Modeling and evaluating longitudinal brain maintenance and cognitive reserve using episodic memory, brain structure, and functional connectivity in older adults.

Cognitive reserve (CR) and brain maintenance (BM) reflect better than expected cognition despite brain pathology and minimal age-related brain changes that explain stable cognition, respectively. Despite being commonly used, joint quantification of these concepts has been limited; our aim is to derive longitudinal CR and BM measures and investigate CR's relationship with education and functional connectivity. We analyzed longitudinal data from 451 participants (241 female, agemean = 68.5 years, follow-upmean = 4.2 years). From a hippocampus change-memory change regression, we modeled BM as memory stability relative to hippocampal stability and CR as better-than-expected memory stability relative to hippocampal atrophy. We examined whether education and resting-state functional connectivity (Default Mode, Executive Control and anterior Salience Network) (1) moderated the hippocampal-memory relationship using linear mixed effects models (LMEs), (2) moderated the hippocampus change-memory change relationship using linear regressions, and (3) were associated with our CR measure using LMEs. Analyses adjusted for demographics and MRI scanner/cohort, using false discovery rate (FDR) for multiple comparisons. From a significant hippocampus change-memory change relationship (β = 0.130, pFDR = 0.016, r2 = 0.282), we derived continuous CR and BM measures, providing subject-level estimates. Salience Network connectivity moderated the hippocampal-memory relationship (β = - 0.070, pFDR = 0.046, marginal r2 = 0.179). No variable moderated the hippocampus change-memory change relationship or correlated with our CR measure, consistent with the measure's longitudinal definition. Our CR and BM measures refine phenotyping of aging trajectories. CR may be expressed through increased Salience Network connectivity, preserving memory level in the presence of hippocampal atrophy. Future work should delineate longitudinal CR mechanisms that decouple memory change from hippocampal atrophy.