One hundred and ninety patients with symptomatic diabetic peripheral neuropathy took part in a double blind multicentre trial of either placebo or tolrestat 200 mg once daily for 6 months. Painful and paraesthetic symptoms, vibration sensory threshold, and nerve conduction velocity (NCV) were assessed as efficacy end-points during the trial. There was an equally marked improvement of painful symptoms during the trial in the tolrestat and placebo groups. A difference in the improvement of paraesthetic symptoms was found however in favour of the placebo group at 24 weeks (p less than 0.02). The deterioration in mean vibration threshold of the tolrestat group was less than placebo at 24 weeks at all 3 sites measured, and reached significance at the carpal site (p less than 0.05). Significant improvements in median motor NCV and in the mean NCV of the four motor nerves were also seen in tolrestat treated patients at 24 weeks compared to placebo (p less than 0.05). In addition, significant changes in favour of tolrestat were seen when the number of motor nerves per patient with NCV increased during the trial was analysed (p less than 0.001). Concordance analysis of patients with increased mean motor NCV and improvement in painful symptoms demonstrated a positive effect for tolrestat compared to placebo (p less than 0.02). Mild reversible elevations of hepatic transaminases were seen in a few patients treated with tolrestat, with no other significant adverse effects. Tolrestat may therefore be helpful in diabetic peripheral neuropathy, where there is little opportunity for therapeutic intervention apart from effort to achieve normoglycaemic control.