We analyzed New South Wales Central Cancer Registry data for 1980-2003, to determine time trends in case fatality from ovarian cancer, after adjusting for stage, histological, and sociodemographic factors, and to consider service-delivery and research implications. After adjusting for covariates, the relative risk (95% confidence limit) of ovarian-cancer death reduced to 0.51 (0.46, 0.57) for 1999-2003 compared with 1980-1983. Relative risks were higher for adenocarcinomas and other specified and unspecified cancers than serous carcinomas, but lower for endometrioid carcinomas, sex cord-stromal and germ cell tumors. The probability of diagnosis with localized as opposed to more advanced disease was lower in older patients, the lowest socioeconomic stratum, women born in non-English-speaking countries, and more recent diagnostic periods. Approximately 61% of ovarian cancers had distant metastases at diagnosis in 1999-2003. Poorer survivals apply to older patients. Research directed at finding an effective screening test for epithelial ovarian cancer remains a priority.