Abstract
Socio-environmental and health-related variables were examined in relation to longitudinal change in select neuroimaging markers through metabolomics. Data from 2255 dementia-free UK Biobank participants were utilized. Statistical analyses involved descriptives, Principal Components Analysis (PCA) for metabolomic data reduction, mixed-effects linear regression models to assess longitudinal change (i.e. empirical Bayes estimators of slope), and Additive Bayesian Networks (ABN). Age was the primary consistent contributor to brain health decline over time, with specific metabolomic markers, mainly "free cholesterol in very large high-density lipoproteins (HDL)", potentially offering protective effects against declines in microstructural integrity, through reduction of or slower pace of increase in mean Orientation Dispersion (ODmean). Air pollution, individual and household-level SES, sex and racial minority status correlated indirectly with brain health through intracranial volumes and time interval between assessments. These insights emphasize using a multifactorial approach to understanding brain aging for predictive models of neurodegeneration.</p>