ESTRO 2020 Abstract book

S325 ESTRO 2020

PD-0543 An externally validated prognostic CT radiomics model for head and neck cancer patients S. Keek 1 , S. Sanduleanu 1 , H.C. Woodruff 1,2 , F. Wesseling 3 , D. Mattavelli 4 , M. Ravanelli 5 , T.K. Hoffman 6 , K. Scheckenbach 7 , C.R. Leemans 8 , P. De Graaf 9 , C. Terhaard 10 , J. Van de Kamer 11 , M. Van der Heijden 12 , G. Calareso 13 , E.S. Gazzani 14 , L. Licitra 15,16 , F. Hoebers 3 , T. Poli 17 , P. Lambin 1,2 1 University of Maastricht GROW Research Institute, Department of Precision Medicine, Maastricht, The Netherlands ; 2 Maastricht University Medical Centre+, Department of Radiology and Nuclear medicine, Maastricht, The Netherlands ; 3 Maastricht University Medical Centre+, Department of Radiation Oncology MAASTRO, Maastricht, The Netherlands ; 4 University of Brescia, Unit of Otorhinolaryngology-Head and Neck Surgery, Brescia, Italy ; 5 University of Brescia, Department of Medicine and Surgery, Brescia, Italy ; 6 University Hospital Ulm, Dept. of Otorhinolaryngology- Head and Neck Surgery, Ulm, Germany ; 7 University Hospital Düsseldorf, Dept. of Otorhinolaryngology- Head and Neck Surgery, Düsseldorf, Germany ; 8 Amsterdam UMC, KNO-heelkunde/Hoofd-halschirurgie, Amsterdam, The Netherlands ; 9 Amsterdam UMC, Department of Radiology, Amsterdam, The Netherlands ; 10 University Medical Center Utrecht, Department of Radiotherapy, Utrecht, The Netherlands ; 11 Antoni van Leeuwenhoek- Netherlands Cancer Institute, Department of Radiotherapy, Amsterdam, The Netherlands ; 12 Antoni van Leeuwenhoek- Netherlands Cancer Institute, Department of Head and Neck Oncology and Surgery, Amsterdam, The Netherlands ; 13 Fondazione IRCCS Istituto Nazionale dei Tumori, Department of Radiology, Milan, Italy ; 14 University of Parma, Department of Radiology, Parma, Italy ; 15 Fondazione IRCCS Istituto Nazionale dei Tumori, Head and Neck Medical Oncology Department, Milan, Italy ; 16 University of Milan, Department of Oncology and Hematology-Oncology, Milan, Italy ; 17 University of Parma, Department of Surgical Sciences, Parma, Italy Purpose or Objective Patients with locoregionally advanced head-and-neck squamous cell carcinoma (HNSCC) have high relapse and mortality rates[WH(1] . Imaging-based decision support could help improve prognosis by preventing unnecessary treatment or escalating treatment for suitably stratified patients. We investigate whether a CT radiomics-based prognostic model for TNM-7 stage-III and –IV HNSCC can stratify patients into low- and high-risk groups, validated on a prospective cohort and external validation dataset. Material and Methods 699 retrospective (training) and 138 prospective (validation) stage III-IV (M0) HNSCC patients from six European hospitals were collected from the BD2Decide database (ClinicalTrials: NCT02832102). 183 external (validation) stage III-IV (M0) HNSCC patients from two Dutch hospitals were collected from the DESIGN database. Patients were treated with surgery and/or (chemo- )radiotherapy. A multivariable Cox regression model was trained to predict overall survival (OS) based on radiomics features derived from the gross tumor volume (GTV) of the primary tumor. ComBat harmonization was used to improve interchangeability of features between the different centers. Patient stratification was assessed through Kaplan-Meier (KM) survival curves and log-rank test for significance (p-value<0.05). The goodness of fit of the model was reported through the concordance-index (CI).

Results A 3-feature radiomic signature (1 shape feature, 2 texture features) is not able to significantly group the prospective validation cohort into two distinct risk groups (p = 0.16), with a CI of 0.66 (figure 1). However, the signature was able to significantly differentiate two risk groups in the external validation cohort in (p < 0.05), although predictive performance in this cohort was much lower with a CI of 0.57 (figure 2).