Sex differences in the effects of prenatal maternal psychosocial stress on white matter microstructural development in early childhood

Using data from the publicly available Healthy Brain and Child Development (HBCD) Study, we plan to examine associations between prenatal exposure to maternal psychosocial stress and longitudinal white matter development in infancy and early childhood, with an emphasis on potential sex differences. To assess maternal psychosocial stress, we will use self-report data from the following instruments: PROMIS Perceived Stress/Social Support, PhenX+ Neighborhood Safety, PhenX+ Discrimination, eHITS, PACES, BFY Benefits/Services, Economic Stress, DSM5 Severity Acute Stress or PTSD. To assess maternal psychosocial stress a comprehensively and efficiently, we will employ principle component analysis (PCA) to reduce the dimensionality of these correlated variables into a smaller set of uncorrelated aggregate components to be used in follow-up analyses. PCA's data reduction qualities allow for fewer models to be run while retaining most of the variance of the original data. To assess white matter microstructural integrity, we will use mean apparent propagator magnetic resonance imaging (MAP-MRI), an advanced neuroimaging technique that uses data from multi-shell diffusion weighted imaging (DWI) to assess the 3D diffusion of water both intracellularly and extracellularly. This allows for the characterization of complex fiber architectures, including crossing or bending fibers, in much more detail compared to conventional single-shell DTI. To assess myelin content and composition, we will use quantitative MRI (qMRI), an advanced neuroimaging technique that is sensitive to myelin loss and neuroinflammation. This technique is complementary to MAP-MRI's ability to characterize microstructure and will provide a more complete picture of white matter development. We will perform linear mixed effects models, with white matter metrics as the outcome and maternal psychosocial stress principal component(s) as the predictor(s) of interest. To formally test for sex differences and longitudinal changes, we will include by-child age and by-child sex interactions. Covariates will include maternal age, race/ethnicity, average household income, employment status, relationship status, or proxies thereof. We will additionally include neuroimaging quality control variables (i.e., head motion), scanner type, and account for site effects. Lastly, we will correct for multiple comparisons where appropriate using false discovery rate (FDR) correction.