ESTRO 2021 Abstract Book

S717

ESTRO 2021

and Quality of Life in Oncology Unit, Besançon, France; 5 Curie Institute, Radiation Oncology, Paris, France; 6 Fondazione Policlinico Universitario Agostino Gemelli IRCCS, Dipartimento di Diagnostica per Immagini, Radioterapia Oncologica ed Ematologia, Rome, Italy; 7 Università degli Studi di Brescia, Dipartimento di Scienze Cliniche e Sperimentali, Brescia, Italy; 8 Neuchâtel Hospital Network, Radiation Oncology, La Chaux- de-Fonds, Switzerland Purpose or Objective Almost 20% of pts with a locally advanced esophageal cancer (LA-EC) receiving exclusive RT-CT would present long-term LC, and could potentially avoid surgery. Aim of this study was to evaluate the predictive value of pre-treatment Positron Emission Tomography–Computed Tomography (PET-CT) radiomic features on locoregional progression free survival after exclusive chemoradiotherapy (CT-RT) in LA-EC pts. Materials and Methods 46 consecutive pts presenting a pathologically confirmed LA-EC were included. All pts underwent PET-CT prior to CT-RT. They received RT 50Gy (2Gy/fr.) with concomitant CT (principally with Carboplatin AUC 2 and Paclitaxel 50 mg/m2, n=28, or FOLFOX, n = 14 [JH1] ). [JH2] A texture analysis was performed for each pt defined by a tumor volume of interest (VOI) on the PET-CT metabolic images before CT-RT. 230 radiomic features were extracted from the VOIs and those correlated with each other at less than 0.9 were introduced in a unsupervised hierarchical cluster analysis to identify the groups of pts expressing similar radiomics features and a heatmap was perfomed (Fig. 1). A Principal component analysis (PCA) was used to select major radiomics features associated with the 2 groups. Primary endpoint was locoregional progression-free survival (LR-PFS), secondary endpoints were overall survival (OS), and progression-free survival (PFS). The free open source MODDICOM library was used for radiomics analyses. Survival was estimated using the Kaplan-Meier method and compared using the log-rank test.

Results Data about response to CT-RT were available for 40 patients. Complete response was obtained in 50% of cases. After a median follow-up time of 17 months, 2-years LR-PFS rate was 35.9% (95%CI 18.9-53.3), and 2-years PFS rate was 21.6% (95%CI 10.0-36.2). Median OS time was 29.4 months (95%CI 21.3-NA). 55 features were retained after correlation analysis. According to the results of the heat map, we have identified 2 groups of patients. Two main radiomic characteristics ("F_rlm_rl_entr"and "F_rlm_2_5D_rl_entr") contributed most to this segregation according to PCA . Median LR-PFS rate was higher in Group 1 than in Group 2 (22.8 vs 9.9 months; HR = 2.64; 95%CI 0.97-7.15; p=0.0573) (Fig. 2). By defining the expression cutoff at the 28 quantile ("F_rlm_2_5D_rl_entr" = 3.3; "F_rlm_rl_entr” = 4.7), those parameters were significantly associated with LR-PFS: patients who expressed “F_rlm_2_5D_rl_entr” <3.3 had a better median LRPFS (29.4 vs 8.2 months; p=0.0343) and patients who expressed “F_rlm_rl_entr_per” < 4,7 had a better median LR-PFS (50.4 vs 9.9 months; p=0.0132). Median OS and PFS rate were higher in Group 1 than in Group 2, but those results were not statistically