Presenter(s)
W. J. Yeo1, S. Nagarajan1, A. Raj1, H. Lin2, W. Yang1, and K. Sheng3; 1University of California, San Francisco, San Francisco, CA, 2Department of Radiation Oncology, University of California San Francisco, San Francisco, CA, 3Department of Radiation Oncology, University of California, San Francisco, San Francisco, CA
Purpose/Objective(s):
Studies have suggested that specific neuronal signals and existing physical brain structures may influence glioma infiltration patterns. We aim to provide novel assessments of glioma infiltration preferences, risks, and survival by functional networks and tracts.
Materials/Methods:
The study population was 1235 patients with incident glioma (mean age 59 years, 41% women), stratified into methylated MGMT glioblastoma (GBM), unmethylated MGMT GBM, and high-grade and low-grade glioma (HGGLGG) subgroups. Glioma was defined with the necrotic, enhancing, and non-enhancing regions. T1 MR images segmented into structural brain regions were registered to Yeo’s functional network atlas (7 classes: visual, somatomotor, dorsal attention, salient, limbic, frontoparietal, and default mode) and Yeh’s tractography atlas (5 classes: association, projection, commissural, cerebellum, cranial nerves). Each brain region was assigned to the network and tract class with the largest volume intersection. Glioma infiltration preference was quantified with prevalence of any intersection amount with each brain region across the population. Risks of glioma infiltration by clinical characteristics (age, sex, tumor volume, and tumor irregularity) were assessed with linear regressions where the outcome was the proportion of brain region infiltrated by tumor. Survival risks for infiltration into each brain region was assessed with Cox proportional-hazards models comprehensively adjusted with clinical characteristics, as well as extent of resection. Significance in associations were defined with the Bonferroni-corrected threshold p-value.
Results:
Across GBM and HGGLGG patients, glioma occurred most prevalently (70%) within the cerebral white matter which corresponds to the default mode network. For unmethylated MGMT GBM, age was slightly protective for glioma fully infiltrating the brain stem (beta -7.4E-4, p = 2.5E-4) and left thalamus proper (beta -2.9E-3, p = 1.6E-4). Both corresponded to the projection tract class, and the former was mapped to the limbic network as well. However, infiltration into the brain stem for unmethylated MGMT GBM was a risk factor for mortality (hazard ratio per 10% region infiltration (HR) 1.7, 95% confidence interval [1.5, 2.0], p = 4.9E-5). For methylated MGMT GBM, the frontal pole, superior frontal gyrus, and supplementary motor cortex on the right hemisphere were risk factors for mortality (HR > 1.15, p < 3.3E-4). All regions belonged to the default mode network and association tract class, apart from the supplementary motor cortex that was mapped to the salient network.
Conclusion:
The default mode network plays an important role in influencing glioma infiltration, and there are differences in risks of infiltration into each network and tract class, as well as their subsequent prognoses, by glioma subtypes. Our study provides insights to the pathophysiology of gliomas and may be used to improve treatment with functional considerations.