The purpose of the study was to reliably identify an early stage of diabetic polyneuropathy (DPN) by measuring injury to epidermal nerve fibers. We compared intraepidermal nerve fiber density (IENFD) at the ankle and thigh of 29 diabetic subjects who had no clinical or electrophysiological evidence of small- or large-fiber neuropathy to that of 84 healthy controls. The mean ankle IENFD of diabetic subjects was 9.1+/-5.0 mm and that of controls, 13.0+/-4.8 mm (P<0.001). The thigh IENFD did not differ significantly. The IENFD ratio (thigh IENFD divided by ankle IENFD) was 2.39+/-1.30 in diabetic subjects and 1.77+/-0.58 in controls (P<0.001), indicating a length-dependent reduction of IENFD in diabetics. Ankle IENFD remained significantly lower and the IENFD ratio higher in diabetic subjects after adjusting for age. Two subjects had parasympathetic dysfunction, two had retinopathy, and two early nephropathy. Age, height, weight, duration of diabetes, and average HbA1c did not influence IENFD among diabetic subjects. We used receiver operating characteristic (ROC) curves to describe and compare the utility of various threshold values of ankle IENFD and IENFD ratio for the diagnosis of early DPN. The sensitivity and specificity of diagnosing DPN using ankle IENFD of less than 10 mm were 72.4% and 76.2%, respectively. Thus, asymptomatic diabetics have a measurable, length-dependent reduction of distal epidermal nerves. Analogous to microalbuminuria in diabetic nephropathy, reliable identification and quantitation of nascent diabetic neuropathy may have potential therapeutic implications.