Abstract Book

S266

ESTRO 37

Conclusion Rectum NTCP models derived from photon therapy differed considerably in how well they fitted prospectively recorded rectum morbidity after PT under the constant RBE assumption. Since five of six models underestimated the clinically observed level of morbidity, this points towards a combined effect of enhanced RBE (above 1.1) and a potential difference associated with the volume effect for rectal morbidity. OC-0511 NTCP modelling and external validation of early side effects for proton therapy of brain tumours A. Dutz 1,2,3 , L. Agolli 1,4 , E.G.C. Troost 1,2,3,4,5 , M. Krause 1,2,3,4,5 , M. Baumann 1,2,3,4,5,6 , A. Lühr 1,2,3 , X. Vermeren 7 , D. Geismar 7,8,9 , B. Timmermann 7,8,9,10 , S. Löck 1,3,4,5 1 OncoRay - Center for Radiation Research in Oncology, Faculty of Medicine and University Hospital Carl Gustav Carus- Technische Universität Dresden, Dresden, Germany 2 Helmholtz-Zentrum Dresden - Rossendorf, Institute of Radiooncology – OncoRay, Dresden, Germany 3 German Cancer Consortium DKTK- partner site Dresden, and German Cancer Research Center DKFZ, Heidelberg, Germany 4 Faculty of Medicine and University Hospital Carl Gustav Carus- Technische Universität Dresden, Department of Radiotherapy and Radiation Oncology, Dresden, Germany 5 National Center for Tumor Diseases NCT, partner site Dresden, Dresden, Germany 6 German Cancer Research Center, DKFZ, Heidelberg, Germany 7 West German Proton Therapy Center Essen WPE, University Hospital Essen, Essen, Germany 8 Clinic for Particle Therapy, University Hospital Essen, Essen, Germany 9 West German Cancer Center WTZ, University Hospital Essen, Essen, Germany 10 German Cancer Consortium DKTK, partner site Essen, Essen, Germany Purpose or Objective To identify patients who are likely to benefit most from proton beam therapy (PBT), based on the potential reduction in normal tissue complication probability (NTCP) compared to photon therapy. The NTCP models required for this decision were developed using clinical data on early side effects for patients with brain tumours Two cohorts of adult patients with brain tumours who received PBT at two different proton therapy centres were included in this study: 113 patients were used for model training and 71 cases were used for external validation. Moreover, volumetric modulated arc therapy (VMAT) plans were retrospectively created for all patients to predict the potential reduction in NTCP when applying PBT. The radiation-induced early side effects alopecia, erythema, pain and fatigue (CTCAE v4.0) were investigated. The occurrence of these side effects was correlated with different dose-volume histogram (DVH) parameters of associated organs at risk (OAR), such as skin and remaining brain. NTCP models were created using logistic regression and the area under the receiver operating characteristic curve (AUC) was used to assess their prognostic ability. Results The NTCP models revealed significant correlations between the incidence of alopecia grade 2 (figure 1) as well as erythema grade≥1 and the DVH parameters D1%, D2%, V15Gy and V20Gy of the skin. The V20Gy models showed a very good discrimination on external validation for both endpoints (AUC ≥ 0.75, figure 2). No correlations between DVH parameters of the remaining brain and the incidence of fatigue or pain were found. Dose comparison who underwent PBT. Material and Methods

scored gastrointestinal (GI) morbidity Grade >= 2 was used as endpoint in the analysis, with a total of 180 events (most of them bleeding). RBE-weighted dose volume histograms assuming RBE = 1.1 were extracted for the rectum for all patients and were subsequently used as input to probit NTCP models using six different rectum model parameter sets including those based on Emami/Burman and from QUANTEC (Table 1). For each model, patients were sorted by NTCP values (low to high) and grouped in 19 groups with 58 patients in each. The mean NTCP of each group was calculated and plotted against the clinically observed frequency of rectal morbidity in each group. The Spearman's rank correlation coefficients for the clinically observed vs the calculated NTCPs for each model were calculated.

Results All photon-based NTCP models gave reasonable fits to our PT data (R 2 in the range 0.60-0.69), but there were clear differences in how well they reproduced the observed level of morbidity (Fig. 1). Model I (QUANTEC) and Model II showed the best agreement with observed morbidity, also with a slope similar to the identity line. For five of six models the data points were well above the identity line. Model III (Emami/Burman), Model IV, and Model VI underestimated the risk of morbidity the most, in particular for patients with the highest risk.