Abstract
BACKGROUND: Although female survivors of Hodgkin lymphoma (HL) have an increased risk of breast cancer (BC), no BC risk prediction model is available. We developed such models incorporating mean radiation dose to the breast or breast quadrant-specific radiation doses. METHODS: Relative risks and age-specific incidence for BC and competing events (mortality or other subsequent cancer) were estimated from 1194 Dutch 5-year HL survivors, treated at ages 11-40 during 1965-2000. Predictors were doses to 10 breast segments or mean breast radiation dose, BC family history, year of and age at HL diagnosis, and ages at menopause and first live birth. Models were independently validated using US Childhood Cancer Survivor Study cohort participants. RESULTS: Predicted absolute BC risks 25 years after HL diagnosis ranged from 1.0% for survivors diagnosed at ages 20-24 with less than 10 Gy mean breast radiation dose and who were menopausal 5 years after HL diagnosis, to 22.0% for survivors 25-29 years at diagnosis, with at least 25 Gy mean breast dose and no menopause within 5 years. In external validation, the observed/expected BC case ratio was 1.19 (95% confidence interval 0.97 to 1.47) for the breast segment-specific dose model, and 1.29 (1.05 to 1.60) for the mean breast dose model. The areas under the receiver operating characteristic curve were 0.68 (0.63 to 0.74) and 0.68 (0.62 to 0.73), respectively. CONCLUSION: Breast segment-specific or mean breast radiation dose with personal and clinical characteristics predicted absolute BC risk in HL survivors with moderate discrimination but good calibration, rendering the models useful for clinical decision-making.