Marginal diffusion slope as a prognostic imaging biomarker of infiltrating phenotype in glioblastoma; A cancer imaging biomarker roadmap study

Abstract Background Conventional MRI protocols fail to probe marginal tumour infiltration in glioblastoma, hindering surgery and radiotherapy planning. This study aimed to demonstrate development and biological validation of a putative imaging biomarker (IB) for characterising glioblastoma infiltration, following principles outlined in the cancer imaging biomarker roadmap. Methods This IB is based upon spatial change in apparent diffusion coefficient (ADC) measures across a macroscopic tumour boundary. We systematically assessed whether ADC slope at the tumour margin (marginal diffusion slope-MDS) could (a) describe an underlying infiltrative phenotype based on reported links between ADC and tumour cellularity validated using a preclinical model, and (b) predict clinical outcome in a single-centre exemplar prospective human cohort study. Results Preclinical results showed a strong, spatially-resolved, negative correlation between marginal ADC and underlying tumour cell density in coregistered MRI-histology datasets from a glioblastoma model. Clinical results (n = 18) showed a positive linear correlation between MDS and clinical outcome, with higher MDS (i.e., steeper marginal ADC slope) associated with longer survival (Pearson’s correlation was 0.636, P < .005). Cox proportional hazard analysis yielded a survival model (P = .013) with MDS significantly associated with overall survival (OS) controlling for age (age P = .35, MDS P = .010). The hazard ratio for each MDS standard deviation was 0.47 (range 0.25–0.89), indicating that higher MDS predicts longer survival. Conclusions In alignment with the Imaging Biomarker Roadmap consensus, we biologically validated MDS as a biomarker of infiltrative phenotype in a preclinical model and demonstrated its predictive value for OS in humans as a prelude to larger clinical validation studies.