Prospective analysis of parametric response map-derived MRI biomarkers: identification of early and distinct glioma response patterns not predicted by standard radiographic assessment

Abstract

Purpose: Currently, radiologic response of brain tumors is assessed according to the Macdonald criteria 10 weeks from the start of therapy. There exists a critical need to identify nonresponding patients early in the course of their therapy for consideration of alternative treatment strategies. Our study assessed the effectiveness of the parametric response map (PRM) imaging biomarker to provide for an earlier measure of patient survival prediction.

Experimental design: Forty-five high-grade glioma patients received concurrent chemoradiation. Quantitative MRI including apparent diffusion coefficient (ADC) and relative cerebral blood volume (rCBV) maps were acquired pretreatment and 3 weeks midtreatment on a prospective institutional-approved study. PRM, a voxel-by-voxel image analysis method, was evaluated as an early prognostic biomarker of overall survival. Clinical and conventional MR parameters were also evaluated.

Results: Multivariate analysis showed that PRM(ADC+) in combination with PRM(rCBV-) obtained at week 3 had a stronger correlation to 1-year and overall survival rates than any baseline clinical or treatment response imaging metric. The composite biomarker identified three distinct patient groups, nonresponders [median survival (MS) of 5.5 months, 95% CI: 4.4-6.6 months], partial responders (MS of 16 months, 95% CI: 8.6-23.4 months), and responders (MS has not yet been reached).

Conclusions: Inclusion of PRM(ADC+) and PRM(rCBV-) into a single imaging biomarker metric provided early identification of patients resistant to standard chemoradiation. In comparison to the current standard of assessment of response at 10 weeks (Macdonald criteria), the composite PRM biomarker potentially provides a useful opportunity for clinicians to identify patients who may benefit from alternative treatment strategies.

Figures

Figure 1

Representative (A,D) contrast-enhanced images and PRM overlays applied to (B,E) ADC and (C,F) rCBV from a (A-C) responding and (D-E) non-responding patients. The responding patient, diagnosed with a glioblastoma, had an overall survival of 17 months, PRMADC+ of 12% and PRMrCBV- of 4%. In contrast, the non-responding subject, also diagnosed with a glioblastoma, had an overall survival of 2 months, PRMADC+ of 1% and PRMrCBV- of 40%. PRM was acquired using pretreatment and 3 week post-treatment initiation imaging data.

Figure 2

Kaplan-Meier survival plots for overall survival, presented as a function of (a) PRMADC+, (b) PRMrCBV- and (c) PRMADC+/PRMrCBV- stratification at week 3 post-treatment initiation and (d) Radiological Response stratification at week 10 post-treatment initiation. The three discrete groups, as categorized by the composite PRM model, had median survivals for non-responders, intermediate and responders of 5.5 months (95% CI: 4.4-6.6 months n=14), 16 months (8.6-23.4 months n=19) and not reached (n=12), respectively.