Baseline multicentric tumors, distant recurrences and leptomeningeal dissemination predict poor survival in patients with recurrent glioblastomas receiving bevacizumab.

Abstract

There are no widely accepted MRI markers that predict treatment outcomes of bevacizumab among patients with recurrent glioblastoma (GB). We aimed to determine if conventional MRI features of recurrent GB predict survival of patients receiving bevacizumab. Patients with recurrent GB were retrospectively included if they received bevacizumab monotherapy between 2008 and 2017 after failure of standard treatment. Their MRI studies obtained at baseline and tumor recurrence, prior to bevacizumab treatment, were evaluated for multiple MRI features including measurable tumor, baseline multicentric tumors, distant recurrence, non-contrast-enhancing tumor, deep white matter invasion, multiple parenchymal tumors, bilateral cerebral involvement, ependymal extension and leptomeningeal dissemination. Predictive values of MRI features and patient characteristics on patient survival were statistically analyzed. A total of 103 patients were included. Baseline multicentric tumors (OR = 4.07; P = 0.042) and distant recurrence (OR = 28.5; P < 0.001) were two significant predictors of 3-month progression-free survival (PFS) rate. Distant recurrence (HR = 3.94; P < 0.001) was the only independent predictor of PFS. Baseline multicentric tumors (HR = 1.97; P = 0.028), distant recurrence (HR = 4.73; P < 0.001) and leptomeningeal dissemination (HR = 2.28; P = 0.044) were three independent predictors of overall survival. Baseline multicentric tumors, distant recurrence and leptomeningeal dissemination predicted poor survival among patients receiving bevacizumab for recurrent GB. Conventional MRI may help selecting patients with recurrent GB for bevacizumab treatment.