After Phase I studies of benzisoquinolinedione (amonafide) in solid tumors identified myelosuppression as the dose-limiting toxicity, we conducted a Phase I study in patients with relapsed or refractory acute leukemia to define the optimal dose. Amonafide was given i.v. over 2-4 h daily for 5 days. The starting dose was 600 mg/m2/day with subsequent escalation to 750, 900, 1100, 1400, and 1800 mg/m2/day. Thirty-eight courses were administered to 24 patients, of whom 12 participated in concomitant pharmacological studies. Nausea and vomiting, transient orange discoloration of the skin, and tinnitus occurred at all dose levels. The latter symptom, along with lightheadedness and flushing, was related to infusion duration; this was increased to 4 h with doses greater than or equal to 900 mg/m2. The dose-limiting toxicities were mucositis and painful skin erythema which occurred in all 4 patients treated with 1800 mg/m2. No remissions occurred. Clearing of peripheral blood blasts occurred in 67% of patients treated with 1100 mg/m2 and in all patients treated with greater than or equal to 1100 mg/m2/day. A decrease in marrow leukemic infiltrate (% blasts x % cellularity) to less than 10% occurred in 15 and 50% of patients treated at these levels, respectively. There were 10 deaths (42%), which were unrelated to dosage. The harmonic mean terminal plasma half-life was 4.6 h (range, 2.5-35.5 h). Three patients had long drug half-lives of 9.7, 16.4, and 35.5 h and each had initial bilirubin levels greater than 1.0 mg/dl. The average urinary excretion of amonafide over 5 days was 3.5% of the total dose. This establishes 1100-1400 mg/m2/day for 5 days as the maximally tolerated dose of amonafide for studies in acute leukemia.