Effect of Excessive Daytime Sleepiness and Long Sleep Duration on All Cause Dementia: A Systematic Review and Meta-analysis

Abstract

Background

Excessive daytime sleepiness (EDS) and long sleep duration are common in older adults and are related to dementia pathology. This study aims to assess the effect of EDS and long sleep duration on all-cause dementia and cognitive decline.

Methods

We identified longitudinal studies assessing the relationship between EDS and/or long sleep duration on cognitive decline/dementia published up to March 24, 2024 from MEDLINE, Embase, PsyINFO, the Cochrane Central Register of Controlled Trials, and PubMed. The inverse-variance-weighted average method and Bayesian multilevel regression models were used to test the effect of EDS and long sleep duration on cognitive decline and dementia risk.

Results

Fifteen studies with 65 501 participants were recruited. Excessive daytime sleepiness was associated with increased risk of cognitive decline and all-cause dementia (risk ratio 1.26, 95% confidence interval [CI]: 1.13–1.41, and risk ratio 1.68, 95% CI: 1.07–2.66, respectively). Long sleep duration increased the risk of all-cause dementia by 29% (95% hazard ratio CI: 0.94–1.77, Pr [hazard ratio > 1] = 0.94); and cognitive decline by 13% (95% risk ratio, CI: 0.92–1.4). Dementia-free time at age 85 was 0.13 years shorter among long sleepers (Pr (∆ Restricted Mean Dementia-Free Time < 0) = 0.98).

Conclusions

Both EDS and long sleep duration were associated with increased risk of all-cause dementia. However, it remains unclear whether these are independent risk factors or merely reflect the same underlying pathological process. Further studies are needed to explore the interaction between these exposures on dementia and their potential as targets for dementia prevention.

The incidence of dementia is increasing globally. Over 55 million people were living with dementia in 2020, and this is estimated to increase to 78 million by 2030 and 139 million by 2050 (1). Interestingly, the evidence from the last 3 decades highlights a reduction in the incidence of dementia in North America and Europe, which might be explained by higher education and better management of vascular risk factors (2). In the last few decades, research has been focusing on the effective prevention of dementia by identifying, and minimizing risk factors. There are several known risk factors for dementia such as hypertension, diabetes, obesity, smoking, high alcohol consumption, and physical inactivity—all of which are traditionally related to metabolic risk (3).

Recently, there has been growing interest in the impact of excessive daytime sleepiness (EDS) and prolonged sleep duration on cognition. These are not traditionally recognized risk factors; however, they are linked to dementia pathology via processes such as vascular disease, neuronal inflammation, cell apoptosis, oxidative stress, brain and hippocampal atrophy, and beta amyloid formation and deposition (4,5).

Previous studies have tested the relationship between the risk of dementia and sleep conditions including sleep duration, insomnia, and sleep behavior disturbances. A systematic review on sleep duration by Fan et al (6) reported an increased risk of dementia among long sleepers, whereas a review by Xu et al (7) suggested a U-shaped relationship between sleep duration and cognitive disorders. Therefore, the exact effect of long sleep duration on dementia remains unclear, as does the effect of EDS.

This study aimed to synthesize the data from longitudinal studies, by conducting a comprehensive systematic review and meta-analysis including the recent findings, to provide high-quality evidence on the associations between EDS, prolonged sleep duration, and the risks of cognitive decline and dementia.

Method

Search Strategy and Selection Criteria

We searched longitudinal studies from MEDLINE, Embase, PsyINFO, the Cochrane Central Register of Controlled Trials, and PubMed from database inception to March 24, 2024 using subject headings and text words for dementia, cognitive decline, cognition, Alzheimer, EDS, daytime sleepiness, hypersomnia, sleep duration, and long sleep duration. The exact search strategy is provided in Supplementary Text 1. No limitations were set for study language or publication status.

We included longitudinal studies which: (i) assessed the relationship between dementia of all cause/cognitive decline as a dichotomous outcome, and EDS as a dichotomous exposure and/or sleep duration per night or over 24 hours as a continuous exposure and which assessed the risk of cognitive decline/dementia of long sleep hours (> 8 h) versus normal sleep hours (7–8 h); (ii) cognition needed to be measured at least twice during follow-up; (iii) participants were adults aged 18 and above who were cognitively healthy at baseline; if cognitive decline was measured, we did not apply this exclusion. Some studies used the samples from a young age group and used different follow-up periods which could have significant effect on the outcome estimates. For example, the outcome prevalence of a study with mean age of 50 years with 5 years follow-up (ie, a mean sample age of 55 years at the follow-up) and that of a study with mean age of 77 years at baseline with 3 years follow-up (ie, a mean sample age of 80 years at follow-up) could vary significantly. This difference in outcome prevalence could have a large impact on the true estimate. To pool the homogenous data, studies with average age of participants less than 65 years at follow-up were excluded from the study given that the estimated prevalence of dementia among people below the age of 65 is low, that is, 0.1% (8). We also excluded cross-sectional studies, case reviews, and case series.

The titles and abstracts of the studies obtained from the search were screened and duplicates were excluded. The full texts of the remaining studies were retrieved. Two authors screened the full-text articles and identified eligible studies for inclusion. The reasons for the exclusion of the ineligible studies were recorded.

Data Extraction and Quality Assessment

We extracted the following data from each included study: country, sample size, number of exposed cases, the number of incident cases of all-cause dementia and cognitive decline, mean age, instrument used to measure the exposures, tool, and criteria used to diagnose dementia and cognitive decline, adjusted confounders, adjusted outcome estimates, outcome prevalence and time of follow-up.

The Newcastle-Ottawa Scale (NOS) was used to assess the quality of the studies (9). The Newcastle-Ottawa Scale assesses the risk of bias in cohort studies based on 3 main domains: Selection, Comparability, and Outcome. It has a total of 9 points; scores of < 5, 5–7, and > 7 was considered as low, moderate, and high quality respectively (9).

Statistical Analysis

The National Sleep Foundation generally recommends 7–8 h of sleep per day (10). For this study, we summarized the risk estimates of higher sleep hours than recommended (> 8 h) compared to the recommended sleep hours (7–8 h). The present statistical analysis outlines 4 separate meta-analyses consisting of 3 standard meta-analyses and 1 survival meta-analysis using Bayesian modeling for meta-analysis of reconstructed survival data (MARS). Data analyses were performed in R 4.3.1 (11) with the ‘metafor’ package (12) for standard meta-analyses and the ‘rjags’ package (13) for MARS.

Standard meta-analysis

The 3 standard meta-analyses combined estimates from studies that examined the relationship between (i) daytime sleepiness and cognitive decline, (ii) long-sleep duration and cognitive decline, and (iii) daytime sleepiness and dementia.

Odds ratios (OR) and hazard ratios (HR) from included studies were converted to risk ratios (RR) using the optimal approximate conversion method (14). The optimal approximate conversions for the dementia outcome assume that the true probability of dementia in both the exposed and unexposed group is within the interval of 2% and 20%, while the approximate conversions for the cognitive decline outcome assume that the true probability of cognitive decline is within the interval of 3% and 30%; it is these assumptions that allow the OR and HR to approximate each other. Not all studies reported the outcome prevalence. Therefore, in order to obtain a conservative confidence interval, we used the maximum bias among all pairwise combinations of outcome prevalences (restricted between 2%–20% or 3%–30%) in the control and exposed groups as a scaling factor. Confidence intervals (CI) of the original estimates were converted to the RR scale using the same optimal approximate conversions and scaled by the maximum bias ratio (the highest bias among all possible within-interval pairwise combinations of outcome probabilities). We also performed sensitivity analyses to examine the effect of specifying different prevalence intervals (Supplementary Table 1). In the worst case, using the maximum bias out of all possible pairwise combinations of outcome prevalences (restricted between 1% and 99%) in the control and exposed group leads to the most conservative 95% CI. The results revealed that for (a) cognitive decline and EDS, and (b) dementia and EDS, the most conservative 95% CI still do not include 1, indicating that these results remain significant even in the worst-case scenario and suggesting the robustness of our findings.

Assuming the true probabilities for each outcome are within the respectively specified intervals, the maximum bias ratio-scaled optimal approximate CIs will have at least 95% coverage of the true RR. Where relevant, studies reporting effect estimates for multiple exposure arms relative to the same reference group had their effect estimates pooled before being converted to RRs. The optimal approximate RRs were pooled across studies using the random-effects method to accommodate the between-study heterogeneity. For meta-analyses with fewer than 5 studies, Hartung–Knapp–Sidik–Jonkman adjustment was applied to correct for bias (15).

Cochran’s Q statistic and Higgins and Thompson’s I² were used to measure heterogeneity; I² values of 0%–40% indicate low heterogeneity, 30%–60% moderate heterogeneity, 50%–90% substantial heterogeneity, and 75%–100% considerable heterogeneity (16).

Meta-analysis of reconstructed survival data

The MARS combined estimates from studies that examined the relationship between long-sleep duration and dementia. The meta-analysis of the effect of long-sleep duration on dementia incidence was performed using a Bayesian multilevel regression model (17,18), which has been developed, validated, and applied in various clinical settings (19,20).

Individual patient survival data was reconstructed from the cumulative hazard curve from Sabia et al (21) using the method of Liu et al (22). Reported HR estimates from the remaining studies were used in the absence of survival curves. Results were reported as cumulative incidence probabilities of dementia by sleep duration starting from age 65, and the difference in restricted mean dementia-free time (∆ RMDFT, defined as RMDFT_{(>8 h sleep, t)—}RMDFT _{(7–8 h sleep, t)}, up to age t), with 95% credible intervals (CrI) (23,24) and posterior probability of ∆ RMDFT < 0, which suggests the chance of a longer dementia-free time for 7–8 h sleep group.

Results

Figure 1 demonstrates the flow diagram of the study selection process. Fifteen longitudinal studies (21,25–38) with a total of 65 501 participants were identified for this systematic review and meta-analysis. Ten studies (21,24–37) reported all-cause dementia and 9 (25,26,29–32,34,35,38) reported cognitive decline. Eight studies (25–32) assessed EDS and 8 (21,31,33–38) assessed sleep duration. The number of participants recruited ranged from 1 041 to 13 888 and the follow-up period ranged from 1 to 16.6 years. The main characteristics of the studies are summarized in Table 1. Out of 15 studies, 2 were of NOS moderate quality whereas the rest were of NOS high quality (Supplementary Table 2).

Figure 1.

Flowchart of selection of studies.

Table 1.

Characteristic of the Included Studies

First Author, Year	Study Cohort	Follow-up Year	Sample Size, Men (%)	Mean Age/Age Range	Sleep Category	Exposure Measurement	Outcome	Outcome Measure	Adjusted Confounders	Outcome Prevalence
										Whole cohort	Exposed cohort
Foley et al (25)	Hawaii	3	2346, 100%	76	EDS	Single item question	Dementia/cognitive decline	DSM III/CASE > 9 points drop from baseline	Age, education, marital status, forced expiratory volume, asthma, number of hours of sleep and napping, benzodiazepines use, COPD, coronary heart disease, stroke, APOE, BMI, depression	6%-dementia 16%-CD	NA
Elwood et al (26)	United Kingdom	10	1225, 100%	62	EDS	Wisconsin sleep questionnaire	Dementia/cognitive decline	National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association criteria/CDR	Age, social class, smoking, alcohol, BMI, neck circumference, angina, ischemic heart disease, chest pain and the use of sleeping tablets.	8%-dementia 14%-CD	NA
Taspanou et al (27)	United States	3	1041, 30%	79	EDS	12 item questionnaire	Dementia	DSM III R	Age, sex, education, ethnicity, APOE, depression, stroke, hypertension, diabetes, heart disease.	3.5% dementia	NA
Smagula et al (28)	United States	10	1951, 39%	78	EDS	Single item question	Dementia	CDR	Age, sex, smoking, alcohol, education, depression, chronic diseases, cardiovascular disease, cerebrovascular disease, diabetes	6%-dementia	9%-dementia
Quesnot and Alperovitch (29)	France	4	1389, 62%	59–71	EDS	Questionnaire	Cognitive decline	MMSE > 3 points drop from baseline	Age, gender, education, BMI, high blood pressure, initial MMSE	13%-CD	NA
Jaussent et al (30)	French	8	4894, 43%	>65	EDS	Single item question	Cognitive decline	DSM IV and MMSE > 4 points drop from baseline	Age, sex, education, depression, BMI, chronic disease, mobility, living alone, apolipoprotein, consumption of fish, fruit, vegetables, prescribed sleep medication	13%-CD	21%-CD
Blackwell et al (31)	United States	3.4	2822, 100%	76	EDS/sleep hours, <5 h, > 5–7, 8, > 8	ESS/Pittsburgh sleep quality index	Cognitive decline	3MS > 5 points drop from baseline, Trails B	Age, clinic site, race, BMI, education, depression, number of comorbidities, presence of impairment of instrumental activities of daily living, benzodiazepine, antidepressant use, self-reported health status, physical activity, alcohol use, and smoking status	18%-CD	NA
Nakakubo et al (32)	Japan	4	2096, 47%	70.8	EDS	Single item question	Cognitive decline	National Centre for Geriatrics and Gerontology Functional Assessment Tool < standard deviation	Age, sex, BMI, education, medication, heart disease, respiratory disease, diabetes, taking sleeping pills or other medication to aid sleep, current drinking habit, smoking, physical activity, gait speed, Geriatric Depression Scale and MMSE	9%-CD	19%
Benito-Leon et al (33)	Spain	3.2	3286, 43%	72.9	Sleep hours, < 5, 6–7, 8, > 9	Total sleep in 24 h	Dementia	DSM IV	Age, gender, education, smoke, alcohol, diabetes mellitus, hypertension, depression.	NA	NA
Lutsey et al (34)	United States	16	1653, 47%	62.7	Sleep hours, < 7, 7–8, 8–9, > 9	Sleep hours per night	Dementia/cognitive decline	DSM V/CDR	Age, sex, center, education, APOE, BMI smoking, physical activity, diabetes, antihypertensive medications, C-reactive protein, systolic blood pressure	3%-dementia 13%-CD	7% - dementia 15%-CD
Chen et al (35)	United States	7	7444, 0%	70.1	Sleep hours, < 5, 6, 7, 8, >9	Sleep hours per night	Dementia/cognitive decline	DSM IV, 3MS > 8 points drop	Age, race, social economic status, smoking, alcohol, physical activities, depression, hypertension, BMI, cardiovascular disease, hypertension, diabetes, hypercholesterolemia	7%-dementia 11%-CD	8% -dementia. 10% for CD
Sabia et al (21)	United Kingdom	21	7959, 67%	50.1	Sleep hours, < 6, 7, >8	Sleep hours per night	Dementia	Record linkage to national patient registers	Age, sex, ethnicity, education, marital status, alcohol, smoking, physical activity, fruit and vegetable consumption, BMI, hypertension, diabetes, cardiovascular disease, depression, central nervous system mediations	5%-dementia	8%-dementia
Wong et al (36)	United Kingdom	16.6	6284, 45%	76.3	Sleep hours, < 7, 7–8, >8	Total sleep in 24 h	Dementia	Dementia screening interview, cognitive tests, self-reported dementia or dementia diagnosis by a doctor	Age, sex, race, ethnicity, education, income, marital status, metropolitan residence, health condition, BMI, Activity of daily living, proxy respondent, depression, anxiety, heart disease, hypertension, diabetes	14%-dementia	NA
Xiong et al (37)	United Kingdom	8	7223, 54%	73	Sleep hours, < 7, 7–8, >8	Average hours of sleep on weekdays	Dementia	IQCODE or physician diagnosed dementia	Age, sex, education, wealth, marriage, wealth, smoking, alcohol, physical activity, BMI, chronic disease, depression, chronic disease	7%-dementia	NA
Xu et al (38)	China	4.1	13888, 27%	61.5	Sleep hours, 3–4, 5, 6, 7, 8, > 9	Total sleep in 24 h	Cognitive decline	DWRT score < 4	Age, sex, occupation, education, smoking, drinking, physical activity, objective health status, self-rated health, waist to hip ratio, daytime napping, morning tiredness, insomnia	7%-dementia	NA

Notes: APOE = apolipoprotein E; BMI = body mass index; CASE = Cognitive Assessment Scale for the Elderly; CD = cognitive decline; CDR = Clinical Dementia Rating Score; DSM = Diagnostic and Statistical Manual of Mental Illnesses; COPD = chronic obstructive pulmonary disease; DWRT = Alzheimer’s Disease Delayed Word Recall Test; ESS = Epworth Sleepiness Scale; IQCODE = Informant Questionnaire on Cognitive Decline in the Elderly; MMSE = Mini Mental State Examination; 3MS = Modified Mini Mental State.

Outcome: Cognitive Decline

Exposure: Excessive Daytime Sleepiness

Six studies (17,18,21–24) with a total of 14 772 people were included to test the effect of EDS on risk of cognitive decline. The pooled data from the 6 included studies indicated that EDS was associated with an increased risk of cognitive decline (RR 1.26, 95% CI: 1.13–1.41, I² 0%). The heterogeneity was not significant (Figure 2A).

Figure 2.

Meta-analysis of risk ratios of cognitive decline and dementia among those with excessive daytime sleepiness relative to nondaytime sleepiness and among long sleep hours (>8 h) relative to reference sleep hours (7–8 h).

Exposure: Long sleep duration

Four included studies (24,26,27,31) tested the relationship between long sleep duration and cognitive decline. The pooled data from these studies totaling 25 807 participants identified that sleeping > 8 h did not statistically significantly increase the risk of cognitive decline (RR 1.13, 95% CI: 0.92–1.40, I² 0%). There was no significant heterogeneity found (Figure 2B).

Outcome: Dementia

Exposure: Excessive Daytime Sleepiness

The pooled data from 4 included studies (25–28), with a total of 6 563 participants, showed a significant association between EDS and dementia (RR 1.68, 95% CI: 1.07–2.66, I² 56.8%) with significant between-study heterogeneity (Figure 2C).

Exposure: Long sleep duration

The data from 6 included studies (21,33–37) with a total 33 849 participants was used for the Bayesian multilevel regression model. The cumulative incidence curve demonstrated a 29% increase (HR: 1.29; 95% CrI: 0.94–1.77) in dementia risk among the group sleeping > 8 h compared to the 7–8 h group (Figure 3A). The probability that the HR estimate is greater than 1 was 94.5%.

Figure 3.

A. Cumulative incidence of dementia from age 65 to 89, by sleep duration, B. RMDFT at 5 year intervals from age 65 up to 89. Note: RMDFT: restricted mean dementia-free time.

By the 20th year from age 65, those in the > 8 h sleep group had on average 0.13 less years of dementia-free time than those in the 7–8 h group; the probability that the ∆ RMDFT is less than 0 was 0.98 (Figure 3B).

Discussion

To our knowledge, this study is the first comprehensive meta-analysis assessing the effect of daytime sleepiness on all-cause dementia, although previous meta-analyses tested the association between sleep problems including sleep duration and dementia risk.

Our pooled result was statistically significant for the effect of EDS on cognitive decline with low between-study heterogeneity (I² = 0%) although there was some variation in point estimates across the 6 included studies (RR varied from 1.07 to 1.55). Of note, 2 out of 6 studies used a validated self-reported questionnaire to measure EDS whereas the others did not. However, both questionnaires used (either validated or not) were self-reported. Different methods were applied to identify cognitive decline. This methodological diversity might explain the differences in the effect sizes of the individual studies.

The between-study heterogeneity of the 4 included studies for the exposure of EDS and the outcome of dementia risk was statistically significant. There was also significant variation in study population, participants’ mean age and characteristics, sample size, the methods to define dementia, and the follow-up period across the 4 studies. Moreover, only 1 out of 4 studies used a validated self-reported questionnaire whereas the other used different self-reported 1 item question to identify EDS. These could have driven the heterogeneity; however, the number of studies was too small to explore the source of heterogeneity formally with subgroup analysis or metaregression.

EDS is related to global and regional cerebral atrophy with cortical thinning, causing accelerated brain aging (39). Moreover, a previous study reported that increased β-amyloid cerebral deposition was also found among participants with EDS (4). It has long been known that EDS is associated with an increased risk of cardiovascular disease (40); for example, the Northern Manhattan study found that EDS was an independent risk factor for stroke and vascular disease (41) which, in turn, is a known risk factor for dementia (42). Furthermore, EDS is a common symptom of obstructive sleep apnea (OSA) (43). The intermittent hypoxia in OSA leads to oxidative stress and increased sympathetic nervous system activity consequently causing cerebral hypoperfusion, blood brain barrier damage, decreased perivascular clearance, hypercoagulability, and endothelial dysfunction resulting in progression of cerebral small vessel disease, atherosclerosis and increased risk of stroke (43). Reduced slow-wave sleep in OSA decreases β-amyloid clearance (43). In addition, the recent systematic review by Kang and colleague revealed that dementia biomarkers such as cerebrospinal fluid Amyloid β-40, blood Amyloid β-42 and blood total-tau were elevated in OSA patients (44). Therefore, EDS is related to dementia via both Alzheimer’s pathology and vascular pathology.

Among the long sleep duration group, a 29% increased risk of dementia was found. There is strong evidence that long sleep duration activates systemic inflammation (5), which is linked to neuroinflammation and consequent amyloid deposition and neuronal cell death (45). Prolonged sleep is also associated with cerebral small vessel disease (46) which is a main cause of cerebrovascular injury leading to vascular dementia (42). Long sleep is also related to cardiovascular disease which is again a major risk factor for vascular dementia (5,39). A previous study reported that sleep is important for β-amyloid clearance from the brain (6), although the link is more complex than just the number of hours slept. Long sleep duration and daytime sleepiness, as well as potentially short sleep (7), are linked to reduced slow-wave sleep during which β-amyloid aggregates are cleared; therefore, these factors are probably better conceptualized as disordered sleep and decreased β-amyloid clearance and increased β-amyloid deposition (47).

Excessive quantity of sleep was related to excessive sleepiness (48) suggesting that prolonged sleep duration might indicate poor-quality sleep and hence lead to EDS. A recent study pointed out the strong interaction between prolonged sleep duration and EDS for dementia risk, that is, long sleep duration with EDS increases the risk of all cause dementia (49). At this point it is unclear whether it is the number of hours of sleep or the quality of sleep, or both, that affect the risk of dementia. This would necessitate a meta-analysis of studies that have used regression models with both terms modeled independently and simultaneously, with an interaction term. It might be better to frame future work in terms of disordered sleep.

It could be argued that long sleep duration might be a prodromal symptom of dementia and the association seen here is simply reverse causation. However, all included studies excluded people with cognitive impairment at baseline and 5 out of 6 included studies had long follow-up periods, from 7 to 16.6 years; these lowered the chance of reverse causation (50). To take the matter further, Xiong et al study performed a sensitivity analysis excluding dementia identified within the first 2 years from baseline (37). Moreover, reverse causation seems less likely given the cumulative incidence curves for dementia in our study do not diverge until very late in the follow-up. By age 85, long sleepers had dementia for on average 0.13 years longer than those who slept 7–8 h per day. Of note, the included studies measured sleep duration only at baseline, potentially leading to measurement error and reducing the power to detect an effect. Moreover, even though we only included studies with participants aged more than 65 years at follow-up, extrapolating over a 20-year time frame may be inexact.

The pooled data from our study showed that long sleep duration increased the risk of cognitive decline by 13% with low between-study heterogeneity (I² = 0%) which was not statistically significant. There were only 4 studies included for this outcome, with widely different sample sizes (1 667–30 518 participants), mean age (60–76 years), patient characteristics and follow-up periods (3–16 years). Of these 4 studies, 2 measured the self-reported sleep duration per day whereas the other 2 measured the self-reported total hours of night sleep, indicating potential clinical and methodological heterogeneity. Of note, we were not able to exclude publication bias due to the small number of included studies which could lead to an exaggeration in heterogeneity (either low or high) and consequently affect the true effect size.

Our meta-analysis found that EDS was associated with increased risk of both cognitive decline and dementia (26% and 68%, respectively). Therefore, EDS may assist in identifying at risk individuals; screening EDS can be both time and cost effective in the primary care setting. Our study finding of an increased risk of dementia (29%) and cognitive decline (13%) among the long sleep duration group also points to the role of assessing both sleep quality and quantity in clinical practice.

Our findings on long sleep duration and dementia risk are similar to the previous meta-analyses by Fan et al. (6) and Xu et al. (7). The former included 7 studies and investigated the association between sleep duration and all-cause dementia. Their study reported that long sleep duration was associated with increased risk of all-cause dementia and Alzheimer’s dementia and the effect of short sleep duration on dementia was not statistically significant. Xu et al. (7) included 17 studies and reported a significantly increased risk of cognitive disorders among those with long sleep duration. Xu et al. study assessed the dementia risk of highest sleep duration versus middle sleep duration. In both Fan et al. and Xu et al. study, the hours of sleep for long sleep duration or highest and middle sleep duration were not clearly defined. Pooling data from studies of variably defined sleep duration and reference sleep groups could bias the outcome estimates. Our study tested the effect of long sleep duration on all-cause dementia and cognitive decline with clearly defined sleep duration (> 8 h for long sleep and recommended sleep of 7–8 h for reference group), and data were pooled from studies of compatible groups. Additionally, we tested the effect of EDS on cognitive decline and all-cause dementia.

The strength of our study is that it is the first meta-analysis testing the effect of EDS and long sleep hours on both cognitive decline and dementia risk. We have also calculated dementia-free time for those with > 8 h sleep versus 7–8 h sleep. Moreover, in our meta-analysis, 8 longitudinal studies were included for daytime sleepiness and 8 studies were included for long sleep duration. Furthermore, the sleep categories (long sleep vs reference sleep group) defined in the studies varied from each other, which can affect the estimates. To deal with these differences, our study applied the National Sleep Foundation’s recommended sleep hours range to define sleep hour categories. The data was then drawn from the studies as per redefined sleep categories and Bayesian multilevel regression model was applied to estimate the effect.

However, we acknowledge some limitations of our study. Firstly, different measures for EDS were used across included studies, that is, some used validated measures such as the Epworth Sleepiness Scale and Wisconsin Sleep Questionnaire, whereas other studies used a 1 item question. Secondly, sleep duration was largely self-reported, which may not reflect actual sleep hours. Therefore, actigraphy devices and wearables sleep-trackers should be considered in the future research to gather objective sleep behavior data and to measure sleep hours accurately. Moreover, some studies measured sleep hours per night and some measured daily sleep hours. Thirdly, there was some heterogeneity in the characteristics of the recruited participants, follow-up periods, measures to identify cognitive decline and dementia, and adjusted variables. Finally, while the studies that assessed dementia excluded participants with dementia at baseline, there is still a possibility of mild cognitive impairment at baseline. These factors could bias the estimates.

In conclusion, both EDS and long sleep hours were associated with increased risk of dementia. Sleeping > 8 h was related to less dementia-free time compared to sleeping 7–8 h. As both exposures are linked to dementia, further studies are required to test the interaction between EDS and long sleep hours on dementia and to assess the role of management of EDS and long sleep duration in reducing the risk of cognitive decline and dementia.

Funding

None.

Conflict of Interest

None.

Author Contributions

K.K.: Conceptualization, Methodology, Validation, Investigation, Writing original draft. X.D.-G.: Writing—Review & Editing, Supervision. J.D.: Software, Formal analysis, Data Curation. X.F.: Software, Formal analysis. Z.T.: Software, Formal analysis. C.F.: Software, Formal analysis. D.B.: Formal analysis, Data Curation. S.Z.: Formal analysis, Data Curation, Writing—Review & Editing. B.R.N.: Writing-Review & Editing, Supervision. J.B.: Methodology, Writing—Review & Editing, Supervision. J.A.: Methodology, Resources, Writing—Review & Editing, Supervision.