Event-free survival for children with acute lymphoblastic leukemia (ALL) now exceeds 80% in the most effective trials. Failures are due to relapse, toxicity, and second cancers such as therapy-related myeloid leukemia or myelodysplasia (t-ML). Topoisomerase II inhibitors and alkylators can induce t-ML; additional risk factors for t-ML remain poorly defined. The occurrence of t-ML among children who had received granulocyte colony-stimulating factor (G-CSF) following ALL remission induction therapy prompted us to examine this and other putative risk factors for t-ML in 412 children treated on 2 consecutive ALL protocols from 1991 to 1998. All children received etoposide and anthracyclines, 99 of whom received G-CSF; 284 also received cyclophosphamide, 58 of whom also received cranial irradiation. There were 20 children who developed t-ML at a median of 2.3 years (range, 1.0-6.0 years), including 16 cases of acute myeloid leukemia, 3 myelodysplasia, and 1 chronic myeloid leukemia. Stratifying by protocol, the cumulative incidence functions differed (P =.017) according to the use of G-CSF and irradiation: 6-year cumulative incidence (standard error) of t-ML of 12.3% (5.3%) among the 44 children who received irradiation without G-CSF, 11.0% (3.5%) among the 85 children who received G-CSF but no irradiation, 7.1% (7.2%) among the 14 children who received irradiation plus G-CSF, and 2.7% (1.3%) among the 269 children who received neither irradiation nor G-CSF. Even when children receiving irradiation were excluded, the incidence was still higher in those receiving G-CSF (P =.019). In the setting of intensive antileukemic therapy, short-term use of G-CSF may increase the risk of t-ML.