Author: Amy Berrington de Gonzalez, David Borrego, Rochelle Curtis, Gretchen L Gierach, Jae Won Jung, Choonik Lee, Choonsik Lee, Matthew Mille, Lene Veiga, Jacqueline Vo 👨🔬
Affiliation: University of Michigan, East Carolina University, Radiation Protection Division, Office of Radiation and Indoor Air, U.S. Environmental Protection Agency, National Cancer Institute, Institute of Cancer Research 🌍
Purpose: Radiotherapy is vital for treating breast cancer, but its long-term health effects remain a concern. Accurate dose reconstruction is crucial for evaluating risks in retrospective epidemiological studies, yet contralateral breast (CB) dose estimates often carry high uncertainty, especially when detailed radiotherapy records are unavailable. This study assessed dosimetric uncertainties in CB doses reconstructed for a subset of breast cancer radiotherapy patients in the National Cancer Institute-Kaiser Permanente (KP) breast cancer survivor cohort treated between 1990 and 2012.
Methods: The analysis included 2,442 breast cancer radiotherapy patients from three KP health systems. Building on previous CB dose reconstructions, the study created multivariate linear regression models and probability mass functions to model key radiotherapy parameters. A Monte Carlo uncertainty propagation method was used to incorporate uncertainty into CB doses, generating 500 plausible dose realizations across five breast sub-regions for each patient. The coefficient of variation (COV) was calculated for each patient from the multiple dose realizations, serving as a measure of dosimetry quality.
Results: The majority of patients had COVs below 0.3. Specifically, 31% (766 patients) had COVs under 0.1, 17% (396–404 patients) had COVs between 0.1 and 0.2, and 28–30% (672–724 patients) had COVs between 0.2 and 0.3. Fewer patients were in the 0.3–0.4 COV range for the upper and lower inner quadrants (26–27 patients) compared to other sub-regions (55–73 patients).
Conclusion: This study quantified dosimetric uncertainties caused by missing treatment parameters and developed dosimetry quality scores for individual CB doses. These scores can support sensitivity analyses in case-control studies of second cancer risk. The uncertainty-integrated dose calculation method developed in this study will be extended to the full NCI-KP cohort, enabling risk evaluation for other adverse health outcomes.