Diffusion MRI Phenotypes Predict Overall Survival Benefit from Anti-VEGF Monotherapy in Recurrent Glioblastoma: Converging Evidence from Phase II Trials

Abstract

Purpose: Anti-VEGF therapies remain controversial in the treatment of recurrent glioblastoma (GBM). In the current study, we demonstrate that recurrent GBM patients with a specific diffusion MR imaging signature have an overall survival (OS) advantage when treated with cediranib, bevacizumab, cabozantinib, or aflibercept monotherapy at first or second recurrence. These findings were validated using a separate trial comparing bevacizumab with lomustine.Experimental Design: Patients with recurrent GBM and diffusion MRI from the monotherapy arms of 5 separate phase II clinical trials were included: (i) cediranib (NCT00035656); (ii) bevacizumab (BRAIN Trial, AVF3708g; NCT00345163); (iii) cabozantinib (XL184-201; NCT00704288); (iv) aflibercept (VEGF Trap; NCT00369590); and (v) bevacizumab or lomustine (BELOB; NTR1929). Apparent diffusion coefficient (ADC) histogram analysis was performed prior to therapy to estimate "ADCL," the mean of the lower ADC distribution. Pretreatment ADCL, enhancing volume, and clinical variables were tested as independent prognostic factors for OS.Results: The coefficient of variance (COV) in double baseline ADCL measurements was 2.5% and did not significantly differ (P = 0.4537). An ADCL threshold of 1.24 μm2/ms produced the largest OS differences between patients (HR ∼ 0.5), and patients with an ADCL > 1.24 μm2/ms had close to double the OS in all anti-VEGF therapeutic scenarios tested. Training and validation data confirmed that baseline ADCL was an independent predictive biomarker for OS in anti-VEGF therapies, but not in lomustine, after accounting for age and baseline enhancing tumor volume.Conclusions: Pretreatment diffusion MRI is a predictive imaging biomarker for OS in patients with recurrent GBM treated with anti-VEGF monotherapy at first or second relapse. Clin Cancer Res; 23(19); 5745-56. ©2017 AACR.

Conflict of interest statement

Conflicts of Interest related to this Manuscript: Benjamin M. Ellingson, Albert Lai, Tracy Batchelor, and Timothy F. Cloughesy are paid consultants, members of the advisory board, and are research grant recipients from Roche/Genentech. Lauren E Abrey is an employee for Roche/Genentech. Dana T. Aftab, Gisela M. Schwab, and Colin Hessel are paid employees and stockholders for Exelixis. Tracy Batchelor is a consultant for Merck, Proximagen/Upsher, Oxigene, Cavion, and Accerta and has research support from Pfizer. Martin van den Bent has received research support from Roche. Marion Smits is a paid independent reviewer for Parexel.

©2017 American Association for Cancer Research.

Figures

Fig 1. Targets for inhibition of VEGF and VEGFR activity using cediranib, bevacizumab, cabozantinib, and aflibercept

Bevacizumab, a humanized monoclonal antibody for VEGF-A, acts within the extracellular domain to inhibit activity through direct inhibition of circulating extracellular VEGF-A, which reduces activity for both VEGFR-1 and VEGFR-2 receptors. Aflibercept, or “VEGF trap”, is a recombinant fusion protein that sequesters VEGF through use of VEGF-binding portions from extracellular domains of VEGFR-1 and VEGFR-2. Cediranib, a pan-tyrosine kinase inhibitor (TKI), acts in the intracellular domain to primarily inhibit VEGFR-1, VEGFR-2, and VEGFR-3 activity. Cabozantinib, also a TKI, acts within the intracellular domain to inhibit VEGF-R1, VEGFR-2, and VEGFR-3 activity.

Fig 2. Contrast enhanced T1-weighted subtraction maps and apparent diffusion coefficient (ADC) histogram analysis in a patient with recurrent GBM

A) Pre-treatment pre-contrast, post-contrast, and T1 subtraction maps in a patient with recurrent GBM. B) T1-subtraction defined tumor segmentation overlaid on ADC map. C) Resulting ADC histogram analysis results in the same patient. Note: Black filled circles indicate ADC measurements extracted from contrast enhancing tumor regions. Red line indicates double Gaussian mixed model fit to the underlying ADC histogram.

Fig 3. Repeatability of ADCL measurements, consistency of resulting ADCL phenotypes for different thresholds, and optimal thresholds for predicting overall survival (OS) in recurrent GBM treated with anti-VEGF therapies

A) Repeated diffusion MR measures of ADCL in recurrent GBM using “double baseline”, repeated pre-treatment examinations, from the cediranib trial. B) Repeatability of ADCL phenotypes (e.g. higher vs. lower ADCL) for different thresholds (solid black line, left y-axis) as well as the proportion of patients higher or lower than this threshold (gray lines, right y-axis). C) Mantel-Haenszel hazard ratios (HRs, solid black line) and 95% confidence intervals (gray area) for OS in recurrent GBM for different ADCL thresholds in patients pooled from all four anti-VEGF therapies. D) Level of significance (p-values) for OS differences for different ADCL thresholds in patients pooled from all four anti-VEGF therapies. E) Mantel-Haenszel HRs for OS for different ADCL thresholds for individual anti-VEGF therapies. F) Level of significance (p-values) for OS differences for ADCL thresholds in individual anti-VEGF therapies.

Fig 4. Univariate log-rank survival analysis applied to Kaplan-Meier (KM) curves obtained for pooled and individual anti-VEGF therapies for an optimal ADCL threshold of 1.24 um2/ms

A) KM data from all anti-VEGF therapies (P<0.0001; HR=0.5303; median OS = 7.7 vs. 11.6 months), B) cediranib (P=0.0489; HR=0.4980; median OS = 5.2 vs. 8.2 months), C) bevacizumab monotherapy (P=0.0050; HR=0.4545; median OS = 9.0 vs. 17.8 months), D) cabozantinib monotherapy (P=0.0107; HR=0.4623; median OS = 7.7 vs. 11.4 months), and E) aflibercept monotherapy (P=0.0017; HR=0.3313; median OS = 6.5 vs. 19.2 months). F) A combination of baseline tumor volume and ADCL phenotype was a strong predictor of OS in anti-VEGF therapies, with “high risk” patients having large volumes and low ADCL (Risk III, Median OS=4.7 months) demonstrating a significantly shorter OS (P<0.0001) compared with “low risk” patients exhibiting small tumors with high ADCL (Risk I, Median OS=14.9 months). G) Risk categorization in the BELOB validation cohort suggested a combination of baseline tumor volume and ADCL was a significant predictor of OS in bevacizumab monotherapy, with “high risk” patients with large volumes and low ADCL (Risk III, Median OS=5 months) demonstrating a significantly shorter OS (P<0.0001; HR=0.2093) compared with “low risk” patients exhibiting small tumors with high ADCL (Risk I, Median OS=10.4 months). H) The same combination biomarker stratified OS in lomustine monotherapy, but with a smaller difference in median OS between risk categories. “High risk” patients with large volumes and low ADCL (Risk III, Median OS=6.2 months) demonstrating a significantly shorter OS (P=0.0390; HR=0.4214) compared with “low risk” patients exhibiting small tumors with high ADCL (Risk I, Median OS=8.3 months). Multivariable Cox regression confirmed that tumor volume was a significant prognostic factor for OS in both treatment arms (P=0.0004 for bevacizumab and P=0.0019 for lomustine) but ADCL phenotypes were only predictive for bevacizumab therapy (P=0.0309 for bevacizumab and P=0.1455).