S2482
Physics - Radiomics, functional and biological imaging, and outcome prediction
ESTRO 2026
specific radiosensitivity, 127 single-nucleotide polymorphisms (SNPs) previously associated with late side-effects were evaluated for association with fibrosis using multivariable Cox with interval-censoring correction to account for gaps in follow-ups. SNPs for inclusion in the polygenic risk score (PRS) were selected until incremental improvement in discrimination (C-index) was <2%, indicating performance saturation. We defined PRS through standardisation of Cox-multivariable prediction and NT_GARD from a 2-variable Cox-model (PRS,EQD2MaxD): NT_GARD=EQD2MaxD+[( β _PRS/ β _EQD2MaxD) × PRS], where [( β _PRS/ β _EQD2MaxD) × PRS] represents the individualised genetic contribution that increases/decreases the physical dose exposure [Avanzo_2012]. We used stratified Kaplan–Meier analyses to evaluate the association of NT_GARD with fibrosis. Results: 231/1881 patients presented with late grade ≥ 2 fibrosis.Five SNPs were identified as associated with fibrosis and used to construct the PRS (rs3213235, rs10517190, rs1503967, rs6839224, rs639923).NT_GARD distribution was broader compared to EQD2MaxD (Figure_1a), highlighting that some sensitive patients experience a biologically- relevant dose much greater than the physical dose.For Kaplan–Meier analyses, we stratified NT_GARD and EQD2MaxD at 10Gy intervals. NT_GARD stratification captures a variation in fibrosis risk with significant Hazard-Ratios of 1.5, 1.3 and 1.4 for increasing NT_GARDs, <50Gy, 50-60Gy, 60-70Gy and >70Gy, respectively (Figure_2a).EQD2MaxD stratification only captures presence of a boost dose, with significant Hazard-Ratio=1.9 for <50Gy vs >50Gy, and no difference between 50-60Gy and >60Gy (Figure_2b).
Conclusion: By explicitly incorporating genetic susceptibility into dose modeling, NT_GARD provides a biologically grounded, patient-specific measure of effective radiation dose. Differences observed in the Kaplan– Meier curves suggest that EQD2MaxD primarily distinguishes between boost and no-boost, and substantial increases in dose raise risks, but fail to explain variability within the boost/no-boost groups.NT_GARD will serve as the exposure variable in causal inference frameworks to advance personalised radiotherapy (example Figure_1b).Funding: EU-FP7 grant 601826 (REQUITE Project). Horizon-EURATOM grant 101166699 (TETRIS). References: [Mukesh_2013] Radiother Oncol. 2013; 108(2):293; Normal tissue complication probability (NTCP) parameters for breast fibrosis: pooled results from two randomised trials[Avanzo_2012] Phys Med. 2012; 28(4):296; Complication probability model for subcutaneous fibrosis based on published data of partial and whole breast irradiation Keywords: Genetic risk, side effects, Breast Cancer Validation of Apparent Diffusion Coefficient as a Quantitative Imaging Biomarker for Pancreatic Cancer on a 1.5T MR-Linac Elame Lionel Rickiel 1 , Madeline Michel 2 , Christelle Bouchart 3,4 , Zelda Paquier 1 1 Medical Physics, Institut Jules Bordet, Hopital Universitaire de Bruxelles, Université Libre de Bruxelles, Brussels, Belgium. 2 Radiation oncology, Institut Jules Bordet, Hopital Universitaire de Bruxelles, Université Libre de Bruxelles, Brussels, Belgium. 3 Radiation oncology, Institut Jules Bordet, brussels, Belgium. 4 Laboratory of Experimental Gastroenterology, Université Libre de Bruxelles, Brussels, Belgium Digital Poster 4061 Purpose/Objective: The implementation of the apparent diffusion coefficient (ADC) as a biomarker on MR-Linac systems remains insufficiently validated for pancreatic cancer
Made with FlippingBook - Share PDF online