Post-Recurrence Outcomes Associated with a Bevacizumab-Containing First-Recurrence Strategy in Glioblastoma: A Propensity-Weighted Single-Center Cohort Study with Competing-Risk Analysis

Background/Objectives: Bevacizumab is widely used at recurrence in glioblastoma, but comparisons are complicated by multimodal treatment selection and progression assessment. We examined outcomes associated with a bevacizumab-containing first-recurrence strategy and explored the components of post-recurrence progression-free survival (prPFS). Methods: This retrospective single-center cohort included 166 patients at first recurrence: 114 received a bevacizumab-containing strategy and 52 a non-bevacizumab strategy. Of these, 163 had evaluable outcomes. Local therapy at first recurrence was recorded in 19.3% and 90.4% of the groups, respectively. Treatment strategy was fixed at first recurrence to reduce immortal-time bias. Analyses used pretreatment covariates, multiple imputation, propensity-score overlap weighting, robust variance estimation, and competing-risk methods. Complete-case adjusted models included 117 patients. Recurrence-specific performance status, corticosteroid exposure, and MGMT status were unavailable. Results: Among 163 patients, 136 deaths and 145 prPFS events occurred. In the overlap-weighted analysis after multiple imputation, a bevacizumab-containing strategy was not associated with improved post-recurrence overall survival (OS; HR 1.40, 95% CI 0.93 to 2.10) or prPFS (HR 0.81, 95% CI 0.55 to 1.18). Competing-risk analysis showed a lower cause-specific hazard of documented second progression (HR 0.29, 95% CI 0.16 to 0.53) and a higher cause-specific hazard of death before documented progression, although its CI included the null (HR 1.65, 95% CI 0.93 to 2.93). Higher baseline neutrophil-to-lymphocyte ratio was associated with poorer OS (HR per doubling 1.49, 95% CI 1.20 to 1.85), but treatment interactions were inconsistent across parameterizations. Conclusions: A bevacizumab-containing first-recurrence strategy was not associated with improved post-recurrence OS or prPFS. Divergent progression and death patterns caution against interpreting progression-based endpoints as direct evidence of disease control in this observational setting.