Abstract Introduction Growing evidence suggests that sleep macro-architecture features show inconsistent associations with cognitive outcomes in aging, including Alzheimer’s disease (AD). These inconsistencies underscore the need to clarify which specific components of sleep architecture are most strongly linked to cognition and AD pathogenesis. Our study aims to comprehensively evaluate multiple sleep architecture features and determine how they connect with cognition and AD biomarkers, using network analysis to map potential pathways. Methods We included 128 participants (23 cognitively normal, 41 mild cognitive impairment, 64 AD) with overnight polysomnography, from which we extracted sleep architecture measures including total sleep time, sleep efficiency, sleep onset latency, REM latency, and stage percentages, plasma AD biomarkers (p-tau181, NfL, BDNF), Aβ PET, and domain-specific cognitive scores. Network analysis was used to quantify conditional interrelationships among sleep architecture, AD biomarkers, and to identify key pathways linking these domains. A partial-correlation–based network was constructed to map associations among sleep architecture, biomarkers, and cognition. Centrality metrics (strength, closeness, betweenness, expected influence) were used to identify key bridge nodes integrating distinct clusters. Results Among all sleep parameters, REM latency emerged as the most central feature in the network. REM latency showed partial correlations of –0.23 with BDNF and +0.23 with Aβ PET, supporting its linkage to AD pathology, and its high closeness centrality indicated that it serves as an efficient integrative node linking sleep, biomarker, and cognitive domains. REM latency also demonstrated one of the highest betweenness values, further underscoring its bridging role in the network. Plasma p-tau181 demonstrated strong negative partial correlations with global cognition (–0.31) and performance on the language domain (–0.27), and its greatest expected influence indicated that tau exerted the strongest net effect across the entire system, shaping not only AD pathology and cognitive performance but also linking back to key sleep features. Consequently, the dominant cross-domain pathway was REM latency → p-tau181 → cognition, supporting a biomarker-mediated model. Conclusion Our results demonstrate that REM latency and plasma p-tau181 form a central bridge between sleep architecture, AD biomarkers, and cognitive impairment. This underscores the critical mechanistic role of REM latency and tau pathology in pathways linking sleep disruption to AD. Support (if any)
Chen et al. (Fri,) studied this question.