A randomized clinical trial was conducted in women with stage III ovarian carcinoma (less than or equal to 1 cm residual lesions), using cyclophosphamide plus cisplatin (CP) with or without doxorubicin. There were 349 evaluable patients, of whom 176 received CP while 173 patients received CP plus doxorubicin (CAP). Hematologic toxicity was almost identical. There was no significant difference in progression-free interval (PFI) (median, 22.7 months and 24.6 months), frequency of negative second-look laparotomy (30.2% and 32.8%), or survival (median, 31.2 months and 38.9 months) between CP and CAP, respectively. Thus, doxorubicin in the dose schedule employed does not improve combination chemotherapy of optimal stage III ovarian carcinoma. Several other findings, independent of treatment arm, were of interest. There was a significant difference in PFI and survival by residual disease category (yes v no) and by grade of differentiation (1 v 2 + 3). In multivariate analysis, age, residual disease at entry, cell type (clear cell carcinoma), and time from surgery to initiation of chemotherapy were significant predictors of survival. There was no difference in outcome comparing those who refused second-look with those who had a second-look.