Study AML-BFM-87 compared prospectively if cranial irradiation could be abandoned by adding two blocks of intensification with high dose ARA-C and VP-16 after consolidation and furthermore, improve prognosis compared to study -83. 210 children were enrolled in study AML-BFM-87 until March 31, 1991. 164 (78%) achieved complete remission. Probabilities for event-free survival (EFS) and event-free interval (EFI) of 5 years were: .45 (SD .04) and .57 (SD .05). In the first 2.5 years of the study irradiation was randomized (n = 31), selected or refused (n = 24). However, during this period irradiation was mandatory in patients with leukocyte count greater than 70,000/mm3, and also in children with initial CNS involvement. Since July 1989 prophylactic cranial irradiation was abandoned. Patients of the group with mandatory irradiation (n = 39) presented with more unfavourable risk parameters than the group of non-irradiated children, who were enrolled in the study after randomisation had been stopped. Nevertheless, results showed in randomized and selected patient groups as well as in the total cohort a longer relapse-free interval (RFI) in irradiated (n = 66) compared to non-irradiated (n = 94) patients (RFI of 5 years: .70, SD .04 vs. .51, SD .07, p less than .05). Relapses in non-irradiated children occurred mainly in the bone marrow and less often in the CNS. The increase in relapse rate was seen especially in non-irradiated patients of the low risk group as defined in study AML-BFM-83 (RFI: .40, SD .14 vs. .79, SD .09 with irradiation, p less than .01). In the high risk group, however, the differences were not significant. Our results suggest that cranial irradiation is an important part of therapy in childhood AML, and that the good prognosis of the low risk group in study AML-BFM-83 was probably based on the combination of intensive chemotherapy and cranial irradiation.