ESTRO 2025 - Abstract Book

S1785

Clinical – Upper GI

ESTRO 2025

References: Teoh S et al. Evaluation of hypofractionated adaptive radiotherapy using the MR Linac in localised pancreatic cancer: protocol summary of the Emerald-Pancreas phase 1/expansion study located at Oxford University Hospital, UK. BMJ Open. 2023 Sep 14;13(9):e068906. 2026 Mini-Oral Development and multicentre validation of an NTCP-model for pneumonia in oesophageal cancer patients after neoadjuvant chemoradiotherapy and surgery Mark L. Frederiks 1 , Bas P.L. Wijnhoven 2 , Ewoud Schuit 3 , Peter S.N. van Rossum 4 , Hanneke W.M. van Laarhoven 5 , Gert J. Meijer 6 , Stella Mook 6 , Joost J. Nuyttens 7 , Heidi Rütten 8 , Bastiaan R. Klarenbeek 9 , Mariska D. den Hartogh 10 , Maaike Berbée 11 , Meindert Sosef 12 , Boudewijn van Etten 13 , Rob H.A. Verhoeven 14 , Johannes A. Langendijk 1 , Christina T. Muijs 1 1 Department of Radiation Oncology, University Medical Center Groningen, Groningen, Netherlands. 2 Department of Surgery, Erasmus MC Cancer Institute, Rotterdam, Netherlands. 3 Department of Epidemiology & Health Economics, University Medical Center Utrecht, Utrecht, Netherlands. 4 Department of Radiation Oncology, Amsterdam UMC, Amsterdam, Netherlands. 5 Department of Medical Oncology, Amsterdam UMC, Amsterdam, Netherlands. 6 Department of Radiation Oncology, University Medical Center Utrecht, Utrecht, Netherlands. 7 Department of Radiation Oncology, Erasmus MC Cancer Institute, Rotterdam, Netherlands. 8 Department of Radiation Oncology, Radboud University Nijmegen Medical Centre, Nijmegen, Netherlands. 9 Department of Surgery, Radboud University Nijmegen Medical Centre, Nijmegen, Netherlands. 10 Department of Radiation Oncology, Radiotherapiegroep, Arnhem, Netherlands. 11 Department of Radiation Oncology, Maastro Clinic, Maastricht, Netherlands. 12 Department of Surgery, Zuyderland Medical Center, Heerlen, Netherlands. 13 Department of Surgery, University Medical Center Groningen, Groningen, Netherlands. 14 Department of Research and Development, Netherlands Comprehensive Cancer Organisation, Utrecht, Netherlands Purpose/Objective: Pneumonia negatively impacts quality of life and mortality in patients with oesophageal cancer (EC) receiving neoadjuvant chemoradiotherapy (nCRT) followed by surgery. Despite the risk posed by radiotherapy, validated prediction models are lacking. Normal Tissue Complication Probability (NTCP)-models are crucial for optimizing radiotherapy treatment plans and reducing complication risks. This study aims to develop and validate an NTCP model for pneumonia using a comprehensive multicentre, multidisciplinary dataset, constructed in collaboration with the Netherlands Cancer Registry, the Dutch Upper GI Cancer Audit and six treatment centres. Material/Methods: Data from EC patients treated with nCRT between 2015-2021 were retrospectively collected. Pneumonia was defined as CTCAE grade ≥ 2 during or within six months post-nCRT. To ensure model performance and generalizability, a one stage internal-external validation procedure was performed and a stratified intercept was estimated across centres 1 . Missing data were addressed, separately for the training and validation set, using multiple imputation. Dose-volume-histogram (DVH)-data were calculated for the heart and lungs, and a principal component (PC)-analysis was used to mitigate its multicollinearity and condense all DVH-values to dose patterns. Results: Model development and validation was performed on data from five centres, including 1447 patients. The incidence of pneumonia was 24%. Selected predictors for the final model are listed in Figure1. The PC-analysis resulted into three components, with PC1 and PC3 as selected predictors. The dose pattern of PC1 refers to the mean dose to the heart and lungs (Figure2). PC3’s dose pattern represents the dose to the lungs (primarily lung V1 to V20) versus the dose to the heart (mainly heart V20 toV40) (Figure2). The model showed fair discrimination across centres and imputation sets, with an average AUC of 0.67. Calibration was good, with an average calibration slope of 0.91 (ideally:1) and calibration-in-the-large of 0.08 (ideally:0). Figure1 depicts centre-specific results.