In total, 281 of the 7711 women who were initially treated for breast cancer between 1954 and 1983 at the Gustave Roussy Institute developed a second malignant neoplasm (SMN) other than second primary breast cancer and nonmelanoma skin cancer at least 1 year after breast cancer treatment. We carried out a nested case-control study to determine the overall relationship between the dose of radiotherapy received at a given anatomical site and the risk of SMN at the same site. In total, 75% of the cases of SMN were previously treated by radiotherapy, as compared to 73% of the controls. In the irradiated patients, the median local dose was higher among cases (3.1 Gy) than among controls (1.3 Gy). More than 40% of the irradiated patients received a local dose of less than 1 Gy. A purely quadratic relationship was observed between the dose of radiation received at an anatomical site and the risk of SMN at this site. According to the quadratic model, the excess risk of SMN was 0.2% (95% CI 0.05-0.5%) when the target organ received 1 Gy. This risk did not differ significantly according to age at the time of radiotherapy (<40 vs >or=40 years). The risk of SMN was 6.7-fold higher for doses of 25 Gy or more than in the absence of radiotherapy. No carcinogenic effect of chemotherapy was observed and a dose-effect relationship between the length of tamoxifen treatment and SMN occurrence was found. This relationship was limited to endometrial cancers and did not modify the relationship with radiation dose. Our results suggest that high radiation doses slightly increase the risk of second malignancies after breast cancer.