The aim was to quantify the risk of post-treatment sarcoma in breast cancer patients. All 122,991 women with a breast cancer from 1958 to 1992 in the Swedish Cancer Register were followed up for soft tissue sarcomas and 116 were found, giving a standardised incidence ratio of 1.9 (95% CI 1.5-2.2). The absolute risk was 1.3 per 10(4) person-years. The sarcomas were located in the breast region or on the ipsilateral arm in 63% (67/106). There were 40 angiosarcomas and 76 sarcomas of other types. In a case-control study, angiosarcoma correlated significantly with lymphoedema of the arm, odds ratio (OR) 9.5 (95% CI 3.2-28.0), but no correlation with radiotherapy was observed. For other types of sarcoma there was a correlation with the integral dose. The dose-response relationship indicated that the risk increased linearly with the integral dose to 150-200 J and stabilised at higher energies. The OR was 2.4 (95% CI 1.4-4.2) for an energy of 50 J, approximately corresponding to the radiation of the breast after breast-conserving surgery. Thus, only oedema of the arm correlated with angiosarcoma, but for other types of sarcoma the integral dose of radiotherapy was a predictor of the risk.