ESTRO38 Congress Report

Radiobiology

4. Immunological contextural basis of a prognostic radiomics signature in head and neck cancers (E38-0734) Dan Ou 1,2,3,4 , Julien Adam 5 , Ingrid Garberis 5,6 , Pierre Blanchard 2 , France Nguyen 2 , Antonin Levy 2,3,4 , Odile Casiraghi 5 , Ralph T.H. Leijenaar 7 , Philippe Gorphe 8 , Ingrid Breuskin 8 , François Janot 8 , Charlotte Robert 2,3,4 , Philippe Lambin 7 ,Stephane Temam 8 , Jean-Yves Scoazec 3,5,6 , Eric Deutsch 2,3,4 , Yungan Tao 2,3,4* 1 Department of Radiation Oncology, Fudan University Shanghai Cancer Center, Shanghai, China, 2 Department of Radiation Oncology, Gustave Roussy Cancer Campus, Villejuif, France, 3 Université Paris Sud, France, 4 INSERM1030 molecular radiotherapy, Villejuif, France, 5 Department of Pathology, Gustave Roussy Cancer Campus, Villejuif, France, 6 INSERM US23/CNRS UMS3655, Molecular analysis, modelling and imaging of cancer disease, Experimental and Translational Pathology, Villejuif, France, 7 The D-Lab: Decision Support for Precision Medicine, GROW - School for Oncology and Developmental Biology, Maastricht University Medical Centre+, Maastricht, The Netherlands, 8 Department of Head and Neck Surgical & Medical Oncology, Gustave Roussy Cancer Campus, Villejuif, France

Overview of abstract In recent years radiomics, defined as the quantification of imaging features, has been increasingly studied and often found to be associated with clinical endpoints. However, the relationship between radiomics and tumor biology are largely unknown. In this study, we sought to explore the immunological contextural basis of a previously developed prognostic radiomic signature in head and neck cancers. What were the three main findings of your research? We categorized patients into four phenotypes based on p16 expression and a 24-feature-based prognostic radiomic signature: p16+/radiolow(PL), p16+/radiohigh(PH) p16-/ radiolow (NL), p16-/radiohigh(NH). Overall survival (OS) among the four phenotypes was significantly different (5-year OS: PL 95.2%, PH 80%, NL 53.7%, NH 20.8%, p < 0.001). We found a significant correlation between the p16-radiomic phenotype and density of CD8+ T cells, FoxP3+ T cells and CD68+ cells, but not PD-L1 expression. Specifically, all patients with high density of all of CD8+ T cells, FoxP3+ T cells and CD68+ macrophages, (suggestive of a favorable immune activated state) exhibited as p16+/ radiolow phenotype. What impact could your research have? Our study demonstrates that radiomic approaches permit non-invasive assessment of immunological characteristic of head and neck tumors, although further validation in an external cohort is required. On one hand, linking radiomic signatures to immunological molecular context in the tumor micro-environment could allow us to understand the biology underlying the prognostic role of radiomics. On the other hand, these initial results imply that radiomics may provide complementary information to HPV/p16 status to further stratify patients and which could play a role in clinical decision-making, eg. the low-risk p16+/radiolow phenotype patients might be potential candidates to receive de-intensified treatment without compromising patient survival.

Is this research indicative of a bigger trend in oncology? Our results illustrate that the molecular characteristics of TME could be reflected by non-invasive radiomic approaches using standard-of-caremedical images. Emerging evidence showed that the immune system plays an important role in carcinogenesis and tumor progression. With more understanding of the biology underlying radiomic predictions, we hypothesize that it would aid in clinical decision-making at low additional cost, especially in the current era, given the increasing use of immunotherapy.

RADIOBIOLOGY | Congress report

44