P01.07.A HUB-RELATED GRAPH METRICS AS MARKER FOR COGNITIVE OUTCOMES IN ADULT GLIOMA SURVIVORS AFTER MULTIMODAL THERAPY

Abstract

Abstract BACKGROUND Although cognitive impairment is a serious and prevalent long-term side effect in adult glioma survivors after therapy, predictive imaging-based markers remain inconsistent to date. Structural brain network topology could be altered by both the tumor and its’ treatment through reorganization of hubs (i.e. highly interconnected nodes). Here, we explored the role of nodal hub-related graph metrics in higher cognitive functioning. MATERIAL AND METHODS 48 WHO grade 2-3 adult glioma survivors (>1 year post-therapy) and 42 matched controls, underwent a cognitive assessment, covering five cognitive domains. Cognitive scores were transformed into w-scores, corrected for age and education. In addition, weighted nodal graph metrics were calculated based on diffusion weighted imaging, using multi-shell multi-tissue constrained spherical deconvolution and anatomically-constrained tractography. Hubs (hub score≥2) were calculated based on 4 nodal metrics, in which the 20% nodes with highest nodal strength and betweenness centrality, and lowest clustering coefficient and characteristic path length received 1 point (score 0-4). Significant differences in nodal graph metrics between patients and controls, and hubs versus non-hubs were tested in a repeated measures ANOVA, using Bonferroni correction for multiple testing. Significant group differences in nodal graph metrics were correlated to cognitive scores per domain. RESULTS In healthy controls 12/78 nodes were defined as hubs, i.e. the bilateral postcentral, precentral, superior frontal, parietofrontal, superior parietal gyri, thalamus and putamen. In glioma survivors, the left superior-parietal region and left putamen had a significantly lower likelihood of being a hub (p<.001). These specific areas showed assortativity to be positively correlated with attention, language and executive function (r(90)>.51, p<.001) and characteristic path length to be negatively correlated with all cognitive domains (r(90)>.51, p<.001). Nodal metrics of clustering coefficient, characteristic path length, local efficiency and assortativity were significantly different between groups in both hub and non-hub regions (both p<.001). Except for betweenness centrality, all nodal metrics strongly correlated with cognitive outcomes, occurring more often in hubs (33-50%) than intermediate hubs (10-33%) or non-hubs (4-12%). CONCLUSION Reorganization of hubs in the brain network is frequently observed in this cohort of glioma survivors compared to healthy controls. Hub connectivity seems to be important in cognitive functioning, more than for non-hub regions. Therefore, structural nodal hub-related graph metrics might be important to consider in predicting cognitive outcomes in glioma survivors.