We assessed whether hypofibrinolytic plasminogen activator inhibitor 1 (PAI-1 activity) showed an independent association with first-trimester miscarriage in the 430 women with polycystic ovary syndrome (PCOS) who had previous pregnancies (from a cohort of 967 women with PCOS). Prospectively, we hypothesized that Glucophage (Bristol-Myers Squibb, Princeton, NJ) promotes successful live births in women with PCOS by lowering PAI-1 activity before conception and maintaining further reductions of PAI-1 activity during the first trimester of pregnancy. We also assessed whether PAI-1 activity levels were independently related to PAI-1 genotype and to modifiable risk factors body mass index (BMI), insulin, and triglyceride. By stepwise logistic regression, with the dependent variable being previous pregnancy outcomes at 3 levels (live birth pregnancies only [n = 208]; both > or =1 live birth and > or =1 first-trimester miscarriage [n = 111]; or first-trimester miscarriages only [n = 71]) and explanatory variables PAI-1 genotype, PAI-1 activity, insulin, homeostasis model assessment of insulin resistance, BMI, and triglyceride, PAI-1 activity was positively associated with first-trimester miscarriage (P = .004). For each 5 IU/mL increment in PAI-1 activity, the risk being in an adverse first-trimester miscarriage category increased (odds ratio, 1.12; 95% confidence interval, 1.04-1.20). Prospectively, from pretreatment to the last preconception visit on Glucophage, in 30 women who subsequently had live births, PAI-1 activity fell 44%, but rose 19% in 23 women with first-trimester miscarriage (P = .03). In the 30 women with live birth pregnancies, median PAI-1 activity fell continuously from pretreatment through the first trimester (from 16.8 to 6.7 IU/mL), whereas PAI-1 activity was either unchanged or rose in women with first-trimester miscarriage. Of the 921 women with PCOS who had 4G5G data, 718 (78%) had 4G4G-4G5G genotypes vs 87 (69%) of 126 normal female controls (chi(2) = 4.95, P = .026). The 4G allele frequency was 53% in women with PCOS vs 46% in controls (chi(2) = 4.3, P = .04). Of the 866 women with PCOS who had PAI-1 activity data, by stepwise regression, positive independent determinants of PAI-1 activity included BMI (partial R(2) = 10.6%, P < .0001), insulin (partial R(2) = 2.8%, P < .0001), triglyceride (partial R(2) = 1.1%, P = .0009), and the 4G4G-4G5G genotype (partial R(2) = 1%, P = .0011). The PAI-1 gene 4G polymorphism is more common in women with PCOS than in normal women and, in concert with obesity, hyperinsulinemia, and hypertriglyceridemia, contributes to treatable, hypofibrinolytic, miscarriage-promoting, high PAI-1 activity. Preconception and first-trimester decrements in PAI-1 activity on Glucophage are associated with live births, whereas increments or no change in PAI-1 activity despite Glucophage appears to be associated with first-trimester miscarriage.