Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and...

8
Effects of Weight Loss, Weight Cycling, and Weight Loss Maintenance on Diabetes Incidence and Change in Cardiometabolic Traits in the Diabetes Prevention Program Diabetes Care 2014;37:27382745 | DOI: 10.2337/dc14-0018 OBJECTIVE This study examined specic measures of weight loss in relation to incident di- abetes and improvement in cardiometabolic risk factors. RESEARCH DESIGN AND METHODS This prospective, observational study analyzed nine weight measures, character- izing baseline weight, short- versus long-term weight loss, short- versus long-term weight regain, and weight cycling, within the Diabetes Prevention Program (DPP) lifestyle intervention arm (n = 1,000) for predictors of incident diabetes and im- provement in cardiometabolic risk factors over 2 years. RESULTS Although weight loss in the rst 6 months was protective of diabetes (hazard ratio [HR] 0.94 per kg, 95% CI 0.90, 0.98; P < 0.01) and cardiometabolic risk factors (P < 0.01), weight loss from 0 to 2 years was the strongest predictor of reduced diabetes incidence (HR 0.90 per kg, 95% CI 0.87, 0.93; P < 0.01) and cardiometabolic risk factor improvement (e.g., fasting glucose: b = 20.57 mg/dL per kg, 95% CI 20.66, 20.48; P < 0.01). Weight cycling (dened as number of 5-lb [2.25-kg] weight cycles) ranged 06 times per participant and was positively associated with incident diabetes (HR 1.33, 95% CI 1.12, 1.58; P < 0.01), fasting glucose (b = 0.91 mg/dL per cycle; P = 0.02), HOMA-IR (b = 0.25 units per cycle; P = 0.04), and systolic blood pressure (b = 0.94 mmHg per cycle; P = 0.01). After adjustment for baseline weight, the effect of weight cycling remained statistically signicant for diabetes risk (HR 1.22, 95% CI 1.02, 1.47; P = 0.03) but not for cardiometabolic traits. CONCLUSIONS Two-year weight loss was the strongest predictor of reduced diabetes risk and improvements in cardiometabolic traits. Obesity is an established risk factor for a range of complex diseases, including type 2 diabetes, cardiovascular disease, several cancers, osteoarthritis, and periodontitis. Intentional weight loss, on the other hand, reduces the risk of most of these dis- eases. A persons body weight can vary considerably in both the short- and long-term. 1 Diabetes Research Center, Massachusetts General Hospital, Boston, MA 2 Department of Medicine, Harvard Medical School, Boston, MA 3 The Biostatistics Center, George Washington University, Rockville, MD 4 MedStar Health Research Institute, Hyattsville, MD, and Georgetown University School of Med- icine, Washington, DC 5 The David Geffen School of Medicine, University of California at Los Angeles, Los Angeles, CA 6 Pennington Biomedical Research Center, Louisi- ana State University, Baton Rouge, LA 7 Veterans Affairs Puget Sound Health Care Sys- tem and University of Washington, Seattle, WA 8 Center for Human Genetic Research, Depart- ment of Medicine, Massachusetts General Hos- pital, Boston, MA 9 Program in Medical and Population Genetics, Broad Institute, Cambridge, MA 10 Department of Medicine, Division of Endocri- nology, Metabolism and Diabetes, University of Colorado Anschutz Medical Campus, Aurora, CO 11 Department of Clinical Sciences, Lund Univer- sity, Malm¨ o, Sweden 12 Department of Nutrition, Harvard School of Public Health, Boston, MA Corresponding author: Linda M. Delahanty, [email protected]. Received 3 January 2014 and accepted 19 June 2014. Clinical trial reg. no. NCT00004992, clinicaltrials .gov. This article contains Supplementary Data online at http://care.diabetesjournals.org/lookup/ suppl/doi:10.2337/dc14-0018/-/DC1. *A list of Diabetes Prevention Program Research Group investigators is provided in the Supple- mentary Data online. The opinions expressed are those of the investi- gators and do not necessarily reect the views of the funding agencies. © 2014 by the American Diabetes Association. Readers may use this article as long as the work is properly cited, the use is educational and not for prot, and the work is not altered. Linda M. Delahanty, 1,2 Qing Pan, 3 Kathleen A. Jablonski, 3 Vanita R. Aroda, 4 Karol E. Watson, 5 George A. Bray, 6 Steven E. Kahn, 7 Jose C. Florez, 1,2,8,9 Leigh Perreault, 10 and Paul W. Franks, 11,12 for the Diabetes Prevention Program Research Group* 2738 Diabetes Care Volume 37, October 2014 EPIDEMIOLOGY/HEALTH SERVICES RESEARCH

Transcript of Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and...

Page 1: Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and Weight Loss Maintenance on Diabetes Incidence and Change in Cardiometabolic Traits in

Effects of Weight Loss, WeightCycling, and Weight LossMaintenance on DiabetesIncidence and Change inCardiometabolic Traits in theDiabetes Prevention ProgramDiabetes Care 2014;37:2738–2745 | DOI: 10.2337/dc14-0018

OBJECTIVE

This study examined specific measures of weight loss in relation to incident di-abetes and improvement in cardiometabolic risk factors.

RESEARCH DESIGN AND METHODS

This prospective, observational study analyzed nine weight measures, character-izing baseline weight, short- versus long-termweight loss, short- versus long-termweight regain, and weight cycling, within the Diabetes Prevention Program (DPP)lifestyle intervention arm (n = 1,000) for predictors of incident diabetes and im-provement in cardiometabolic risk factors over 2 years.

RESULTS

Although weight loss in the first 6 months was protective of diabetes (hazard ratio[HR] 0.94 per kg, 95% CI 0.90, 0.98; P < 0.01) and cardiometabolic risk factors (P <

0.01), weight loss from 0 to 2 years was the strongest predictor of reduced diabetesincidence (HR 0.90 per kg, 95% CI 0.87, 0.93; P < 0.01) and cardiometabolic risk factorimprovement (e.g., fasting glucose: b =20.57 mg/dL per kg, 95% CI20.66,20.48;P < 0.01). Weight cycling (defined as number of 5-lb [2.25-kg] weight cycles) ranged0–6 times per participant and was positively associated with incident diabetes (HR1.33, 95% CI 1.12, 1.58; P < 0.01), fasting glucose (b = 0.91mg/dL per cycle; P = 0.02),HOMA-IR (b = 0.25 units per cycle; P = 0.04), and systolic blood pressure (b = 0.94mmHg per cycle; P = 0.01). After adjustment for baseline weight, the effect of weightcycling remained statistically significant for diabetes risk (HR 1.22, 95% CI 1.02, 1.47;P = 0.03) but not for cardiometabolic traits.

CONCLUSIONS

Two-year weight loss was the strongest predictor of reduced diabetes risk andimprovements in cardiometabolic traits.

Obesity is an established risk factor for a range of complex diseases, including type 2diabetes, cardiovascular disease, several cancers, osteoarthritis, and periodontitis.Intentional weight loss, on the other hand, reduces the risk of most of these dis-eases. A person’s body weight can vary considerably in both the short- and long-term.

1Diabetes Research Center, MassachusettsGeneral Hospital, Boston, MA2Department of Medicine, Harvard MedicalSchool, Boston, MA3The Biostatistics Center, George WashingtonUniversity, Rockville, MD4MedStar Health Research Institute, Hyattsville,MD, and Georgetown University School of Med-icine, Washington, DC5The David Geffen School of Medicine, Universityof California at Los Angeles, Los Angeles, CA6Pennington Biomedical Research Center, Louisi-ana State University, Baton Rouge, LA7Veterans Affairs Puget Sound Health Care Sys-tem and University of Washington, Seattle, WA8Center for Human Genetic Research, Depart-ment of Medicine, Massachusetts General Hos-pital, Boston, MA9Program in Medical and Population Genetics,Broad Institute, Cambridge, MA10Department of Medicine, Division of Endocri-nology, Metabolism and Diabetes, University ofColorado Anschutz Medical Campus, Aurora, CO11Department of Clinical Sciences, Lund Univer-sity, Malmo, Sweden12Department of Nutrition, Harvard School ofPublic Health, Boston, MA

Corresponding author: Linda M. Delahanty,[email protected].

Received 3 January 2014 and accepted 19 June2014.

Clinical trial reg. no. NCT00004992, clinicaltrials.gov.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc14-0018/-/DC1.

*A list of Diabetes Prevention Program ResearchGroup investigators is provided in the Supple-mentary Data online.

The opinions expressed are those of the investi-gators and do not necessarily reflect the views ofthe funding agencies.

© 2014 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered.

Linda M. Delahanty,1,2 Qing Pan,3

Kathleen A. Jablonski,3 Vanita R. Aroda,4

Karol E. Watson,5 George A. Bray,6

Steven E. Kahn,7 Jose C. Florez,1,2,8,9

Leigh Perreault,10 and Paul W. Franks,11,12

for the Diabetes Prevention Program

Research Group*

2738 Diabetes Care Volume 37, October 2014

EPIDEM

IOLO

GY/HEA

LTHSERVICES

RESEA

RCH

Page 2: Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and Weight Loss Maintenance on Diabetes Incidence and Change in Cardiometabolic Traits in

Factors associated with fluctuations inbody weight include many biopsychoso-cial processes, such as genetic predispo-sition, thoughts, and emotions, andbarriers in a person’s social or culturalenvironment (1–3). Regaining weightafter intentional weight loss is common(4,5). Although the extent and rate ofweight regain varies greatly from oneperson to the next, few studies haveconsidered the dynamic nature ofchanges in body weight in relation todiabetes and cardiometabolic risk(6,7). Studies that seek to determinethe dynamic relationships between var-iations in body weight and susceptibilityto disease may yield novel insights intothe role of obesity in chronic diseaseand thus give rise to tailored interven-tions that seek to mitigate disease risk.The purpose of this analysis was to

compare the effects of weight lossvariables on diabetes incidence and car-diometabolic risk factor levels in partic-ipants who received the DiabetesPrevention Program (DPP) weight lossintervention. We also sought to estab-lish an optimized set of weight loss var-iables that are most strongly related tothe risk of type 2 diabetes.

RESEARCH DESIGN AND METHODS

ParticipantsThe design, methods, and main out-comes of the DPP have been publishedpreviously (4). In the DPP, 3,234 partic-ipants aged 25 years or older with ele-vated fasting glucose (5.3–6.9 mmol/L,or #6.9 mmol/L for American Indians),impaired glucose tolerance (glucose7.8–11.0 mmol/L 2 h after a 75-g oralglucose load), and elevated BMI ($22kg/m2 for Asian Americans, $24 kg/m2

for others) were randomly assigned toone of three treatments: intensive life-style intervention, metformin, or pla-cebo control (4). This analysis waslimited to participants randomized tothe intensive lifestyle intervention. Par-ticipants included those who had com-pleted at least 2 years of the lifestyleintervention to allow sufficient time toexamine fluctuations in body weight af-ter the initial 6 months of weight losswith the core curriculum. Of the 1,079DPP participants originally randomizedto intensive lifestyle, 63 were excludeddue to less than 2 years of total partici-pation and another 16 were excludeddue to missing measurements, leaving

1,000 participants in the current analy-sis (93% of the total lifestyle cohort).Measurements for fasting glucose,HOMA-insulin resistance (IR), triglycer-ides, and weight at baseline were notsystematically different between the79 excluded subjects (SupplementaryTable 1) and the 1,000 participants in-cluded in the analysis (Table 1).

Useof thiazidediuretics orb-adrenergicantagonists was an exclusion criterion,but other classes of antihypertensiveagents were permitted. New medicaltherapies initiated by nonstudy physiciansafter enrollment were not restricted.Study procedures and documents wereapproved by institutional review boardsat all participating sites. All participantsprovided written informed consent fortheir participation. The study was per-formed in accordance with the Declara-tion of Helsinki.

InterventionThe DPP lifestyle intervention has beenpreviously described in detail (8). Inbrief, the lifestyle intervention had twospecific goals: to facilitate $7% ofbody weight loss and to achieve $150min/week of moderate intensity physicalactivity. Participants were given caloriegoals to promote weight loss deter-mined by initial body weight and fatgram goals that were based on 25% ofcalories from fat. Lifestyle coaches metwith participants individually over thefirst 24 weeks to review a 16-sessioncore curriculum that focused on diet,physical activity, and strategies for be-havioral modification. After the first 6months, lifestyle coaches offered tai-lored individual sessions at least onceevery 2 months and group classes/cam-paigns three times per year to help im-prove physical activity levels and sustainweight loss.

Outcomes MeasuresSubjects were weighed in light clothingwithout shoes at each intervention andoutcome visit. Diabetes incidencewas de-termined by fasting plasma glucose levelsevery 6 months and 2-h glucose levelsfrom 75-g oral glucose tolerance testsperformed annually. Diabetes eventsthrough the 1.5-year checkup (beforethe 2-year follow-up) are excluded fromour analysis for the diabetes outcome (n =48), and only those between year 2 andthe end of DPP (31 July 2001) are consid-ered (n = 952). Therefore, the risk of

diabetes conditional on the fact that aparticipant was diabetes-free before the2-year follow-up ismodeled. Cardiometa-bolic risk factor assessments were donefor fasting glucose (mg/dL), insulin resis-tance (HOMA-IR) (9,10), fasting triglycer-ides (mg/dL), and systolic blood pressure(SBP; mmHg). Cardiometabolic outcomemeasuresweremadeat the 2-year follow-up assessment.

Weight Loss VariablesTo determine which aspects of weightloss were most strongly related to theincidence of diabetes and levels of car-diometabolic risk factors, nine separatevariables for weight or weight loss overspecific time intervals were examined.These included:

c Baseline weight (kg)c Weight loss (kg) from baseline (imme-

diate loss) to:

+ 1 month

+ 3 months

+ 6 months

+ 2 yearsc Weight loss (kg) up to 24 months (de-

layed loss):

+ From 6 to 24 months

+ From 18 to 24 monthsc Weight cycling: A completeweight cy-

cle is defined as a 5-lb (2.25-kg) ormore weight loss from the highestweight since last cycle and then a5-lb (2.25-kg) or more weight regainfrom the lowest weight since thelast cycle. Weight cycling for a givenparticipant is the number of suchcycles in the 2-year period startingfrom the DPP baseline.

c Average weight in the first 2 years:Theweighted average of the availableweight measures from baseline toyear 2, where each weight observa-tion is weighted by the time durationbetween this measure and its previ-ous measure.

Statistical MethodsThe relationships of the weight variableswith diabetes incidence and cardiome-tabolic risk factor levels were estimatedusing Cox proportional hazards and lin-ear regression models, respectively. Co-variates included age at randomization(years), race, sex, and the correspondingbaseline measure of glucose, HOMA-IR,SBP, and triglyceride level. The Pearsoncorrelation coefficient between thebaseline weight and weight loss from

care.diabetesjournals.org Delahanty and Associates 2739

Page 3: Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and Weight Loss Maintenance on Diabetes Incidence and Change in Cardiometabolic Traits in

1, 3, and 6 months and 2 years to base-line, weight cycles, and average weightin the first 2 years were significantly dif-ferent from zero (P , 0.01; Table 2).Therefore, baseline weight was not con-trolled for due to the moderate-to-highdegree of correlations between baselineweight and some of the other measuresof weight.To assess the effects of weight cycling

beyond baseline weight, 2-year weight

loss, and average weight, we performedsecondary analyses where baselineweight, 2-year weight loss, and averageweight were adjusted in addition to num-ber of 5-lb weight cycles. Assessments ofweight loss from 18 to 24 months werereplaced in the Cox models by a variablequantifying weight loss in the most re-cent 6 months, which was treated as atime-varying coefficient. Outliers in thecardiometabolic outcomes (one in fasting

glucose, one in HOMA-IR, and one inSBP) were deleted before linear regres-sion analyses were performed for thecorresponding outcomes.

One model was fitted for each combi-nation of outcome and weight measure.For the same outcome, we comparedthe size of the coefficients correspond-ing to different standardized weightmeasures. For example, we fitted ninemodels for the same fasting glucose

Table 1—Weight variables and cardiometabolic risk factors of the study cohort

Mean SD Minimum 25th percentile Median 75th percentile Maximum

BaselineFasting glucose (mg/dL) 106.27 8.15 76.00 100.00 105.00 112.00 139.00HOMA-IR 6.99 4.33 0.56 4.15 6.05 8.89 54.46Triglyceride (mg/dL) 163.42 97.38 31.00 98.00 139.00 198.00 796.00SBP (mmHg) 123.42 14.63 84.00 112.00 122.00 132.00 175.00

At year 2Fasting glucose (mg/dL) 104.92 14.78 74.00 97.00 103.00 110.00 331.00HOMA-IR 5.97 4.67 0.46 3.20 4.87 7.31 60.03Triglyceride (mg/dL) 140.95 84.00 31.00 86.00 118.00 169.00 672.00SBP (mmHg) 120.12 14.72 84.00 110.00 119.00 129.00 209.00

Baseline weight (kg) 93.70 20.29 48.60 79.00 90.90 104.58 192.20

Weight loss from baseline to (kg)1 month 1.38 2.66 217.17 20.18 1.27 2.83 15.853 months 4.40 3.87 29.62 1.79 4.13 6.51 25.276 months 6.84 5.62 213.25 3.20 6.45 9.60 35.552 years 5.39 7.56 219.15 1.05 4.50 8.50 77.50

Weight loss from (kg)18 months to 2 years 21.38 5.73 227.60 24.35 21.20 1.13 49.206 months to 2 years 20.61 3.18 217.95 22.25 20.60 0.90 33.60

5-lb weight cycles (n) 1.45 0.99 0 1 1 2 6

Average weight (kg) 83.03 19.87 44.93 74.02 84.57 98.66 188.68

All body composition variables are raw weight loss values in the units of kilograms. The study sample consisted of 1,000 intensive lifestyle participantswith at least 2 years of follow-up and excluded participants who developed diabetes during those 2 years (n = 952).

Table 2—Pearson partial correlation coefficients between the nine weight metrics conditional on age at enrollment

BLweight

BL to 1month

BL to 3months

BL to 6months

18 months to 2years

BL to 2years

6 months to 2years

5-lb cycles(n)

Averageweight

BL weight 1 0.10 0.26 0.33 0.01 0.20 20.06 0.35 0.960.00 ,0.01 ,0.01 0.78 ,0.01 0.10 ,0.01 ,0.01

BL to 1 month 1 0.80 0.60 20.07 0.38 20.10 0.05 20.06,0.01 ,0.01 0.03 ,0.01 0.00 0.13 0.08

BL to 3 months 1 0.86 20.08 0.56 20.12 0.06 0.04,0.01 0.02 ,0.01 0.00 0.09 0.24

BL to 6 months 1 20.10 0.67 20.12 0.05 0.08,0.01 ,0.01 0.00 0.14 0.02

18 months to2 years

1 0.26 0.45 20.11 0.03,0.01 ,0.01 0.00 0.39

BL to 2 years 1 0.66 20.10 20.06,0.01 0.00 0.09

6 months to2 years

1 20.19 20.16,0.01 ,0.01

5-lb cycles (n) 1 0.38,0.01

BL, baseline.

2740 Weight Change Affects Diabetes Incidence Diabetes Care Volume 37, October 2014

Page 4: Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and Weight Loss Maintenance on Diabetes Incidence and Change in Cardiometabolic Traits in

outcome and obtained nine coefficientestimates for the nine standardizedweight measures. Weight measureswere standardized by the correspondingethnic- and sex-specific SDs, which yielddirectly comparable regression coeffi-cient estimates. Therefore, the regres-sion coefficient of each measure ofweight is the difference in the outcomecorresponding to 1 SD change in thegiven weight measure. Pairwise com-parisons among the nine coefficientswere made. We then tested whetherthe difference between a pair of coeffi-cients varied from zero.Themost predictive single weight var-

iable was selected, whose model wascalled the core model. To determinewhether a combination of the weightvariables could be defined that im-proved the predictive accuracy for inci-dent diabetes of the core model,expanded models that included addi-tional weight variables besides the co-variates in the core model wereselected, and the predictive ability ofdifferent models was assessed consider-ing both statistical significance and clin-ical importance. Statistical selection ofcandidate weight measures was doneby stepwisemodel selections, with a sig-nificance level for variable entry of P =0.25 and a significance level for variableremoval of P = 0.15. The three stron-gest predictors of diabetes identifiedthrough the stepwise model selectionranked by effect sizes are 1) weightloss from baseline to year 2, 2) averageweight in the first 2 years, and 3) weightloss in the first month after randomiza-tion. None of these variables was collin-ear (variance inflation factor ,2).We also estimated the predictive ac-

curacy of the logistic model includingthe top threeweightmeasures (11). Fur-thermore, three sets of sensitivity anal-yses were tested to determine theprognostic power of weight loss frombaseline to year 2 on diabetes risk byadding dietary and physical activitymeasurements to the model. In the firstsensitivity analysis, the average daily ca-loric intake calculated from food fre-quency questionnaires at baseline andyear 1 was added to the Cox model of2-year weight loss and demographic var-iables. In the second sensitivity analysis,average physical activity measured inmetabolic hours per week at baselineand year 1 was added in addition to

the same core model. In the third sensi-tivity analysis, average daily caloric in-take and weekly physical activity inmetabolic hours per week were bothadded. All calculations were done usingSAS 9.2 software (SAS Institute, Cary,NC).

RESULTS

At baseline, this subgroup of lifestyleparticipants was 51 (SD 11) years oldon average, 68% were women, and54% were white, 18% African American,16% Hispanic American, 5% NativeAmerican, and 6% Asian American byself-reported ethnicity. Participants’mean BMI was 33.77 (SD 6.56) kg/m2.Other weightmeasures and cardiometa-bolic risk factors are outlined in Table 1.The cohort achieved a mean weight lossof 6.84 (SD 5.62) kg at 6 months andsustained a mean weight loss of 5.39(SD 7.56) kg at 2 years. After completionof the core curriculum at 6 months,there was a modest mean weight regainof 0.61 (SD 3.18) kg over the ensuing 18months, with more rapid weight gain(mean 1.38 [SD 5.73] kg) between 18months and 24 months. From baselineto 2 years, there were small changes infasting glucose and systolic blood pres-sure, with greater changes in HOMA-IRand triglyceride levels. The averagenumber of 5-lb (2.25-kg) weight cycleswas 1.45, with a range of 0–6 weightcycles over 2 years. Weight cycling wasmore frequent in men than in women(1.59 vs. 1.39 cycles/year, P , 0.01),younger participants (1.60 cycles/yearfor age ,45 years vs. 1.40 for 45–59years, and 1.35 for $60 years, P =0.01), and in African Americans thanother ethnic groups (1.68 vs. 1.00–1.44cycles/year, P , 0.01).

Table 2 summarizes the pairwise par-tial correlation coefficients betweenweight variables conditional on age atenrollment. Baseline body weight andseveral weight loss variables were sig-nificantly correlated with one another.Higher baseline weight correlatedwith a larger number of weight cycles(r = 0.35, P , 0.01), weight loss (6months: r = 0.33, P , 0.01; 2 years: r =0.20, P , 0.01), and a larger averageweight in the first 2 years (r = 0.96, P ,0.01), but not weight regain between 6and 24 months (P = 0.10) or from 18 to24months (P = 0.78). More initial weightloss (e.g., baseline to 1-month weight

loss) was associated with greater subse-quent weight loss from baseline to 24months (r = 0.38, P , 0.01) and lessweight loss/greater weight regain be-tween 6 and 24 months (r = 20.10,P , 0.01) but not with more weight cy-cling (number of 5-lb (2.25-kg) weightcycles; P = 0.13). Weight cycling was cor-related with average weight in the first 2years (r = 0.38, P , 0.001), less weightloss/greater weight regain between 6and 24 months (r = 0.19, P , 0.01) andbetween 18 and 24 months (r = 20.11,P, 0.01), and less long-termweight lossat 2 years (r = 20.10, P , 0.01).

There were 99 incident diabetesevents ascertained during a medianfollow-up of 3 years (range 2–4.5 years).Table 3 reports the estimated effectsof a 1-unit (kg or weight cycle) changein the weight loss variable on diabetesincidence and glucose concentrations,HOMA-IR, SBP, and triglyceride concen-trations at 2 years. Figure 1 shows thecomparative effects of 1-SD change spe-cific to ethnic/sex of the various bodyweight variables on diabetes incidenceand cardiometabolic outcomes at 2years. Greater weight loss in the first 6months and from 6 months to 2 yearswere both predictive of reduced diabe-tes incidence; however, overall weightloss from baseline to 2 years was thestrongest single predictor of lower dia-betes incidence. The effects of 2-yearweight loss on diabetes risk were notaffected by further adjusting for covari-ates measuring caloric intake and phys-ical activity (data not shown).

Weight cycling (defined as number of5-lb weight cycles) was predictive of anincreased incidence of diabetes. More-over, the effect of weight cycling ondiabetes risk remained statistically sig-nificant in models adjusted for baselineweight (hazard ratio [HR] 1.22, 95% CI1.02, 1.48; P = 0.03) or 2-year weightloss (HR 1.22, 95% CI 1.02, 1.47; P =0.03). The effect of the number of 5-lbweight cycles was no longer significantwhen baseline weight and 2-year weightloss were both included in the samemodel (HR 1.11, 95% CI 0.91, 1.35; P =0.29) or when average weight over 2years was included in the model (HR1.15, 95% CI 0.95, 1.39; P = 0.15).

The expandedmodel including demo-graphic variables, 2-year weight loss,1-month weight loss, and average weightadds significant predictive power to the

care.diabetesjournals.org Delahanty and Associates 2741

Page 5: Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and Weight Loss Maintenance on Diabetes Incidence and Change in Cardiometabolic Traits in

core model (P , 0.01 compared withthe core model). The areas under thecurve (AUCs) of the core model (demo-graphics and 2-year weight loss) and theexpanded models described above were0.72 and 0.76, respectively. The AUC inthe model including the top threeweight measures was 0.76, which wasessentially the same as the logisticmodel including all candidate weightmeasures (AUC 0.76; P = 0.90 for differ-ence between AUCs). Weight cyclingraised the risk of diabetes (HR 1.33,95% CI 1.12, 1.58; P = 0.001) but didnot significantly add to the predictiveaccuracy of the core model (P = 0.28).

Cardiometabolic OutcomesWeight loss was predictive of improvedcardiometabolic outcomes in both thefirst 6 months and from 6 months to 2years, but again, was not as predictive astotal weight loss from baseline to 2years. Weight loss from 18 months to2 years was a significant predictor of im-proved triglyceride levels and insulinsensitivity at 2 years but was not a sig-nificant predictor of the other cardiome-tabolic traits or diabetes incidence.Weight cycling was associated withworsening of most cardiometabolic riskfactor levels except for fasting triglycer-ides: fasting glucose (b = 0.91, 95% CI0.17, 1.65 mg/dL per kg; P = 0.02),HOMA-IR (b = 0.25, 95% CI 0.01, 0.49per kg; P = 0.04), and SBP (b = 0.94,95% CI 0.19, 1.68 mmHg per kg; P =0.01); however, these effects were nolonger statistically significant when ad-justed for baseline weight or 2-yearweight loss.

CONCLUSIONS

This study expands on previous researchby examining the relative importance ofseveral measures of body weight as pre-dictors of diabetes and changes in car-diometabolic risk factors and identifyingan optimized set of weight loss variablesassociated with diabetes risk. We com-pared the strength and magnitude ofthe associations for various weight lossvariables with the change in cardiome-tabolic levels and diabetes incidence toquantify the short- and long-term time-dependent effects of weight loss, recentweight loss, degree of exposure to ex-cess weight, and weight cycling in thefirst 2 years of the DPP trial. Althoughweight loss during the first 6 months of

Table

3—Estim

atedeffects(H

Rorco

efficien

t)ofthebodyweightdynamicsondiabetesinciden

ceandca

rdiometabolicoutcomes

Diabetes

inciden

ceFastingglucose

(mg/dL)

HOMA-IR

Triglycerides

(mg/dL)

SBP(m

mHg)

HR

95%CI

Pb

95%CI

Pb

95%CI

Pb

95%CI

Pb

95%CI

P

BLweigh

t1.02

1.01,

1.03

,0.01

0.08

0.04,

0.12

0.01

0.04

0.02,

0.05

,0.01

0.10

20.11,

0.30

0.36

0.09

0.05,

0.12

,0.01

BLto

1month

0.93

0.86,

1.00

0.05

20.47

20.75,

0.20

,0.01

20.28

20.37,

0.19

,0.01

21.34

22.83,

0.14

0.08

20.35

20.64

,0.07

0.01

BLto

3months

0.91

0.86,

0.97

,0.01

20.67

20.86,

20.47

,0.01

20.27

20.33,

20.21

,0.01

21.77

22.81

,20.73

,0.01

20.37

20.56,

20.17

,0.01

BLto

6months

0.94

0.90,

0.98

,0.01

20.59

20.72,

20.46

,0.01

20.22

20.26,

20.18

,0.01

21.68

22.38

,20.99

,0.01

20.35

20.48,

20.21

,0.01

BLto

2years

0.90

0.87,

0.93

,0.01

20.57

20.66,

20.48

,0.01

20.20

20.23,

20.17

,0.01

22.27

22.76

,21.78

,0.01

20.30

20.40,

20.21

,0.01

Last

6months

0.95

0.89,

1.01

0.12

20.22

20.45,

0.01

0.06

20.11

20.18,

20.04

,0.01

21.85

23.07

,20.64

,0.01

20.15

20.39

,0.08

0.19

6monthsto

2years

0.91

0.88,

0.95

,0.01

20.46

20.58,

20.33

,0.01

20.15

20.19,

20.11

,0.01

22.38

23.03

,21.72

,0.01

20.20

20.33,

20.07

,0.01

5-lb

cycles

(n)

1.33

1.12,

1.58

,0.01

0.91

0.17,

1.65

0.02

0.25

0.01,

0.49

0.04

2.57

21.35,

6.50

0.20

0.94

0.19,

1.68

0.01

Average

weightin

2years

1.05

1.03,

1.07

,0.01

0.30

0.22,

0.38

,0.01

0.14

0.10,

0.16

,0.01

0.68

0.26

,1.10

,0.01

0.26

0.18,

0.34

,0.01

BL,baseline.ThePvalues

areoutputfrom

individualregressionsforthespecificoutcomeandweightmetriccombinationwithoutadjustmen

tsformultipletests.Allweightmeasuresareinkilogram

sexceptfor

number

of5-lb

weightcycles.

2742 Weight Change Affects Diabetes Incidence Diabetes Care Volume 37, October 2014

Page 6: Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and Weight Loss Maintenance on Diabetes Incidence and Change in Cardiometabolic Traits in

the lifestyle intervention was a majorpredictor of reduced incidence of diabe-tes, the strongest predictor of reduceddiabetes risk and improvements in car-diometabolic traits was the overall 2-year weight loss: every kilogram ofweight loss in the first 6 months of thetrial corresponded with a 6% reduction

in risk of diabetes, whereas every kilo-gram of weight loss from baseline to 2years corresponded with a 10% de-crease in the risk of diabetes. Our resultsdiffer from those previously reported byHamman et al. (12) because we exam-ined the development of diabetes fromyear 2 until study end and excluded

those who had less than 2 years offollow-up and those who developed di-abetes within the first 1.5 years of theintervention.

For all cardiometabolic traits, theoverall weight loss at 2 years was a bet-ter predictor than average weight inthe first 2 years. Interestingly, the initial

Figure 1—Estimated effects (HR or coefficient) and the corresponding 95% CIs of body weight dynamics on diabetes risks (A), fasting glucose (B),HOMA-IR (C), triglyceride levels (D), and SBP (E).

care.diabetesjournals.org Delahanty and Associates 2743

Page 7: Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and Weight Loss Maintenance on Diabetes Incidence and Change in Cardiometabolic Traits in

6-month weight loss appears to be a bet-ter predictor for the reduction in fastingblood glucose, SBP, and HOMA-IR thanthe most recent 6-month weight loss,whereas the most recent 6-monthweight loss was a predictor of both lowertriglyceride levels and improved insulinsensitivity.We also identified an optimized set of

weight loss variables for predicting inci-dent diabetes in the DPP, which includes1-month weight loss, 2-year weight loss,and either initial body weight or expo-sure to excess body weight over 2 years.These results suggest that there areadded benefits to early rapid weightloss in the first month in the context ofsuccessful 2-year weight loss results,particularly for those who have higherinitial body weights.Our results are consistent with previ-

ous reports that have shown that weightlosses that are notmaintained are not aseffective in reducing the risk of diabetesas those that are maintained (12). Ourfindings regarding the importance of thepattern of weight loss are also consis-tent with findings from the Look AHEAD(Action for Health in Diabetes) trial, an-other long-term intentional weight lossstudy in people with diabetes. In LookAHEAD, greater overall weight losseshad the strongest relationship with im-proved cardiovascular disease risk fac-tors at 4 years; however, a pattern oflarger month-to-month weight loss dur-ing the first year was also predictive ofgreater improvements in glycated he-moglobin, HDL cholesterol, and SBP at4 years, independent of total weightloss (6). These studies and our data em-phasize that early larger weight lossesthat are sustained seem to confer addedbenefits for reducing the risk of diabetesand improving cardiometabolic out-comes in the medium-term.Weight cycling and the average

weight in the first 2 years were bothamong the top three predictors of in-creased incidence of diabetes, implyingthat continued efforts to reduce expo-sure to excess body weight are impor-tant in reducing the risk of diabetes. Theeffect of weight cycling on diabetes riskwas still significant even after adjustingfor baseline weight or 2-year weightchange, but not after adjusting for aver-age weight in the first 2 years. Impor-tantly, however, in a translationalsetting, one would not know the long-

term weight change before or duringthe intervention period, and so focusingon reducing weight cycling and prevent-ing weight gain are logical and com-plimentary objectives for diabetesprevention. Moreover, a patient’s base-line weight is clearly not the focus of anintervention. Thus, in the context oftranslation, the correlations amongbaseline weight, average weight, andweight cycling are important for a clini-cian to consider because heavier pa-tients are more likely to weight cycleand develop diabetes. Nevertheless,this does not mean diabetes risk is notaffected by weight cycling in these pa-tients or that one should not focus onreducing weight cycling for diabetesprevention. By contrast, we concludefor the cardiometabolic traits thatweight cycling does not convey effectsthat are independent of baseline weightor 2-year weight change.

Our findings generally agree withthose reported elsewhere that indicatethat weight cycling is associated withweight regain, increased risk of hyper-tension and diabetes, and elevations inother cardiometabolic traits (2,13–17).In contrast to previous studies thathave categorized the magnitude andfrequency of weight cycling (13,15,17)(e.g., 5–9, 10–19, 20–49, and $50 lb)(13), we measured the effects of thenumber of weight cycles (gain andloss) of at least 5 lb because most DPPparticipants did not have 10-lb weightcycles over 2 years of follow-up. Al-though our findings show that overallweight status is an important determinantof diabetes and cardiometabolic risk andthat overall weight loss lowers risk, avoid-ing weight regain of more than 5 lb alsoappears to be clinically important.

A strength of our study is that we ex-amined and compared the relative impor-tance of a variety of weight loss variablesin a well-characterized, ethnically diversecohort of people with impaired glucosetolerance that was pursuing intentionalweight loss. We also used weight changedata from actual clinic visits rather thanfrom self-reported weights to assessweight cycling.

However, these findings must also beinterpreted with several limitations inmind. Many estimated effects are statis-tically significant but may not be clinicallyrelevant due to the large sample size.The results from analyses focused on

estimating the HRs of each weight vari-able on diabetes and cardiometabolicoutcomes are not adjusted for multipletesting. This is because extensive existingevidence supports an effect of bodyweight change on diabetes risk; thus,the prior probability of association islikely to be greater than the null, whichis not considered in conventionalmultipletest correction procedures. By clearly de-fining the number of tests performed andpresenting nominal P values, we providethe necessary information for readers todetermine for themselves the possibilitythat these associations represent truepositive effects. Potential confoundingfactors include changes in diet or activitythat vary within the intervention andcould affect development or worseningof cardiometabolic risk factors. Althoughwe measured changes in body weight,weight cycling, and magnitude of expo-sure to excess body weight, we did notexamine body fat distribution by waist cir-cumference or visceral fat, andweight as aproxy for change in adiposity may lackprecision. Because the DPP is a multieth-nic cohort of people at high risk for diabe-tes, these resultsmaynot begeneralizableto individuals at a lower risk of diabetes.

These findings have several importantclinical implications. First, current pro-grams that focus on translating the firstmonths of the DPP lifestyle interventionneed to consider offering maintenanceprograms to maximize sustainability ofweight loss, improvements in cardiome-tabolic risk factors, and the potential toprevent or delay diabetes. Early rapidweight loss that is sustained may conferadditional benefits for reducing the riskof diabetes. Although lapses in eatingbehavior may lead to weight regainand some weight cycling, the ability torefocus on weight loss behaviors andachieve weight loss overall appears tobe most important for diabetes preven-tion and improved cardiometabolic out-comes. Nevertheless, minimizing weightcycling is also important for reducing di-abetes risk and for psychological well-being and perceived self-efficacy relatedto weight loss. Efforts to promoteweight loss should seek to achieve thisin a manner that is steady and consis-tent over time.

In conclusion, the overall 2-yearweightloss among the DPP lifestyle interven-tion participants was the strongest pre-dictor of reduced risk for diabetes and

2744 Weight Change Affects Diabetes Incidence Diabetes Care Volume 37, October 2014

Page 8: Effects of Weight Loss, Weight - Diabetes Care · Effects of Weight Loss, Weight Cycling, and Weight Loss Maintenance on Diabetes Incidence and Change in Cardiometabolic Traits in

improvements in cardiometabolic traitscompared with eight other body weightvariables. Early rapid weight loss in thecontext of successful 2-year weight lossalso provided added benefit in diabetesrisk reduction. In the context of the DPP’sintentional weight loss program, weightcyclingwas associatedwithoverallweightregain, increased risk of diabetes, and el-evations in SBP, insulin resistance, andfasting glucose. Weight cycling conferredan increased risk for diabetes indepen-dent of baseline weight or 2-year weightchange but not when both were consid-ered simultaneously or when averageweight over 2 years was considered. Theeffects of weight cycling on the cardio-metabolic traits were not statistically sig-nificant when baseline weight or 2-yearweight were conditioned upon. Definingthe weight loss variables that are moststrongly associated with the reduced in-cidence of diabetes and improvementsin cardiometabolic traits may have impli-cations for the choice of approaches instudies that examine genetic influenceson weight loss and in the translation offindings from studies such as the DPP toclinical practice.

Acknowledgments. The DPP investigatorsgratefully acknowledge the commitment anddedication of the participants of the DPP.Funding. The National Institute of Diabetesand Digestive and Kidney Diseases (NIDDK) ofthe National Institutes of Health (NIH) providedfunding to the clinical centers and the coordi-nating center for the design and conduct of thestudy and the collection, management, analysis,and interpretation of the data (U01-DK-048489). The Southwestern American IndianCenters were supported directly by the NIDDKand the Indian Health Service. The GeneralClinical Research Center Program, National Cen-ter for Research Resources, and the Departmentof Veterans Affairs supported data collection atmany of the clinical centers. Funding for datacollection and participant support was alsoprovided by the Office of Minority Health, theNational Institute of Child Health and HumanDevelopment, the National Institute on Aging,the Centers for Disease Control and Prevention,the Office of Research on Women’s Health, and

the American Diabetes Association. Bristol-Myers Squibb and Parke-Davis provided medi-cation. This research was also partly supportedby the intramural research program of theNIDDK. LifeScan Inc., Health o meter, HoechstMarion Roussel, Inc., Merck-Medco ManagedCare, Inc., Merck and Co., Nike Sports Market-ing, Slimfast Foods Co., and Quaker Oats Co.donated materials, equipment, or medicines forconcomitant conditions. McKesson BioServicesCorp., Matthews Media Group, Inc., and theHenry M. Jackson Foundation provided supportservices under subcontract with the coordi-nating center. A complete list of centers, in-vestigators, and staff can be found in theSupplementary Data online. K.A.J. (the DPP Ge-netics Program) has a research grant from NIH(R01-DK072014). J.C.F. has received NIH grants(R01-DK072041). P.W.F. is supported by theDepartment of Clinical Sciences, Lund Universityand the Department of Nutrition, Harvard Uni-versity. P.W.F. has received speaking honorariafrom academic and nonprofit organizations.Duality of Interest. L.M.D. has a financialinterest in Omada Health, a company that devel-ops online behavior-change programs, with afocus on diabetes. These interests were reviewedand are managed by Massachusetts GeneralHospital and Partners HealthCare in accordancewith their conflict-of-interest policies. J.C.F. hasreceived consulting honoraria fromNovartis, Lilly,and Pfizer. P.W.F. has received speaking hono-raria from Novo Nordisk and has research grantsfrom Novo Nordisk.Author Contributions. L.M.D., Q.P., and P.W.F.researched the data and wrote the manu-script. K.A.J., V.R.A., K.E.W., G.A.B., S.E.K., andJ.C.F., and L.P. reviewed and edited the manu-script. P.W.F. is the guarantor of this work and,as such, had full access to all the data in thestudy and takes responsibility for the integrityof the data and the accuracy of the dataanalysis.

References1. Papas MA, Alberg AJ, Ewing R, Helzlsouer KJ,Gary TL, Klassen AC. The built environment andobesity. Epidemiol Rev 2007;29:129–1432. Teixeira PJ, Going SB, Sardinha LB, LohmanTG. A review of psychosocial pre-treatment pre-dictors of weight control. Obes Rev 2005;6:43–653. Hill JO, Galloway JM, Goley A, et al. Scientificstatement: sociological determinants of predia-betes and type 2 diabetes. Diabetes Care 2013;36:1–104. Knowler WC, Barrett-Connor E, Fowler SE,et al.; Diabetes Prevention Program ResearchGroup. Reduction in the incidence of type 2 di-abetes with lifestyle intervention or metformin.N Engl J Med 2002;346:393–403

5. Wing RR; Look AHEAD Research Group. Long-term effects of a lifestyle intervention onweightand cardiovascular risk factors in individualswith type 2 diabetes mellitus: four-year resultsof the Look AHEAD trial. Arch Intern Med 2010;170:1566–1575

6. Neiberg RH, Wing RR, Bray GA, et al.; LookAHEAD Research Group. Patterns of weightchange associated with long-term weightchange and cardiovascular disease risk factorsin the Look AHEAD Study. Obesity (Silver Spring)2012;20:2048–2056

7. Maruthur NM,Ma Y, Delahanty LM, et al.; forthe Diabetes Prevention Research Group. Earlyresponse to preventive strategies in the Diabe-tes Prevention Program. J Gen InternMed 2013;28:1629–1636

8. Diabetes Prevention Program (DPP) ResearchGroup. The Diabetes Prevention Program (DPP):description of lifestyle intervention. DiabetesCare 2002;25:2165–21719. Matthews DR, Hosker JP, Rudenski AS, NaylorBA, Treacher DF, Turner RC. Homeostasis modelassessment: insulin resistance and beta-cell func-tion from fasting plasma glucose and insulin con-centrations inman. Diabetelogia 1985;28:412–41910. Haffner S, Temprosa M, Crandall J, et al.;Diabetes Prevention Program Research Group.Intensive lifestyle intervention or metformin oninflammation and coagulation in participantswith impaired glucose tolerance. Diabetes2005;54:1566–157211. DeLong ER, DeLong DM, Clarke-Pearson DL.Comparing the areas under two or more corre-lated receiver operating characteristic curves:a nonparametric approach. Biometrics 1988;44:837–84512. Hamman RF, Wing RR, Edelstein SL, et al. Ef-fect of weight loss with lifestyle intervention onrisk of diabetes. Diabetes Care 2006;29:2102–210713. Field AE, Byers T, Hunter DJ, et al. Weightcycling, weight gain, and risk of hypertension inwomen. Am J Epidemiol 1999;150:573–57914. Guagnano MT, Pace-Palitti V, Carrabs C,Merlitti D, Sensi S. Weight fluctuations couldincrease blood pressure in android obesewomen. Clin Sci (Lond) 1999;96:677–68015. Guagnano MT, Ballone E, Pace-Palitti V,et al. Risk factors for hypertension in obesewomen. The role of weight cycling. Eur J ClinNutr 2000;54:356–36016. Waring ME, Eaton CB, Lasater TM, LapaneKL. Incident diabetes in relation to weight pat-terns during middle age. Am J Epidemiol 2010;171:550–55617. French SA, Folsom AR, Jeffery RW, ZhengW, Mink PJ, Baxter JE. Weight variability andincident disease in older women: the IowaWomen’s Health Study. Int J Obes Relat MetabDisord 1997;21:217–223

care.diabetesjournals.org Delahanty and Associates 2745