Sex steroid hormones have been linked to fractures in older women. To test the hypothesis that hormones measured over the menopausal transition predict fractures. Seven US clinical centers. Two thousand nine hundred sixty women (average age, 46.4 ± 2.7 years) who had at least two repeat hormone measures and prospective information on fractures. Fasting serum was collected annually for hormone assays. Estradiol (E2) was measured with a modified direct immunoassay. FSH and SHBG were measured with two-site chemiluminescence immunoassays. Hormones were lagged (visit year -1) and transformed using log base 2. Incident fractures were ascertained at each annual visit. All medications including hormone therapy were time varying covariates. Discrete survival methods were used. Five hundred eight (17.2%) women experienced an incident fracture over an average follow up of 8.8 ± 4.4 years. Women who experienced an incident fracture were more likely to be white, report high alcohol intake and diabetes, and less likely to report premenopausal status at baseline. A woman whose log E2 was twice that of another had a 10% lower risk of fracture independent of covariates, relative risk (95% CI) = 0.90 (0.82, 0.98). Neither FSH nor SHBG were associated with fractures. Serum E2 levels may help to identify women at higher risk of fractures over the menopausal transition. However, hormone assays must be standardized across laboratories for clinical implementation and further work is needed to define E2 thresholds.