The early occurrence of five cases of acute myeloid leukemia (AML) in children treated for primary rhabdomyosarcoma on the Intergroup Rhabdomyosarcoma Study III (IRS III) has prompted this report. These patients received cyclophosphamide and four received etoposide in addition to other agents. There were 1,062 eligible patients entered on IRS III between 1984 and 1991. Following surgery, treatment consisted of multiagent chemotherapy and radiotherapy in select clinical groups. Median follow-up time is 3.7 years (range 0-7.4 years). Incidence densities and odds ratios for AML were calculated for various treatment groups. Five cases of secondary AML have been reported through August 1992. A single case of osteogenic sarcoma was reported in the same period and a patient with myelodysplastic syndrome has occurred since that time. Median time to development of AML was 39 months. Incidence density of AML for patients receiving neither cyclophosphamide nor etoposide was 0, for those receiving cyclophosphamide but no etoposide it was 7.6, and when both agents were given, it was 51.6. The odds ratios of AML for the latter two groups indicated a risk of AML which was seven times higher in the patients who received both agents. A history of breast cancer was present in all five families of patients with AML and several other cancers had occurred in three families. Preliminary analysis suggests a possible causal role for low-dose etoposide in addition to that assumed for cyclophosphamide in the early development of AML among pediatric patients treated for rhabdomyosarcoma.