In 74 women with polycystic ovary syndrome, treated for 4 years with metformin (MET) and diet, we prospectively assessed whether, and to what degree, weight loss, reduction of insulin resistance, and amelioration of coronary heart disease risk factors could be sustained. We hypothesized that response to MET-diet would not differ by pretreatment body mass index (BMI) classes <25 (normal), > or =25 to <30 (overweight), > or =30 to <40 (obese), and > or =40 (extremely obese). [table: see text] Metformin-diet was successful in producing stable approximately 8% weight reduction for all 4 years (trend P < .0001). Percentage of reductions in weight on MET-diet was significant (P < .05) and did not differ among the 3 highest BMI categories (> or =40, > or =30 to <40, > or =25 to <30), but were not significant in the normal-weight category (BMI, <25). On MET-diet, median homeostasis model assessment of insulin resistance (HOMA-IR) was 33% lower than entry at 1 year, 50% at 2 years, 51% at 3 years, and 50% at 4 years (trend, P < .0001). On MET-diet, median low-density lipoprotein cholesterol (LDL-C) was 6% lower than entry at year 1, 6% at year 2, 7% at year 3, and 11% at year 4 (trend P < .0001). On MET-diet, median high-density lipoprotein cholesterol (HDL-C) was 3% higher than entry at year 2, 8% higher at year 3, and 11% higher at year 4 (trend P < .0001). Percentage of reductions in HOMA-IR, LDL-C, triglyceride, and systolic blood pressure, and increments in HDL-C did not differ (P > .1) in the 4 BMI categories. By stepwise regression, weight loss was a significant (P < or = .01) positive explanatory variable for reduction in HOMA-IR for all 4 follow-up years. Metformin-diet in women with polycystic ovary syndrome effectively and safely reduces weight and LDL-C while raising HDL-C, and maintains these outcomes stable over 4 years.