Intraepidermal nerve fiber density (IENFD) is a sensitive measure of small fiber injury, and holds promise as a clinical trial endpoint measure. A total of 48 punch biopsies were obtained from 22 patients. Tissue was sectioned and stained with PGP9.5. The relative intertrial variability (RIV) of IENFD measurements for each section and punch made by two different observers was determined (interobserver variability). Intraobserver variability (same observer measuring twice) was determined for 50% of the sections and punches. Sections from 12 punch biopsies were also stained at a second laboratory. The effect of the number of sections counted and processing site on reproducibility was investigated. A total of 223 sections were analyzed. The mean IENFD was 6.7 fibers/mm. Mean (+/-standard deviation) interobserver variability was 9.6%+/-9.4 for each biopsy site and 10.2%+/-11.9 for individual sections. Mean intraobserver variability was 9.6%+/-8.9 for biopsies, and 8.8%+/-9.0 for sections. There was no significant difference in IENFD for tissue stained at different laboratories. Intraclass correlation coefficients were greater than 0.98 for each comparison. There was no relationship between absolute IENFD and reproducibility. Reproducibility was highest when four sections were counted. IENFD measurement is highly reproducible. At least four sections should be analyzed. Reliability does not vary with severity of disease. These findings suggest IENFD may be a useful endpoint measure in future neuropathy treatment trials.