To investigate the roles of radiation therapy and chemotherapy in the occurrence of subsequent leukemia after childhood cancer. We analyzed data from a case-control study with 35 cases and 140 controls. The active bone marrow (ABM) was segmented into 19 compartments, and the radiation dose was estimated in each. The chemotherapy drug doses were also estimated to enable adjustments. Models capable of accounting for radiation dose heterogeneity were implemented for analysis. Univariate analysis showed a significant trend in the increase of secondary leukemia risk with radiation dose, after accounting for dose heterogeneity (P=.046). This trend became nonsignificant after adjustment for doses of epipodophyllotoxins, alkylating agents, and platinum compounds and the first cancer on multivariate analysis (P=.388). The role of the radiation dose appeared to be dwarfed, mostly by the alkylating agents (odds ratio 6.9, 95% confidence interval 1.9-25.0). Among the patients who have received >16 Gy to the ABM, the radiogenic risk of secondary leukemia was about 4 times greater in the subgroup with no alkylating agents than in the subgroup receiving ≥10 g/m(2). Notwithstanding the limitations resulting from the size of our study population and the quite systematic co-treatment with chemotherapy, the use of detailed information on the radiation dose distribution to ABM enabled consideration of the role of radiation therapy in secondary leukemia induction after childhood cancer.