P17.39.A IMPACT OF FRAILTY ON SURGICAL OUTCOMES IN PATIENTS AFFECTED BY GLIOBLASTOMA: A MULTICENTER ANALYSIS WITH MULTIVARIATE AND MACHINE LEARNING APPROACHES.

Abstract

Abstract BACKGROUND Frailty, a multidimensional syndrome characterized by reduced physiological reserves and increased vulnerability to stressors, has emerged as a critical determinant in treatment decision-making and is considered a determining factor in surgical eligibility for patients suffering of glioblastoma, frequently leading to the exclusion of frail individuals from potentially beneficial surgical interventions. Additionally, frailty is increasingly recognized as an important issue in the management of patients affected by glioblastoma (GB), potentially influencing choices and affecting evaluations regarding surgical treatment. However, the extent to which frailty independently affects survival outcomes remains unclear. This study investigates the relationship between frailty and prognosis in patients suffering of glioblastoma undergoing surgical resection. MATERIAL AND METHODS A retrospective multicenter cohort study was conducted, including 894 patients with IDH-wildtype Glioblastoma who underwent surgery across eight Italian hospitals. Patients were classified as non-frail (n=763) or moderate-to-severely frail (n=131) based on validated frailty indices. Survival analyses were performed using Kaplan-Meier estimates, univariate and multivariate Cox regression models, and Machine Learning (ML) algorithms. Principal Component Analysis (PCA) was applied to explore heterogeneity within the frail patient group. RESULTS Kaplan-Meier analysis showed a lower survival probability among frail patients (p=0.05). Univariate Cox regression confirmed an association between frailty and reduced survival (p<0.001). However, after adjusting for age, MGMT methylation status, postoperative Karnofsky Performance Status (KPS), and treatment protocols, frailty was not an independent predictor of survival in multivariate analysis (p=0.98). ML models corroborated this finding, showing no improvement in 12-month survival prediction when frailty status was included (AUC=0.77). PCA identified two subgroups within the frail population with significantly different survival trajectories (log-rank p<0.001), while no meaningful clustering was observed among non-frail patients. CONCLUSION Frail patients with glioblastoma are often deemed ineligible for surgery due to presumed higher perioperative risks and limited survival benefits. In this study we demonstrated that frailty is associated with poorer survival outcomes in univariate analyses but does not independently influence prognosis when other clinical variables are considered. Identifying subgroups among frail patients with differing survival outcomes may refine surgical decision-making and supports a more individualized approach to the management of patients affected by glioblastoma.

Ähnliche Arbeiten