BACKGROUND An excess risk of second malignancies has been reported in survivors of Ewing's sarcoma. We examined a multiinstitutional data base to reevaluate the risk among survivors of Ewing's sarcoma and to identify possible causal factors. METHODS Information was derived from a data base that included 266 survivors of Ewing's sarcoma. Cumulative incidence rates of second malignancies were calculated. Contributions of clinical features, type and dose of chemotherapy, and cumulative radiation dose to the risk of second malignancies were evaluated. RESULTS After a median follow-up duration of 9.5 years (range, 3.0 to 30), 16 patients have developed second malignancies, which included 10 sarcomas (five osteosarcomas, three fibrosarcomas, and two malignant fibrous histiocytomas) and six other malignancies (acute myeloblastic leukemia, acute lymphoblastic leukemia, meningioma, bronchioalveolar carcinoma, basal cell carcinoma, and carcinoma-in-situ of the cervix). The median latency to the diagnosis of the second malignancy was 7.6 years (range, 3.5 to 25.7). The estimated cumulative incidence rates at 20 years for any second malignancy and for secondary sarcoma were 9.2% (SD = 2.7%) and 6.5% (SD = 2.4%), respectively. The cumulative incidence rate of secondary sarcoma was radiation dose-dependent (P = .002). No secondary sarcomas developed among patients who had received less than 48 Gy, while the absolute risk of secondary sarcoma was 130 cases per 10,000 person-years of observation among patients who had received > or = 60 Gy. CONCLUSION The overall risk of second malignancies after Ewing's sarcomas is similar to that associated with treatment for other childhood cancers. The radiation dose-dependency of secondary sarcomas justifies modification in therapy to reduce radiation doses.