ESTRO 2025 - Abstract Book

S1113

Clinical – Head & neck

ESTRO 2025

4086

Digital Poster Outcome Prediction Before Definitive Chemoradiation in Older Adults With Head-and-Neck Cancer Using Artificial Neural Networks Sebastian Norbert Marschner 1,2 , Elia Lombardo 1 , Erik Haehl 1 , Carmen Kut 3 , Marlen Haderlein 4 , Alexander Fabian 5 , Carolin Senger 6,7 , Daniel R. Dickstein 8 , Victor Lewitzki 9 , Sujith Baliga 10 , Jens Von der Grün 11,12 , Eric Chen 13 , Constantinos Zamboglou 14,15,16 , Arnulf Mayer 11 , Panagiotis Balermpas 12 , Harry Quon 3 , Carmen Stromberger 17,18 , Anca-Ligia Grosu 14 , Guillaume Landry 1 , Franziska Walter 1 , Claus Belka 1,2,19 , Nils H. Nicolay 14,20,21 , Alexander Rühle 14,20,21 1 Department of Radiation Oncology, University Hospital, LMU Munich, Munich, Germany. 2 German Cancer Consortium (DKTK) Partner Site Munich, German Cancer Research Center (DKFZ), Heidelberg, Munich, Germany. 3 Department of Radiation Oncology, Johns Hopkins University, Baltimore, MD, USA. 4 Department of Radiation Oncology, Universitätsklinikum Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany. 5 Department of Radiation Oncology, University Hospital Schleswig-Holstein, Campus Kiel, Kiel, Germany. 6 Department of Radiation Oncology, Charité - Universitätsmedizin Berlin; Freie Universität Berlin, Humboldt Universität zu Berlin, Berlin, Germany. 7 German Cancer Consortium (DKTK) Partner Site Berlin, German Cancer Research Center (DKFZ), Berlin, Germany. 8 Department of Radiation Oncology, Icahn School of Medicine at Mount Sinai, New York, USA. 9 Department of Radiation Oncology, University Hospital Würzburg, Würzburg, Germany. 10 Department of Radiation Oncology, Ohio State University Wexner Medical Center, Columbus, Ohio, USA. 11 Department of Radiotherapy and Oncology, Goethe University Frankfurt, Frankfurt am Main, Germany. 12 Department of Radiation Oncology, University Hospital Zurich, University of Zurich, Zurich, Switzerland. 13 Department of Radiation Oncology, University Hospitals Seidman Cancer Center, Cleveland, Ohio, USA. 14 Department of Radiation Oncology, University of Freiburg - Medical Center, Freiburg, Germany. 15 German Cancer Consortium (DKTK) Partner Site Freiburg, German Cancer Research Center (DKFZ), Heidelberg, Freiburg, Germany. 16 German Oncology Center, European University of Cyprus, Limassol, Cyprus. 17 Department of Radiotherapy and Oncology, Charité - Universitätsmedizin Berlin; Freie Universität Berlin, Humboldt-Universität zu Berlin; and Berlin Institute of Health, Berlin, Germany. 18 German Cancer Consortium (DKTK) Partner Site Berlin, German Cancer Research Center (DKFZ), Heidelberg, Berlin, Germany. 19 Bavarian Cancer Research Center, (BZKF), Munich, Germany. 20 Department of Radiation Oncology, University of Leipzig, Leipzig, Germany. 21 Cancer Center Central Germany (CCCG), Partner Site Leipzig, Leipzig, Germany Purpose/Objective: The elderly population (aged >65) with head-and-neck squamous cell carcinoma (HNSCC) is underrepresented in clinical trials, leading to uncertainty regarding the efficacy of cisplatin-based chemoradiotherapy due to the comorbidities and the overall limited life expectancy. This study aimed to develop two machine learning models— one to predict overall survival (OS) and another for progression-free survival (PFS)—to distinguish between elderly high-risk and low-risk patients with locally advanced HNSCC undergoing definitive chemoradiation. Our goal was to identify patients at higher risk of mortality following treatment, enabling clinicians to make more informed decisions tailored to each patient’s risk profile. Material/Methods: Utilizing data from the SENIOR cohort (NCT05337631), an international registry now comprising 1,281 elderly patients (aged >65) with locoregionally advanced HNSCC, we developed two separate artificial neural networks (ANNs) to classify patients with definitive chemoradiation into high-risk (<2 years OS/PFS) and low-risk (>2 years OS/PFS) groups. The model was trained and internally validated on a cohort of 738 patients for OS and 770 patients for PFS. For external validation, an independent testing cohort from a different institution was used, consisting of 84 patients for OS analysis and 89 patients for PFS analysis. Performance evaluation was conducted using log-rank statistical tests, ROC-AUC and PR-AUC metrics on an independent testing cohort comprising patients from a different institution. SHAP (Shapley Additive exPlanations) values were employed for model interpretability.

Results: