Abstract
OBJECTIVES: This study aimed to assess the prognostic significance of CD40LG and a related radiomics model in high-grade gliomas. METHODS: This retrospective cohort study utilized data from TCGA (n = 298) and TCIA (n = 89) following STROBE guidelines. From The Cancer Genome Atlas (TCGA), HGGs with genomic and clinical data were analyzed to establish CD40LG's prognostic value through Kaplan-Meier survival analysis and multivariate Cox regression. A radiomic model, based on TCGA data and matched MRI T1 images from The Cancer Imaging Archive (TCIA), was built to predict CD40LG levels. Radiomic features were extracted via PyRadiomics, filtered by 1000-repeat LASSO regression, and validated through 5-fold cross-validation. An independent cohort (n = 182) tested the model's prognostic utility. Subsequently, a prognostic model and nomogram were developed. RESULTS: Kaplan-Meier curves indicated a significant association between CD40LG expression and overall survival. CD40LG emerged as a crucial risk factor in both univariate and multivariate analyses. Immune cell infiltration analyses highlighted CD40LG's connection to the tumor immune microenvironment. A radiomic model, constructed using LASSO regression and five features, successfully predicted CD40LG expression pre-surgery. Combining the model's Rad-scores with clinical data, we created an effective prognostic model. CONCLUSIONS: CD40LG expression correlates with high-grade glioma prognosis. Our MRI-based radiomic signature predicted CD40LG expression and prognosis, offering potential guidance for treatment decisions and future research.