To compare the therapeutic effects of the alpha-glucosidase inhibitor miglitol (BAY m 1099), the sulfonylurea glibenclamide, and placebo on parameters of metabolic control and safety in patients with NIDDM that is inadequately controlled by diet alone. After a 4-week placebo run-in period, 201 patients in 18 centers in 4 countries were randomized in a double-blind manner to miglitol (50 mg t.i.d., followed by 100 mg t.i.d.), glibenclamide (3.5 mg q.d/b.i.d.), or placebo for 24 weeks. Efficacy criteria were changes from baseline of HbA1c, fasting and postprandial blood glucose and insulin levels, body weight, and serum triglycerides. Efficacy was assessed in 119 patients who completed the full protocol, and the results were similar to those obtained in 186 patients who fulfilled the validity criteria for analysis. Compared with placebo, mean baseline-adjusted HbA1c decreased by 0.75% (P = 0.0021) and 1.01% (P = 0.0001) in the miglitol and glibenclamide treatment groups, respectively. Blood glucose decreased slightly in the fasting state and considerably in the postprandial state in both treatment groups but not in the placebo group. Fasting insulin levels increased slightly (NS) in all treatment groups; however, postprandial insulin levels decreased with miglitol, while increasing markedly with glibenclamide (P = 0.0001 between all treatment groups). Gastrointestinal side effects (flatulence and diarrhea) occurred mostly in the miglitol-treated patients, while some glibenclamide-treated patients had symptoms suggestive of hypoglycemia. Miglitol monotherapy is effective and safe in NIDDM patients. Compared with glibenclamide, it reduced HbA1c less effectively and caused more gastrointestinal side effects. On the other hand, glibenclamide, unlike miglitol, tended to cause hypoglycemia, hyperinsulinemia, and weight gain, which are not desirable in patients with NIDDM.