Cisplatin is generally accepted to be the most active cytotoxic agent for the treatment of ovarian cancer but the optimum dose remains unclear. We have performed a randomised trial to assess the importance of cisplatin dose in the treatment of advanced epithelial ovarian cancer. Patients were randomly assigned treatment with 50 mg/m2 (low dose) or 100 mg/m2 (high dose) cisplatin plus 750 mg/m2 cyclophosphamide, for a maximum of six cycles with intervals of 3 weeks. We planned to recruit 300 patients, but an interim analysis on the first 165 indicated a highly significant survival difference (p = 0.0008). Recruitment was therefore stopped and the trial patients were followed-up for 12 months longer. The relative progression rate (high-dose/low-dose) after 12 months' extra follow-up was 0.55 (95% confidence interval 0.37-0.81, p = 0.003) and the relative death rate 0.53 (0.34-0.81, p = 0.003). Overall median survival was 69 weeks in the low-dose group and 114 weeks in the high-dose group. Residual disease extent before chemotherapy had an important influence--patients with lesions of less than 2 cm did best; if given high-dose cisplatin their median survival was 3 years. 56 low-dose and 45 high-dose patients completed six cycles of chemotherapy; 15 and 9 patients, respectively, were withdrawn early because of progressive disease and treatment was stopped in 6 and 25, respectively, because of unacceptable side-effects or patient refusal. Toxic effects were significantly greater in the high-dose group, especially those on the nervous system and ears, alopecia, vomiting, and anaemia. Although the higher dose of cisplatin clearly leads to better results in terms of survival, its overall clinical benefit in the management of ovarian cancer will depend on further improvements in measures to alleviate toxic effects.