This article aims to test the hypothesis that the risk of epithelial ovarian cancer (EOC) in women with endometriosis might be changed by enrolling different population. A nationwide 14-year historic cohort study using the National Health Insurance Research Database (NHIRD) of Taiwan and the Registry for Catastrophic Illness Patients was conducted. A total of 239,385 women aged between 20 and 51 years, with at least 1 gynecologic visit after 2000, were analyzed. Cases included women with a diagnosed endometriosis, which was established along a spectrum from at least 1 medical record of endometriosis (recalled endometriosis) to tissue-proved ovarian endometriosis (n = X). Controls included women without any diagnosis of endometriosis (n = 239,385 - X). We used Cox regression, and computed hazard ratios (HRs) with 95% confidence intervals (95% CI) to determine the risk of EOC in patients. The EOC incidence rates (IRs, per 10,000 person-years) of women with endometriosis ranged from 1.90 in women with recalled endometriosis to 18.70 in women with tissue-proved ovarian endometrioma, compared with those women without any diagnosis of endometriosis (0.77-0.89), contributing to crude HRs ranging from 2.59 (95% CI, 2.09-3.21; P < 0.001) to 24.04 (95% CI, 17.48-33.05; P < 0.001). After adjustment for pelvic inflammatory disease, infertility, Charlson co-morbidity index, and age, adjusted HRs were ranged from the lowest of 1.90 (95% CI, 1.51-2.37; P < 0.001) in recalled endometriosis to the highest of 18.57 (95% CI, 13.37-25.79; P < 0.001) in tissue-proved ovarian endometrioma, which was inversely related to the prevalence rate of endometriosis (from the highest of 30.80% in recalled endometriosis to the lowest of 1.54% in tissue-proved ovarian endometrioma). The risk of EOC in women with endometriosis varied greatly by different criteria used. Women with endometriosis might have a more apparently higher risk than those reported by systematic review and meta-analysis.