Glioblastoma: a method for predicting response to antiangiogenic chemotherapy by using MR perfusion imaging--pilot study.

Abstract

To derive a magnetic resonance (MR)-based imaging metric that reflects local perfusion changes resulting from the administration of angiogenic-inhibiting chemotherapy in patients with recurrent glioblastoma multiforme (GBM). In this retrospective Institutional Review Board-approved HIPAA-compliant study, 16 patients (12 men, four women; mean age, 51.8 years + or - 15.1 [standard deviation]) with recurrent GBM received bevacizumab every 3 weeks (15 mg per kilogram of body weight) as part of a clinical trial. Baseline MR images were acquired, and follow-up images were acquired every 6 weeks thereafter until tumor progression or death. Imaging included perfusion and T1-weighted contrast material-enhanced MR imaging. Perfusion images were analyzed both with and without correction for contrast material leakage. The volumes of interest were selected as enhancing voxels on T1-weighted contrast-enhanced MR images. Relative cerebral blood volume (rCBV) maps were created from analysis of MR perfusion images. The volumes of interest were used to calculate the following parameters: size, mean rCBV, mean leakage coefficient K(2), and hyperperfusion volume (HPV), which is the fraction of the tumor with an rCBV higher than a predetermined threshold. Percent change in each parameter from baseline to first follow-up was compared with time to progression (TTP) by using a Cox proportional hazards model with calculation of hazard ratios. The most significant hazard ratio was seen with a DeltaHPV cutoff of rCBV greater than 1.00 (hazard ratio, 1.077; 95% confidence interval: 1.026, 1.130; P = .002). The only significant ratios greater than one were those that resulted from perfusion calculated as mean rCBV and DeltaHPV. The ratios were also higher after correction for leakage. This pilot study derived an imaging metric (HPV) that reflects local perfusion changes in GBMs. This metric was found to show a significantly improved correlation to TTP as compared with more commonly used metrics.