P12.53.B DEVELOPMENT AND VALIDATION OF A PRACTICAL PROGNOSTIC MODEL FOR GLIOBLASTOMA INTEGRATING ICAM1 EXPRESSION, RADIOLOGIC FEATURES, AND CLINICAL DATA

Abstract BACKGROUND Glioblastoma (GBM) is a highly aggressive brain tumour with limited prognostic tools that are broadly implementable in routine clinical settings. Current models often depend on molecular platforms not universally accessible. We aimed to develop and validate a clinically practical prognostic model for GBM that integrates ICAM1 protein expression, imaging-derived tumor morphometrics, and clinical features. MATERIAL AND METHODS We analyzed ICAM1 expression in GBM at the DNA methylation, RNA, and protein levels using public datasets and institutional cohorts, confirming its association with tumor grade and poor survival. ICAM1 protein levels were then quantified via digital pathology in a multi-institutional cohort of 197 newly diagnosed, treatment-naïve GBM patients. Tumor surface irregularity (SIR) was calculated from pre-operative contrast-enhanced MRI using open-source imaging tools. Age at diagnosis was also recorded. These three variables—ICAM1 expression, SIR, and age—were integrated into a Clinical-Imaging-Molecular (CIM) model using multivariable Cox regression. The model was validated in an independent external GBM cohort (n = 20). Survival prediction was assessed using Kaplan-Meier curves, ROC analysis, and concordance index (c-index). RESULTS ICAM1 expression was associated with DNA hypomethylation and significantly correlated with reduced survival at transcriptomic and proteomic levels. Pathway analysis identified associations with immune activation and angiogenesis. In the clinical cohort, high ICAM1 expression, elevated tumor SIR, and older age independently predicted worse survival. The CIM model combining these factors stratified patients into distinct prognostic groups, achieving strong performance (c-index = 0.779, AUC = 0.778). External validation confirmed the model’s predictive utility (median survival difference: 9.1 months, P = 0.009). CONCLUSION We present a validated and clinically applicable prognostic model for GBM that integrates ICAM1 expression, imaging features, and age. The CIM model is scalable, interpretable, and feasible for implementation across academic and community settings, providing a robust tool for patient stratification, treatment planning, and trial design.