In a study designed to compare response rates of patients with stage III epithelial ovarian carcinoma to ifosfamide and carboplatin, 152 patients were randomised to receive either sequential therapy with three cycles of ifosfamide followed by three cycles of carboplatin, or to six cycles of single agent carboplatin. Ifosfamide was given every 3 weeks in a dose of 5 gm m-2 as a 24 h infusion with mesna, 1 gm m-2 by i.v. bolus prior to ifosfamide, 3 gm m-2 with ifosfamide, and 1 gm m-2 as an 8 h infusion after ifosfamide. Carboplatin was given in a dose of 400 mg m-2 by short i.v. infusion every 4 weeks. Sixty-eight evaluable patients were randomised to sequential ifosfamide/carboplatin, and 67 to single agent carboplatin. Median follow-up is 36 months (range 5.5-82.3). After three cycles of treatment two patients in the ifosfamide/carboplatin arm achieved complete remission (CR), and 12 partial remission (PR) for an overall response rate of 29%, whereas in the carboplatin arm ten patients achieved CR, and 23 PR, for an overall response rate of 63% (P = 0.0008). Seven of 15 patients with progressive disease, and nine of 20 patients with stable disease at the initial response evaluation, following three cycles of ifosfamide, subsequently responded to carboplatin therapy so that the final response rate to the complete regimen was 65% for the ifosfamide/carboplatin arm, compared to 71% for the carboplatin arm (NS). For the ifosfamide/carboplatin arm, median recurrence free survival and overall survival were 14.1 months and 18.7 months. Corresponding figures for the carboplatin arm were 14.5 months and 21.5 months (NS). Both treatments were generally well tolerated. However 47% of patients in the ifosfamide/carboplatin arm developed alopecia sufficient to require a wig, compared to only 2% in the carboplatin arm. Ifosfamide is clearly less effective, and more toxic than carboplatin. Ifosfamide failures can however be effectively salvaged by subsequent carboplatin treatment. Ifosfamide cannot be recommended for single agent therapy in ovarian carcinoma, however the combination of carboplatin plus ifosfamide might be a suitable treatment to be tested in a future randomised study against carboplatin alone.