Both short and long sleep durations have been associated with an increased risk of future dementia [1, 2]. Transitioning to long sleep duration may be a marker of ongoing neurodegeneration [1]. However, the mechanisms linking short sleep time to dementia risk are unclear; short sleep duration may directly relate to Alzheimer’s Disease (AD) pathology by limiting the opportunity for glymphatic clearance [3], or may associate with dementia independent of AD pathology through shared risk factors, such as hypertension [4] or genetic vulnerability for AD [5]. In a study of older adults, self-reported short sleep duration (≤6 vs. 7–8 h) was associated with higher amyloid beta (Aβ) burden measured by positron emission tomography (PET) scans [6]. However, other AD biomarkers (e.g. tau) were not examined. Since associations between sleep and AD may be bidirectional, it is important to establish these relationships in younger cognitively unimpaired individuals, to determine the extent to which sleep time is associated with the earliest AD biomarker changes.
This preliminary study aimed to examine if short nighttime sleep duration was associated with amyloid, tau, and neurodegeneration (A/T/N) cerebrospinal fluid (CSF) biomarkers of AD in a community-based cohort of middle-aged adults with a family history of dementia. We hypothesized that shorter sleep would be associated with lower Aβ42 (which decreases with increasing AD biomarker severity) [7] and higher total-tau, phosphorylated-tau 181 (p-tau), and neurofilament light chain (NfL) levels.
The Healthy Brain Project (HBP) is a community-based study that tracks the cognitive health of Australians with annual online assessments [8]. Eligible participants include those aged 40–70 years, residing in Australia, and fluent in English. Participants were self-referred and recruited through various sources, such as social media and advertisements. A total of 82 participants completed a biomarker sub-study, including a lumbar puncture and 2 weeks of actigraphy. Participants with self-reported cognitive impairment, neurodegenerative disease, or dementia were excluded, as were those taking medications for dementia or AD. The cohort was enriched to over-represent APOE ε4 carriers (38%) and 66 participants completed all measures and were included in the study. The study was approved by the Melbourne Health Human Research Ethics Committee.
CSF samples were obtained by single pass lumbar puncture in the L3/L4 or L4/L5 interspace at the Royal Melbourne Hospital (Australia), processed, and stored at −80º. The biochemical analyses of Aβ42, t-tau, and p-tau181 from thawed CSF were conducted using the Roche Elecsys immunoassay, whereas NfL (a nonspecific marker of neurodegeneration) was measured using ELISA (UmanDiagnostics). All biochemical analyses were performed by the National Dementia Diagnostics Laboratory (The Florey Institute, University of Melbourne, Australia). A total of 22 (33%) participants had CSF Aβ42 scores above the maximum limit of detection and were assigned the maximum detectable score (1700 pg/mL).
Sleep was averaged from 17 consecutive days of wrist-worn actigraphy (Phillip Respironics Actiwatch Spectrum Plus) and a sleep diary. The first 2 nights of data were removed to avoid any impact of the lumbar puncture on sleep. Where there was a discrepancy of over 1 h between diary and Actiwatch, sleep duration was derived from the sleep diary. Participants with less than 10 days of actigraphy were excluded from analyses (n = 3).
Short sleep was related to the CSF biomarkers adjusting for age, sex, and body mass index (BMI; Model 1). A second model included an additional adjustment for APOE ε4 status (Model 2), and a third model included adjustments for average sleep efficiency, daytime nap duration, and the sleep regularity index (SRI), in addition to age, sex, BMI, and APOE ε4 status. The SRI [9] was calculated as the probability of being in the same state (asleep or awake) between each contiguous 2-day pair, averaged over the full recording period. A score of 0 indicated completely irregular sleep and 100 indicated perfectly regular sleep. A small number of participants had missing data for APOE ε4 genotype (n = 3) and SRI (n = 4). Multiple imputation by predictive mean matching (producing 10 imputed datasets) was used prior to analysis (R package mice) [10]. Our analysis code is available online: https://osf.io/ekj6f/.
The mean age of the sample was 59; 66% were female, 97% white, and 41% APOE ε4 carriers (Table 1). Eight participants (12%) had short sleep duration (≤ 6 h) and 58 had normal sleep duration (>6 and <9 h; Table 1), not inclusive of naps. No participants displayed long sleep duration (≥9 h). No participants reported sleep disorders. One participant with normal sleep duration reported sleep medication use. Twelve (18%) participants had CSF amyloid levels ≤1000 pg/mL, indicative of abnormal amyloid [11].
Table 1.
Sample demographics stratified by sleep duration
| Variable | Sleep duration | p † | |
|---|---|---|---|
| Short | Normal | ||
| (≤ 6 h) | (>6 and <9 h) | ||
| N | 8 | 58 | |
| Age, years | 55.7 (5.8) | 58.9 (6.9) | .18 |
| Education, years | 16.2 (3.6) | 16.0 (3.4) | .87 |
| Females, N (%) | 4 (50) | 39 (67) | .34 |
| APOE ε4 carrier, N (%)‡ | 4 (67) | 23 (40) | .23 |
| MMSE scores | 28.5 (1.6) | 28.9 (1.2) | .57 |
| Aβ1-42 positive*, N (%) | 4 (50) | 8 (14) | .023 |
| Aβ1-42 (pg/mL) | 1081 (463) | 1451 (313) | .060 |
| T-tau (pg/mL) | 175 (78.6) | 198 (64.6) | .44 |
| P-tau-181 (pg/mL) | 14.8 (7.52) | 16.5 (6.20) | .47 |
| NfL (pg/mL) | 552 (164) | 620 (317) | .55 |
| Average Sleep Duration, hours | 5.53 (0.54) | 7.27 (0.66) | <.001 |
| Average Onset Latency, mins | 23.2 (29.3) | 16.8 (9.8) | .56 |
| Average WASO, mins | 33.0 (19.8) | 38.1 (13.6) | .50 |
| Average Awakenings (count) | 25.8 (6.7) | 33.2 (9.2) | .017 |
| Average nap time (minutes) | 35.3 (47.1) | 8.14 (11.5) | .15 |
| Average sleep mid-point (time) | 3:25 am | 2:59 am | .27 |
| Average sleep efficiency (%) | 79.2 (7.07) | 85.7 (4.60) | .036 |
| Sleep regularity index | 73.2 (8.35) | 75.8 (7.05) | .43 |
All values are presented as mean (SD) unless stated otherwise, highest sample sizes reported. NfL = neurofilament light. MMSE = Mini Mental State Examination; CDR-SOB = Dementia Rating Scale Sum of Boxes Score; WASO = Wake after sleep onset.
*Persons with CSF amyloid levels <1000 pg/L/
†Welch t-test for continuous variables and chi-squared test for categorical variables.
‡APOE4-ε4 genotype unavailable in n = 3.
Short sleepers had lower CSF Aβ42 (M ± SD = 1081 ± 463 pg/mL) than those who slept >6 h (M ± SD = 1451 ± 313 pg/mL). In a Tobit regression model, suitable for outcome variables with censoring due to detection limits, the between-group difference was estimated as −440 pg/mL (95% CI = −795, −85.4; p = .015), adjusted for age, sex, and BMI (Figure 1). This was equivalent to a Cohen’s d of −0.99 on the (latent) uncensored outcome scale, a large effect. For clinical context, some studies have reported a smaller group difference in CSF Aβ42 measured with Elecsys when comparing participants with MCI to unimpaired controls [12].
Figure 1.
Biomarkers of Alzheimer’s disease by Sleep Duration. (A) Association of sleep duration with CSF Aβ1-42; (B) Association of sleep duration with CSF t-tau; (C) Association of sleep duration with CSF p-tau; (D) Association of sleep duration with CSF NfL. Models are adjusted for age, sex, and BMI. Left panels show estimated mean AD biomarkers in short sleepers (<6 h) and normal sleepers (>6 and <9 h). Right panels show estimated relationship between sleep duration, modeled as a continuous variable using a restricted cubic spline, and AD biomarkers. In all plots, model-estimated biomarker levels are calculated for females at the sample mean age (58 years). Error bars and gray ribbon represent 95% confidence interval for left and right panels, respectively. CSF Aβ1-42 is on the latent (uncensored) outcome scale.
In post hoc age, sex, and BMI-adjusted Tobit models, total sleep time was modeled as a continuous rather than binary variable. A restricted cubic spline, with knots placed at the 5th, 35th, 65th, and 95th quantiles, was used to allow for non-linearity in the relationship between sleep time and Aβ42 levels. Aβ42 levels were lowest amongst short sleepers, with such levels increasing to and peaking at around 7 h before decreasing as sleep duration increased to 9 h (Figure 1). Using this model, estimated mean Aβ42 levels were 361 pg/L lower (bootstrapped 95% CI = −781, −4.40) for the median short sleepers (5.8 h) compared with the median normal sleepers (7.2 h). There was, however, only weak evidence that sleep duration was associated with Aβ42 in the global test of the spline term (χ² (3) = 7.6, p = .085).
Aside from Aβ42, there was no evidence that other AD biomarkers that typically become abnormal later in the AD disease course differed by sleep duration. In ANCOVA models adjusting for age, sex, and BMI, there was no evidence of difference between short and normal sleepers in levels of t-tau (MD = 4.71, 95% CI = −43.1, 52.5; p = .84), p-tau (MD = 0.72, 95% CI = −3.96, 5.40; p = .76), or NfL (MD = −20.7, 95% CI = −188, 229; p = .84). Results were not meaningfully different in models including spline terms for total sleep time (Figure 1 and Table 2). Results were similar after additional adjustments APOE ε4 status (Table 2). The relationship between short sleep and lower Aβ42 strengthened after additional adjustments for average sleep efficiency, daytime nap duration, and SRI, in addition to age, sex, BMI, and APOE ε4 status (Table 2).
Table 2.
Association between sleep duration and Alzheimer’s disease biomarkers
| Outcome | Total sleep time as continuous variable | Short sleep as binary variable (≤6 h vs. >6 h) | ||
|---|---|---|---|---|
| Estimate (95% CI)* | p † | Estimate (95% CI) | p | |
| Aβ42 (pg/mL)‡ | ||||
| Model 1 | −361 (−781, −4.40) | .085 | −440 (−795, −85.4) | .015 |
| Model 2 | −333 (−750, 50.1) | .099 | −394 (−755, −32.7) | .033 |
| Model 3 | −449 (−1000, 25.3) | .055 | −554 (−946, −163) | .006 |
| t-tau (pg/mL) | ||||
| Model 1 | 6.50 (−34.5, 47.5) | .97 | 4.71 (−43.1, 52.5) | .84 |
| Model 2 | 3.05 (−38.0, 44.1) | .99 | −0.92 (−49.0, 47.2) | .97 |
| Model 3 | 2.68 (−54.7, 60.1) | .93 | 5.00 (−45.8, 55.8) | .99 |
| p-tau181 (pg/mL) | ||||
| Model 1 | 0.93 (−3.07, 4.94) | .91 | 0.72 (−3.96, 5.40) | .76 |
| Model 2 | 0.56 (−3.44, 4.56) | .95 | 0.11 (−4.58, 4.81) | .96 |
| Model 3 | 1.04 (−3.90, 5.99) | .95 | 0.85 (−4.75, 6.44) | .76 |
| NfL (pg/mL) | ||||
| Model 1 | 143 (−31.4, 317) | .34 | 20.7 (−188, 229) | .84 |
| Model 2 | 125 (−49.1, 299) | .30 | −7.92 (−218, 202) | .94 |
| Model 3 | 6.32 (−199, 211) | .067 | −130 (−372, 112) | .29 |
When treated as a continuous variable, total sleep time is modeled using a 3 degree of freedom restricted cubic spline with knots placed at the 5th, 35th, 65th, and 95th percentiles.
Model 1: Adjusted for age, sex, and BMI;
Model 2: Adjusted for age, sex, BMI, and APOE-e4;
Model 3: Adjusted for age, sex, BMI, APOE-e4, daily average nap time, sleep regularity index, and average sleep efficiency (total sleep time/ time in bed).
* As nonlinear effects for total sleep time are estimated, contrasts are between mean biomarker levels between those with representative short sleep (5.8 h) and those with representative normal sleep (7.2 h). The estimate thus indicates the mean change in biomarker concentration for decreasing sleep duration from 7.2 to 5.8 h.
† Global test of the spline term with the likelihood ratio test.
‡ Estimates are on latent (without censoring due to detection limit) scale.
The results of this preliminary study indicate that, in non-demented adults, short sleep was associated with lower CSF Aβ42. These results contrast with a previous study by Ju et al. [13] in which sleep time was not significantly associated with CSF Aβ42 amongst 145 individuals aged 45–75 years. However, in that previous study [13], total sleep time was treated as a continuous variable and CSF Aβ42 as a discrete variable. We demonstrated that the association between sleep time and amyloid was nonlinear and this may explain the contrasting findings. Potentially, a reduced opportunity for the glymphatic clearance of cerebral Aβ, which is more active during slow-wave sleep [3], may lead to an aggregation of cerebral Aβ over time. Alternatively, as the genetic risk of AD is associated with short sleep [5] genetic factors may cause both short sleep and Aβ accumulation. As the causes of short sleep duration are unclear, we cannot exclude that the association between sleep time and CSF Aβ42 was due to uncontrolled bias such as confounding.
Our analysis also revealed that CSF t-tau, p-tau, and NfL did not differ by sleep duration. Declining Aβ levels in the CSF precede other AD-related changes, including changes in CSF tau and NfL [3]. It is possible that associations between sleep time, tau, and neurodegeneration may emerge in older adults with more advanced AD pathology. The small sample size of our study also means that we may have been underpowered to detect subtle differences in these biomarker levels. Additionally, due to the cross-sectional design, we were unable to determine if short sleep was associated with the progression of AD biomarkers. Larger prospective studies are needed to confirm our findings and determine if sleep duration is related to AD progression.
Acknowledgments
We thank all HBP participants for their commitment and dedication to helping advance research into the early detection and causation of dementia.
Contributor Information
Madeline Gibson, The Turner Institute for Brain and Mental Health, Monash University, Clayton, Victoria, Australia.
Jessica Nicolazzo, The Turner Institute for Brain and Mental Health, Monash University, Clayton, Victoria, Australia.
Marina Cavuoto, The Turner Institute for Brain and Mental Health, Monash University, Clayton, Victoria, Australia.
Ella Rowsthorn, The Turner Institute for Brain and Mental Health, Monash University, Clayton, Victoria, Australia.
Lachlan Cribb, The Turner Institute for Brain and Mental Health, Monash University, Clayton, Victoria, Australia.
Lisa Bransby, The Turner Institute for Brain and Mental Health, Monash University, Clayton, Victoria, Australia.
Rachel Buckley, Melbourne School of Psychological Sciences, University of Melbourne, Parkville, Victoria, Australia; Department of Neurology, Massachusetts General Hospital and Harvard Medical School, Boston, MA, USA; Center for Alzheimer Research and Treatment, Department of Neurology, Brigham and Women’s Hospital, Boston, MA, USA.
Nawaf Yassi, Department of Medicine and Neurology, Melbourne Brain Centre at The Royal Melbourne Hospital, University of Melbourne, Parkville, Victoria, Australia; Population Health and Immunity Division, The Walter and Eliza Hall Institute of Medical Research, Parkville, Victoria, Australia.
Stephanie Yiallourou, The Turner Institute for Brain and Mental Health, Monash University, Clayton, Victoria, Australia.
Amy Brodtmann, Department of Medicine and Neurology, Melbourne Brain Centre at The Royal Melbourne Hospital, University of Melbourne, Parkville, Victoria, Australia; Department of Neurology, Austin Health, Melbourne, Victoria, Australia; Eastern Cognitive Disorders Clinic, Eastern Health, Monash University, Clayton, Victoria, Australia.
Dennis Velakoulis, Neuropsychiatry at The Royal Melbourne Hospital, Parkville, Victoria, Australia.
Dhamidhu Eratne, Neuropsychiatry at The Royal Melbourne Hospital, Parkville, Victoria, Australia.
Garun S Hamilton, Monash Lung, Sleep, Allergy and Immunology, Monash Health, Clayton, Victoria, Australia; School of Clinical Sciences, Monash University, Clayton, Victoria, Australia.
Matthew T Naughton, Department of Respiratory Medicine, Alfred Health and Central Clinical School, Melbourne, Victoria, Australia.
Yen Ying Lim, The Turner Institute for Brain and Mental Health, Monash University, Clayton, Victoria, Australia.
Matthew P Pase, The Turner Institute for Brain and Mental Health, Monash University, Clayton, Victoria, Australia; Harvard T.H. Chan School of Public Health, Boston, MA, USA.
Funding
Dr Pase is supported by a National Heart Foundation of Australia Future Leader Fellowship (GTN102052) with sleep and dementia research funding from the National Health and Medical Research Council of Australia (GTN2009264; GTN1158384), National Institute on Aging (R01 AG062531-01A1), and Alzheimer’s Association (2018-AARG-591358). Dr Cavuoto and Dr Pase are supported by a Dementia Australia Research Foundation award (Lucas’ Papaw Remedies Project Grant). The Healthy Brain Project (healthybrainproject.org.au) is funded by the National Health and Medical Research Council (NHMRC; GNT1158384, GNT1147465, GNT1111603, GNT1105576, GNT1104273, GNT1158384, and GNT1171816), the Alzheimer’s Association (AARG-17-591424, AARG-18-591358, and AARG-19-643133), the Dementia Australia Research Foundation, the Yulgilbar Alzheimer’s Research Program, and the Charleston Conference for Alzheimer’s Disease. Dr Lim is supported by an NHMRC Career Development Fellowship (GNT1162645). Dr Buckley is supported by a National Institutes of Health K99-R00 award (K99AG061238) and an Alzheimer’s Association Research Fellowship (AARF-20-675646).
Disclosure Statement
None declared.
Data Availability
The data supporting the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
References
- 1. Westwood AJ, et al. Prolonged sleep duration as a marker of early neurodegeneration predicting incident dementia. Neurology. 2017;88(12):1172–1179. doi: 10.1212/wnl.0000000000003732. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Sabia S, et al. Association of sleep duration in middle and old age with incidence of dementia. Nat Commun. 2021;12(1):2289. doi: 10.1038/s41467-021-22354-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Selkoe DJ, et al. The amyloid hypothesis of Alzheimer’s disease at 25 years. EMBO Mol Med. 2016;8(6):595–608. doi: 10.15252/emmm.201606210. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Wang Q, et al. Short sleep duration is associated with hypertension risk among adults: a systematic review and meta-analysis. Hypertens Res. 2012;35(10):1012–1018. doi: 10.1038/hr.2012.91. [DOI] [PubMed] [Google Scholar]
- 5. Leng Y, et al. Genetic risk of Alzheimer’s disease and sleep duration in non-demented elders HHS Public access author manuscript. Ann Neurol. 2021;89(1):177–181. doi: 10.1002/ana.25910. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Winer JR, et al. Association of short and long sleep duration with Amyloid-β burden and cognition in aging. JAMA Neurol. 2021;78(10):1187–1196. doi: 10.1001/jamaneurol.2021.2876. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Wattmo C, et al. Cerebro-spinal fluid biomarker levels: Phosphorylated tau (T) and total tau (N) as markers for rate of progression in Alzheimer’s disease. BMC Neurol. 2020;20(1):1–12. doi:10.1186/s12883-019-1591-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Lim YY, et al. The healthy brain project: an online platform for the recruitment, assessment, and monitoring of middle-aged adults at risk of developing Alzheimer’s disease. J Alzheimers Dis. 2019;68(3):1211–1228. doi: 10.3233/jad-181139. [DOI] [PubMed] [Google Scholar]
- 9. Phillips AJK, et al. Irregular sleep/wake patterns are associated with poorer academic performance and delayed circadian and sleep/wake timing. Sci Rep. 2017;7(1):3216. doi: 10.1038/s41598-017-03171-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. van Buuren S, et al. Journal of Statistical Software mice: Multivariate Imputation by Chained Equations in R, vol. 45. 2011; http://www.jstatsoft.org/. Accessed 1 December 2022. [Google Scholar]
- 11. Roche Diagnostics, Roche. Elecsys ® β-Amyloid (1-42) CSF: ElectroChemiLuminescence Immunoassay (ECLIA) for the in Vitro Quantitative Determination of ß-Amyloid (1-42) in Human Cerebrospinal Fluid (CSF); 2020. [Google Scholar]
- 12. van Harten AC, et al. Detection of Alzheimer’s disease amyloid beta 1-42, p-tau, and t-tau assays. Alzheimers Dement. 2022;18(4):635–644. doi: 10.1002/ALZ.12406. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Ju Y, et al. Sleep quality and preclinical Alzheimer disease. JAMA Neurol. 2013;70(5):587–593. doi:10.1001/jamaneurol.2013.2334. [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.
Data Availability Statement
The data supporting the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.