The effects of beraprost sodium (BPS), a stable prostacyclin analogue, on the tail flick (TF) latency were investigated in streptozotocin-induced (STZ;55mg/kg, i.p.) diabetic male Sprague-Dawley (SD) rats and in spontaneously diabetic WBN/Kob rats. The SD rats were divided into 5 groups, i.e., (I) normal, (II) diabetic control, diabetic treated with (III) BPS (10 micrograms/kg/day), (IV) BPS (30 micrograms/kg/ day), and (V) aldose reductase inhibitor (ARI; epalrestat, 50 micrograms/kg/day). The drugs were administered orally. At 12 weeks, TF latency was significantly longer in untreated diabetic rats than in normal rats. After 4 weeks treatment, BPS significantly improved the abnormality in TF latency dose-dependently. But ARI did not normalize the response. The 45 weeks male WBN/Kob rats were divided into 2 groups: diabetic control and diabetic treated with BPS at 30 micrograms/kg/day, p.o., respectively. Untreated, age-matched wistar rats were served as the normal group. At 61 weeks, TF latency was significantly longer in control WBN/Kob rats than in normal wistar rats in time-dependent manner. After 16 weeks treatment, BPS significantly normalized the prolongation in TF latency. In in vivo experiments, BPS significantly increased the cyclic AMP (cAMP) content in sciatic nerves from normal rats dose-dependently. In STZ-induced diabetic rats, cAMP content in sciatic nerves were significantly reduced, and 4 weeks treatment of BPS significantly restored this reduced cAMP content. It was suggested that BPS may be effective on diabetic neuropathy by, at least in part, maintenance of cAMP contents in the nerves.