Short-term trials with the antioxidant thioctic acid (TA) appear to improve neuropathic symptoms in diabetic patients, but the long-term response remains to be established. Therefore, Type 1 and Type 2 diabetic patients with symptomatic polyneuropathy were randomly assigned to three treatment regimens: (1) 2 x 600(mg of TA (TA 1200), (2) 600)mg of TA plus placebo (PLA) (TA 600) or (3) placebo and placebo (PLA). A trometamol salt solution of TA of 1200 or 600 mg or PLA was intravenously administered once daily for five consecutive days before enrolling the patients in the oral treatment phase. The study was prospective, PLA-controlled, randomized, double-blind and conducted for two years. Severity of diabetic neuropathy was assessed by the Neuropathy Disability Score (NDS) and electrophysiological attributes of the sural (sensory nerve conduction velocity (SNCV), sensory nerve action potential (SNAP)) and the tibial (motor nerve conduction velocity (MNCV), motor nerve distal latency (MNDL)) nerve. Statistical analysis was performed after independent reviewers excluded all patients with highly variable data allowing a final analysis of 65 patients (TA 1200: n = 18, TA 600: n = 27; PLA: n = 20). At baseline no significant differences were noted between the groups regarding the demographic variables and peripheral nerve function parameters for these 65 patients. Statistically significant changes after 24 months between TA and PLA were observed (mean +/- SD) for sural SNCV: +3.8 +/- 4.2 m/s in TA 1200, +3.0+/-3.0m/s in TA 600, -0.1+/-4.8m/s in PLA (p < 0.05 for TA 1200 and TA 600 vs. PLA); sural SNAP: +0.6+/-2.5 microV in TA 1200, +0.3+/-1.4 microV in TA 600, -0.7 +/- 1.5 microV in PLA (p = 0.076 for TA 1200 vs. PLA and p < 0.05 for TA 600 vs. PLA), and in tibial MNCV: +/- 1.2 +/- 3.8 m/s in TA 1200, -0.3 +/- 5.2 m/s in TA 600, 1.5 +/- 2.9 m/s in PLA (p < 0.05 for TA 1200 vs. PLA). No significant differences between the groups after 24 months were noted regarding the tibial MNDL and the NDS. We conclude that in a subgroup of patients after exclusion of patients with excessive test variability throughout the trial, TA appeared to have a beneficial effect on several attributes of nerve conduction.