Structural brain markers are differentially associated with neurocognitive profiles in socially marginalized people with multimorbid illness

Homeless and marginally housed individuals constitute a socially impoverished population characterized by high rates of multimorbid illness that includes polysubstance use, viral infection, and psychiatric illness. Their extensive exposure to risk factors is associated with numerous poor outcomes, yet little is known about structural brain integrity and its association with neurocognition in this population. In Study 1, we conducted a cluster analysis to re-construct three previously derived subgroups with distinct neurocognitive profiles in a large sample of socially marginalized persons (N = 299). Cluster 1 (n = 87) was characterized as highest functioning overall, whereas Cluster 3 (n = 103) was the lowest functioning neurocognitively, with a relative strength in decision-making. Cluster 2 (n = 109) fell intermediate to the other subgroups, with a relative weakness in decision-making. Next, we examined the association between complementary fronto-temporal cortical brain measures (gyrification, cortical thickness) and neurocognitive profiles using multinomial logistic regression. Chi-square tests and ANOVAs differentiated subgroups on proxy measures of neurodevelopment and acquired brain insult/risk exposure. We found that greater frontal and temporal gyrification and more proxies of aberrant neurodevelopment were associated with Cluster 3 (lowest functioning subgroup). Further, age moderated the association between orbitofrontal cortical thickness and neurocognition, with positive associations in older adults, and negative associations in younger adults. Finally, greater acquired brain insult/risk exposure was associated with the cluster characterized by selective decision-making impairment (Cluster 2), and the higher functioning cluster (Cluster 1). In Study 2, we examined the association between white matter integrity and neurocognitive profiles using multinomial logistic regression and Tract-based Spatial Statistics. We found significantly lower fractional anisotropy (FA), with corresponding increased axial and radial diffusivity (AD, RD) in widespread and bilateral brain regions of Cluster 3. Differences in RD were more prominent compared to AD. Altogether, our findings highlight the unique pathways to neurocognitive impairment in a heterogeneous population and help to clarify the vulnerabilities confronted by different subgroups.