Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome

7
Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome Brent Graham, MD, Glenn Regehr, PhD, Gary Naglie, MD, James G. Wright, MD From the University Health Network, Hospital for Sick Children, Toronto Rehabilitation Institute, Toronto, Ontario, Canada; and the University of Toronto, Toronto, Ontario, Canada. Purpose: To develop clinical diagnostic criteria for carpal tunnel syndrome (CTS) that modeled the clinical diagnostic practices of experts. Methods: Fifty-seven clinical findings associated with CTS had been ranked previously in order of diagnostic importance using Delphi as a method of establishing consensus among a panel of expert clinicians. The 8 most highly ranked criteria then were placed into all possible combinations to create 256 unique case histories. Two new panels of experts rated these case histories. One panel made a binary evaluation as to whether the case history did or did not represent CTS. This allowed the development of a logistic regression model that had the probability of carpal tunnel syndrome as the dependent variable and the weighted diagnostic criteria as the independent variables. This model then was validated against the judgments of the second panel of clinicians who estimated the probability of CTS for each of the same case histories. Results: The correlation between the probability of CTS predicted by the model and the panel of clinicians was 0.71. Conclusions: The most important clinical diagnostic criteria for CTS as identified from a larger pool of potential diagnostic items through a consensus approach using Delphi were weighted and found to correlate well with the judgments of a new panel of clinicians. By improving the consistency of the diagnosis of CTS these criteria should lead to more effective treatment and a better understanding of the effect of workplace exposures in the development of this condition. A methodology that emphasizes a rigorous approach to item generation and item reduction through expert consensus, followed by validation, may represent a template for establishing consensus among experts on other controversial clinical issues. (J Hand Surg 2006;31A:919.e1–919.e7. Copyright © 2006 by the American Society for Surgery of the Hand.) Type of study/level of evidence: Diagnostic, Level I. Key words: Carpal tunnel syndrome, consensus, diagnosis, logistic regression. C arpal tunnel syndrome (CTS) is diagnosed commonly, with prevalence estimates in the general population varying from 0.1% to 9.2%. 1–5 This variation in prevalence estimates is caused in part by different diagnostic criteria for CTS. 1,6 –10 Misdiagnosis, which probably is frequent, may lead to the inappropriate use of electrodiagnostic tests and, more importantly, inappropriate treatment. The inaccurate diagnosis of CTS has been identified as one of the most common causes of treatment failure for CTS. 11,12 There is no gold standard for establishing a diag- nosis of CTS. The diagnosis of CTS has been based on both clinical findings 8 –10 and the results of elec- trodiagnostic testing. 6,7,13 Although electrodiagnos- tic studies often are assumed to represent a diagnostic gold standard for CTS, 14 –16 this assumption may be questioned for at least 3 reasons. First, because elec- trodiagnostic study results are abnormal only when compression is sufficiently severe to cause structural abnormalities in the median nerve, 17,18 nerve con- duction test results may be normal despite the pres- The Journal of Hand Surgery 919.e1

Transcript of Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome

Page 1: Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome

C9cCmtTaf

Development and Validation of DiagnosticCriteria for Carpal Tunnel Syndrome

Brent Graham, MD, Glenn Regehr, PhD, Gary Naglie, MD,James G. Wright, MD

From the University Health Network, Hospital for Sick Children, Toronto Rehabilitation Institute, Toronto,Ontario, Canada; and the University of Toronto, Toronto, Ontario, Canada.

Purpose: To develop clinical diagnostic criteria for carpal tunnel syndrome (CTS) thatmodeled the clinical diagnostic practices of experts.Methods: Fifty-seven clinical findings associated with CTS had been ranked previously inorder of diagnostic importance using Delphi as a method of establishing consensus among apanel of expert clinicians. The 8 most highly ranked criteria then were placed into all possiblecombinations to create 256 unique case histories. Two new panels of experts rated these casehistories. One panel made a binary evaluation as to whether the case history did or did notrepresent CTS. This allowed the development of a logistic regression model that had theprobability of carpal tunnel syndrome as the dependent variable and the weighted diagnosticcriteria as the independent variables. This model then was validated against the judgments ofthe second panel of clinicians who estimated the probability of CTS for each of the same casehistories.Results: The correlation between the probability of CTS predicted by the model and the panelof clinicians was 0.71.Conclusions: The most important clinical diagnostic criteria for CTS as identified from a largerpool of potential diagnostic items through a consensus approach using Delphi were weightedand found to correlate well with the judgments of a new panel of clinicians. By improving theconsistency of the diagnosis of CTS these criteria should lead to more effective treatment anda better understanding of the effect of workplace exposures in the development of thiscondition. A methodology that emphasizes a rigorous approach to item generation and itemreduction through expert consensus, followed by validation, may represent a template forestablishing consensus among experts on other controversial clinical issues. (J Hand Surg2006;31A:919.e1–919.e7. Copyright © 2006 by the American Society for Surgery of the Hand.)Type of study/level of evidence: Diagnostic, Level I.Key words: Carpal tunnel syndrome, consensus, diagnosis, logistic regression.

nottgqtca

arpal tunnel syndrome (CTS) is diagnosedcommonly, with prevalence estimates in thegeneral population varying from 0.1% to

.2%.1–5 This variation in prevalence estimates isaused in part by different diagnostic criteria forTS.1,6–10 Misdiagnosis, which probably is frequent,ay lead to the inappropriate use of electrodiagnostic

ests and, more importantly, inappropriate treatment.he inaccurate diagnosis of CTS has been identifieds one of the most common causes of treatment

ailure for CTS.11,12 d

There is no gold standard for establishing a diag-osis of CTS. The diagnosis of CTS has been basedn both clinical findings8–10 and the results of elec-rodiagnostic testing.6,7,13 Although electrodiagnos-ic studies often are assumed to represent a diagnosticold standard for CTS,14–16 this assumption may beuestioned for at least 3 reasons. First, because elec-rodiagnostic study results are abnormal only whenompression is sufficiently severe to cause structuralbnormalities in the median nerve,17,18 nerve con-

uction test results may be normal despite the pres-

The Journal of Hand Surgery 919.e1

Page 2: Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome

esatcttcsphCai

fistdtcrtCoica

MA(ttffpbeulottwtt

ttsa

ib5drmrapbac

drspbrt

Fd

919.e2 The Journal of Hand Surgery / Vol. 31A No. 6 July–August 2006

nce of clinically significant median nerve compres-ion. Thus improvement is observed in many patientsfter surgery for CTS despite normal electrodiagnos-ic test results.19,20 Second, it is assumed that nerveonduction velocities are distributed normally withhresholds for abnormality defined as the lower 2.5%o 5% of values; however, the distribution of nerveonduction velocities is skewed substantially towardlower conduction velocities in the asymptomaticopulation,21,22 resulting in a large proportion ofealthy individuals being mislabeled as affected byTS. Third, the cut-off point used for determiningbnormality in the measurement of nerve conductions highly variable.14,23,24

Some clinical experts rely exclusively on clinicalndings and do not use electrodiagnostic study re-ults to make the diagnosis of CTS.25 Thus symp-oms and signs play a major role in establishing theiagnosis of CTS. Clinicians, however, are inconsis-ent in the importance they ascribe to the variouslinical findings.26 Therefore an important step ineducing inconsistencies and misdiagnoses would behe development of standardized clinical criteria forTS that could be applied consistently. The objectivef this study was to establish consensus among clin-cal experts on the most important clinical diagnosticriteria for CTS, give these criteria weighted values,nd then validate the resulting diagnostic instrument.

aterials and Methodsdiagnostic scale for CTS was developed in 3 stages

Fig. 1). For the process of item generation 57 po-ential diagnostic criteria for CTS were identifiedhrough a combination of literature review, key in-ormant interviews, and focus groups with cliniciansrom various clinical backgrounds.27 This pool ofotential diagnostic items was rated for importancey a panel of clinicians who met our criteria for beingxperts in the clinical diagnosis of CTS. An individ-al was considered an expert if he or she had pub-ished reports in peer-reviewed journals on the topicf CTS management or had a national reputation inhe field of peripheral nerve conditions. Ranking ofhese criteria in order of diagnostic importance thusas established using the Delphi technique, a process

hat allowed a group consensus to be establishedhrough repeated assessments of the criteria.28

The decision about the number of items fromhis list that should be included in a final diagnos-ic scale required a balance between comprehen-iveness and feasibility. Examination for common-

lity and redundancy among the 20 highest-ranked s

tems showed that several of these could be com-ined together. For example, the items ranked 1, 3,, and 7 in importance all related to the nature andistribution of the sensory disturbance. Itemsanked 2 and 4 described denervation of the thenarusculature. Items ranked 8, 12, and 19 were

elated to coexisting medical conditions. Items 15nd 16 described the response to common thera-eutic interventions for CTS. This process of com-ining similar clinical items to create broader di-gnostic constructs allowed the top 20 items to beollapsed into 8 major criteria (Table 1).

The relative importance of the final 8 items wasetermined by establishing weights for each crite-ion. The objective was to develop a logistic regres-ion model with the dependent variable being therobability of CTS and the independent variableseing the items in the diagnostic scale. This processequired 2 new groups of 16 clinicians each. None ofhese individuals had participated in any of the earlier

Item generation Literature review

Focus groupsKey informant interviews

57 items

Item reduction

Delphi technique

Ranking of items in order ofdiagnostic importance

Weighting of top 8 items

512 CTS and non-CTS case histories

16 clinical expertsIs this CTS? yes/no

Regression model predicting the probability of CTS

Validation512 CTS and non-CTS case histories16 clinical expertsWhat is the probability of CTS?

VAS scale

igure 1. Summary of the steps in the development of aiagnostic scale for CTS.

tages of the project.

Page 3: Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome

itctaAassatzgtniawphbo

scwewtcsrTwvTho

aTmbdwavpms

ROnaPuBjee

dpcbm0

DAeaitdtdtc

cdc

Graham et al / Diagnostic Criteria for CTS 919.e3

These clinicians examined case histories contain-ng the 8 diagnostic criteria in all possible combina-ions.29,30 This required the creation of 256 uniquease histories. This approach was taken so that all ofhe 8 criteria could be evaluated as they occurredlone in a case or together with the other criteria.nother 256 non-CTS case histories also were cre-

ted to present the clinicians with conditions that areeen commonly in the same patient population pre-enting with CTS or that frequently are misdiagnoseds CTS. These included cases representing cubitalunnel syndrome, de Quervain’s tenosynovitis, trape-iometacarpal joint osteoarthritis, flexor tendon trig-ering, osteoarthritis of the wrist, lateral epicondyli-is, flexor carpi radialis tendinitis, and radial sensoryerve compression. Each case history had a standard-zed format that included a randomly assigned patientge, gender, and occupation except in instances inhich age or gender was intrinsic to the item (eg,regnancy as a diagnostic factor). Two typical caseistories are shown in Appendix A (Appendix A maye viewed at the Journal’s Web site, www.jhandsurg.rg).Each of the 32 clinicians evaluated 16 randomly

elected CTS and 16 non-CTS case histories. Theases were allocated to each rater so that all featuresere uncorrelated and equally represented within

ach set of cases. This approach also ensured thatithin 1 rater, 1 factor would not increase spuriously

he importance of another factor through inadvertentorrelations. This complicated approach was neces-ary to the goal of developing an unbiased logisticegression model to predict the probability of CTS.he use of actual patients rather than case historiesould have introduced a number of uncontrollableariables that would have thwarted this objective.he use of vignettes such as those used in our studyas been shown to be more effective than historiesbtained from patient charts.31

One panel of 16 clinicians made a binary judgments to whether or not each case represented CTS.hese data were used to develop a logistic regressionodel with the dependent variable being the proba-

ility of CTS and the predictor variables being the 8iagnostic criteria. A second panel of 16 cliniciansho evaluated the same case histories rated the prob-

bility of CTS for each case history on a 10-cmisual analog scale (VAS). These ratings were com-ared with the predictions of the logistic regressionodel. The process of reviewing the case histories is

ummarized in Figure 2. c

esultsnly 6 of the 8 criteria (numbness in the medianerve distribution, nocturnal numbness, weakness/trophy of the thenar musculature, Tinel’s sign,halen’s test, loss of 2-point discrimination) contrib-ted significantly (p � .05) to the model (Appendix

may be viewed at the Journal’s Web site, www.handsurg.org). By applying the resulting formula toach of the cases a CTS probability was predicted forach case by the model.

The probability of CTS for each case history pre-icted by the model then was compared with therobability of CTS independently assigned to eachase by the clinicians using the VAS. The correlationetween the probability of CTS predicted by theodel and that estimated by the clinical experts was

.71 (Fig. 3).

iscussionfailing of contemporary clinical medicine in gen-

ral is the lack of clearly stated, widely agreed on,nd formally established diagnostic criteria for many,f not most, medical conditions. Diagnosis, an essen-ial component of clinical decision making, oftenetermines which patients will be treated and whatreatment they will receive. Thus standardization ofiagnostic practices for commonly diagnosed condi-ions is of fundamental importance to the practice oflinical medicine.

The term gold standard, which often is used inlinical medicine, implies a criterion for definitiveiagnosis; however, gold standards exist only byonsensus. If there is no consensus on the diagnostic

Table 1. Final List of Unweighted ClinicalDiagnostic Criteria

Numbness and tingling in the median nerve distributionNocturnal numbnessWeakness and/or atrophy of the thenar musculatureTinel’s signPhalen’s testLoss of 2-point discriminationAmeliorating/exacerbating factors*

Improvement by splinting and/or steroid injectionWorsening with activities such as driving and

strenuous hand useCoexisting medical conditions*

PregnancyDiabetesHypothyroidism

*These factors did not contribute significantly to the logisticregression model.

riteria for a condition then a standard of diagnosis

Page 4: Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome

ctntcess

tfeooCi

ntctdsuTCpbinobcw

ses fo

Fv0

919.e4 The Journal of Hand Surgery / Vol. 31A No. 6 July–August 2006

annot be established. The existence of agreement ishe key issue. This is true even when a tissue diag-osis is made because unless there is agreement onhe meaning of the histologic findings the signifi-ance remains in doubt. In the diagnosis of CTSlectrodiagnostic tests often are used as the goldtandard criterion for this condition. There is no con-ensus on this issue, however, and thus electrodiagnos-

51

sesac STC 652

16 panelists evaluating 32 cases (16 CTS, 16 non-CTS)

Binary evaluation. Is CTS present, yes/no?

Regression analysis to develop weighted diagnostic scale

Figure 2. Schema for dstribution of ca

r = 0.71

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

-3 -2 -1 0 1 2 3

Standardized VAS values of specialist clinicians

Scores predicted by

weighted diagnostic

scale

igure 3. Standardized VAS scores from specialist cliniciansersus scores predicted by the weighted diagnostic scale. r �

m.71.

ic tests cannot be considered a gold standard criterionor the diagnosis of CTS. This is not to assert thatlectrodiagnostic tests never play a role in the diagnosisf CTS but rather to suggest that they cannot on theirwn be considered the criterion for the diagnosis ofTS in all circumstances, which is necessary if these

nvestigations are to be considered the gold standard.Our study has developed standardized clinical diag-

ostic criteria for CTS that have been validated againsthe opinions of clinical experts. The correlation with thelinical experts of 0.71 was substantial, explaining morehan 50% of the variance in the sample. Standardizediagnostic criteria such as those established by ourtudy have many uses. First, standard criteria could besed to determine the population prevalence of CTS.he use of valid standardized diagnostic criteria forTS may decrease the wide variance in populationrevalence caused by variable case definitions. This haseen attempted in the past32 but the case definitionsntended for large-scale epidemiologic studies often areot applicable for clinical use in individuals. The studyf etiologic factors such as workplace exposures woulde aided by the adoption of standardized diagnosticriteria because variability in the identification of CTSould be limited.Second, nonexpert clinicians such as those in pri-

sesac STC-non 652

16 panelists evaluating 32 cases (16 CTS, 16 non-CTS)

VAS evaluation. What is the probability of CTS?

alidation of weighted scale

r weighting criteria in the final scale.

2 cases

V

ary care settings could use standardized criteria to

Page 5: Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome

epibstFtveplWawgbEitriminirptt

ustooepTvfccsrhebatis

bbttabDItetaconoaeste

tlttctcsefttefhac

CmnsstfCtb

Graham et al / Diagnostic Criteria for CTS 919.e5

stablish the probability of CTS in a manner ap-roaching the practices of clinical experts. Depend-ng on the probability of CTS the action taken mighte the initiation of nonsurgical treatment, referral forurgery, further investigation with electrodiagnosticesting, or consideration of an alternative diagnosis.or example, if the probability of CTS is high then

reatment with either nonsurgical or surgical inter-entions probably can be instituted safely withoutlectrodiagnostic testing. At a high level of clinicalrobability for CTS electrodiagnostic tests are notikely to add much greater certainty to the diagnosis.

hen the probability of CTS is very low it may beppropriate to consider a different diagnosis. Evenhen electrodiagnostic test results are positive in thisroup the posttest probability of CTS still is likely toe insufficient to recommend surgical intervention.stimates of probability that are intermediate may

ndicate a need for further investigation with elec-rodiagnostic testing. This is the group in whom theesults of electrodiagnostic testing are most likely tonfluence decision making about treatment. Further-ore electrodiagnostic test data could be interpreted

n a Bayesian context using the output of the diag-ostic instrument as an estimate of pretest probabil-ty. The ability to report the electrodiagnostic testesults as a posttest probability may improve theirrecision beyond the current standard of establishinghe CTS diagnosis based on comparison with ahreshold for nerve conduction velocity.

Third, standardized diagnostic criteria also could besed to establish guidelines for instituting treatmentuch as surgery. For example surgical decompression ofhe carpal tunnel might be considered if the probabilityf CTS exceeds a defined threshold value. This thresh-ld might vary depending on the clinical context—forxample if there has been treatment that failed or if theatient is receiving workers’ compensation benefits.he decision to institute treatment is based on the se-erity of the condition to be treated and a variety ofactors associated with treatment including the risk foromplications, morbidity, and cost. As such it is diffi-ult to determine precisely a critical threshold at whichurgical treatment should be recommended. The crite-ion used to make important clinical decisions shouldave a reliability of at least 0.80.33 It might be consid-red that a CTS probability of at least this level mighte necessary to start treatment. A determination of thectual threshold for recommending different forms ofreatment requires further study; however, the first steps to base the diagnosis of CTS on standardized criteria

o that these comparisons can be made. c

Our study has some limitations. Because they areased on expert opinion these diagnostic criteria maye relevant only to current thinking on the topic ofhe clinical diagnosis of CTS. As knowledge growshe views of experts may change through time and as

result an instrument based on expert opinion mayecome obsolete without constant re-evaluation. Theiagnostic and Statistical Manual of Mental Disorders

V34 is an example of an expert-based diagnostic systemhat has had successive revisions aimed at capturing thevolution of the concepts of mental disease. Althoughhe diagnostic criteria for conditions in the Diagnosticnd Statistical Manual of Mental Disorders havehanged over time, they have been based consistentlyn a consensus of experts. In general although theature of a consensus may evolve, the principle ofbtaining consensus should remain the key consider-tion in establishing diagnostic criteria. It may be nec-ssary to determine periodically whether expert consen-us is changing. As evidence linking outcomes withhese diagnostic criteria accumulates the reliance onxpert consensus probably will decrease.

A second limitation of our diagnostic instrument ishat the validation process was performed using simu-ated case histories created for this purpose. Althoughhe use of clinical vignettes is a more valid approachhan the abstraction of clinical data from patientharts,31 the instrument requires testing in actual pa-ients.35 The reliability of some of the individual clini-al criteria is unknown and may be variable in differentettings; however, the output of the instrument, whichxpresses the diagnosis in probabilistic terms, allowsor flexibility on the part of the clinician. In other wordshe clinician may readjust the probability predicted byhe instrument if aspects of the history or physicalxamination for CTS are equivocal or if there are otheractors not included in the instrument that are thought toave diagnostic implications. These questions can benswered only in studies on patients. These studiesurrently are underway in our center.

We have developed clinical diagnostic criteria forTS based on expert consensus. The diagnostic instru-ent correlates well with the judgments of expert cli-

icians. The diagnostic instrument could be used totandardize the clinical diagnosis of CTS and thishould increase the accuracy of diagnosis, improve theargeting of specific treatments, and facilitate the per-ormance of epidemiologic studies of the risk factors forTS. The role and interpretation of electrodiagnostic

est results in the diagnosis of CTS may be definedetter as a result of the use of standardized diagnostic

riteria. We believe that a formal and methodologically
Page 6: Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome

stw

R1

9va(

aa

3b

R

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

2

3

3

3

3

3

3

919.e6 The Journal of Hand Surgery / Vol. 31A No. 6 July–August 2006

ound approach to developing diagnostic criteria similaro that used in this study could be used in other contextshen consensus on diagnosis is lacking.

eceived for publication November 4, 2005; accepted in revised form March4, 2006.

Supported by the Physicians’ Services Incorporated Foundation (grant7-52), the Mary Trimmer Chair in Geriatric Medicine Research, Uni-ersity of Toronto (G.N.), and the R.B. Salter Chair of Surgical Researchnd an Investigator of the Canadian Institute for Health ResearchJ.G.W.).

No benefits in any form have been received or will be received fromcommercial party related directly or indirectly to the subject of this

rticle.Corresponding author: Brent Graham, MD, Toronto Western Hospital,

99 Bathurst St, FP-174, Toronto, Ontario M5T 2S8, Canada; e-mail:[email protected] © 2006 by the American Society for Surgery of the Hand0363-5023/06/31A06-0008$32.00/0doi:10.1016/j.jhsa.2006.03.005

eferences1. Katz JN, Larson MG, Sabra A, Krarup C, Stirrat CR, Sethi

J, et al. The carpal tunnel syndrome: diagnostic utility of thehistory and physical examination findings. Ann Intern Med1990;112:321–327.

2. Levine DW, Simmons BP, Koris MJ, Daltroy LH, Hohl GG,Fossel AH, Katz JN. A self-administered questionnaire for theassessment of severity of symptoms and functional status in carpaltunnel syndrome. J Bone Joint Surg 1993;75A:1585–1592.

3. Atroshi I, Gummesson C, Johnsson R, Ornstein E, RanstamJ, Rosen I. Prevalence of carpal tunnel syndrome in a generalpopulation. JAMA 1999;282:153–158.

4. de Krom MCTFM, Kester ADM, Knipschild PG, Spaans F.Risk factors for carpal tunnel syndrome. Am J Epidemiol1990;132:1102–1110.

5. Tanaka S, Wild DK, Seligman PJ, Behrens V, Cameron L,Putz-Anderson V. The US prevalence of self-reported carpaltunnel syndrome: 1988 National Health Interview Surveydata. Am J Public Health 1994;84:1846–1848.

6. Katz JN, Larson MG, Fossel AH, Liang MH. Validation ofa surveillance case definition of carpal tunnel syndrome.Am J Public Health 1991;81:189–193.

7. Katz JN, Simmons BP. Clinical practice. Carpal tunnel syn-drome. N Engl J Med 2002;346:1807–1812.

8. Cummings K, Maizlish M, Rudolph L, Dervin K, Ervin A.Occupational disease surveillance: carpal tunnel syndrome.MMWR Morb Mortal Wkly Rep 1989;38:485–489.

9. Margolis W, Kraus JF. The prevalence of carpal tunnelsyndrome symptoms in female supermarket checkers. J Oc-cup Med 1987;29:953–956.

0. Silverstein BA, Fine LJ, Armstrong TJ. Occupational factorsand carpal tunnel syndrome. Am J Ind Med 1987;11:343–358.

1. Hunt TR, Osterman AL. Complications of the treatment ofcarpal tunnel syndrome. Hand Clin 1994;10:63–71.

2. Kessler FB. Complications of the management of carpaltunnel syndrome. Hand Clin 1986;2:401–406.

3. de Krom MCTFM, Knipschild PG, Kester ADM, Spaans F.Efficacy of provocative tests for diagnosis of carpal tunnelsyndrome. Lancet 1990;335:393–395.

4. Nathan PA, Keniston RC, Meadows KD, Lockwood RS.Predictive value of nerve conduction measurements at the

carpal tunnel. Muscle Nerve 1993;16:1377–1382.

5. Boniface SJ, Morris I, MacLeod A. How does neurophysi-ologic assessment influence the management and outcome ofpatients with carpal tunnel syndrome? Br J Rheum 1994;33:1169–1170.

6. Golding DN, Rose DM, Selvarajah K. Clinical tests forcarpal tunnel syndrome: an evaluation. Br J Rheum 1986;25:388–390.

7. Campion D. Electrodiagnostic testing in hand surgery.J Hand Surg 1996;21A:947–956.

8. Wilcox MS, Bilbao A. Sensitivity of electrophysiologicalstudies and the carpal tunnel syndrome. Muscle Nerve 1993;16:1265–1266.

9. Braun RM, Jackson WJ. Electrical studies as a prognosticfactor in the surgical treatment of carpal tunnel syndrome.J Hand Surg 1994;19A:893–900.

0. Grundberg A. Carpal tunnel decompression in spite of nor-mal electromyography. J Hand Surg 1983;8:348–349.

1. Goodgold J. A statistical problem in diagnosis of carpaltunnel disease. Muscle Nerve 1994;17:1490–1491.

2. Robinson LR, Temkin NR, Fujimoto WY, Stolov WC. Ef-fect of statistical methodology on normal limits in nerveconduction studies. Muscle Nerve 1991;14:1084–1090.

3. Jackson DA, Clifford JC. Electrodiagnosis of mild carpal tunnelsyndrome. Arch Phys Med Rehabil 1989;70:199–204.

4. Stevens JC. AAEE minimonograph #26: the electrodiagno-sis of carpal tunnel syndrome. Muscle Nerve 1987;10:99–113.

5. Duncan KH, Lewis RC, Foreman KA, Nordyke ME. Treat-ment of carpal tunnel syndrome by members of the Ameri-can Society for Surgery of the Hand: results of a question-naire. J Hand Surg 1987;12A:384–391.

6. Graham B, Dvali L, Regehr G, Wright JG. Variations indiagnostic criteria for carpal tunnel syndrome among On-tario specialists. Am J Ind Med 2006;49:8–13.

7. Graham B. The development of diagnostic criteria for carpaltunnel syndrome. MSc: University of Toronto, 2003:52–60.

8. Graham B, Regehr G, Wright JG. Delphi as a method toestablish consensus for diagnostic criteria. J Clin Epidemiol2003;56:1150–1156.

9. Gustafson DH, Fryback DG, Rose JH, Yick V, Prokop CT,Detmer DE, Moore J. A decision theoretic methodology forseverity index development. Med Decis Making 1986;6:27–35.

0. Fishburn WR. Experiences-in-being. J Am Coll Health As-soc 1970;18:288–291.

1. Peabody JW, Luck L, Glassman P, Dresselhaus TR, Lee M.Comparison of vignettes, standardized patients and chartabstraction: a prospective validation of 3 methods for mea-suring quality. JAMA 2000;283:1715–1722.

2. Rempel D, Evanoff B, Amadio PC, de Krom M, Franklin G,Gray R, et al. Consensus criteria for the classification ofcarpal tunnel syndrome in epidemiologic studies. Am JPublic Health 1998;88:1447–1451.

3. Bland JM, Altman DG. Statistics notes: Cronbach’s alpha.BMJ 1997;314:572–573.

4. American Psychiatric Association Task Force on DSM-IV. Diag-nostic and statistical manual of mental disorders: DSM-IV. 4th ed.Washington, DC: American Psychiatric Association, 1994.

5. Jones TV, Gerrity MS, Earp J. Written case simulations: dothey predict physician behaviour? J Clin Epidemiol 1990;

43:805–815.
Page 7: Development and Validation of Diagnostic Criteria for Carpal Tunnel Syndrome

APPwdsT

t

tos

v

y

t

sg

t

nsse

vc

a

t

AfP

wbPcx

Graham et al / Diagnostic Criteria for CTS 919.e7

ppendix A. Two Typical Case Historiesrovided to the Panelistsatient number 522 is a 32-year-old male who presentsith a chief complaint of numbness/tingling in theistribution of the median nerve. There are no otherymptoms. The symptoms are made worse by grasping.here has been some improvement with splinting.The past history includes no other relevant condi-

ions.The physical examination shows a positive Phalen’s

est and a Tinel’s sign over the median nerve at the levelf the carpal tunnel The remainder of a comprehen-ive physical examination is within normal limits.

On the basis of only the clinical information pro-ided, does this patient has carpal tunnel syndrome?The rater giving a binary response classified this as

es.The rater responding on the VAS scored the his-

ory as 76.Case number 773 is a 63-year-old male who pre-

ents with a chief complaint of numbness and tin-ling in the ring and small fingers.The past history includes hypothyroidism. The pa-

ient is retired. n

The physical examination shows increased numb-ess with passive flexion of elbow, positive Tinel’sign over cubital tunnel, and normal intrinsic muscletrength. The remainder of a comprehensive physicalxamination is within normal limits.

On the basis of only the clinical information pro-ided, estimate the probability that this patient hasarpal tunnel sydrome?

The rater giving a binary response classified thiss no.

The rater responding on the VAS scored this his-ory as 12.

ppendix B. Logistic Regression Formulaor the 6-Item Model Predicting therobability of CTS

p(CTS) ⁄ 1 � p(CTS) � eb 0 �b1x1 �b2x2 �b3x3 �b4x4 �b5x5 �b6x6

here b0 � �2.14; variable present � 1, absent � 0;

1 � 1.44; x1 � thenar atrophy; b2 � 1.44; x2 �halen’s test; b3 � 1.30; x3 � loss of 2-point dis-rimination; b4 � 1.16; x4 � Tinel’s sign; b5 � 1.16;

5 � nocturnal numbness; b6 � 1.03; x6 � numb-

ess, median nerve distribution.