MRI Features for Differentiation of Meningioma DNA Methylation Groups

BACKGROUND: DNA methylation profiling is a predictor of meningioma behavior and outcomes. We aimed to identify qualitative and quantitative MRI features to distinguish between three meningioma methylation groups: Merlin-intact, Immune-enriched, and Hypermitotic, each with distinct clinical outcomes, biologic features, and therapeutic vulnerabilities.

MATERIALS AND METHODS: Preoperative MRIs were retrospectively analyzed in meningiomas with previous DNA methylation profiling. Pearson's Chi-square, Fisher exact, and ANOVA tests were used to compare features between the three groups. ROC AUCs were used to assess the accuracy in discriminating between groups.

RESULTS: 165 patients (54 years ± 14 SD; 58 men) were analyzed. 60 meningiomas were Merlin-intact, 55 Immune-enriched, and 50 Hypermitotic. Qualitative reduced diffusion ( p< .001), nADC ( p< .001), T2WI signal intensity ( p= .02), T1 CE volume ( p<.005), and tumor site ( p<.001) varied between the groups. Merlin-intact meningiomas had higher T2WI signal intensity than Immune-enriched tumors (1.97 ± 0.98 vs 1.63 ± 0.45, p= .04). Hypermitotic meningiomas had the highest proportion of tumors with qualitative reduced diffusion (67%) and lowest nADC values (1.07 ± 0.14) compared to Merlin-intact (18%, p< .001; 1.41 ± 0.30, p< .001) and Immune-enriched (31%, p= .02; 1.29 ± 0.29, p= .002) meningiomas. The presence of qualitative reduced diffusion (AUC 0.71, p= .001) and lower nADC (AUC 0.82, p< .001) were able to predict Hypermitotic meningiomas. Merlin-intact tumors were predicted by the absence of qualitative reduced diffusion (AUC 0.66, p= 0.01), higher nADC (AUC 0.74, p< .001), and higher T2WI signal intensity (AUC 0.64, p= .047). Hypermitotic tumors (64.3 cm 3 ± 49.1) had larger T1CE volumes than Merlin-intact (42.5 cm 3 ± 37.9, p= .02) and Immune-enriched (38.2 cm 3 ± 37.7, p=<.002) tumors, with tumor size a predictor of Hypermitotic (AUC 0.65, p=.003) and Immune-enriched (AUC 0.62, p=.02) meningiomas. Merlin-intact tumors were predicted by presence at the skull base (AUC 0.67, p<.001) while Immune-enriched tumors were predicted by location outside of the skull base (AUC 0.61; 95% CI 0.70, 0.52, p= .02).

CONCLUSIONS: MR imaging has the potential to discriminate between different molecular groups of meningioma and to serve as a surrogate non-invasive marker of tumor behavior.