Sixty-three evaluable patients with advanced breast cancer were randomly allocated to receive three-week intravenous courses of carminomycin (18 mg/m2) or 4'-epidoxorubicin (90 mg/m2). The former yielded one (3%) partial response for nine weeks among 29 patients whereas, in the other arm, nine (27%) of 34 patients achieved partial response for a median of 28 weeks (range, nine to 36 weeks; p less than 0.02). The major toxic effect of these anthracyclines was leukopenia with median white blood cell nadirs of 1,600/microL (range, 300-4,000/microL) versus 1,800/microL (range, 500-4,300/microL), respectively. Acute nonhematologic toxic effects were qualitatively similar but carminomycin produced significantly less gastrointestinal intolerance and alopecia. Patients whose disease failed to respond to first-line anthracycline received doxorubicin (60 mg/m2) every three weeks. Four partial responses were obtained among 19 patients previously treated with carminomycin. Following 4'-epidoxorubicin therapy, one of 12 evaluable patients also attained partial response. Survival curves were not affected by the initial treatment option. Carminomycin has marginal activity against breast cancer whereas 4'-epidoxorubicin deserves further evaluation of its therapeutic index relative to doxorubicin. The design used in this trial appears attractive for prompt phase II evaluation of anthracycline analogs.