ESTRO 2020 Abstract Book

S126 ESTRO 2020

The model is based on modern normal tissue complication probability (NTCP) models (e.g. for xerostomia [2]). This daily slot allocation strategy is compared (in terms of average NTCP values) to a threshold-based PT patient selection in which patients are selected for whole PT treatment if a slot is available at start of fractionated treatment and their ∆NTCP exceeds a threshold (5%, 10%, 15%). To simulate many patients, the doses in relevant OARs (e.g. contralateral parotid gland) are sampled from a 2D gaussian distribution (Figure 1) derived from the OAR doses of 45 HNC patients for which IMRT and IMPT plans were previously created [1] and rescaled to a standard of care prescription (1.8 Gy to PTV, 2.3 Gy to GTV).

[1] Jakobi A. et al., IJROBP, 92.5 (2015): 1165-1174 [2] Houweling A.C. et al. IJROBP, 76.4 (2010): 1259-1265 PH-0243 Joint optimization of combined photon- carbon ion treatments for infiltrative tumors A.B.A. Bennan 1 , M. Bangert 1 , J. Unkelbach 2 1 German Cancer Research Center DKFZ, Medical Physics in Radiation Oncology, Heidelberg, Germany ; 2 University Hospital Zurich, radiation oncology, Zurich, Switzerland Purpose or Objective Carbon ions show higher RBE and reduced fractionation effect near the Bragg peak, which has advantages for the treatment of radioresistant gross tumor volumes (GTV) However, carbon ions face limitations for treating infiltrative disease when normal tissues inside the clinical target volume (CTV) must be protected through fractionation. These characteristics gave rise to combined photon-carbon ion treatments in which carbon ions give a boost to the GTV while photons irradiate the CTV. Here, we investigate a novel strategy to jointly optimize both photon and carbon ion fractions in order to determine the optimal combination of both modalities for Glioblastoma (GBM) cases. Material and Methods Joint optimization of pencil beam intensities for carbon ions and of fluence maps for photons is performed based on the cumulative biological effect (BE) of the two modalities, i.e., the negative log of the survival fraction from the linear quadratic cell survival model. The carbon ion effect is calculated based on the Local Effect Model; constant radiobiological parameters alpha and beta are assumed for photons. Dose prescriptions and normal tissue constraints are adopted from the CLEOPATRA-GBM trial, i.e., 25 photon fractions deliver 50Gy to the CTV and a 6 carbon ion fractions deliver 18Gy (RBE) to the GTV. For joint optimization, the total prescriptions are converted to BE; additional objectives are to limit the dose received by normal brain to an effect equivalent to a photon dose of 50Gy in 25 fractions, a 5mm CTV margin conformity objective, and minimization of the mean BE in normal brain. A photon alpha/beta of 10 and 2 is assumed for tumor and normal tissue, respectively. Results Figure 1 shows the optimized combination of carbon ion fractions (b), photon fractions (d), and their cumulative effect converted to equi-effective dose in 2Gy photon fractions (f). For comparison, sub-figures (a), (c) and (e) show the treatment scheme used in the CLEOPATRA trial. Corresponding cumulative Effect Volume Histograms are shown in Figure 2. Overall, the optimal combination exhibits better conformity, for comparable target coverage and reveals that the dose delivered to the CTV (outside of the GTV) is almost entirely delivered by photons. Compared to the simple combination of the CLEOPATRA trial, the optimized combination increases the carbon ion dose contribution in the centre of the GTV by more than 100% and reduces the photon dose accordingly. At the boundary of GTV and CTV, the photon contribution is increased, resulting in a 16% reduction of mean effect to the normal tissue in the CTV.

Results The daily slot allocation strategy leads to a higher reduction of the average NTCP values for xerostomia than the threshold-based PT patient selection as shown in Figure 2 for any number of available proton slot per day. If all patients receive only photons or only protons, the average NTCP values for xerostomia are 16.9% and 6.3%, respectively. If 3 proton slots are available per day for HNC patients, the average NTCP value for xerostomia is 12.5% for the daily slot allocation strategy and 14.0% for the threshold-based PT patient selection which would select patients with 10% ΔNTCP threshold. The NTCP benefit of 1.5% can be explained by two considerations: 1) combined proton-photon treatments make optimal use of all proton slots whereas patient selection strategies face a trade-off between leaving slots unused or blocking slots for future patients with higher benefit; 2) on the convex part of the NTCP curve , the first proton fractions delivered are the most benefcial.

Conclusion Limited proton therapy resources can be more efficiently utilized, from a global health system perspective, with combined proton-photon treatments with daily allocation of proton slots compared to single-modality treatments with optimal patient selection.