The addition of acarbose to insulin treatment was evaluated in 14 Type 1 (insulin-dependent) diabetic patients assessed conventionally (blood glucose profile and HbA1c measurement) and with an artificial B-cell. Their metabolic control was poor, fasting blood glucose 10.7 +/- 0.3 (+/- SE) mmol l-1, mean daily blood glucose 9.7 +/- 0.3 mmol l-1, and HbA1c 9.6 +/- 0.2% (normal range 5.0-6.1%). They were of normal body weight (body mass index 22.5 +/- 0.3 kg m-2), and were C-peptide deficient (fasting 0.08 +/- 0.02 nmol l-1). In addition to their usual insulin therapy (46.9 +/- 3.5 U day-1 in three pre-meal injections), they received 100 mg acarbose or placebo three times a day for 6 weeks in a randomized double-blind crossover design. On the last day of either acarbose or placebo treatment, the usual insulin therapy was discontinued and an artificial B-cell was used for insulin delivery, programmed for euglycaemia. Placebo or acarbose was continued before meals. Acarbose reduced mean daily blood glucose concentrations (8.5 +/- 0.3 vs 9.7 +/- 0.3 mmol l-1, p = 0.002) and HbA1c levels (8.3 +/- 0.1 vs 9.6 +/- 0.2%, p less than 0.001). A significant reduction in insulin requirement after meals was found with the artificial B-cell, 25.1 +/- 2.5 (first treatment acarbose) and 24.1 +/- 2.9 U (first treatment placebo) with acarbose and 40.0 +/- 2.5 and 35.6 +/- 2.9 U with placebo (p less than 0.001). These results suggest that acarbose could usefully be administered to Type 1 diabetic patients to ameliorate glucose control and reduce insulin requirement.