Research on the effect of body mass index (BMI) on ovarian cancer survival is inconsistent, but previous studies did not consider the possible impact of ascites, bowel obstruction, or cachexia, which commonly occur in late-stage disease. We evaluated the association of BMI, before and around the time of diagnosis, with overall and disease-specific survival in a cohort study of primary invasive epithelial ovarian cancers diagnosed from 2000 to 2013 in Kaiser Permanente Northern California (KPNC) (n=1184). Deaths were identified through December 2014, with a median follow-up of 37 months. Proportional hazards regression was used to estimate overall and ovarian cancer-specific mortality, accounting for prognostic variables including age at diagnosis, race, stage, grade, histology, comorbidities, treatment, post-treatment CA125 levels, ascites, and bowel obstruction. There was no evidence of an association between BMI and overall or ovarian cancer-specific survival. However, we found strong effect modification by stage (Pinteraction<0.01). Compared with normal prediagnosis BMI (18.5-24.9 kg m-2), for women who were obese before diagnosis (BMI⩾35 kg m-2) ovarian cancer-specific survival was lower among those diagnosed at stages I/II (hazard ratio (HR): 3.40; 95% confidence interval (CI): 1.16-9.99), but increased among those diagnosed with stage IV disease (HR: 0.58; 95% CI: 0.35-0.96). Associations were attenuated after excluding those diagnosed with cachexia (n=82) and further adjustment for ascites and bowel obstruction, with no evidence of effect modification by these factors. Associations of obesity with ovarian cancer survival may differ by stage, with decreased survival among those with localised disease and increased survival among those with late-stage disease. Stage-specific effects of obesity on survival suggest a tailored approach to improve prognosis may be appropriate.