BACKGROUND: Bevacizumab has proven beneficial in patients with recurrent glioblastoma (rGB), however its impact on tumor blood flow and oxygenation status remains controversial. Here we examine dynamic susceptibility-weighted contrast-enhanced MRI (DSC-MRI) derived Gaussian-normalized, relative cerebral blood volume and flow (nrCBV, nrCBF) parametric-response-maps (PRMs), for predicting response to bevacizumab in patients with rGB. METHODS: A total of 71 patients diagnosed with rGB underwent conventional-anatomic MRI and DSC-MRI at baseline and at first follow-up after bevacizumab initiation. PRMs were created by a multi-step (non-linear) registration of the patient's post- to pre-treatment scan and voxel-wise subtraction between nrCBV and nrCBF maps. Intratumoral voxels were stratified as increased [PRM(+)] or decreased [PRM(-)] if they exceeded a threshold representing the 95% confidence interval in the normal-appearing brain. Correlation with PFS and OS was performed using Cox-proportional-hazard models. RESULTS: The hazards for progression and death significantly increased with (I) higher baseline nrCBV (HR = 1.86, p < 0.01; HR = 1.52, p < 0.01) and nrCBF values (HR = 1.78, p < 0.01; HR = 1.86, p < 0.01), (II) higher PRM(-) of nrCBV (HR = 1.03, p = 0.01; HR = 1.02, p = 0.03) and nrCBF (HR = 1.04, p < 0.01; HR = 1.03, p < 0.01), but (III) not for higher PRM(+) of nrCBV and nrCBF, and (IV) not for the relative change in mean nrCBV and nrCBF, confirming the superiority of the PRM-approach. The magnitude of PRM(-) for both nrCBV and nrCBF significantly increased for higher baseline values (p < 0.01, respectively). CONCLUSION: Pre-treatment hemodynamic parameters are the principal determinant of BEV-response in rGB and emerge as a potential predictive imaging biomarker for selecting those patients with rGB who may maximally benefit from BEV-treatment. Although the magnitude of PRM(-) is a function of the corresponding pre-treatment parameter, the finding of higher PRM(-) and lack of PRM(+) in BEV non-responders may correspond to a pro-invasive, non-angiogenic evasion.