Electrodiagnosis of diabetic peripheral polyneuropathy. A multivariate analytic approach

Traditional univariate comparisons of nerve conduction data against standard norms may produce conflicting estimates of the presence or absence of a diabetic neuropathy, depending upon the data obtained and the specific nerves sampled. Alternatively, a multivariate analytic approach, using discriminant functions, provides a useful single measure of the degree of neuropathy determined as a weighted combination of the available data. The weights are derived from the linear discriminant function, which maximizes statistical separation of diabetic and nondiabetic subject groups. In this study, 12 electrophysiologic attributes are used to generate a single discriminant function that clearly separates diabetic from nondiabetic subjects and is interpretable as a neuropathic index. Each individual's index of diabetic neuropathy (I) can be quantified as follows: (Formula: see text) where Ai is the original electrophysiologic attribute (i = 12), ai is the coefficient for each attribute that defines the discriminant function and C is a constant specific for that function. For the first time, the degree of diabetic neuropathy can thus be quantified for purposes of comparison and correlation with other quantifiable clinical/somatic measures of diabetes. The index allows for a higher percentage of type II diabetic patients to be classified as neuropathic than previously described and enables determination of degree of neuropathy is affected individuals by an interpolative method.