Genomic determinants of radiation necrosis in diffuse glioma patients.

2053 Background: Diffuse gliomas are the most common primary brain tumors, requiring multimodal therapy. Radiotherapy (RT) represents a key pillar of treatment; however, radiation-induced brain toxicities remain a major challenge. Radiation necrosis (RN) is a delayed complication that can impair clinical outcomes and quality of life. Emerging evidence suggests that tumor biology, together with clinical factors, can influence radiation toxicity in a variety of cancer entities. However, in diffuse gliomas, tumor-intrinsic biomarkers predisposing to radiation toxicities remain poorly defined. We hypothesized that integrating clinical features with tumor genomic data could identify patients at increased risk of developing RN. Methods: We analyzed a cohort of 943 adult patients with diffuse gliomas treated between 2014 and 2024 at MD Anderson. All tumors were classified according to the WHO 2021 criteria into three diagnoses: GBM, astrocytoma (astro), and oligodendroglioma (OD). RN occurring at least 12 weeks after completion of RT was identified based on pathology and/or advanced brain tumor imaging (ABTI), which included conventional MRI, perfusion imaging and spectroscopy. A multivariate regression model was employed to identify clinical or genetic predictors for RN. Included covariates were demographics, diagnosis, and common genetic alterations. To address missing values, multiple imputation was utilized. Results: Sixty patients with RN were identified: 51 GBM (85%), 6 astro (10%), and, 3 OD (5%); compared with 883 controls: 682 GBM (77%), 140 astro (16%), and 61 OD (7%). Age, gender, and diagnoses were not associated with RN. We evaluated the top 22 altered genes and MGMT status. In the overall cohort, IDH1 (OR 0.15 [95% CI, 0.02–0.93]; p=0.042) and MDM2 (OR 0.11 [95% CI, 0.01–1.00]; p=0.050) alterations were associated with reduced RN risk, while ATR alterations were associated with increased RN risk (OR 6.05 [95% CI, 0.85–33.1]; p=0.049) on multivariable analysis. In the GBM subgroup, CDK4 (OR 3.32 [95% CI, 1.04–9.97]; p=0.042) and MSH6 alterations (OR 6.05 [95% CI, 1.12–33.1]; p=0.037) were associated with increased RN risk, whereas MDM2 alterations remained associated with reduced risk (OR 0.07 [95% CI, 0.01–0.69]; p=0.023). In the IDH -mutant diffuse glioma subgroup (RN n=9; controls n=201), ATRX alterations were associated with lower RN risk (OR 0.10 [95% CI, 0.01–0.73]; p=0.023). MGMT status was not associated with RN risk. Conclusions: In this cohort, RN risk was not associated with demographic factors, tumor diagnosis, or MGMT status, but was associated with specific tumor genomic alterations. The enrichment of DNA damage response–related genes among those associated with RN suggests a potential biological link between tumor genomics and RN toxicity. Prospective validation will be required to determine the utility of these findings for genome-based risk stratification.