Commercial Analytic Dose Calculation Algorithms Dramatically Overpredict Proton Therapy Dose to Thoracic Tumors in Lung Credentialing Phantom

Abstract

To better understand the poor agreement between calculated and measured doses in the IROC Houston lung phantom. IROC’s anthropomorphic lung phantom is used for proton approval and credentialing for NCI-funded trials. The phantom contains TLD and film to measure the delivered dose. A motor moves the phantom target insert within the thorax, mimicking a human breathing cycle. This phantom provides a consistent QA tool that assesses comparability in dose delivery between participating proton centers. The overall pass rate for this lung phantom is only 63%, and those irradiations that do pass, do so marginally, despite the generous 7%/4mm acceptance criteria. To understand the role of the dose calculation algorithm in this disagreement, five institutions that previously irradiated the phantom with their clinical analytic dose calculation algorithms were asked to recalculate dose to the phantom with a Monte Carlo algorithm. These recalculations were collected and compared to measurements. Monte Carlo recalculations showed a striking improvement in agreement with the physical measurements, compared to the clinical analytic algorithm calculations. The analytic algorithms calculate, on average, the dose to the center of the target to be 7% higher than TLD measurements, whereas Monte Carlo algorithms are, on average, only 1.5% higher. The difference is even greater at the edge of the GTV, where analytic algorithms overpredict dose by up to 27% as compared to physical measurements. Monte Carlo recalculations agreed within 5% of the measurements. Within the PTV, one case showed a 41% overpredition of dose by the analytic algorithm. All comparisons showed a region of at least 10% dose discrepancy between the analytic and Monte Carlo algorithms inside the target volume. Previous studies comparing in-house Monte Carlo to a commercial analytic algorithm also showed a systematic overestimation of dose by the analytic algorithm, both in lung and in other disease sites. Analytic algorithms are doing a poor job predicting proton dose in lung tumors. There are many uncertainties in proton dose delivery, but eliminating this one uncertainty will reduce the overall uncertainty. For the safety of patients and the efficacy of treatment, proton therapy centers should switch to Monte Carlo or other advanced dose calculation algorithms for treatment planning for thoracic tumors.