In the last few years some studies assessed the effects of attenuation of hyperinsulinemia and insulin resistance, obtained by insulin sensitizing agents, in women with polycystic ovary syndrome (PCOS), suggesting potential scope for these drugs in treating the whole spectrum of reproductive, endocrine, and metabolic abnormalities found in such subjects. However, the results of these studies, mostly uncontrolled and short-term, are still inconclusive, and there is no long-term follow-up. In the present study, 23 PCOS subjects [mean (+/- SE) body mass index 30.0+/-1.1 kg/m2] were randomly assigned to double-blind treatment with metformin (500 mg tid) or placebo for 6 months, while maintaining their usual eating habits. Before and after treatment, menstrual history, endocrine and metabolic profiles, serum 17-hydroxyprogesterone response to GnRH-agonist testing, and insulin sensitivity measured by the glucose clamp technique were assessed. Eighteen of these women, as well as 14 additional PCOS patients, were subsequently given metformin in an open trial for 11.0+/-1.3 months (range 4-26), to assess long-term effects of treatment and baseline predictors of metformin efficacy on reproductive abnormalities. After metformin treatment, mean frequency of menstruation improved (P = 0.002), due to striking amelioration of menstrual abnormalities in about 50% of subjects. Women given metformin showed reduced plasma insulin (at fasting: P = 0.057; during the clamp studies: P<0.01) and increased insulin sensitivity (P<0.05). Concurrently, ovarian hyperandrogenism was attenuated, as indicated by significant reductions in serum free testosterone (P<0.05) and in the 17-hydroxyprogesterone response to GnRH-agonist testing (P<0.05). No changes were found in the placebo group. Only comparable minor changes in body mass index were found both in the metformin group and in the placebo group. In the open, long-term trial 17 women (54.8%) showed striking improvements of their menstrual abnormalities and were considered as responders. Logistic regression analysis of baseline characteristics in responders and nonresponders showed that plasma insulin, serum androstenedione, and menstrual history were independent predictors of the treatment's clinical efficacy. In 10 subjects whose menses proved regular after treatment, the great majority of cycles became ovulatory (32 out of 39 assessed, 79%). In conclusion, in women with PCOS metformin treatment reduced hyperinsulinemia and hyperandrogenemia, independently of changes in body weight. In a large number of subjects these changes were associated with striking, sustained improvements in menstrual abnormalities and resumption of ovulation. Higher plasma insulin, lower serum androstenedione, and less severe menstrual abnormalities are baseline predictors of clinical response to metformin.