Sleep duration, sleep efficiency, and amyloid β among cognitively healthy later-life adults: a systematic review and meta-analysis

Background Abnormal amyloid β (Aβ) deposits in the brain are a hallmark of Alzheimer’s disease (AD). Insufficient sleep duration and poor sleep quality are risk factors for developing AD. Sleep may play a role in Aβ regulation, but the magnitude of the relationship between sleep and Aβ deposition remains unclear. This systematic review examines the relationship between sleep (i.e., duration and efficiency) with Aβ deposition in later-life adults. Methods A search of PubMed, CINAHL, Embase, and PsycINFO generated 5,005 published articles. Fifteen studies met the inclusion criteria for qualitative syntheses; thirteen studies for quantitative syntheses related to sleep duration and Aβ; and nine studies for quantitative syntheses related to sleep efficiency and Aβ. Results Mean ages of the samples ranged from 63 to 76 years. Studies measured Aβ using cerebrospinal fluid, serum, and positron emission tomography scans with two tracers: Carbone 11-labeled Pittsburgh compound B or fluorine 18-labeled. Sleep duration was measured subjectively using interviews or questionnaires, or objectively using polysomnography or actigraphy. Study analyses accounted for demographic and lifestyle factors. Based on 13 eligible articles, our synthesis demonstrated that the average association between sleep duration and Aβ was not statistically significant (Fisher’s Z = -0.055, 95% CI = -0.117 ~ 0.008). We found that longer self-report sleep duration is associated with lower Aβ (Fisher’s Z = -0.062, 95% CI = -0.119 ~ -0.005), whereas the objectively measured sleep duration was not associated with Aβ (Fisher’s Z = 0.002, 95% CI = -0.108 ~ 0.113). Based on 9 eligible articles for sleep efficiency, our synthesis also demonstrated that the average association between sleep efficiency and Aβ was not statistically significant (Fisher’s Z = 0.048, 95% CI = -0.066 ~ 0.161). Conclusion The findings from this review suggest that shorter self-reported sleep duration is associated with higher Aβ levels. Given the heterogeneous nature of the sleep measures and outcomes, it is still difficult to determine the exact relationship between sleep and Aβ. Future studies with larger sample sizes should focus on comprehensive sleep characteristics and use longitudinal designs to better understand the relationship between sleep and AD. Supplementary Information The online version contains supplementary material available at 10.1186/s12877-024-05010-4.


Background
Alzheimer's disease (AD) is a progressive neurodegenerative disease affecting one in ten adults over the age of 65 worldwide, which poses a considerable economic challenge [1].More than 6.5 million older Americans suffered from AD in 2022, and the estimated cost for AD is $321 billion worldwide [2].By 2050, the number of AD cases in the US is expected to reach 12.7 million individuals.Neurodegenerative processes associated with AD result in the accumulation of senile plaques and pathologic changes in Amyloid β (Aβ) throughout the brain, cerebrospinal fluid (CSF), and serum [3,4].AD biomarkers may be present even decades before clinical AD symptoms appear [5].However, few effective disease-modifying treatments exist to delay the onset of AD symptoms.Thus, there is a pressing need to identify modifiable risk factors and develop novel interventions to decrease the risk of AD.
Alterations in sleep duration and efficiency can lead to numerous consequences for health and well-being and increase the risk of AD [6].Current guidelines state that healthy sleep is a sleep duration of 7 or more hours per night for adults between 18 and 60, 7-9 h for adults between 61 and 64, and 7-8 h for 65 years and older [7][8][9].However, later-life adults who are older than 50 typically experience less than 7 h of sleep duration and 85% sleep efficiency compared to younger adults [10].In particular, slow-wave sleep declines significantly with age [10].In addition to the changes in sleep structures, sleep disorders including insomnia and sleep-disordered breathing increase with age [10].Additionally, individuals with mild cognitive impairment or AD often experience disruptions in sleep and experience sundown syndrome.However, this condition often occurs years prior to impairment [11][12][13].
Sleep maintains brain and neural homeostasis [14].During sleep, the brain controls Aβ peptide regulation [15], clears neurotoxins including Aβ plaques [16], and decreases systematic inflammation [17].Thus, reductions in sleep duration or disruptions during sleep can influence the pathological changes of Aβ.Numerous recent papers and reviews focusing on the overall direction of sleep have suggested that sleep fragmentation or disruption is associated with AD via Aβ or tau pathology [15,[18][19][20][21].These findings suggest that improving sleep efficiency and optimal sleep quantity could be an opportunity to prevent and delay AD pathology by decreasing Aβ deposition and tau hyperphosphorylation. Insomnia, sleep disordered breathing, and sleep fragmentation have been found to be associated with the risk of developing AD and related dementia [6].However, researchers do not fully understand what the optimal sleep duration is to prevent AD.
Prior studies have speculated that shorter sleep duration can be associated with Aβ levels, because both total and partial sleep deprivation have been shown to increase Aβ levels in plasma [22,23], CSF [24,25], or brain [26][27][28].For instance, Zhao et al. (2019) found that chronic sleep restriction was associated with increases in Aβ in a mouse model [28].Kang et al. (2009) suggested that acute sleep deprivation can increase Aβ levels in animals via orexin regulation [29].To better understand the magnitude of the relationship between sleep metrics and Aβ, a meta-analysis and/or systematic review is needed.However, few meta-analyses and systematic reviews have specifically focused on how sleep duration and/ or efficiency matters for Aβ accumulation in human studies on adults in later life.The purpose of this systematic review is to focus on the current state of science on how sleep duration/efficiency is associated with Aβ in the brain, CSF, and serum in older adults.

Methods
The purpose of this study is to conduct a systematic review to evaluate how sleep duration and efficiency are associated with Aβ.The study was registered a priori with the International Prospective Register of Systematic Reviews (PROSPERO; registration no.CRD42021266789).
This review was conducted following the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement for reporting systematic reviews and meta-analyses [30].Search strategies were developed with the assistance of a health sciences librarian with expertise in searching for systematic reviews.The flow diagram in Fig. 1 provides details on the search strategy and the number of articles each database yielded.Comprehensive strategies, including both index and keyword methods, were devised by the librarian and the primary author for the following databases: PubMed, CINAHL (EBSCO platform), Embase (Elsevier platform), and PsycINFO (EBSCO platform).To maximize sensitivity, no pre-established database filters other than the English language filter were used.The full Pub-Med search strategy, as detailed in Supplemental Table A, was also adapted for the other databases.In addition to the database searches, references and cited papers of the 1,156 relevant papers were located using the Scopus database.
Inclusion criteria for the qualitative synthesis were as follows: 1) observational studies with a longitudinal or cross-sectional design, 2) includes exposure variables of sleep duration and/ or sleep efficiency using subjective or objective measures, 3) has Amyloid β plaques (e.g., Aβ, Aβ 42 , Aβ 40 , Aβ 42 /Aβ 40 ) as the outcome, 4) a human study of adults aged 3 50 years old, and 5) recruited (or included) cognitively healthy individuals.An additional inclusion criterion for the quantitative synthesis was studies that reported sufficient data for examining the effect sizes, such as Pearson's correlation (r), regression coefficient (β), means, standard deviations, t, F, or X 2 values.We excluded studies 1) not written in English, 2) interventional studies, 3) non-peer reviewed papers, proceedings, editorials, and reviews, and 4) the study sample focused only on neurological conditions or sleep disorders.For the quantitative synthesis, we excluded studies that lacked or had inadequate inferential statistical results for calculating the effect size.
The initial search yielded 6,987 articles.After removing 1,982 duplicate articles, 5,005 articles were imported to the web-based systematic review application, Rayyan.The level of agreement between authors was determined using Cohen's κ statistics.Four authors (CM, KV, HD, MZ) screened the abstracts and titles of the 5,005 articles based on the eligibility criteria (k = 0.49).Then, additional articles were removed leaving 62 full-text articles that were reviewed by four authors (CM, AS, YC, MZ) (k = 0.97).A total of 15 articles met the inclusion criteria for the qualitative synthesis and 13 studies met the criteria for quantitative synthesis of sleep duration, and 9 studies for sleep efficiency (Fig. 1 ).Disagreements were resolved through discussion among all authors until consensus was reached.

Quality analysis
The risk bias of the selected papers was assessed independently by two reviewers using the National Institute of Health Study Quality Assessment Tool (2019) for Observational Cohort and Cross-Sectional Studies (https:// www.nhlbi.nih.gov/ health-topics/ study-quali ty-asses smenttools).The internal validity of the studies were assessed based on 14 domains: 1) bias due to an unclear purpose, 2) bias due to an unclear specification of the population, 3) bias due to ineligible participants, 4) bias due to recruitment from a different population, 5) bias due to unclear power justification, 6) bias due to measure timing, 7) bias due to time frame, 8) bias due to outcome level, 9) bias due to invalid exposure measure, 10) bias due to frequency of the assessments, 11) bias due to the outcome measure, 12) bias due to an unblinded assessment, 13) bias due to loss during follow up, and 14) bias due to statistical analysis and confounding.For each domain, we categorized the risk of bias as either low or high risk.We rated an item "unclear risk" if there was no information about the risk of bias.

Statistical analysis
We aggregated the effect sizes across the studies and calculated the publication bias, overall effect sizes, and Q statistics using Comprehensive Meta-Analysis (CMA) version 4 software (Biostat, Inc).We also calculated the effect sizes using Fisher's Z as the effect size index after examining the available information on the correlation between sleep duration/sleep efficiency and Aβ using Fisher's Z = 0.5*Log(1 + Corr)/(1-Corr) [31].We used a Q statistic to evaluate the heterogeneity of variance.We also calculated the I 2 index using I 2 = 100% × (Q − degree of freedom)/Q to identify how the variance in observed effects reflected the variance in true effects rather than by random error.The random-effects model was applied in the current study because we expected that the sampling distribution varied across the studies and parameters were drawn from random variables [32][33][34][35][36].We also ran subgroup analysis based on the sleep measure (i.e., self-report vs. objective).We used meta-regression analyses to investigate quantitative relationships between the dependent variable and covariate (i.e., sex).To consider the possibility of sampling bias from all possible samples, we assessed the studies for publication bias.First, we visually inspected the studies for symmetry of the funnel plot (Supplemental Figure A, B).Second, we ran Tweedie's Trim and Fill test to ensure that the publication bias could not reverse our estimate of the effect sizes [31].

Results
Fifteen articles were included in the qualitative synthesis portion of this review.Thirteen studies were included for the quantitative synthesis of sleep duration and Aβ levels, and nine studies were included for quantitative synthesis of sleep efficiency and Aβ levels.Table 1 summarizes the study sample, design, and assessments of each study.Table 2 summarizes the study results and relevant information.Table 3 summarizes meta-analysis results.

Quality assessment
Figure 2 illustrates the assessment of the risk of bias categories.Among the 15 studies, three had a moderate to high risk of bias due to measurement timing, the outcome measure, exposure measure, or the population and participants.Two of the 15 articles had a risk of bias related to a small sample size and population without any power justification.Two of the studies reported a high risk of bias related to the measurement timing, timeframe, and outcome and exposure variable.Ten of the 15 articles had a risk of bias related to the exposure measure using a self-report sleep question or questionnaire.Five articles had a risk of bias due to the limited number of confounding variables.

Sleep measures
Both subjective and objective sleep measures were used in the reviewed studies (Table 1).Five studies used objective measurements, including polysomnography (PSG) [39,42] and actigraphy [37,43,45].Overnight data were collected in studies with PSG.Of the three articles assessing sleep duration and sleep efficiency with actigraphy, two of the studies used Actiwatch 2 (Phillips Respironics) [37,43].Ettore et al. (2019) used a three-axis accelerometer (GT3X + , Actigraph Corp, Pensacola, FL).All actigraphy data were collected in 60-s epochs, and the duration of the actigraphy recording ranged from 6 to 14 days.
Ten of the reviewed studies used subjective measurements including the Pittsburgh Sleep Quality Index (n = 6) [41,46,[48][49][50][51], a standardized interview to assess sleep duration (n = 3) [38,44,47], and the sleep scale in the Medical Outcomes Study (n = 1) [40].The interviews included questions related to the duration of nighttime sleep, daytime sleep, total sleep time (daytime and nighttime sleep), and sleep efficiency [44].Six articles used sleep duration as a categorical variable.Spira

Association between exposure and outcomes
Table 2 describes the findings of the reviewed studies for the qualitative synthesis.Five of the 15 articles [38,[47][48][49][50] found that shorter sleep duration was associated with higher Aβ.However, four of the studies reported the reverse association between sleep duration and PETmeasured global and regional Aβ burden.Winer et al., (2021) found that self-reported shorter sleep duration was associated with greater 18 F-florbetapir-PET brain imaging derived DVR Aβ burden (β = -0.01;p = 0.005).Spira et al. ( 2014) also reported that shorter sleep duration was associated with greater Aβ burden, measured by mean cortical [C-11] PiB PET derived DVR (cDVR; β = 0.08, p = 0.005) and precuneus DVR (β = 0.11, p = 0.007).Longer total sleep time was associated with reduced 18 F-florbetapir-PET brain imaging derived SUVr global Aβ (β = -0.005;p = 0.03), reduced medial orbitofrontal Aβ (β = -0.009;p < 0.001), and reduced anterior cingulate Aβ (β = -0.011;p < 0.001).Sleep duration longer than 8 h was associated with having higher amyloid burden compared to sleep duration shorter or equal to 8 h (Odds Ratio = 4.167; p = 0.020) [50].In addition to studies using PET, findings from a Chinese sample by Liu et al. (2021) also found that shorter sleep duration was associated with higher plasma Aβ 42 (β = 0.495, p = 0.021) and Aβ 42 /Aβ 40 ratio (β = 0.101, p < 0.001) [48].Fu et al. (2021) also found a non-linear relationship indicating a decrease in CSF Aβ42 with shorter or longer sleep duration, with the extreme point being 6.23 h of sleep [49].However, Blackman et al. (2023)'s cross-sectional analysis and longitudinal analysis did not identify any association between sleep duration and Aβ42 although the authors found significant associations between sleep characteristics and CSF P-tau and t-tau [51].

Quantitative synthesis
Thirteen eligible articles were used for the quantitative synthesis because the remaining studies were not included due to a lack of reporting information to calculate Fisher's Z value.Because one study included both objective and subjective sleep data, we included objective sleep information to calculate the overall associations.However, we included the effect size information based on the self-report measures when we conducted the subgroup analysis.Figure 3 and Table 3 summarize the quantitative synthesis of the articles.
The findings demonstrate that the average association between sleep duration and Aβ was not statistically significant (Fisher's Z = -0.055,95% CI = -0.117~ 0.008) (Fig. 3A).The Z-value was -1.720 with p = 0.085.As shown in Table 3, for heterogeneity, our results indicate a Q-value of 44.44 with 12 degrees of freedom.The amount of between-study variance in the observed effect was less than we expected based on sampling error alone.The I 2 statistic was 73%, indicating that 73% of the variance in observed effects reflect the variance in true effects rather than sampling error.Tau reflects the standard deviation of the true effect size, which is 0.007 in Fisher's Z units.
For the subgroup analysis (Fig. 3B) based on the sleep measurement, we found that the average association between self-reported sleep duration and Aβ was significant.This finding indicates that longer self-reported sleep duration is associated with lower Aβ (Fisher's Z = -0.062,95% CI = -0.119~ -0.005) with a Z-value of -2.146 (p-value = 0.032) in nine studies.However, the average association between objectively measured sleep duration and Aβ was not statistically significant (Fisher's As shown in the standard error funnel plot by Fisher's Z (Supplemental Figure A), the plot is slightly asymmetric, indicating that there could be minor publication bias from the included studies.This might be due to either our inability to identify studies with non-significant finings or failing to report non-significant findings [31].Thus, we conducted an analysis using Tweedie's Trim and Fill method for the overall relationship between sleep duration and Aβ, which demonstrated that even if we removed one study, the effect size remained statistically insignificant (Fisher's Z = -0.054,95% CI = -0.117~ 0.008).This finding may imply that some of the articles may not have presented the findings of non-significant results.
For sleep efficiency, the findings from nine studies demonstrated that the average association between sleep efficiency and Aβ was not statistically significant (Fisher's Z = 0.048, 95% CI = -0.066~ 0.161) (Fig. 4A). Figure 4 and Table 3 summarize the quantitative synthesis of the articles.The Z-value was 0.823 with p = 0.410.The Q-value is 66.532 with 8 degrees of freedom and p < 0.001.Using a criterion alpha of 0.100, we can reject the null hypothesis that the true effect size is the same in all of these studies.The I 2 statistic was 88% and Tau-squared was 0.024 in Fisher's Z units, and Tau was 0.154 in Fisher's Z units.For the subgroup analysis based on the sleep measurement (Fig. 4B), we found that the average association between self-report sleep efficiency and Aβ was not significant (Fisher's Z = -0.007,95% CI = -0.126~ 0.113) with Z value of -0.107 (p-value = 0.915) in five studies.However, the average association between objectively measured sleep efficiency and Aβ was not statistically significant (Fisher's Z = 0.085, 95% CI = -0.054~ 0.225) with a Z value of 1.199 (p-value = 0.230) in five studies.However, the meta-regression results indicated that higher proportion of female was associated with the higher correlation of sleep efficiency and Aβ (coefficient = 1.746, 95% CI = 0.345 ~ 3.136) with Z-value of 2.44 (p-value = 0.015).As shown in the standard error funnel plot by Fisher's Z (Supplemental Figure B), the plot is slightly asymmetric, indicating that there could be minor publication bias from the included studies.This bias could be due to either our inability to identify studies with non-significant Fig. 4 Forest plot of Fisher's Z score for the association between sleep efficiency with Amyloid beta finings or failing to report non-significant findings [31].We ran Tweedie's Trim and Fill method for the overall relationship between sleep duration and Aβ, which demonstrated that even if we remove one study, the effect size remains statistically insignificant (Fisher's Z = 0.048, 95% CI = -0.066~ 0.161).This finding may imply that some of the articles might not have presented the findings due to non-significant results.

Discussion
This review synthesized fifteen studies for the qualitative synthesis and 13 studies for the quantitative synthesis focusing on sleep duration and Aβ and 9 studies for focusing on sleep efficiency and Aβ.This review adds to the current literature with an overview of the measurements and findings related to sleep and Aβ levels.Our meta-analysis findings indicate that there was a significant inverse relationship between self-report sleep duration and Aβ levels, indicating that self-report shorter sleep duration was associated with greater Aβ levels, with Fisher's Z = -0.062(95% CI = -0.119~ -0.005, p = 0.032).However, we did not find a significant overall relationship between sleep duration and Aβ levels.We also did not find a significant overall relationship between sleep efficiency and Aβ levels.Due to the heterogeneity among the published studies, no firm conclusions could be drawn.
We only found a significant inverse relationship between self-report sleep duration and Aβ levels.Prior research has demonstrated that chronic sleep restriction or deprivation of slow wave sleep can alter the diurnal fluctuation of CSF Aβ levels [24,25,52,53].Sleep deprivation may also impair human memory consolidation, in part by reducing the synthesis of proteins needed to support synaptic plasticity [14,[54][55][56].In a meta-analysis by Wu and colleagues (2018), the authors suggested that there is a U-shaped relationship between sleep duration and cognitive disorders.Compared to the reference group (7 -8 h per day), individuals with short or long duration had a higher risk of developing cognitive disorders, such as Alzheimer's disease or dementia [57].Both shorter sleep duration (< 7 h / night) and poor subjective sleep quality are important for cognitive function [58] and brain structures and functions [59,60].Furthermore, higher amyloid burden may be associated with worse sleep recollection in the self-report measures.However, more studies are needed using larger sample sizes, and a prospective design.Future studies on the magnitude of correlations of null results can shed light on the true relationship between sleep duration and Aβ.
The overall quantitative synthesis of sleep (duration and sleep efficiency) and Aβ revealed an effect size of -0.055 and 0.048, but they were not statistically significant, indicating that sleep may not be a primary factor in Aβ accumulation.Alternatively, these results may be due to moderators (i.e., APOE4, sex, age, family history, or unmeasured moderators), heterogenous outcome types of Aβ that may not provide consistent ideas, or publication bias due to insignificant results not being favorable for publication.There was considerable heterogeneity in the methods used in the reviewed studies to identify the relationship between sleep (duration and efficiency) and Aβ burden.It is intriguing that the results from the subgroup analysis differed when studying sleep duration and Aβ in studies using subjective and objective sleep measures.Future studies are needed to capture habitual sleep duration using an actigraphy with verification using a sleep diary for 7 to 14 days [61].The use of a polysomnography would also provide insights on the structure of sleep and sleep disorders [62][63][64].
The majority of the studies used sleep duration as a continuous variable using multi-variate linear regression models.Winer [47,48].Both studies indicated that a shorter sleep duration compared to the standard 7-8 h a night or 7 h or more sleep is associated with greater Aβ.Among the reviewed studies, only Fu and colleagues reported a non-linear relationship between sleep duration and CSF measured Aβ 42 demonstrating lower Aβ 42 values for shorter or longer sleep and the highest for 6.23 h [49].Future studies using comprehensive and accurate assessments of sleep as well as a non-linear model would provide deeper insights on recommendations for sleep duration.
In addition to the considerable variability in sleep measurements, Aβ was also measured in different ways: CSF, serum sample, and PET to quantify the Aβ burden.Studies using PET used different tracers (e.g., 11C-PiB, 18F-florbetapir, 18F-flutemetamol) as well as different quantification methods.Most studies focused on global Aβ burden in the brain, but assessing both the overall levels of Aβ in PET as well as specific regional deposition could help us understand areas of the brain that may be affected more than other areas.This variability across the measurements prevented us drawing strong conclusions.However, it is promising in the current field of science to review data across different measures of Aβ accumulation.Although AD can be diagnosed at an autopsy [65], the US National Institute on Aging and Alzheimer's Association has suggested using Aβ as well as tau and neurodegeneration to define and diagnose AD in both symptomatic and asymptomatic stages [66].Increased accessibility to biomarkers and the potential for blood biomarkers or additional biomarkers in addition to Aβ would provide further information about the underlying disease progression in the future.
In addition to duration and efficiency of sleep, other sleep dimensions could be important factors for Aβ accumulation.The reviewed studies also identified a positive link between Aβ and different sleep characteristics including less adequate sleep, more sleep problems, and greater somnolence based on participants' self-report perceptions [40], sleep quality [38], frequent napping [37], longer sleep latency [41,45], greater sleep fragmentation [45], a higher apnea hypopnea index, and slow wave sleep time [39,42].These results may indicate that different dimensions of sleep could contribute more to Aβ burden than quantity of sleep.
The reviewed studies accounted for various demographic and clinical confounders in the multivariate models.Interestingly, we found that sex distribution could impact the relationship between sleep efficiency and Aβ although sex was not a significant moderator on the relationship between sleep duration and Aβ.Most of the reviewed studies accounted for age and sex, which are well-known confounders [10,[67][68][69].Further stratified analysis or moderation analysis based on sex and age group will provide more accurate understanding of the covariates.Individuals with sleep disorders, underlying health or psychological conditions, medications, genetic factors, social determinants, high fat diets, and physical activity have different sleep quantity and quality [70][71][72].These factors may also increase the amount of Aβ accumulation [73][74][75][76][77][78][79][80].These confounding factors may play critical roles in determining the association between sleep duration and Aβ burden.However, these studies used a cross-sectional design, which prevents us from determining causal relationships.Specifically, some of the studies did not measure sleep and Aβ burden in a similar time period, so the results may not reflect a direct link between the two factors.Although researchers have speculated that the relationship between sleep and AD pathology could be bidirectional, there is limited evidence to support the longitudinal relationships [21,81].
The strength of this review is that we examined current evidence related to sleep and Aβ.However, there are a few limitations.First, the study did not test for moderating effects of age, sex, or APOE4 status.Second, the current study only included publications written in English even though some important findings may have been published in different languages.Third, this review focused on sleep duration and efficiency, but other specific sleep characteristics could have more influence on Aβ pathology.

Conclusions
The results of this systematic review suggest an inverse association between self-report sleep duration and Aβ.However, the relationships between sleep duration and Aβ accumulation as well as sleep efficiency and Aβ accumulation should be interpreted with caution.Researchers would greatly benefit from more studies using a longitudinal design, comprehensive sleep measure, a broad range of biomarkers, and larger sample sizes to advance scholarly understanding of the relationship between sleep and AD.
Fisher's Z score for association between sleep (duration and efficiency) with Amyloids beta Upper limit Z-value P-value Lower limit Upper limit Tau TauSq Q-value df (Q) P-value Isleep efficiency was associated with a 41% reduction of Aβ positivity (Odds Ratio = 0.59, < 0.001).Ju et al. (2013) reported that individuals with a low CSF Aβ 42 level (≤ 500 pg/mL), which is indicative of amyloid deposition in the brain, had worse sleep efficiency than those with a normal CSF Aβ 42 level (80.4% vs. 83.7%,p = 0.08), although sleep duration was not statistically significant.Liu et al. (2021) reported that sleep efficiency was negatively associated with the plasma Aβ42 level (β = − 0.025, 95% CI − 0.037 ~ − 0.013, p = 0.001) and Aβ42/Aβ40 ratio

Fig. 2
Fig. 2 Risk of bias summary

Fig. 3
Fig. 3 Forest plot of Fisher's for the association between sleep duration with Amyloid β

Table 1 Study characteristics First author (year) Study Design Sample name, country Sample size / female n (%) Mean age ± SD (years)
et al. coded the categorized sleep variable in a continuous manner by coding 0 for sleep duration longer than 7 h a night, and 3 for sleep duration of 5 h or less.The other two papers by Gabelle et al. and Winer et al. categorized sleep duration as shorter (6 h or less), normal (6-7 h and 6-8 h), and longer (≥ 7 h or ≥ 9 h) sleep duration.Chu et al.