Eighty eight patients with FIGO stage IIb/c (postoperative residual) or III/IV epithelial ovarian cancer were randomised to receive cycles of cyclophosphamide (600 mg/m2) with either iproplatin (240 mg/m2), cisplatin (100 mg/m2) or carboplatin (300 mg/m2). A total of six cycles were given at monthly intervals. Patients were well-balanced for major prognostic factors. There was no significant difference in overall response rate (iproplatin arm, 64.3%, cisplatin arm 72.7% and carboplatin arm 66.7%). There were more complete remissions on the carboplatin arm, 45.8% compared with 21.4% on iproplatin and 22.7% on cisplatin but the difference was not statistically significant (p = 0.11). With a median follow up of 50 months the median survival for the iproplatin arm is 18 months, for the cisplatin arm 19 months and for the carboplatin arm 24 months (p = 0.15). Toxicity was greatest with cisplatin and least with carboplatin. Myelotoxicity limited the dose delivery of iproplatin as measured by total dose, dose intensity and dose intensity product. Carboplatin is at least as effective and less toxic than cisplatin when used in conjunction with cyclophosphamide for the treatment of ovarian carcinoma, and this analogue has been selected for dose intensification studies in this tumour at the Christie Hospital.