Thirty-nine patients with untreated acute promyelocytic leukemia (APL) were randomly allocated to receive rubidazone (zorubicin) 200 mg/m2/d, days 1 to 4 plus cytarabine (Ara C) 200 mg/m2/d, days 1 to 7 (arm A, 21 patients), or amsacrine (Amsa) 150 mg/m2/d, days 1 to 4 plus Ara C 200 mg/m2/d, days 1 to 7 (arm B, 18 patients). Prophylaxis of disseminated intravascular coagulation was made by platelet transfusions and heparin. In case of leukemic resistance, patients received a second course with 2 days of rubidazone (arm A) or Amsa (arm B) and 3 days of Ara C. Patients who achieved complete remission (CR) received three consolidation courses with the two drugs used for induction and maintenance therapy for 3 years. Two patients in arm A and one in arm B were allografted in first CR. Initial characteristics were similar in both arms. In arm A, 18 patients (86%) reached CR, two had hypoplastic death, and one had leukemic resistance after two courses. In arm B, 12 patients (66%) achieved CR, two had early death (CNS bleeding, one case; ventricular fibrillation, one case), and four had resistant leukemia after two courses. The difference in CR rate between the two arms was not significant. In arm A, disease-free survival (DFS) showed a plateau at 54.3% after 34 months (95% confidence interval [CI], 32.1% to 74.9%), with eight CRs longer than 34 months. In arm B, DFS was significantly shorter (P less than .03), showing a plateau at 16.7% after 38 months (95% confidence interval, 4.7% to 44.6%), and only two prolonged CRs were seen. The difference in DFS remained significant after censoring allografted patients and patients who died in CR (one in arm A, two in arm B). Our results suggest that Amsa-Ara C combinations may be inferior to anthracycline-Ara C combinations in the treatment of APL, because they seem to provide shorter DFS and, possibly, a higher incidence of initial leukemic resistance. However, studies with larger numbers of patients are required.