The present study is the first population-based investigation that compared the performance of various nerve conduction tests in the diagnosis of CTS. The results revealed that measurement of median nerve distal motor latency, digit-wrist sensory latency, and wrist-palm sensory conduction velocity as well as comparison of sensory conduction velocity in the carpal tunnel with that in the forearm had similar diagnostic power while median-ulnar sensory latency difference demonstrated higher accuracy.
No consensus exists regarding the type and number of nerve conduction tests needed to establish the neurophysiological diagnosis in CTS. Moreover, there is no consensus on the definition of abnormality [13]. In addition to median nerve motor and wrist-digit sensory latency measurements, numerous new tests have been successively introduced to improve the sensitivity of nerve conduction tests [1]. Performing multiple nerve conduction tests on an individual would increase the likelihood of obtaining a false positive result [14]. Measurements of wrist-palm sensory conduction or median-ulnar comparison have been considered superior to distal motor and digit-wrist sensory latency measurements, particularly in detecting patients with mild CTS [15].
The strengths of the present study lie in its population-based design, the independent clinical and electrophysiological assessments, and the use of ROC curves comparing the overall performance of different diagnostic tests rather than their performance at specific cutoff values. A possible limitation of this study, also shared by almost all previous studies, is the use of the clinical diagnosis of CTS as the criterion standard against which electrodiagnostic tests were assessed. However, no superior criterion standard for the assessment of nerve conduction tests is currently available, and the analysis involved comparison of different tests in the same population. The diagnosis of CTS, which determined the sensitivity of the nerve conduction tests, was made by the examining surgeon before testing and was supported independently by a validated hand diagram. Estimation of specificity was based on the results of nerve conduction tests performed on completely asymptomatic persons randomly selected from a general population. The examining surgeon verified the absence of symptoms in the hands confirming the control persons' responses to the questionnaire.
A discrepancy exists between the results of this study and those of many previous reports regarding the sensitivity and specificity of nerve conduction tests and, in particular, the similar performance of median nerve motor and sensory conduction tests. Although most previous clinical studies have shown sensory conduction tests to be more sensitive than motor conduction tests, this finding has not been consistent [2]. The discrepancy shown might have been caused by several factors such as methodological differences, with the majority of previous studies being based on (1) referred patients with CTS and/or small convenience samples of controls (in those that did include controls) as opposed to random population-based samples, (2) non-blinded as opposed to blinded neurophysiological examinations and interpretations, and (3) comparisons of sensitivity of different tests in patients using certain cutoff values as opposed to comparison of their overall performance in persons with CTS and controls. The cutoff value would determine the sensitivity of a certain test; for distal motor latency the cutoff values used in different studies have ranged from 3.8 ms to 4.6 ms [4]. The requirement, stated in the practice parameter, that each neurophysiological laboratory should have its own "reference values" [5] does not facilitate a standardized assessment of diagnostic tests. Difference in the demographic characteristics and disease severity among the patient populations studied is yet another factor that could explain the differences in the reported sensitivity for nerve conduction tests in CTS. Although the present study, like other studies that have shown moderate sensitivity for nerve conduction tests, may face the frequently stated argument that the "most sensitive" tests were not used, this argument ignores the limitations of these diagnostic tests [13].
The performance of nerve conduction tests in this population-based investigation does not necessarily apply to their performance in the diagnosis of CTS in clinical settings. The sensitivity estimates in the present study may apply to populations that are similar to the sample examined, being representative of a general population with a wide spectrum of disease severity. The results of the tests in this population showed few persons with absent sensory responses (Table 1), more commonly seen in patients with severe CTS.
Although it is recognized that patients with typical CTS might have normal nerve conduction test results, the exact size of this group has not been specifically investigated. In a recent consensus report on CTS, the authors stated that consensus could not be reached regarding the patients with typical symptoms but normal nerve conduction test results [16]. In two prospective clinical studies that used the clinical diagnosis of CTS as well as complete symptom relief after surgery as the criterion standard, the sensitivity of nerve conduction testing was 78% and 74%, respectively [17, 18]. The statement made in the practice parameter that the sensitivity of nerve conduction tests exceeds 85% at a specificity of 95% [5] appears to lack strong supportive evidence; this can only be determined with a high degree of certainty by performing appropriately designed randomized clinical studies and testing large population random samples of asymptomatic persons [13].
The present study showed a relatively high level of false positive test results (18%) for nerve conduction tests using a cutoff value considered optimal according to the ROC analysis. The most common method of estimating specificity has been to perform nerve conduction tests on a group of "normal" persons who in almost all previous studies have been recruited among hospital employees or similar convenience samples. The normal values have then been calculated based on the mean values plus 2 to 3 standard deviations recorded in these control groups. Owing to the nature of this method, high specificity is obtained and values of 95% to 97% are usually reported [5]. However, when a test the specificity of which had been calculated with this method in a control group is then used in a different population of "normal" persons, the specificity obtained might be markedly worse.
False positive nerve conduction test results in CTS have been reported previously. In a population-based study from England, a mail survey of 1000 persons (age, 18 to 75 years) found that symptoms of numbness, tingling, or pain in at least 2 of the 3 radial fingers were reported by 18% of the 648 responders [19]. Electrophysiological examination of about half of the symptomatic persons and of 40 randomly selected asymptomatic persons showed median neuropathy (defined as distal motor latency above 4.5 ms or wrist-digit sensory latency above 3.7 ms) in only 18% of those with symptoms of CTS and in 20% of the asymptomatic persons [19]. In a study of 50 asymptomatic persons (mean age, 34 years) the specificity of median-ulnar wrist-palm latency difference of 0.5 ms was reported to be 100% [14]; the use of the same test and cutoff value in a subsequent study of 1021 job applicants gave false positive test results in 16% [20]. In another study of 824 workers (mean age, 38 years) the same test and cutoff value also showed abnormal test results in 16% of the asymptomatic workers (cutoff value of 0.8 ms also gave false positive test results) [21]. Surprisingly, in the recent extensive literature review [4], on which the updated practice parameter was based, the above-cited studies that have shown a high rate of abnormal nerve conduction test results among completely asymptomatic persons were not included.
In the present study, a cutoff value derived from the standard deviations calculated for the control sample would have yielded a markedly worse sensitivity. In the absence of a universally accepted criterion standard for the diagnosis of CTS and as in other non-perfect diagnostic tests, a cutoff value should be chosen that yields sensitivity and specificity levels that can be considered acceptable for clinical purposes. The results of the ROC analyses showed that, with acceptable levels of specificity, the sensitivity of nerve conduction tests would not be very high. These findings, similar to a previous report [22] support the view that the diagnosis of CTS ought to be clinical, with nerve conduction tests used to provide objective evidence when necessary, and to support the diagnosis in less typical cases. The findings are, however, based on a clinical diagnosis made by an experienced hand surgeon. Because, in clinical practice, physicians with varying experience manage patients with CTS, the diagnostic role of nerve conduction tests becomes more important.
Relying solely on the clinical examination in making the diagnosis of CTS might lead not only to missing the diagnosis of CTS in some patients but also to incorrect diagnosis and unnecessary surgery in others. On the other hand, relying solely on nerve conduction tests might lead to some patients who actually have CTS being denied surgical treatment because of their normal test results. Despite the limitations of nerve conduction testing, it is the only tool currently available that can provide direct evidence of median neuropathy at the carpal tunnel to strongly support the diagnosis of CTS in symptomatic patients. Nerve conduction testing also provides an assessment of the severity of median neuropathy, which can be helpful in making decisions concerning type of treatment. Furthermore, in clinical research assessing treatment efficacy in CTS, nerve conduction testing can be useful as part of the inclusion criteria to improve the overall diagnostic certainty; this can minimize the risk of misclassification that reduces the study's ability to detect true differences. However, the results of such studies may not be generalized to patients who are diagnosed on clinical grounds only, which is probably more common in clinical practice.
Predictive values of diagnostic tests are dependent upon disease prevalence. The low positive predictive value for nerve conduction testing is not unusual in a disease with a prevalence rate of less than 5% [8] in the general population. This suggests that nerve conduction tests are not suitable for screening purposes among non-patient populations. The positive predictive value would be much higher when the tests are used on patients presenting with symptoms of numbness and tingling in the hands.
Besides possessing higher diagnostic accuracy than that of isolated median nerve motor and sensory conduction tests, measurement of median-ulnar latency difference has the advantage of being less influenced by factors such as age, height, weight, and hand temperature. It has been suggested that ulnar nerve involvement might be present in CTS as shown by symptoms and sensory loss involving the ulnar fingers and the resolution of these findings after carpal tunnel release [23, 24]. The similarity of ulnar nerve motor and sensory latencies among the persons with clinically certain CTS and the controls shown in the present study suggests that ulnar nerve conduction abnormalities in association with CTS are uncommon.
The standardized hand symptom diagram [6, 7] demonstrated high sensitivity but poor specificity in the diagnosis of CTS among this population-based sample of persons with numbness and tingling in the hand. Consequently, if the diagram is used among populations for the purpose of screening for CTS, the final diagnosis ought to be confirmed through a detailed history taken by a physician.
In conclusion, nerve conduction testing has moderate sensitivity and specificity and a low positive predictive value in population-based CTS. Among the various motor and sensory nerve conduction tests, measurement of median-ulnar sensory latency difference appears to have the highest diagnostic accuracy.