Linking sleep microstructure, blood markers of inflammation and metabolism, and cognition: mediation analysis in the osteoporotic fractures in men study

The sleep electroencephalographic (EEG) microstructure is explicitly related to brain functions, such as sleep spindle and memory consolidation. On the other hand, given the crosstalk between the central nervous system and the body, including inflammation regulation and metabolic systems, there is a gap in understanding the bidirectional relationship between sleep and these systems that contribute to cognitive health. We used data from the Osteoporotic Fractures in Men (MrOS) Study, a six-site cohort study of community-dwelling men 65 years or older. We analyzed 1898 participants who participated in the Sleep Visit and Visit 2. We analyzed 41 sleep EEG microstructures and 9 blood markers. The outcome was the Modified Mini-Mental State Examination (3MS) in Visit 2. We used partial Spearman's correlation to investigate the pairwise associations and performed 3MS prediction. We then performed mediation analyses using each blood marker as the exposure and each sleep EEG microstructure as the mediator, and then the other way around. Sleep EEG microstructures were more strongly correlated with 3MS than blood markers in general. The best Pearson's correlation between the actual and predicted 3MS scores was 0.45 (95% confidence interval 0.38-0.47) using sleep EEG microstructures and covariates, which provided little improvement compared to using covariates alone (0.43, 0.39-0.48). Leptin and 3MS score were associated (Spearman's rho = 0.053, p = 0.02) while adjusting for covariates. The association between leptin and 3MS score was mediated by fast spindle density (indirect effect = 0.030, p = 0.02) and spindle-slow oscillation (SO) overlap (indirect effect = 0.029, p = 0.02). No blood marker mediated the association between sleep EEG microstructure and 3MS. Instead, the following sleep EEG microstructures had significant direct associations with 3MS independent of the blood markers: theta power during N1 and REM (direct effect = - 0.57 and - 0.46, p = 0.0002 and 0.007, respectively), spindle density (direct effect = 0.39, p = 0.006), and spindle-SO coupling overlap (direct effect = 0.29, p = 0.01). The blood markers of inflammation and metabolism were less predictive of and indirectly associated with global cognition compared to the sleep EEG microstructures. Future work is needed to confirm these results in an experimental setting.