williamson JD, FurbergCD: Therapeutic benefits of ACE inhib-itors and other antihypertensive drugs inpatients with type 2 diabetes. DiabetesCare 23:888–892, 2000

53. Tatti P, Pahor M, Byington RP, Di MauroP, Guarisco R, Strollo G, Strollo F: Out-come results of the Fosinopril VersusAmlodipine Cardiovascular Events Ran-domized Trial (FACET) in patients withhypertension and NIDDM. Diabetes Care21:597–603, 1998

54. Hansson L, Lindholm LH, Niskanen L,Lanke J, Hedner T, Niklason A, Luoman-maki K, Dahlof B, de Faire U, Morlin C,Karlberg BE, Wester PO, Bjorck JE: Ef-fect of angiotensin-converting-enzymeinhibition compared with conventionaltherapy on cardiovascular morbidityand mortality in hypertension: the Cap-topril Prevention Project (CAPPP) ran-domised trial. Lancet 353:611–616,1999

55. Estacio RO, Jeffers BW, Hiatt WR, Big-gerstaff SL, Gifford N, Schrier RW: Theeffect of nisoldipine as compared withenalapril on cardiovascular outcomes inpatients with non-insulin-dependent di-abetes and hypertension. N Engl J Med338:645–652, 1998

56. Effects of ramipril on cardiovascular andmicrovascular outcomes in people withdiabetes mellitus: results of the HOPEstudy and MICRO-HOPE substudy:Heart Outcomes Prevention EvaluationStudy Investigators. Lancet 355:253–259, 2000

57. Bosch J, Yusuf S, Pogue J, Sleight P, LonnE, Rangoonwala B, Davies R, OstergrenJ, Probstfield J: Use of ramipril in pre-venting stroke: double blind random-ised trial. BMJ 324:699–702, 2002

58. Tuomilehto J, Rastenyte D, BirkenhagerWH, Thijs L, Antikainen R, Bulpitt CJ,Fletcher AE, Forette F, Goldhaber A,Palatini P, Sarti C, Fagard R: Effects ofcalcium-channel blockade in older pa-tients with diabetes and systolic hyper-tension: Systolic Hypertension inEurope Trial Investigators. N Engl J Med340:677–684, 1999

59. Iso H, Jacobs DR Jr, Wentworth D, Nea-ton JD, Cohen JD: Serum cholesterol lev-els and six-year mortality from stroke in350,977 men screened for the multiplerisk factor intervention trial. N Engl J Med320:904–910, 1989

60. Laws A, Marcus EB, Grove JS, Curb JD:Lipids and lipoproteins as risk factors for

coronary heart disease in men with ab-normal glucose tolerance: the HonoluluHeart Program. J Intern Med 234:471–478, 1993

61. Kagan A, Popper JS, Rhoads GG: Factorsrelated to stroke incidence in HawaiiJapanese men: the Honolulu HeartStudy. Stroke 11:14–21, 1980

62. Shahar E, Chambless LE, RosamondWD, Boland LL, Ballantyne CM, McGov-ern PG, Sharrett AR: Plasma lipid profileand incident ischemic stroke: the Ath-erosclerosis Risk in Communities (ARIC)study. Stroke 34:623–631, 2003

63. Harmsen P, Rosengren A, Tsipogianni A,Wilhelmsen L: Risk factors for stroke inmiddle-aged men in Goteborg, Sweden.Stroke 21:223–229, 1990

64. Cholesterol, diastolic blood pressure,and stroke: 13,000 strokes in 450,000people in 45 prospective cohorts: pro-spective studies collaboration. Lancet346:1647–1653, 1995

65. Collins R, Armitage J, Parish S, Sleigh P,Peto R: MRC/BHF Heart ProtectionStudy of cholesterol-lowering with sim-vastatin in 5963 people with diabetes: arandomised placebo-controlled trial.Lancet 361:2005–2016, 2003

66. Colhoun HM, Betteridge DJ, DurringtonPN, Hitman GA, Neil HA, LivingstoneSJ, Thomason MJ, Mackness MI, Charl-ton-Menys V, Fuller JH: Primary preven-tion of cardiovascular disease withatorvastatin in type 2 diabetes in the Col-laborative Atorvastatin Diabetes Study(CARDS): multicentre randomised pla-cebo-controlled trial. Lancet 364:685–696, 2004

67. Amarenco P, Bogousslavsky J, CallahanA 3rd, Goldstein LB, Hennerici M, Ru-dolph AE, Sillesen H, Simunovic L, Sza-rek M, Welch KM, Zivin JA: High-doseatorvastatin after stroke or transientischemic attack. N Engl J Med 355:549–559, 2006

68. Centers for Disease Control (CDC): TheSurgeon General’s 1989 report on re-ducing the health consequences ofsmoking: 25 years of progress. MMWRMorb Mortal Wkly Rep 38 (Suppl. 2):1–32, 1989

69. Mokdad AH, Marks JS, Stroup DF, Ger-berding JL: Correction: actual causes ofdeath in the United States, 2000. JAMA293:293–294, 2005

70. Mokdad AH, Marks JS, Stroup DF, Ger-berding JL: Actual causes of death in theUnited States, 2000. JAMA 291:1238–1245, 2004

71. Wilson PW, D’Agostino RB, Sullivan L,Parise H, Kannel WB: Overweight andobesity as determinants of cardiovascu-lar risk: the Framingham experience.Arch Intern Med 162:1867–1872, 2002

72. Suk SH, Sacco RL, Boden-Albala B,Cheun JF, Pittman JG, Elkind MS, PaikMC: Abdominal obesity and risk of isch-

emic stroke: the Northern ManhattanStroke Study. Stroke 34:1586–1592,2003

73. Jood K, Jern C, Wilhelmsen L, Rosen-gren A: Body mass index in mid-life isassociated with a first stroke in men: aprospective population study over 28years. Stroke 35:2764–2769, 2004

74. Song YM, Sung J, Davey Smith G, Ebra-him S: Body mass index and ischemicand hemorrhagic stroke: a prospectivestudy in Korean men. Stroke 35:831–836, 2004

75. Abbott RD, Behrens GR, Sharp DS, Ro-driguez BL, Burchfiel CM, Ross GW,Yano K, Curb JD: Body mass index andthromboembolic stroke in nonsmokingmen in older middle age: the HonoluluHeart Program. Stroke 25:2370–2376,1994

76. Rexrode KM, Hennekens CH, WillettWC, Colditz GA, Stampfer MJ, Rich-Edwards JW, Speizer FE, Manson JE: Aprospective study of body mass index,weight change, and risk of stroke inwomen. JAMA 277:1539–1545, 1997

77. Kurth T, Gaziano JM, Rexrode KM, KaseCS, Cook NR, Manson JE, Buring JE:Prospective study of body mass indexand risk of stroke in apparently healthywomen. Circulation 111:1992–1998,2005

78. Kurth T, Gaziano JM, Berger K, Kase CS,Rexrode KM, Cook NR, Buring JE, Man-son JE: Body mass index and the risk ofstroke in men. Arch Intern Med 162:2557–2562, 2002

79. Megnien JL, Denarie N, Cocaul M, Si-mon A, Levenson J: Predictive value ofwaist-to-hip ratio on cardiovascular riskevents. Int J Obes Relat Metab Disord 23:90–97, 1999

80. Walker SP, Rimm EB, Ascherio A, Kawa-chi I, Stampfer MJ, Willett WC: Bodysize and fat distribution as predictors ofstroke among US men. Am J Epidemiol144:1143–1150, 1996

81. Lu M, Ye W, Adami HO, Weiderpass E:Prospective study of body size and riskfor stroke amongst women below age60. J Intern Med 260:442–450, 2006

82. Oster G, Thompson D, Edelsberg J, BirdAP, Colditz GA: Lifetime health and eco-nomic benefits of weight loss amongobese persons. Am J Public Health 89:1536–1542, 1999

83. Ravera M, Ratto E, Vettoretti S, Viazzi F,Leoncini G, Parodi D, Tomolillo C, DelSette M, Maviglio N, Deferrari G, Pon-tremoli R: Microalbuminuria and sub-clinical cerebrovascular damage inessential hypertension. J Nephrol 15:519–524, 2002

84. Gerstein HC, Mann JF, Yi Q, Zinman B,Dinneen SF, Hoogwerf B, Halle JP,Young J, Rashkow A, Joyce C, Nawaz S,Yusuf S: Albuminuria and risk of cardio-vascular events, death, and heart failure

Air and Kissela


in diabetic and nondiabetic individuals.JAMA 286:421–426, 2001

85. Yuyun MF, Khaw KT, Luben R, WelchA, Bingham S, Day NE, Wareham NJ:Microalbuminuria and stroke in a Britishpopulation: the European ProspectiveInvestigation into Cancer in Norfolk(EPIC-Norfolk) population study. J In-tern Med 255:247–256, 2004

86. Marre M, Lievre M, Chatellier G, MannJF, Passa P, Menard J: Effects of low doseramipril on cardiovascular and renaloutcomes in patients with type 2 diabe-tes and raised excretion of urinary albu-min: randomised, double blind, placebocontrolled trial (the DIABHYCARstudy). BMJ 328:495, 2004

87. Slowik A, Turaj W, Iskra T, Strojny J,Szczudlik A: Microalbuminuria in non-diabetic patients with acute ischemicstroke: prevalence, clinical correlates,and prognostic significance. CerebrovascDis 14:15–21, 2002

88. Lasaridis AN, Sarafidis PA: Diabetic ne-phropathy and antihypertensive treat-ment: what are the lessons from clinicaltrials? Am J Hypertens 16:689–697,2003

89. Isomaa B, Almgren P, Tuomi T, ForsenB, Lahti K, Nissen M, Taskinen MR,Groop L: Cardiovascular morbidity andmortality associated with the metabolicsyndrome. Diabetes Care 24:683–689,2001

90. Ninomiya JK, L’Italien G, Criqui MH,Whyte JL, Gamst A, Chen RS: Associa-tion of the metabolic syndrome with his-tory of myocardial infarction and strokein the Third National Health and Nutri-tion Examination Survey. Circulation109:42–46, 2004

91. Kurl S, Laukkanen JA, Niskanen L, Laak-sonen D, Sivenius J, Nyyssonen K, Sa-lonen JT: Metabolic syndrome and therisk of stroke in middle-aged men. Stroke37:806–811, 2006

92. McNeill AM, Rosamond WD, GirmanCJ, Golden SH, Schmidt MI, East HE,Ballantyne CM, Heiss G: The metabolicsyndrome and 11-year risk of incidentcardiovascular disease in the Atheroscle-rosis Risk in Communities study. Diabe-tes Care 28:385–390, 2005

93. Hogikyan RV, Galecki AT, Pitt B, HalterJB, Greene DA, Supiano MA: Specificimpairment of endothelium-dependentvasodilation in subjects with type 2 dia-betes independent of obesity. J Clin En-docrinol Metab 83:1946–1952, 1998

94. Steinberg HO, Chaker H, Leaming R,Johnson A, Brechtel G, Baron AD: Obe-sity/insulin resistance is associated withendothelial dysfunction: implicationsfor the syndrome of insulin resistance.J Clin Invest 97:2601–2610, 1996

95. Maejima K, Nakano S, Himeno M, TsudaS, Makiishi H, Ito T, Nakagawa A, Kigo-shi T, Ishibashi T, Nishio M, Uchida K:

Increased basal levels of plasma nitricoxide in type 2 diabetic subjects: rela-tionship to microvascular complica-tions. J Diabetes Complications 15:135–143, 2001

96. McCarty MF: Up-regulation of endothe-lial nitric oxide activity as a central strat-egy for prevention of ischemic stroke:just say NO to stroke! Med Hypotheses55:386–403, 2000

97. Cosentino F, Rubattu S, Savoia C, Ven-turelli V, Pagannonne E, Volpe M: En-dothelial dysfunction and stroke.J Cardiovasc Pharmacol 38 (Suppl. 2):S75–S78, 2001

98. Nazir FS, Alem M, Small M, Connell JM,Lees KR, Walters MR, Cleland SJ:Blunted response to systemic nitric ox-ide synthase inhibition in the cerebralcirculation of patients with type 2 diabe-tes. Diabet Med 23:398–402, 2006

99. Cellek S, Anderson PN, Foxwell NA: Ni-trergic neurodegeneration in cerebral ar-teries of streptozotocin-induced diabeticrats: a new insight into diabetic stroke.Diabetes 54:212–219, 2005

100. Endres M, Laufs U, Huang Z, NakamuraT, Huang P, Moskowitz MA, Liao JK:Stroke protection by 3-hydroxy-3-methylglutaryl (HMG)-CoA reductaseinhibitors mediated by endothelial nitricoxide synthase. Proc Natl Acad Sci U S A95:8880–8885, 1998

101. Gil-Nunez AC, Villanueva JA: Advan-tages of lipid-lowering therapy in cere-bral ischemia: role of HMG-CoAreductase inhibitors. Cerebrovasc Dis 11(Suppl. 1):85–95, 2001

102. Plehn JF, Davis BR, Sacks FM, RouleauJL, Pfeffer MA, Bernstein V, Cuddy TE,Moye LA, Piller LB, Rutherford J, Simp-son LM, Braunwald E: Reduction ofstroke incidence after myocardial infarc-tion with pravastatin: the Cholesteroland Recurrent Events (CARE) study: theCare Investigators. Circulation 99:216–223, 1999

103. Blauw GJ, Lagaay AM, Smelt AH, West-endorp RG: Stroke, statins, and choles-terol: a meta-analysis of randomized,placebo-controlled, double-blind trialswith HMG-CoA reductase inhibitors.Stroke 28:946–950, 1997

104. Collins R, Armitage J, Parish S, Sleight P,Peto R: Effects of cholesterol-loweringwith simvastatin on stroke and othermajor vascular events in 20536 peoplewith cerebrovascular disease or otherhigh-risk conditions. Lancet 363:757–767, 2004

105. Aoki I, Shimoyama K, Aoki N, HomoriM, Yanagisawa A, Nakahara K, Kawai Y,Kitamura SI, Ishikawa K: Platelet-depen-dent thrombin generation in patientswith diabetes mellitus: effects of glyce-mic control on coagulability in diabetes.J Am Coll Cardiol 27:560–566, 1996

106. Meigs JB, Mittleman MA, Nathan DM,

Tofler GH, Singer DE, Murphy-SheehyPM, Lipinska I, D’Agostino RB, WilsonPW: Hyperinsulinemia, hyperglycemia,and impaired hemostasis: the Framing-ham Offspring Study. JAMA 283:221–228, 2000

107. Davi G, Gennaro F, Spatola A, CatalanoI, Averna M, Montalto G, Amato S, No-tarbartolo A: Thrombin-antithrombinIII complexes in type II diabetes melli-tus. J Diabetes Complications 6:7–11, 1992

108. Romano M, Guagnano MT, Pacini G,Vigneri S, Falco A, Marinopiccoli M,Manigrasso MR, Basili S, Davi G: Associ-ation of inflammation markers withimpaired insulin sensitivity and coag-ulative activation in obese healthywomen. J Clin Endocrinol Metab 88:5321–5326, 2003

109. Mertens I, Van Gaal LF: Obesity, haemo-stasis and the fibrinolytic system. ObesRev 3:85–101, 2002

110. Ballantyne CM, Hoogeveen RC, Bang H,Coresh J, Folsom AR, Chambless LE,Myerson M, Wu KK, Sharrett AR, Boer-winkle E: Lipoprotein-associated phos-pholipase A2, high-sensitivity C-reactiveprotein, and risk for incident ischemicstroke in middle-aged men and womenin the Atherosclerosis Risk in Commu-nities (ARIC) study. Arch Intern Med165:2479–2484, 2005

111. Adeniyi A, Folsom AR, Brancati FL, Des-vorieux M, Pankow JS, Taylor H: Inci-dence and risk factors for cardiovasculardisease in African Americans with diabe-tes: the Atherosclerosis Risk in Commu-nities (ARIC). J Natl Med Assoc 94:1025–1035, 2002

112. Engstrom G, Stavenow L, Hedblad B,Lind P, Eriksson KF, Janzon L,Lindgarde F: Inflammation-sensitiveplasma proteins, diabetes, and mortalityand incidence of myocardial infarctionand stroke: a population-based study.Diabetes 52:442–447, 2003

113. Mandal S, Sarode R, Dash S, Dash RJ:Hyperaggregation of platelets detectedby whole blood platelet aggregometry innewly diagnosed noninsulin-dependentdiabetes mellitus. Am J Clin Pathol 100:103–107, 1993

114. Tschoepe D, Roesen P, Kaufmann L,Schauseil S, Kehrel B, Ostermann H,Gries FA: Evidence for abnormal plateletglycoprotein expression in diabetes mel-litus. Eur J Clin Invest 20:166–170, 1990

115. Trovati M, Mularoni EM, Burzacca S,Ponziani MC, Massucco P, Mattiello L,Piretto V, Cavalot F, Anfossi G: Impairedinsulin-induced platelet antiaggregatingeffect in obesity and in obese NIDDMpatients. Diabetes 44:1318–1322, 1995

116. Halushka PV, Rogers RC, Loadholt CB,Colwell JA: Increased platelet throm-boxane synthesis in diabetes mellitus.J Lab Clin Med 97:87–96, 1981

117. Macko RF, Kittner SJ, Epstein A, Cox



DK, Wozniak MA, Wityk RJ, Stern BJ,Sloan MA, Sherwin R, Price TR, Mc-Carter RJ, Johnson CJ, Earley CJ, Buch-holz DW, Stolley PD: Elevated tissueplasminogen activator antigen andstroke risk: the Stroke Prevention InYoung Women Study. Stroke 30:7–11,1999

118. Lindgren A, Lindoff C, Norrving B, Ast-edt B, Johansson BB: Tissue plasmino-gen activator and plasminogen activatorinhibitor-1 in stroke patients. Stroke 27:1066–1071, 1996

119. Ridker PM, Hennekens CH, StampferMJ, Manson JE, Vaughan DE: Prospec-tive study of endogenous tissue plas-minogen activator and risk of stroke.Lancet 343:940 –943, 1994

120. Teal PA: Recent clinical trial results withantiplatelet therapy: implications instroke prevention. Cerebrovasc Dis 17(Suppl. 3):6–10, 2004

121. Lalouschek W, Lang W, Mullner M:Current strategies of secondary preven-tion after a cerebrovascular event: the Vi-enna stroke registry. Stroke 32:2860–2866, 2001

122. A comparison of two doses of aspirin (30mg vs. 283 mg a day) in patients after atransient ischemic attack or minor isch-emic stroke: the Dutch TIA Trial StudyGroup. N Engl J Med 325:1261–1266,1991

123. Swedish Aspirin Low-Dose Trial (SALT)of 75 mg aspirin as secondary prophy-laxis after cerebrovascular ischaemicevents: the SALT Collaborative Group.Lancet 338:1345–1349, 1991

124. A randomized trial of aspirin and sulfin-pyrazone in threatened stroke: the Cana-dian Cooperative Study Group. N EnglJ Med 299:53–59, 1978

125. Diener HC, Cunha L, Forbes C, SiveniusJ, Smets P, Lowenthal A: EuropeanStroke Prevention Study. 2. Dipyridam-ole and acetylsalicylic acid in the sec-ondary prevention of stroke. J Neurol Sci143:1–13, 1996

126. A randomised, blinded, trial of clopi-dogrel versus aspirin in patients at risk ofischaemic events (CAPRIE): CAPRIESteering Committee. Lancet 348:1329–1339, 1996

127. Bhatt DL, Fox KA, Hacke W, Berger PB,Black HR, Boden WE, Cacoub P, CohenEA, Creager MA, Easton JD, Flather MD,Haffner SM, Hamm CW, Hankey GJ,Johnston SC, Mak KH, Mas JL, Montale-scot G, Pearson TA, Steg PG, SteinhublSR, Weber MA, Brennan DM, Fabry-Rib-audo L, Booth J, Topol EJ: Clopidogreland aspirin versus aspirin alone for theprevention of atherothrombotic events.N Engl J Med 354:1706–1717, 2006

128. Diener HC, Bogousslavsky J, Brass LM,Cimminiello C, Csiba L, Kaste M, LeysD, Matias-Guiu J, Rupprecht HJ: Aspirinand clopidogrel compared with clopi-

dogrel alone after recent ischaemicstroke or transient ischaemic attack inhigh-risk patients (MATCH): random-ised, double-blind, placebo-controlledtrial. Lancet 364:331–337, 2004

129. Wagenknecht LE, D’Agostino R Jr, Sav-age PJ, O’Leary DH, Saad MF, HaffnerSM: Duration of diabetes and carotidwall thickness: the Insulin ResistanceAtherosclerosis Study (IRAS). Stroke 28:999–1005, 1997

130. Brohall G, Oden A, Fagerberg B: Carotidartery intima-media thickness in pa-tients with type 2 diabetes mellitus andimpaired glucose tolerance: a systematicreview. Diabet Med 23:609–616, 2006

131. Matsumoto K, Sera Y, Nakamura H,Ueki Y, Miyake S: Correlation betweencommon carotid arterial wall thicknessand ischemic stroke in patients with type2 diabetes mellitus. Metabolism 51:244–247, 2002

132. Chlumsky J, Charvat J, Empana JP,Zureik M, Gariepy J, Courbon D, Dart-igues JF, Ritchie K, Tzourio C, Alpero-vitch A, Ducimetiere P, Brohall G, OdenA, Fagerberg B, Tsivgoulis G, VemmosKN, Spengos K, Papamichael CM, Cim-boneriu A, Zis V, Zakopoulos N, Ma-vrikakis M, Matsumoto K, Sera Y,Nakamura H, Ueki Y, Miyake S: Echo-cardiography and carotid sonography indiabetic patients after cerebrovascularattacks. J Int Med Res 34:689–694, 2006

133. Weir CJ, Murray GD, Dyker AG, LeesKR: Is hyperglycaemia an independentpredictor of poor outcome after acutestroke? Results of a long-term follow upstudy. BMJ 314:1303–1306, 1997

134. Pulsinelli WA, Levy DE, Sigsbee B,Scherer P, Plum F: Increased damage af-ter ischemic stroke in patients with hy-perglycemia with or without establisheddiabetes mellitus. Am J Med 74:540–544, 1983

135. Toni D, Sacchetti ML, Argentino C, Gen-tile M, Cavalletti C, Frontoni M, FieschiC: Does hyperglycaemia play a role onthe outcome of acute ischaemic strokepatients? J Neurol 239:382–386, 1992

136. Moulin T, Tatu L, Crepin-Leblond T,Chavot D, Berges S, Rumbach T: The Be-sancon Stroke Registry: an acute strokeregistry of 2,500 consecutive patients.Eur Neurol 38:10–20, 1997

137. Williams LS, Rotich J, Qi R, Fineberg N,Espay A, Bruno A, Fineberg SE, TierneyWR: Effects of admission hyperglycemiaon mortality and costs in acute ischemicstroke. Neurology 59:67–71, 2002

138. Kronmal RA, Barzilay JI, Smith NL, PsatyBM, Kuller LH, Burke GL, Furberg C:Mortality in pharmacologically treatedolder adults with diabetes: the Cardio-vascular Health Study, 1989–2001.PLoS Med 3:e400, 2006

139. Bruno A, Levine SR, Frankel MR, BrottTG, Lin Y, Tilley BC, Lyden PD, Brod-

erick JP, Kwiatkowski TG, Fineberg SE:Admission glucose level and clinical out-comes in the NINDS rt-PA Stroke Trial.Neurology 59:669–674, 2002

140. Bruno A, Biller J, Adams HP Jr, ClarkeWR, Woolson RF, Williams LS, HansenMD: Acute blood glucose level and out-come from ischemic stroke: Trial ofORG 10172 in Acute Stroke Treatment(TOAST) Investigators. Neurology 52:280–284, 1999

141. Kiers L, Davis SM, Larkins R, Hopper J,Tress B, Rossiter SC, Carlin J, Ratnaike S:Stroke topography and outcome in rela-tion to hyperglycaemia and diabetes.J Neurol Neurosurg Psychiatry 55:263–270, 1992

142. Parsons MW, Barber PA, Desmond PM,Baird TA, Darby DG, Byrnes G, TressBM, Davis SM: Acute hyperglycemiaadversely affects stroke outcome: amagnetic resonance imaging and spec-troscopy study. Ann Neurol 52:20 –28,2002

143. Szczudlik A, Slowik A, Turaj W, Wyr-wicz-Petkow U, Pera J, Dziedzic T,Trabka-Janik E, Iskra T: Transient hy-perglycemia in ischemic stroke pa-tients. J Neurol Sci 189:105–111, 2001

144. Topic E, Pavlicek I, Brinar V, Korsic M:Glycosylated haemoglobin in clarifica-tion of the origin of hyperglycaemia inacute cerebrovascular accident. DiabetMed 6:12–15, 1989

145. de Falco FA, Sepe Visconti O, Fucci G,Caruso G: Correlation between hyper-glycemia and cerebral infarct size in pa-tients with stroke: a clinical and X-raycomputed tomography study in 104 pa-tients. Schweiz Arch Neurol Psychiatr 144:233–239, 1993

146. Kushner M, Nencini P, Reivich M, RangoM, Jamieson D, Fazekas F, ZimmermanR, Chawluk J, Alavi A, Alves W: Relationof hyperglycemia early in ischemic braininfarction to cerebral anatomy, metabo-lism, and clinical outcome. Ann Neurol28:129–135, 1990

147. Baird TA, Parsons MW, Phanh T,Butcher KS, Desmond PM, Tress BM,Colman PG, Chambers BR, Davis SM:Persistent poststroke hyperglycemia isindependently associated with infarctexpansion and worse clinical outcome.Stroke 34:2208–2214, 2003

148. Ribo M, Molina C, Montaner J, RubieraM, Delgado-Mederos R, Arenillas JF,Quintana M, Alvarez-Sabin J: Acute hy-perglycemia state is associated withlower tPA-induced recanalization ratesin stroke patients. Stroke 36:1705–1709,2005

149. Gentile NT, Seftchick MW, Huynh T,Kruus LK, Gaughan J: Decreased mortal-ity by normalizing blood glucose afteracute ischemic stroke. Acad Emerg Med13:174–180, 2006

150. Hillen T, Coshall C, Tilling K, Rudd AG,

Air and Kissela


McGovern R, Wolfe CD: Cause of strokerecurrence is multifactorial: patterns,risk factors, and outcomes of stroke re-currence in the South London StrokeRegister. Stroke 34:1457–1463, 2003

151. Johnston SC, Sidney S, Bernstein AL,Gress DR: A comparison of risk factorsfor recurrent TIA and stroke in patientsdiagnosed with TIA. Neurology 60:280–285, 2003

152. Staaf G, Lindgren A, Norrving B: Puremotor stroke from presumed lacunar in-farct: long-term prognosis for survivaland risk of recurrent stroke. Stroke 32:2592–2596, 2001

153. Eriksson SE, Olsson JE: Survival and re-current strokes in patients with differentsubtypes of stroke: a fourteen-year fol-low-up study. Cerebrovasc Dis 12:171–180, 2001

154. Hankey GJ, Jamrozik K, Broadhurst RJ,Forbes S, Burvill PW, Anderson CS,Stewart-Wynne EG: Long-term risk offirst recurrent stroke in the Perth Com-munity Stroke Study. Stroke 29:2491–2500, 1998

155. Petty GW, Brown RD Jr, Whisnant JP,Sicks JD, O’Fallon WM, Wiebers DO:Survival and recurrence after first cere-bral infarction: a population-basedstudy in Rochester, Minnesota, 1975through 1989. Neurology 50:208–216,

1998156. Alter M, Sobel E, McCoy RL, Francis ME,

Davanipour Z, Shofer F, Levitt LP, Mee-han EF: Stroke in Lehigh Valley: risk fac-tors for recurrent stroke. Neurology 37:503–507, 1987

157. Hier DB, Foulkes MA, SwiontoniowskiM, Sacco RL, Gorelick PB, Mohr JP, PriceTR, Wolf PA: Stroke recurrence within 2years after ischemic infarction. Stroke22:155–161, 1991

158. Sacco RL, Foulkes MA, Mohr JP, WolfPA, Hier DB, Price TR: Determinants ofearly recurrence of cerebral infarction:the Stroke Data Bank. Stroke 20:983–989, 1989

159. Sacco RL, Shi T, Zamanillo MC, Karg-man DE: Predictors of mortality and re-currence after hospitalized cerebralinfarction in an urban community: theNorthern Manhattan Stroke Study. Neu-rology 44:626–634, 1994

160. Soda T, Nakayasu H, Maeda M, KusumiM, Kowa H, Awaki E, Saito J, NakashimaK: Stroke recurrence within the first yearfollowing cerebral infarction: TottoriUniversity Lacunar Infarction PrognosisStudy (TULIPS). Acta Neurol Scand 110:343–349, 2004

161. Lee AH, Somerford PJ, Yau KK: Risk fac-tors for ischaemic stroke recurrence afterhospitalisation. Med J Aust 181:244–

246, 2004162. Berthet K, Neal BC, Chalmers JP, Mac-

Mahon SW, Bousser MG, Colman SA,Woodward M: Reductions in the risks ofrecurrent stroke in patients with andwithout diabetes: the PROGRESS Trial.Blood Press 13:7–13, 2004

163. Alter M, Lai SM, Friday G, Singh V, Ku-mar VM, Sobel E: Stroke recurrence indiabetics: does control of blood glucosereduce risk? Stroke 28:1153–1157, 1997

164. Jorgensen HS, Nakayama H, Reith J,Raaschou HO, Olsen TS: Stroke recur-rence: predictors, severity, and prog-nosis: the Copenhagen Stroke Study.Neurology 48:891– 895, 1997

165. American Diabetes Association/Ameri-can College of Cardiology: The Diabetes-Heart Disease Link: Surveying Attitudes,Knowledge and Risk. Report of commis-sioned RoperASW Survey, conductedAug-Oct 2001. Released Feb 19, 2002.The Make the Link! Initiative. www.diabetes.org/makethelink. Accessed 10April 2007

166. Bhatt DL, Steg PG, Ohman EM, HirschAT, Ikeda Y, Mas JL, Goto S, Liau CS,Richard AJ, Rother J, Wilson PW: Inter-national prevalence, recognition, andtreatment of cardiovascular risk factorsin outpatients with atherothrombosis.JAMA 295:180–189, 2006



Lessons From the Avandia ControversyA new paradigm for the development of drugs to treat type 2 diabetes

ROBERT I. MISBIN, MD

Sulfonylureas used to be the only oralagents available in the U.S. to treatpatients with type 2 diabetes. The

biguanide phenformin had been usedwidely in the 1950s and 1960s but wasremoved from the market in 1977 be-cause of lactic acidosis (1). Other bigua-nides, including metformin, were felt tobe safer and had been used extensivelyelsewhere in the world. But a biguanidedid not become available in the U.S. until1995 (2).

The year 1995 marked a sea change inavailablilty of drugs to treat type 2 diabe-tes. The approval of metformin, a bigua-nide, was followed by the approval ofacarbose, a galactosidase inhibitor that,like metformin, had been widely used inEurope. A new drug application for acar-bose had been rejected by the Food andDrug Administration (FDA) 3 years ear-lier, but in 1995 the regulatory environ-ment had become friendlier, owing inlarge part to the finding of the DiabetesControl and Complications Trial (DCCT)that aggressive treatment of hyperglyce-mia prevented or delayed complicationsof diabetes (3). Approval of the thiazo-lidinedione (TZD) troglitazone and thenon-sulfonylurea insulin secretagog repa-glinide followed soon after, along with ad-ditional members of the newly approvedclasses.

In 2005, the uninterrupted succes-sion of new drug approvals came to anend. Muraglitazar was a dual peroxisomeproliferator–activated receptor agonist. Itwas different chemically from the TZDsand fibrates but was believed to combine

the desirable pharmacological effects ofboth. Muraglitazar was highly effective inlowering A1C and had favorable effectson serum lipids. But contrary to what onemight expect, muraglitazar appeared toincrease the risk of adverse events relatedto cardiac ischemia (4). This might havebeen considered an isolated event were itnot for the recent brouhaha that rosiglita-zone (Avandia) also increased the risk ofcardiac ischemia.

The Avandia affair began with a meta-analysis showing that rosiglitazone in-creased the risk of myocardial infarction.Its release by the New England Journal ofMedicine online on 21 May 2007 with anaccompanying editorial (5,6) generated ajournalistic blitzkrieg against rosiglita-zone and the FDA (7). A congressionalcommittee hearing took place even beforethe article appeared in print. But an FDAadvisory committee concluded by a voteof 22 to 1 on 30 July 2007 that the evi-dence against rosiglitazone was insuffi-cient to recommend that it be withdrawnfrom the market.

The Avandia affair undermined theconfidence that patients have in the drugsthey take and in the physicians who pre-scribe those drugs. It cast further doubton the FDA’s ability to protect patientsfrom harm. One hopes that the delibera-tions of the advisory committee will putthe issues into perspective. But the FDAshould not ignore the perceived short-comings in the regulatory process that al-lowed the Avandia affair to get out ofhand.

Even before the Avandia affair, Dr.

David Nathan had expressed skepticismabout the recent approval of sitagliptin,apparently unimpressed by the extensivebody of data that was available on theFDA Web site (8). What the FDA shouldtake away from Dr. Nathan’s criticism andfrom the Avandia affair is that the time hascome to reassess what should be expectedof a new drug to treat diabetes. That a newdrug is more effective than placebo inlowering glucose levels is no longerenough. New drugs should be tested incomparison with other antidiabetesagents that are already in use (9). A planshould be in place at the time of approvalthat will determine what benefit and harmcan be expected from chronic use (10). Inthe following sections, I highlight prob-lems with the approval process that areillustrated by Avandia and suggest a newparadigm for the development of drugs totreat type 2 diabetes.

Lessons from the Avandiacontroversy

Reducing A1C may not be enough. TheDCCT and UKPDS (UK Prospective Dia-betes Study) showed that intensive treat-ment to lower A1C reduced the risk ofnephropathy, retinopathy, and neuropa-thy in patients with type 1 (3) and type 2diabetes (11,12). Clinical benefit has alsobeen reported for acarbose (13,14). Basedon the DCCT, the FDA accepted A1C as asurrogate marker for approval of newdrugs to treat diabetes. The standards ofapproval were discussed at an FDA advi-sory committee meeting in March 1998.Although a guidance to the industry forthe development of drugs to treat diabeteswas never issued by the FDA, the mainpoints of the draft guidance discussed bythe advisory committee were largely in-corporated into the standards used by theEuropean Agency for the Evaluation ofMedicinal Products (EMEA), which havebeen posted on the EMEA Web site since2002.

In contrast to older agents, use ofTZDs has not been shown to decrease therisk of the complications of diabetes.Clinical benefit was presumed to follow asa consequence of reduction in A1C. In theabsence of direct evidence of benefit, the

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

From the Division of Endocrinology and Metabolism, Food and Drug Administration, Silver Spring, Mary-land.

Address correspondence and reprint requests to Robert I. Misbin, MD, 711 Schindler Dr., Silver Spring,MD 20903. E-mail: [email protected].

Received and accepted for publication 1 October 2007.The opinions expressed in this article represent those of the author and do not necessarily reflect the

official position of the Food and Drug Administration.Abbreviations: ADOPT, A Diabetes Outcome Progression Trial; DCCT, Diabetes Control and Compli-

cations Trial; FDA, Food and Drug Administration; TZD, thiazolidinedione.A table elsewhere in this issue shows conventional and Systeme International (SI) units and conversion

factors for many substances.DOI: 10.2337/dc07-1908© 2007 by the American Diabetes Association.The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby


R e v i e w s / C o m m e n t a r i e s / A D A S t a t e m e n t sC O M M E N T A R Y


report alleging increased risk of myocar-dial ischemia with rosiglitazone was par-ticularly unsettling.Outcome trials need to be considered.The anomaly that muraglitazar and ros-iglitazone decreased glucose levels butappeared to increase the risk of cardiacischemia has led Psaty and Furburg (6) tocall for large, long-term randomized clin-ical trials to be completed as soon as pos-sible after approval of a new antidiabetesagent to identify the health benefits andrisk of the new treatment. These authorshave previously cautioned against the un-critical acceptance of surrogate end points(15) and have recently pointed out thatthe American Diabetes Association ac-knowledges that lowering A1C to preventmacrovascular disease is based on epide-miological studies rather than controlledclinical trials (16).

Although I recognize the desirabilityof long-term outcome trials, I have littledoubt that hyperglycemia per se is harm-ful. Requiring that an outcomes trial becompleted before approval would delaythe approval of new drugs. In my judg-ment, approval of drugs to treat type 2diabetes should continue to be based onchange in A1C, but a dedicated safety trialshould be completed before approval.Additional safety data should accrue fromextension of pivotal trials for 2 years be-yond the date of approval.

It may be appropriate for the FDA torequire a commitment to begin an out-comes trial as a condition of approval. Butthis requirement should be driven bysafety concerns related to the drug inquestion. It is unreasonable to expect thata single drug manufacturer should be ex-pected to bear the burden of answeringfundamental questions such as the natureof the relationship between diabetes andheart disease. The ACCORD (Action toControl Cardiovascular Risk in Diabetes)trial, funded by the National Institutes ofHealth, is attempting to address suchissues.

Performance of a successful outcomestrial presupposes the knowledge of whatoutcomes need to be measured, what thepopulation of interest is, how frequentlythose outcomes occur in the populationto be tested, and what the appropriatecomparators are. There are two basic ap-proaches to this type of trial. The manu-facturers of pioglitazone and rosiglitazoneeach employed different approaches tomeasuring long-term outcomes. Neitherhas been successful (17,18).

One approach is a placebo-controlled

trial in which the new agent is added tobackground therapy. The problem withthis design is that any benefit observedwith the new drug can be attributed tobetter control of hyperglycemia per serather than to a specific action of the newdrug. Thus, the PROactive (PROspectivepioglitAzone Clinical Trial in macroVas-cular Events) study largely confirmed thefinding from the UK PDS that loweringblood pressure and glucose levels re-duced the risk of complications of diabe-tes (19). An alternative approach is tocompare two treatment regimens be-lieved to have equivalent effectiveness forcontrol of hyperglycemia. A problem hereis that there is no gold standard for com-parison. In the RECORD (RosiglitazoneEvaluated for Cardiac Outcomes and Reg-ulation of Glycaemia in Diabetes) trial,rosiglitazone plus metformin was com-pared with a sulfonylurea plus met-formin. But there is no body of evidencethat sulfonylureas plus metformin reducecardiovascular end points. Furthermore,good patient care can be expected to re-duce statistical power because the adverseevent of interest is less likely to occur thanpreviously thought.

As the medical officer at the FDA whoinitially reviewed the Avandia applica-tion, I recommended that a postmarket-ing safety trial be a condition of approval.This recommendation was based on animbalance of cardiac ischemic events,weight gain, and lipid alterations in con-trolled trials of 6–12 months’ duration(20). That a safety trial was not performedwas noted by Congressman Waxman tobe a failure of FDA management (21). Onthe other hand, the postmarketing trialthat was performed, A Diabetes OutcomeProgression Trial (ADOPT), has provideduseful safety data even though its primaryobjective was to assess durability of effi-cacy. Troglitazone had been removedfrom the market because of an unaccept-ably high risk of liver failure (22). ButADOPT showed that chronic use of ros-iglitazone was safe to the liver. Of partic-ular interest is the finding from ADOPTthat rosiglitazone increases the risk offracture in postmenopausal women (23).The same was found for pioglitazone inPROactive, a long-term study designed toexamine cardiac effects (24). The in-creased risk of fracture was unexpectedand illustrates that a postmarketing safetystudy should cast a broad net.Combination therapy trials should bereevaluated. Troglitazone was initiallyapproved to be used in combination with

insulin. Indications for monotherapy anduse with sulfonylureas followed. Troglita-zone was never labeled to be added tometformin monotherapy. By contrast, theinitial approval for rosiglitazone was foradd-on to metformin monotherapy. Thus,the path chosen for the development of ros-iglitazone seems largely to have been drivenby a desire to fulfill a void in the market(TZD plus metformin) rather than by differ-ences in pharmacology between troglita-zone and rosiglitazone.

That sponsors seek to develop newdrugs to fill a marketing niche can bepartially attributed to the FDA. Accord-ing to the standards currently employed,clinical trials are needed for each of thesituations in which the drugs will be used:monotherapy and combinations withmetformin, sulfonylureas, insulin, etc.This approach needs to be reconsidered.There are no examples of approved drugsthat were effective as monotherapy or incombination with metformin but werenot effective in combination with otheragents. Thus, requiring efficacy trials foreach situation appears to be unnecessary.By contrast, safety issues have emerged insome situations but not others. For exam-ple, congestive heart failure emerged as aproblem in the trial of rosiglitazone withinsulin but not in trials of rosiglitazonemonotherapy. It should also be noted thatit is problematic to evaluate the efficacy ofa new drug in insulin-treated patients be-cause of the need to adjust the dose ofinsulin based on changes in glycemia. Forthese reasons, combination trials with in-sulin should be structured to evaluatesafety.

New paradigm for the developmentof drugs to treat type 2 diabetes

Trial design. Approval of new drugs totreat type 2 diabetes should recognize therecommendations of the American Diabe-tes Association that patients with type 2diabetes should generally be treated withmetformin at the time of diagnosis (25).Implementation of this recommendationwill make it difficult to conduct placebo-controlled monotherapy trials. Drug-naıve patients will become increasinglyscarce and ethical considerations preventthem from remaining untreated. For thesereasons, the period of placebo-only treat-ment should be 8–12 weeks (Table 1,trial A). The pivotal monotherapy trialshould be a 6- to 12-month comparisonof the new drug, drug X, to metformin(trial B). A second pivotal trial (trial C)

Lessons from Avandia: a new paradigm


should compare drug X to placebo in pa-tients who are receiving metformin butwhose hyperglycemia is not adequatelycontrolled.

A safety trial that does not exclude pa-tients who may be at risk of adverse eventsshould be completed before approval. Asillustrated by the trials with rosiglitazone,cardiac adverse events are more likely to befound in insulin-treated patients. Trial D inthe proposed paradigm compares one or

more doses of drug X (including the maxi-mal dose to be marketed) with placebo inpatients who are receiving insulin with orwithout oral agents. The insulin doseshould be adjusted to account for the activ-ity of drug X such that there be little differ-ence between the arms with respect tochange in glucose levels. Under these cir-cumstances, safety issues that emerge canbe attributed to drug X and not to differ-ences in glycemic control. The primary

end point should be reports of serious ad-verse events. A secondary end point,change in insulin dose, should assess ef-ficacy. Subset analyses should be basedon background oral antidiabetes treat-ment. In some circumstances, enrich-ment with patients using TZDs or otheroral agents may be appropriate.

A safety trial has not previously beenrequired before approval of oral antidia-betes agents currently marketed in theU.S. However, concerns about vascularevents in postpartum women taking bro-mocriptine led the FDA to require that asafety trial be completed before approvalof Cycloset. Details of this trial have beenpublished (26).

Trials A–D should provide the dataon efficacy and safety to form the basis fora new drug application. Extensions ofthese trials should continue during thetime it takes the FDA to review the appli-cation (10 months for a standard reviewand 6 months for a priority review). Asalready required, a safety update on on-going trials should be submitted to theFDA before regulatory action. Extensionsof these trials for an additional 2 yearsshould be required for approval.

The extension trials will provide long-term comparisons of drug X to metformin(trials A and B), metformin plus drug Xversus metformin plus standard therapy(trial C), and insulin plus drug X versusinsulin plus standard therapy (trial D).New trials may be required to begin afterapproval to examine safety issues thatemerged during the FDA review.Approval. Drug X should be approved ifits efficacy is superior to that of placebo(trials A and C), noninferior to metformin(trial B), noninferior to standard therapywith respect to serious adverse events(trial D), and does not have other serioussafety problems. It might still be appro-priate to approve drug X even if it is lesseffective than metformin, particularly if itrepresents a novel mechanism of actionand raises no safety concerns.

The FDA does not have the authorityto require that a new drug be superior toexisting drugs. On the other hand, theFDA is not required to have proof that anew drug is unsafe to deny approval andshould set a high standard for drugs thatoffer no advantage over existing therapy.Even for a novel agent, a few cases of a rarebut life-threatening event, such as agran-ulocytosis, may be sufficient to preventapproval.Labeling. New drugs should be labeledfor treatment of hyperglycemia in patients

Table 1—Paradigm for development of new drugs to treat type 2 diabetes

Suggested pivotal trialsTrial A 8- to 12-week monotherapy trial in treatment-naïve

patients to study effects of three of more doses ofdrug X versus placebo.

Primary end point can be A1C, fasting,postprandial, or mean daily glucose dependingon the circumstances.

Trial B 6- to 12-month active comparator trial in patientsto study effects of two or more doses of drug Xversus metformin.

Primary end point is A1C. Events related to cardiacischemia should be adjudicated.

Trial C 6- to 12-month add-on trial in patients onmetformin to study effects of two or more dosesof drug X versus placebo.

Primary end point is A1C.Trial D 6- to 12-month add-on trial in patients on insulin

to study effects of one or more doses of new drugversus placebo. Dose of insulin is titratedaccording to the standard of practice.

Primary end point is reporting of serious adverseevents. Events related to cardiac ischemia shouldbe adjudicated.

Suggested extensions to pivotal trials(at least 2 years beyond approval)Trial A Patients on drug X are continued on drug X.

Patients on placebo are started on metformin.Rescue due to hyperglycemia with insulin

secretagogue, insulin sensitizer, or insulin asappropriate depending on the nature of drug X.Events related to cardiac ischemia should beadjudicated.

Trial B Patients on drug X are continued on drug X.Patients on metformin are continued onmetformin.

Rescue with insulin secretagogue, insulin sensitizer,or insulin as appropriate depending on thenature of drug X.

Trial C Patients on drug X are continued on drug X.Patients on placebo are started on insulinsecretagog or insulin sensitizer as appropriatedepending on the nature of drug X.

Rescue with insulin.Trial D Patients on drug X are continued on drug X.

Patients on placebo are continued on placebo.Insulin is given as needed for glycemic control.

Patients who cannot achieve adequate control arewithdrawn.

Misbin


with type 2 diabetes, but no claim for re-ducing the risk of complications shouldbe allowed unless directly supported bydata. There should be no distinction be-tween first- and second-line therapy orbetween monotherapy and combinationtherapy. Data from pivotal trials shouldbe shown in the label to provide insightinto what clinical effects can be expectedunder what conditions. Gaps, uncertain-ties, or inconsistencies should be ac-knowledged, such as lack of informationin certain populations. In some circum-stances, sponsors should be required toperform postmarketing trials to addressremaining issues. Failure to completethese trials should be posted on the FDAWeb site. From the clinical review andaction letter posted by the FDA on its Website, professional organizations can de-velop recommendations about how a newdrug should be used in relation to otherapproved agents.

Looking aheadTable 1 is a summary of a suggested par-adigm for clinical trials to develop newdrugs for type 2 diabetes. With betterknowledge and experience, this paradigmmay need to change. But the goal shouldremain the same. New drugs should con-tinue to be developed to reduce the globalburden of diabetes and its complications.Regulators should set high standards butshould also be pragmatic. Fear of uncer-tainty should not be allowed to stifleinnovation.

References1. Misbin RI: Phenformin associated lactic

acidosis: pathogenisis and treatment. AnnIntern Med 87:591–595, 1977

2. Misbin RI: The phantom of lactic acidosisdue to metformin in patients with type 2diabetes. Diabetes Care 27:1791–1793,2004

3. Diabetes Control and Complications TrialResearch Group: The effect of intensivetreatment of diabetes on the development

and progression of long-term complica-tions of insulin-dependent diabetes mel-litus. N Engl J Med 329:977–986, 1993

4. Golden J: Muraglitazar safety analysis.Presented at the Food and Drug Adminis-tration Metabolic and Endocrine AdvisoryCommittee Meeting, 9 September 2005,Silver Spring, Maryland

5. Nissen SE, Wolski K: Effect of rosiglita-zone on the risk of myocardial infarctionand death from cardiovascular causes.N Engl J Med 356:2467–2471, 2007

6. Psaty BM, Furberg CD: Rosiglitazone andcardiovascular risk. N Engl J Med 356:2522–2524, 2007

7. Bloomgarden ZT: The Avandia debate.Diabetes Care 30:2401–2410, 2007

8. Nathan DM: Finding new treatments ofdiabetes: how many, how fast. . . howgood? N Engl J Med 356:437–440, 2007

9. Misbin RI: A possible drug fix? Washing-ton Post, 24 Aug 1998: Sect. A, p. 18

10. Misbin RI: Evaluating the safety of diabe-tes drugs: perspective of a Food and DrugAdministration insider. Diabetes Care 28:2573–2576, 2005

11. UK Prospective Diabetes Study (UKPDS)Group: Intensive blood glucose controlwith sulfonylureas or insulin comparedwith conventional treatment and risks ofcomplications in patients with type 2 di-abetes (UKPDS 33). Lancet 352:837–853,1998

12. UK Prospective Diabetes Study (UKPDS)Group: Intensive blood glucose controlwith metformin complications in over-weight patients with type 2 diabetes (UK-PDS 34). Lancet 352:854–865, 1998

13. Chiasson JL, Josse RG, Gomis R, HanefeldM, Karasik A, Laakso M; STOP-NIDDMTrial Research Group: Acarbose treatmentand the risk of cardiovascular disease andhypertension in patients with impairedglucose tolerance. JAMA 290:486–494,2003

14. Hanefield M, Cagatay M, Petrowitsch T,Neuser D, Petzinna D, Rupp M: Acarbosereduces the risk of myocardial infarctionin type 2 diabetic patients: meta-analysisof seven long-term studies. Eur Heart J 25:10–16, 2004

15. Psaty BM, Weiss NS, Furburg CSD, Ko-epsell TD, Siscovick DS, Rosendaal FR,Smith NL, Heckbert SR, Kaplan RC, Lin

D, Fleming TR, Wagner EH: Surrogateend points, health outcomes, and thedrug-approval process for the treatmentof risk factors for cardiovascular disease.JAMA 282:786–790, 1999

16. Psaty BM, Furberg CD: Rosiglitazone andthe FDA. N Engl J Med. 29 August 2007[Epub] (DOI: 10.1056/NEJMc076347)

17. Nathan DM: Rosiglitazone and cardiotox-icity: weighing the evidence. N Engl J Med357:64–66, 2007

18. Skyler JS: PROactive: a sad tale of inap-propriate analysis and unjustified inter-pretation. Clinical Diabetes 24:63–64,2006

19. Fonseca V, Jawa A, Asnani S: Commen-tary: the PROactive Study: the glass is halffull. J Clin Endocrinol Metab 91:25–27,2006

20. Food and Drug Administration: MedicalOfficers Review of Avandia. Rockville, MD,Food and Drug Administration, 1999 (ap-plication no. 021071)

21. Committee on Oversight and Govern-ment Reform: Hearing on FDA’s Role inEvaluating Safety of Avandia [articleonline], 2007. Available from http://oversight.house.gov/story.asp?ID�1325.Accessed 6 June 2007

22. Misbin RI: Troglitazone: associated he-patic failure. Ann Intern Med 130:330,1999

23. Kahn SE, Haffner SM, Heise MA HermanWH, Holman RR, Jones NP, Kravatz BG,Lachman JM, O’Neill MC, Zinman B, Vib-erti G: Glycemic durability of rosiglita-zone, metformin or glyburide mono-therapy. N Engl J Med 355:2427–2443,2006

24. Food and Drug Administration: MedicalOfficers Review of Actos. Rockville, MD,Food and Drug Administration, 2007 (ap-plication no. 021073)

25. American Diabetes Association: Stan-dards of medical care in diabetes—2007(Position Statement). Diabetes Care 30(Suppl. 1):S4–S41, 2007

26. Scranton RE, Gaziano JM, Rutty D, Cin-cotta MA: A randomized, double-blind,placebo-controlled trial of safety and tol-erability during treatment of type 2 diabe-tes with either Cycloset or placebo. BMCEndocr Disord 7:3, 2007

Lessons from Avandia: a new paradigm


Diabetes and ObesityPart 1

ZACHARY T. BLOOMGARDEN, MD

Perspectives on the News commentaries arenow part of a new, free monthly CME activ-ity. The Mount Sinai School of Medicine,New York, New York, is designating this ac-tivity for 2.0 AMA PRA Category 1 credits. Ifyou wish to participate, review this articleand visit www.diabetes.procampus.net tocomplete a posttest and receive a certificate.The Mount Sinai School of Medicine is ac-credited by the Accreditation Council forContinuing Medical Education (ACCME) toprovide continuing medical education forphysicians.

This is the fourth in a series of articlesbased on presentations at the Amer-ican Diabetes Association’s 67th Sci-

entific Sessions, 22–26 June 2007 inChicago, discussing aspects of the interre-lationships between diabetes and obesity.

ObesityGerald Reaven (Stanford, CA) argued thatobesity is not synonymous with insulinresistance. Measuring the steady-stateplasma glucose (SSPG) during infusion ofglucose and insulin to characterize insulinsensitivity, there is a continuous distribu-tion of levels with a six- to eightfold vari-ation from least to most insulin sensitivein the apparently normal population.SSPG correlates with both waist circum-ference and BMI in men and women but,Reaven noted, with “enormous variabil-ity” (1), only explaining �25% of thevariability in this measure. BMI and waistare similar in their power to identify indi-viduals with abnormal SSPG, as well as inpredicting abnormalities of glucose, tri-glyceride, HDL cholesterol, and otherparameters associated with insulin resis-tance. A study of individuals of Malay,Chinese, and Indian ethnicity showedthat metabolic syndrome frequently oc-curs without satisfying criteria for abdom-

inal obesity (2), with metabolic syndromesimilarly only being moderately associ-ated with directly measured visceral fat.Furthermore, not all obese individualshave insulin resistance, and those obeseindividuals showing metabolic benefits ofweight loss belong to the insulin-resistantsubset in the lowest tertile of insulin sen-sitivity. Comparing adipocyte cell sizedistributions of insulin-resistant versusinsulin-sensitive obese individuals, theformer have a greater proportion of smalladipocytes, contradicting earlier conceptsthat large fat cells were metabolically lessefficient. Preadipocytes from insulin-resistant obese individuals appear lesscapable of differentiating into mature adi-pocytes, perhaps explaining this finding.

In a study presented at the AmericanDiabetes Association’s 67th annual meet-ing, relevant to the concept of adipocytesubpopulations contributing to insulinresistance, Rittig et al. (abstract 18) mea-sured perivascular brachial artery fat withmagnetic resonance imaging, finding asignificant correlation of this adipocytedepot with reduced insulin sensitivity,measured from euglycemic-hyperinsu-linemic glucose clamp study (abstractnumbers refer to the American DiabetesAssociation Scientific Sessions, Diabetes56 [Suppl. 1], 2007).

Steven Schneider (New Brunswick,NJ) presented a series of observationscomplementary to Reaven’s discussion,pertaining to cardiometabolic risk in thenonoverweight insulin-resistant patient, atype he termed metabolically obese nor-mal weight (MONW) (3). Consideringsuch individuals to be no more than 10%over ideal body weight and to exhibit hy-perinsulinemia, he suggested other char-acteristics to include increased fat cell size(perhaps contradicting Reaven’s observa-

tion that reduced adipocyte size may berelated to insulin resistance), increasedblood pressure, and increased triglyceridelevels. Such individuals often are off-spring of type 2 diabetic parents, them-selves developing type 2 diabetes atrelatively young ages, having history ofmyocardial infarction and of cholesterolcholelithiasis. Criteria for the MONWstate are similar to those for metabolicsyndrome (4), including hyperinsuline-mic individuals with normal weight andmultiple cardiovascular disease (CVD)risk factors (5). An alternative approach isto identify nonobese hypertensive indi-viduals, recognizing this to be a groupcharacterized by increased insulin and tri-glyceride levels and by decreased insulinsensitivity (6). In the U.S. National Healthand Nutrition Surveys, MONW consti-tute a large number of at-risk individualsin the U.S. population (7). A Canadianstudy evaluating normal-weight individ-uals with features of insulin resistanceshowed a tripling in risk of CVD (8).

Schneider asked how knowledge ofthe existence of this group should influ-ence our thinking, specifically addressingthe usefulness of relying on simple mea-sures of body weight, as the MONW con-cept implies that a large number ofnormal-weight individuals would benefitfrom interventions now thought appro-priate for obese individuals. He discussedthe usefulness of measures of adipose tis-sue other than that of total fat mass andthe question of whether measures of insu-lin resistance and hyperinsulinemiawould be useful in ascertainment of theseabnormalities.

Clinically, we are not readily able toassess adiposity. The 75-kg man at age 53years typically has 7 kg more fat and lesslean mass than he had at the same weightat age 25 years, without apparent differ-ence in physiognomy. Nondiabetic,nonobese offspring of type 2 diabetic par-ents will be found to have increased fatmass despite normal BMI (9). The con-cept of abnormal fat distribution dates toVague’s differentiation between benignand metabolic obesity, with relatively fewmetabolic abnormalities in the formergroup, the latter exhibiting the pattern ofincreased abdominal fat. There is clearly

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

Zachary T. Bloomgarden, MD, is a practicing endocrinologist in New York, New York, and is affiliated withthe Division of Endocrinology, Mount Sinai School of Medicine, New York, New York.

Abbreviations: CCK, cholecystokinin; CVD, cardiovascular disease; DPP, Diabetes Prevention Program;ER, endoplasmic reticulum; I�B, inhibitor of �B kinase; IRS, insulin receptor substrate; JNK, c-Jun NH2-terminal kinase; MONW, metabolically obese normal weight; SSPG, steady-state plasma glucose; TNF,tumor necrosis factor; UPR, unfolded protein response.

DOI: 10.2337/dc07-zb12© 2007 by the American Diabetes Association.

R e v i e w s / C o m m e n t a r i e s / A D A S t a t e m e n t sP E R S P E C T I V E S O N T H E N E W S


an association between weight gain inadulthood and increased CVD and diabe-tes risks. Furthermore, regardless of thedegree of obesity, the presence of greaterabdominal fat is associated with greaterreduction in insulin sensitivity (10).These observations have been confirmedin children (11), Asian Indians (12), andJapanese (13). Although fasting insulinpredicts visceral obesity (14), Schneidercommented that we cannot be certain thatabdominal fat distribution is not a markerrather than a mediator of reduced insulinsensitivity. The concept of fat repartition-ing is based on the notion that the pres-ence of fat stores in liver and muscle isassociated with metabolic abnormality.Lean offspring of patients with type 2 di-abetes have increased intramuscular lipidconcentration (15), and associations havebeen demonstrated between insulin resis-tance and increased intramyocellularlipid (16), as well as with increased he-patic insulin stores (17). A particular ex-ample exempl i fy ing the MONWsyndrome is HIV/protease inhibitor lipo-dystrophy, associated with increased in-tramyocellular and hepatic fat stores (18).Schneider pointed out that MONW al-ways appears to be associated with de-creased aerobic capacity (19), withoffspring and other first-degree relativesof type 2 diabetic individuals having de-creased aerobic capacity before develop-ment of diabetes (20,21). A possibility isthat, rather than this being caused by aphysically inactive lifestyle, poor aerobicexercise capacity might be associated withintrinsic abnormality of fat oxidation con-tributing to insulin resistance and, perhaps,to weight gain. There is intriguing evidenceof mitochondrial dysfunction in thin off-spring of type 2 diabetic parents (15).

Another relevant observation pertainsto abnormal androgen metabolism. Mildtestosterone deficiency has been demon-strated in type 2 diabetic patients, with anassociation with low testosterone and lowsex hormone binding globulin (22). Pop-ulation surveys have shown this hor-monal pattern to be a predictor ofmetabolic syndrome (23), particularly in-creasing the likelihood of metabolic syn-drome in nonobese individuals. Whetherhypogonadism causes insulin resistanceor certain effects of insulin resistance leadto abnormality of circulating androgens isnot known. Certainly, Schneider con-cluded, it is important to develop ap-proaches to the identification of MONWindividuals, who exhibit high CVD risk,which often is not well treated. Energy

restriction and increased physical activitymay be of therapeutic benefit in at least asubset of such individuals, with a numberof studies suggesting that exercise im-proves insulin action in nonobese, as wellas obese, individuals with and without di-abetes (24).

Steven Blair (Columbia, SC) contin-ued the line of observation, addressingthe question of whether fitness protectsagainst obesity. Sedentary lifestyle isstrongly associated with type 2 diabetes,although many existing studies have beenlimited in relying on self-reported physi-cal activity and diabetes. Blair reviewed aset of studies carried out over the pastthree decades, with maximal treadmill ex-ercise testing for assessment of aerobic ca-pacity. In a study of 7,442 men examinedon two occasions, who had normal elec-trocardiogram and glucose tolerance andwere without CVD, at 6 years’ follow-upthe risk factor–adjusted likelihood of im-paired fasting glucose doubled with lowfitness (25). Contrary to what one mightexpect, the number of obese individualsexhibiting acceptable physical fitness isnot inconsequential, one study showingthat 25% of women with BMI �36 kg/m2

were physically fit (26).The association between fitness and

glycemic abnormality was similar in over-weight and normal-weight groups. Blairreviewed further studies showing thatmetabolic syndrome incidence similarlyis inversely proportional to maximal ex-ercise capacity both in men and women.Among 1,200 individuals with impairedfasting glucose, fitness was associatedwith better outcome at follow-up, withsimilar studies in diabetic individualsshowing fitness to protect against CVDmortality (27). Lack of physical activityand the proxy measure—the number ofhours of television watching weekly—areassociated with worse CVD outcomes, al-though one cannot be certain that the lackof activity itself is the mediator of adverseoutcome. Mortality benefit is associatedwith greater physical fitness in diabeticindividuals both with normal weight andwith obesity (28). Taking into accountboth obesity and fitness, among fit indi-viduals, there was no increased mortalitywith obesity; similarly, among unfit indi-viduals, there was no increased mortalitywith obesity. “I’m not going to say ‘forgetBMI,’” Blair concluded, “but we do needto pay more attention to fitness.”

Daniel Porte (San Diego, CA) pre-sented further views of the relationshipbetween obesity and diabetes, noting that

the effect of insulin in the brain is to de-crease food intake, in a sense opposing itseffect in the periphery, where it leads tofat storage. “The same molecule that actsin the periphery,” he said, “is playing acounterregulatory role for the regulationof food intake in the CNS [central nervoussystem].” Comparing animals duringoverfeeding, underfeeding, and ad libfeeding, body weight shows strong corre-lation with insulin levels (29). After theover- or underfeeding periods in thisstudy, all animals had free access to food,and weight in all three groups rapidlyequalized, which suggests that weight isclosely regulated and that insulin mightbe a mediator of body weight stability.The hyperbolic relationship between in-sulin sensitivity and body weight in hu-mans (30) is, of course, characteristic of aregulatory signal. Such observations ledto the hypothesis of a feedback loop be-tween insulin and central effectors offeeding, in which a central insulin sensoracts to decrease food intake (31). Directintracerebroventricular administration ofinsulin in primates leads to rapid weightloss over a 20-day period (32), supportingthis concept.

Central regulation of food intake iscomplex. Studies of cholecystokinin(CCK) administration reveal suppressionof food ingestion with stable body weight(33). The compensation for CCK was anincrease in meal frequency to stabilize cal-orie intake. Intracerebroventricular insu-lin administration potentiates the effect ofCCK in decreasing food intake (34), sug-gesting that there are both short-termcontrols of food intake, influenced by var-ious environmental, learned, and biolog-ical factors, and long-term controls,largely determined by biological factorspromoting long-term stability of body fatmass, with the major candidates for thelatter signals being insulin and leptin.Many short-term signals arise in the gut,either as hormonal factors or being re-layed through vagal signals. Porte sug-gested, then, that there is a negativefeedback system combining central andperipheral control of food intake, whichmay be considered programmable ratherthan exhibiting a hard wired set point(35). The afferent signals are, in general,circulating factors, including insulin,with insulin crossing the blood-brainbarrier by a specific receptor-mediatedtransport system, as well as leptin, glu-cocorticoids, ghrelin, protein YY (3–36),glucagon-like peptide (GLP)-1, and amy-lin. These may be considered afferent sig-

Perspectives on the News


nals of body adiposity, which are alsoregulated by nonadiposity factors. Singlemeal sizes are regulated by vagal neuralinputs and by circulating hormones suchas ghrelin and CCK, arising from the gas-trointestinal tract, as well as what may betermed hedonic factors impacting the ce-rebral cortex. Central integration occursin the hypothalamus; via amines such asnorepinephrine, serotonin, dopamine,and neuropeptides Y; Agouti-related pro-tein; melanocyte-stimulating hormone;cocaine-amphetamine–related transcript;orexin; and other factors. Central effer-ents lead to changes in food intake andenergy expenditure. Obesity and ca-chexia, then, may be considered disordersof this regulatory system.

In animal models not expressingbrain insulin receptors, obesity develops,as well as disruption of fertility (36). Sim-ilarly, reduction of hypothalamic insulinreceptors increases food intake and fatmass in a dose-related fashion (37). Insu-lin activates the phosphatidylinositol3-kinase pathway in the hypothalamus asit does in other insulin-responsive tissues(38). Animals not expressing insulin re-ceptor substrate (IRS)-2 in the brain havea phenotype similar to that of those lack-ing the insulin receptor but, in addition,show an increase in circulating leptin lev-els (39). Porte suggested that leptin levelsprovide a measure of total fat mass, whileinsulin levels, increasing with the degreeof insulin resistance, furnish a measure ofvisceral fat mass, with complex overallbrain circuitry of insulin effect and that ofother satiety signals (35,40). There also isevidence of acute central insulin effects,with intracerebroventricular administra-tion of insulin antibodies increasing he-pat ic glucose product ion duringeuglycemic-hyperinsulinemic clamp,suggesting that part of the effect of insulinon the liver is centrally mediated (41).Porte summarized that an expanded viewof insulin resistance includes the conceptthat reduced brain action of insulin, aswell as of leptin, leads to positive energybalance causing obesity, while reducedinsulin action in the periphery leads tohyperglycemia via deficiency in glucoseuptake. There is fascinating evidence thatintranasal insulin administration maylead to weight loss under certain circum-stances (42), which may be interpreted toindicate enhanced brain uptake via thisroute. These considerations of the appe-tite-suppressing effect of insulin may ex-plain the apparently lesser degree ofweight gain when insulin detemir rather

than NPH is administered for glycemiccontrol (43). Detemir may have a prefer-ential effect on phosphorylation of hypo-thalamic IRS-2, perhaps caused byselective uptake of detemir into the brainwhen administered in a quantity produc-ing similar peripheral effect to that of hu-man insulin (44).

Inflammation: basis of obesity anddiabetesGokhan Hotamisligil (Boston, MA), receiv-ing the American Diabetes Association’sDistinguished Achievement Award, re-minded listeners that worldwide therewere 200 million individuals with obesityin 1995 and 300 million in 2000, withprojections for 500 million obese individ-uals in 2010, covering every area of theworld, from the U.S. to sub-Saharan Af-rica. Obesity is associated with insulin re-sistance, type 2 diabetes, fatty liver, CVD,hypertension, and dyslipidemia, as wellas with airway disorders, musculoskeletalconditions, neurodegenerative diseases,Alzheimer’s—now being considered“diabetes of the brain”—and with malig-nancy. An important underlying mecha-nism appears to be the inflammatoryrelationship between metabolism and obe-sity may be relevant. Insulin action in-volves a complex signaling pathway.Increased tumor necrosis factor (TNF)�expression by white adipose tissue de-rived both from mature adipocytes andfrom stromal vascular cells occurs in a va-riety of states of obesity and of diabetes,with neutralization of TNF� improvinginsulin sensitivity (45). Mice not express-ing TNF� are protected from obesity-induced insulin resistance, leading toimproved glucose tolerance (46), withTNF� signaling acting on IRS-1 to in-crease serine phosphorylation, antagoniz-ing the insulin receptor signaling effect ofincreased IRS-1 tyrosine phosphoryla-tion. This appears to be a mechanism ofaction of many inflammatory cytokines,as well as explaining some aspects of lip-id-induced inhibition of insulin action.There are multiple potential serine phos-phorylation sites, so that the identifica-tion of specific effects of inflammatorycytokines on IRS-1 is extraordinarily com-plex. c-Jun NH2-terminal kinases (JNKs)are intracellular mediators of inhibition ofinsulin action by TNF� (47). Another in-hibitory kinase involved in this pathway,downstream of TNF� signaling, is the in-hibitor of �B kinase (I�K), which plays arole in beneficial effects of aspirin (48).

Hotamisligil summarized, “The story

so far: IRS-1 serine phosphorylation oc-curs in insulin resistance and is critical.JNK regulation appears as mediator andtherefore is expected to be a mechanismof insulin resistance in obesity and in type2 diabetes.” JNK is markedly increased inobesity in muscle, liver, and adipose tis-sue (49). Obesity and type 2 diabetes leadto activation of the JNK1 isoform, a phe-nomenon not observed in an animalmodel deficient in this enzyme (50). Obe-sity is associated with a high degree offatty infiltration of the liver, which is to alarge extent blocked by absence of JNK1.JNK and I�K act, then, in the develop-ment of insulin resistance, although it isnot certain whether I�K acts in the samefashion on IRS-1. This explanation of mo-lecular mechanisms shows the centralrole of inflammation in insulin resistanceand suggests potential approaches totreatment. Furthermore, there is evidenceof a role of JNK in atherosclerosis, andthere is now evidence that mutations inJNK binding proteins can cause diabetes(51), with the potential that small mole-cules inhibiting JNK may lead to newtherapies for these processes (52).

JNK and I�K may themselves activatethe inflammatory cytokine response, po-tentially causing a “mini vicious loop.” Anunanswered question is, “How do we getto [over] production of inflammatory cy-tokines in the first place?” Proinflamma-tory effects of excessive levels of nutrients,perhaps acting via reactive oxygen speciesor by causing mitochondrial dysfunction,appears to be an important initiatingmechanism. Hotamisligil hypothesizedthat signals leading to alterations in JNK,transmittal of stress responses, or transla-tion of metabolic stresses into inflamma-tory pathways may all take place at theorganelle level and be communicated viaintracytoplasmic signaling pathways.Dysfunction of the endoplasmic reticu-lum (ER) appears to be critical formetabolic disease. The ER is a vast mem-branous network covering much of thecell, acting as the synthetic organelle ofcells, and plays a role in quality control byremoving misfolded proteins. Excessivedemand leads to the state termed “ERstress,” in which cellular responses re-ferred to as the “unfolded protein re-sponse” (UPR) occur, a program activatedby signals including excess protein loadand a variety of other alterations in nutri-ents, energy, or calcium flux. UPR occurswith ER sensing of pathogen stress aswell, suggesting that the ER may bethought of as an integrating system. UPR

Bloomgarden


involves nuclear transcription factorATF-6, which controls gene expression ofprotein chaperones, used by the ER tocompensate for stresses. Insulin responseelement-1 autophosphorylation activatesX-box binding protein-1, another factorcausing ER adaptation, and another ki-nase, PERK, leads to nuclear factor �Bactivation, and suppresses protein syn-thesis, a negative feedback loop benefi-cially reducing the ER stress response.Given this background, a critical observa-tion was the discovery that JNK is acti-vated in the setting of ER stress,suggesting a role of the ER in activation ofmetabolic stress, with obesity potentiallya condition of ER stress. XBP-1, in statesof obesity, is associated with increasedphosphorylation of JNK (53). ER stress,then, may lead to insulin response ele-ment-1– dependent JNK activation, re-sulting in insulin resistance, offering apotential route to the improvement of in-sulin sensitivity in type 2 diabetes, hepaticsteatosis, and atherosclerosis. Mutationsin UPR genes disrupt ER homeostasis andcause diabetes in animal models and inhumans, while pharmacological modifi-cation of ER responses with chemicalchaperones appears to represent an effec-tive approach in animal models of diabe-tes. Obesity-induced JNK activation canbe reduced to normal levels by adminis-tration of chemical chaperones, whichdecrease adipose tissue TNF�, interleu-kin-6, and other inflammatory mediatorlevels, reducing inflammation while notreducing fat mass.

If initiation of insulin resistance doestake place within the cell at the organellelevel, in addition to the ER, mitochondrialand/or nuclear abnormalities may be fur-ther causes. Hotamisligil discussed as-pects of the linkage between nutrition andmetabolism. Nutrient and pathogen-sensing pathways are integrated, as aremetabolic and inflammatory pathways.“The way we respond to bugs,” he said, “isnot all that different from the way we re-spond to food.” The fat body of the fruitfly, which corresponds to liver and hema-topoietic tissues in mammals, also showssimilarity to mammalian adipose tissue(54). The immune and metabolic sys-tems, then, may be considered to have“once [been] part of the same organ struc-ture . . . As such, these organs maintaintheir metabolic heritage.” Metabolic andinflammatory cell types appear in closeproximity both in adipose tissue and inliver. Under normal circumstances, anovert inflammatory response is not seen,

suggesting controlling mechanisms.Thus, an important question is not whatcauses obesity to be associated with in-flammation, but rather, what prevents aninflammatory response from occurringfollowing normal nutrient ingestion?

Molecular mediators are present inadipocytes and respond to nutritionalcues. STAMP2 was found from a genomescan, appearing to act as an insulin-sensitizing agent, showing linkage toprostate cancer and highly expressed inwhite adipose tissue. In normal lean ani-mals, STAMP2 is stimulated in visceraladipose tissue by feeding (55). In obeseanimals, this regulation by feeding is lost,with baseline levels increased in all tis-sues. The normal regulatory pathwaysmay then not be active in obesity, perhapsexplaining its association with inflamma-tion, as animals not expressing STAMP2show visceral adipose tissue macrophageinfiltration and increased expression ofinterleukin-6, eventually developing thephenotype of metabolic syndrome withinsulin resistance and glucose intoler-ance. In a variety of ways, then, nutrientand energy status is linked to health. Justas undernutrition is almost always associ-ated with immune system suppression,overnutrition is associated with immuneactivation, suggesting the importance ofoptimal nutrition. If such a nutritionalstatus is achievable, these mechanisms“could be exploited to assist” in reversingthe abnormalities associated with exces-sive nutrient intake.

Genetic aspects of obesityAt a symposium on genetic aspects of obe-sity, Alan Shuldiner (Baltimore, MD) dis-cussed findings from genetics studies ofthe Amish. He recalled Claude Bouchard’sconcepts of gene-nutrition and gene-physical activity interactions in the con-text of factors leading to development oftype 2 diabetes, from the viewpoint thatlifestyle/environmental factors lead tooutcomes modulated by genetic heteroge-neity, with unpredictable degrees of lin-earity or nonlinearity in responses tovarious genetic differences. As an exam-ple, he cited the findings of the Nurses’Health Study that polymorphisms of thealcohol dehydrogenase 3 gene modulatealcohol-induced increases in HDL choles-terol. Individuals having the �1/2 or �2/2genotype have slower alcohol metabolismand show greater degrees of increase inHDL with alcohol ingestion. Similarly,there must be genetic determinants ofweight gain. This was shown particularly

clearly in the overfeeding experiment car-ried out by Bouchard two decades ago.Twelve pairs of monozygotic twins werefollowed for 100 days while ingesting 840calories daily in excess of their weightmaintenance requirement, calculatedduring a 2-week basal period under sed-entary circumstances (56). Weight gain,fat mass, and abdominal visceral fat oftwin pairs showed high degrees of corre-lation, with the genetic effect explainingapproximately half of the variance in in-crease in visceral adiposity. The degree ofintrapair resemblance in fasting insulinlevel increased after overfeeding, and thisassociation remained stable over 5 years.The resting metabolic weight increased by�10% with weight gain, again showingstrong correlation between twins.

Shuldiner noted that gene-nutrientinteractions are seen in the GET READIStudy of African American siblings, onewith LDL cholesterol above the 50th per-centile, on a high-carbohydrate, high-protein, low-fat, low-cholesterol diet,with increased fiber, potassium, magne-sium, and calcium. Of the 170 individualsin the study, HDL decreased by 5 mg/dland apolipoprotein A1 by 9 mg/dl, withmarked variability, but great similaritybetween siblings, suggesting a gene-environment interaction. In a weight lossstudy, the Quebec Negative Energy Bal-ance Study, seven pairs of monozygotictwins with increased body fat exercisedfor 15 min twice daily to achieve a 580calorie daily energy deficit for 100 days,showing strong correlation between twinsin change in weight, fat mass, and visceralfat levels. Similarly, a study of 14 pairs ofobese monozygotic twins treated for 28days with a 400 calorie per day dietshowed strong correlations in the degreeof weight loss and of fat loss. Three genesappear to account for 15% of the variancein weight gain, the genes for the ADRA2�-adrenoceptor, the NR3C1 glucocorti-coid receptor, and lipoprotein lipase.Activity of phosphofructokinase, oxoglu-tarate dehydrogenase, and cytochromeoxidase are markers of mitochondrialactivity that may explain some of thesephenomena.

In the HERITAGE family study, an-other gene-exercise interaction study, 99families including 483 sedentary individ-uals without CVD, hypertension, or dia-betes completed a thrice weekly exerciseprogram, with a mean increase in energyexpenditure of 380 calories per day (57).There was a within family association ofcalorie expenditure and of fasting insulin,



cardiac output, and a variety of other pa-rameters. Gene markers of this phenom-enon included the angiotensinogen andACE genes (58). Thus, in a variety of set-tings, the benefits of physical activity andboth benefits and adverse effects ofchanges in diet/nutrients are modulatedby genetic heterogeneities.

Molly Bray (Houston, TX) further dis-cussed the genetics of exercise-inducedchanges in body composition, noting thevariability of response to diet interven-tions. Exercise-induced weight loss ap-pears to be less, but more lasting, thanthat occurring with diet. A number ofchanges occur in gene expression with ex-ercise. Muscle contraction decreases ATP,increases AMP, stimulating AMP kinase,endothelial nitric oxide synthase, p38,and acyclCoA carboxylase, leading to in-creased muscle glucose uptake and con-sequently improving insulin sensitivity(59). Exercise increases the efficiency offuel processing, increasing the metabolicrate. Genes may have different effects withand without exercise, as shown by vari-ability among individuals in responses toexercise of HDL, VO2max, heart rate, andnumerous other physiologic parametersexhibiting familial components (60). The�2 adrenoreceptor gene polymorphismGLN27GLU is associated with increasedobesity and greater likelihood of hyper-tension in physically inactive individuals,but the deleterious allele fails to show anadverse phenotype in individuals engag-ing in regular physical activity (61). In theAtherosclerosis Risk In Communities co-hort of 15,792 individuals, among Afri-can Americans, polymorphisms in theG-protein �3 subunit showed oppositeeffect in likelihood of obesity among indi-viduals with and without high levels ofphysical activity, with similar effects ofthe gene on hypertension. Particular ad-verse effect was associated with the com-bination of obesity and lack of physicalactivity, with TT homozygotes appearingto particularly benefit from lifestyle inter-vention. Altogether, Bray noted, some200 genes have been found to be associ-ated with effects of physical activity. In afurther study of whether genes influenceexercise adherence, college students wereinvited to participate in a thrice weeklyexercise program, with variability in ad-herence to the program and in weight re-sponse. Age, sex, baseline activity, BMI,and self-motivation assessed by question-naire did not correlate with these mea-sures, while the degree of adherence andthe change in waist were associated with

waist circumference and with the adiponec-tin, Agouti-related protein, dopamine D4receptor, peroxisome proliferator–activatedreceptor � coactivator 1, and proopiomel-anocortin genes, suggesting genetic deter-minants enjoyment of regular exercise.Thus, as we need to better characterizegenes affecting metabolism, it may also beuseful to understand genes affecting whathas been considered the psychological re-sponse to exercise training.

Jose Florez (Rockville, MD) reviewednew concepts of the transcription factor7–like 2 (TCF7L2) gene, focusing on itsrelationship to metabolic measures andthe response to interventions in the Dia-betes Prevention Program (DPP). TCF7L2was discovered as a diabetes-related genein 2003 (62), with subsequent recogni-tion that a polymorphism in TCF7L2 con-ferred risk of type 2 diabetes inpopulations in Iceland, Demnark, and theU.S. (63). TCF7L2 acts in a signal trans-duction pathway, leading to decreasedphosphorylation of the cytoplasmic adhe-sion and nuclear signaling protein �-cate-nin. Mice not expressing TCF7L2 have adefect in gastrointestinal tract endocrinecells (64), which may reduce GLP-1 tran-scription (65). Human studies have con-firmed an association of TCF7L2 withreduction in insulin secretion (66). In theDPP, 3,234 individuals with impairedglucose tolerance were enrolled, withmetformin and lifestyle decreasing devel-opment of diabetes by 31 and 58%, re-spectively (67). Stratification of DPPparticipants by TCF7L2 genotype atrs79702 showed an explanation of diabe-tes risk, with carriers of the genotype hav-ing decreased insulin secretion (68). Theincreased diabetes risk was confirmed in arecent meta-analysis of multiple studies(69). The DPP lifestyle intervention in-cluded modest weight loss and an exer-cise program. Studying the lifestyleintervention-genotype interaction, thelifestyle intervention was particularly ef-fective in those with the high-risk TT ge-notype, while the placebo group havingthe TT genotype had the highest diabetesrisk. In contrast, metformin treatment re-sponse did not appear to be affected bythis polymorphism.

Florez found that TT carriers hadsomewhat higher insulin sensitivity, al-though lower insulin secretion, notingthat the apparent increase in insulin sen-sitivity may be an artifact of ascertainmentand enrollment of individuals with im-paired glucose tolerance not having dia-betes in the DPP. Similarly, the finding

that the T-allele was more likely to be seenin lean controls and the C-allele in obesecases again may be related to the selectionprocess. Population studies suggest thatthe T-allele is associated with decreasedinsulin secretion but not with reduced in-sulin sensitivity (70). Thus, the mecha-nism of the diabetogenic effect of TCF7L2polymorphisms appears to be the associ-ation of the T gene with decreased insulinsecretion and decreased incretin effect.Increased �-cell TCF7L2 expression ap-pears to be associated with increased�-cell production but impaired process-ing of proinsulin, with preserved periph-eral insulin sensitivity, but with decreasedsuppression of hepatic glucose produc-tion by insulin, indicating hepatic insulinresistance. Current studies do not showthat GLP-1 levels vary by TCF7L2 geno-type but rather that GLP-1–induced insu-lin secretion is decreased in T carriers.Full understanding of the effect ofTCF7L2 on GLP-1 expression, on pancre-atic �-cell gene expression, and in theliver and gastrointestinal tract is not yetavailable, and the molecular mechanismfor impaired insulin processing and de-creased incretin effect with TCF7L2 vari-ants has not been established. Anotherimportant question is whether the assess-ment of TCF7L2 genotype would be use-ful in disease prediction and whethersuch information would allow more spe-cific treatment.

In a study of another genetic variantrelated to diabetes presented at the Amer-ican Diabetes Association’s meeting, Pow-ers et al. (abstract 271) reported that thers1750330 polymorphism in the T-box15 gene, involved in embryonic de-velopment, was associated with obesityand increased waist circumference amongwomen, but not in men, among 697healthy, Caucasian, nondiabetic individ-uals. The T-box15 gene appears to inhibitadipocyte development, with the poly-morphism suggesting a genetic variantwith decreased activity, hence associatedwith increased adipocyte developmentand fat accumulation increasing the like-lihood of obesity.

References1. Ford ES, Mokdad AH, Giles WH: Trends

in waist circumference among U.S. adults.Obes Res 11:1223–1231, 2003

2. Lee J, Ma S, Heng D, Tan CE, Chew SK,Hughes K, Tai ES: Should central obesitybe an optional or essential component ofthe metabolic syndrome? Ischemic heartdisease risk in the Singapore Cardiovascu-

Bloomgarden


lar Cohort Study. Diabetes Care 30:343–347, 2007

3. Ruderman NB, Schneider SH, BerchtoldP: The “metabolically-obese,” normal-weight individual. Am J Clin Nutr 34:1617–1621, 1981

4. Ruderman N, Chisholm D, Pi-Sunyer X,Schneider S: The metabolically obese,normal-weight individual revisited. Dia-betes 47:699–713, 1998

5. Zavaroni I, Bonora E, Pagliara M,Dall’Aglio E, Luchetti L, Buonanno G,Bonati PA, Bergonzani M, Gnudi L, Pas-seri M, Reaven G: Risk factors for coro-nary artery disease in healthy personswith hyperinsulinemia and normal glu-cose tolerance. N Engl J Med 320:702–706, 1989

6. Ferrari P, Weidmann P, Shaw S, GiachinoD, Riesen W, Allemann Y, Heynen G: Al-tered insulin sensitivity, hyperinsulin-emia, and dyslipidemia in individualswith a hypertensive parent. Am J Med 91:589–596, 1991

7. St-Onge MP, Janssen I, Heymsfield SB:Metabolic syndrome in normal-weightAmericans: new definition of the metabol-ically obese, normal-weight individual.Diabetes Care 27:2222–2228, 2004

8. St-Pierre AC, Cantin B, Mauriege P,Bergeron J, Dagenais GR, Despres JP,Lamarche B: Insulin resistance syndrome,body mass index and the risk of ischemicheart disease. CMAJ 172:1301–1305,2005

9. Osei K, Cottrell DA, Orabella MM: Insulinsensitivity, glucose effectiveness, andbody fat distribution pattern in nondia-betic offspring of patients with NIDDM.Diabetes Care 14:890–896, 1991

10. Carey DG, Jenkins AB, Campbell LV,Freund J, Chisholm DJ: Abdominal fatand insulin resistance in normal and over-weight women: direct measurements re-veal a strong relationship in subjects atboth low and high risk of NIDDM. Diabe-tes 45:633–638, 1996

11. Freedman DS, Srinivasan SR, Burke GL,Shear CL, Smoak CG, Harsha DW, Web-ber LS, Berenson GS: Relation of body fatdistribution to hyperinsulinemia in chil-dren and adolescents: the Bogalusa HeartStudy. Am J Clin Nutr 46:403–410, 1987

12. Raji A, Seely EW, Arky RA, Simonson DC:Body fat distribution and insulin resis-tance in healthy Asian Indians and Cau-casians. J Clin Endocrinol Metab 86:5366–5371, 2001

13. Katsuki A, Sumida Y, Urakawa H, Gaba-zza EC, Murashima S, Maruyama N, Mo-rioka K, Nakatani K, Yano Y, Adachi Y:Increased visceral fat and serum levels oftriglyceride are associated with insulin re-sistance in Japanese metabolically obese,normal weight subjects with normal glu-cose tolerance. Diabetes Care 26:2341–2344, 2003

14. Boyko EJ, Leonetti DL, Bergstrom RW,

Newell-Morris L, Fujimoto WY: Low in-sulin secretion and high fasting insulinand C-peptide levels predict increasedvisceral adiposity: 5-year follow-up amonginitially nondiabetic Japanese-Americanmen. Diabetes 45:1010–1015, 1996

15. Petersen KF, Dufour S, Shulman GI: De-creased insulin-stimulated ATP synthesisand phosphate transport in muscle of in-sulin-resistant offspring of type 2 diabeticparents. PLoS Med 2:e233, 2005

16. Jacob S, Machann J, Rett K, Brechtel K,Volk A, Renn W, Maerker E, Matthaei S,Schick F, Claussen CD, Haring HU: Asso-ciation of increased intramyocellular lipidcontent with insulin resistance in leannondiabetic offspring of type 2 diabeticsubjects. Diabetes 48:1113–1119, 1999

17. Seppala-Lindroos A, Vehkavaara S, Hak-kinen AM, Goto T, Westerbacka J, Sovi-jarvi A, Halavaara J, Yki-Jarvinen H: Fataccumulation in the liver is associatedwith defects in insulin suppression of glu-cose production and serum free fatty acidsindependent of obesity in normal men.J Clin Endocrinol Metab 87:3023–3028,2002

18. Gan SK, Samaras K, Thompson CH, Krae-gen EW, Carr A, Cooper DA, ChisholmDJ: Altered myocellular and abdominalfat partitioning predict disturbance in in-sulin action in HIV protease inhibitor-re-lated lipodystrophy. Diabetes 51:3163–3169, 2002

19. Schneider SH, Amorosa LF, Khachadur-ian AK, Ruderman NB: Studies on themechanism of improved glucose controlduring regular exercise in type 2 (non-insulin-dependent) diabetes. Diabetologia26:355–360, 1984

20. Thamer C, Stumvoll M, Niess A, Tschrit-ter O, Haap M, Becker R, Shirkavand F,Bachmann O, Rett K, Volk A, Haring H,Fritsche A: Reduced skeletal muscle oxy-gen uptake and reduced beta-cell func-tion: two early abnormalities in normalglucose-tolerant offspring of patients withtype 2 diabetes. Diabetes Care 26:2126–2132, 2003

21. Nyholm B, Mengel A, Nielsen S, Skjaer-baek C, Moller N, Alberti KG, Schmitz O:Insulin resistance in relatives of NIDDMpatients: the role of physical fitness andmuscle metabolism. Diabetologia 39:813–822, 1996

22. Selvin E, Feinleib M, Zhang L, RohrmannS, Rifai N, Nelson WG, Dobs A, Basaria S,Golden SH, Platz EA: Androgens and di-abetes in men: results from the Third Na-tional Health and Nutrition ExaminationSurvey (NHANES III). Diabetes Care 30:234–238, 2007

23. Kupelian V, Page ST, Araujo AB, TravisonTG, Bremner WJ, McKinlay JB: Low sexhormone-binding globulin, total testos-terone, and symptomatic androgen defi-ciency are associated with development ofthe metabolic syndrome in nonobese

men. J Clin Endocrinol Metab 91:843–850,2006

24. Ronnemaa T, Mattila K, Lehtonen A, Kal-lio V: A controlled randomized study onthe effect of long-term physical exerciseon the metabolic control in type 2 diabeticpatients. Acta Med Scand 220:219–224,1986

25. Wei M, Gibbons LW, Mitchell TL, Ka-mpert JB, Lee CD, Blair SN: The associa-tion between cardiorespiratory fitnessand impaired fasting glucose and type 2diabetes mellitus in men. Ann Intern Med130:89–96, 1999

26. Farrell SW, Braun L, Barlow CE, ChengYJ, Blair SN: The relation of body massindex, cardiorespiratory fitness, and all-cause mortality in women. Obes Res 10:417–423, 2002

27. Wei M, Gibbons LW, Kampert JB, Nicha-man MZ, Blair SN: Low cardiorespiratoryfitness and physical inactivity as predic-tors of mortality in men with type 2 dia-betes. Ann Intern Med 132:605–611,2000

28. Church TS, Cheng YJ, Earnest CP, BarlowCE, Gibbons LW, Priest EL, Blair SN: Ex-ercise capacity and body composition aspredictors of mortality among men withdiabetes. Diabetes Care 27:83–88, 2004

29. Bernstein IL, Lotter EC, Kulkosky PJ,Porte D Jr, Woods SC: Effect of force-feeding upon basal insulin levels of rats.Proc Soc Exp Biol Med 150:546–548, 1975

30. Kahn SE, Prigeon RL, McCulloch DK,Boyko EJ, Bergman RN, Schwartz MW,Neifing JL, Ward WK, Beard JC, PalmerJP, et al: Quantification of the relationshipbetween insulin sensitivity and beta-cellfunction in human subjects: evidence fora hyperbolic function. Diabetes 42:1663–1672, 1993

31. Woods SC, Porte D Jr: Relationship be-tween plasma and cerebrospinal fluid in-sulin levels of dogs. Am J Physiol 233:E331–E334, 1977

32. Woods SC, Lotter EC, McKay LD, Porte DJr: Chronic intracerebroventricular infu-sion of insulin reduces food intake andbody weight of baboons. Nature 282:503–505, 1979

33. West DB, Fey D, Woods SC: Cholecysto-kinin persistently suppresses meal sizebut not food intake in free-feeding rats.Am J Physiol 246:R776–R787, 1984

34. Figlewicz DP, Stein LJ, West D, Porte D Jr,Woods SC: Intracisternal insulin alterssensitivity to CCK-induced meal suppres-sion in baboons. Am J Physiol 250:R856–R860, 1986

35. Porte D Jr, Seeley RJ, Woods SC, BaskinDG, Figlewicz DP, Schwartz MW: Obe-sity, diabetes and the central nervous sys-tem. Diabetologia 41:863–881, 1998

36. Bruning JC, Gautam D, Burks DJ, GilletteJ, Schubert M, Orban PC, Klein R, KroneW, Muller-Wieland D, Kahn CR: Role ofbrain insulin receptor in control of body



weight and reproduction. Science 289:2122–2125, 2000

37. Obici S, Feng Z, Karkanias G, Baskin DG,Rossetti L: Decreasing hypothalamic insu-lin receptors causes hyperphagia and in-sulin resistance in rats. Nat Neurosci5:566–572, 2002

38. Niswender KD, Morrison CD, Clegg DJ,Olson R, Baskin DG, Myers MG Jr, SeeleyRJ, Schwartz MW: Insulin activation ofphosphatidylinositol 3-kinase in the hy-pothalamic arcuate nucleus: a key media-tor of insulin-induced anorexia. Diabetes52:227–231, 2003

39. Burks DJ, Font de Mora J, Schubert M,Withers DJ, Myers MG, Towery HH, Al-tamuro SL, Flint CL, White MF: IRS-2pathways integrate female reproductionand energy homeostasis. Nature 407:377–382, 2000

40. Schwartz MW, Morton GJ: Obesity: keep-ing hunger at bay. Nature 418:595–597,2002

41. Obici S, Zhang BB, Karkanias G, RossettiL: Hypothalamic insulin signaling is re-quired for inhibition of glucose produc-tion. Nat Med 8:1376–1382, 2002

42. Hallschmid M, Benedict C, Schultes B,Fehm HL, Born J, Kern W: Intranasal in-sulin reduces body fat in men but not inwomen. Diabetes 53:3024–3029, 2004

43. Hermansen K, Davies M, Derezinski T,Martinez Ravn G, Clauson P, Home P: A26-week, randomized, parallel, treat-to-target trial comparing insulin detemirwith NPH insulin as add-on therapy tooral glucose-lowering drugs in insulin-naive people with type 2 diabetes. Diabe-tes Care 29:1269–1274, 2006

44. Hennige AM, Sartorius T, Tschritter O,Preissl H, Fritsche A, Ruth P, Haring HU:Tissue selectivity of insulin detemir actionin vivo. Diabetologia 49:1274–1282, 2006

45. Hotamisligil GS, Shargill NS, SpiegelmanBM: Adipose expression of tumor necrosisfactor-alpha: direct role in obesity-linkedinsulin resistance. Science 259:87–91, 1993

46. Uysal KT, Wiesbrock SM, Marino MW,Hotamisligil GS: Protection from obesity-induced insulin resistance in mice lackingTNF-alpha function. Nature 389:610–614, 1997

47. Aguirre V, Uchida T, Yenush L, Davis R,White MF: The c-Jun NH(2)-terminal ki-nase promotes insulin resistance duringassociation with insulin receptor sub-strate-1 and phosphorylation of Ser(307).J Biol Chem 275:9047–9054, 2000

48. Yuan M, Konstantopoulos N, Lee J, Han-sen L, Li ZW, Karin M, Shoelson SE: Re-versal of obesity- and diet-induced insulinresistance with salicylates or targeted dis-ruption of Ikkbeta. Science 293:1673–1677, 2001

49. Hirosumi J, Tuncman G, Chang L, Gor-gun CZ, Uysal KT, Maeda K, Karin M,Hotamisligil GS: A central role for JNK inobesity and insulin resistance. Nature

420:333–336, 200250. Tuncman G, Hirosumi J, Solinas G,

Chang L, Karin M, Hotamisligil GS: Func-tional in vivo interactions between JNK1and JNK2 isoforms in obesity and insulinresistance. Proc Natl Acad Sci U S A 103:10741–10746, 2006

51. Waeber G, Delplanque J, Bonny C,Mooser V, Steinmann M, Widmann C,Maillard A, Miklossy J, Dina C, Hani EH,Vionnet N, Nicod P, Boutin P, Froguel P:The gene MAPK8IP1, encoding islet-brain-1, is a candidate for type 2 diabetes.Nat Genet 24:291–295, 2000

52. Bennett BL, Sasaki DT, Murray BW,O’Leary EC, Sakata ST, Xu W, Leisten JC,Motiwala A, Pierce S, Satoh Y, BhagwatSS, Manning AM, Anderson DW:SP600125, an anthrapyrazolone inhibitorof Jun N-terminal kinase. Proc Natl AcadSci U S A 98:13681–13686, 2001

53. Ozcan U, Cao Q, Yilmaz E, Lee AH, Iwa-koshi NN, Ozdelen E, Tuncman G, Gor-gun C, Glimcher LH, Hotamisligil GS:Endoplasmic reticulum stress links obe-sity, insulin action, and type 2 diabetes.Science 306:457–461, 2004

54. Tong Q, Dalgin G, Xu H, Ting CN, LeidenJM, Hotamisligil GS: Function of GATAtranscription factors in preadipocyte-adi-pocyte transition. Science 290:134–138,2000

55. Wellen KE, Fucho R, Gregor MF, Furu-hashi M, Morgan C, Lindstad T, Vaillan-court E, Gorgun CZ, Saatcioglu F,Hotamisligil GS: Coordinated regulationof nutrient and inflammatory responsesby STAMP2 is essential for metabolic ho-meostasis. Cell 129:537–548, 2007

56. Bouchard C, Tremblay A, Despres JP,Nadeau A, Lupien PJ, Theriault G, Dus-sault J, Moorjani S, Pinault S, Fournier G:The response to long-term overfeeding inidentical twins. N Engl J Med 322:1477–1482, 1990

57. Bouchard C, An P, Rice T, Skinner JS,Wilmore JH, Gagnon J, Perusse L, LeonAS, Rao DC: Familial aggregation ofVO(2max) response to exercise training:results from the HERITAGE FamilyStudy. J Appl Physiol 87:1003–1008, 1999

58. Rankinen T, Gagnon J, Perusse L, Chag-non YC, Rice T, Leon AS, Skinner JS, Wil-more JH, Rao DC, Bouchard C: AGTM235T and ACE ID polymorphisms andexercise blood pressure in the HERITAGEFamily Study. Am J Physiol Heart CircPhysiol 279:H368–H374, 2000

59. Hildebrandt AL, Pilegaard H, Neufer PD:Differential transcriptional activation ofselect metabolic genes in response to vari-ations in exercise intensity and duration.Am J Physiol Endocrinol Metab 285:E1021–E1027, 2003

60. Bouchard C, Rankinen T: Individual dif-ferences in response to regular physicalactivity. Med Sci Sports Exerc 33 (6Suppl.):S446–S451, 2001

61. Meirhaeghe A, Helbecque N, Cottel D,Amouyel P: Beta2-adrenoceptor genepolymorphism, body weight, and physi-cal activity. Lancet 353:896, 1999

62. Reynisdottir I, Thorleifsson G, Benedikts-son R, Sigurdsson G, Emilsson V, Einars-dottir AS, Hjorleifsdottir EE, OrlygsdottirGT, Bjornsdottir GT, Saemundsdottir J,Halldorsson S, Hrafnkelsdottir S, Sigur-jonsdottir SB, Steinsdottir S, Martin M,Kochan JP, Rhees BK, Grant SF, FriggeML, Kong A, Gudnason V, Stefansson K,Gulcher JR: Localization of a susceptibil-ity gene for type 2 diabetes to chromo-some 5q34–q35.2. Am J Hum Genet 73:323–335, 2003

63. Grant SF, Thorleifsson G, Reynisdottir I,Benediktsson R, Manolescu A, Sainz J,Helgason A, Stefansson H, Emilsson V,Helgadottir A, Styrkarsdottir U, Magnus-son KP, Walters GB, Palsdottir E, Jonsdot-tir T, Gudmundsdottir T, Gylfason A,Saemundsdottir J, Wilensky RL, ReillyMP, Rader DJ, Bagger Y, Christiansen C,Gudnason V, Sigurdsson G, Thorsteins-dottir U, Gulcher JR, Kong A, StefanssonK: Variant of transcription factor 7-like 2(TCF7L2) gene confers risk of type 2 dia-betes. Nat Genet 38:320–323, 2006

64. Korinek V, Barker N, Moerer P, van Don-selaar E, Huls G, Peters PJ, Clevers H:Depletion of epithelial stem-cell compart-ments in the small intestine of mice lack-ing Tcf-4. Nat Genet 19:379–383, 1998

65. Drucker DJ: The biology of incretin hor-mones. Cell Metab 3:153–165, 2006

66. Lyssenko V, Lupi R, Marchetti P, DelGuerra S, Orho-Melander M, Almgren P,Sjogren M, Ling C, Eriksson KF, LethagenAL, Mancarella R, Berglund G, Tuomi T,Nilsson P, Del Prato S, Groop L: Mecha-nisms by which common variants in theTCF7L2 gene increase risk of type 2 dia-betes. J Clin Invest 117:2155–2163, 2007

67. Knowler WC, Barrett-Connor E, FowlerSE, Hamman RF, Lachin JM, Walker EA,Nathan DM, Diabetes Prevention Pro-gram Research Group: Reduction in theincidence of type 2 diabetes with lifestyleintervention or metformin. N Engl J Med346:393–403, 2002

68. Florez JC, Jablonski KA, Bayley N, PollinTI, de Bakker PI, Shuldiner AR, KnowlerWC, Nathan DM, Altshuler D, DiabetesPrevention Program Research Group:TCF7L2 polymorphisms and progressionto diabetes in the Diabetes PreventionProgram. N Engl J Med 355:241–250, 2006

69. Florez JC: The new type 2 diabetes geneTCF7L2. Curr Opin Clin Nutr Metab Care10:391–396, 2007

70. Loos RJ, Franks PW, Francis RW, BarrosoI, Gribble FM, Savage DB, Ong KK,O’Rahilly S, Wareham NJ: TCF7L2 poly-morphisms modulate proinsulin levelsand beta-cell function in a British Europidpopulation. Diabetes 56:1943–1947, 2007

Bloomgarden


OBSERVATIONS

Effects of MaternalDiabetes on VisualEvoked Potentialsand EarlyPsychomotorDevelopment of theOffspring

P lanning and intensive treatment ofdiabetes in pregnancy has resultedin dramatic improvements in out-

comes in terms of congenital malforma-tions or perinatal morbidity (1), but it isstill not clear to what extent maternal di-abetes affects cognitive development ofnewborns. In infants of diabetic mothers(IDMs), significant relations have been re-ported between poor maternal metabolicregulation during pregnancy and poorchild intellectual performance, despite anachievement of overall healthy neuropsy-chological functioning (2,3). Evoked po-tentials are commonly used in infants toanalyze maturational processes and clini-cal problems (4).

To evaluate the effects of maternal di-abetes on brain functions of the offspring,we analyzed psychomotor development(Brunet-Lezine test) and visual evokedpotentials (VEPs) in 40 2-month-oldIDMs (21 males and 19 females); 24mothers had type 1 diabetes, 3 had type 2

diabetes, and 13 had gestational diabetes.VEPs were recorded by standard proce-dure (5) within 2 h of administration ofthe development test and blinded to theresults of the development test. Norma-tive VEP data were obtained from 63healthy infants matched for age and sex.The most stable component IV was usedfor statistical analysis. The mean develop-ment quotient (DQ) of IDMs was 94.7 �19.3, and the DQ adjusted for gestationalage (ADQ) was 121.2 � 66.9, with alower subtest score on speech (115.6 �74.0). Six patients (15%) had DQ scores�80, but only one of them had an abnor-mal ADQ. VEPs showed a mean latencysignificantly higher in IDMs than in con-trol subjects on both hemispheres (right:200.2 � 33.8 vs. 155.6 � 29.0 ms, P �0.001; left: 197.9 � 35.5 vs. 155.3 �30.3 ms, P � 0.001), with abnormal re-sponses in four cases (10%). No relationswere found between VEPs and develop-mental scores; one infant had both abnor-mal VEPs and ADQ.

In conclusion, IDMs show delayedVEPs, while their mean developmentalquotient is normal, even if abnormal val-ues may occur in some cases. These datasuggest that maternal diabetes may havesubtle negative effects on brain functionsof offspring. VEPs seem to be suitable toanalyze subclinical neurophysiologicchanges that could represent markers forsubsequent developmental risk.

MARIO BRINCIOTTI, MD1

MARIA MATRICARDI, MD, PHD1

ANTONIETTA COLATRELLA, MD2

FRANCESCO TORCIA, MD3

FRANCESCO FALLUCCA, MD2

ANGELA NAPOLI, MD2

From the 1Department of Child Neuropsychiatryand Rehabilitation Sciences, Faculty of Medicine I,“Sapienza” Rome University, Rome, Italy; the 2De-partment of Clinical Sciences, Diabetes Unit, Facultyof Medicine II, “Sapienza” Rome University, Rome,Italy; and the 3Department of Gynecology, Perina-tology, and Child Health, Faculty of Medicine II,“Sapienza” Rome University, Rome, Italy.

Address correspondence to Mario Brinciotti, MD,Child Neuropsychiatry and Rehabilitation Sciences,“Sapienza” Rome University, Via dei Sabelli, 108,00185 Rome, Italy. E-mail: [email protected].

DOI: 10.2337/dc07-1070© 2007 by the American Diabetes Association.

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Willhoite MB, Bennert HW, Palomaki GE,

Zaremba MM, Herman WH, Williams JR,Spear NH: The impact of preconceptioncounseling on pregnancy outcomes: theexperience of the Maine Diabetes in Preg-nancy Program. Diabetes Care 16:450–455, 1993

2. Rizzo T, Metzger B, Burns WJ, Burns K:Correlations between antepartum mater-nal metabolism and intelligence of off-spring. N Engl J Med 325:911–916, 1991

3. Silverman B, Tizzo T, Cho N, Mettzger B:Long-term effects of the intrauterine envi-ronment. Diabetes Care 21 (Suppl. 2):B142–B149, 1998

4. Taylor MJ, McCulloch DL: Visual evokedpotentials in infants and children. J ClinNeurophysiol 9:357–372, 1992

5. American Electroencephalographic Soci-ety guidelines in electroencephalography,evoked potentials, and polysomnography.J Clin Neurophysiol 11:1–147, 1994

O N L I N E L E T T E R S

e128 DIABETES CARE, VOLUME 30, NUMBER 12, DECEMBER 2007

OBSERVATIONS

Metformin in HeartFailure

W e recently learned of an impor-tant change in the prescribinginformation for metformin, pre-

viously contraindicated in patients withtype 2 diabetes accompanied by heart fail-ure (HF) requiring pharmacological ther-apy. The contraindication was introducedto the label soon after the drug’s U.S.release, when case reports emerged ofHF patients taking metformin who de-veloped lactic acidosis. It was neverclear, however, whether either the HFor the medication were contributingfactors. Subsequent analyses revealed aminiscule risk of lactic acidosis associ-ated with metformin therapy, as com-pared with that of an earlier biguanide,phenformin (1).

In 2005, two studies (2,3) sug-gested that metformin was safe and mayalso provide advantages in the very pa-tients in whom the drug was not to beused. These observational studies foundsignificantly lower mortality risks intype 2 diabetic patients treated withmetformin compared with other agents,after adjustment for differences in pa-tient mix. With this information, theFood and Drug Administration (FDA)

alerted metformin manufacturers that alabel change was appropriate. Over theensuing months, the prescribing informa-tion for all generic products was updated,eliminating the HF contraindication. InNovember 2006, the final package insert(Glucophage) was likewise modified.

For years, the wisdom of metformin’sHF contraindication had been chal-lenged, since it was essentially based oncase reports in the absence of clinical trialvalidation and since glucose-loweringoptions for such patients were otherwiselimited. Moreover, there was biologicalplausibility of benefit, particularly in pa-tients with insulin resistance (4). The re-cent FDA endorsement increases the oralpharmacotherapy choices in HF patientswith diabetes. Renal dysfunction andmetabolic acidosis remain contraindica-tions; HF is still in the label’s “Warnings”section; it should obviously not be used inthose with acute or unstable symptoms.

We bring this information to yourreadership’s attention because we our-selves were surprised to hear of this majorlabel change, with significant implica-tions for patient care. After discussionswith our colleagues, it was clear that wewere not alone in our ignorance. We con-ducted an informal poll of 118 endocri-nologists and diabetes educators 10months after the label change was final-ized; of 101 respondents, 95 (94%) weresimilarly unaware. This episode raisesconcerns regarding clinician notificationwhen prescribing information is updated

after patents expire—in stark contrast tothe recent thiazolidinedione label changes,which were widely disseminated.

SILVIO E. INZUCCHI, MD1

FREDERICK A. MASOUDI, MD, MSPH2

DARREN K. MCGUIRE, MD, MHSC3

From 1Yale University, New Haven, Connecticut;the 2Denver Health Medical Center, Denver, Colo-rado; and the 3University of Texas, SouthwesternMedical Center, Dallas, Texas.

Address correspondence to Silvio E. Inzucchi,MD, Endocrinology, LLCI-101, Yale UniversitySchool of Medicine, New Haven, CT 06520-8020.E-mail: [email protected].

This letter is being simultaneously published in2007 in Diabetes Care and the American Heart Jour-nal. Copyright 2007 by the American Diabetes As-sociation, Inc., and Elsevier.


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Misbin RI: The phantom of lactic acidosis

due to metformin in patients with diabe-tes. Diabetes Care 27:1791–1793, 2004

2. Masoudi FA, Inzucchi SE, Wang Y,Havranek EP, Foody JM, Krumholz HM:Thiazolidinediones, metformin, and out-comes in older patients with diabetes andheart failure. Circulation 111:583–590,2005

3. Eurich DT, Majumdar SR, McAlister FA,Tsuyuki RT, Johnson JA: Improved clini-cal outcomes associated with metforminin patients with diabetes and heart failure.Diabetes Care 28:2345–2351, 2005

4. Masoudi FA, Inzucchi SE: Diabetes melli-tus and heart failure. Am J Cardiol 99:113B–132B, 2007


DIABETES CARE, VOLUME 30, NUMBER 12, DECEMBER 2007 e129

OBSERVATIONS

IsletAutotransplantationRestores NormalGlucose Tolerance ina Patient WithChronic Pancreatitis

Impaired glucose tolerance is a frequentcomplication of chronic pancreatitis(CP) that over time leads to diabetes

(1). Intractable abdominal pain from CPthat can no longer be managed by oralpancreatic enzymes and by narcotic anal-gesics is treated by surgery. To avoid dia-betes onset following pancreatectomy,combined islet autotransplantation hasbeen offered to the patients who sufferedfrom CP in Leicester, U.K., since 1994(2,3) (http://hpb.org.uk/hpbunitspecialist.php?page_id�11).

Here, we present a case report of apatient that suffered from CP and hadan abnormal oral glucose tolerance test(OGTT) before pancreatectomy (show-ing diabetes and impaired glucose tol-erance) but since surgery has hadnormal results. Indeed, OGTT results ofstimulated blood glucose were 11.9 and8 mmol/l at 19 and 11 months presur-gery, respectively.

At the time of the surgery, the patientwas a 43-year-old woman. The etiology ofCP was idiopathic, and the patient suf-fered from chronic epigastric pain formore than 2 years before surgery. On 28May 2002, the patient underwent a totalpancreatectomy combined with islet au-totransplantation during an 8-h operation(including the islet preparation process,as previously described [2,3]). Theweight of the pancreas was 55 g, and after

digestion, 14 ml pancreatic tissue (tissue-packed volume) was obtained. A total of232,721 islets, corresponding to 57,296islet equivalents (IEQ), 18% free of acinartissue, was infused into the left branch ofthe portal vein.

The patient remains insulin indepen-dent 5 years after autoislet transplantationwith 954.93 IEQ/kg body wt. Intravenousinsulin was stopped 8 days postopera-tively; her blood glucose levels were mon-itored every 4 h and remained stablethroughout her hospital stay (21 days).She did not experience any perioperativecomplications. By 2 years postoperation,she began to experience regular postpran-dial hypoglycemic episodes, which is awell-known prolonged insulin responseof transplanted islets due to a defectiveglucagon secretion (4,5).

Annual follow-up visits revealednormal OGTT results. Her fasting C-peptide levels have remained detect-able, and the C-peptide production inresponse to an oral glucose load is sig-nificantly higher than basal levels (1.44and 5.02 ng/ml, respectively, at 5 yearspostsurgery).

To our knowledge, this is the first casereport of a CP patient, with abnormalOGTT before surgery, who has become in-sulin independent after autoislet transplan-tation with less than 1,000 IEQ/kg body wt.These results show that it is possible to re-store normal glucose tolerance with totalpancreatectomy combined with islet auto-transplantation in CP patients with border-line diabetes. Furthermore, these resultssupport the concept that insulin indepen-dence can be achieved in patients with isletcell autograft of a low number of islets. In-deed, from our series, no correlation wasfound between the islet number and theoutcome of the islet transplant in terms ofinsulin independence (2).

SEVERINE ILLOUZ, PHD1

M’BALU WEBB, MSC1

CRISTINA POLLARD, BA1

PATRICK MUSTO, FCA(SA)2

KIERAN O’REILLY, MBCHB3

DAVID BERRY, MD1

ASHLEY DENNISON, MD1

From the 1Department of Hepatobiliary and Pancre-atic Surgery, University Hospital of Leicester, Leic-ester, U.K.; the 2Department of Anesthesiology,University Hospital of Leicester, Leicester, U.K.; andthe 3Department of Pathology, University Hospitalof Leicester, Leicester, U.K.

Address correspondence to Dr. Severine Illouz,PhD, Department of Hepatobiliary and PancreaticSurgery, Leicester General Hospital, GwendolenRoad, Leicester , LE5 4PW, U.K. E-mai l :[email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Banks PA: Epidemiology, natural history,

and predictors of disease outcome inacute and chronic pancreatitis. Gastroin-test Endosc 56 (Suppl. 6):S226–S230,2002

2. Clayton HA, Davies JE, Pollard CA, WhiteSA, Musto PP, Dennison AR: Pancreatec-tomy with islet autotransplantation forthe treatment of severe chronic pancreati-tis: the first 40 patients at the LeicesterGeneral Hospital. Transplantation 76:92–98, 2003

3. White SA, Davies JE, Pollard C, Swift SM,Clayton HA, Sutton CD, Wemyss-HoldenS, Musto PP, Berry DP, Dennison AR: Pan-creas resection and islet autotransplanta-tion for end-stage chronic pancreatitis.Ann Surg 233:423–431, 2001

4. Kendall DM, Teuscher AU, Robertson RP:Defective glucagon secretion during sus-tained hypoglycemia following successfulislet allo- and autotransplantation in hu-mans. Diabetes 46:23–27, 1997

5. Paty BW, Ryan EA, Shapiro AM, Lakey JR,Robertson RP: Intrahepatic islet trans-plantation in type 1 patients does not re-store hypoglycemic hormonal counter-regulation or symptom recognition afterinsulin independence. Diabetes 51:3428–3434, 2002



OBSERVATIONS

Possible Relevanceof HLA-DRB1*0403Haplotype in InsulinAutoimmuneSyndrome Inducedby �-Lipoic Acid,Used as a DietarySupplement

Insulin autoimmune syndrome (IAS) ischaracterized by frequent hypoglyce-mic attacks associated with the pres-

ence of autoantibodies to insulin inpatients who have not received insulin in-jections. Approximately half of IAS pa-tients have a medication history beforeonset, and over 90% of the agents are sul-fydryl compounds such as methimazole,mercaptopropionyl glycine, or glutathi-one. In addition to these compounds,�-lipoic acid (ALA), which is widely usedas a health supplement, is associated witha risk of IAS induction, as previously re-ported in Diabetes Care (1) and otherjournals (2,3). DRB1*0406 is reportedlythe most common and DRB1*0403 thenext most common HLA haplotype con-ferring susceptibility to IAS (4). As forALA-induced IAS, all three reported caseshave the DRB1*0406 but not theDRB1*0403 haplotype (1–3). However,we recently observed a case of IAS, possi-

bly induced by ALA, in a patient who hasthe DRB1*0403 haplotype.

The patient, a 45-year-old woman,lapsed into hypoglycemic coma 1 monthafter starting to take ALA. She had nottaken any of the other aforementionedsulfydryl compounds. She exhibitedmarked hyperinsulinemia (fasting plasmaglucose 88 mg/dl, serum immunoreactiveinsulin 13,240 �U/ml, and serum C-peptide immunoreactivity 2.93 ng/ml).Antibodies to insulin were detected withan insulin binding ratio of 81.2%. Anti-body affinity was low, while binding ac-tivity was high, as commonly observed inIAS. Based on these results, she was diag-nosed as having IAS possibly induced byALA. However, she has the DRB1*0403,not the DRB1*0406, haplotype.

This is the first report of a patient withALA-induced IAS having the DRB1*0403haplotype. Since the DRB1*0403 haplo-type is reportedly associated with IAS in-duced by other sulfydryl compounds, it islikely to confer susceptibility to ALA-induced IAS. Although IAS was a rela-tively rare cause of hypoglycemia in thepast, ALA has become more widely avail-able as a dietary supplement for treatingobesity and diabetes complications. Fur-thermore, in contrast to the very low prev-alence of DRB1*0406 in ethnic groupsother than East Asians, DRB1*0403 wasfound to be widely distributed across vari-ous populations (5). We should thereforebe more aware of ALA-induced IAS.

TETSUYA YAMADA, MD, PHD

JUNTA IMAI, MD, PHD

YASUSHI ISHIGAKI, MD, PHD

YOSHINORI HINOKIO, MD, PHD

YOSHITOMO OKA, MD, PHD

HIDEKI KATAGIRI, MD, PHD

From the Department of Diabetes and Metabolism,Tohoku University Hospital, Sendai, Japan.

Address correspondence to Dr. Hideki Katagiri,Department of Diabetes and Metabolism, TohokuUniversity Hospital, Seiryo, Aoba, Sendai, 980-8575, Japan. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Ishida Y, Ohara T, Okuno Y, Ito T, Hirota

Y, Furukawa K, Sakaguchi K, Ogawa W,Kasuga M: �-Lipoic acid and insulin au-toimmune syndrome. Diabetes Care30:2240–2241, 2007

2. Takeuchi Y, Miyamoto T, Kakizawa T,Shigematsu S, Hashizume K: Insulin au-toimmune syndrome possibly caused byalpha lipoic acid. Intern Med 46:237–239,2007

3. Furukawa N, Miyamura N, Nishida K,Motoshima H, Taketa K, Araki E: Possiblerelevance of alpha lipoic acid contained ina health supplement in a case of insulinautoimmune syndrome. Diabetes Res ClinPract 75:366–367, 2007

4. Uchigata Y, Hirata Y: Insulin autoimmunesyndrome (IAS, Hirata disease). Ann MedInterne (Paris) 150:245–253, 1999

5. Uchigata Y, Hirata Y, Omori Y, IwamotoY, Tokunaga K: Worldwide differences inthe incidence of insulin autoimmune syn-drome (Hirata disease) with respect to theevolution of HLA-DR4 alleles. Hum Immu-nol 61:154–157, 2000



COMMENTS ANDRESPONSES

The Role of Iron inDiabetes and ItsComplications

Reponse to Swaminathan et al.

M arkers of fatty liver such as �-glu-tamyltransferase (GGT) are inde-pendently associated with an

increased risk of type 2 diabetes (1). Somerecent studies have shown that hyperfer-ritinemia may also predict new-onset type2 diabetes (2).

We assessed the cross-sectional rela-tionships between ferritin, GGT, and glu-cose intolerance status in a large cohort ofadults. We performed a retrospectiveanalysis on the database of our clinicalchemistry laboratory to retrieve results ofserum ferritin, GGT, lipids, glucose (fast-ing plasma glucose [FPG]), and C-reac-tive protein (high-sensitivity C-reactiveprotein [hs-CRP]) tests, which were per-formed on the whole cohort of outpatientadults (aged �35 years) consecutively re-ferred by general practitioners for routineblood testing over the past 2 years. Fast-ing GGT, FPG, and lipids were measuredby standard enzymatic procedures(Roche Diagnostics), ferritin by a chemi-luminescence assay (DiaSorin-Liaison),and hs-CRP by a nephelometric assay(Dade-Behring).

We used separate multivariable logis-tic regression analyses to examine the in-teraction relationships with impairedfasting glycemia (impaired fasting glucose

[IFG] as defined by an FPG value �5.6mmol/l) or diabetes (FPG value �7.1mmol/l) as the dependent variables pre-dicted from ferritin quartiles (�42, 42–80, 80–156, and �156 �g/l) within thequartiles of GGT (�16, 16–25, 26–35,and �36 units/l). Adjusting variableswere sex, age, lipids, and hs-CRP.

Cumulative results of FPG and fer-ritin were retrieved for 2,637 individuals.After excluding subjects with C-reactiveprotein �10 mg/l (because inflammationmay increase ferritin) and those with verylow ferritin, which might be due to anemia(�15 �g/l), and very high ferritin, whichmight be due to hemochromatosis (�400�g/l in men and �300 �g/l in women), thefinal study population consisted of 2,449subjects (63% female) with a mean � SD(range) age of 61.8 � 15 years (35–107).Overall, 161 (6.6%) subjects had a FPGvalue �7.1 mmol/l, and 559 (22.8%) sub-jects had IFG. Mean GGT and ferritin con-centrations were 33 � 46 units/l and 108 �84 �g/l, respectively.

Although the prevalence rates of fer-ritin quartiles increased steadily acrossIFG/diabetes categories (ranging from 17to 27% for IFG and from 4 to 8% for di-abetes; P � 0.0001), these prevalences re-markably varied by GGT quartiles. AsGGT increased, the prevalence rates offerritin quartiles across IFG/diabetes cat-egories strengthened (P � 0.001 for inter-action). For example, within the lowestGGT quartile, ferritin quartiles were not as-sociated with IFG (ranging from 12.7 to14.5%) or diabetes (from 1.2 to 1.5%), incontrast to the highest GGT quartile,wherein the prevalence rates ranged from19.2 to 28.3% for IFG and from 9.4 to13.5% for diabetes (P �0.01). These resultsremained significant even after adjustmentfor sex, age, lipids, and hs-CRP.

Our findings, although only correla-tive in nature, indicate that ferritin is as-sociated with a greater frequency of IFGor diabetes only among those with high-normal GGT (�36 units/l), not in thosewith low-normal GGT, and suggest thatferritin itself might not be a sufficient riskfactor for developing IFG/diabetes. Theassociation between increased GGT andglucose intolerance might be explained bysome underlying, biological, mechanismssuch as enhanced oxidative stress, insulinresistance, and fatty liver (3).

GIOVANNI TARGHER, MD1

MASSIMO FRANCHINI, MD2

MARTINA MONTAGNANA, MD3

GIUSEPPE LIPPI, MD3

From the 1Section of Endocrinology, Department ofBiomedical and Surgical Sciences, University Hospi-tal of Verona, Verona, Italy; the 2Service of Immu-nohematology and Transfusion, Civil Hospital,Verona, Italy; and the 3Section of Clinical Chemis-try, Department of Biomedical and MorphologicalSciences, University Hospital of Verona, Verona,Italy.

Address correspondence to Dr. Giovanni Targher,Section of Endocrinology, Department of Biomedi-cal and Surgical Sciences, University of Verona, Os-pedale Civile Maggiore, Piazzale Stefani, 1, 37126Verona, Italy. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Adams LA, Angulo P, Lindor KD: Nonal-

coholic fatty liver disease. CMAJ 172:899–905, 2005

2. Swaminathan S, Fonseca VA, Alam MG,Shah SV: The role of iron in diabetes andits complications. Diabetes Care 30:1926–1933, 2007

3. Targher G: Nonalcoholic fatty liver dis-ease, the metabolic syndrome and the riskof cardiovascular disease: the plot thick-ens. Diabet Med 24:1–6, 2007




On Real-TimeEstimates of BloodGlucose Levels

Response to Trevino

T revino (1) has misunderstood ourdescription of the feedback aboutblood glucose levels offered to pa-

tients in our trial of a real-time telemedi-cine system (2). He has also attributed aquotation from an article by Bode et al. (3)to our report (2). Trevino calls for the useof better statistical methods to deal withtemporal correlations between the re-peated glucose values used in real-timemonitoring. While we have sympathywith this message, we do not wish to seeour trial data misrepresented.

Patients using our telemedicine sys-tem received real-time feedback consist-ing of time series graphs and color-codedhistograms of blood glucose measure-ments. They did not receive, as suggestedin Trevino’s letter, information aboutmeans and SD of blood glucose measure-ments. This suggestion may have arisenfrom a misunderstanding about the appli-cation of statistical methods used in ourstudy to analyze the A1C measurements,

which were the primary outcome of thetrial.

Trevino also attributes the followingstatement to our article: “Overtreating hy-poglycemia has resulted in a marginallysignificant increase in the frequency ofhyperglycemic excursions.” This was notan outcome of our trial, and the quotedtext is from an article by Bode et al. (3).Unfortunately, this misattribution is re-peated by Garg (4), who argues that “thefact that, using traditional finger-stickmethods, patients ‘overtreating hypogly-cemia’ suffer ‘unrecognized hypo- andhyperglycemia’ is precisely because in-termittent monitoring with finger-sticksdoes not give them enough informa-tion.” In fact, in its correct context, thequotation on “overtreating hypoglyce-mia” from Bode et al. (3) describes anoutcome observed during use of theGuardian continuous glucose monitor-ing system rather than finger-stickmethods.

While we have followed this corre-spondence with interest, our trial is not avalid exemplar of the statistical issue dis-cussed by Trevino (1), since patients werenot presented with means or SDs of bloodglucose. We hope that this confusion hasnot distracted from discussion of the im-portant issue of the statistical analysis ofblood glucose time series.

OLIVER J. GIBSON, MENG1

ANDREW J. FARMER, FRCGP2

PATRICK E. MCSHARRY, DPHIL1

LIONEL TARASSENKO, DPHIL1

From the 1Department of Engineering Science, Uni-versity of Oxford, Oxford, U.K.; and the 2Division ofPublic Health and Primary Health Care, Universityof Oxford, Oxford, U.K.

Address correspondence to Oliver J. Gibson, De-partment of Engineering Science, University of Ox-ford, Parks Road, Oxford, OX1 3PJ, U.K. E-mail:[email protected].

O.J.G. has received consulting fees from t� Medi-cal, a company that provides technology to assist pa-tients with chronic diseases in their self-management.L.T. has served on an advisory panel for and holdsstock in t� Medical.


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Trevino G: On real-time estimates of

blood glucose levels (Letter). DiabetesCare 30:e34, 2007. DOI: 10.2337/dc06-2577

2. Farmer AJ, Gibson OJ, Dudley C, BrydenK, Hayton PM, Tarassenko L, Neil A: Arandomized controlled trial of real-timetelemedicine support on glycemic controlin young adults with type 1 diabetes (IS-RCTN 46889446). Diabetes Care 28:2697–2702, 2005

3. Bode B, Gross K, Rikalo N, Schwartz S,Wahl T, Page C, Gross T, Mastrototaro J:Alarms based on real-time sensor glucosevalues alert patients to hypo- and hyper-glycemia: the Guardian Continuous Mon-itoring System. Diabetes Technol Ther6:105–113, 2004

4. Garg SK: On real-time estimates of bloodglucose levels (Letter). Diabetes Care 30:e35–e36, 2007. DOI: 10.2337/dc07-0011




On Real-TimeEstimates of BloodGlucose Levels

Response to Gibson et al.

I would like to extend my apologies toGibson et al. (1) for having misunder-stood their description of the feedback

about blood glucose levels offered to pa-tients in their trial of a real-time telemedi-cine system (2) and for incorrectlyattributing a quotation from an article byBode et al. (3) to their report. I, too, do notwish to see the interesting results of theFarmer et al. (2) trial data misrepresented.It was (and still is) not my intent to eitherblemish or deny the results of Farmer etal. or to assume an antagonistic posturetoward the efforts of the authors.

However, as to their application ofstatistical methods to analyze A1C mea-surements, at the recent American Diabe-

tes Association 2007 Scientific Sessions itwas reported that using A1C to assess gly-cemic control will soon be replaced bymean blood glucose (MBG) (4). This willpresumably “add clarity for diabetic pa-tients looking to manage their disease”(5). Since this will require the replace-ment of one set of reference values (thoseused to define the standard MBG �f(A1C) relationship) with a second set(those from the patient’s own history), itwill also call for the use of better statisticalmethods to deal with the temporal corre-lations between the repeated glucose val-ues used in real-time monitoring.Averages of uncorrelated data are not in-terchangeable with averages of correlateddata. Further, I concur with Gibson et al.(1) in my hope that this confusion has notdistracted from discussion of the impor-tant issue of the statistical analysis ofblood glucose time series.

GEORGE TREVINO, PHD

From CHIRES, Inc., San Antonio, Texas.Address correspondence to George Trevino,

CHIRES, P.O. Box 201481, San Antonio, TX 78220-8481. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Gibson OJ, Farmer AJ, McSharry PE,

Transsenko L: On real-time estimates ofblood glucose levels (Letter). DiabetesCare 30:e133, 2007. DOI: 10.2337/dc07-1668.

2. Farmer AJ, Gibson OJ, Dudley C, BrydenK, Hayton PM, Tarassenko L, Neil A: Arandomized controlled trial of real-timetelemedicine support on glycemic controlin young adults with type 1 diabetes (IS-RCTN 46889446). Diabetes Care 28:2697–2702, 2005

3. Bode B, Gross K, Rikalo N, Schwartz S,Wahl T, Page C, Gross T, Mastrototaro J:Alarms based on real-time sensor glucosevalues alert patients to hypo- and hyper-glycemia: the Guardian Continuous Mon-itoring System. Diabetes Technol Ther6:105–113, 2004

4. Nathan D: International A1c-derived av-erage glucose study (Presentation). Amer-ican Diabetes Association’s 67th ScientificSessions, Chicago, IL, 22–26 June 2007

5. O’Riordan M: Average blood glucoseinstead of HbA1c? Change appears to becoming for diabetes care [article online],July 2007. Heartwire. Available fromh t t p : / / w w w . m e d s c a p e . c o m / v i e warticle/559262.




Vitamin D,ParathyroidHormone Levels, andthe Prevalence ofMetabolic Syndromein Community-Dwelling OlderAdults

Response to Reis et al.

W e read with interest the recent ar-ticle by Reis et al. (1), who con-cluded that neither parathyroid

hormone (PTH) nor 25-hydroxyvita-min-D [25(OH)D] level predicts any ofthe three biochemical components (hy-perglycemia, hypertriglyceridemia, andlow HDL cholesterol) of metabolic syn-drome. Since 25(OH)D and PTH mayplay a role in the etiology of metabolicsyndrome, we retrospectively analyzedassociations between results of PTH,25(OH)D, fasting plasma glucose (FPG),triglycerides, and HDL cholesterol tests,which were performed on an entire co-hort of outpatients aged �35 years whowere consecutively referred to our clinicallaboratory by general practitioners forroutine blood testing over the past 2years.

Venous blood from outpatients wasroutinely collected in the morning on fast-ing subjects. FPG, HDL cholesterol, andtriglycerides were measured by enzymaticprocedures on a Roche/Hitachi Modular

System P (Roche Diagnostics, Mannheim,Germany), whereas PTH and 25(OH)Dwere measured by chemiluminescenceassays on Liaison (DiaSorin, Vercelli, It-aly). Participants were classified as havingabnormal values of the biochemical com-ponents of the metabolic syndrome if theyhad triglycerides �1.69 mmol/l (150 mg/dl), HDL cholesterol �1.3 mmol/l (50mg/dl), and FPG �6.1 mmol/l (110 mg/dl). Significance of differences and fre-quency distribution of values wereassessed by the Kruskal-Wallis and the �2

tests (for categorical variables), respec-tively. Multiple linear regression analysiswas used to test the association betweenFPG, HDL cholesterol, triglycerides (sin-gularly entered as the dependent vari-able), PTH, and 25(OH)D. Because of thelow number of results retrieved (n �112), male patients were excluded fromstatistical analysis. Cumulative results forthe above-mentioned parameters were re-trieved for 738 female outpatients aged�35 years (mean � SD age 65 � 11years).

The study population was clusteredin tertiles of PTH (�50, 50–71, and �71ng/l) and 25(OH)D (�56, 56–127, and�127 nmol/l). No significant differenceswere observed in the value distributionamong the lower, medium, and top ter-tiles of PTH (FPG: 5.6 � 1.3, 5.4 � 1.1,and 5.5 � 1.1 mmol/l, respectively, P �0.994; triglycerides: 1.2 � 1.6, 1.2 � 1.5,and 1.2 � 1.6 mmol/l, P � 0.996; andHDL cholesterol: 1.7 � 1.3, 1.7 � 1.3,and 1.7 � 1.3 mmol/l, P � 0.580) and25(OH)D (FPG: 5.5 � 1.2, 5.4 � 1.2, and5.5 � 1.3 mmol/l, P � 0.497; triglycer-ides: 1.2 � 1.6, 1.2 � 1.5, and 1.2 � 1.5mmol/l, P � 0.361; and HDL cholesterol:1.6 � 1.3, 1.8 � 1.3, and 1.7 � 1.3mmol/l, P � 0.052). Accordingly, no sig-nificant differences were observed in the

frequency of abnormal values among thelower, medium, and top tertiles of PTH(FPG: 20, 15, and 18%, respectively, P �0.647; triglycerides: 15, 15, and 20%,P � 0.549; and HDL cholesterol: 18, 15,and 15%, P � 0.800) and 25(OH)D(FPG: 20, 15, and 19%, P � 0.622; trig-lycerides: 17, 15, and 19%, P � 0.753;and HDL cholesterol: 20, 12, and 15%,P � 0.291). Finally, neither PTH nor25(OH)D was significantly associatedwith FPG, triglyceride, or HDL choles-terol in multiple linear regression analy-sis. The results of this retrospectiveanalysis confirm no effect of PTH and25(OH)D concentrations on individualbiochemical components of the metabolicsyndrome in women.

GIUSEPPE LIPPI, MD1

MARTINA MONTAGNANA, MD1

GIOVANNI TARGHER, MD2

GIAN CESARE GUIDI, MD1

From the 1Sezione di Chimica Clinica, Dipartimentodi Scienze Morfologico-Biomediche, Universita diVerona, Italy; and the 2Sezione di Endocrinologia eMalattie del Metabolismo, Dipartimento di ScienzeBiomediche e Chirurgiche, Universita di Verona,Italy.

Address correspondence to Prof. Giuseppe Lippi,MD, Sezione di Chimica Clinica, Dipartimento diScienze Morfologico-Biomediche, Universita degliStudi di Verona, Ospedale Policlinico G.B. Rossi,Piazzale Scuro, 10, 37134 Verona, Italy. E-mail:[email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Reis JP, von Muhlen D, Kritz-Silverstein

D, Wingard DL, Barrett-Connor E: Vita-min D, parathyroid hormone levels, andthe prevalence of metabolic syndrome incommunity-dwelling older adults. Diabe-tes Care 30:1549–1555, 2007




Vitamin D,ParathyroidHormone Levels, andthe Prevalence ofMetabolic Syndromein Community-Dwelling OlderAdults

Response to Lippi et al.

W e thank Lippi et al. (1) for theircomment to our recent article (2)examining the cross-sectional re-

l a t i on o f 25-hydroxyv i t amin D[25(OH)D] and parathyroid hormone(PTH) with metabolic syndrome and its

components in a community-dwellingsample of older adults. Although Lippi etal. claim we found no association of25(OH)D or PTH levels with componentsof metabolic syndrome, we would like toclarify that in men we did indeed observedose-response associations of 25(OH)Dand PTH with fasting hyperglycemia thatwere not explained by numerous poten-tial confounders. However, we failed toconfirm these findings in women.

In agreement with our results, Lippiet al. report no association of 25(OH)D orPTH levels with the prevalence of hyper-glycemia, hypertriglyceridemia, or lowHDL concentrations among olderwomen. Taken together, these inconsis-tent findings highlight the need for well-performed prospective studies todetermine what role, if any, 25(OH)D orPTH plays in the development of meta-bolic syndrome or its components. Opti-mally, these studies should includeindividuals residing in areas exposed tovarying amounts of direct sunlight to en-sure adequate variability in 25(OH)Dlevels.

JARED P. REIS, PHD1

DENISE VON MUHLEN, MD, PHD2

From the 1Department of Epidemiology, JohnsHopkins Bloomberg School of Public Health, Balti-more, Maryland; and the 2Department of Family andPreventive Medicine, University of California, SanDiego, La Jolla, California.

Address correspondence to Jared P. Reis, PHD,Department of Epidemiology, Johns HopkinsBloomberg School of Public Health, 2024 E. Monu-ment St., Suite 2-600, Baltimore, MD 21287. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Lippi G, Montagnana M, Targher G, Guidi

GC: Vitamin D, parathyroid hormone lev-els, and the prevalence of metabolic syn-drome in community-dwelling olderadults. Diabetes Care 30:e135, 2007. DOI:10.2337/dc07-1605

2. Reis JP, von Muhlen D, Kritz-SilversteinD, Wingard DL, Barrett-Connor E: Vita-min D, parathyroid hormone levels, andthe prevalence of metabolic syndrome incommunity-dwelling older adults. Diabe-tes Care 30:1549–1555, 2007




Strong AssociationBetween TimeWatching Televisionand Blood GlucoseControl in Childrenand AdolescentsWith Type 1Diabetes

Response to Margeirsdottir et al.

W e reply to Margeirsdottir et al.(1), reporting our experience in237 type 1 diabetic Italian pa-

tients (139 male), aged (mean � SD)16.8 � 7.1 years (range 1.8–36.3) withmean disease duration 9.0 � 6.3 years(1.0–27.6), mean A1C 7.7 � 1.1% (5.2–13.5), and mean insulin requirement0.82 � 0.24 units � kg�1 � day�1 (0.24–1.58).

Between January and March 2007, inall patients we evaluated by direct or tele-phone interviews the influence of leisuretime in TV watching and computer use,daily snacking, and weekly hours of exer-cise. Moreover, degree of metabolic con-tro l , BMI SD score , and insul inrequirement were considered. Six pa-tients (2.5%) watched TV �1 h/day(mean A1C 7.2 � 0.6%), 55 patients(23.2%) watched TV 1–2 h/day (meanA1C 7.1 � 0.7%), 67 patients (28.3%)watched TV 2–3 h/day (mean A1C 7.4 �

0.8%), 78 patients (32.9%) watched TV3–4 h/day (mean A1C 8.0 � 1.2%), and31 patients (13.1%) watched TV �4h/day (mean A1C 8.9 � 1.3%). In ourItalian patients, A1C levels were posi-tively related to hours of TV watching(P � 0.001) and to frequency of snacking(P � 0.0001).

On the contrary, time spent using thecomputer did not significantly influencedegree of metabolic control. Like Mar-geirsdottir et al., we found no significantcorrelation between A1C levels and timefor exercise (1). BMI SD score was higherin patients watching TV �2 h/day (n �176; 74.3%) than in the remaining pa-tients (P � 0.01). These data could beexplained by the lower energy expendi-ture and snacking behavior due to TVwatching (1). Moreover, TV watching caneasily be accompanied by eating, and thefrequent food-related advertisements de-signed to invoke feelings of hunger mayresult in higher food intake (2). Computeruse keeps both hands occupied and is notaccompanied by food advertisements andis therefore less prone to induce snacking(2). In healthy schoolchildren, TV watch-ing is associated with soft drink consump-tion (2) and increased BMI (3), andsedentary behavior raises the risk of obe-sity and type 2 diabetes. Our results con-firm the association between time spentwatching TV and poor metabolic control,increased BMI SD score, and insulin re-quirement—factors associated withhigher cardiovascular risk.

In young type 1 diabetic patients, acareful educational intervention aimed toavoid excess TV watching and junk foodintake is mandatory in order to fight in-sulin resistance and metabolic syn-drome—also recently reported in type 1

diabetes and characterizing so-called“double diabetes” (4).

ALESSANDRO GIANNATTASIO, MD1

FRANCESCA LUGANI, MD1,

ANGELA PISTORIO, MD, PHD2

NICOLA MINUTO, MD1

RENATA LORINI, MD1

GIUSEPPE D’ANNUNZIO, MD1

From the 1Pediatric Clinic, University of Genoa, Re-gional Center for Pediatric Diabetes, Genoa, Italy;and the 2Epidemiology and Biostatistic Unit, IRCCSG. Gaslini, Genoa, Italy.

Addres s cor re spondence to Giusepped’Annunzio, MD, Pediatric Clinic, G. Gaslini Insti-tute, Largo Gaslini 5, 16147, Genova, Italy. E-mail:[email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Margeirsdottir HD, Larsen JR, Brunborg

C, Sandvik L, Dahl-Jorgensen K, for theNorwegian Study Group for ChildhoodDiabetes: Strong association between timewatching television and blood glucosecontrol in children and adolescents withtype 1 diabetes. Diabetes Care 30:1567–1570, 2007

2. Giammattei J, Blix G, Marshak HH, Wol-litzer AO, Pettitt DJ: Television watchingand soft drink consumption: associationswith obesity in 11- to 13-year-old school-children. Arch Pediatr Adolesc Med 157:882–886, 2003

3. Kaur H, Choi WS, Mayo MS, Harris KJ:Duration of television watching is associ-ated with increased body mass index. J Pe-diatr 143:506–511, 2003

4. Kilpatrick ES, Rigby AS, Atkin SL: Insulinresistance, the metabolic syndrome, andcomplication risk in type 1 diabetes:“double diabetes” in the Diabetes Controland Complications Trial. Diabetes Care30:707–712, 2007




Effect ofPeriodontitis onOvert Nephropathyand End-Stage RenalDisease in Type 2Diabetes

Response to Shultis et al.

The research study by Shultis et al.(1), which sought to establish a cor-relation between periodontal dis-

ease and overt nephropathy and end-stage renal disease in type 2 diabetes,raises several concerns.

First, while the analysis of the dataattempted to preclude those with knownrisk factors, it fell short of accounting formany other common complications of di-abetes and their direct and indirect im-pact on oral diseases. Diabetes risk factorsand complications that might be associ-ated with alveolar bone changes and werenot considered in this research investiga-tion are secondary hyperparathyroidism,musculoskeletal and rheumatologic dis-orders, bone metabolism alterations, eat-ing disorders, nutritional disorders, self-

care behaviors, psychosocial problems,and thyroid disease.

Second, the assessment and diagnosisof periodontal diseases in this study is in-consistent with the definition currentlyadopted by the American Academy ofPeriodontology. According to their posi-tion statement on the classification ofperiodontal disease, “to arrive at a peri-odontal diagnosis, the dentist must relyupon such factors as: 1) the presence orabsence of clinical signs of inflammation(e.g., bleeding on probing); 2) probingdepths; 3) extent and pattern of loss of clin-ical attachment and bone; 4) patient’s med-ical and dental histories; and 5) presence orabsence of miscellaneous signs and symp-toms, including pain, ulceration, andamount of observable plaque and calculus.”The quantitative assessment of only attach-ment loss by measuring the distance be-tween the cementoenamel junctions to thelevel of alveolar bone in the interproximalarea of the tooth is not considered an ac-ceptable means for the diagnosis and assess-ment of periodontal disease (2).

To a certain extent, dental researchersinvestigating the systemic relationships ofperiodontal diseases tend to downplaythe complex mechanism, pathophysiol-ogy, and complications of diabetes. Muchremains to be learned about the complexinterrelationship between diabetes andoral diseases, and it is unrealistic for thistype of research to encompass these fac-tors. However, multiple factors that mightinfluence the risk of oral health problems

should always be acknowledged in per-forming research of this type. This willlead to a better understanding of theoral-systemic relationship that justifi-ably exists.

In summary, utilizing an assessmentinstrument that does not agree with theprofessional criteria for diagnosis of peri-odontal disease and equating the two isunacceptable. Failing to allude to thesedifferences and an attempt to establish arelationship to chronic kidney diseaseand end-stage renal disease without ac-knowledging other possible mecha-nisms does not substantiate this study’sconclusions.

FRANK N. VARON, DDS

From private practice, Omaha, Nebraska; and theNebraska Diabetes Control Advisory Committee,Lincoln, Nebraska.

Address correspondence to Frank N. Varon,5360 South 72nd St., Ralston, Nebraska 68127. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Shultis WA, Weil EJ, Looker HC, Curtis

JM, Shlossman M, Genco RJ, KnowlerWC, Nelson RG: Effect of periodontitis onovert nephropathy and end-stage renaldisease in type 2 diabetes. Diabetes Care30:306–311, 2007

2. American Academy of Periodontology:Diagnosis of periodontal diseases. J Peri-odontol 74:1237–1247, 2003




Metabolic Syndromein HypertensivePatients: CorrelationBetweenAnthropometric Dataand LaboratoryFindings

Response to Bulhoes and Araujo

R ecently, Bulhoes and Araujo (1) re-ported a high prevalence of meta-bolic syndrome in 102 Brazilian

hypertensive patients, especially whenusing International Diabetes Federation(IDF) criteria for the metabolic syndrome(2). We have some methodological criti-cisms regarding their study: blood pres-sure levels were not described; criteriaadopted for diagnosis of hypertensionwere not those recommended by the Sev-enth Report of the Joint National Com-mittee on the Prevention, Detection,Evaluation, and Treatment of High BloodPressure (3); and data on waist circumfer-ence and HDL were not presented indi-vidually for men and women.

Considering that all patients had highblood pressure, besides abnormal waistcircumference—an obligatory feature forthe IDF diagnosis—only one more com-

ponent was required to classify patients asbelonging to the metabolic syndromegroup. Actually, the high prevalence ofmetabolic syndrome was already ex-pected because patients were selected bythe presence of one metabolic syndromecomponent. This comment is supportedby demonstration of a high prevalence ofmetabolic syndrome when patients wereselected by the presence of diabetes, an-other metabolic syndrome component. Ina sample of consecutive 212 type 2 dia-betic outpatients (80.8% hypertensive;pressure 142 � 22/83 � 11 mmHg), theprevalence of metabolic syndrome (IDFcriteria) was 81.3%, similar to that re-ported by Bulhoes and Araujo (82.4%;P � 0.911). The age (60.7 � 10.5 vs.60.1 � 10.3 years; P � 0.635) and BMI(28.7 � 4.2 vs. 28.6 � 4.1 kg/m2; P �0.857) of their and our patients, respec-tively, were not different. Nevertheless,because the prevalence of metabolic syn-drome increases with age (4), this aspectmust also be considered.

When we evaluated a sample of 110hypertensive (3) patients (blood pressure135 � 15/87 � 10 mmHg; BMI � 33.9 �4.7 kg/m2) attending the Internal Medi-cine Outpatient Clinic, a rather low fre-quency of metabolic syndrome (IDFcriteria) of 51.8% was observed. Thiscould be explained by a lower age (44.2 �7.6 years) compared with that of the pa-tients in the Bulhoes and Araujo studyand also compared with that of our type 2diabetic patients (P � 0.0001 for bothcomparisons); even the latter were over-weight or obese. In conclusion, a highprevalence of metabolic syndrome is pre-

dictable when patients are selected by thepresence of one metabolic syndromecomponent, especially in relatively oldpatients.

TICIANA C. RODRIGUES, MD

CAROLINE K. KRAMER, MD

THAIS STEEMBURGO, RD

VALESCA DALL’ALBA, RD

MIRELA J. AZEVEDO, MD

From the 1Endocrine Division, Hospital de Clınicasde Porto Alegre, Universidade Federal do RioGrande do Sul, Porto Alegre, Brazil.

Address correspondence to Mirela Jobim de Aze-vedo, MD, Hospital de Clınicas de Porto Alegre, RuaRamiro Barcelos, 2350-Predio 12-4° andar, 90035-003, Porto Alegre, Brazil. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Bulhoes K, Araujo L: Metabolic syndrome

in hypertensive patients: correlation be-tween anthropometric data and labora-tory findings. Diabetes Care 30:1624–1626, 2007

2. Alberti KGMM, Zimmet P, Shaw J: Meta-bolic syndrome: a new world-wide defini-tion: a consensus statement from theInternational Diabetes Federation. DiabetMed 23:469–480, 2006

3. Chobanian AV, Bakris GL, Black HR,Cushman WC, Green LA, Izzo JL Jr, JonesDW, Masterson BJ, Oparil S, Wright JT Jr,Roccella EJ: The seventh report of theJoint National committee on Prevention,Detection, Evaluation, and Treatment ofHigh Blood Pressure: the JNC 7 report.JAMA 289:2560–2572, 2003

4. Eckel RH, Grundy SM, Zimmet PZ. Themetabolic syndrome. Lancet 365:1415–1428, 2005




Consensus Statementon the WorldwideStandardization ofthe Hemoglobin A1CMeasurement: theAmerican DiabetesAssociation,EuropeanAssociation for theStudy of Diabetes,InternationalFederation of ClinicalChemistry andLaboratoryMedicine, and theInternationalDiabetes Federation

Response to the ConsensusCommittee

I read with keen interest the statementof the Consensus Committee on thepending worldwide standardization of

hemoglobin A1C measurement (1) andfelt compelled to offer a more “believable”reason why the data supporting thepremise that A1C assay reflects averageglycemia over the preceding fewmonths “are not exceptionally robust.”The Consensus Committee contendsthat “glucose concentrations were notmeasured frequently enough to com-pute a true ‘average.’”

The more believable reason is this: avariation of parameter analysis was car-ried out in ref. 2, and it was shown thereinthat A1C curves decay in a manner “op-posite” to the way they should, as themagnitude of the “disappearance con-stant” changes. It was reported there thatthis inverted decay suggests the particular“nonphysical” behavior that A1C percent-age is decreasing faster, while plasma glu-cose level is decreasing slower, thusdiscrediting the weighted-average rela-tionship now believed to be the corner-stone for the assessment of diabetes care.

That work led to a second effort (3)where the genesis of the flaw revealed inref. 2 was identified. Specifically, Shi et al.(4) derived a differential equation be-tween the fraction of glycated protein hand glucose level G and then solved it toobtain h � 1 � exp[�k � G(t)dt]. Theythen invoke the condition that the frac-tion of glycated protein is “small” andsubsequently linearize the solution toread h � k � G(t)dt. The linearized ver-sion, however, is not a “solution” to theirderived differential equation. In effect,linearizing the solution unacceptably dis-torts the physics. Specifically, it elimi-nates the nonlinearity that manifeststhrough the coupling Gh and reduces abehavior where cause and effect are notproportional to one another (nonlinear)to one where cause and effect are propor-tional to one another (linear). For exam-ple, doubling G in h � k � G(t)dtimmediately doubles h; doubling G in h �1 � exp[�k � G(t)dt] does not. The first-order nonlinear estimate of the solution,viz. h � k � G(t)dt � 0.5k2[� G(t)dt]2, will(for obvious reasons) yield more palatableresults.

In short, a better understanding of therelationship between A1C and averageblood glucose cannot be attained usingexclusively “frequent capillary measure-

ments and continuous glucose monitor-ing” (1). Any approach that avoidsaddressing the evident nonlinearity isdoomed from inception. Further, com-puting a “true ‘average’” cannot beachieved using garden-variety arithmeticaveraging (5). Glucose values drawn as atime series from one individual are inher-ently correlated; standard averaging tech-niques require that the analyzed data beuncorrelated.

GEORGE TREVINO, PHD

From CHIRES, Inc., San Antonio, Texas.Address correspondence to George Trevino,

PhD, CHIRES, Inc., P.O. Box 201481, San Antonio,TX 78220. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Consensus Committee: Consensus state-

ment on the worldwide standardization ofthe hemoglobin A1C measurement: theAmerican Diabetes Association, EuropeanAssociation for the Study of Diabetes, In-ternational Federation of Clinical Chem-istry and Laboratory Medicine, and theInternational Diabetes Federation. Diabe-tes Care 30:2399–2400, 2007

2. Trevino G: On the weighted-average rela-tionship between plasma glucose andHbA1c (Letter). Diabetes Care 29:466,2006

3. Trevino G: On A1c and its dependence onPG level (Letter). Diabetes Res Clin Pract73:111–112, 2006

4. Shi K, Tahara Y, Noma Y, Yasukawa K,Shima K: The response of glycated albu-min to blood glucose change in the circu-lation in streptozotocin-diabetic rats:comparison of theoretical values with ex-perimental data. Diabetes Res Clin Pract17:153–160, 1992

5. Trevino G: On the independence of intra-individual reference values (Letter). ClinChem Lab Med 44:512, 2006




Corneal Sensitivity IsReduced and Relatesto the Severity ofNeuropathy inPatients withDiabetes

Response to Tavakoli et al.

I read with interest the study by Tava-koli et al. (1) and found it to be inter-esting and useful in the management of

diabetes for preventing further complica-tions; however, I would like to make a fewcomments. In this study (1), whether pa-tients with diabetes who had a previoushistory of corneal trauma, contact lensusers, or those with cataract were ex-

cluded was not mentioned. These fac-tors may alter corneal sensitivity.

In a previous study (2), it was seenthat the noncontact corneal aesthesiom-eter was able to assess the corneal sensationthreshold in an accurate and repeatablemanner, and that the Cochet-Bonnet aes-thesiometer has serious deficiencies in itsdesign that limit its ability to accuratelymeasure the corneal sensitivity at low-threshold stimulus. However, in thepresent study (1), the authors have founda good correlation between the Cochet-Bonnet aesthesiometer and the noncon-tact corneal aesthesiometer, and thisappears to be very useful for daily prac-tice. Confocal microscopy appears toallow early detection of beginning neu-ropathy, as decreases in nerve fiber bun-dle counts precede the impairment ofcorneal sensitivity (3).

The present study, along with confo-cal microscopy, will be useful for detect-ing neuropathy in diabetic patients, aswell as in those with impaired glucose tol-erance, so that an early intervention can

be done to prevent further progression ofcomplications.

SANDIP KUMAR DASH, MD, DNB, MNAMS

From Neurology, Apollo Hospitals, Dhaka, Bang-ladesh.

Address correspondence to Dr. Sandip KumarDash, Neurology, Apollo Hospitals, Dhaka, 1229Bangladesh. E-mail: [email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. TavakoliM , Kallinikos PA, Efron N, Boul-

ton AJM, Malik RA: Corneal sensitivity isreduced and relates to the severity of neu-ropathy in patients with diabetes. DiabetesCare 30:1895–1897, 2007

2. Murphy PJ, Lawrenson JG, Patel S, Mar-shall J: Reliabilty of the non-contact cor-neal aesthesiometer and its comparisonwith the Cochet-Bonnet aesthesiometer.Ophthalmic Physiol Opt 18:532–539, 1998

3. Rosenberg Tervo TM, Immonen IJ, MullerLJ, Gronhagen-Riska C, Vesaluoma MH:Corneal structure and sensitivity in type 1diabetes mellitus. Invest Ophthalmol Vis Sci41:2915–2921, 2000




Corneal Sensitivity IsReduced and Relatesto the Severity ofNeuropathy inPatients WithDiabetes

Response to Dash

D r. Dash (1) has correctly identifiedpotential confounders when mea-suring corneal sensitivity. Accord-

ingly, exclusion criteria in our study wereas follows: previous ophthalmic surgeryaffecting the cornea (including refractivesurgery), previous ocular trauma, currentuse of topical ophthalmic medications,contact lens wear, chronic dry eye, orother ocular disease that can affect thecornea. This information was not detailedin our article (2) as a result of the stringentword limitations of Brief Reports in Dia-betes Care.

It is true that the Cochet-Bonnetaesthesiometer (CB-E) is unable to ac-curately measure corneal sensitivity at

low-stimulus thresholds, and we weresurprised to find that the CB-E ex-ceeded the capacity of the noncontactcorneal esthesiometer (NCCE) to detectcorneal sensitivity loss in patients withmild neuropathy (2). Nevertheless,both the CB-E and the NCCE were ableto detect clear differences in cornealsensitivity across the broad range ofneuropathic severity examined in ourstudy (2). Our finding of a significantrelationship between neuropathic se-verity and corneal sensitivity usingthese techniques ought not be surpris-ing in view of our previous demonstra-tions of a significant relationshipbetween neuropathic severity (assessedusing conventional techniques of quan-titative sensory testing and nerve elec-trophysiology) and the density of nervesin the sub-basal layer of the cornea(3,4).

We thank Dr. Dash for his interest inour work and his endorsement of ourdemonstration of the clinical utility ofCB-E (2), NCCE (2), and corneal confocalmicroscopy (3,4) as novel noninvasive oph-thalmic markers of diabetic neuropathy.

MITRA TAVAKOLI, MSC1

PANAGIOTIS A. KALLINIKOS, PHD1

NATHAN EFRON, PHD2

ANDREW J.M. BOULTON, MD, FRCP1

RAYAZ A. MALIK, PHD, MRCP1

From the 1Division of Cardiovascular Medicine,University of Manchester and Manchester Royal In-firmary, Manchester, U.K.; and the 2Institute ofHealth and Biomedical Innovation, QueenslandUniversity of Technology, Brisbane, Australia.

Address correspondence to Dr. Rayaz A. Malik,Division of Cardiovascular Medicine, University ofManchester, Manchester, M13 9NT, U.K. E-mail:[email protected].


● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

References1. Dash SK: Corneal sensitivity is reduced

and relates to the severity of neuropathyin patients with diabetes (Letter). DiabetesCare 30:e142, 2007. DOI: dx.doi.org/10.2337/dc07-1555

2. Tavakoli M, Kallinikos PA, Efron N, Boul-ton AJM, Malik RA: Corneal sensitivity isreduced and relates to the severity of neu-ropathy in patients with diabetes. DiabetesCare 30:1895–1897, 2007

3. Malik RA, Kallinikos P, Abbott CA, vanSchie CHM, Morgan PB, Efron N, BoultonAJM: Corneal confocal microscopy: anon-invasive surrogate of nerve fibredamage and repair in diabetic patients.Diabetologia 46:683–688, 2003

4. Quattrini C, Tavakoli M, Jeziorska M,Kallinikos P, Tesfaye S, Finnigan J, Mar-shall A, Boulton AJM, Efron N, Malik RA:Surrogate markers of small fiber damagein human diabetic neuropathy. Diabetes56:2148–2154, 2007



Diabetes Care - lib.ajaums.ac.irlib.ajaums.ac.ir/booklist/382463.pdf · Management of Type 2...

Documents

Transcript of Diabetes Care - lib.ajaums.ac.irlib.ajaums.ac.ir/booklist/382463.pdf · Management of Type 2...