Methods

We studied a cohort of 1108 newly employed workers in several industries in St. Louis, USA, between July 2004 and October 2006.6 7 Subjects were eligible if they were newly hired workers >18 years old, starting a new full-time job (>30 h/week) or changing from temporary to permanent employment status. Subjects were excluded if they had a current or previous diagnosis of CTS or peripheral neuropathy, if they reported a contraindication to nerve conduction studies (NCS), or were pregnant. Pregnant women were not recruited because nerve conduction results are commonly altered by pregnancy. Recruitment occurred during employee orientations, classes at apprenticeship programmes or at the time of employer-mandated post-offer, pre-placement screening. Our subjects were recruited from eight employers and three trade unions, representing manufacturing, construction, biotechnology and healthcare. Testing included a symptom questionnaire, structured physical examination and NCS of the median and ulnar sensory nerves bilaterally. Finger symptoms occurring more than three times or lasting more than 1 week in the last year were classified in three levels of specificity for CTS: (1) any symptoms in the fingers; (2) finger symptoms of burning, pain, numbness or tingling (“specific symptoms in fingers”); (3) distribution of symptoms resulting in a rating of “classic” or “probable” for CTS on a Katz hand diagram.6 8 The physical examination included two common provocative tests for CTS (Phalen's and Tinel's tests), and Semmes–Weinstein sensory testing using a 2.83-mm monofilament to determine sensitivity to light touch. NCS were performed with the NC-Stat automated nerve conduction testing device (NEUROMetrix, Inc., Waltham, MA, USA) and a conservative definition for median nerve abnormality (sensory median–ulnar latency difference ≥97.5th percentile, or median distal sensory or motor latency >99.8th percentile).9

Our referent case definition of CTS (“gold standard”) required a combination of median nerve abnormality and associated symptoms. Symptoms were classified as “classic” or “probable” on a modified hand diagram described by Katz et al.2 We compared sensitivities and specificities, pre-test and post-test probabilities, and likelihood ratios10 of three diagnostic strategies for each hand compared with the case definition reference: (1) symptoms alone, (2) physical examination testing alone and (3) physical examination only if symptoms were present.

Likelihood ratios are an alternative statistic for summarizing diagnostic accuracy. Likelihood ratio is the ratio of the probability of a specific test result in people who do have the disease divided by the probability in people who do not. It can be calculated easily from sensitivity and specificity: for a positive test, sensitivity divided by 1−specificity; for a negative test, 1−sensitivity divided by specificity. Calculation of post-test probabilities was made by applying the likelihood ratio to the pretest probability (prevalence) of disease (box 1).11

Results

Thus, 2970 potentially eligible workers were invited to join the study and 1108 (37.3%) participated. The cohort included 435 apprentice construction workers, 478 hospital workers (mostly in housekeeping, food service or laboratory positions), 158 workers in computer or laboratory jobs and 37 in other positions. There was wide variability in previous work history, with 258 different job titles reported for the job held before our study.7 The study group was 65.1% male, with a mean age of 30.8 years (SD 10.3) and a mean body mass index of 28.5 (SD 6.6).

The prevalence of CTS based on specific symptoms and median nerve abnormalities at the nerve conduction study (reference) in our cohort was 1.2% for the right hand and 1.0% for the left hand. As expected, physical examination alone did not accurately predict CTS; any symptoms in the fingers had higher post-test probabilities and likelihood ratio than physical examination alone (table 1). The likelihood ratio of physical examination was only slightly improved by testing those subjects with finger symptoms. The post-test probability of positive testing, even in subjects with specific symptoms, was far less than 50% (best=33%). If a confirmatory NCS testing strategy was based on questionnaire findings of specific symptoms, it would have led to 97 of 1108 workers receiving NCS (8.8%, 90 right hand, 59 left hand, two hands with CTS missed); if based on specific symptoms and Semmes–Weinstein sensory testing, it would have led to 51 NCS (4.6%, 47 right hand, 22 left hand, 10 hands with CTS missed). Although specificity was high for many tests, there was a low likelihood ratio for negative testing due to the low pretest probability of disease (prevalence), meaning that the test did not meaningfully alter the post-test probability of disease.

Discussion

Our study found that physical examination, alone or in combination with symptoms, was not predictive of CTS. Post-test probabilities for all physical examination strategies were far less than 50% when compared with a CTS case definition requiring typical symptoms and median nerve abnormalities. Although some authors have concluded that physical examination is useful for CTS surveillance in epidemiologic studies,3 4 12 our results and other population-based studies of CTS suggest that physical examination is minimally useful in the case definition of CTS in population-based research studies.2 4 8 13 14 We suggest using specific symptoms as a first level screening tool, and NCS as a confirmatory test (reference), as a testing strategy for CTS in epidemiologic studies. Actually, our study does not address the utility or yield of physical examination manoeuvres in a clinical setting. There are many reviews, books and papers about the utility of physical examination for the diagnosis of CTS in the clinical setting.15–19 Because our data come from a large cohort of workers who had not sought clinical care, our data are relevant to the question of case definition in population studies. These results were based on screening a large population for clinically unreported CTS, and our results may not be applicable to clinical or compensation settings, where the prevalence and severity of disease are higher than in our study population.