Four hundred eleven women with metastatic breast cancer were randomly assigned to receive either 60 mg/m2 doxorubicin (130 patients), 320 mg/m2 bisantrene (146 patients), or 14 mg/m2 mitoxantrone (135 patients). The doses were given intravenously every 3 weeks with a cross-over design to determine their relative efficacy and toxicity. To be eligible, patients must have had one previous chemotherapy regimen, and patients who were estrogen receptor positive must have failed endocrine therapy. There were 365 patients assessable for response and 399 assessable for toxic effects. The median age was 57 years; 18% were premenopausal or perimenopausal. Visceral dominant disease was present in 66% of the patients. Ninety-seven percent of the patients had a disease-free interval from diagnosis to first recurrence of less than 1 year. The response rate was 28% with doxorubicin, 13% with bisantrene, and 14% with mitoxantrone (P = .004). Median time to treatment failure was 133 days with doxorubicin, 66 days with bisantrene, and 68 days with mitoxantrone (logrank P = .06). The median survival was 315 days for doxorubicin, 290 days for bisantrene, and 177 days for mitoxantrone (logrank P = .04), although survival at 2 years was similar for all three agents. There were five responses in the 66 patients crossed over to doxorubicin and one response each for patients crossed over to bisantrene (39 patients) or mitoxantrone (63 patients). Toxicity leading to discontinuance of therapy was more common with doxorubicin, and discontinuance of therapy was due primarily to patient's request or cardiotoxicity. The major dose-limiting toxic effect for all three agents was leukopenia. Nausea and vomiting, mucositis, and alopecia were more severe with doxorubicin. Congestive heart failure developed in nine patients treated with doxorubicin, zero patients treated with bisantrene, and two patients treated with mitoxantrone. A decrease in the left ventricular ejection fraction, as defined by moderate to severe Alexander grade changes, was more common in patients treated with doxorubicin (doxorubicin-treated patients = 20%, bisantrene-treated patients = 5%, and mitoxantrone-treated patients = 10%). This study demonstrates that bisantrene and mitoxantrone have only modest activity in metastatic breast carcinoma. The activity of doxorubicin is greater than that of the other two agents, but at a cost of increased toxicity.