This report describes the long-term follow-up data of three consecutive Dutch Childhood Oncology Group acute myeloid leukemia (AML) protocols. A total of 303 children were diagnosed with AML, of whom 209 were eligible for this report. The first study was the AML-82 protocol. Results were inferior (5-year probability of overall survival (pOS) 31%) to other available regimes. Study AML-87 was based on the BFM-87 protocol, with prophylactic cranial irradiation in high-risk patients only, and without maintenance therapy. This led to a higher cumulative incidence of relapse than that reported by the Berlin-Frankfurt-Münster (BFM), but survival was similar (5-year pOS 47%), suggesting successful retrieval at relapse. The subsequent study AML-92/94 consisted of a modified BFM-93 protocol, that is, without maintenance therapy and prophylactic cranial irradiation. However, all patients were to be transplanted (auto- or allogeneic), although compliance was poor. Antileukemic efficacy was offset by an increase in the cumulative incidence of nonrelapse mortality, especially in remission patients, and survival did not improve (5-year pOS 44%). Our results demonstrate that outcome in childhood AML is still unsatisfactory, and that further intensification of therapy carries the risk of enhanced toxicity. Our patients are currently included in the MRC AML studies, based on the results of their AML 10 trial.