Dynamic susceptibility contrast MRI measures of relative cerebral blood volume continue to show promise as an early response marker in the setting of bevacizumab treatment

We read with interest Whitney Pope's recent editorial¹ in Neuro-Oncology on the paper by Kickingereder et al² addressing the utility of relative cerebral blood volume (rCBV) derived from susceptibility-contrast perfusion-weighted MRI for determining therapeutic response in bevacizumab-treated recurrent glioblastoma multiforme (GBM). Kickingereder and Pope contend that although pretreatment rCBV may be a predictive marker of overall survival (OS), there is no evidence supporting post-bevacizumab change in rCBV from baseline as an early response marker. We caution against premature dismissal of rCBV change as an early predictor of treatment response.

Not discussed in Pope's editorial was a recent publication by Schmainda et al³ demonstrating, in a multicenter trial organized by the American College of Radiology Imaging Network (ACRIN), that bevacizumab-treated recurrent GBM patients with positive versus negative changes in tumor rCBV from pretreatment baseline had highly significant differences in OS. Although these results appear contradictory to those of Kickingereder as well as another recent paper by Schmainda et al,⁴ they are actually consistent and potentially offer new insights regarding treatment response and the optimal timing of imaging. Specifically, rCBV changes measured 2 and 16 weeks, but not 8 weeks, after bevacizumab initiation in the ACRIN study predicted OS. Similarly, Kickingereder² (Table 2) and Schmainda⁴ (Fig. 5b) found that 8-week changes in rCBV were nonpredictive of OS. Therefore, these studies are not inconsistent, and the conclusion that posttherapy rCBV changes are generally nonpredictive of OS appears unfounded. It may all depend on the time at which posttreatment imaging is performed. It remains unclear why 2- and 16-week measurements may be more meaningful than 8-week measurements, but the pathophysiology of GBM is complex and dynamic, and the prognostication of rCBV measures may vary at different time points.

We analyzed the pretherapy results from the multicenter ACRIN trial and found that contrary to the single-center data from Kickingereder² and Schmainda,⁴ baseline rCBV did not predict OS-1 (AUC = 0.55–0.62 for all 21 patients with perfusion MRI). However, intersite perfusion MRI methodologies (eg, MRI equipment manufacturer, field strength, spin- or gradient-echo technique, echo time) were more variable than for single-center studies. This result suggests that change in rCBV may be more consistent than absolute rCBV in the multicenter setting. It is quite possible that if consistency between sites is improved, baseline rCBV may someday be shown to predict OS in a multicenter trial; but at present, change from baseline may be a more robust measurement in this setting. That being said, we agree with Pope's contention that lack of standardization hampers the application of perfusion imaging but caution against viewing an arterial input function region of interest for normalization of rCBV as a panacea. Such normalization will vary by hematocrit and selection method,⁵ with recent evidence that rCBV repeatability is worse when an arterial input function is used for its computation.⁶ Transformation of rCBV to a standardized scale⁷ provides a viable alternative and has been successfully used in a multicenter setting.³

Kickingereder and Pope also postulated that pretreatment rCBV measures alone are sufficient to predict response to bevacizumab. Schmainda⁴ found that although low baseline rCBV predicted longer OS, subsequent change in posttreatment rCBV further substratified patients. Patients with high baseline rCBV may have shorter OS than patients with low rCBV, but may relatively benefit from bevacizumab. The ACRIN 2-week results suggest that rCBV change identifies those patients for whom this is true, providing a modulation of the inherent OS predicted by baseline rCBV. In fact, after adjusting for baseline rCBV in the ACRIN data, rCBV change at week 2 still correlates significantly with OS. A recent sub-analysis of the ACRIN perfusion data presented at the 2014 meeting of the Radiology Society of North America demonstrated that patients not progressing on conventional postcontrast T1-weighted imaging could be substratified into relatively short and long OS based on rCBV change from baseline at 2 weeks posttreatment initiation.⁸ It is therefore premature to discount the added value of posttreatment change in rCBV, and at the very least a larger multicenter trial exploring both baseline rCBV and 2-week posttreatment change in rCBV is merited and may be practice changing if the additional time point is found to be beneficial.

We are grateful for the opportunity to address important points raised by Kickingereder and Pope regarding rCBV measures for the evaluation of bevacizumab response in recurrent GBM. Our combined multiyear experience with perfusion MRI and recent involvement with the ACRIN clinical trials have further strengthened our belief that rCBV has promise as an important early response marker in the setting of bevacizumab treatment, and we are hopeful that the additional evidence provided herein will encourage support for its continued evaluation.

Funding

This study was supported by NIH/NCI (grant no. R01 CA082500 and U01 176110 to K.M.S).

Conflict of interest statement. Kathleen M. Schmainda has ownership interest in Imaging Biometrics LLC.