The Sensitivity and Specificity of Nerve Conduction Studies for Diagnosis of Carpal Tunnel Syndrome: A Systematic Review

Abstract

Background: The utility of nerve conduction studies (NCS) for diagnosis of carpal tunnel syndrome (CTS) has continued to be a subject of debate. Proponents of NCS assume a high sensitivity and specificity; however, many are unaware of the actual literature on this topic and the cutoff values commonly used for diagnosis. The purpose of this systematic review of the literature is to report the sensitivity and specificity of NCS for diagnosis of CTS in various studies. Methods: A literature review of PubMed and EMBASE databases was performed for all articles on NCS for diagnosis of CTS. The outcome of interest was the sensitivity and/or specificity of the NCS distal motor latency (DML) or distal sensory latency (DSL) cutoff value used to diagnose CTS in each study. Results: A total of 3066 total articles were screened and 21 were included in the review after assessment by two independent reviewers. The mean cut-off value for DSL was 3.37 ms (range 2.8-4 ms) and the mean cutoff value for DML was 4.28 ms (range 3.8-4.6 ms). Weighted mean DSL sensitivity was 73.4% and weighted mean DSL specificity was 93.6%. Weighted mean DML sensitivity was 56.2% and weighted mean DML specificity was 95.8%. Conclusions: There is significant variation in the cutoff values used for both DSL and DML. The wide range of cut-off values makes it difficult to interpret the literature, and there is a lack of high-quality studies with control groups using a priori cut-off values for diagnosis.

Introduction

Carpal tunnel syndrome (CTS) is a common entrapment neuropathy occurring due to compression of the median nerve at the wrist. This condition can lead to significant morbidity and may affect one’s daily lifestyle due to symptoms such as decreased grip strength, hand and digit numbness, tingling, and intermittent pain. In order to decrease the burden of this condition on the lives of those affected, it is important to have a modality for diagnosis that is both sensitive and specific. The utility of nerve conduction studies (NCS) for diagnosis of CTS has continued to be a subject of debate. Reasonable physicians may disagree on the utility of these tests for routine cases of CTS. Proponents of the routine use of NCS point to the high sensitivity and specificity reported in the literature. Detractors of their routine use reference studies showing relatively high rates of false-negative and false-positive tests.^1,2 A thorough understanding of the current literature may facilitate further discussions on the utility of NCS for diagnosis of CTS. In particular, understanding common cut-off values used for diagnosis may allow physicians to compare values used in their local NCS labs versus national standards.

The sensitivity and specificity of a diagnostic test depends on several factors. The patient population can greatly influence the accuracy of a diagnostic test with high prevalence populations falsely elevating the sensitivity and specificity. The reference standard, particularly for studies involving NCS, is critical when determining the sensitivity and specificity of the test. A restrictive definition such as atrophy of the thenar eminence will decrease the sensitivity while a lax definition such as hand pain or hand numbness will elevate the sensitivity. Finally, the cut-off values for a positive diagnosis may vary widely and greatly affect sensitivity and specificity. By choosing a lower distal sensory latency (DSL) and/or distal motor latency (DML) cut-off, one can increase the sensitivity of the exam. By choosing a higher value, one can increase the specificity of the exam. Most studies attempt to optimize sensitivity and specificity by using receiver operating curves. When analyzing the literature, it is important to examine these factors.

Despite a large volume of studies reporting the sensitivity and specificity of NCS, many of these studies are of small sample size and are classified into a lower level of evidence. A recent meta-analysis by Chen and colleagues³ identified over 2400 articles pertaining to NCS and the median nerve. However, only 1 sensory and 1 motor article met inclusion criteria for a high-quality study to be included in the analysis. Based on these findings, a systematic review of all studies is indicated. The purpose of this systematic review of the literature is to report the sensitivity and specificity of various NCS cutoff values for diagnosis of CTS in the hopes of guiding future treatment and management.

Materials and Methods

Study Selection

A comprehensive literature search of the EMBASE and PubMed electronic scientific databases was performed using the search terms “carpal tunnel syndrome” and “nerve conduction study OR electrodiagnostic testing.” Articles from this search underwent a two-step screening process by two independent reviewers: (1) article titles and abstracts were screened; and (2) articles which appeared to meet the eligibility criteria underwent full article review.

Eligibility Criteria

Inclusion criteria for articles for this review: (1) English-language articles; (2) levels of evidence I through IV; and (3) inclusion of sensitivity and/or specificity of NCS DML or DSL cutoff value used to diagnose CTS. Exclusion criteria included: (1) non-English-language articles; (2) patients in study with peripheral neuropathy; (3) patients in study with cervical radiculopathy; and (4) non-applicable data.

Data Abstraction/Analysis

A PubMed search of the above-defined search terms on December 26, 2017, yielded 1179 English-language results dating from November 1964 to December 2017. These results were combined with 1887 English-language articles identified via an EMBASE search of the same search terms on January 17, 2018, with articles dating from December 1962 to January 2018. After screening these total 3066 records and removing duplicate articles that were included in both PubMed and EMBASE, 53 articles were fully reviewed for inclusion. From citations in these articles, an additional 7 articles were analyzed, and finally 21 articles were determined to meet the inclusion criteria to be included in this study (Figure 1).^4-24 An independent reviewer extracted the sensitivity and/or specificity of the DML and/or DSL cut-off values as well as the number of hands affected and unaffected by CTS that were involved in each study.

Figure 1.

Record screening process.

Results

The studies included in this review used a variety of methods to determine the cutoff values for diagnosis of CTS. Several studies determined the DML and DSL in a group of control subjects and then used a value of two standard deviations above the control mean as the cut-off value. Other studies used a clinical diagnosis of a physician as the reference standard.

A total of 21 articles met inclusion criteria and reported results on 1149 control hands and 3254 hands with a diagnosis of CTS. A control group was utilized in 14 studies. Sensitivity was reported in all 21 studies and specificity was reported in 7 studies. Both DML and DSL cutoffs were reported in 11 studies while others reported only one of the two.

DSL sensitivity was reported in 19 studies with a total of 1914 CTS hands and 1015 control hands, while DSL specificity was only reported in 6 studies with a total of 504 CTS hands and 460 control hands. The mean cutoff value utilized for DSL was 3.37 ms (range 2.8-4 ms). The weighted mean sensitivity and specificity of the DSL values used were 73.4% and 93.6%, respectively.

DML sensitivity was reported in 19 studies with a total of 2963 CTS hands and 949 control hands, while DML specificity was reported in just 6 studies with a total of 468 CTS hands and 430 control hands. The mean cutoff value utilized for DML was 4.28 ms (range 3.8-4.6 ms). The mean sensitivity and specificity of the DML values used were 56.2% and 95.8%, respectively.

Discussion

This systematic review found the sensitivity and specificity of DSL to be approximately 75% and 93%, respectively. The sensitivity and specificity of DML was found to be approximately 65% and 95%, respectively. One issue with the published literature is that it tends to look at these tests in isolation, rather than combined, as would occur in clinical practice. Regardless, the published literature does document a high specificity for NCS, an important factor for a confirmatory test.

There is significant variation in the cutoff values used for both DSL and DML. The DSL cutoff ranged from 2.8 to 4.0 ms and the DML cut-off ranged from 4.1 to 4.6 ms. The wide range of cutoff values makes it difficult to interpret the literature. Chang et al reported a specificity of 100% using a DSL cut-off of 2.9 ms while Tseng et al reported a specificity of 100% using a DSL cut-off of 4.0 ms. It would be interesting to use a cutoff of 2.9 ms using the clinical series from Tseng et al and determine how this affects the sensitivity and specificity of the test.

One of the largest series in the literature is from the Mayo Clinic, published in 1987.²⁰ In this series, the sensitivity of a DSL cutoff of 3.5 ms was 64% and the sensitivity of a DML cutoff of 4.6 ms was 38%. The authors note that severity matters when determining the diagnostic accuracy of NCS. If the study population is a group of subjects with mostly mild CTS, the sensitivity will be low when compared to a study population of subjects with severe CTS. Despite being the largest series in this systematic review, it is of low quality as it is retrospective, does not have a control group, and the data are not presented clearly.

Kimura¹¹ reported on NCS findings in 122 control hands and 172 CTS hands. Mean DML was 3.6 (SD ± 0.36) ms and mean DSL was 2.82 (SD ± 0.28) ms in control hands. In 172 CTS hands, the mean DML was 4.9 (SD ± 1.03) ms and mean DSL was 3.70 (SD ± 0.59) ms. While these differences appear to be large, closer inspection demonstrates that 2 SD from the CTS mean includes the mean values for the control group. The authors note that based solely on the sensory latency, only 108/172 (63%) of hands had abnormal DSL. When considering both DML and DSL, 159/172 (92%) of hands had abnormal NCS. Moon et al¹⁴ evaluated 80 subjects with a clinical diagnosis of CTS and 80 asymptomatic controls using a DML cut-off of 4.4 ms and a DSL cut-off of 3.5 ms. The authors found a sensitivity of 56% for DML and 45% for DSL. Unfortunately, the authors did not report the specificity of NCS despite having a control group. Khosrawi and Dehghan¹⁰ evaluated 100 hands of healthy volunteers and 64 hands of subjects with a clinical diagnosis of CTS. The authors reported a sensitivity of 87% and specificity of 91% for DSL (using a cut-off of 3.6 ms) and sensitivity of 70% and specificity of 100% for DML (using a cut-off of 4.2 ms).

Chang and colleagues⁷ specifically excluded more severe cases of CTS, including any subjects with thenar atrophy and/or abnormal pin-prick sensation in the fingers. The authors reasoned that more severe cases will always be positive on NCS. The authors note that for the 50 control subjects, all findings were within normal limits for DML and DSL, resulting in a specificity of 100%. In the 100 subjects with clinical evidence of CTS, only 72 subjects had an abnormal DML and 74 had an abnormal DSL. These findings reinforce the tenet that DML and DSL cannot be interpreted in isolation. In subjects where there is clinical suspicion of CTS and there is normal DML and DSL, additional testing is indicated.

Most of the studies included in this series that included a control group used the control group to define the 95% confidence intervals that were then used to determine the cutoff values for the CTS group (Uzar et al,²² Prakash et al,¹⁷ Chang et al)⁸. This will artificially elevate the specificity of NCS as the control group, by definition, was within the 95% confidence interval for a negative diagnosis. Gerawarapong⁹ specifically defined the control group as subjects with normal NCS findings. It is misleading to then report the specificity of the test as 100% when the definition of the group is normal NCS. Atroshi et al²⁵ found a specificity of 82% when using predetermined cutoffs for CTS in a study of subjects in the general population.

This systematic review has highlighted several limitations with the published literature regarding NCSs. First, there is a lack of high-quality studies with control groups using a priori cutoff values for diagnosis. Second, the majority of studies compare new tests to standard DML and DSL, reporting the sensitivity and specificity of the individual tests rather than a combination of the tests as would be reported in clinical practice. Third, many of these studies are performed in centers with a high prevalence for CTS, thereby inflating the sensitivity and specificity of the tests. These limitations should not necessarily be interpreted as a reason to not use NCS, but to encourage high-quality research using contemporary methods to provide a more accurate sensitivity and specificity for NCS.