Isocitrate dehydrogenase 1 gene (IDH-1) mutations have been shown to be present in over 80% of secondary glioblastoma (GBM), but only in 5-12% of primary GBM.¹ Given that patients with IDH-1 mutated tumors demonstrate an increased survival benefit compared to wild-type tumors following further resection of non-enhancing disease,² we hypothesized that the non-enhancing components of IDH-1 mutated and wild-type GBM differed significantly. We therefore investigated if the non-enhancing lesion demonstrated intrinsic differences between IDH-1 mutated and wild-type tumors using multimodality imaging data.

54 patients were imaged on a Siemens 3T Magnetom Trio with conventional anatomical, diffusion tensor (DTI), dynamic susceptibility contrast perfusion (DSC) and 1H chemical shift imaging (CSI) performed. All images were coregistered to the T2-weighted image using the FLIRT toolbox in FSL. The fluid attenuated inversion recovery (FLAIR) region of interest (ROI) was determined excluding the contrast-enhancing tumor and necrosis, thus forming the non-enhancing component for analysis (Figure 1).

Analysis was performed on three levels of data resolution (Figure 2). The first and second levels applied a mixed model design to DTI voxelwise and spectroscopy voxelwise data (Figure 2a and 2b respectively). The mixed model method enables robust fitting in the presence of unbalanced data, inter-subject differences and missing data. The third level of analysis approached the FLAIR VOI as a whole by averaging all variables across the total volume and modelling using a generalised linear model (Figure 2c). Significance was assessed at the 5% level with Bonferroni correction for multiple testing. Area under the receiver operator characteristics curve (AUC) was computed following multiple logistic regression. All statistical analysis was performed with 64-bit R version 3.0.2.

The FLAIR peritumoural region showed significant increases in choline relative to creatine in IDH-1 mutated tumours (Estimate 0.190 – 0.205, p = 0.004-0.008), and choline relative to N-acetyl-aspartate (Estimate 0.510–0.801, p = 0.008–0.037) compared to IDH-1 wild type tumours. Conversely, significantly lower levels of glutamate and glutamine were observed IDH-1 mutated tumours (Estimate -0.603 – -0.878, p = 0.007–0.014). These findings with age provided an area under the curve of 0.943. We propose two hypotheses to explain this finding and the improved survival in patients with IDH-1 mutated tumors. Firstly, invasive margins of IDH-mutated tumors are more sensitive to treatment. This correlates with in vitro findings,³ and in vivo IDH-1 mutated tumours with a higher Cho/NAA in the peritumoral margin may be more sensitive to cytotoxic therapies.⁴ Secondly, the reduced concentration of glutamate and glutamine identified in IDH-1 mutated tumors is neuroprotective relative to the IDH-1 wild-type. Glutamate is known to be released in excitotoxic concentrations in glioma cells and in GBM, and is involved in the mechanism of glioma invasion.⁵Across three levels of statistical analysis, IDH-1 mutated tumors demonstrated significantly different metabolite profiles in the peritumoral region than the IDH-1 wild type with high predictive power when combined with age.