Fifty-eight acute promyelocytic leukemia (APL) patients (11 newly diagnosed and 47 relapsed) were studied for arsenic trioxide (As2O3) treatment. Clinical complete remission (CR) was obtained in 8 of 11 (72.7%) newly diagnosed cases. However, As2O3 treatment resulted in hepatic toxicity in 7 cases including 2 deaths, in contrast to the mild liver dysfunction in one third of the relapsed patients. Forty of forty-seven (85.1%) relapsed patients achieved CR. Two of three nonresponders showed clonal evolution at relapse, with disappearance of t(15;17) and PML-RARalpha fusion gene in 1 and shift to a dominant AML-1-ETO population in another, suggesting a correlation between PML-RARalpha expression and therapeutic response. In a follow-up of 33 relapsed cases over 7 to 48 months, the estimated disease-free survival (DFS) rates for 1 and 2 years were 63.6% and 41.6%, respectively, and the actual median DFS was 17 months. Patients with white blood cell (WBC) count below 10 x 10(9)/L at relapse had better survival than those with WBC count over 10 x 10(9)/L (P =.038). The duration of As2O3-induced CR was related to postremission therapy, because there was only 2 of 11 relapses in patients treated with As2O3 combined with chemotherapy, compared with 12 of 18 relapses with As2O3 alone (P =.01). Reverse transcription polymerase chain reaction (RT-PCR) analysis in both newly diagnosed and relapsed groups showed long-term use of As2O3 could lead to a molecular remission in some patients. We thus recommend that ATRA be used as first choice for remission induction in newly diagnosed APL cases, whereas As2O3 can be either used as a rescue for relapsed cases or included into multidrug consolidation/maintenance clinical trials.