Abstract:
|
Women who were treated with chest radiotherapy for Hodgkin lymphoma (HL) are at increased risk of developing breast cancer (BC). To provide patients with accurate estimates of their BC risk we developed absolute risk prediction models. Models were estimated by combining hazard ratios (HRs) estimated from a nested case-control study conducted within the Dutch HL survivors cohort and general population incidence rates. We developed one model that used patient-level radiation dose and one with detailed information on dose to specific breast locations. For the latter, we modeled the breast cancer hazard as a sum of location-specific hazards that were the product of a baseline hazard, exponential covariate effects and a linear excess relative risk term HR = 1+bD with location-specific dose D. Due to the limited sample size, we corrected HRs using a Firth type bias correction. Finally, the competing hazard for death was obtained from the cohort using a Cox regression model, with the Breslow estimate for the corresponding baseline hazard. We show model performance based on internal cross-validation and based on independent data from the US Childhood Cancer Survivor Study cohort.
|