This cohort study examines data from a sample of participants in the Framingham Heart Study who completed 2 overnight polysomnography studies to assess whether reductions in slow-wave sleep are associated with the risk of incident dementia.
Key Points
Question
Does the percentage of slow-wave sleep decline with aging, and are intra-individual declines associated with dementia risk?
Findings
This cohort study involving 346 participants from the Framingham Heart Study found that slow-wave sleep percentage declined with aging and Alzheimer disease genetic risk, with greater reductions associated with the risk of incident dementia.
Meaning
Slow-wave sleep loss may be a dementia risk factor.
Abstract
Importance
Slow-wave sleep (SWS) supports the aging brain in many ways, including facilitating the glymphatic clearance of proteins that aggregate in Alzheimer disease. However, the role of SWS in the development of dementia remains equivocal.
Objective
To determine whether SWS loss with aging is associated with the risk of incident dementia and examine whether Alzheimer disease genetic risk or hippocampal volumes suggestive of early neurodegeneration were associated with SWS loss.
Design, Setting, and Participants
This prospective cohort study included participants in the Framingham Heart Study who completed 2 overnight polysomnography (PSG) studies in the time periods 1995 to 1998 and 2001 to 2003. Additional criteria for individuals in this study sample were an age of 60 years or older and no dementia at the time of the second overnight PSG. Data analysis was performed from January 2020 to August 2023.
Exposure
Changes in SWS percentage measured across repeated overnight sleep studies over a mean of 5.2 years apart (range, 4.8-7.1 years).
Main Outcome
Risk of incident all-cause dementia adjudicated over 17 years of follow-up from the second PSG.
Results
From the 868 Framingham Heart Study participants who returned for a second PSG, this cohort included 346 participants with a mean age of 69 years (range, 60-87 years); 179 (52%) were female. Aging was associated with SWS loss across repeated overnight sleep studies (mean [SD] change, −0.6 [1.5%] per year; P < .001). Over the next 17 years of follow-up, there were 52 cases of incident dementia. In Cox regression models adjusted for age, sex, cohort, positivity for at least 1 APOE ε4 allele, smoking status, sleeping medication use, antidepressant use, and anxiolytic use, each percentage decrease in SWS per year was associated with a 27% increase in the risk of dementia (hazard ratio, 1.27; 95% CI, 1.06-1.54; P = .01). SWS loss with aging was accelerated in the presence of Alzheimer disease genetic risk (ie, APOE ε4 allele) but not hippocampal volumes measured proximal to the first PSG.
Conclusions and Relevance
This cohort study found that slow-wave sleep percentage declined with aging and Alzheimer disease genetic risk, with greater reductions associated with the risk of incident dementia. These findings suggest that SWS loss may be a modifiable dementia risk factor.
Introduction
Inadequate sleep may be a dementia risk factor. Most notably, in experimental models, sleep deprivation promotes the accumulation of amyloid β1 and the release and spread of tau,2 the 2 proteins that aggregate in Alzheimer disease (AD). Sleep augments the clearance of metabolic waste from the brain.3 This function, termed glymphatic clearance, is optimized in the deepest non–rapid eye movement (REM) sleep stage, N3 or slow-wave sleep (SWS).3 More SWS disruption over a single night correlates with greater increases in cerebrospinal fluid amyloid-β40.4 Whereas such mechanisms are intriguing, relationships between acute sleep disruption and changes in circulating amyloid β and tau proteins do not necessarily imply that chronic SWS disruption can cause the cascade of AD pathology, leading to dementia. It remains uncertain whether chronic reductions in SWS (measured longitudinally) are associated with dementia risk in humans. Furthermore, whether dementia-related processes contribute to reductions in SWS in cognitively healthy individuals is unclear. These missing pieces of the puzzle are crucial to our understanding of how age-related changes in sleep relate to clinical dementia and may be important for informing understanding of dementia pathophysiology and therapeutic interventions.
We leveraged data within a unique community-based cohort with repeated overnight polysomnographic (PSG) sleep studies and uninterrupted surveillance for incident dementia. First, we used longitudinal data to examine how sleep, particularly SWS, changed with aging. Next, we investigated whether within-person changes in SWS percentage were associated with the risk of later-life dementia up to 17 years later. Lastly, to explore bidirectional associations between sleep and neurodegeneration, we examined whether genetic risk for AD or brain volumes suggestive of early neurodegeneration were associated with further SWS loss. Figure 1 shows the study design.
Figure 1. Overview of the Study Time Points.
Exposure variables are shown in light blue and outcomes in medium blue. A, Aim 1 investigates annualized changes in percentage of slow-wave sleep (SWS) measured between the first and second polysomnography (PSG) in association with the risk of incident dementia over the next 17 years of follow-up (n = 347 with 52 incident dementia cases). B, Aim 2 investigates the association between genetic risk of Alzheimer disease dementia and annualized change in SWS percentage (n = 345). C, Aim 2 also investigates if low hippocampal volumes were associated with annualized changes in SWS percentage between the 2 PSG studies (n = 237). Brain magnetic resonance imaging (MRI) was performed once between 1999 and 2003 but before the second PSG. The mean (SD) time interval between the brain MRI and second PSG was 2.23 (0.80) years (maximum, 4.0 years).
Methods
Participants
The study sample comprised participants from the community-based Framingham Heart Study Offspring and Omni cohorts.5 The Offspring cohort, which includes children of the Original cohort and their spouses, commenced in 1971. Surviving participants continue to be examined approximately every 4 years. In 1994, the Omni cohort was established to represent the growing diversity in the town of Framingham and includes African American, Hispanic, Asian, Indian, and Pacific Islander individuals and persons of Native American descent. Omni cohort examinations coincide with that of the Offspring cohort.
Proximal to the sixth Offspring and first Omni cohort examination cycle (1995-1998), 999 participants completed overnight PSG as part of the multicenter Sleep Heart Health Study.6 Proximal to the seventh Offspring and second Omni cohort examinations (1998-2001), 868 participants returned for a second Sleep Heart Health Study PSG study. We studied a subset of participants who, at the time of the follow-up PSG, were at least 60 years old and free from dementia and other significant neurological diseases (eg, multiple sclerosis). Sample selection is shown in the eFigure in Supplement 1. All participants provided written informed consent. The institutional review board at Boston University Medical Center approved the study. This report follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
Sleep Assessments
In-home all-night PSG was performed from 1995 to 1998 and again from 2001 to 2003, a mean (SD) of 5.2 (0.4) years apart (range, 4.8-7.1 years). The portable PSG system included electroencephalograms, electrooculograms, electrocardiograms (C3/A1 and C4/A2), chin electromyogram, oximetry, chest wall, and abdomen inductance plethysmography, and nasal/oral airflow (thermistry). Complete methods, including scoring guidelines and reliability, have been published.6,7,8 Sleep staging was scored in 30-second epochs, according to Rechtshaffen and Kales.9 We extracted sleep metrics from the first and second PSG. We then calculated the difference between the 2 time points as the annualized change in each metric. The primary variable of interest was the change in SWS percentage. Changes in other sleep variables were calculated as described in the eMethods in Supplement 1.
Dementia Case Ascertainment
The incidence of dementia was monitored through uninterrupted surveillance of all participants (eMethods in Supplement 1).10 Dementia was diagnosed with the Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition, Text Revision).11
Ascertainment of Hippocampal Volumes
Hippocampal atrophy is a hallmark of preclinical AD and low hippocampal volume cross-sectionally is associated with poorer memory, worse clinical function, and higher dementia risk, when modeled as the bottom quintile in the Framingham Heart Study.12 Therefore, as a proxy for neurodegeneration, we classified participants with the smallest age- and sex-specific hippocampal volumes on brain magnetic resonance imaging (MRI) as a percentage of intracranial volume. Quintiles of hippocampal volume were calculated separately for participants aged 60 to 70 years and 70 to 87 years. Participants in the bottom quintile were compared with participants in the top 4 quintiles. Because we were interested in whether small hippocampal volumes were accompanied by changes in SWS, we only included participants who had a brain MRI scan before their second PSG. The mean (SD) time interval between the brain MRI and second PSG was 2.23 (0.80) years (maximum interval, 4.0 years). MRI details are provided in the eMethods in Supplement 1.
AD Risk Genes
Full genotyping details are provided in the eMethods in Supplement 1. Briefly, we combined 23 single-nucleotide polymorphisms (excluding APOE) known to be associated with AD into a polygenic risk score.13,14 Participants were categorized with high genetic AD risk if they were above the 80th percentile of the polygenic risk score, intermediate genetic AD risk between the 20th and 80th percentiles, and low genetic AD risk below the 20th percentile. APOE genotypes were also obtained and examined separately. Participants were classified as either homozygous for ε3, having at least 1 copy of the ε4 allele, or as ε3/ε2 or ε2/ε2 carriers.
Demographic and Clinical Data
Data were obtained on clinical and demographic characteristics, including self-reported race, to characterize the cohorts. At each PSG, participants self-reported the use of any medications to help them sleep (never vs at least once a month), as well as the use of antidepressants and anxiolytics.
Statistical Analyses
Changes in Sleep and Risk of Incident Dementia
We calculated the annualized change in each sleep metric between the 2 PSGs (formula provided in the eMethods in Supplement 1). After checking and confirming that the proportional hazards assumption was not violated, we related annualized change in SWS percentage to the risk of incident all-cause dementia and AD dementia using Cox proportional hazards regression models. For these analyses, annualized changes in SWS percentage were reverse-scored to aid interpretability such that dementia hazard ratios were expressed per percentage decrease in SWS percentage per year. Follow-up for incident dementia was from the time of the second PSG to the time of incident dementia, death, or the date the participant was last known to be dementia-free, through 2018. All models were adjusted for age (at baseline for dementia follow-up, ie, the second PSG), sex, and cohort (Offspring or Omni). Model 2 included additional adjustment for covariates thought to be causally related to sleep or dementia, including positivity for at least 1 APOE ε4 allele, smoking status, and use of sleeping medication, antidepressants, and anxiolytics (sleep medications, antidepressants, and anxiolytics were each separately coded as 0 or 1, where 1 indicated use at either PSG).
Sensitivity and Moderation Analyses
We repeated the analysis examining changes in SWS percentage and the risk of incident dementia after doing each of the following:
Censoring the first 2 years of follow-up (to limit reverse causality).
Adding an additional model 2 adjustment for SWS percentage at the time of the first PSG.
Adding additional model 2 adjustments for the apnea-hypopnea index at the first PSG, annualized change in the index between PSGs, oxygen desaturation at the first PSG (sleep time with oxygen saturation <90%), and annualized change in oxygen desaturation.
Adding additional model 2 adjustments for self-reported physical activity measured by the Physical Activity Index15 and overall vascular risk factor burden measured by the Framingham Stroke Risk Profile16 (proximal to baseline).
Adding additional adjustments for Epworth Sleepiness Scale scores at the first PSG and annualized change in Epworth Sleepiness Scale scores (ie, between PSG 1 and 2).
Adding additional model 2 adjustments for total sleep time at the first PSG and annualized change in total sleep time between PSGs.
When examining the associations between SWS loss and dementia, we also explored for interactions by sex and REM sleep percentage (eMethods in Supplement 1).
AD Genetic Risk, Hippocampal Atrophy, and Changes in Sleep
Although poor sleep is common in patients with dementia,17 the extent to which AD genetic risk or brain changes associated with neurodegeneration contribute to SWS loss remains unknown. Linear regression models were performed to examine whether AD risk genes or hippocampal volume were associated with changes in SWS percentage, adjusting for age, sex, and cohort. The models with hippocampal volume as the exposure also included an additional adjustment for APOE ε4 carrier status. For polygenic risk score, low genetic AD risk (<20th percentile of the score distribution) was used as the reference category. For the APOE genotype, the ε3 homozygous group was used as the reference category. For hippocampal volume, quintiles 2 to 5 were used as the reference category.
Mediation Analysis
We examined whether the association between APOE ε4 positivity (vs the ε3 homozygous group) and incident dementia was mediated by annualized changes in SWS percentage. Therefore, we derived estimators of direct and indirect effects with the binary outcome of incidence of dementia, using the Cox proportional hazards model adjusted for age, sex, and cohort, and the continuous mediator (SWS) was modeled using linear regression.18
Analyses were conducted using SAS version 9.4 (SAS Institute). Missing data were assumed to be missing completely at random and were excluded from analyses. Results were deemed significant if P < .05 (2-tailed).
Results
Changes in SWS With Aging
The study sample comprised 346 participants with a mean age of 69 years (range, 60-87 years), 179 (52%) were female, 27 (8%) were Black, 18 (5%) were Hispanic, 10 (3%) were Pacific Islanders, and 291 (84%) were White (Table 1). For context, the mean percentage of SWS at baseline of 18.1% was comparable with population norms.19,20 In our study, the mean (SD) percentage of time in SWS decreased by 0.57 (1.48) units per year (Figure 2A and eTable 1 in Supplement 1); the rate of SWS loss accelerated nominally from age 60 years onwards before peaking at ages 75 to 80 years and slowing thereafter (Figure 2B). In contrast, with aging, the percentage of time in sleep stages N1 and N2 increased slightly, and REM percentage remained relatively stable (Figure 2A). Wake after sleep onset and apnea-hypopnea index increased while sleep maintenance efficiency decreased (eTable 1 in Supplement 1). Total sleep time was stable across the 2 assessments. As shown in Table 1, individuals who experienced a decline in SWS percentage were more likely to have cardiovascular disease, be an APOE ε4 carrier, and take medications affecting sleep.
Table 1. Cohort Characteristics for the Dementia Study Sample.
| Characteristic | No. (%) | ||||
|---|---|---|---|---|---|
| Overall (N = 346) | Dementia cases on follow-up | Decline in slow-wave sleepa | |||
| Yes (n = 52) | No (n = 294) | Yes (n = 222) | No (n = 124) | ||
| Age, mean (SD),y | 69 (6) | 74 (5) | 69 (6) | 70 (6) | 69 (6) |
| Sex | |||||
| Female | 179 (52) | 35 (67) | 144 (49) | 116 (52) | 63 (51) |
| Male | 167 (48) | 17 (33) | 150 (51) | 106 (48) | 61 (49) |
| Education | |||||
| No high school degree | 27 (8) | 5 (10) | 22 (7) | 20 (9) | 7 (6) |
| High school degree | 97 (28) | 13 (26) | 84 (29) | 60 (28) | 37 (30) |
| Some college | 113 (33) | 20 (40) | 93 (32) | 68 (31) | 45 (36) |
| College graduate | 105 (31) | 12 (24) | 93 (32) | 70 (32) | 35 (28) |
| Race and ethnicityb | |||||
| Black | 27 (8) | 1 (2) | 26 (9) | 15 (7) | 12 (10) |
| Hispanic | 18 (5) | 4 (8) | 14 (5) | 13 (6) | 5 (4) |
| Pacific Islander | 10 (3) | 1 (2) | 9 (3) | 7 (3) | 3 (2) |
| White | 291 (84) | 46 (88) | 245 (83) | 187 (84) | 104 (84) |
| BMI, median (IQR)c | 28 (25-31) | 27 (25-31) | 28 (25-31) | 28 (25-31) | 28 (25-30) |
| Systolic BP, mean (SD), mm Hg | 129 (18) | 133 (20) | 128 (17) | 129 (19) | 128 (15) |
| BP treatment | 131 (38) | 24 (46) | 107 (37) | 81 (37) | 50 (40) |
| Diabetes | 46 (13) | 9 (18) | 37 (13) | 28 (13) | 18 (15) |
| Prevalent CVD | 50 (14) | 11 (21) | 39 (13) | 39 (18) | 11 (9) |
| Current smoker | 23 (7) | 3 (6) | 20 (7) | 14 (6) | 9 (7) |
| APOE ε4 carrier | 75 (22) | 20 (39) | 55 (19) | 54 (25) | 21 (17) |
| Medication used | |||||
| Antidepressants | 42 (12) | 5 (10) | 37 (13) | 30 (14) | 12 (10) |
| Benzodiazepine | 30 (9) | 5 (10) | 25 (9) | 23 (10) | 7 (6) |
| Sleeping medication | 59 (17) | 9 (17) | 50 (17) | 39 (18) | 20 (16) |
Abbreviations: APOE, apolipoprotein E; BMI, body mass index; BP, blood pressure; CVD, cardiovascular disease.
Defined as a decrease in the percentage of slow-wave sleep between the first and second polysomnography.
Race and ethnicity were self-reported.
Calculated as weight in kilograms divided by height in meters squared.
Use of sleep medications, antidepressants, and benzodiazepines was coded as yes if use was specified at either polysomnography. Other variables were adjudicated at the cohort examination most proximal to the second polysomnography.
Figure 2. Changes in Sleep and Risk of Dementia.
A, Median (IQR) percentage change in each sleep stage per year (N = 346). B, Median (IQR) change in slow-wave sleep (SWS) by 5-year age increments. Results are shown for changes in SWS measured as a percentage of total sleep time (TST) and minutes. Sample sizes are 109 (age 60-65 years), 86 (>65-70 years), 79 (>70-75 years), 61 (>75-80 years), and 15 (>80 years). C, Association between annualized change in each sleep stage and incident all-cause dementia risk, adjusting for age, sex, and cohort (N = 346). For comparison across sleep stage, hazard ratios (HRs) are expressed per percentage increase in each sleep stage per year. Elsewhere, HRs for SWS are expressed per percentage decrease in SWS per year. D, Changes in each sleep stage stratified by individuals who eventually developed dementia (n = 52) and those who did not (n = 295) on follow-up. E, Individual slopes of change in SWS, stratified by individuals who eventually developed dementia (n = 52) and those who did not (n = 295) on follow-up. Participants were classified as having stable SWS if they demonstrated less than a 2.5% change from their baseline value. Otherwise, participants were categorized as increasing or decreasing. F, Cumulative incidence of all-cause dementia by changes in SWS percentage (n = 347). For illustrative purposes, we compare persons who demonstrated a decline in SWS percentage over time vs those whose SWS percentage remained stable or increased. Changes in sleep were measured between the 2 polysomnography studies over a maximum of 7.1 years (mean [SD], 4.8 [0.4] years). Dementia follow-up commenced at the time of the second polysomnography and continued for a maximum of 17 years (mean, 12 years). REM indicates rapid eye movement.
Declines in SWS and Dementia Risk
Of the 346 participants in the study sample, we observed 52 cases of incident dementia over a mean (SD) follow-up of 12 (4) years from the second PSG (maximum, 17 years). Of the incident cases, 44 were consistent with AD dementia. In Cox regression models adjusted for age, sex, and cohort, each percentage decrease in SWS per year was associated with a 28% increase in the risk of all-cause dementia and a 32% increase in the risk of AD dementia (Table 2). After further adjustments for positivity for at least 1 APOE ε4 allele, smoking status, sleeping medication use, antidepressant use, and anxiolytic use, each percentage decrease in SWS per year was associated with a 27% increase in the risk of all-cause dementia and a 32% increase in the risk of AD dementia (Table 2). Results were not meaningfully changed in a series of sensitivity analyses that (1) censored the first 2 years of dementia follow-up or included further adjustments for (2) SWS percentage at the time of the first PSG, (3) apnea-hypopnea index and sleep time with oxygen saturation less than 90%, (4) vascular risk factors and physical activity, (5) Epworth sleepiness scale scores, or (6) total sleep time (eTable 2 in Supplement 1). Changes in SWS percentage for individuals who eventually did and did not develop dementia are displayed in Figure 2D and E; the mean (SD) decline in SWS percentage was twice as large in incident dementia cases (−1.03 [1.42] per year) vs noncases (−0.48 [1.48] per year). Figure 2F shows the cumulative incidence of dementia for individuals who experienced a decline in SWS percentage relative to those who did not. The association between declining SWS percentage and incident dementia was not modified by changes in REM sleep or sex (all P > .10 for interaction). A post hoc analysis showing risk of dementia by quartiles of SWS percentage change is shown in eTable 3 in Supplement 1.
Table 2. Annualized Change in Slow-Wave Sleep and Risk of Dementia.
| Modela | All dementia | Alzheimer disease dementia | ||||
|---|---|---|---|---|---|---|
| No. of cases/total No. | HR (95% CI)b | P value | No. of cases/total No. | HR (95% CI)b | P value | |
| 1 | 52/346 | 1.28 (1.07-1.53) | .006 | 44/346 | 1.32 (1.09-1.59) | .005 |
| 2 | 51/331 | 1.27 (1.06-1.54) | .01 | 43/331 | 1.32 (1.08-1.62) | .006 |
Abbreviation: HR, hazard ratio.
Model 1 adjusts for age, sex, and cohort. Model 2 includes additional adjustments for positivity for at least 1 APOE ε4 allele, smoking status, sleeping medication use, antidepressant use, and anxiolytic use (at either polysomnography).
HRs are expressed per percentage decrease in slow-wave sleep per year.
Correlates of SWS Loss
The characteristics of the study sample with brain MRI before the second PSG are shown in eTable 4 in Supplement 1. In our sample, 232 participants (67%) were homozygous for the ε3 allele (reference), 76 (22%) had at least 1 ε4 allele, and 37 (11%) were ε3/ε2 or ε2/ε2 carriers. In linear regression adjusted for age, sex, and cohort, APOE ε4 carriers displayed greater declines in SWS relative to ε3 homozygotes (β [SE] = −0.40 [0.20]; P = .04). We did not observe any difference in SWS decline between the ε2 (ie, ε3/ε2 or ε2/ε2) and ε3 homozygote group (β [SE] = −0.26 [0.26]; P = .32). The AD polygenic risk score excluding APOE was not associated with changes in SWS (high [top 20%] vs low [bottom 20%] polygenic risk scores, β [SE] = −0.23 [0.28]; P = .41; moderate [percentile 20 through 80] vs low-risk scores, β [SE] = −0.17 [0.23]; P = .46). Changes in SWS percentage by APOE status are shown in Figure 3A.
Figure 3. Changes in Slow-Wave Sleep (SWS) by APOE Allele Status and Hippocampal Volume.
A, Median (IQR) annualized change in SWS percentage stratified by APOE allele status, including ε3 homozygotes (n = 232), ε3/ε2 or ε2/ε2 carriers (n = 37), and ε4 allele carriers (ε4/ε4, ε4/ε2, or ε4/ε3; n = 76). B, Median (IQR) annualized change in SWS percentage stratified by low (quintile 1 [n = 49]) and normal (quintiles 2-5 [n = 192]) hippocampal volume on magnetic resonance imaging. Changes in sleep were measured between the 2 polysomnography studies over a maximum of 7.1 years (mean [SD], 5.2 [0.4] years). All values are unadjusted.
Hippocampal volume (quintile 1 vs 2-5) was not associated with changes in SWS percentage, adjusting for age, sex, cohort, and APOE ε4 allele status (β [SE] = −0.40 [0.25]; P = .11) (Figure 3B).
SWS Reductions as a Mediating Factor and Other Sleep Metrics
These data raise the question of whether reductions in late-life SWS partly account for the increase in dementia risk imposed by the APOE ε4 allele. That is, declining SWS may hasten dementia onset in ε4 carriers. In a causal mediation analysis, 17% of the effect of APOE ε4 positivity (vs ε3/ε3) on incident dementia risk was mediated through changes in SWS percentage. However, the natural indirect effect was not statistically significant (hazard ratio, 1.11; 95% CI, 0.98-1.26; P = .11), indicating that late-life change in SWS percentage was not a significant mediator of the association between APOE ε4 allele carrier status and incident dementia in this sample (eTable 5 in Supplement 1).
In post hoc analyses, neither changes in other aspects of sleep macro architecture, sleep fragmentation, continuity, nor sleep apnea were associated with dementia risk (Figure 2C and eTables 6 and 7 in Supplement 1).
Discussion
These data from a community-based sample demonstrate that SWS percentage declines over time in adults 60 years and older and that greater declines are associated with the risk of future dementia. APOE ε4 genotype was independently associated with greater SWS percentage loss with aging. However, greater SWS percentage decline was associated with a higher risk of incident dementia independent of APOE ε4 status. Although we cannot rule out residual confounding, these data suggest SWS loss as a dementia risk factor.
These data extend an emerging body of research suggesting that SWS may be critically important for clearing metabolic waste, including the principal proteins that aggregate in AD.3 Some studies have shown associations between slow-wave activity and amyloid burden in humans whereas others suggest that sleep spindle–slow wave coupling, rather than the quantity of overnight slow waves per se, is associated with amyloid accumulation.21,22,23 Further, 1 study linked REM sleep microstructure to amyloid deposition in older adults.24 Studies, including ours, have failed to demonstrate an association between SWS percentage measured at a single time point and risk of cognitive decline or dementia.25,26 Thus, it may be the decline in SWS percentage, rather than individual differences at a given time, that are important for determining dementia risk. More severe SWS loss with aging may lead to a sustained dampening of glymphatic clearance, increasing the risk for amyloid plaque formation and the spread of cortical tau tangles and leading to neurodegeneration and dementia. In addition to glymphatic mechanisms, lower SWS has also been related to the development of vascular risk factors27 and may contribute to dementia through the accumulation of cerebrovascular disease. However, adjustment for vascular risk factors did not attenuate our findings.
Previous findings have been mixed in linking the APOE polymorphism to sleep disturbances at a single time point.28,29,30 Here, we demonstrate that the APOE ε4 allele (vs ε3/ε3) was associated with greater SWS loss. Because both APOE and the sleep-wake cycle are involved in amyloid production and or clearance,1,3,31 further research is required to determine their synergistic effects on dementia risk. Further research is also needed to determine the components of SWS microarchitecture that are particularly important for dementia, such as slow oscillations, and whether such sleep-dementia associations are modified by sex or other biological variables. Although several sleep wearables provide readouts of N3 sleep (ie, deep sleep), such readouts tend to be inaccurate when compared with PSG.32 Further refining wearable-device algorithms to detect and differentiate N3 sleep may yield several benefits, particularly as large-scale longitudinal tracking of N3 may lead to new research opportunities in the study of cognitive aging.
Limitations
The current study is not without limitations. We did not have gold-standard AD biomarkers available. It would be interesting to examine whether SWS loss is associated with the accumulation of amyloid β or tau. Although hippocampal volume was not related to prospective changes in SWS in our study, this does not rule out the possibility that other more sensitive AD markers may have different associations with changes in SWS (eg, brain amyloid burden or specific hippocampal subfields more affected by AD pathology). Lastly, as our study is observational, we are unable to determine whether SWS loss causes dementia. Although we conducted several analyses to investigate potential bidirectional associations, it is possible that preclinical dementia disrupts the homeostatic mechanisms that regulate SWS. However, experimental studies have provided mechanisms to suggest that SWS loss could increase dementia risk via effects on glymphatic clearance,3 vascular risk factors,27 or both. Therefore, efforts are needed to establish whether SWS enhancement can limit cognitive decline and neurodegeneration in at-risk individuals.
Conclusions
This cohort study found that SWS percentage declined with aging and Alzheimer disease genetic risk, with greater reductions associated with the risk of incident dementia. These data propose SWS loss as a potential modifiable dementia risk factor.
eMethods
eTable 1. Summary of sleep parameters across baseline and follow-up sleep studies
eTable 2. Slow wave sleep loss (per one percent decline / year) and incident dementia in pre-specified sensitivity analyses
eTable 3. Risk of dementia by quartiles of slow wave sleep change
eTable 4. Characteristics of the MRI analysis sample (N=237)
eTable 5. Mediation model to examine if changes in slow wave sleep mediate the association between APOE Ɛ4 status and incident dementia
eTable 6. Annualized change in sleep architecture and the risk of dementia
eTable 7. Sleep fragmentation/disturbances and risk of dementia
eFigure. Study Flow Diagram
eReferences
Data sharing statement
References
- 1.Kang J-E, Lim MM, Bateman RJ, et al. Amyloid-beta dynamics are regulated by orexin and the sleep-wake cycle. Science. 2009;326(5955):1005-1007. doi: 10.1126/science.1180962 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Holth JK, Fritschi SK, Wang C, et al. The sleep-wake cycle regulates brain interstitial fluid tau in mice and CSF tau in humans. Science. 2019;363(6429):880-884. doi: 10.1126/science.aav2546 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Xie L, Kang H, Xu Q, et al. Sleep drives metabolite clearance from the adult brain. Science. 2013;342(6156):373-377. doi: 10.1126/science.1241224 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Ju YS, Ooms SJ, Sutphen C, et al. Slow wave sleep disruption increases cerebrospinal fluid amyloid-β levels. Brain. 2017;140(8):2104-2111. doi: 10.1093/brain/awx148 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Tsao CW, Vasan RS. Cohort profile: the Framingham Heart Study (FHS): overview of milestones in cardiovascular epidemiology. Int J Epidemiol. 2015;44(6):1800-1813. doi: 10.1093/ije/dyv337 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Quan SF, Howard BV, Iber C, et al. The Sleep Heart Health Study: design, rationale, and methods. Sleep. 1997;20(12):1077-1085. [PubMed] [Google Scholar]
- 7.Whitney CW, Gottlieb DJ, Redline S, et al. Reliability of scoring respiratory disturbance indices and sleep staging. Sleep. 1998;21(7):749-757. doi: 10.1093/sleep/21.7.749 [DOI] [PubMed] [Google Scholar]
- 8.Redline S, Sanders MH, Lind BK, et al. ; Sleep Heart Health Research Group . Methods for obtaining and analyzing unattended polysomnography data for a multicenter study. Sleep. 1998;21(7):759-767. doi: 10.1093/sleep/21.7.759 [DOI] [PubMed] [Google Scholar]
- 9.Rechtschaffen A, Kales A. A manual of standardized terminology, techniques, and scoring system for sleep stages of human subjects. US Government Printing Office; 1968. [DOI] [PubMed]
- 10.Satizabal CL, Beiser AS, Chouraki V, Chêne G, Dufouil C, Seshadri S. Incidence of dementia over three decades in the Framingham Heart Study. N Engl J Med. 2016;374(6):523-532. doi: 10.1056/NEJMoa1504327 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.American Psychiatric Association . Diagnostic and Statistical Manual of Mental Disorders. 4th ed, text revision. American Psychiatric Association; 2000. [Google Scholar]
- 12.Weinstein G, Beiser AS, Decarli C, Au R, Wolf PA, Seshadri S. Brain imaging and cognitive predictors of stroke and Alzheimer disease in the Framingham Heart Study. Stroke. 2013;44(10):2787-2794. doi: 10.1161/STROKEAHA.113.000947 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Peloso GM, Beiser AS, Satizabal CL, et al. Cardiovascular health, genetic risk, and risk of dementia in the Framingham Heart Study. Neurology. 2020;95(10):e1341-e1350. doi: 10.1212/WNL.0000000000010306 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.van der Lee SJ, Wolters FJ, Ikram MK, et al. The effect of APOE and other common genetic variants on the onset of Alzheimer’s disease and dementia: a community-based cohort study. Lancet Neurol. 2018;17(5):434-444. doi: 10.1016/S1474-4422(18)30053-X [DOI] [PubMed] [Google Scholar]
- 15.Tan ZS, Spartano NL, Beiser AS, et al. Physical activity, brain volume, and dementia risk: the Framingham Study. J Gerontol A Biol Sci Med Sci. 2017;72(6):789-795. doi: 10.1093/gerona/glw130 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Dufouil C, Beiser A, McLure LA, et al. Revised Framingham Stroke Risk Profile to reflect temporal trends. Circulation. 2017;135(12):1145-1159. doi: 10.1161/CIRCULATIONAHA.115.021275 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Cipriani G, Lucetti C, Danti S, Nuti A. Sleep disturbances and dementia. Psychogeriatrics. 2015;15(1):65-74. doi: 10.1111/psyg.12069 [DOI] [PubMed] [Google Scholar]
- 18.Vansteelandt S, Bekaert M, Lange T. Imputation strategies for the estimation of natural direct and indirect effects. Epidemiol Methods. 2012;1(1):131-158. doi: 10.1515/2161-962X.1014 [DOI] [PubMed] [Google Scholar]
- 19.Pase MP, Harrison S, Misialek JR, et al. Sleep architecture, obstructive sleep apnea, and cognitive function in adults. JAMA Netw Open. 2023;6(7):e2325152-e2325152. doi: 10.1001/jamanetworkopen.2023.25152 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Boulos MI, Jairam T, Kendzerska T, Im J, Mekhael A, Murray BJ. Normal polysomnography parameters in healthy adults: a systematic review and meta-analysis. Lancet Respir Med. 2019;7(6):533-543. doi: 10.1016/S2213-2600(19)30057-8 [DOI] [PubMed] [Google Scholar]
- 21.Winer JR, Mander BA, Helfrich RF, et al. Sleep as a potential biomarker of tau and β-amyloid burden in the human brain. J Neurosci. 2019;39(32):6315-6324. doi: 10.1523/JNEUROSCI.0503-19.2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Chylinski D, Van Egroo M, Narbutas J, et al. Timely coupling of sleep spindles and slow waves linked to early amyloid-β burden and predicts memory decline. Elife. 2022;11:e78191. doi: 10.7554/eLife.78191 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Mander BA, Marks SM, Vogel JW, et al. β-Amyloid disrupts human NREM slow waves and related hippocampus-dependent memory consolidation. Nat Neurosci. 2015;18(7):1051-1057. doi: 10.1038/nn.4035 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.André C, Champetier P, Rehel S, et al. ; Medit-Ageing Research Group . Rapid eye movement sleep, neurodegeneration, and amyloid deposition in aging. Ann Neurol. 2023;93(5):979-990. doi: 10.1002/ana.26604 [DOI] [PubMed] [Google Scholar]
- 25.Pase MP, Himali JJ, Grima NA, et al. Sleep architecture and the risk of incident dementia in the community. Neurology. 2017;89(12):1244-1250. doi: 10.1212/WNL.0000000000004373 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Song Y, Blackwell T, Yaffe K, Ancoli-Israel S, Redline S, Stone KL; Osteoporotic Fractures in Men (MrOS) Study Group . Relationships between sleep stages and changes in cognitive function in older men: the MrOS Sleep Study. Sleep. 2015;38(3):411-421. doi: 10.5665/sleep.4500 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Javaheri S, Zhao YY, Punjabi NM, Quan SF, Gottlieb DJ, Redline S. Slow-wave sleep is associated with incident hypertension: the Sleep Heart Health Study. Sleep. 2018;41(1):zsx179. doi: 10.1093/sleep/zsx179 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Thakre TP, Mamtani MR, Kulkarni H. Lack of association of the APOE epsilon 4 allele with the risk of obstructive sleep apnea: meta-analysis and meta-regression. Sleep. 2009;32(11):1507-1511. doi: 10.1093/sleep/32.11.1507 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Tranah GJ, Yaffe K, Nievergelt CM, et al. ; Osteoporotic Fractures in Men Study (MrOS) Research Group . APOEε4 and slow wave sleep in older adults. PLoS One. 2018;13(1):e0191281-e0191281. doi: 10.1371/journal.pone.0191281 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Blackman J, Love S, Sinclair L, Cain R, Coulthard E. APOE ε4, Alzheimer’s disease neuropathology and sleep disturbance, in individuals with and without dementia. Alzheimers Res Ther. 2022;14(1):47. doi: 10.1186/s13195-022-00992-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Liu CC, Zhao N, Fu Y, et al. ApoE4 accelerates early seeding of amyloid pathology. Neuron. 2017;96(5):1024-1032.e3. doi: 10.1016/j.neuron.2017.11.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Haghayegh S, Khoshnevis S, Smolensky MH, Diller KR, Castriotta RJ. Accuracy of wristband Fitbit models in assessing sleep: systematic review and meta-analysis. J Med Internet Res. 2019;21(11):e16273. doi: 10.2196/16273 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods
eTable 1. Summary of sleep parameters across baseline and follow-up sleep studies
eTable 2. Slow wave sleep loss (per one percent decline / year) and incident dementia in pre-specified sensitivity analyses
eTable 3. Risk of dementia by quartiles of slow wave sleep change
eTable 4. Characteristics of the MRI analysis sample (N=237)
eTable 5. Mediation model to examine if changes in slow wave sleep mediate the association between APOE Ɛ4 status and incident dementia
eTable 6. Annualized change in sleep architecture and the risk of dementia
eTable 7. Sleep fragmentation/disturbances and risk of dementia
eFigure. Study Flow Diagram
eReferences
Data sharing statement