Pathways from chronic stress to cognitive decline: A multi-cohort mediation analysis of inflammation and...

Chronic stress is a well-established risk factor for poorer cognitive aging, but the biological mechanisms underlying these associations remain unclear. Inflammatory processes have been proposed as key mechanisms, but evidence across large, independent cohorts is limited. We conducted statistical mediation analyses to examine whether inflammatory biomarkers and immune-related gene expression show indirect effect estimates consistent with mediating associations between chronic stress and cognitive performance in two population-based cohorts: Midlife in the United States Study (MIDUS; N = 209-276) and UK Biobank (UKB; N up to 132,068). Chronic stress was assessed using theory-driven composites of childhood adversity and adult loneliness/social isolation. Mediators included five inflammatory biomarkers (interleukin-6, C-reactive protein, tumor necrosis factor-α, E-selectin, and intercellular adhesion molecule-1) and genes comprising the Conserved Transcriptional Response to Adversity. Cognitive outcomes indexed memory, executive function, processing speed, reasoning, and fluid intelligence. Mediation models tested indirect effect estimates, with age moderation. Across cohorts, childhood physical and emotional/psychological abuse were the most consistent predictors of poorer cognitive performance. In the UKB, statistical results were consistent with partial mediation by C-reactive protein of associations between childhood physical and emotional abuse with fluid intelligence, and by intercellular adhesion molecule-1 of the association between childhood emotional abuse and processing speed after correction for multiple testing. Gene-expression indirect effect estimates for TNF in MIDUS were positive and nominally significant, they did not survive FDR correction but suggested involvement of stress-related inflammatory signaling pathways. Independent of mediation, robust direct associations of childhood maltreatment with lower fluid intelligence and slower processing speed were observed across cohorts. Inflammation was associated with, but did not fully account for, stress-related differences in cognitive function in adulthood, highlighting the need to consider parallel biological and psychosocial pathways.