Purpose: Glutathione S-transferase theta (GSTTI) and mu (GSTMI) genes are polymorphic, the genes being absent in approximately 15% and 50% of the population, respectively. Because glutathione S-transferases may be involved in the metabolism of chemotherapy drugs, we hypothesized that presence or absence of the genes may influence the outcome of treatment for childhood acute myeloid leukemia (AML). Patients and Methods: We genotyped GSTTI and GSTMI in 306 children with AML receiving chemotherapy on Children's Cancer Group therapeutic studies. Outcomes were compared in those with and without GSTTI and GSTMI genes. Results: Patients with the GSTTI-negative genotype had reduced survival compared with those with at least one GSTTI allele (GSTTI positive) (52% v 40% at 5 years; log-rank P = .05). A multivariate model of survival adjusted for age group, sex, WBC count, chloroma, CNS involvement, and French-American-British group confirmed the increased risk of death in the GSTTI-null cases (relative risk, AQ 1.6; P = .02). The frequency of death in remission was increased in GSTTI-negative cases compared with GSTTI-positive cases (24% v 12%, log-rank P = .05). The frequency of relapse from end of induction was similar in GSTTI-negative and GSTTI-positive cases (38% v 35%, logrank P = .5). Conclusion: Children who lacked GSTTI had greater toxicity and reduced survival after chemotherapy for AML compared with children with at least one GSTTI allele. If confirmed in further studies, GSTTI genotype might be useful in selecting appropriate chemotherapy regimens for children with AML.