S190
Clinical - Biomarkers of clinical response
ESTRO 2026
prospective trial, aimed to corroborate its potential in selecting patients for treatment personalization. The study was granted by AIRC (grant IG25951)
Purpose/Objective: The radiomic feature Statistical Percentile 10% (p10) extracted from [ ¹⁸ F]FDG-PET images after induction chemotherapy (pre-RT) was previously found to be a robust predictor of distant relapse-free survival (DRFS) in locally advanced pancreatic cancer (LAPC) and temporally and externally validated. The present study aimed to maximize the prognostic role of p10 when combined with CA19.9 (GICA) values taken at different times before Radiotherapy Material/Methods: A total of 287 LAPC patients (211 PET-positive and 76 PET-negative) were included. p10 was directly computed for PET-positive cases, while for PET- negative patients was arbitrarily set equal to 0. For 23 PET-positive patients, it was not possible to extract p10 due to different technical reasons, reducing the number of available patients to 264. GICA values after induction chemotherapy were available for 169/264 patients. Optimal cut-offs were determined using Youden’s index applied to Kaplan–Meier (KM) analyses. Cox models were performed both in univariate and bivariate configurations, combining p10 with GICA values taken before induction chemotherapy (GICA_preCHT), just before Radiotherapy (GICA_pre- RT), and at the nadir value (GICA_nadir) in continuous and dichotomized form. Models’ performances were evaluated through p-values, hazard ratios (HR), and C- index; KM results were summarized by comparing median distant progression-free survival (DRFS). Results: The optimal p10 cut-off of 2.71 yielded the strongest prognostic separation. In univariate Cox analysis, p10 > 2.71 showed HR=2.63 (95% CI 1.72–4.01;p< 0.0001; C-index = 0.55), confirmed by KM analysis (P < 0.0001) with median DRFS of 4.0 vs. 10.9 months for p10 ≥ or <2.71 respectively (HR=4.37). Among GICA variables, GICA_pre-RT>80 was the strongest predictor (HR=1.83; p=0.0016; C-index=0.58). The best bivariate model resulted from the combination of p10 and GICA_pre- RT (p< 0.0001; C-index=0.68) taken as continuous variables: the best predictive index cut-off value (PI>0.76) resulting from Cox analysis identified high- and low-risk groups with median DRFS equal to 4.3 vs. 15.0 months (p<0.0001; HR=3.57). Figures 1 and 2 showed the DRFS curves stratified according to the best cut-off values for p10 and the combined model
References: [1] Mori M, Passoni P, Incerti E, et al. Training and validation of a robust PET radiomic-based index to predict distant-relapse-free-survival after ra-dio- chemotherapy for locally advanced pancreatic cancer. Radiother Oncol. 2020; 153:258-264. doi: 10.1016/j.radonc.2020.07.003.[2] Vincenzi MM, Mori M, Passoni P, et al. Temporal Validation of an FDG-PET- Radiomic Model for Distant-Relapse-Free-Survival After Radio-Chemotherapy for Pancreatic Adenocarcinoma. Cancers. 2025; 17(6):1036. doi: 10.3390/cancers17061036 .[3] Vincenzi MM, Tummineri R, Boldrini L et al. External validation of an FDG-PET-Radiomic model for disstant-relapse-free survival after radio-chemotherapy for pancreatic cancer (Submitted) Keywords: pancreatic cancer, PET biomarkers, radiomics Poster Discussion 2807 Role of salivary S100A7 as prognostic biomarker in Oral Squamous Cell Carcinoma (OSCC) Manju Parkavi 1 , Smriti Suri 1 , Arnab Pal 1 , Sushmita Ghoshal Chakrabarti 2
respectively. Conclusion:
p10 confirmed to be a robust, clinically interpretable, predictor of DRFS in LAPC. The combination of p10 with GICA_pre_RT largely improves the prediction performances, translating into a large separation between low/high-risk groups, once stratifying patients according to the optimal cut-off. Current biomarker is currently under investigation into a
Made with FlippingBook - Share PDF online