Assessment of robustness against setup uncertainties using probabilistic scenarios in lung cancer: a comparison of proton with photon therapy

We compared the sensitivity of intensity modulated proton therapy (IMPT) and photon volumetric modulated arc therapy (VMAT) plans to setup uncertainties in locally advanced non-small cell lung cancer (NSCLC) using probabilistic scenarios. Minimax robust (MM) and planning target volume (PTV) optimised IMPT and VMAT nominal plans were created with physical dose of 70 Gy in 35 fractions in 10 representative patients. Using population data of setup errors, a fractionated treatment course was simulated, summed (D ) and compared to the nominal plan. Three treatment-course simulations were done for each plan. Target robustness criteria were: dose deviation of ≤5% to clinical target volume (CTV) D and CTV V ≥ 99.9%. Voxelwise simulation repeatability was analysed using Bland-Altman plots. Acceptable limits of agreement were 2% of the prescription dose. All D met target robustness criteria. While fraction VMAT and MM-IMPT doses were excellent, simulated fraction doses in PTV-IMPT were suboptimal. Almost all (>99%) of VMAT and MM-IMPT fraction doses met both target robustness criteria. For PTV-IMPT, only 96.9 and 80.3% of fractions met CTVD and V criteria respectively. Simulation repeatability was excellent (limits of agreement range: 0.41-1.1 Gy) with strong positive correlations. When considering the whole treatment course, setup errors do not influence robustness irrespective of planning techniques used. However, on a fraction level, VMAT and MM-IMPT plans are superior compared to PTV-IMPT plans. Probabilistic analysis provides a fast and practical method for evaluating VMAT and IMPT plan sensitivity against setup uncertainty. VMAT and robust-optimised IMPT plans have comparable sensitivity to setup uncertainties in conventionally fractionated treatment for NSCLC.