Associations among sleep quality, sleep duration, and Alzheimer's disease biomarkers: A systematic review and meta‐analysis

Abstract

INTRODUCTION

Although sleep disturbances are widely recognized as risk factors for cognitive decline and Alzheimer's disease (AD), their influence on AD biomarkers remains unclear. This study aimed to clarify whether sleep quality or sleep duration affect amyloid beta (Aβ) and tau levels in plasma, cerebrospinal fluid (CSF), and positron emission tomography (PET) in non‐demented populations.

METHODS

PubMed, Web of Science, and Embase were systematically searched up to February 2025.

RESULTS

In total, 30 studies were included comprising 14,997 subjects. Individuals with poor sleep quality exhibited greater PET Aβ burden and higher Aβ42 levels in plasma than those with good sleep quality. Shorter sleep duration was associated with higher Aβ burden on PET. However, no association between either sleep quality or sleep duration and tau levels was found.

DISCUSSION

Sleep may be a modifiable marker of early AD management by modulating Aβ levels.

Highlights

• lPoor sleep quality and shorter sleep duration were significantly associated with higher amyloid beta (Aβ) burden detected by positron emission tomography (PET) in non‐demented populations. Poor sleep quality was also associated with elevated Aβ42 levels in plasma.

• lNo significant associations were found between sleep quality or sleep duration and tau levels in plasma, cerebrospinal fluid, or PET.

• lInterventions targeting sleep could serve as a viable and low‐cost prevention strategy for early management of Alzheimer's disease.

1. BACKGROUND

Dementia has emerged as one of the most significant global public health challenges of the generation, generating an enormous economic burden worldwide. ¹ Alzheimer's disease (AD), the most common type of dementia, is characterized by the extracellular accumulation of amyloid beta (Aβ) plaques and intracellular hyperphosphorylation of tau protein. ² Moreover, Aβ and tau are crucial biomarkers for predicting the occurrence and progression of AD.

Approximately one third of AD cases are attributed to modifiable and treatable risk factors, underscoring the importance of early prevention. ³ , ⁴ Among these factors, sleep disturbances have become an increasingly crucial aspect of AD prevention. ⁵ A study involving 1.1 million people indicated that almost one fourth of the people sleep less than the recommended duration for their specific age group, and ≈ 15% to 20% of older adults suffer from different insomnia symptoms. ⁶

Patients with sleep disturbances have a 1.19 times higher risk of developing all‐cause dementia and a 1.65 times higher risk of cognitive impairment than those without. ⁷ , ⁸ Aβ pathology begins a decade or more before the appearance of clinical symptoms. Similarly, sleep disruption emerges before the cognitive symptoms of AD manifest and progresses alongside cognitive dysfunction and the development of AD, serving as an indicator of a more rapid cognition decline. ⁹ Additionally, AD patients with severe cognitive symptoms exhibit greater sleep disruption than those with milder cognitive symptoms. ¹⁰

While sleep disturbances are recognized as risk factors for dementia, conclusions regarding their impact on different biomarkers of AD remain inconsistent. Some studies suggest that sleep disturbances affect Aβ and tau levels. ¹¹ , ¹² , ¹³ , ¹⁴ However, there is also evidence that doubts the association between sleep and Aβ ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ or tau ¹⁵ , ¹⁶ , ¹⁷ , ¹⁹ , ²⁰ levels. Therefore, this study examined the association of sleep quality and sleep duration with Aβ and tau in plasma and cerebrospinal fluid (CSF), and on positron emission tomography (PET). Focusing on early prevention and considering the potential influence of cognitive status on sleep conditions, this meta‐analysis was conducted only for adults without cognitive impairment. Our research may provide valuable insights for developing public health strategies aimed at early AD prevention through public sleep interventions.

2. METHODS

2.1. Selection criteria and search strategy

A systematic review and meta‐analysis were performed following the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses guidelines. ²¹ The study protocol for this review was registered in the International Prospective Register of Systematic Reviews (CRD42024499506). Two reviewers (C.L.C. and M.Y.Z.) systematically searched PubMed, Web of Science, and Embase up to February 2025 for studies reporting associations of sleep with Aβ and tau. Disagreements between the two reviewers were resolved by a third researcher (L.S.). The following search terms were used: “sleep” AND “Alzheimer's disease” OR “amyloid β” OR “tau” (see Table S1 in supporting information for the full strategy). We also searched reference lists of included and review articles for additional studies to include.

The inclusion criteria were as follows: (1) observational studies reporting that sleep quality or duration affects Aβ and tau levels in plasma, CSF, or detected by PET; (2) peer‐reviewed full‐text studies; and (3) English publications.

The exclusion criteria were as follows: (1) animal or in vitro studies; (2) case reports; (3) reviews; (4) pediatric studies; (5) studies only involving participants with cognitive impairment, dementia, or diseases associated with the levels of Aβ and tau, such as Parkinson's disease and fibromyalgia; ²² , ²³ , ²⁴ , ²⁵ , ²⁶ and (6) studies involving individuals with certain sleep disorders, such as obstructive sleep apnea (OSA), narcolepsy, and sleep movement disorders.

2.2. Definition of sleep duration, sleep quality, and biomarkers categories

For sleep duration, we considered both instrument‐measured and self‐reported sleep duration. Subjective assessment tools, such as the Pittsburgh Sleep Quality Index and the Insomnia Severity Index, were used to assess sleep quality, along with measures of sleep onset latency or sleep continuity (sleep efficiency, wake after sleep onset, and wakefulness). ²⁷ The primary outcome parameters of interest were Aβ and tau in blood, CSF, and PET, including Aβ40, Aβ42, Aβ42/Aβ40 ratio, total Aβ, phosphorylated tau (p‐tau), total tau (t‐tau), p‐tau/Aβ42, t‐tau/Aβ42, as well as Aβ and tau pathology detected by PET.

2.3. Assessment of bias in individual studies

Bias risk within individual studies was independently evaluated by two researchers (C.L.C. and M.Y.Z.). The Agency for Healthcare Research and Quality checklist was used to evaluate the quality of cross‐sectional studies. ²⁸ The checklist comprises 11 items, with each having three responses, “yes,” “no,” or “unclear.” Based on the number of “yes” responses, the articles were categorized as low quality (< 5 “yes” responses), medium quality (5–8 “yes” responses), or high quality (> 8 “yes” responses). The Newcastle–Ottawa scale was used for longitudinal studies with a maximum of nine stars, ²⁹ assessing the quality based on the selection of population, comparability of groups, and ascertainment of outcome.

2.4. Data extraction

Two researchers (C.L.C. and M.Y.Z.) independently extracted relevant data from the included studies. Discrepancies were resolved through discussion until a consensus was reached. The following study characteristics were extracted: (1) first author, (2) journal name and year published, (3) country or continent, (4) study center or hospital, (5) sample size, (6) mean age, (7) sex distribution, (8) body mass index, (9) education, and (10) covariates (Table 1). In longitudinal studies, only cross‐sectional or baseline data were extracted.

TABLE 1.

Characteristics of included studies.

First author & Publication (year)	Study center	Country/continent	Sample size	Age (mean ±  SD in years)	Sex/gender (% women)	Education (mean ±  SD in years)	APOE ε4 carrier (%)	BMI (mean ±  SD)	Sleep method	Sleep measure	Biomarker method	Biomarker measure	Adjustment
Ju et al., 2013 40	Washington University Knight Alzheimer Disease Research Center (ADRC)	America	142	65.6 ± 8.2	59.2	NA	36.6	NA	Actigraphy for 2 weeks	Sleep efficiency	CSF: ELISA	CSF: Aβ42	Age, sex, APOE ε4 allele
Spira et al., 2013 50	The Baltimore Longitudinal Study of Aging (BLSA)	America	66	78.2 ± 7.9	47.1	16.8 ± 2.3	28.6	27.2 ± 4.1	Standardized interview for sleep duration and the five‐item Women's Health Initiative Insomnia Rating Scale (WHIIRS)	Sleep duration and WHIIRS total	PET PiB: ref. cerebellar gray matter Region: the superior, middle and inferior frontal and orbitofrontal, superior parietal, supramarginal and angular gyrus regions, precuneus, superior, middle and inferior occipital, superior, middle and inferior temporal, anterior, middle and posterior cingulate regions	PET: Aβ burden DVR	Age, sex, race, center for epidemiological survey‐depression scale score, BMI, APOE ε4, history of cardiovascular or pulmonary disease, and current sleep medication use
Branger et al., 2016 51	The Multimodal Imaging of Early‐Stage AD study	France	51	64.1 ± 10.6	54.9	12.4 ± 3.7	27	24.5 ± 2.9	Sleep questionnaire	Self‐reported sleep latency(mins)	PET FBP: ref. cerebellar gray matter Region: prefrontal areas that included the anterior cingulate cortex	PET: Aβ burden SUVR	Age, APOE ε4, BMI, anxiety, and depression scores
Brown et al., 2016 52	The Australian Imaging, Biomarkers and Lifestyle (AIBL) study of aging	Australia	184	75.5 ± 6.1	58.7	>12:113 (61.4%)	22.82	26.4 ± 4.4	PSQI	Global PSQI Score and Sleep duration	PET PiB(N = 68): ref. the cerebellar cortex; PET FMM(N = 58), PET FBP(N = 58): ref. the whole cerebellum and the pons Region: the frontal, superior parietal, lateral temporal, occipital, and anterior and posterior cingulate regions	PET: Aβ burden SUVR	Age, BMI, CVD risk, GDS and APOE ε4
Sprecher et al., 2017 41	The Wisconsin Registry for Alzheimer's Prevention (WRAP)	America	101	63.76 ± 6.18	65.30	16.6 ± 2.8	29.7	28.92 ± 6.2	Medical Outcomes Study (MOS)	Sleep adequacy and sleep problems	CSF: ELISA	CSF: Aβ42, Aβ42/Aβ40, p‐tau, t‐tau, p‐tau/Aβ42, t‐tau/Aβ42	Sex, age at CSF sample, CSF assay batch and time between CSF sample and sleep assessment
Chen et al., 2018 20	Daping Hospital clinical	China	46 (23 insomniacs and 23 controls)	Insomnia patients: 53.96  ±  5.60 Normal control: 52.61  ± 5.63	65.22	Insomnia patients: 12.65 ± 3.94 NC: 12.57 ± 3.91	0	NA	PSQI; Insomnia DSM‐IV‐R criteria	PSQI scores	CSF: ELISA	CSF: Aβ40, Aβ42, p‐tau, and t‐tau	NA
Hwang et al., 2018 53	The Korean Brain Aging Study for the Early Diagnosis and Prediction of Alzheimer's Disease (KBASE)	Korea	133	68.05 ± 7.68	53.4	12.05 ± 4.56	18.8	24 ± 3.15	Actigraphy for 8 consecutive days	Sleep latency, sleep efficiency, WASO, Total sleep time	PET PIB: ref. cerebellar  Region: the frontal, lateral parietal, posterior cingulate‐precuneus, and lateral temporal regions	PET: Aβ burden SUVR	Age, sex, APOE ε4, and GDS score
Ettore et al., 2019 54	The InveStIGation of Alzheimer's Predictors in Subjective Memory Complainers (INSIGHT‐pre AD) cohort	France	68	76.7 ± 3.52	70.6	NA	NA	NA	Actigraphy	Sleep latency, sleep efficiency, WASO, awakening, average awakenings(mins), total sleep time (min)	PET FBP: ref. in pons and whole cerebellum Region: posterior cingulum, anterior cingulum, orbito‐frontal cortex, precuneus, temporal, and parietal lobe	PET: Aβ burden SUVR	Age, sex, MMSE, and depression
Kam et al., 2019 42	NIH‐supported longitudinal studies on normal aging and biomarkers of AD at NYU	America	50	67.2  ±  7.3	54	16.7  ±  2.1	34	25.4  ±  3.5	In‐lab nocturnal polysomnography (NPSG) and actigraphy for seven consecutive days	WASO, sleep efficiency and total sleep time (habitual)	CSF: ELISA	CSF: Aβ42, t‐tau, p‐tau and t‐tau/Aβ42	Age, sex and APOE ε4 status
Shokouhi 2019 55	The Alzheimer's Disease Neuroimaging Initiative (ADNI) database	America	35	77 ± 7	65.71	NA	28	NA	NPI‐sleep	NPI sleep quality	PET FBP: ref. whole cerebellum; Region: posterior cingulate, precuneus, medial orbitofrontal and cortical regions (frontal, anterior/posterior cingulate, lateral parietal, lateral temporal) PET FTP: ref. cerebellar gray matter Region: left and right entorhinal and brainstem	PET: Aβ burden SUVR and tau burden SUVR	Age, sex, and APOE
Gao et al., 2020 36	The cluster sampling method for register villagers over 40 years old in the Qubao village near Xi'an	China	1459	57.4 ±  9.7	59	7.0 ± 4.9	12.6	25.3 ±  3.5	PSQI	PSQI total score;	Plasma: ELISA	Plasma: Aβ40, Aβ42 and Aβ42/Aβ40	Age, sex, years of education, smoking, drinking, history of hypertension, history of heart disease, cerebrovascular disease, BMI, FBG, TC, TG, HDL‐c, and LDL‐c
Lysen et al., 2020 17	Rotterdam Study cohort	Netherlands	4712 Actigraphy:849	71.1 (IQR 66.1–77.2) Actigraphy: 66.7 (IQR 63.7‐73.1)	57 Actigraphy: 51	Primary education: 536 (11%), 53 (6%) (actigraphy) Lower/intermediate or lower vocational: 2088 (44%), 368 (43%) (actigraphy) Higher or intermediate vocational: 1436 (31%), 281 (33%) (actigraphy) Higher vocational or university: 652 (14%), 147 (18%) (actigraphy)	NA	27.6  ±  4.1 Actigraphy: 27.9  ±  4.0	PSQI and actigraphy for seven days	Self‐rated: PSQI score and sleep duration Actigraphy: sleep onset latency, WASO, sleep efficiency	Plasma: Simoa	Plasma: Aβ40, Aβ42 and t‐tau	Age and sex, educational level, batch, time interval between measurement of sleep and biomarkers, alcohol consumption, employment status, smoking status, body mass index, presence of hypertension, presence of diabetes mellitus, total serum cholesterol level, history of cardiovascular disease, and possible sleep apnea
Xu et al., 2020 19	The Chinese Alzheimer's Biomarker and LifestylE (CABLE) study	China	736	62.3 ± 10.5	58.6	9.8 ± 4.3	14	NA	PSQI	PSQI score	CSF: ELISA	CSF: Aβ42, Aβ42/Aβ40, p‐tau, t‐tau, p‐tau/Aβ42 and t‐tau/Aβ42	Age, sex, education, APOE ε4 status and MMSE score.
López‐García et al., 2021 43	The Valdecilla Study for Memory and Brain Aging	Spain	127	65.47  ±  6.33	70.1	Tertiary no. (%): 40 (34.5) Secondary no. (%): 45 (38.8) Primary no. (%): 31 (26.7)	30.7	NA	Actigraphy for seven nights and Oviedo Sleep questionnaire	Total sleep time and Oviedo Sleep questionnaire (Subjective sleep quality impression and Insomnia score)	CSF: ELISA	CSF: Aβ42/Aβ40, p‐tau, and t‐tau	Age, sex, and APOE ε4 status
Winer et al., 2021 13	The Anti‐Amyloid Treatment in Asymptomatic Alzheimer's Disease (A4) study	America, Canada, Australia, and Japan	4417	71.3 ± 4.7	59.2	16.6 ± 2.8	32.3	27.5 ± 5.1	Sleep questionnaire	Hours of sleep per night	PET FBP: ref. whole cerebellum  Region: global cortical	PET: Aβ burden SUVR	Age, sex, self‐identified race/ethnicity, years of education, number of APOE ε2 alleles, number of APOE ε4 alleles and cognitive performance
Blackman et al., 2022 44	The European Prevention of Alzheimer's Dementia (EPAD) Longitudinal Cohort Study	Europe	1168	64.7  ±  7.1	58.1	14.8 ± 3.5	36.8	26.3 ±  4.4	PSQI	PSQI score	CSF: ECLIA	CSF: Aβ42, p‐tau, and t‐tau	Age, sex, site of data collection, APOE ε4 carriership, anxiety (State‐Trait Anxiety Inventory) and log (CSF p‐tau) levels/log (CSF Aβ42)
Liu et al., 2022 37	Cognitive Disorders Clinics in the First People's Hospital of Foshan and communities	China	305	69.07 ± 6.37	59.67	0, n (%): 10 (3.28%) 1–6, n (%): 90 (29.51%) ≥7, n (%): 205 (67.21%)	14.10%	23.41 ± 2.98	PSQI	Sleep efficiency and sleep duration	Plasma: ELISA	Plasma: Aβ40, Aβ42 and Aβ42/Aβ40	Age, sex, education, APOE ε4, body mass index, exercise frequency, diabetes, hypertension, triglyceride, fasting blood glucose, mini‐mental state examination, and the geriatric depression scale
Naismith et al., 2022 45	The European Prevention of Alzheimer's Disease (EPAD) Longitudinal Cohort Study	Europe	1240	65.34  ±  7.11	56.5	NA	35.7	26.22 ±  4.30	PSQI	PSQI score and sleep duration	CSF: ECLIA	CSF: p‐tau/Aβ42	Age, sex, depressive symptoms, APOE ε4, vascular risk, hippocampal volume, and WMH volume
Aslanyan et al., 2023 56	The Anti‐Amyloid Treatment in Asymptomatic Alzheimer's Disease (A4) study	America, Canada, Australia and Japan	4322 (FTP N = 443)	71.3 ± 4.7	59.5	16.58 ± 2.83	33.2	NA	Self‐report questionnaires	Night‐time sleep duration	PET FBP: ref. cerebellum Region: medial orbitofrontal cortex, anterior cingulate cortex, posterior cingulate cortex, and precuneus PET FTP: ref. cerebellar gray matter Region: entorhinal cortex, amygdala, parahippocampus, fusiform gyrus, and inferior and middle temporal gyri	PET: Aβ burden SUVR and tau burden SUVR	Age, sex, education, and APOE ε4
Chu et al., 2023 38	The Shanghai Sixth People's Hospital	China	335	64.4 ± 7.8	62.39	IQR: 12 (10–15)	19.4	23.73  ±  3.17	PSQI	Sleep duration, sleep disturbance and PSQI score	Plasma: Simoa; PET FBP: ref. the cerebellar crus Region: frontal gyrus, lateral parietal gyrus, lateral temporal gyrus, medial temporal gyrus, posterior cingulate gyrus, and precuneus	Plasma: Aβ40, Aβ42, Aβ42/Aβ40, p‐tau, and t‐tau; PET: Aβ burden SUVR	Age, sex, education, BMI, APOE ε4 positive
Cook et al., 2023 16	The African Americans Fighting Alzheimer's in Midlife (AA‐FAiM) study	America	147	63.2  ±  8.51	69.4	14.9 ± 2.6	41.5	33.3 ±  7.6	The Medical Outcomes Study Sleep Scale	Sleep duration	Plasma: liquid chromatography‐tandem mass spectrometry (LC‐MS/MS)	Plasma: Aβ42, and Aβ42/Aβ40	Age at visit, self‐identified gender, years of education, baseline body mass index, and APOE ε4 allele count
Du et al., 2023 15	The Wisconsin Registry for Alzheimer's Prevention (WRAP)	America	220	63.6 ± 5.6	66	16.2 ± 2.1	38.6	28.6 ± 5.7	The Medical Outcomes Study (MOS) Sleep Scale	Sleep hours, the Sleep Problems Index and sleep adequacy	PET PIB: ref. cerebellar cortex Region: global cortical	PET: Aβ burden DVR	Age, sex, APOE ε4, family history of AD, and body mass index
Gibson et al., 2023 46	The Healthy Brain Project (HBP)	Australia	N = 66 Short sleep: N = 8 Normal sleep: N = 58	Short(N = 8): 55.7  ± 5.8 Normal(N = 58): 58.9  ±  6.9	66	Short N = 8: 16.2 ± 3.6 Normal N = 58: 16.0 ± 3.4	38	NA	17 days of wrist‐worn actigraphy	Sleep duration	CSF: ECLIA	CSF: Aβ42, p‐tau, and t‐tau	Adjusted for age, sex, BMI, APOE ε4, daily average nap time, sleep
Nicolazzo et al., 2023 47	The Healthy Brain Project (HBP)	Australia	63	59  ±  7	67	Only participants with ISI (N = 58): 17 ± 4	43	NA	Actigraphy for 17 days (N = 63) Insomnia Severity Index (N = 58)	Actigraphy: sleep onset latency, WASO and awakenings; Insomnia Severity Index	CSF: ECLIA	CSF: Aβ42, p‐tau, t‐tau and t‐tau/Aβ42	Age, sex, and APOE ɛ4 genotype
Baril et al., 2024 57	The Framingham Heart Study Third Generation cohort at their third clinic examination (2016–2019)	America	271	53.6 ± 8.0	49.1	Unfinished high school 3 (1) High school degree 26 (10) Some college 74 (27) College graduate 167 (62)	23	26.8 ± 5.1	Question “Numbers of hours that you typically sleep.”	Sleep duration	PET PIB and FTP: ref. cerebellar cortex Region: PIB: frontal, lateral, and retrosplenial cortices; PET FTP: entorhinal and composite rhinal cortices, medial orbitofrontal gyrus, and rostral anterior cingulate cortex.	PET: Aβ burden SUVR and tau SUVR	Age, sex, time between sleep assessment and neuroimaging, PET camera, APOE ε4 carriers, depression, diabetes, hypertension, and prevalent cardiovascular disease
de Oliveira et al., 2024 48	The Behavioral Neurology Section of Hospital São Paulo	Brazil	27	78.5 ± 8.7	66.6	5.81 ± 4.72	14.8	NA	Questionnaires	Sleep duration	CSF: ELISA	CSF: Aβ42/Aβ40, p‐tau/ Aβ42 and t‐tau/Aβ42	Age at inclusion in the study, lifetime alcohol use, lifetime smoking, lifetime living with sanitation, and APOE ε4
Nguyen Ho et al., 2024 58	Rotterdam Study cohort	Netherlands	319	69.24 ± 5.25	47.0	Primary: 21 (6.6%) Lower: 89 (27.9%) Intermediate: 87 (27.3%) Higher: 122 (38.2%)	28.2	27.31 ± 3.87	Actiwatch 4 (Cambridge Technology) or aGeneActiv (Activinsights) for 7 consecutive days and nights; Sleep diary	Total sleep time, sleep latency, WASO, sleep efficiency and sleep quality (diary)	PET: FBP: ref. the cerebellar PET Region: frontal, cingulate, lateral parietal, and lateral temporal regions	PET: Aβ burden SUVR	Age, sex, carrying at least 1 APOE ɛ4, time between PET and actigraphy, type of actigraph, education, self‐reported possible sleep apnea, self‐reported sleep medication, body mass index, hypertension, diabetes, current smoking, physical activity, depressive symptoms, and paid employment status
Rosenblum et al., 2024 39	The Max Planck Institute of Psychiatry	German	60	39.80 ± 16.05	53.0	NA	NA	NA	Polysomnography	WASO	Plasma: ELISA	Plasma: Aβ40 and Aβ42	Age
Stankeviciute et al., 2024 59	The Harvard Aging Brain Study (HABS)	America	39	74.6 ± 8.69	61.5	16.5 ± 2.75	NA	NA	PSG	Total sleep time, sleep efficiency and WASO	PET FTP: ref.  cerebellar gray region; Region: the bilateral entorhinal cortex, inferior temporal cortex, middle temporal region, and fusiform regions	PET: tau burden SUVR	Age and sex
Zhang et al., 2024 49	Knee replacement surgery participants	China	64	61.6 ± 8.5	75.0	Primary school: 23 (35.9%) Junior high school: 17 (26.6%) Senior high school and above: 24 (37.5%)	NA	18.5‐24:17(26.6%); 25‐28: 30(46.9%); ≥28: 17(26.6%)	PSQI	PSQI score	CSF: ELISA	CSF: Aβ42	Age and BMI

Abbreviations: Aβ, amyloid beta; AD, Alzheimer's disease; AHI, apnea–hypopnea index; APOE, apolipoprotein E; BMI, body mass index; CDR, Clinical Dementia Rating; CSF, cerebrospinal fluid; CVD, cardiovascular disease; DSM‐IV, Diagnostic and Statistical Manual of Mental Disorders, 4th edition; DVR, distribution volume ratio; ECG, electrocardiogram; ECLIA, enzyme‐linked immunoculture assay; ELISA, enzyme‐linked immunosorbent assay; FBB, (18)F‐florbetaben; FBG, fast blood glucose; FBP, (18)F‐florbetapir; FTP, (18)F‐flortaucipir; GDS, Geriatric Depression Scale; HDL‐c, high‐density lipoprotein; HDRS, Hamilton Depression Rating Scale; IQR, interquartile range; LDL‐c, low‐density lipoprotein; MMSE, Mini‐Mental State Examination; NC, normal control; NIH, National Institutes of Health; NPI, Neuropsychiatric Inventory; PET, positron emission tomography; PiB, (11)C‐Pittsburgh compound B; PLMI, periodic limb movement index; PSQI, Pittsburgh Sleep Quality Index; p‐tau, phosphorylated tau; Simoa, single‐molecule array assay; SUVR, standard uptake ratio; TC, total cholesterol; TG, triglyceride; t‐tau, total tau; VRS, Verbal Rating Scale; WASO, wake after sleep onset; WHIIRS, Women's Health Initiative Insomnia Rating Scale; WMH, white matter hyperintensities.

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed the literature using traditional sources (e.g., PubMed, Embase, and Web of Science). Most reviews and meta‐analyses focus on the relationship between sleep and cognition and dementia, but few of them mention how sleep disturbances affect the level of dementia‐related biomarkers.

2. Interpretation: Our studies suggested that non‐demented individuals with poor sleep quality and shorter sleep duration tend to have greater amyloid beta (Aβ) burden detected by positron emission tomography. Poor sleep quality was also associated with higher Aβ42 in plasma.

3. Future directions: Further studies are required to investigate the interactions and the underlying mechanisms between different sleep disorders and macro/micro sleep parameters on Alzheimer's disease (AD) biomarkers. Future work should also explore AD prevention strategies from a sleep management perspective.

2.5. Data analysis

Pearson correlation coefficient (r) was extracted to calculate the Fisher z score. When r was not available, other reported statistics, such as Spearman ρ and regression β, were collected and transformed into Fisher z score. ³⁰ Articles that cannot provide calculatable results of Fisher z score were not included in the analysis. If available, effect sizes adjusted for covariates were selected. Sleep parameters indicating better sleep conditions with higher values (e.g., sleep efficiency and sleep adequacy) were reversed in the analysis. Regarding sleep duration, only continuous data were included in the main analysis. The difference between long/short versus normal sleep duration was conducted in the subgroup analysis. A fixed‐effect meta‐analysis was conducted when one study contained multiple sleep parameters or biomarker levels (e.g., different brain regions on PET or sleep questionnaire items), while data analyses across studies applied the random‐effect model.

Heterogeneity was assessed using Cochran Q test and I ² statistic. ³¹ The I ² statistic represented the percentage of variability in effect sizes owing to actual differences among studies rather than sampling error. I ² values of 75%, 50%, and 25% corresponded to high, moderate, and low levels of heterogeneity, respectively. The statistical significance threshold was set at a p value of 0.05. The potential publication bias was tested by visual inspection of funnel plots and Egger test. ³² , ³³ Significant asymmetry in funnel plots may indicate the presence of possible publication bias. When sufficient studies for each biomarker were available (N ≥ 6), meta‐regression was conducted to assess the potential impact of age, sample size, sex, and percentage of apolipoprotein E (APOE) ε4 carriers within studies. ³⁴ Sensitivity analysis was conducted using the leave‐one‐out method by omitting one study each time to identify the influence of each study (N ≥ 4). The statistical analyses were performed using the Comprehensive Meta‐Analysis software (version 3) and the Meta package in R software (version 4.2.3). ³⁵

3. RESULTS

3.1. Search results and quality assessment

The initial search of the PubMed, Web of Science, and Embase resulted in 19,340 studies (Figure 1) as of February 8, 2025. After eliminating duplicates, 13,096 articles underwent screening based on titles and abstracts, with 145 full‐text articles retrieved for further investigation. In total, 30 studies met the inclusion criteria, with 6 studies on plasma biomarkers, ¹⁶ , ¹⁷ , ³⁶ , ³⁷ , ³⁸ , ³⁹ 12 focusing on CSF biomarker levels, ¹⁹ , ²⁰ , ⁴⁰ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ and 13 on PET biomarkers ¹³ , ¹⁵ , ³⁸ , ⁵⁰ , ⁵¹ , ⁵² , ⁵³ , ⁵⁴ , ⁵⁵ , ⁵⁶ , ⁵⁷ , ⁵⁸ , ⁵⁹ (Table 1 and Table S2 in supporting information).

FIGURE 1.

Flow diagram of the literature search.

Table 1 summarizes the main characteristics of the included studies, comprising 27 cross‐sectional and three longitudinal studies. The studies were published during 2013 through 2024, encompassing 14,997 participants in total with an average age of 67.9 years, and a female proportion of 58.3%. Additionally, 22 studies adjusted their analyses for the APOE genotype as a covariate.

The cross‐sectional studies had an average score of 8.5 ± 1.0 out of 11 on the Agency for Healthcare Research and Quality checklist (Table 2). The Newcastle–Ottawa scale used for longitudinal studies (Table 3) presented an average score of 7.7 ± 0.6, suggesting a good overall quality.

TABLE 2.

The Agency for Healthcare Research and Quality (AHRQ) checklist for cross‐sectional studies.

First author & year	1	2	3	4	5	6	7	8	9	10	11	Total
Ju et al., 2013 40	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	N	9/11
Spira et al., 2013 50	Y	Y	U	Y	Y	N	Y	Y	Y	Y	N	8/11
Branger et al., 2016 51	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	N	9/11
Brown et al., 2016 52	Y	Y	U	Y	Y	N	Y	Y	Y	Y	N	8/11
Sprecher et al., 2017 41	Y	Y	Y	Y	Y	Y	Y	Y	Y	N	N	9/11
Chen et al., 2018 20	Y	Y	Y	U	Y	N	Y	Y	Y	Y	N	8/11
Hwang et al., 2018 53	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	N	9/11
Ettore et al., 2019 54	Y	Y	U	Y	Y	N	Y	Y	Y	U	N	7/11
Kam et al., 2019 42	Y	Y	U	Y	Y	N	Y	Y	Y	Y	N	8/11
Shokouhi 2019 55	Y	Y	U	Y	Y	N	Y	Y	Y	Y	N	8/11
Gao et al., 2020 36	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	N	10/11
Lysen et al., 2020 17	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	N	9/11
Xu et al., 2020 19	Y	Y	Y	U	Y	Y	Y	Y	U	N	N	7/11
López‐García et al., 2021 43	Y	Y	U	U	Y	N	Y	Y	Y	Y	N	7/11
Winer et al., 2021 13	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	N	10/11
Liu et al., 2022 37	Y	Y	U	U	Y	N	Y	Y	Y	Y	N	7/11
Naismith et al., 2022 45	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	N	9/11
Aslanyan et al., 2023 56	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	N	10/11
Chu et al., 2023 38	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	N	9/11
Du et al., 2023 15	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	N	10/11
Gibson et al., 2023 46	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	N	9/11
Nicolazzo et al., 2023 47	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	N	9/11
de Oliveira et al., 2024 48	Y	Y	Y	Y	Y	Y	Y	Y	Y	N	N	9/11
Nguyen Ho et al., 2024 58	Y	Y	Y	Y	Y	Y	Y	Y	Y	N	N	9/11
Rosenblum et al., 2024 39	Y	Y	Y	N	N	Y	Y	Y	Y	N	N	7/11
Stankeviciute et al., 2024 59	Y	Y	Y	N	Y	Y	Y	Y	Y	N	N	8/11
Zhang et al., 2024 49	Y	Y	Y	N	Y	Y	Y	Y	Y	N	N	8/11

Note: Yes = Y; No = N; Unclear = U; Agency for Healthcare Research and Quality (AHRQ) checklist:(1) Define the source of information (survey, record review); (2) List inclusion and exclusion criteria for exposed and unexposed subjects (cases and controls) or refer to previous publications; (3) Indicate time period used for identifying patients; (4) Indicate whether subjects were consecutive if not population based; (5) Indicate if evaluators of subjective components of study were masked to other aspects of the status of the participants; (6) Describe any assessments undertaken for quality assurance purposes (e.g., test/retest of primary outcome measurements); (7) Explain any patient exclusions from analysis; (8) Describe how confounding was assessed and/or controlled; (9) If applicable, explain how missing data were handled in the analysis; (10) Summarize patient response rates and completeness of data collection; and (11) Clarify what follow‐up, if any, was expected and the percentage of patients for which incomplete data or follow‐up was obtained.

TABLE 3.

The Newcastle–Ottawa scale (NOS) for longitudinal studies.

	Selection				Comparability		Outcome
First author & year	Representative of the exposed cohort	Selection of external control	Ascertainment of exposure	Outcome of interest not present at the start of the study	Main factor	Additional factor	Assessment of outcomes	Sufficient follow‐up time	Adequacy of follow‐up	Total
Blackman et al., 2022 44	☆	☆	☆	☆	☆	☆	☆	☆		8/9
Cook et al., 2023 16	☆	☆	☆		☆	☆	☆	☆		7/9
Baril et al., 2024 57	☆	☆	☆	☆	☆	☆	☆	☆		8/9

3.2. Association of sleep quality and biomarkers

Random‐effect models were used to examine the association between sleep quality and biomarkers across 24 studies with 10,132 participants (Table 4). A total of 10 studies were included in the PET biomarker analysis, revealing a significant correlation between poor sleep quality and high Aβ burden (Fisher z = 0.153, k = 9, 95% confidence interval [CI] = 0.005 to 0.302, p = 0.042). No significant association was found between sleep quality and tau burden (Fisher z = 0.002, k = 2, 95% CI = −0.003 to 0.007, p = 0.443; Figure 2).

TABLE 4.

Association of sleep quality and Aβ and tau in non‐demented individuals.

Method	Biomarker	No. of studies	No. of participants	Fisher z	95% CI	p‐value	I 2	Q
PET	Aβ	9	1411	0.153	0.005–0.302	0.042 *	0.985	544.491
	tau	2	77	0.002	−0.003–0.007	0.443	0.740	3.853
Plasma	Aβ40	5	6871	0.028	−0.039–0.096	0.410	0.801	20.086
	Aβ42	5	6871	0.007	0.003–0.011	<0.001 ***	0.401	6.681
	Aβ42/Aβ40	3	2099	−0.025	−0.075–0.024	0.319	0.715	7.017
	t‐tau	2	5047	−0.010	−0.058–0.037	0.669	0.310	1.450
CSF	Aβ42	8	2370	0.579	−0.381–1.538	0.237	0.935	107.699
	Aβ42/Aβ40	3	964	−0.020	−0.099–0.059	0.619	0.213	2.543
	p‐tau	7	2291	0.000	−0.000–0.000	0.433	0.373	9.567
	t‐tau	7	2291	−0.018	−0.084–0.047	0.584	0.608	15.317
	p‐tau/Aβ42	3	2077	0.018	−0.022–0.058	0.378	0.835	12.143
	t‐tau/Aβ42	4	950	0.004	−0.041–0.049	0.860	0.970	99.638

Abbreviations: Aβ, amyloid beta; CI, confidence interval; CSF, cerebrospinal fluid; PET, positron emission tomography; p‐tau, phosphorylated tau; t‐tau, total tau.

^* p < 0.05.

^*** p < 0.001.

FIGURE 2.

Forest plot of meta‐analytic results on the association of sleep quality and Aβ and tau detected by PET in non‐demented individuals using the random‐effect model. The results are expressed as Fisher z score and 95% CI. Aβ, amyloid beta; CI, confidence interval; PET, positron emission tomograph.

For plasma, significant correlations were found between poor sleep quality and elevated Aβ42 levels (Fisher z = 0.007, k = 5, 95% CI = 0.003 to 0.011, p < 0.001). In contrast, Aβ40 (Fisher z = 0.028, k = 5, 95% CI = −0.039 to 0.096, p = 0.410), Aβ42/Aβ40 ratio (Fisher’ z = −0.025, k = 3, 95% CI = −0.075 to 0.024, p = 0.319), and t‐tau (Fisher z = −0.010, k = 2, 95% CI = −0.058 to 0.037, p = 0.669) showed no significant associations with sleep quality (Table 4 and Figure S1 in supporting information). For CSF, no significant correlation of sleep quality and biomarkers was found (Aβ42: Fisher z = 0.579, k = 8, 95% CI = −0.381 to 1.538, p = 0.237; Aβ42/Aβ40: Fisher z = −0.020, k = 3, 95% CI = −0.099 to 0.059, p = 0.619; p‐tau: Fisher z = 0.000, k = 7, 95% CI = −0.000 to 0.000, p = 0.433; t‐tau: Fisher z = −0.018, k = 7, 95% CI = −0.084 to 0.047, p = 0.584; p‐tau/Aβ42: Fisher z = 0.018, k = 3, 95% CI = −0.022 to 0.058, p = 0.378; t‐tau/Aβ42: Fisher z = 0.004, k = 4, 95% CI = −0.041 to 0.049, p = 0.860; Table 4 and Figure S1 in supporting information).

3.3. Association of sleep duration and biomarkers

Random‐effect models were used to examine the relationship between sleep duration and biomarkers across 16 studies involving 11,035 participants (Table 5 and Table S3 in supporting information). There were six studies on Aβ and two studies on tau pathology measured by PET for continuous sleep duration (Figure 3). Shorter sleep duration was associated with greater Aβ burden detected by PET (Fisher z = −0.002, k = 6, 95% CI = −0.003 to −0.001, p = 0.002). No association was found between PET tau and sleep duration (Fisher z = −0.000, k = 2, 95% CI = −0.000 to 0.000, p = 0.748).

TABLE 5.

Association of sleep duration and Aβ and tau in non‐demented individuals.

Method	Biomarker	No. of studies	No. of participants	Fisher z	95% CI	p‐value	I 2	Q
PET	Aβ	6	5246	−0.002	−0.003 to −0.001	0.002 **	0.000	4.169
	tau	2	482	−0.000	−0.000–0.000	0.748	0.151	1.177
Plasma	Aβ40	3	5352	0.001	−0.006–0.009	0.740	0.000	1.399
	Aβ42	4	5499	−0.027	−0.080–0.025	0.304	0.660	8.832
	Aβ42/Aβ40	3	787	−0.007	−0.022–0.008	0.381	0.943	35.205
	t‐tau	2	5047	0.018	−0.035–0.071	0.507	0.367	1.581
CSF	Aβ42/Aβ40	2	154	0.000	−0.000–0.001	0.337	0.921	12.653
	t‐tau	2	177	−0.130	−0.369–0.109	0.286	0.906	10.629
	p‐tau/Aβ42	2	1267	0.004	−0.001–0.009	0.117	0.000	0.145
	t‐tau/Aβ42	2	77	0.002	−0.011–0.015	0.806	0.333	1.500

Abbreviations: Aβ, amyloid beta; CI, confidence interval; CSF, cerebrospinal fluid; PET, positron emission tomography; p‐tau, phosphorylated tau; t‐tau, total tau.

^** p < 0.01.

FIGURE 3.

Forest plot of meta‐analytic results on the association of sleep duration and Aβ and tau detected by PET in non‐demented individuals using the random‐effect model. The results are expressed as Fisher z score and 95% CI. Aβ, amyloid beta; CI, confidence interval; PET, positron emission tomography.

For fluid biomarkers, no significant associations were found between continuous sleep duration and Aβ and tau levels (Table 4 and Figure S2 in supporting information). The following biomarkers were examined, including plasma Aβ40 (Fisher z = 0.001, k = 3, 95% CI = −0.006 to 0.009, p = 0.740), plasma Aβ42 (Fisher z = −0.027, k = 4, 95% CI = −0.080 to 0.025, p = 0.304), plasma Aβ42/Aβ40 ratio (Fisher z = −0.007, k = 3, 95% CI = −0.022 to 0.008, p = 0.381), plasma t‐tau (Fisher z = 0.018, k = 2, 95% CI = −0.035 to 0.071, p = 0.507), CSF Aβ42/Aβ40 ratio (Fisher z = 0.000, k = 2, 95% CI = −0.000 to 0.001, p = 0.337), CSF t‐tau (Fisher z = −0.130, k = 2, 95% CI = −0.369 to 0.109, p = 0.286), CSF p‐tau/Aβ42 (Fisher z = 0.004, k = 2, 95% CI = −0.001 to 0.009, p = 0.117), and CSF t‐tau/Aβ42 (Fisher z = 0.002, k = 2, 95% CI = −0.011 to 0.015, p = 0.806).

Owing to the limited number of studies, subgroup analyses comparing categorical sleep duration (long/short vs. normal sleep duration) were conducted for the Aβ burden detected by PET within only two studies, and no significant association was observed (Table S3 in supporting information).

3.4. Publication bias, meta‑regression analysis, and sensitivity analysis

Visual inspection of the funnel plot and the Egger test for each biomarker were conducted to detect potential publication bias (Figure S3 in supporting information). Meta‐regression analyses were conducted on biomarkers with a sufficient number of studies (N ≥ 6), considering factors such as age, sex, sample size, percentage of APOE ε4 carrier, and publication year. The meta‐regression found that there were no significant effects of the above factors (Table S4 in supporting information). We used the leave‐one‐out method to assess the robustness of associations. The estimated effects between sleep quality and plasma Aβ42 and PET Aβ were different when studies were excluded individually in the leave‐one‐out analysis (Figures S4 and S5 in supporting information).

4. DISCUSSION

This study pooled data from 30 studies involving nearly 15,000 participants to explore the impact of sleep quality and sleep duration on Aβ and tau levels. Moreover, by focusing on a population of adults without cognitive impairment, our findings underscore sleep as a preventive measure for cognitive dysfunction. The findings suggest that sleep quality and sleep duration could significantly influence Aβ levels, with no evidence of effects on tau levels. Specifically, poor sleep quality was linked to higher plasma Aβ42 levels and greater PET Aβ burden. Short sleep duration was linked to increased Aβ burden detected by PET. Reducing modifiable risk factors is a key strategy in AD prevention. ⁶⁰ Our findings contribute to this approach by highlighting the potential interventions from a sleep perspective, including managing both sleep quality and sleep duration to control the levels of preclinical AD biomarkers, thereby lowering the risk of AD onset.

A prior meta‐analysis explored the association between sleep and Aβ, focusing on sleep duration and sleep efficiency in older adults. ⁶¹ In contrast to the previous study, our review expands to include tau and enlarges the population age range to adulthood. We conducted stratified analyses by categorizing studies based on plasma, CSF, and PET data rather than pooling them, and examined the associations with sleep quality and sleep duration independently. Additionally, we excluded studies that might influence AD biomarkers or sleep to ensure the reliability of our results. ,  ,  

Previous meta‐analyses have also demonstrated that sleep disturbances and abnormal sleep duration increase the risk of cognitive impairment and dementia. ⁷ , ⁸ , ⁶² , ⁶³ , ⁶⁴ , ⁶⁵ Higher nighttime restlessness is related to worse memory performance, while shorter sleep onset latency is associated with higher executive function. ⁶³ Abnormal sleep duration is found to be associated with poorer executive function, verbal memory, and working memory capacity. ⁶⁵ Moreover, sleep disorders such as insomnia and OSA, along with sleep parameters such as sleep fragmentation, are associated with increased AD risk. ⁸ , ⁶² In addition, OSA and excessive daytime sleepiness can also influence the levels of AD‐related biomarkers. ⁶⁶ , ⁶⁷ Our research complements the continuity of disease progression, where sleep quality and sleep duration affect AD biomarkers, ultimately contributing to cognitive impairment and potentially leading to dementia.

In our meta‐analysis, we found associations between sleep and Aβ rather than tau within individuals without cognitive impairment. We observed that PET Aβ burden and plasma Aβ42 levels were influenced by sleep quality and a significant relationship between shorter sleep duration and greater Aβ burden on PET. Though the observed effect size was minimal and suggested that the impact may be limited, the results remained robust. Also, further studies using categorical classifications of sleep duration (long/short vs. normal sleep duration) are needed to thoroughly investigate whether a J‐shaped relationship exists between sleep duration and Aβ levels. In addition, the studies we included were published only after 2013, as we did not impose any restrictions on publication years. Earlier research focusing on sleep and Aβ and tau was predominantly experimental, focusing more on animal studies rather than observational studies in humans. ¹² , ⁶⁸

Other biomarkers with only one study were not included in the meta‐analysis. Higher CSF Aβ40 levels were significantly correlated with worse sleep quality, and higher plasma p‐tau levels were associated with sleep latency over 30 minutes in the Aβ PET–positive group and with moderate/severe sleep disturbances in the Aβ PET–negative group. ²⁰ , ³⁸ Continuous sleep duration showed a negative correlation with CSF p‐tau and a positive correlation with plasma p‐tau levels. ³⁸ , ⁴³ Shorter sleep duration versus normal sleep duration (< 6 hours vs. 6–9 hours) was associated with lower CSF Aβ42, and was not associated with CSF p‐tau and CSF t‐tau. ⁴⁶

In our observation, poor sleep conditions are linked to high Aβ levels. Various studies have explored potential mechanisms underlying this connection. Sleep can effectively facilitate the clearance of metabolic byproducts during wakefulness, as Aβ levels exhibit a circadian rhythm, increasing during wakefulness and decreasing during sleep ¹² through processes like lymphatic transport ⁶⁹ and synaptic activity. ⁷⁰ In sleep deprivation experiments, even a minor degree of sleep loss increases soluble Aβ levels and Aβ burden. ⁷¹ , ⁷² , ⁷³ Except for macroscopic sleep parameters, factors such as slow‐wave activity during non–rapid eye movement sleep and rapid eye movement sleep duration are also important factors influencing Aβ levels, ⁷⁴ , ⁷⁵ and lower theta power during rapid eye movement sleep is connected with greater Aβ burden. ⁷⁶ While evidence exists, the underlying mechanisms and common pathways by which sleep quality and sleep duration affect AD biomarkers still require further investigation.

Notably, the relationship between sleep and AD is generally considered bidirectional and mutually influential. ⁷⁷ , ⁷⁸ The presence of Aβ and tau may conversely impact sleep conditions. ⁷⁷ , ⁷⁸ After the formation of Aβ plaque, the sleep–wake cycle deteriorates significantly, and the diurnal fluctuation of Aβ subsides in individuals without dementia. ¹² , ⁷⁹ Aβ pathology is also linked to disrupted slow‐wave activity and the impairment of memory consolidation. ⁸⁰

Our study's strength lies in its comprehensive examination of the impact of sleep quality and duration on Aβ and tau in plasma, CSF, and PET, categorizing analyses specifically among those without dementia. However, our study has several limitations. First, only cross‐sectional data are included in the analysis, which prevents us from obtaining a causal relationship between sleep and biomarkers. Second, there was considerable heterogeneity among studies, particularly given the relatively small number of articles on certain biomarkers. Moreover, restricted by the limited number of studies, subgroup analyses of population characteristics for each biomarker were not performed. The leave‐one‐out analysis indicated that the associations between sleep quality and plasma Aβ42 and PET Aβ require additional studies to achieve more robust results. Third, sleep quality and sleep duration may interact with each other. Extreme sleep duration often suggests a worsened sleep quality and vice versa.

To conclude, our study broadens the current understanding of how sleep influences Aβ and tau levels, demonstrating that sleep quality and sleep duration are associated with Aβ levels among people without cognitive impairment. For non‐demented individuals, maintaining good sleep conditions may be crucial for preventing AD by regulating the production and clearance of Aβ. Further research focusing on sleep as a modifiable intervention strategy for dementia prevention is warranted.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

Supporting information

Supporting Information

Supporting Information

Supporting Information

Chen C‐L, Zhang M‐Y, Wang Z‐L, et al. Associations among sleep quality, sleep duration, and Alzheimer's disease biomarkers: A systematic review and meta‐analysis. Alzheimer's Dement. 2025;21:e70096. 10.1002/alz.70096