Contributions
Abstract: P502
Type: Poster Sessions
Abstract Category: Pathology and pathogenesis of MS - MRI and PET
Background: Coordinate based meta-analysis of voxel-based morphometry (VBM) studies of grey matter (GM) density shows consistent regional atrophy in MS; where the coordinates reported by the studies form dense clusters. Between regions the effect size (reported Z scores normalised by the number of subjects), which quantifies relative degree of atrophy, is correlated. These regions may form networks of GM atrophy.
Objective: use novel Coordinate Based Network Meta-Analysis (CBNMA) to look for GM atrophy networks.
Methods: Search for VBM studies comparing MS or CIS to healthy controls. Use CBNMA to locate regions (clusters of reported coordinates, which form the network nodes) where atrophy is reported consistently across the studies, then estimate a correlation (of effect size to form the network edges) matrix between the regions; computed by fitting (pairwise) bivariate normal distributions to the effect sizes. The likelihood ratio test provides asymptotic p-values for each edge, and thresholding the p-values by false discovery rate (FDR) produces an adjacency matrix that defines network connections. Research into the thresholds needed to create brain networks is ongoing. Here FDR was chosen to create, as far as possible, symmetric networks.
Results: 43 studies met the inclusion criteria and reported 523 foci in total. Two separate networks were identified. An FDR of 0.1 was chosen as it produced symmetric networks. Other FDR levels resulted in the same nodes (regions of consistent GM atrophy), but some edges were missing at lower FDR and extra edges added at higher FDR. Network 1 involved (bilaterally) the thalamus, Postcentral Gyrus (Brodmann area 3), and the Claustrum. In this network the most connected nodes were the postcentral gyri, where atrophy was correlated with atrophy in every other node. The second network involved bilateral insula.
Conclusions: Multiple GM regions are reported as atrophied consistently by VBM studies comparing CIS and MS to healthy controls. These regions form independent networks of atrophy, where knowing the degree of atrophy in one region can indicate atrophy in others from the same network. Clinical trials using GM atrophy as an outcome might then test for significant networks, rather than significant voxels or regions of atrophy. The single statistical test of a network would require no correction for multiple comparisons, and thus be more powerful necessitating smaller sample sizes.
Disclosure: Tench: nothing to disclose
Singh: nothing to disclose
Tanasescu: nothing to disclose
Constantinescu: nothing to disclose
Abstract: P502
Type: Poster Sessions
Abstract Category: Pathology and pathogenesis of MS - MRI and PET
Background: Coordinate based meta-analysis of voxel-based morphometry (VBM) studies of grey matter (GM) density shows consistent regional atrophy in MS; where the coordinates reported by the studies form dense clusters. Between regions the effect size (reported Z scores normalised by the number of subjects), which quantifies relative degree of atrophy, is correlated. These regions may form networks of GM atrophy.
Objective: use novel Coordinate Based Network Meta-Analysis (CBNMA) to look for GM atrophy networks.
Methods: Search for VBM studies comparing MS or CIS to healthy controls. Use CBNMA to locate regions (clusters of reported coordinates, which form the network nodes) where atrophy is reported consistently across the studies, then estimate a correlation (of effect size to form the network edges) matrix between the regions; computed by fitting (pairwise) bivariate normal distributions to the effect sizes. The likelihood ratio test provides asymptotic p-values for each edge, and thresholding the p-values by false discovery rate (FDR) produces an adjacency matrix that defines network connections. Research into the thresholds needed to create brain networks is ongoing. Here FDR was chosen to create, as far as possible, symmetric networks.
Results: 43 studies met the inclusion criteria and reported 523 foci in total. Two separate networks were identified. An FDR of 0.1 was chosen as it produced symmetric networks. Other FDR levels resulted in the same nodes (regions of consistent GM atrophy), but some edges were missing at lower FDR and extra edges added at higher FDR. Network 1 involved (bilaterally) the thalamus, Postcentral Gyrus (Brodmann area 3), and the Claustrum. In this network the most connected nodes were the postcentral gyri, where atrophy was correlated with atrophy in every other node. The second network involved bilateral insula.
Conclusions: Multiple GM regions are reported as atrophied consistently by VBM studies comparing CIS and MS to healthy controls. These regions form independent networks of atrophy, where knowing the degree of atrophy in one region can indicate atrophy in others from the same network. Clinical trials using GM atrophy as an outcome might then test for significant networks, rather than significant voxels or regions of atrophy. The single statistical test of a network would require no correction for multiple comparisons, and thus be more powerful necessitating smaller sample sizes.
Disclosure: Tench: nothing to disclose
Singh: nothing to disclose
Tanasescu: nothing to disclose
Constantinescu: nothing to disclose