The pivotal research project by Darby et al. (NEJM 2013) attempted to quantify the risk of cardiovascular disease after radiation therapy in a large population of breast cancer patients. The Darby group did not have access to detailed cardiac radiation dose-volume data because the majority of their patients were treated in the era before 3D conformal planning. To overcome this issue the Darby group used a common approach in radiation epidemiology research that included using a proxy measure of mean heart dose that relied on the radiation plan of a single "representative" woman with breast cancer. The purpose of this study was to estimate how the Darby group's results would vary had they chosen a different "representative" woman. Materials/Methods: We retrospectively studied 16 women with breast cancer treated with whole breast radiation on the left side between 2013 and 2016. We re-planned each patient by creating 2-field tangential radiation plans using the same field angles for each patient based on classic anatomic approaches using breast borders defined on exam at the time of simulation to define these radiation fields. We normalized the dose to 50 Gy in 2 Gy fractions, and for each patient we contoured the heart using the Michigan heart atlas by Feng et al. (IJROBP 2011) and determined the mean heart dose. The Darby paper found a linear relationship between mean heart dose and risk of toxicity e with a 7.4% increased risk for every additional 1 Gy mean heart dose. We tested how choosing different "representative" women would influence this risk-to-dose relationship. Results: The mean heart dose for patients in our cohort ranged from 0.5-9.3 Gy. The average heart dose was 1.5 Gy with a standard deviation of 2.2 Gy. Therefore, had the Darby analysis chosen different "representative" women, the risk-to-dose relationship could have ranged from 1.2%-22.4% increased risk for every additional 1 Gy mean heart dose. Conclusion: The use of a "representative" breast cancer patient to estimate heart dose in larger radiation epidemiology research does not take individual anatomy into account and can lead to considerable error when estimating the relationship between heart dose and cardiovascular toxicity. Depending on the "representative" woman chosen the radiation-toxicity relationship could vary by more than an order of magnitude. For accurate estimation of radiation-induced cardiac toxicity detailed individual patient dosimetric data must be used whenever possible.

No abstract