ESTRO 2021 Abstract Book

S383

ESTRO 2021

Gemelli Molise Hospital – Università Cattolica del Sacro Cuore, Campobasso, Italy; 5 Istituto di Radiologia, Università Cattolica del Sacro Cuore, Roma, Italy; 6 Medical Physics Unit, Gemelli Molise Hospital – Università Cattolica del Sacro Cuore, Campobasso, Italy; 7 Radiation Oncology, San Rossore Private Hospital, Pisa, Italy; 8 Radiotherapy Unit, Ospedale Belcolle, Viterbo, Italy; 9 Institute of Oncology Vojvodina, Sremska Kamenica, Medical Faculty, Novi Sad University, Novi Sad, Serbia; 10 Unit of Radiation Therapy, IRCCS “Casa Sollievo della Sofferenza”, San Giovanni Rotondo, Italy; 11 UOC di Radioterapia, Dipartimento di Scienze Radiologiche, Radioterapiche ed Ematologiche, Fondazione Policlinico Universitario A. Gemelli IRCCS, Roma, Italy; 12 Radiation Oncology, Campus Biomedico University, Rome, Italy; 13 Radiation Oncology, Azienda Ospedaliero Universitaria Careggi, Florence, Italy; 14 Department of Radiotherapy, Policlinico San Martino, University of Genoa, Genoa, Italy; 15 Radiotherapy Unit, Azienda Ospedaliera San Gerardo, Monza, Italy; 16 Radiation Therapy, Candiolo Cancer Institute - FPO, IRCCS Candiolo, Candiolo, Italy; 17 Medical Oncology, IRCCS Azienda Ospedaliero-Universitaria di Bologna, Bologna, Italy; 18 Department of Radiotherapy, Azienda Ospedaliera Universitaria Integrata, Verona, Italy Purpose or Objective International guidelines report a wide range of options in locally advanced pancreatic cancer (LAPC): chemotherapy (CHT), chemoradiation (CRT), or stereotactic body radiotherapy (SBRT) +/- CHT. Purpose of this analysis was to retrospectively compare these approaches in terms of the following outcomes: local control (LC), distant metastasis free survival (DMFS), and overall survival (OS), and to propose a predictive model of LC in LAPC. Materials and Methods LAPC patients from a multicentric retrospective database (PAULA-1), treated with definitive CHT, or CRT, or SBRT+/- CHT were included. Kaplan-Meier survival curves were compared with the log-rank test. A multivariate analysis (Cox’s) was performed to identify predictors of LC, OS and DMFS. A predictive model for LC in LAPC was developed based on a random forest machine learning method. Results Median follow-up was 16.6 months (range: 3.0-92.0). Of 419 LAPC included, 298 (71.1%) were treated with CRT, 65 (15.5%) with CHT and 56 (13.4%) with SBRT. At univariate and multivariate analysis, pancreatic body site (p=0.002) and SBRT+/- CHT treatment (p=0.042) were both significantly correlated with improved LC. At univariate analysis and multivariate analysis, tumor site in the tail (p= 0.043) and ECOG 2 status (p=0.009) were both significantly correlated to improved OS and worse OS, respectively. At multivariable analysis, increased CA19-9 negatively impacted on LC, DMFS, and OS. Finally, the LC predictive model AUC was 68% (CI 58.7%-77.4%) (Figure 1). Figure 1: Multivariate analysis of the effects of patient characteristics and treatments on local control

Conclusion Results for SBRT +/CHT are comparable to definitive CHT and CRT, in terms of OS and DMFS. Moreover,