Abstract
BACKGROUND AND PURPOSE: The purpose of this study was to investigate whether treatment information from medical records can be used to estimate radiation doses to heart and lungs retrospectively in pediatric patients receiving spinal irradiation with conventional posterior fields. MATERIAL AND METHODS: An algorithm for retrospective dosimetry in children treated with spinal irradiation was developed in a cohort of 21 pediatric patients with available CT-scans and treatment plans. We developed a multivariable linear regression model with explanatory variables identifiable in case note review for retrospective estimation of minimum, maximum, mean and V(10%)-V(80%) doses to the heart and lungs. Doses were estimated for both linear accelerator (Linac) and (60)Co radiation therapy modalities. RESULTS: Age and spinal field width were identified as statistically significant predictors of heart and lung doses in multivariable analyses (p < 0.01 in all models). Models showed excellent predictive performance with R(2) = 0.70 for mean heart dose and 0.79 for mean lung dose, for Linac plans. In leave-one-out cross-validation analysis the average difference between predicted and actual mean heart dose was 6.7% and 7.6% of the prescription dose for Linac and (60)Co plans, respectively, and 5.2% and 4.9% for mean lung dose. Due to the small sample size and large inter-patient variation in heart and lung dose, prospective studies validating these findings are highly warranted. CONCLUSIONS: The models presented here provide retrospective estimates of heart and lung doses for historical cohorts of pediatric patients, thus facilitating studies of long-term adverse effects of radiation.