After primary surgery, 125 patients with epithelial ovarian cancer (International Federation of Gynaecology and Obstetrics [FIGO] 1c + IIb + IIc = 22 patients, FIGO III = 82 patients, FIGO IV = 21 patients) were randomly allocated to receive PC (cisplatin 50 mg/m2 + cyclophosphamide 600 mg/m2 on day 1 every 28 days) (corrected) or PAC (PC + doxorubicin 45 mg/m2). After six cycles, patients clinically disease-free or with resectable residual disease were submitted to second-look surgery. After restaging, patients in surgical complete response (CR) stopped treatment while those responding partially (PR) received six more courses; patients whose disease progressed were excluded from the study. Among patients with measurable disease, the following clinical response rates were observed: PC = 20% CR, 34.3% PR, 14.3% stable disease, and 31.4% progression; PAC = 40.6% CR, 15.6% PR, 12.5% stable disease, and 31.3% progression. In the 75 patients submitted to second look, the results have been the following: PC = 39.5% CR, 36.8% PR, 7.9% stable disease, and 15.8% progression; PAC = 62.2% CR, 18.9% PR, 10.8% stable disease, and 8.1% progression. The difference in surgical complete response in favor of the PAC regimen is significant (P less than .05). Median survival and progression-free survival were 800 and 400 days, respectively, for PAC arm; median survival and progression-free survival were 680 and 380 days, respectively, for PC. These differences are not significant. Probability of survival was affected by FIGO stage, amount of residual disease, histology, performance status, and response at second look, while no influence was observed according to grade of tumor differentiation and age. Our results demonstrate the usefulness of doxorubicin in terms of surgical CR.