From 1978 until 1988, 116 patients with epithelial ovarian cancer were entered onto one of three consecutive prospective clinical trials involving cisplatin-based combination chemotherapy. They had the following characteristics: (1) stage III or IV disease, (2) grade 2 or 3 tumors, and (3) optimally debulked tumors (residual disease < or = 2 cm). The purpose of the study was to investigate the influence of duration of chemotherapy on survival. The treatment plans were as follows: Trial 1, 12 cycles of cisplatin/melphalan (43 patients); Trial 2, 12 cycles of cisplatin/cyclophosphamide (24 patients); and Trial 3, 6 cycles of cisplatin/cyclophosphamide (49 patients). The total dose of cisplatin was 60 mg/m2 in the first trial and 50 mg/m2 in the second and third trials. Median survival times for the three groups were 58, 29, and 35 months, respectively (NS). Median progression-free survival (PFS) times were 37, 23, and 15 months, respectively (P = 0.0008). Combining patients from the first two trials, the median PFS for patients receiving 12 planned cycles of chemotherapy was 30 months versus 15 months for patients receiving 6 planned cycles (P = 0.0004). Using a forward stepwise Cox proportional hazard model, the use of 12 cycles of therapy and melphalan predicted increased PFS (P = 0.0001 and P = 0.0002, respectively). In view of these results, the lack of published data supporting the superiority of 6 over 12 cycles of chemotherapy, and the rather recent availability of less toxic maintenance therapy (i.e., carboplatin), we believe that a multiinstitutional trial comparing the 6-cycle regimen with more prolonged chemotherapy is justifiable.