Despite substantial progress in the management of childhood acute myeloid leukemia (AML), only about 50% of patients are cured by intensive chemotherapy. The long-term results of clinical trials may reveal principles that can guide the development of future therapy. From 1980 to 2000, 251 patients <15 years of age with newly diagnosed AML were enrolled on one of the five consecutive St Jude AML studies. The median age of the 128 boys and 123 girls was 6.2 years; 193 were white, 45 black, and 13 of other racial groups. With the exception of one protocol (AML-83), outcomes improved in general over the two decades. The estimated 5-year event-free survival (+/-s.e.) was 30.8+/-5.6% for AML-80; 11.1+/-4.3% for AML-83; 35.9+/-7.4% for AML-87; 43.5+/-6.2% for AML-91; and 45.0+/-11.1% for AML-97. Resistant or relapsed AML caused the great majority of treatment failures. Increasing the intensity of chemotherapy (AML-87) did not improve outcome, partially because of toxicity, nor did prolonging postremission therapy by adding sequential myeloablative (AML-80) or nonmyeloablative (AML-83) chemotherapy cycles. We conclude that subtype-specific therapies are needed to replace the 'one size fits all' strategy of the past two decades.