Between March 1985 and January 1987, 103 women with histologically proven stage III-IV ovarian carcinoma were randomly allocated to groups receiving monthly intravenous regimens of 1 g of cyclophosphamide/m2 plus either 60 mg of cisplatin (CDDP)/m2 or 150 mg of carboplatin (CBDCA)/m2 for 1 year unless disease progressed earlier. The groups were well balanced according to the stratification factors (age, histologic differentiation, extent of residual disease, and performance score), and both treatments were well tolerated and produced similar median first-course leukopenia (2,200 and 2,000 cells/microL) and thrombocytopenia (220,000 and 202,500 cells/microL). The CBDCA regimen was less emetogenic. After an interim analysis in January 1987 revealed superior progression-free survival for the group of 53 patients receiving CDDP (P = .005), the study was closed to further accrual. Those 24 patients still receiving CBDCA were encouraged to cross over to the CDDP-based regimen and 21 of them did. Following treatment crossover, the relative risk of death associated with original allocation to CBDCA receded from 1.79 to 0.97, indicating success of the salvage treatment using the CDDP-based regimen. This aborted study demonstrated the superiority of CDDP over CBDCA when the two platinum compounds were compared at equally myelosuppressive low doses in combination with 1 g of cyclophosphamide/m2. If CDDP is to be supplanted by CBDCA, larger, more myelosuppressive doses of CBDCA will be required. The platinum drug antitumor effect is a critically important therapeutic feature of this combination.