P11.09.B PREDICTIVE MODEL FOR RADIATION NECROSIS AFTER STEREOTACTIC RADIOSURGERY FOR BRAIN METASTASES BASED ON THE V12 PARAMETER AND INFLAMMATION-RELATED PLASMA BIOMARKERS

Abstract BACKGROUND Cerebral radiation necrosis (RN) is a deleterious adverse event following stereotactic radiosurgery (SRS) for brain metastases (BMs). Although the volume of brain tissue receiving 12 Gy (V12) is a recognised predictor of RN, its predictive performance remains suboptimal. Prior authors demonstrated that radiation-induced brain injury can be traced in peripheral blood. We sought to identify predictive biomarkers of symptomatic RN in liquid biopsies from BMs patients at the time of SRS and to develop a risk prediction model integrating both V12 and biomarker data. MATERIAL AND METHODS A case-control study was designed using plasma samples from BMs patients treated with SRS. Olink 96 Target Immuno-Oncology Panel was performed to analyse 92 immune-related human proteins using oligonucleotide-bound antibodies and real-time polymerase-chain reaction. Statistical analysis was conducted using GraphPad Prism 9.0. Receiver-operating characteristic (ROC) curves were performed for selected biomarkers. Multivariate Cox proportional hazard (PH) regression was conducted to assess predictor variables related to patients, SRS treatment and baseline plasma biomarkers, and the subsequent development of RN. Multiplex immunophenotyping in RN tissue was performed in available cases to validate the presence of a pro-inflammatory microenvironment. RESULTS A total of 47 patients were included, comprising 21 individuals who developed RN post-SRS and 26 matched controls with successful ablation of BMs without RN. Plasma analysis at the time of SRS revealed significantly elevated CD8a levels and significantly decreased MUC-16 and IL-8 levels (p<0.05) in patients who later developed RN. Multivariate Cox PH analysis identified V12 (p<0.01), IL-8 (p<0.01), and CXCL11 (p<0.001) as independent predictors of RN. Development of a necrosis prognostic index (NPI) combining V12, MUC-16, CD8a, CXCL11, and IL-8 substantially improved predictive performance, with the ROC area under the curve (AUC) increasing from 0.77 (V12 alone) to 0.91 (p=0.097). Histopathological analysis confirmed enhanced inflammatory infiltrates (CD4+, CD8+, GFAP+STAT3+, and CD68+iNOS+ cells) in RN specimens, supporting an enhanced inflammatory response. CONCLUSION The predictive accuracy of the V12 dosimetric parameter for RN can be significantly augmented by integrating plasma-derived inflammatory biomarkers measured at the time of SRS. Prospective validation of this model may enable early identification of BMs patients at higher risk of radiation-induced brain toxicity, facilitating the development of personalised SRS treatment strategies to mitigate RN incidence.