To detect relapse acute myeloid leukemia (AML) treatment protocols have called for bone marrow exams every 2 to 4 months in remission. To investigate the effect of replacing this policy with one calling for marrows only when blood count is abnormal (platelets < 100,000, neutrophils < 1,000, circulating blasts > 0%, unrelated to prior chemotherapy), we reviewed the records of all 444 patients with AML whose disease recurred ( > or = 5% marrow blasts unrelated to prior chemotherapy) for the first time between 1980 and 1995. The 375 patients with adequate follow-up were classified as (1) simultaneous-- blood count abnormal when relapse noted without a normal marrow intervening between first abnormal count and relapse, 289 patients (77% of 375), (2) marrow first--blood count normal when relapse noted, 60 patients (16%), or (3) blood first--a normal marrow intervened between first abnormal blood count and relapse, 26 patients (7%). Interval between marrow exams and blood counts did not differ in the three groups (a 25-patient sample of the 289 patient simultaneous group was analyzed as representative of this group) with marrows done at a median of once monthly and blood counts at a median of once weekly from complete remission (CR) date to relapse date. The three groups also had similar first CR duration, and pretreatment cytogenetics. CR rates following salvage chemotherapy were 32% to 33% in the simultaneous and marrow first groups and 17% in the blood first group. We conclude that routine marrow exams for morphology are not needed in the great majority of AML patients in first CR.