Response to Stevens and Harper

We thank Stevens and Harper (1) for their response to our study and for identifying citations that help to highlight the important issues associated with research in this field. The aim of our study (2) was to determine the best standard nerve conduction study (NCS) marker of clinical carpal tunnel syndrome (CTS), in diabetic subjects, under the hypothesis that at least one parameter would reliably identify CTS. Surprisingly, the results indicated that standard NCS techniques fail to reliably distinguish the presence or absence of CTS in subjects with diabetes. From the results, we were able to infer that the electrophysiological changes at the wrist most likely arise from diffuse nerve injury associated with diabetes rather than signifying the specific symptomatic entrapment of the median nerve. Stevens and Harper express two main concerns with the study: the methods of electrophysiological evaluation and the potential for misclassification bias in clinical CTS cases.

Stevens and Harper recommend and cite specific methods for which they suggest exclusion from the NCS evaluation invalidates our study results. We strongly feel, however, that the recommended methods either do not diverge significantly from our protocol or do not have reasonable support from the literature. Further ratio testing such as median-to-radial latencies may be of interest, although similar techniques using median-to-ulnar and median-to-sural ratios were conducted in our study and were nondiscriminatory for symptomatic CTS. In this regard, neither were median-to-ulnar motor latency ratios. The authors cite the study of Vogt et al. (3), in which the results appear contradictory: the difference in abductor pollicis brevis and abductor digiti minimi (ADM) latencies, which might be expected to be consistent with the more divergent difference in lumbrical and interosseous latencies (l-I DIFF), were minor. Furthermore, it appeared that a diffuse increase in ADM and lumbrical latencies existed across groups whether CTS or diabetic polyneuropathy (DPN) were present exclusively or in combination. l-I DIFF values are associated with large variance, and statistical comparisons were not provided. Furthermore, the comparison relevant to this discussion was underpowered (only 30 subjects with DPN and 22 subjects with DPN and coexistent symptomatic CTS were studied). Although this observation may be a very important one, the study from which it arose is not sufficient to suggest that a study excluding this technique is invalid or inappropriate. The study by Hansson (4) compared segmental median nerve conduction measurements between three mutually exclusive groups: those with CTS and without diabetes, those with diabetes and CTS and without DPN, and those with DPN excluding coexistent CTS. Although we partially evaluated segmental conductions (at the elbow and the wrist) in our study, we feel it necessary to highlight this article as one that we strongly considered in our study design—we aimed to achieve a method that would discriminate CTS regardless of diabetes or DPN status and thus made no effort to create an artificial case-control comparison in which conditions were mutually exclusive. We feel that the studies by Vogt et al. and Hansson have provided very important contributions, but they did not offer convincing evidence for us to undertake methodology that diverges from standard techniques. Furthermore, our study aimed to answer a separate question altogether: in a population of diabetic patients with varying degrees of DPN, can NCS discriminate CTS findings? Excluding or minimizing those with both conditions serves to limit the generalizability of the clinical method we aimed to identify. We do, however, encourage investigators to consider nonstandard techniques in the context of clinical studies that are similarly generalizable, but, based on our study results, remain skeptical.

Drs. Stevens and Harper very appropriately raise the issue of misclassification bias. The inclusion criteria in our report are based on those recommended by the American Academy of Neurology (5) for diagnosis of CTS, and thus were considered appropriate. The requirement that four of six clinical criteria be present to establish a diagnosis of clinical CTS, with the requirement in all subjects for a discrepancy between lower limb and upper limb paresthesiae, are reasonable in this clinical context. As Stevens and Harper note, symptoms of CTS, even in those without diabetes, vary widely (6), and thus a diagnostic instrument for clinical CTS requires the ability to select from a list of criteria. Our criteria included the predominance of radial digit involvement, the presence of nocturnal awakening, and the precipitation of paresthesiae by activity as suggested. The presence of cervical or shoulder pain with radiation and of ulnar nerve palsies were excluded based on clinical neurological evaluation. Although we agree that misclassification bias is a potential factor in the negative study results, we feel comfortable that its contribution is minor for the reasons stated, and we note further that the criteria we used compare favorably with the criteria used in the references cited by Stevens and Harper. Furthermore, increasing the stringency for case definition may serve to create misclassification of control subjects (and thus a biased and deceptively elevated specificity), and restriction of case and control subjects would have even further limited the generalizability of a potential diagnostic method.

In summary, the criticisms and citations raised by Stevens and Harper highlight the difficulty in designing a diagnostic study for a syndromic diagnosis without an unequivocal gold standard. The usual selection of subjects limits generalizability of results and thus fails to answer broad clinical questions. We therefore remain comfortable with and stress the importance of our inferences: the ubiquitous nature of nerve injury, and not the absence of distal median nerve changes, defeated the diagnostic ability of NCS to identify symptomatic CTS in this diabetic population.