Association of β-Amyloid Burden With Sleep Dysfunction and Cognitive Impairment in Elderly Individuals With Cognitive Disorders

Key Points Question Is β-amyloid deposition in the brain associated with sleep dysfunction and cognition in elderly individuals with cognitive disorders? Findings In this survey study of 52 participants aged 65 years and older, β-amyloid deposition in the precuneus was associated with the number of nocturnal awakenings, whereas β-amyloid deposition in the brainstem was associated with daytime sleepiness. Nocturnal awakenings, but not daytime sleepiness, were associated with poor cognition, and β-amyloid deposition was indirectly associated with cognitive impairment via nocturnal awakenings. Meaning In elderly individuals with cognitive disorders, a mechanism that involves disruption of nighttime sleep may underlie the association between β-amyloid deposition and cognitive impairment.


Introduction
Sleep dysfunction is associated with cognitive decline in the aging population. 1,2 Increasing evidence shows that sleep and circadian rhythm disturbances predispose the brain to accumulation of β-amyloid (Aβ), 1,3 a protein metabolite that impairs neuronal function and is linked to numerous cognitive disorders, including Alzheimer disease (AD), dementia with Lewy bodies, Parkinson disease dementia, cerebrovascular dementia, and frontotemporal dementia. 4

,5 Studies 6-12 in humans have
shown that a variety of poor sleep indicators, including prolonged sleep latency, increased sleep fragmentation, decreased total sleep time, and excessive daytime sleepiness, are associated with increased Aβ deposition in the brain. Mechanistic studies have also demonstrated that sleep disruption promotes Aβ accumulation by simultaneously upregulating Aβ synthesis 13,14 and interfering with Aβ clearance. 14,15 Importantly, the relationship between sleep dysfunction and Aβ appears to be bidirectional because elevated levels of Aβ in the brain also impair slow-wave sleep, 16 thereby exacerbating sleep problems. 17,18 The synergistic relationship between Aβ abnormalities and sleep dysfunction has been shown to interfere with memory consolidation and recall. 9,16 Moreover, individuals with normal cognition experiencing poor sleep quality have increased brain Aβ burden and are at higher risk of developing dementia later in life. 8,10,19,20 Notably, patients with cognitive disorders have profound disturbances in sleep architecture and circadian rhythms. 2,21 However, studies examining the association between Aβ and sleep have, thus far, largely been performed in individuals with normal cognition, and whether Aβ continues to play an important role in sleep dysfunction after the onset of cognitive decline is unclear. Therefore, this study aims to examine the associations between Aβ abnormalities, subjective measures of sleep quality, and cognitive function in elderly individuals with cognitive disorders.

Participants
All participants in this study were patients with cognitive disorders who received care at the Cognitive Disorders and Comprehensive Alzheimer Disease Center at Thomas Jefferson University Hospital, a major referral center in the Delaware Valley. The center is an outpatient clinic that serves patients who are living independently, with family or caregivers, or in assisted living facilities. Therefore, hospitalized patients or those living in nursing homes were not included in the study, even though there were no specific exclusion criteria for these conditions. Inclusion criteria included age 65 years or older and a diagnosis of mild cognitive impairment (MCI) or dementia based on criteria established by the Diagnostic and Statistical Manual of Mental Disorders (Fifth Edition) and/or the National Institutes of Aging and the Alzheimer Association. Patients with normal cognition or subjective complaints unverified by cognitive testing were excluded from the study. Other major exclusion criteria included cancer requiring active therapy, hip or pelvic fracture within the 12 months before enrollment, and loss to follow-up. A sample of 80 patients who underwent Aβ positron emission tomography (PET) imaging at Thomas Jefferson University Hospital as part of the multicenter Imaging Dementia-Evidence for Amyloid Scanning Study (ClinicalTrials.gov identifier: NCT02420756) were originally considered for this study. Of this sample, 67 patients were deemed eligible for the study on the basis of the aforementioned criteria.
This study was approved by the institutional review board at Thomas Jefferson University Hospital and was categorized as minimal risk. Verbal informed consent was obtained from participants as they were contacted via phone for the study. The informed consent process included education about protected health information, the purpose of the study, data to be gathered by the investigators, risks of the study, steps taken to ensure participant privacy and confidentiality, and information on how to withdraw consent. Participants were told that there was no monetary compensation for the study. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 22 Data collection and analysis occurred between November 2018 and March 2019. Data pertaining to race/ethnicity were extracted from electronic health records and were based on the participants' own responses to previous health questionnaires. Race/ethnicity was subsequently assessed as a potential confounding factor in the analyses.

Cognitive Assessment
Cognition was assessed via the participants' most recent performance on the Mini-Mental State Examination (MMSE). The MMSE is a 30-point questionnaire that is widely used to evaluate the severity of cognitive impairment in individuals with cognitive disorders. 23 The questionnaire tests multiple cognitive domains, including orientation, attention and calculation, language, verbal memory, and visuospatial planning. Score cutoffs for the MMSE are as follows: greater than or equal to 24 for normal cognition, between 19 and 23 for MCI, between 10 and 18 for moderate cognitive impairment, and less than or equal to 9 for severe cognitive impairment. 24

Sleep Quality Assessment
Eligible participants and their family or caregivers were contacted via phone for sleep quality assessment; 52 participants (77.6%) gave verbal informed consent to participate in the study.
Reasons for refusing consent included not wanting to share personal information over the phone ( 25 We made certain that participants with MMSE scores less than 24 had at least 1 family member or caregiver present during the questionnaire to corroborate their responses; those with higher scores were also encouraged to respond jointly with family members to improve response accuracy.

Brain Imaging and Regional Analysis
For Aβ PET imaging, participants were injected intravenously with fluorine 18-labeled florbetaben tracer (Neuraceq; Piramel Imaging), and imaging was performed approximately 60 minutes later as per a standard protocol using either PET-magnetic resonance imaging or PET-computed tomography. Image processing was performed using MIMNeuro software (version 6.6, MIM Software), and Aβ deposition in different brain regions was calculated via standardized uptake value ratio. Regions of interest (ROIs) were based on susceptibility to Aβ deposition 19,26 and included the anterior and posterior cingulate gyrus, medial frontal lobe, medial and lateral temporal lobe, precuneus, superior parietal lobe, thalamus, and brainstem.

Statistical Analysis
The deposition of Aβ across ROIs was analyzed via the Welch analysis of variance with the Games-Howell post hoc test, which helps to detect differences between samples with unequal variances.
Associations between Aβ deposition, different sleep measures, and MMSE performance were analyzed via multiple linear regression. Data in regression analyses were modeled by least-squares regression lines, with each line having a slope equivalent to the regression coefficient (B), also known as regression weight. A Wald-type 95% CI for B was then calculated from the SE of the sampling distribution of data points from the slope of the regression line. The strength of association between variables of interest was determined by calculating the coefficient of determination (R 2 ). Forward stepwise regression was used to control for potentially confounding variables; this multivariate regression method incorporates variables in stepwise fashion until the model fit is no longer significantly improved via the F-test. Therefore, variables that make statistically significant
Subsequent regional analysis revealed that nocturnal awakenings had the strongest association with

Mediation Analysis of Aβ Deposition, Nocturnal Awakenings, and MMSE Performance
Given the much stronger association of Aβ deposition with nocturnal awakenings vs poor MMSE performance, we hypothesized that Aβ did not have a direct association with poor cognition but instead was associated with cognitive impairment indirectly via nocturnal awakenings as an intermediary. This hypothesis is supported by previous studies 9, 16 showing that sleep modulates the association between Aβ and cognition in individuals with normal cognition. To test whether a

Discussion
In patients with cognitive disorders, daytime sleepiness was associated with Aβ burden in the brainstem, whereas nocturnal awakenings were associated with Aβ burden in the precuneus. Importantly, only nocturnal awakenings were associated with poor performance on the MMSE.
Moreover, an indirect association between Aβ and cognitive impairment was found to rely on nocturnal awakenings as an intermediary.
The associations of different sleep behaviors with Aβ deposition in specific brain regions are worth noting. For example, the association between daytime sleepiness and Aβ deposition in the brainstem of individuals with cognitive impairment has not been previously reported to our knowledge. Given the important role of the brainstem in attention and arousal, 37   associations were not observed in the present study, which may, in part, reflect the limitations of the MMSE in detecting subtle cognitive changes that occur in early stages of dementia. Therefore, future evaluation with more comprehensive neuropsychological testing is needed. Another possibility to consider is that a mechanistic shift may occur during the transition from normal aging to disease. In other words, factors leading to daytime sleepiness, which are likely important for the development of dementia in individuals with normal cognition, may evolve after the onset of dementia to have more complex associations with cognitive decline.
The finding that nocturnal awakenings are negatively associated with MMSE performance is consistent with previous polysomnographic data showing that wakefulness after sleep onset is associated with cognitive impairment. 39 In addition, the results of this study provide novel insight by highlighting the role that Aβ may play in nocturnal awakenings and demonstrating that the association of Aβ with cognitive impairment relies on nocturnal awakenings as an intermediary. The indirect association of Aβ with cognitive impairment in patients with cognitive disorders may be important to consider given the failures of antiamyloid immunotherapy to improve cognition in clinical trials. 40 Our data suggest that a combinational approach targeting both Aβ and sleep dysfunction may be necessary. Notably, recent studies 41,42 indicate that tau may be a proximal mediator of sleep dysfunction, suggesting that therapeutic strategies targeting both Aβ and tau may be beneficial. Data from animal models of disease 18,43 also demonstrate the contribution of corticothalamic circuit dysfunction to wakefulness after sleep onset, a mechanism that needs to be further explored in humans.

Limitations
Although mediation analysis is a powerful tool that can disentangle direct and indirect associations among multiple variables of interest, an important point to keep in mind is that the analysis uses observational data to draw pathways of association, not causation. In particular, mediation analysis does not assess for the presence or absence of unmeasured variables, which can potentially be confounding and limit interpretation of causality. Therefore, the results of this study should only be interpreted as observational and associative, and a future interventional study would be needed to establish a causative pathway between Aβ, nocturnal awakenings, and cognitive impairment. We also acknowledge that the amount of time elapsed between MMSE testing and sleep questionnaire administration is a potential source of variability in the study. Other limitations include not assessing a wider range of sleep behaviors, such as napping. Sleep disorders other than sleep apnea were also not assessed. The lack of objective sleep data in this study precluded detailed evaluation of the effect of nocturnal awakenings on different sleep stages. Future studies should evaluate the relative contributions of rapid eye movement and non-rapid eye movement sleep disturbances to cognitive impairment. 16,39,41 Longitudinal studies with larger sample sizes are also necessary to explore how sleep dysfunction evolves over time.

Conclusions
In elderly individuals with cognitive disorders, Aβ accumulation is indirectly associated with cognitive impairment via nocturnal awakenings as an intermediary. Future investigation of the mechanism underlying this indirect association may be crucial for improved understanding of cognitive decline in disorders associated with Aβ accumulation.