Acarbose, an alpha-glucosidase inhibitor, delays absorption of carbohydrate in the gut, thereby lowering postprandial glucose levels. Safety data on this drug have been gathered in a series of studies on animals and in extensive clinical trials in humans. Although an initial long term feeding study in rats showed an excess of renal tumours at very high dosages of acarbose (up to 300 mg/kg bodyweight daily), further evaluation with similar studies in rats, hamsters, and dogs indicated that the problem was related to carbohydrate malabsorption. With adequate glucose intake and in gavage studies, no difference in tumour incidence between placebo- and acarbose-treated groups was seen. From 1976 to 1989, safety data on acarbose were obtained in approximately 8800 patients in 2 separate groups of clinical trials, the Bayer International Clinical Data Pool and the American phase III trials. Almost all adverse experiences, as reported by 56 to 76% of patients on acarbose vs 32 to 37% of patients on placebo, were related to the digestive system and included diarrhoea, flatulence, bloating and nausea. Most symptoms were of mild to moderate intensity and tended to improve with time. In the American trials a small but significant increase in liver transaminases was seen, 3.8% in acarbose-treated patients vs 0.9% in controls together with a 1% increase in anaemia in the acarbose group. Overall, acarbose was well tolerated and the adverse experience profile was clinically acceptable.