Sleep duration and health in adults: an overview of systematic reviews

Abstract: The objective of this overview of systematic reviews was to examine the associations between sleep duration and health outcomes in adults. Four electronic databases were searched in December 2018 for systematic reviews published in the previous 10 years. Included reviews met the a priori determined population (community-dwelling adults aged 18 years and older), intervention/exposure/comparator (various levels of sleep duration), and outcome criteria (14 outcomes examined). To avoid overlap in primary studies, we used a priority list to choose a single review per outcome; reviews that examined the effect of age and those that looked at dose–response were prioritized. A total of 36 systematic reviews were eligible and 11 were included. Reviews included comprised 4 437 101 unique participants from 30 countries. Sleep duration was assessed subjectively in 96% of studies and 78% of studies in the reviews were prospective cohort studies. The dose–response curves showed that the sleep duration that was most favourably associated with health was 7–8 h per day. Modification of the effect by age was not apparent. The quality of the evidence ranged from low to high across health outcomes. In conclusion, the available evidence suggests that a sleep duration of 7–8 h per day is the one most favourably associated with health among adults and older adults. (PROSPERO registration no.: CRD42019119529.)


Introduction
Sleep is increasingly recognized as a critical component of cognitive, emotional, and physical health. Healthy sleep is characterized by adequate duration, good quality, appropriate timing, and the absence of sleep disorders (Buysse 2014;Chaput and Shiau 2019). Insufficient sleep has become a public health concern in many countries given its high prevalence and association with mortality and morbidity (Liu et al. 2016;Chaput et al. 2017). Not only can insufficient sleep adversely impact health, it can also lead to mistakes in the workplace, lower psychomotor performance, decreased work productivity, and increased risk of car accidents (Institute of Medicine (US) Committee on Sleep Medicine and Research 2006). Thus, insufficient sleep poses a substantial burden to our health and economic sectors in disability and injury each year.
Sleep duration varies across the lifespan and shows an inverse association with age (Chaput et al. 2018). Sleep duration recommendations issued by public health bodies are important for surveillance, help inform policies and interventions, and can be used to educate the general public about healthy sleep (Chaput 2019). For example, the National Sleep Foundation recommends 7-9 h of sleep per day for adults (aged 18-64 years) and 7-8 h for older adults (aged ≥65 years) (Hirshkowitz et al. 2015) while the American Academy of Sleep Medicine and Sleep Research Society recommends ≥7 h per night on a regular basis for adults aged 18-60 years (Watson et al. 2015). These American organizations used very similar guideline development processes but had different experts at the table, which resulted in slightly different sleep duration recommendations.
Given the large body of evidence linking sleep duration and health outcomes in the adult population, an overview of reviews that examine the amount of sleep most favourably associated with overall health as evidenced with dose-response curves is needed to better inform public health guidelines. An overview of reviews (rather than a review of primary studies) is needed at this stage because it allows to leverage existing research, reduces redundancy in research, and enhances efficiency in the review process considering limited timelines and resources. Furthermore, such an overview of reviews needs to examine the effect of age on the associations between sleep duration and health outcomes to determine if recommendations should be different for older compared with younger adults. Therefore, the present work is the first to provide an overview of systematic reviews on sleep duration and health outcomes in adults aged 18 years and older. The main objective is to determine the sleep duration associated with overall health in adults and examine the impact of age. Findings from this review will inform public health guidelines around sleep duration and identify future research needs.

Protocol and registration
The present overview of reviews was registered a priori with the International Prospective Register of Systematic Reviews (PROSPERO; registration no. CRD42019119529; available from http://www.crd. york.ac.uk/PROSPERO/display_record.asp?ID=CRD42019119529), and was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement for reporting systematic reviews and meta-analyses (Moher et al. 2009).

Eligibility criteria
The Participants, Interventions, Comparisons, Outcomes, and Study design (PICOS) framework (Schardt et al. 2007) was followed to identify key study concepts in the research question a priori and to facilitate the search process.

Population
The population of interest was community-dwelling adults aged 18 years and older, including apparently healthy adults, adults with obesity, adults with metabolic syndrome, or adults who have had one or more falls in the past year. This also included studies that, among their participant pool, included adults with a chronic condition (e.g., heart disease, diabetes, cancer). Reviews including mixed populations, that is, comprising studies with both individuals who met and those who did not meet the eligibility criteria, were included if the results pertaining to the population of interest were reported separately. If results for the population of interest were not reported separately, studies with a mixed population were included if 80% or more of the study population met the inclusion criteria (or if the sample average fit within the criteria). For example, a systematic review with no subgroup analyses that included some studies from the general population, and some from disease-specific populations, would be included if 80% or more of the participants in the systematic review were from those studies performed in the general population. Exclusion criteria included pregnant women, residents in long-term care, patients in acute care or a hospital setting, people who were unable to move under their own power, and elite athletes (e.g., varsity/ provincial level athletes or Masters' athletes). We also excluded studies that targeted exclusively shiftworkers and individuals with a sleep disorder or diagnosed disease (e.g., insomnia, type 2 diabetes, depression) to keep the focus on the general adult population and not specific clinical populations.

Intervention (exposure)
The intervention or exposure was sleep duration. Sleep duration could be reported in different ways in the studies and could include total sleep duration (i.e., per 24-h period, including naps, or nighttime sleep duration only). Reviews were eligible if they included primary studies that used objective (e.g., polysomnography, actigraphy/accelerometry) or subjective (e.g., self-report) measures of sleep duration (or both). For systematic reviews that included primary experimental studies, the interventions must have targeted sleep duration exclusively and not multiple health behaviours (e.g., both sleep and diet).

Comparison
Various levels of sleep duration were used for comparison. However, a comparator or control group was not required for inclusion.

Study designs
Published or in-press peer-reviewed systematic reviews (as defined by The Cochrane Collaboration (Higgins and Green 2011)) with or without meta-analyses were eligible for inclusion. No grey literature was eligible for inclusion. Systematic reviews that did not receive a "yes" or "partial yes" for items 4 (adequacy of literature search) and 9 (risk of bias from individual studies being included in the review) on A MeaSurement Tool to Assess systematic Reviews (AMSTAR 2) assessments were excluded as these characteristics were considered critical flaws (Shea et al. 2017). For the purpose of this overview, systematic reviews must have searched at least 2 relevant databases and provided a key word and/or search strategy. Justifying language restrictions was not required, as long as the review was published in English or French.

Information sources and search strategy
A research librarian with expertise in systematic review searching created the electronic search strategy and a second research librarian reviewed it. The complete search strategies are available in Supplement S1. 1 The following databases were searched using the Ovid interface: MEDLINE, EMBASE, and PsycINFO. CINAHL was also searched using the EBSCO platform. Searches were conducted the week of December 18, 2018. Studies published in the previous 10 years only (December 18, 2008, until December 18, 2018 were searched to manage scope, reduce overlap, and with a goal to include the most recent body of evidence to inform current public health recommendations around sleep duration in Canada. The search was rerun on October 14, 2019, in case relevant reviews may have been published since then. Reference lists of included studies were also checked.

Study selection
Bibliographic records were extracted and imported into the Reference Manager Software (Thompson Reuters, San Francisco, Calif., USA) for removal of duplicate references. In level 1 screening, titles and abstracts of potentially relevant articles were screened by 2 independent reviewers using Covidence (Veritas Health Innovation, Melbourne, Australia). In level 2 screening, full-text copies of articles were obtained for those meeting the initial screening criteria. If an article was included by 1 reviewer and not the other at the title and abstract stage, the article was obtained for further review. Two independent reviewers examined all full-text articles. Any discrepancies were resolved with a discussion and consensus between the 2 reviewers or by a third reviewer if required.
After having the pool of eligible systematic reviews, we aimed to retain 1 review per outcome to reduce overlap in primary studies. The review that was retained was selected using a priority list that was designed to select the review that would provide the best evidence to inform sleep duration recommendations. Prioritization of systematic reviews to identify the best review for each outcome was as follows. First, studies that reported direct outcome measures were prioritized over studies that reported indirect markers of the outcome measures. Second, we prioritized reviews that examined the effect of age (e.g., if the effects were different in adults aged 18-64 years vs. ≥65 years) and dose-response (i.e., optimal sleep duration from dose-response curves). Third, if there was more than 1 review that addressed these criteria, we selected the review that was of highest quality based on full AMSTAR 2 assessment. Finally, if there were multiple reviews of high-quality we prioritized the most recent review. If a review did not address, or addressed only 1 of "age" or "dose-response", we considered whether estimates of effect from separate reviews could be included to address these components. For example, if no reviews assessed the effect of age for the mortality outcome, we considered including estimates of effect from 1 systematic review that included adults aged 18-64 years, and estimates of effect from a second systematic review that included adults aged 65 years or older. The same strategy was applied for "dose-response".
If estimates of effect from more than 1 systematic review needed to be included for a given outcome (i.e., to be able to evaluate the effect of age and/or dose-response), we assessed and reported on the degree of overlap in primary studies between systematic reviews using the corrected covered area (CCA) (Pieper et al. 2014). The extent of primary study overlap among the systematic reviews was interpreted as either slight (0%-5%), moderate (6%-10%), high (11%-15%), or very high (>15%) (Pieper et al. 2014). The degree of overlap was reported but not used as an exclusion criterion if papers analyzed the data differently (e.g., by age in 1 review and by dose-response in another review).
If there were no systematic reviews for a critical outcome, a de novo review was planned. De novo reviews were not planned if there were no systematic reviews for outcomes deemed important but not critical.
Detailed information about the methodology used to prepare this overview of reviews is provided in another paper published in this issue of the journal ).

Data extraction
Microsoft Excel was used for data extraction. Data extraction was completed by 1 reviewer and verified by another reviewer. Where multiple models were reported, results from the most fully adjusted models were extracted. Important study features (e.g., author, publication year, country, study design, sample size, age, exposure, outcome, results, covariates, and quality of the evidence) were extracted. The optimal sleep duration was also extracted from dose-response curves, where available. We also extracted differences in effect by age, sex, race/ethnicity, socioeconomic status, weight status, and/or chronic disease status, where available. Reviewers were not blinded to the authors or journals when extracting data.

Quality assessment of included systematic reviews
Two reviewers independently assessed the methodological quality of each systematic review using the AMSTAR 2 rating scale (Shea et al. 2017). AMSTAR 2 contains 16 items to appraise the methodological aspects of reviews. All assessments were discussed and agreed upon based on discussion among the 2 reviewers (or in consultation with a third reviewer, if required). AMSTAR 2 ratings are "high", "moderate", "low", and "critically low" quality. For the purposes of this review, we considered the following items noncritical: 2 (pre-registration of protocol), 3 (explanation of included study designs), 7 (justification for exclusion of individual studies), and 10 (reporting sources of funding for individual studies). The item on conflict of interest (COI) requires that COI for the systematic review and all primary studies be assessed. We modified this item to assess whether potential COI was documented only for the review itself.

Quality assessment of primary studies within included systematic reviews
The quality assessment of primary studies, performed by the authors of the systematic reviews, was extracted and reported. The information was reported as indicated by the systematic review authors (e.g., "the authors concluded that the evidence was low to moderate quality"). In the event of a de novo search for a particular outcome, quality assessment using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework was planned.

Synthesis of results
Results were summarized via narrative synthesis, grouped by outcome. Results were described as reported by the systematic review authors, such as reporting available summary estimates and confidence intervals as well as the number of primary studies and participants that contributed to each available estimate. Tables were used to ensure consistency of data presentation across studies. A narrative description of any subgroup analyses by age, sex, race/ethnicity, socioeconomic status, weight status, and/or chronic disease status was also summarized.

Description of studies
As reported in Fig. 1, a total of 2315 records were identified through database searches and 1257 records remained after removing duplicates. After titles and abstracts were screened, 130 full-text articles were obtained for further review and 36 articles met the inclusion criteria and were thus eligible for inclusion. Reasons for exclusion were as follows: wrong outcome (n = 34), wrong study design (n = 23), duplicate publication (n = 18), wrong exposure (n = 15), and wrong population (n = 4). A total of 11 articles were included in this overview of reviews after using the priority list to avoid overlap and with a goal to keep the best review per outcome. Among the 36 articles that were eligible for inclusion, reasons for excluding articles were the absence of dose-response relationships reported (n = 12), the lower priority of outcomes reported (i.e., directedness) (n = 8), and the lower quality reviews using AMSTAR 2 (n = 5). See Supplement S2 1 for the complete list of full-text articles excluded and the list of 25 reviews that were not chosen for this overview of reviews. Of note, the updated search strategy conducted in October 2019 did not result in relevant systematic reviews that would have been included in this overview of reviews.
Characteristics of studies sorted by outcome are summarized in Table 1 (critical outcomes) and Table 2 (important outcomes). Overall, the 11 reviews included 4 437 101 unique participants from 30 countries. The study designs used in the reviews were prospective cohort studies (78% of studies), cross-sectional studies (19% of studies), and case-control studies (3% of studies). Sleep was assessed subjectively in 96% of studies and objectively in 4% of studies. Sleep duration (hours) was reported "per night" or "per day" in the studies. The quality of the evidence, as reported in the reviews by the authors, ranged from low to high across health  Table 1.

High
One noncritical weakness (authors did not report on the sources of funding for the studies included) Overall, this review reported that short sleep duration, defined as a duration less than 6 hh/d, was associated with a 12% absolute increase in mortality risk compared with normal sleep.
Quality scores of included studies were all considered "high" quality on the Newcastle-Ottawa scale as they ranged from 7/9 to 9/9 (7-9 = high quality). Mean score = 8.0
Overview of key systematic reviews that examined the relationship between sleep duration and important health outcomes in adults.

High
One noncritical weakness (authors did not report on the sources of funding for the studies included) Overall, short sleep defined as the duration of less than 6 h was associated with a 38% absolute increase in the incidence of obesity Variation of effect: Compared with normal sleep duration, short sleep duration was associated with a significant increase in the incidence of obesity only among participants Incidence of obesity (mean weight gain relative to reference category, BMI ≥25 kg/m 2 and ≥30 kg/m 2 , body weight increase >5 kg, and WC of ≥90 cm for men and ≥85 cm for women) A meta-analysis of 13 articles revealed that compared with normal sleep, long sleep duration was significantly associated with increased incidence of obesity at an RR of 1.08 (95% CI: 1.02-1.15, p = 0.010, I 2 = 0%) Quality scores of included studies ranged from 5/9 to 9/9 on the Newcastle-Ottawa scale (4-6 = moderate quality, 7-9 = high quality). Mean score = 6.1

High
One noncritical weakness (authors did not report on the sources of funding for the studies included) Overall, long sleep duration was associated with an 8% absolute increase in the incidence of obesity Variation of effect: Long sleep duration, compared with normal sleep duration, was associated with a significant increase in the incidence of obesity only among participants <65 y (RR = 1.09, 95% CI: 1.00-1.18), and not among those ≥65 y (RR = 1.45, 95% CI: 0.87-2.42). In comparison with normal sleep duration, long sleep duration was associated with a significant increase in the incidence of obesity only among female (RR = 1.12, 95% CI: 1.00-1.26 for female vs. RR = 1.10,

Incidence of hypertension
was defined as a blood pressure ≥140/90 mm Hg, taking anti-hypertensive medications, or selfreported This meta-analysis containing 9 articles revealed that compared with 7 h of sleep (reference category), 6 h and ≤5 h of sleep per night were associated with higher incidence of hypertension, while 9 h of sleep per night was associated with a decreased risk of incidence of hypertension. There was no significant difference between the 7 h and >9 h groups Quality scores of included studies ranged between 6/9 and 8/9 on the Newcastle-Ottawa scale (4-6 = moderate quality and 7-9 = high quality). Mean score = 7.1

Low
One critical flaw with noncritical weaknesses (authors did not report some of the information needed in the literature search strategy and did not provide a list of the full texts that were excluded along with the justifications) Overall, this dose-response metaanalysis showed that hypertension incidence was higher with shorter sleep (<6 h/d) durations than 7 to 9 h of sleep Variation of effect: Variation of effect by age, sex, race/ethnicity, socioeconomic status, weight status, or chronic disease status not reported outcomes. Our assessment of the AMSTAR 2 ratings ranged from low to high for the systematic reviews.

Mortality
We included 3 reviews examining the association between sleep duration and mortality in adults (Table 1). Yin et al. (2017) showed that compared with 7 h of sleep per day, a 1-h decrease in sleep duration was associated with 6% increased risk of all-cause mortality and a 1-h increase in sleep duration was associated with a 13% increased risk of all-cause mortality (N = 241 107 adults in 43 articles). Itani et al. (2017) showed that short sleep duration (<6 h/day) was associated with a 12% absolute increase in mortality risk compared with normal sleep (N = 1 298 970 adults in 38 articles) while Jike et al. (2018) showed that long sleep duration (>8 or 9 h/day) was associated with a 39% absolute increase in mortality compared with normal sleep (N = 1 296 470 adults in 37 articles). These last 2 reviews were published by the same authors; they decided to publish 1 systematic review on "short sleep" and another one on "long sleep" in 2 different publications instead of 1. The extent of primary study overlap was thus very high (CCA: 65.7%). The quality of the evidence and AMSTAR 2 ratings were moderate-to-high. Wang et al. (2016) reported that compared with 7 h of sleep per day, a 1-h decrease in sleep duration was associated with an 11% increased risk of incident cardiovascular heart disease and a 1-h increase in sleep duration was associated with 7% increased risk of incident cardiovascular heart disease (N = 517 440 adults in 17 articles) ( Table 1). The quality of the evidence was moderate-tohigh and the AMSTAR 2 rating was moderate. Shan et al. (2015) showed that compared with 7 h per day, a 1-h decrease in sleep duration was associated with 9% increased risk of incident type 2 diabetes and a 1-h increase in sleep duration was associated with 14% increased risk of incident type 2 diabetes (N = 482 502 adults in 10 articles) ( Table 1). The quality of the evidence was moderate-to-high and the AMSTAR 2 rating was moderate. Itani et al. (2017) reported that short sleep duration (<6 or ≤7 h/day) was not significantly associated with the incidence of depression compared with normal sleep duration (N = 16 257 adults in 2 articles) ( Table 1). Jike et al. (2018) reported that long sleep duration (>8 h/day) was not significantly associated with the incidence of depression compared with normal sleep duration (N = 15 204 adults in 1 article). These 2 reviews were published by the same authors but were split into 2 publications (1 focusing on short sleep and 1 focusing on long sleep). The extent of primary study overlap was thus very high (CCA: 50%). The quality of the evidence was moderate-to-high and the AMSTAR 2 rating was high. Wu et al. (2018) showed that the lowest incident risk of cognitive disorders was found at a sleep duration of 7 to 8 h per day (N = 22 187 adults in 9 articles) ( Table 1). The quality of the evidence was moderate-to-high and the AMSTAR 2 rating was low. Lo et al. (2016) reported that extreme sleep durations (both short and long) were associated with cognitive decline compared with the reference sleep of 7 to 8 h/day (N = 97 264 adults in 18 articles) ( Table 1). The quality of the evidence was judged "satisfactory" with a modified version of the Downs and Black Quality Index score system and the AMSTAR 2 rating was low.

Variation of the effect by sex or other factors
Another objective of this review was to examine if the associations between sleep duration and outcomes differed not only by age but also by sex, race/ethnicity, socioeconomic status, weight status, and/or chronic disease status by looking at subgroup analyses and/or effect modification. No reviews reported information about a possible variation of effect by socioeconomic status, weight status, or chronic disease status. Among the reviews that reported a possible variation of effect by sex, n = 9 reported no variation of effect by sex while n = 1 (Jike et al. 2018) reported that the association between long sleep duration and incident obesity was significant among female (RR = 1.12, 95% confidence interval (CI): 1.00-1.26) but not male (RR = 1.10, 95% CI: 0.97-1.25) participants. With regard to race/ethnicity, 1 study reported no variation of effect (Wu et al. 2018) while another study (Wu and Sun 2017) found a stronger association between short sleep duration and falls among Caucasians (odds ratio (OR) = 1.69, 95% CI: 1.36-2.08) compared with Asians (OR = 1.23, 95% CI: 1.04-1.45).

Discussion
This overview of reviews synthesized peer-reviewed scientific evidence from 11 systematic reviews examining the associations between sleep duration and health outcomes in adults aged 18 years and older. The reviews included over 4 million participants from 30 countries, and the majority of studies (78%) were prospective cohort studies that relied on self-reported sleep (96% of studies) for the assessment of sleep duration. The overall quality of evidence (primary studies contained within the reviews) ranged from low to high across outcomes, and AMSTAR 2 ratings (quality of systematic reviews) also ranged from low to high. Collectively, a U-shaped association between sleep duration and health outcomes was observed. Evidence from dose-response curves showed that the sleep duration that was most favourably associated with the health outcomes examined was around 7-8 h per day. Among the critical outcomes examined, no reviews reported a possible modification of the effect by age.
A key observation when looking at the available evidence was the reliance on self-reported sleep duration measures (96% of studies). Sleep questions typically used in population health surveys tend to overestimate actual sleep duration compared with objective measures and can introduce inaccuracies (Girschik et al. 2012). Using an objective measure when assessing sleep may be especially important for adults with poor sleep quality because time in bed, as assessed with sleep duration questionnaire items, is not the same as actual sleep duration as not all time in bed is spent asleep (Chaput et al. 2018). There is a growing popularity in the use of actigraphy/accelerometry in epidemiologic research, which may have implications for sleep duration recommendations in the future (e.g., lower optimal sleep duration if objectively assessed compared with self-reported time in bed for example). With advances in wearable health technologies (e.g., Fitbit, Apple Watch), measures of sleep duration can be readily derived and tracked by the general public and future studies will need to re-examine dose-response curves with the use of other and more objective sleep assessments (and actigraphy assessment is not without limitations).
A goal of this overview of reviews was to provide evidence that could be used to better inform sleep duration recommendations for public health guidance. Currently, the National Sleep Foundation recommends 7-9 h of sleep per day for adults (18-64 years) and 7-8 h for older adults (≥65 years) (Hirshkowitz et al. 2015) while the American Academy of Sleep Medicine and Sleep Research Society recommends ≥7 h per night for adults aged 18-60 years (Watson et al. 2015). Our findings are more in line with the National Sleep Foundation recommendations and suggest that 7-8 h of sleep per day is the optimal amount based on doseresponse curves. While public health guidelines are informed by the best available evidence, they are also informed by expert consensus, stakeholder consultations, and consideration of values and preferences, applicability, feasibility, and equity (Tremblay et al. 2016). For example, the well-known inter-individual variability in sleep needs (Chaput et al. 2018) means that 9 h of sleep per day may well be the optimal duration for a given individual; it may thus be ill-advised to recommend less sleep to this individual even if the sleep duration recommendations say "7-8 h per day". This highlights the potential complexity of conveying such a message in public health because it is never a one-size-fits-all approach and we need to adapt our recommendations on a caseby-case basis.
Our findings clearly show the presence of a U-shaped association between sleep duration and health outcomes, with both short and long sleep durations being associated with adverse health outcomes. While there is a large body of evidence providing biological plausibility for short sleep as causally related to adverse health, the role of long sleep is unclear (Chaput et al. 2018). Long sleep is generally associated with other health problems (e.g., depression, chronic pain, obstructive sleep apnea), which may confound the association between sleep duration and health. Thus, the observed association between long sleep and poor health may reflect reverse causation and residual confounding (Knutson and Turek 2006;Stamatakis and Punjabi 2007). The absence of plausible biological mechanisms by which long sleep per se can cause adverse health was a key factor for the American Academy of Sleep Medicine and Sleep Research Society to recommend a threshold value (≥7 h) for sleep duration rather than a range. However, long sleep duration may be a useful marker of other possible health problems and may also be indicative of poor sleep efficiency (i.e., spending a lot of time in bed but with low quality sleep).
Primary studies included in the systematic reviews synthesized herein included both nighttime sleep duration and 24-h sleep duration depending on the tool used, and all presented results in terms of "hours per day". A meta-analysis of prospective cohort studies using both nighttime sleep duration and 24-h sleep duration reported similar findings with all-cause mortality in adults (Shen et al. 2016). However, recent systematic reviews that examined daytime napping only and mortality concluded that naps longer than 30 min per day were associated with an increased risk of mortality (da Silva et al. 2016;Yamada et al. 2015;Zhong et al. 2015). Napping is also discouraged among people with insomnia as it can reduce the homeostatic sleep drive and perpetuate nighttime insomnia (Robbins et al. 2019). However, naps are encouraged to counteract sleep loss and achieve more sleep in some populations such as shiftworkers (Faraut et al. 2017). Napping is a heterogeneous and complex sleep habit that is beyond the scope of this overview of reviews and that has been recently reviewed in detail elsewhere (Faraut et al. 2017).
There were no relevant systematic reviews identified that examined the associations between sleep duration and health-related quality of life, work productivity, or physical activity/sedentary behaviour. We also excluded reviews with clinical populations (e.g., exclusively in patients with sleep disorders) to focus on community-dwelling adults in the general population. The present overview of reviews focused on sleep duration only and did not include other important characteristics of sleep health such as sleep quality, sleep timing, sleep consistency, daytime alertness, and the absence of sleep disorders (Buysse 2014;Chaput and Shiau 2019). Future studies are thus needed to broaden the scope of this overview of reviews.
We also examined whether the associations between sleep duration and health outcomes varied as a function of age, sex, race/ ethnicity, socioeconomic status, weight status, and/or chronic disease status. Unfortunately, many systematic reviews did not conduct subgroup analyses, making it difficult to know whether the findings observed can be generalized to all community-dwelling adults. However, among the reviews that included subgroup analyses, the findings were not different based on age or sex of the participants.
A number of limitations should be highlighted. First, the available evidence largely consisted of studies relying on self-reported sleep (96% of studies), with different categorization of sleep durations across studies. Second, many reviews did not report subgroup analyses, so it is difficult to determine whether the relationships observed apply broadly to all adults in the general population. Third, this review excluded studies that targeted clinical populations exclusively as well as those that examined other characteristics of sleep health. Fourth, reviews pooled studies that assessed sleep duration "per night" and "per day" together so the influence of napping on the observed relationships is not known. Fifth, a single measure of sleep duration at 1 point in time may not fully capture the chronic effects of sleep duration over time when related to long-term disease incidence. Sixth, this review only included systematic reviews published in English or French and conducted over the past 10 years. However, excluding non-English publications from evidence syntheses does not impact conclusions according to a recent meta-epidemiologic study (Nussbaumer-Streit et al. 2020). Finally, the risk of publication bias (i.e., an overrepresentation of studies with significant findings) cannot be discarded.

Conclusion
A comprehensive body of evidence supports the presence of a U-shaped association between sleep duration and health outcomes in adults. Dose-response curves showed that the sleep duration that was most favourably associated with the health outcomes that were examined was around 7-8 h per day in adults, with no apparent modification of the effect by age in the few studies that looked at it. The present overview of reviews is important to inform public health recommendations around healthy sleep duration of adults. Given the growing interest in using actigraphy/accelerometry in population health research, future studies will need to assess the impact of this approach on the measured associations between sleep duration and health outcomes. Future work is also needed to better examine whether sleep duration recommendations should vary between adults and older adults.