Abstract Book

S157

ESTRO 37

OC-0301 NTCP-model based patient selection for hypofractionated prostate treatment - A computer simulation R. Bijman 1 , A.W. Sharfo 1 , W. Schillemans 1 , W. Heemsbergen 1 , M. Witte 2 , F. Pos 2 , L. Incrocci 1 , B. Heijmen 1 1 Erasmus MC Cancer Center, Radiation Oncology, Rotterdam, The Netherlands 2 Antoni van Leeuwenhoek Hospital, Radiation Oncology, Amsterdam, The Netherlands Purpose or Objective External beam radiotherapy (EBRT) is one of the primary treatment modalities for prostate cancer. In the last years, hypofractionated EBRT has gained increasing popularity for prostate cancer treatment, largely related to reduced hospital visits for the patients and lower overall costs. In case hypofractionation has a (slightly) enhanced risk for late toxicity compared to conventional fractionation, this has to be balanced against the advantages. To this purpose, automated planning could be used to generate for each individual patient a plan for conventional fractionation and a plan for hypofractionation. Patient selection could then be based on differences in calculated NTCPs (similar to model- based patient selection for proton therapy). In this work, we have performed computer simulations to establish percentages of patients that could be treated with hypofractionation. These percentages were determined as a function of accepted thresholds for increases in calculated normal tissue complication probabilities (NTCP) relative to conventional fractionation. Material and Methods Data from the randomized phase 3 HYPRO trial were used in this study. In this trial, 820 patients were treated with either 39x2 Gy or 19x3.4 Gy. Overall, reported incidences of grade ≥ 2 genitourinary (GU) and gastrointestinal (GI) toxicity (RTOG/EORTC criteria) at 3 years were 40.2% and 19.8%, respectively. For 724 patients, reported late toxicities and the clinically delivered dose distributions and clinical baseline parameters were available and used to derive NTCP models for these toxicities, using logistic regression. For this study, fully automated treatment plan generation was then used to generate for all patients a VMAT plan for conventional fractionation and a plan for hypofractionation with similar planning constraints relative to the prescribed dose. When the differences in GU and GI NTCP (ΔNTCP) between the two fractionation schemes were below pre-defined thresholds, the patient was assumed eligible for hypofractionation. Percentages of patients being selected for hypofractionation were established as a function of ΔNTCP thresholds for GI and GU complications. Results Differences in PTV coverage between the two arms were not significant (ΔV 95% = 0.03%, p = .83). Figure 1 shows that with accepted NTCP increases of 4.7% and 7.0% for GU and GI toxicity, respectively, 95% of the patients could be treated with hypofractionation. Accepted GU and GI NTCP increases of 4.3% and 0% would result in only 5% of patients treated with hypofractionation. For acceptable GU and GI NTCP increases of 4.6% and 2.4%, 50% of patients could be treated with hypofractionation.

GTV, stomach, duodenum, liver, bowel bag, spinal cord and kidneys were contoured on the simulation CT scan by two radiation oncologists. CTV was considered equal to GTV. PTV was generated from CTV, adding an isotropic 3 mm margin and excluding areas overlapping with OARs. IMRT plans with 20 equal distant angular beams were elaborated for the MR-Linac, while dual arc VMAT plans were optimised for the SBRT Linac. Prescription dose was 40Gy in 5 fractions (normalisation= 40Gy at 50% of PTV). All plans had to strictly respect the following OARs constraints: V33Gy<1cc and V25Gy<20cc for duodenum, stomach and bowel bag, V12Gy<50% for liver and kidneys, V25Gy for spinal cord. The dosimetric comparison evaluated the target coverage (D98 for CTV and V95 for PTV), dose homogeneity (HI) according to ICRU 83, conformity index (CI) according to ICRU 62 and dose diffusion (V10Gy, V20Gy). Statistical significance was estimated calculating the Wilcoxon Mann Whitney test for paired samples. Results Seven plans were considered clinical acceptable (V95 PTV>95%). No statistical significance was observed for D98 CTV (p-value =0.13) and V95 PTV (p=0.11) (see Figure 1).

V105 values were always inferior to 2% for both CTV and PTV in all cases analysed. No significant difference was observed for OARs irradiation, dose constraints were always met. The analysis of dose diffusion is reported in Fig.2.

In six cases the low dose values are more spread in the MR-Linac plans, even if no statistical significance was observed. The plan quality indicators are also comparable: the CI mean value calculated for PTV was 1.21±0.16 for MR- Linac and 1.17±0.7 for SBRT Linac. The PTV HI mean value was 0.10±0.03 for MR-Linac, 0.07±0.04 for SBRT Linac. Conclusion The dosimetric performance between MR and SBRT Linac are comparable for SBRT pancreas treatment. Plans calculated with SBRT Linac show less dose diffusion, probably due to the different optimisation software, that is more complex but also slower in processing in comparison to MR-Linac one (optimisation mean time is 2 min for MR-Linac, 20 min for SBRT Linac). Considering the MR-ART need to optimise the treatment plan while the patient is on the couch, MRIdian Linac allows to elaborate SBRT plans clinically and dosimetrically satisfying.