This 5-year surveillance study assessed the tolerability and safety of acarbose in patients with diabetes. A total of 2035 patients were enrolled of whom 1954 were classified as having Type 2 diabetes. The study was open with no control group. Physicians had sole control of the acarbose doses prescribed. Fasting blood glucose levels, 2-h postprandial glucose levels, HbA(1) or HbA(1c) and other clinical parameters, such as lipids and liver enzyme levels, were also assessed as measures of efficacy and safety. One-third of the patients received acarbose as monotherapy and two-thirds in combination with other glucose-lowering treatment. The concomitant diseases were also assessed. Doses of acarbose were low in the majority of the patients and well tolerated. The incidence of acarbose-associated side effects was 4.7%. No sustained adverse changes in laboratory measures occurred. Over the 5 years, HbA(1) and glycated haemoglobin (HbA(1c)) decreased by 2.4 and 1.8% points, respectively, and the mean fasting glucose and 2-h postprandial glucose decreased by 2.7 and 3.4 mmol/l. Mean body weight was reduced by 0.9 kg. The results suggest that when used in long-term day-to-day management of diabetes, acarbose is well tolerated and can improve glycaemic control as monotherapy, as well as in combination therapy. In a high-risk patient group acarbose proved to be a safe drug.