OS6.1 Glioma anatomic location and clinical phenotype relate to regional healthy brain network signature. (19th September 2018)

• Record Type: Journal Article
• Authors: Numan, T; Derks, J; de Witt Hamer, P C; Gerstner, E R; Stufflebeam, S M; Alexander, B; van Dijk, K R A; Cagney, D N; Reijneveld, J C; Douw, L
• Is Part Of: Neuro-oncology. Volume 20(2018)Supplement 3
• Page Start: iii225
• Page End: iii226
• Publication Date: 2018-09-19
• Subjects: Brain Neoplasms -- Periodicals; Brain -- Tumors -- Periodicals; Brain -- Cancer -- Periodicals; Nervous system -- Cancer -- Periodicals; 616.99481
• DOI: 10.1093/neuonc/noy139.038
• Languages: English
• ISSNs: 1522-8517

Abstract

Abstract: Background: In network neuroscience, patterns of communication between brain regions are investigated. Particularly high connectivity (' hubness ') may predispose certain regions to pathology and may impact clinical symptoms. For example, Alzheimer's pathology and cognitive decline are related to hubness . The relation between hubness, glioma location and clinical phenotypes is currently unknown. We explored whether (1) glioma location associates with regional hubness of the healthy brain, and (2) whether Lesional hEalthy Network SignaturE (LENSE), which describes the regional healthy hubness of the tumor location of an individual patient, relates to tumor volume, overall survival (OS) and performance status. Material and Methods: Glioma patients (LGG n=46, HGG n=37), n=21 IDH mutant, 1p19q codeleted (type I), n=30 IDH mutant, non-codeleted (type II), and n=15 IDH wildtype (type III), n=17 molecular subtyping not available) were included at VU University Medical Center. Tumor masks were manually drawn on 3DT1-weighted MRI and concatenated into a cohort tumor occurrence map, which was parcellated into 210 cortical regions (Brainnetome atlas). Healthy regional hubness was obtained from 96 healthy controls using preprocessed fMRI. Pearson's correlations between time series from all 210 regions defined connectivity strength. Regional hubness (the sum of all connections per region) was non-parametrically correlated to regional tumor occurrence. This analysis was repeated in a second cohort of 121 glioma patients (collected at Massachusetts General Hospital) and a cohort of brain metastases (n=17, collected at Dana-Farber Cancer Institute). LENSE was quantified by averaging the healthy hubness of the regions within the tumor mask of each individual patient. Results: Locations with high glioma occurrence had low hubness in healthy controls and vice versa (R=-0.45, p=1.9e-10), which was replicated in the second glioma cohort (R=-0.43, p=2.0e-09), but not in metastases patients (R=0.06, p=0.40). Higher healthy hubness of the tumor location related to larger tumor volume (R=0.54, p=1.8e-07). Within type I patients, higher healthy hubness of the tumor location related to lower KPS (p=0.071) and shorter OS (HR=1.36 (95% confidence interval 1.02–1.82), p=0.04). Conclusion: Glioma locations are generally marked by lower healthy regional hubness, suggesting (1) 'beneficial glioma-outgrow' properties in low hubness regions, (2) protective properties in high hubness regions, or (3) both. However, when gliomas do occur in high hubness regions, they represent a poorer clinical phenotype, particularly within type I gliomas. Our results suggest that LENSE may convey additional prognostic value in (subpopulations of) glioma patients.