Peripheral neuropathy in the lower extremity is clinically important in the general population. Polyneuropathy can have a variety of causes, including exposure to toxins, metabolic disorders, or infection. The detection of mild peripheral neuropathy may require careful clinical examination and/or the use of electrodiagnostic testing. The American Academy of Neurology and others have offered consensus definitions of polyneuropathy ¹. Quick and accurate screening and clinical diagnosis rely on high sensitivity and specificity of the methods employed. The sensitivity and specificity of the techniques employed significantly affect the outcomes of clinical screening and are also important for epidemiological research of peripheral neuropathy in the lower extremity.

Several tools and procedures such as electrodiagnostic testing, quantitative sensory tests, physical examination procedures, body diagrams and symptom questionnaires have been employed for screening and epidemiological research of peripheral neuropathy. Each of these tools and procedures has strengths and weaknesses. Both electrodiagnostic testing and quantitative vibratory quantitative sensory tests are highly reproducible and complementary to each other ^2–4. However, the role of quantitative vibratory tests in diagnosis of distal neuropathy needs to be studied further ¹. The advantage of electrodiagnostic testing is that it provides an objective measure of peripheral nerve function, which clinical psychophysical examinations do not offer. Therefore, to detect peripheral neuropathy in the lower extremities, sural nerve conduction testing is considered to be an appropriate tool ⁵. However, electrodiagnostic testing requires specific equipment and training for examiners.

The objective of this study was to assess the agreement between electrodiagnostic testing, monofilament testing, a lower extremity symptom survey and a body diagram in identification of possible peripheral neuropathy in the feet. The analysis for the assessment was carried out in a population of dental professionals.

Subjects were recruited during the Michigan Dental Association (MDA) annual conventions held in 2009 (n=232) and 2010 (n=283). They represented a convenience sample of dental professionals who attended the conventions and were recruited for a gene-environment study that investigated the relationship of nerve conduction tests with urinary and hair mercury biomarkers in dental professionals. Each participant signed a written informed consent approved by the University of Michigan Institutional Review Board (HUM00027621).

Abnormal sural nerve function was defined by two separate criteria used by the University of Michigan Electroneuromyography Laboratory: 1) age-adjusted peak latency >4.1ms (20–60 years old) or >5ms (>60 years old); 2) age-adjusted amplitude ≤ 6μV (20–60 years old) or ≤ 5μV (>60 years old). In addition to age adjustment, the corrected peak latency was defined by adjusting to a standard temperature of 32 °C based on the following formula: latency _corrected=latency _initial− 0.3msec*(32°C-temperature in °C) ⁹. No temperature adjustment was applied to amplitudes, because in our data temperature was not a predictor of the sural amplitude in multivariate linear regression analysis (not shown). Due to missing values (n=56) for foot temperature, the sample size for peak latency was smaller than that for amplitude.

Since nerve conduction and monofilament tests were conducted only on the right foot, all analyses describe results only in the right foot. All analyses were performed using SAS 9.2 (SAS Institute, Inc. Cary, NC). Agreement of electrodiagnostic findings with symptoms consistent with neuropathy in the feet and monofilament results, respectively was assessed by kappa coefficients. Separate kappa coefficients were calculated for all subjects, diabetic subjects, and non-diabetic subjects. Kappa results were interpreted as excellent (>0.75), fair to good (0.40~0.75) and poor (<0.40) ¹⁰. Pearson chi-square tests or Fisher exact tests were performed to assess the association of electrodiagnostic findings with symptoms consistent with neuropathy in the feet and monofilament results. Using sural nerve function (peak latency and/or amplitude) as the gold standard, sensitivity and specificity of various combinations of the other tests (body diagram, symptom questionnaire, and monofilament) were calculated. To reflect the clinical relevance of the nerve function as continuous measurements, we also calculated the mean nerve function stratified by various combinations of the clinical tests discussed above. Normality tests showed the distributions were not normal for temperature- and age-adjusted peak latency (Shapiro-Wilk = 0.98; p<0.0001) and amplitude (Shapiro-Wilk = 0.94; p<0.0001). The mean amplitude and corrected peak latency were compared between test strata using non-parametric Mann-Whitney tests.

The prevalence of self-reported diabetes was approximately 4% (Table 1). Diabetic subjects were significantly older and had higher BMI than non-diabetic subjects. The proportions of positive findings for the various symptom criteria (Table 2) are shown in Table 3. Except for peak latency, the proportions of positive findings in the diabetic subjects were higher than those in the non-diabetic subjects. In general, the proportion of abnormal nerve function findings was usually higher than that of abnormal monofilament results.

The main results were summarized in Table 4. Overall, kappa values (−0.12–0.44) were mostly poor, sensitivity was low, and specificity was high. Monofilament testing appeared to perform slightly better than symptom surveys among non-diabetic subjects. Not surprisingly, results among diabetic subjects were somewhat better than among non-diabetic subjects. However, the number of diabetic subjects was small, so the confidence intervals tend to be broad, and none of the results achieved statistical significance. Chi-square test statistics and p-values showed that significant associations in all subjects occurred only in those combinations whose kappa values were among the highest. In Supplemental Tables 1–3, we also showed kappa values, sensitivity and specificity for the combinations of monofilament findings and symptom results in comparison with abnormal amplitude or peak latency, abnormal peak latency alone, or abnormal amplitude alone, respectively. We observed similar patterns in the low kappa coefficients, high sensitivity and low specificity.

Comparisons of means of nerve function between screening test strata were performed using Mann-Whitney tests (See Supplemental Tables 4–5). Given the small numbers, none of the means differed significantly for either sural amplitude or peak latency among diabetic subjects. Results for all subjects and for non-diabetic subjects were similar. The means of peak latency did not differ significantly for any screening outcome defined purely on the basis of the body diagram with or without symptoms. In contrast, the comparison of mean amplitudes differed significantly in the expected direction for the body diagram and for most combinations of the body diagram with symptoms. The mean peak latency among all subjects with a positive monofilament test (mean = 4.13 ms) was significantly greater than among those with a negative monofilament test (mean = 3.51 ms; p = 0.05). Consistent with the latency results, the mean amplitude among all subjects with a positive monofilament test (mean=6.71 uV) was significantly lower than among those with a negative monofilament test (mean=13.12 uV; p < 0.0001). Despite these differences, the overlap of the distributions of nerve test results when stratified by screening test outcome was considerable (See Supplemental Figures 1a–2b).

We assessed the agreement between electrodiagnostic testing results (sural nerve peak latency and amplitude), monofilament findings and symptoms consistent with neuropathy in the right foot of a non-random convenience sample of dental professionals. Overall, the low kappa coefficients showed poor agreement between electrodiagnostic tests and the other procedures. Kappa coefficients in diabetic subjects were somewhat higher than in non-diabetic subjects.

Using nerve conduction as the gold standard, the sensitivity and specificity of various combinations of the screening tests were mostly low and high, respectively. The results of mean differences in nerve function (See Supplemental Tables 4–5) revealed some significant differences, but the clinical utility of these differences may be limited by the observed considerable overlap of the distributions of ‘normal’ and ‘abnormal’ screening results. These results highlighted the potential influence of nerve function cutoff values and the potential importance of using nerve function as a continuous versus dichotomous outcome in comparing test procedures. Unlike some other clinical procedures that produce binary outcomes, nerve conduction is measured on a continuum. Peripheral neuropathy, defined by measured nerve function, is therefore a continuum. This might have had an impact on the results of poor agreement between the screening procedures, although the direction of such an impact was not clear. However, as was shown in Supplemental Figure 1, there was considerable overlap of the distributions of nerve conduction parameters among those with normal and abnormal screening test results based on monofilaments, body diagrams, or symptoms.

The study examined the agreement between different screening tools and procedures for neuropathy in the lower extremity in a non-random convenience sample of volunteers. A similar study that investigated such agreement in the upper extremity also reported relatively poor agreement between physical examinations, electrodiagnostic findings and symptoms consistent with carpal tunnel syndrome ¹¹. The results of the present study point to a need for further development and evaluation of the methods used to screen for neuropathy in the feet. The low prevalence of positive findings highlighted the challenge of developing a screening tool for peripheral neuropathy for use in non-clinical populations, because positive predictive value and negative predictive value are a function of prevalence, and not just sensitivity and specificity. Overall, the results demonstrated a low discriminatory power between the screening procedures for identifying persons with impaired sural nerve function in a non-random convenience sample of volunteers.