P3.17-18 Towards Individualizing Patient Selection in Proton Therapy for Lung Cancer Using the Model-Based Approach

Abstract

Intensity modulated proton therapy (IMPT) has the potential to reduce dose to adjacent normal tissue compared to photon therapy. Although dosimetric advantages can be seen, these results need to translate to clinically meaningful benefit. We aim to identify patients who would benefit from IMPT over photon volumetric modulated arc therapy (VMAT) in locally advanced lung cancer using recently published normal tissue complications probabilities (NTCP) models for heart and lung. VMAT and robust optimised-IMPT plans were assessed to a physical dose of 70Gy in 35 fractions according to RTOG 1308 protocol. Risk estimates of grade 3+ cardiac toxicities and dyspnoea were calculated based on NTCP models which incorporated dose metrics and individual patients’ baseline risk-factors (RF) (cardiac: pre-existing heart disease (HD), lung: patient-reported outcome dyspnoea grade). Risk estimates were calculated and compared based on the pre-treatment patients’ RF and in the hypothetical scenarios where all patients had baseline HD or grade 1 dyspnoea. The difference between the probability estimates of VMAT and IMPT were determined (ΔNTCP). The models were applied in the different scenarios to ten proxy patients. These patients were retrospectively identified to ensure different anatomical locations (lobes and mediastinal/hilar regions) were represented (GTV median: 106cc, range: 15–404cc). Five patients had planning target volume (PTV) overlapping with the heart. There was no significant difference in target coverage between VMAT and IMPT. Protons delivered a lower dose and to heart and lung in all patients (p<0.05, lung: V20, V5, mean lung dose; heart: V30, V5 and mean heart dose). When the risk estimates were calculated according to the patients’ baseline RF, only three had an estimated ΔNTCP≥5%- all based on cardiac toxicities (AR VMAT vs IMPT: 38% vs 10%, 13% vs 10%, 18% vs 9%). Of the three, two patients had pre-existing HD. The relative risk reduction with IMPT varied between none to 0.74. Using ΔNTCP≥5% as a threshold, when all patients were considered to either have baseline HD or grade 1 dyspnoea, the same three patients were selected based on cardiac toxicities and two based on lung toxicity. Predictors of ΔNTCP≥5% for cardiac toxicities were pre-existing cardiac disease, tumour volume extending to the level of T7 vertebrae or below, degree of overlap between PTV and heart. In an unselected patient group, not all patients would benefit from IMPT. Although the validity of photon NTCP models in proton is unclear, appropriate models could be used to aid randomised control trial design.