LONGITUDINAL SLEEP INSTABILITY CONTRIBUTES TO COGNITIVE DECLINE AND ALZHEIMER'S PATHOLOGY: FINDINGS FROM THE SEATTLE LONGITUDINAL STUDY

Abstract Introduction Although emerging data suggest a link between sleep duration, aging, and dementia diagnosis, the impact of longitudinal patterns of sleep behavior on Alzheimer’s disease-associated pathology and cognitive decline remains unclear. Methods Utilizing longitudinal data on self-reported sleep duration collected within the Seattle Longitudinal Study (n=535 subjects, age 75.18 +/- 11.6 years), we first used a Cox proportional hazard (CPH) regression model to evaluate the overall effect of mean sleep duration, longitudinal changes in sleep duration, and variability in sleep duration on the development of cognitive decline, including APOE4 genotype, sex, years of education, and depression as covariates within the model. In a secondary study, within the subset of subjects for whom postmortem histopathological assessment has been conducted (n=105 subjects, age 86.6 +/- 8.0 years), we used supervised machine learning to predict Alzheimer’s pathology, measured by CERAD neuritic plaque score and Braak stage. Results Multivariable CPH regression analysis demonstrated that there was a significant association between cognitive decline risk and APOE status (hazard ratio 2.28, 95% CI 1.41-3.69, p<0.005), depression (hazard ratio 1.09, 95% CI 1.04-1.14, p<0.005), sleep stability (hazard ratio 2.78, 95% CI 1.43-5.43, p<0.005), and sleep restlessness (hazard ratio 0.70, 95% CI 0.51-0.97, p<0.03). Notably, sleep variability across time and not median sleep duration was significantly associated with cognitive impairment. Adding these sleep parameters to the model improved model performance (base model concordance = 0.66, model including sleep = 0.72). Sleep parameters also improved performance on models predicting Alzheimer’s disease-associated pathology. Conclusion While these findings support a mixed effect of sleep duration on cognitive decline and the development of Alzheimer’s-associated neuropathology, they further suggest that variability in longitudinal sleep duration may exert a previously unappreciated influence on these pathological processes. Future studies should seek to replicate these findings in an independent cohort and to assess whether sleep variability of different time scales (weeks, years, decades) exerts similar effects on these outcomes. Support (If Any)