Abstract
BACKGROUND: Glioblastoma (GB) is the most aggressive brain tumor, characterized by rapid progression and poor prognosis. Despite initial multimodal treatment options, therapeutic options become more limited upon recurrence. Consequently, recurrent IDH-wildtype GB is associated with poor survival outcomes, with limited data to guide optimal therapeutic strategies. This study presents the largest meta-analysis to date, pooling institutional data with individual patient data (IPD), addressing progression-free survival (PFS) and overall survival (OS) from timepoint after re-resection. METHODS: Institutional data and data from literature (2016-2024) were analyzed to evaluate PFS and OS in relation to the extent of resection (EoR). Survival data from identified studies were extracted from Kaplan-Meier curves with the R package IPDfromKM. Additionally, a retrospective analysis of institutional data was conducted, assessing for PFS and OS in 53 patients. EoR was dichotomized as suggested by the RANO group into residual contrast-enhancing tumor volume (CE-RTV) ≤ 1 cm(3) or > 1cm(3). RESULTS: A total of 442 IPD were included in this meta-analysis. Among them, 331 patients (74.9%) underwent neurosurgical treatment with CE-RTV ≤ 1cm(3), while 111 patients (25.1%) had CE-RTV > 1 cm(3). Pooled analysis indicated a significant reduction in OS after re-resection with CE-RTV > 1 cm(3) compared to CE-RTV ≤ 1cm(3) (HR: 1.731, 95% confidence interval (CI): 1.342-2.234, p < 0.0001). While re-resection with CE-RTV ≤ 1cm(3) was associated with longer OS (14.4 months) compared to CE-RTV > 1 cm(3) (8.8 months) (p < 0.0001), PFS did not differ between the two groups (CE-RTV ≤ 1 cm(3): 7.2 months compared to CE-RTV > 1cm(3): 5.8 months) (p = 0.76). CONCLUSION: Across pooled IPD, maximal safe resection at re-resection operationalized as GTR or RANO class 1 and 2 was significantly associated with longer overall survival (OS). Where volumetric assessment is available, achieving a postoperative CE-RTV ≤ 1 cm(3) may be a reasonable pragmatic target, however this threshold was not directly measured in all included cohorts and should be interpreted as hypothesis-generating.