Objective To test the efficacy and tolerability of metformin. Research Design and Methods An 8-mo double-blind placebo-controlled parallel-group trial was performed at University hospital diabetic clinics on 60 patients with non-insulin-dependent diabetes mellitus (NIDDM) treated by diet alone. Metformin was administered and built up to a maximum dosage of 1 g three times daily. Results Mean HbA1 fell from 11.7 ± 0.4 to 10.3 ± 0.4% (means ± SE) on metformin but rose from 11.8 ± 0.4 to 13.3 ± 0.4% on placebo (P 1 than placebo without weight gain or significant unwanted effects.