A comparative analysis of different cognitive reserve estimates and their relation to connectivity in resting‐state functional MRI

Abstract Background Cognitive reserve (CR) is a concept that addresses the differences in trajectories of cognitive decline while facing neurodegenerative changes in the brain. Common proxies of CR include demographic or socio‐behavioral measures such as sex and educational/occupational attainment. Recent studies have quantified CR using a residual approach, a direct and cross‐sectional measurement of the discrepancy between estimated cognition and neurodegenerative burden. The associations between these two approaches remain largely unclear. Methods Data of 801 participants from the German DELCODE study was used, including 215 healthy controls and 586 individuals with increased risk of AD dementia (61 subjective cognitive decline, 150 mild cognitive impairment and 75 aged 80 years and older with a first‐degree relative with AD). To differentiate similar clusters in tested lifestyle factors, we conducted a dimension reduction method using principal component analysis and consequently derived 8 component factor scores. A residualized memory CR score (CR MEM ) was calculated as residuals from a multilinear regression model. Functional connectivity was analyzed using resting‐state fMRI data. Results Partial correlation analyses revealed significant positive correlations between CR MEM and component factors 1 (retirement and living), 3 (educational attainment) and 4 (social network and occupational attainment) in the entire cohort and at risk for AD dementia individuals. Higher CR MEM was associated with higher functional connectivity between the hippocampus and regions related to the default‐mode network. Factor scores 3 and 4 were associated with increased rostral hippocampus‐right angular gyrus functional connectivity. A mediation analysis showed a significant indirect effect of factor 3 on the relationship between CR MEM and rostral hippocampus‐right angular gyrus functional connectivity. Conclusion CR‐favorable socio‐behavioral factors are correlated with higher CR MEM in healthy aging and AD‐risk groups. Educational attainment might present a rather specific proxy of CR for individuals at AD‐risk. In contrast, mid‐life (occupational attainment) and late‐life factors (retirement years, social network and living status) predict CR both in AD‐risk and healthy aging groups. However, only in the AD‐risk group, hippocampal functional connectivity was higher in individuals with higher CR MEM , mediated in favor of higher educational and female sex, suggesting a close relationship to neural reserve.