The role of early magnetic resonance imaging in predicting survival on bevacizumab for recurrent glioblastoma: Results from a prospective clinical trial (CABARET)

Authors:
Publisher website
Google Scholar
Abstract
BACKGROUND: Bevacizumab has been associated with prolonged progression‐free survival for patients with recurrent glioblastoma; however, not all derive a benefit. An early indicator of efficacy or futility may allow early discontinuation for nonresponders. This study prospectively assessed the role of early magnetic resonance imaging (eMRI) and its correlation with subsequent routine magnetic resonance imaging (MRI) results and survival.METHODS: Patients were part of a randomized phase 2 clinical trial (CABARET) comparing bevacizumab with bevacizumab plus carboplatin for recurrent glioblastoma. eMRI was conducted after 4 weeks in the trial (after 2 treatments with bevacizumab [10 mg/kg every 2 weeks]). The results were compared with the results of the subsequent 8‐week MRI standard.RESULTS: For 119 of 122 patients, eMRI was available, and 111 had subsequent MRI for comparison. Thirty‐six (30%) had an early radiological response, and 17 (14%) had progressive disease. The concordance between eMRI and 8‐week MRI was moderate (κ = 0.56), with most providing the same result (n = 79 [71%]). There was strong evidence that progression‐free survival and overall survival were predicted by the eMRI response (both P values < .001). The median survival was 8.6 months for an eMRI response, 6.6 months for stable disease, and 3.7 months for progressive disease; the hazard ratio (progressive disease vs stable disease) was 3.4 (95% confidence interval, 1.9‐6.0). Landmark analyses showed that eMRI progression was a strong predictor of mortality independent of other potential baseline predictors.CONCLUSIONS: In this study, early progression on MRI appears to be a robust marker of a poor prognosis for patients on bevacizumab. Cancer 2017;123:3576‐82. © 2017 American Cancer Society.
Funding Information
  • National Health and Medical Research Council (1037786)

This publication has 27 references indexed in Scilit: