Date Published: June 13, 2019
Publisher: Public Library of Science
Author(s): Kristina M. Gicas, Andrea A. Jones, William J. Panenka, Chantelle Giesbrecht, Donna J. Lang, Fidel Vila-Rodriguez, Olga Leonova, Alasdair M. Barr, Ric M. Procyshyn, Wayne Su, Alexander Rauscher, A. Talia Vertinsky, Tari Buchanan, G. William MacEwan, Allen E. Thornton, William G. Honer, Pei-Ning Wang.
Cognition is impaired in homeless and vulnerably housed persons. Within this heterogeneous and multimorbid group, distinct profiles of cognitive dysfunction are evident. However, little is known about the underlying neurobiological substrates. Imaging structural covariance networks provides a novel investigative strategy to characterizing relationships between brain structure and function within these different cognitive subgroups.
Participants were 208 homeless and vulnerably housed persons. Cluster analysis was used to group individuals on the basis of similarities in cognitive functioning in the areas of attention, memory, and executive functioning. The principles of graph theory were applied to construct two brain networks for each cognitive group, using measures of cortical thickness and gyrification. Global and regional network properties were compared across networks for each of the three cognitive clusters.
Three cognitive groups were defined by: higher cognitive functioning across domains (Cluster 1); lower cognitive functioning with a decision-making strength (Cluster 3); and an intermediate group with a relative executive functioning weakness (Cluster 2). Between-group differences were observed for cortical thickness, but not gyrification networks. The lower functioning cognitive group exhibited higher segregation and reduced integration, higher centrality in select nodes, and less spatially compact modules compared with the two other groups.
The cortical thickness network differences of Cluster 3 suggest that major disruptions in structural connectivity underlie cognitive dysfunction in a subgroup of people who have a high multimorbid illness burden and who are vulnerably housed or homeless. The origins, and possible plasticity of these structure-function relationships identified with network analysis warrant further study.
Excessive multimorbidity is a prominent feature of marginalized populations worldwide, including persons who are homeless or vulnerably housed . Common co-occurring illnesses include psychosis, polysubstance use, HIV infection, and traumatic brain injury [2–4], which may reflect a set of genetic, environmental, and developmental vulnerabilities that predispose individuals to significant cognitive dysfunction. Previous studies in marginalized populations have shown substantial cognitive heterogeneity and impairment across domains [5–7] and this was linked with regional alterations in cortical thickness and gyrification , and to variation in white matter microstructure . However, due to the scope of illness burden among marginalized persons, there are likely to be more widespread alterations to structural brain integrity, which warrants further exploration. Given that multimorbidity is an emergent global health issue , there is impetus for extending our knowledge of structure-function relationships by characterizing distinct neural and cognitive phenotypes that reflect the consequences of multiple co-occurring psychiatric and physical illnesses, as opposed to those associated with single diagnostic categories.
We identified three distinct profiles of cognitive functioning that were associated with underlying differences in cortical brain network topology in a large sample of homeless and vulnerably housed adults with significant multimorbidity. Using structural covariance network analyses of complementary cortical parameters, we found significant between-group differences in network coefficients for cortical thickness. Specifically, we found that Cluster 3 (the lowest cognitive functioning group) was significantly higher on measures of segregation (clustering coefficient, transitivity, local efficiency, modularity) and demonstrated reduced network integration (longer characteristic path length, lower global efficiency) compared to the other groups for the cortical thickness networks. All networks demonstrated a small-world architecture. Regionally, Cluster 3 exhibited higher node betweeness and had a greater number of hubs overall compared to Clusters 1 and 2. No differences were observed between Clusters 1 and 2 in any comparison. At minimum density, five modules were identified for each cortical thickness network, with noticeably different patterns of node distribution across modules within each group.