In glioblastoma, tumor progression appears to be triggered by expression of VEGF, a regulator of blood vessel permeability. Bevacizumab is a monoclonal antibody that inhibits angiogenesis by clearing circulating VEGF, resulting in a decline in the contrast-enhancing tumor, which does not always correlate with treatment response. Our objectives were: (1) to evaluate whether changes in DSC perfusion MRI-derived leakage could predict survival in recurrent glioblastoma, and (2) to estimate whether leakage at baseline was related to treatment outcome. We retrospectively analyzed DSC perfusion MRI in 24 recurrent glioblastomas treated with bevacizumab as second line chemotherapy. Leakage at baseline and changes in maximum leakage between baseline and the first follow-up after treatment were selected for quantitative analysis. Survival univariate analysis was made constructing survival curves using Kaplan-Meier method and comparing subgroups by log rank probability test. Leakage reduction at 8weeks after initiation of bevacizumab treatment had a significant influence on overall survival (OS) and progression-free survival (PFS). Median OS and PFS were 2.4 and 2.8months longer for patients with leakage reduction at the first follow-up. Higher leakage at baseline was associated with leakage reduction after treatment. Odds ratio of treatment response was 9 for patients with maximum leakage at baseline>5. Leakage decrease may predict OS and PFS in recurrent glioblastomas treated with bevacizumab. Leakage reduction postulates as a potential biomarker for treatment response evaluation. Leakage at baseline seems to predict response to treatment, but was not independently associated with survival.