ESTRO 2024 - Abstract Book

S1341

Clinical - Head & neck

ESTRO 2024

Stefania Volpe 1,2 , Maria G Vincini 1 , Mattia Zaffaroni 1 , Sara Raimondi 3 , Liliana Belgioia 4 , Elisabetta Bonzano 5 , Sara Colombo 5 , Alessia Di Rito 6 , Domenico A Romanello 7 , Camilla Satragno 4 , Daniela Sibio 8 , Federico Mastroleo 1,9 , Daniela Alterio 1 , Roberto Orecchia 10 , Sara Gandini 3 , Barbara A Jereczek-Fossa 1,2 1 European Institute of Oncology IRCCS, Department of Radiation Oncology, Milan, Italy. 2 University of Milan, Department of Oncology and Hemato-Oncology, Milan, Italy. 3 European Institute of Oncology IRCCS, Department of Experimental Oncology, Milan, Italy. 4 University of Genoa, Department of Health Sciences (DISSAL), Genoa, Italy. 5 Fondazione IRCCS Policlinico San Matteo, Department of Radiation Oncology, Pavia, Italy. 6 "Mons. Dimiccoli" Hospital, Radiotherapy Unit, Barletta, Italy. 7 MedAustron Ion Therapy Center, Department of Radiation Oncology, Vienna, Austria. 8 ASST Grande Ospedale Metropolitano Niguarda, Department of Radiation Oncology, Milan, Italy. 9 University of Piemonte Orientale (UPO), Department of Translational Medicine, Novara, Italy. 10 European Institute of Oncology IRCCS, Scientific Directorate, Milan, Italy

Purpose/Objective:

The increasing availability of radiomic studies has led to the creation of several scoring systems, to assess the quality of evidence through a systematic approach. However, the reproducibility of these scores has never been tested. Hence, this work aims to assess the inter-observer agreement among Radiation Oncologists (ROs) and non clinical professionals in scoring available literature on radiomic applications in magnetic resonance imaging (MRI)- based studies for nasopharyngeal cancer (NPC). Two popular scores were considered, namely the Luo score and the Radiomic Quality Score (RQS).

Material/Methods:

In January 2023, a PRISMA-compliant systematic review was performed on the Embase, Scopus and NCBI Pubmed datasets with the following inclusion criteria: 1) adult series including NPC only, 2) studies describing oncological and/or toxicities outcomes, and 3) availability of full-text original research articles. Therefore, manuscripts on diagnostic applications of radiomics were not considered, as well as systematic reviews and metanalysis, which were however screened for cross-referencing. Works meeting the inclusion criteria were independently rated by four ROs, one statistician and one biotechnologist. All raters had dedicated experience in radiomics and/or NPC treatment. Inter-observer agreement among all the readers and between ROs and non-clinical researchers was assessed by the interclass correlation coefficient (ICC) with 95% confidence intervals (CIs). The Bland-Altman method was implemented to provide the average difference between clinical and non-clinical scores with 95% limits of agreement; these were then compared by using the Wilcoxon signed rank test. P-values< 0.05 were considered statistically significant. Analyses were performed with the SAS software v.9.4. This study was endorsed by the AIRO (Italian Association of Radiation Oncology and Clinical Oncology), and it is registered on PROSPERO (ID CRD42022324919).

Results:

Of the 819 identified unique records, 31 were included in the analysis. The median number of included patients was 152, and contrast-enhanced T1-weighted images were the most commonly studied, followed by T2-weighed sequences. Image pre-processing was implemented in the majority of cases (22/31). Considering scoring, all raters were able to successfully complete their task. Their agreement was higher for the RQS (0.79; 95% CI: 0.69-0.88) than