Amonafide is a new imide derivative of naphthalic acid. The drug had demonstrated significant activity in preclinical studies and some activity in Phase I trials. The drug is extensively metabolized and detected in plasma and urine. Its toxicity has previously been correlated to the formation of an active metabolite, N-acetyl-amonafide. Amonafide was chosen for inclusion in the Cancer and Leukemia Group B (CALGB) master metastatic breast cancer protocol. CALGB 8642 randomizes previously untreated metastatic breast cancer patients either to one of several Phase II agents given for up to four cycles and then followed by standard cyclophosphamide-doxorubicin-5-fluorouracil, or to immediate treatment with standard cyclophosphamide-doxorubicin-5-fluorouracil. The end point of CALGB 8642 is to assess the difference in survival, toxicity, and overall response when limited exposure to Phase II agents precedes standard chemotherapy. This report deals only with amonafide as a Phase II agent. Comparisons with the cyclophosphamide-doxorubicin-5-fluorouracil arm will not be addressed. Patients had to have histologically documented measurable breast cancer and a performance status of 0-1. Patients could not have had prior chemotherapy for metastatic disease. Prior adjuvant chemotherapy was permitted. Patients could not have visceral crisis. Amonafide was given at 300 mg/m2/day i.v. for 5 days, and repeated at 21-day intervals for a maximum of four cycles. Escalation and reduction in dose was mandated dependent on hematotoxicity or lack thereof. Toxicity was primarily hematological and bimodal: 32% had grade 3 or 4 leukopenia and 24% had grade 3 or 4 thrombocytopenia; 22% had no leukopenia and 44% had no thrombocytopenia. The response rate was 18%, including one complete response. When response was analyzed by hematological toxicity, there was a 35.7% response if patients had leukopenia grade 3/4 (versus 8.3%, P = 0.08). There was a 50% response if patients had thrombocytopenia grade 3/4 (versus 7.1%, P = <0.01). We conclude that amonafide is somewhat active in previously untreated breast cancer patients. There may be a steep dose-response curve, based on the significant correlation between myelosuppression and response. Rates of responses in patients adequately dosed (i.e., with significant hematotoxicity) with amonafide ranged from 35 to 50%. Further studies will incorporate individualized dosing based on pretreatment acetylator phenotyping.