Identifying the specific aspects of sleep that relate to incident dementia is the first step towards the development
of sleep interventions to reduce dementia risk. Detailed overnight sleep studies, known as polysomnography
(PSG), provide the gold-standard assessment of sleep. As obtaining PSG is burdensome, studies with PSG tend
to enroll a limited number of participants and consequently have limited statistical power to detect small but
potentially important associations between sleep and dementia. We propose to curate data from 5 large
population-based cohorts (Atherosclerosis Risk in Communities, Cardiovascular Health Study, Framingham
Heart Study, Osteoporotic Fractures in Men, and the Study of Osteoporotic Fractures) with methodologically
consistent sleep studies and neurocognitive outcomes. By combining study-level data in meta-analysis, we
propose the following aims: Aim 1 is to examine the aspects of sleep that relate to a higher risk of incident
Alzheimer's disease (AD) dementia (N=2776, 499 incident cases). We will capture 134 sleep metrics,
measuring all aspects of sleep neurophysiology. We will then identify clusters and calculate the first principal
component from each cluster as the exposers. We will further assess the association between cluster specific
sleep metrics and outcomes using least absolute shrinkage and selection operator (LASSO) regression 1.a. We
will examine each aspect of sleep neurophysiology with respect to the risk of incident AD dementia, after
accounting for known confounders. 1.b. We will leverage our statistical power to explore differences by age
decades, sex, and genetic risk (e.g., APOE e4 positivity). Aim 2 is to examine the aspects of sleep (defined
in Aim 1) that relate cross-sectionally to dementia endophenotypes. As poor sleep is potentially modifiable,
it is important to know whether poor sleep is related to preclinical phenotypes of dementia—a time when
dementia risk may still be malleable. Brain atrophy on MRI and subtle deficits in cognitive ability precede
dementia diagnosis by up to a decade. We will relate each sleep marker to general and domain-specific cognitive
performance (N=6723) as well as brain volume (total brain and hippocampal) and brain injury (white matter
disease, silent infarcts) on MRI (N=1157). Aim 3 is to examine whether changes in sleep neurophysiology
over ~6 years predict incident dementia (N=1558, 275 events), cognition (N=3065), or brain volume
(N=763). Leveraging repeated PSGs ~6 years apart, we will examine if changes in sleep neurophysiology relate
to incident AD dementia, brain volume, or cognitive function. Our large analysis of community-based participants
from across the U.S. will provide the most robust evidence yet on the associations between sleep and AD
dementia risk. Moreover, leveraging our large pooled sample size to examine subgroup differences (e.g., by age
decades, sex and APOE) and the comprehensive investigation of sleep neurophysiology, including innovative
sleep measures (e.g., spindle density), may inform therapeutic strategies for dementia prevention by identifying
subgroups most at risk, new biomarkers to improve dementia risk stratification, and novel biological pathways.