OBJECTIVE: To assess adherence to prescribing guidelines, continuation rates, population effects on glycemic control, and occurrence of lactic acidosis during the first 20 months of the availability of metformin in a large health maintenance organization. RESEARCH DESIGN AND METHODS: A retrospective cohort study was performed in the 90,000-member diabetes registry of Kaiser Permanente, northern California. Principal study measures were the proportions of patients started on metformin who met prescribing guidelines (previously on sulfonylureas, HbA1c, obesity, creatinine), the change in HbA1c at 6 months after starting metformin, and hospitalization rates for lactic acidosis. RESULTS: A total of 9,875 patients received metformin during this interval. At least 74% were previously treated with sulfonylureas alone, 81% had baseline HbA1c > or = 8.5%, 71% were obese, and 99% had a serum creatinine < or = 1.5 mg/dl. Among patients on sulfonylureas at baseline, those starting metformin had significantly lower HbA1c levels 6 months later than those not started, after adjustment for age, sex, and the higher baseline levels in those started (adjusted difference: 0.5%, P < 0.0001). Patients starting metformin as initial monotherapy also improved significantly, but patients previously treated with insulin (with or without sulfonyl-ureas) had slightly higher follow-up HbA1c levels than similar patients not starting metformin. Continuation of metformin at 12 months was significantly higher for patients previously treated with sulfonylureas than other groups. One probable case of lactic acidosis was identified during 4,502 person-years on metformin. CONCLUSIONS: Adherence to prescribing guidelines was relatively high during metformin's first 20 months of availability. Glycemic control improved substantially for patients previously treated with sulfonylureas. Lactic acidosis was rare.