Evidence for an association between indicators of adiposity and survival after ovarian cancer has been inconsistent. A prospective cohort study was conducted in China to examine the relationship between ovarian cancer survival and body mass index (BMI). From the 214 patients recruited in 1999 to 2000 with histopathologically confirmed invasive epithelial ovarian cancer, 207 patients or their close relatives (96.7% of cases) were traced and followed to 2003. Deaths were recorded and Cox proportional hazards regression was used to obtain hazard ratios (HR) and 95% confidence intervals (95% CI) from multivariate models. Reduced survival was observed among patients with BMI &GE; 25 kg/m(2) at 5 years before diagnosis (P = 0.001). There were 98 (59.8%) of 164 patients with BMI < 25 kg/m(2) survived to the time of interview compared with only 15 women (34.9%) among the 43 patients whose BMI was &GE; 25 kg/m(2). The HRs significantly increased with higher BMI at 5 years before diagnosis but not at diagnosis nor at age 21 years. The adjusted HR was 2.33 (95% CI, 1.12-4.87) for BMI of &GE; 25 versus < 20 kg/m(2), with a significant dose-response relationship. The HR was 3.31 (95% CI, 1.26-8.73) among patients who had been overweight or obese at age 21 years, but a linear dose-response was not found. We conclude that premorbid BMI may have independent prognostic significance in ovarian cancer.