ESTRO 2020 Abstract book

S883 ESTRO 2020

stratify patients from the validation cohort into groups with significantly different LRC (p=0.07), while the model based on imaging features and the combined model led to significant patient stratifications in validation (p=0.04, see figure). Selected imaging features were weakly correlated to GTV (R²≤0.3).

Germany ; 6 Department of Nuclear Medicine, Faculty of Medicine and University Hospital Carl Gustav Carus- Technische Universität Dresden, Dresden, Germany ; 7 Helmholtz-Zentrum Dresden-Rossendorf, PET Center- Institute of Radiopharmaceutical Cancer Research, Dresden, Germany ; 8 Department of Radiotherapy, Hospital Dresden-Friedrichstadt, Dresden, Germany ; 9 Department of Radiooncology and Radiotherapy, Charité University Hospital, Berlin, Germany ; 10 German Cancer Research Center DKFZ, Heidelberg and German Cancer Consortium DKTK partner site Berlin, Berlin, Germany ; 11 Department of Radiotherapy- Medical Faculty, University of Duisburg-Essen, Essen, Germany ; 12 German Cancer Research Center DKFZ, Heidelberg and German Cancer Consortium DKTK partner site Essen, Essen, Germany ; 13 Department of Radiotherapy and Oncology, Goethe-University Frankfurt, Frankfurt, Germany ; 14 German Cancer Research Center DKFZ, Heidelberg and German Cancer Consortium DKTK partner site Frankfurt, Frankfurt, Germany ; 15 Department of Radiotherapy and Radiation Oncology, Ludwig-Maximilians-Universität, Munich, Germany ; 16 German Cancer Research Center DKFZ, Heidelberg and German Cancer Consortium DKTK partner site Munich, Munich, Germany ; 17 Department of Radiation Oncology, Technische Universität München, Munich, Germany ; 18 Department of Radiation Oncology, Faculty of Medicine and University Hospital Tübingen- Eberhard Karls Universität Tübingen, Tübingen, Germany ; 19 German Cancer Research Center DKFZ, Heidelberg and German Cancer Consortium DKTK partner site Tübingen, Tübingen, Germany ; 20 German Cancer Research Center, Deutsches Krebsforschungszentrum DKFZ, Heidelberg, Germany Purpose or Objective To improve personalised treatment of locally advanced head and neck squamous cell carcinoma (HNSCC), a comprehensive biomarker trial was initiated by the German Cancer Consortium - Radiation Oncology Group (DKTK-ROG). Biomarkers that are identified in the retrospective part of this trial will be validated in a prospective validation cohort. Thus, the aim of this retrospective study was to define prognostic radiomic models for loco-regional control (LRC) after primary radiochemotherapy (pRCTx) using pre-treatment computed tomography (CT) imaging, before subsequent A multicentre cohort including 318 locally advanced HNSCC patients was collected and divided into training and validation cohorts (233 and 85 patients, respectively). All patients underwent a non-contrast-enhanced CT scan for treatment-planning and received pRCTx at one of six DKTK-ROG partner sites [1,2]. 458 imaging features were derived from the primary gross tumour volume (GTV); 16 clinical features were considered, including age, T and N stage, GTV and HPV status. Three prognostic models were developed for predicting LRC based on: (i) imaging features, (ii) clinical features, and (iii) a combination of both feature classes. Different machine learning algorithms were compared based on the training cohort as outlined in [2]. Model performance was measured using the concordance index (C-Index). Subsequently, patients were stratified into groups of low and high risk of loco-regional recurrence using the median risk value. Differences in LRC were assessed by the log-rank test. Results Elastic-net Cox regression models, after minimum redundancy maximum relevance feature selection, performed best in training and achieved the following C- indices in validation: (i) 0.62±0.05 (mean±std) for the model based on imaging features, (ii) 0.62±0.03 for clinical features, and (iii) 0.67±0.04 for the combination of imaging and clinical features. The model based on the selected clinical features GTV and N stage could not prospective validation. Material and Methods

Conclusion We developed and validated models that predict LRC for patients with locally advanced HNSCC treated by pRCTx. Combining CT imaging with clinical features improved model performance. In future, the developed models will be validated on the prospective HNSCC cohort of the DKTK- ROG once follow-up is complete. After combination with validated molecular biomarkers, they may be applied in a planned interventional trial on individual dose prescription. References: [1] Linge et al. Radiother Oncol 121: 364 (2016). [2] Leger et al. Sci Rep 7: 13206 (2017). PO-1541 Predictive treatment planning with SIP (simultaneously integrated protection) based on TCP and NTCP B. Thomann 1 , D. Baltas 1 , I. Sachpazidis 1 , T. Fechter 1 , A. Grosu 2 , T. Brunner 3 , E. Gkika 2 1 University Medical Center Freiburg, Department of Radiation Oncology- Division of Medical Physics, Freiburg, Germany ; 2 University Medical Center Freiburg, Department of Radiation Oncology, Freiburg, Germany ; 3 University Medical Center Magdeburg, Department of Radiation Oncology, Magdeburg, Germany is defined as the overlapping region of the PTV and the planning risk volume (PRV), both resulting from organ movement (4D-CT) and safety margins. The target dose in PTV SIP is reduced in order to lower the normal tissue complication probability (NTCP) while the target dose in the remaining, dominant PTV dom is increased in order to maintain a sufficient tumour control probability (TCP). The concept has previously been validated with in silico simulations and on patient data. The aim of this study is to transfer the TCP- and NTCP-based planning with SIP into For a group of 10 patients with pancreas carcinoma the GTV was defined in every respiratory phase of a 4D-CT. The ITV was created as the union volume of these GTVs in the average CT and the PTV resulted from a 4 mm margin around the ITV. PTV SIP was defined as the overlap between the PTV and the PRV of the small bowel. The original treatment plans gave 48 Gy to PTV dom (EQD2 = 56 Gy) and 42 Gy (EQD2 = 47.25 Gy) to PTV SIP in 12 fractions. Previous studies have shown the benefit of the SIP concept compared to a standard, homogeneous dose prescription. In this study, additional treatment plans were created, keeping the dose objective for PTV SIP at EQD2 = 47.25 Gy but varying the dose objective for PTV dom to identify the optimal dose prescription in terms of TCP and NTCP. TCP (for GTV pancreas) and NTCP (for small bowel) were Purpose or Objective The PTV SIP the clinical workflow. Material and Methods