- Email: [email protected]
Sleep duration and risk of stroke: a dose–response meta-analysis of prospective cohort studies
Accepted Manuscript Sleep duration and risk of stroke: a dose–response meta-analysis of prospective cohort studies Qiao He, Hao Sun, Xiaomei Wu, Peng Zhang, Huixu Dai, Cong Ai, Jingpu Shi PII:
S1389-9457(16)30329-X
DOI:
10.1016/j.sleep.2016.12.012
Reference:
SLEEP 3263
To appear in:
Sleep Medicine
Received Date: 8 September 2016 Revised Date:
17 November 2016
Accepted Date: 5 December 2016
Please cite this article as: He Q, Sun H, Wu X, Zhang P, Dai H, Ai C, Shi J, Sleep duration and risk of stroke: a dose–response meta-analysis of prospective cohort studies, Sleep Medicine (2017), doi: 10.1016/j.sleep.2016.12.012. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Sleep duration and risk of stroke: a dose–response meta-analysis of prospective cohort studies
Qiao He a, b, Hao Sun a, b, Xiaomei Wu a, b, Peng Zhang a, b, Huixu Dai a, b, Cong Ai a, b,
a
RI PT
Jingpu Shi a, b, *
Department of Clinical Epidemiology, Institute of Cardiovascular Diseases and
Center of Evidence Based Medicine, The First Affiliated Hospital, China Medical
Center of Evidence Based Medicine, Liaoning Province & China Medical University,
Shenyang, China *
Corresponding author.
M AN U
b
SC
University, Shenyang, China
Department of Clinical Epidemiology,
Institute of
Cardiovascular Diseases and Center of Evidence Based Medicine, The First Affiliated Hospital, China Medical University, No.155 Nanjing Bei Street, Heping District,
TE D
Shenyang, Liaoning 110001, China.
AC C
EP
E-mail address: [email protected] (J. Shi).
1
ACCEPTED MANUSCRIPT ABSTRACT Objectives: Suboptimal sleep duration has been considered to increase the risk of stroke incidence. Thus we aimed to conduct a dose–response meta-analysis to examine the association between sleep duration and stroke incidence.
RI PT
Methods: We searched PubMed, Web of science and the Cochrane Library to identify all prospective studies evaluating the association of sleep duration and nonfatal and/or fatal stroke incidence. Then, restricted cubic spline functions and piecewise linear functions were used to evaluate the nonlinear and linear dose–response association
SC
between them.
Results: We included a total of 16 prospective studies enrolling 528,653 participants
M AN U
with 12,193 stroke events. Nonlinear dose–response meta-analysis showed a J-shaped association between sleep duration and total stroke with the lowest risk observed with sleeping for 7 h. Considering people sleeping for 7 h as reference, long sleepers had a higher predicted risk of total stroke than short sleepers (the pooled risk ratios (95% confidence intervals): 4 h: 1.17 (0.99–1.38); 5 h: 1.17 (1.00–1.37); 6 h: 1.10
TE D
(1.00–1.21); 8 h: 1.17 (1.07–1.28); 9 h: 1.45 (1.23–1.70); 10 h: 1.64 (1.4–1.92); pnonlinearity<0.001). Short sleep durations were only significantly associated with nonfatal stroke and with total stroke in the subgroups of structured interview and
EP
non-Asian countries. Additionally, we found a slightly decreased risk of ischemic stroke among short sleepers. For piecewise linear trends, compared to 7 h, every 1-h increment of sleep duration led to an increase of 13% (the pooled risk ratios (95%
AC C
confidence intervals): 1.13 (1.07–1.20); p<0.001) in risk of total stroke. Conclusion: Both in nonlinear and piecewise linear dose–response meta-analyses, long sleep duration significantly increased the risk of stroke incidence.
Keywords Stroke Sleep duration Meta-analysis Prospective studies 2
ACCEPTED MANUSCRIPT 1. Introduction Stroke is the second-leading cause of death and a leading cause of disability worldwide [1]. The Global Burden of Disease (GBD) 2013 study showed that, in 2013, there were globally almost 25.7 million stroke survivors; and over the
RI PT
1990–2013 period, there was a statistically significant increase in the absolute number of DALYs due to ischemic stroke (IS), and deaths, survivors and incident events of as a result of both IS and hemorrhagic stroke [2]. Therefore, it is important to look for the risk factors of stroke, especially those that can be
SC
ameliorated by modified lifestyles.
Sleep, an important role in human life, takes up approximately one-third of our
M AN U
lifetime. An optimal sleep duration (7–8 h) predicts fewer deaths [3]. In recent decades, suboptimal sleep duration was widely demonstrated to be associated with bad health outcomes such as obesity, diabetes, metabolic syndrome and hypertension which are risk factors of stroke [4–7]. Thus, accumulating epidemiological studies have begun to investigate whether habitual sleep duration is related to stroke. Several
TE D
primary studies have put forward a curvilinear relationship between sleep duration and stroke [8,9], where both short and long sleep duration led to an increase in the risk of stroke. But discrepancies exist in the literatures that support this association. Some
EP
conclude that either short or long sleep duration is associated with stroke [10–13], while others found that a null association exists between them [14–18]. Four meta-analyses found that both short and long sleep duration were significant
AC C
predictors of stroke [11,19–21], but most of them used a traditional two-category model, which led to massive loss of information for sleep durations in primary studies divided into more than two categories. This we conducted a dose–response meta-analysis based on prospective studies to explore the association between sleep duration and stroke incidence.
2. Materials and Methods We performed this study following the guidelines recommended by the meta-analysis of observational studies in epidemiology (MOOSE) checklist [22]. 3
ACCEPTED MANUSCRIPT 2.1 Data source and searches We searched PubMed, Web of Science and the Cochrane Library. PubMed search terms were (((stroke) or cerebral infarction) or cerebrovascular accident) and sleep duration. Similar search terms were used for Web of Science and the Cochrane
RI PT
Library (for detailed search queries, see Table S1). The search included all relevant studies published before 3 November 2016 with no language limitations. We also screened the reference lists of relevant review articles and included studies for
SC
additional information.
2.2 Study selection
M AN U
We included studies investigating the association of between sleep duration and nonfatal and/or fatal stroke incidence that met the following criteria: (1) the study design was prospective; (2) the population were adults (age > 18 years); (3) the mean or median follow-up duration was more than 2 years; (4) there were available effect estimates (risk ratio (RR), hazard ratio (HR)), and 95% confidence intervals (CIs) for
TE D
at least three quantitative categories of sleep duration; (5) subjects had no stroke history at the baseline. Furthermore, if multiple publications were available for a study,
EP
we included the report with the longest follow-up.
2.3 Data extraction and quality assessment Two of our authors extracted the following data independently and in duplicate, using
AC C
a pre-defined standardized data-extraction form: name of first authors, publication year, country of origin of the population studied, study design, follow-up period, participant characteristics (sex and age), exposure and outcome assessment, sleep duration categories, sample size (numbers of participants and incident cases ) in each category, covariates adjusted in the multivariable analysis, and effect size (RRs or HRs) with 95% CI for all categories of sleep duration. When studies had several adjusted models, we extracted those that had the maximum extent of adjustment for potentially confounding variables. The differences between these were resolved 4
ACCEPTED MANUSCRIPT through discussion and consensus with another of our authors. We used ‘hours’ as the common scale of sleep duration. Quality assessment was performed according to the Newcastle-Ottawa Quality Assessment Scale (NOS), which is a validated scale for non-randomized studies in
RI PT
meta-analyses. The NOS tool contains nine items, four for selection of participants and measurement of exposure, two for comparability of cohorts on the basis of the design or analysis, and three for assessment of outcomes and adequacy of follow-up, and each item is assigned with a star if a study meets the criteria of the item [23]. We
SC
considered a study to be of high quality if it scored more than 6.
M AN U
2.4 Statistical analysis
Firstly, we separately summarized the RRs for the shortest and longest sleep duration in each included study compared to its own reference category to explore possible association between sleep duration and total stroke incidence using the random effects meta-analysis [24]. Heterogeneity was evaluated by I2 and Cochran’s Q [25], and it
TE D
was classified as low (I2 < 25%), moderate (25% ≤ I2 < 50%) or high (≥ 50%) Potential publication bias was evaluated by funnel plots and by Egger’s test and Begg’s test [26].
EP
For dose–response meta-analysis, considering the log relative risk as an independent variable and the exposure level as a dependent variable, we first adopted restricted cubic splines with five knots at 1st, 25th, 50th, 75th, and 99th percentiles of
AC C
exposure distribution in the dose–response meta-regression model to evaluate potential nonlinear trends between sleep duration and total stroke, nonfatal stroke, and fatal stroke [27]. Generalized least-square method was used to estimate the parameters [28,29]. In each of the included studies, we assigned the reported median or mean sleep duration of each category as the category sleep duration. When a study reported only the range of sleep duration for a category, we used the average value of the lower and upper bounds of that category. When the shortest or the longest category was open-ended, we assumed that the open-ended interval length had the same length as the adjacent interval. Additionally, the number of stroke cases and participants (or 5
ACCEPTED MANUSCRIPT person-years) of each category of sleep duration was required, and these numbers were inferred based on total number of cases and the reported risk estimates if the study didn’t report them [30]. The p-value for nonlinearity was calculated by testing the null hypothesis that coefficients of the second, third and fourth spline were zero
RI PT
[29]. Then, we set 7 h as the reference category to predict RRs and 95% CI of stroke under different sleep durations [10]. Finally, if a U-, J- or S-shape or their reverse association was observed, we treated the slope as two piecewise with the cut point of 7 h to show the linear trends, respectively [31,32].
SC
If the study only reported information by sex, we considered them as two cohorts. All the regression coefficient and results were pooled in a random-effects model to
M AN U
account for heterogeneity among studies [33]. At the same time, we conducted subgroup analyses of nonlinearity and linearity based on stroke subtype, sex, location, the assessment of sleep duration, excluding studies that enrolled special populations because they were considered to be potential residual confounders. Analyses were performed with STATA version 12.0 (StataCorp, College Station, TX) and all tests
3. Results
TE D
were two-sided with a significance level of 0.05.
EP
3.1 Literature search and study characteristics A detailed description of how studies were selected is presented in Fig. 1. Briefly, from the preliminary literature search, a total of 825 titles and abstracts were
AC C
identified. After excluding 786 irrelevant records, we read 39 full-text articles for further assessment. Eventually, we identified 12 articles that met all criteria for this meta-analysis, which enrolled a total of 528,653 participants with 12,193 stroke events [10,11,13–17,34–38]. The characteristics and quality score of the individual studies are shown in Table 1. Among 12 included studies, four only reported information on separate sexes [14,17,36,37]. Thus, we eventually had 16 cohorts. The mean or median follow-up duration ranged from 7.8 to 14.7 years. The sample size ranged from 2282 to 98,634. The study population was the general population except in one study that enrolled 6
ACCEPTED MANUSCRIPT middle-aged male workers in a light metal factory of Japan [16]. The age at the entry of population ranged from 25 to 98. Two studies reported total stroke, nonfatal and fatal stroke [11,14], four fatal stroke [34–37], three nonfatal stroke [10,15,17] and three combined [13,16,38]. Four studies also supplied information about different
RI PT
stroke types [11,13,34,37]. The exposure of sleep duration in all studies was self-reported either by structured interview or self-administered questionnaires.
3.2 Shortest and longest sleep duration and the risk of total stroke
SC
The original reference category in each included study was slightly different, as were the shortest and longest sleep duration. The reference duration of five studies are
M AN U
at 7 h, two at 6–8 h, five at 7–8 h. For shortest sleep duration, six studies were set to ≤6 h, four to ≤5 h and 1 to ≤4 h. The pooled RR for the shortest sleep duration for total stroke incidence was 1.10 (95% CI: 0.97–1.24; Fig. 2) with moderate heterogeneity (I2 = 49.2%, p = 0.014). No significant publication bias was found (Begg’s test p = 0.44, Egger’s test p = 0.48). For longest sleep duration, five were set to ≥8 h, five
TE D
to ≥9 h and one to ≥10 h, and the corresponding pooled RR was 1.37 (95% CI: 1.23–1.54; Fig. 2) with significant heterogeneity (I2 = 55.7%, p = 0.004) and no
EP
publication bias exsited (Begg’s test p = 1.00, Egger’s test p = 0.62).
3.3 The nonlinear association between sleep duration and stroke Figures 3–5 show the nonlinear curves between self-reported sleep duration and
AC C
total stroke, nonfatal stroke, and fatal stroke, respectively. A J-shaped association was found between sleep duration and total stroke incidence with lowest risk observed at sleeping for 7 h. When we considered 7 h as the reference category, the risk of total stroke was higher among long sleepers than that among short sleepers. The pooled RRs (95% CI) were as follows: 4 h: 1.17 (0.99–1.38); 5 h: 1.17 (1.00–1.37); 6 h: 1.10 (1.00–1.21); 8 h: 1.17 (1.07–1.28); 9 h: 1.45 (1.23–1.70); 10 h: 1.64 (1.4–1.92) (pnonlinearity<0.001, Table 2). For nonfatal stroke, a U-shaped association including five cohorts was observed with borderline significance of nonlinearity, and the predict RRs were as follows: 5 h: 1.43 (1.13–1.80); 6 h: 1.30 (1.09–1.54); 8 h: 1.08 (0.98–1.20); 7
ACCEPTED MANUSCRIPT 9 h: 1.51 (1.15–1.99) (pnonlinearity = 0.05, Table 2). The trend of fatal stroke was similar to that of total stroke and we only found a significantly increased risk among long sleepers (4 h: 1.08 (0.91–1.29); 5 h: 1.09 (0.92–1.28); 6 h: 1.05 (0.95–1.16); 8 h: 1.15 (1.05–1.26); 9 h: 1.43 (1.19–1.71); 10 h: 1.65 (1.36–2.00); pnonlinearity<0.001, Table 2).
RI PT
The nonlinear analyses of subgroups are shown in Table 2. J-shaped associations were found in most subgroups, and the risk for short sleepers did not reach significance
except
in
the
subgroups
of
non-Asian
countries
with
the
structured-interview. Additionally, we observed a slightly protective effect of IS
SC
among short sleepers compared to people sleeping for 7 h, though it was nonsignificant (4 h: 0.92 (0.70–1.21); 5 h: 0.91 (0.70–1.19); 6 h: 0.94 (0.82–1.07)).
M AN U
After excluding the study by Hamazaki et al. [16], which enrolled a special population (workers), the results compared to that of overall analysis attenuated slightly.
3.4 The linear association between sleep duration and strokes
All nonlinear analyses found a J-shaped or U-shaped association between sleep
TE D
duration and stroke except the IS one, so we treated the slope as two piecewise with the cut point of 7 h to show the linear trends, respectively. Among people who slept more than 7 h, every 1 h of sleep duration increment led to an increase of 13% (RR
EP
(95% CI) = 1.13 (1.07–1.20); p<0.001) in risk of total stroke and 12% (RR (95% CI) = 1.12 (1.04–1.21)) of fatal stroke. Sleeping less than 7 h did not show a statistically significant risk of each stroke outcome. The risk by sleeping more than
AC C
7 h were also significant in most of subgroups (Table 3).
4. Discussion
In this dose–response meta-analysis of 16 prospective studies, we found a J-shaped
association between sleep duration and total stroke incidence, and long sleepers had a significantly higher risk of total, nonfatal, and fatal stroke incidence compared to people sleeping for 7 h. Short sleep durations were only significantly related with nonfatal stroke, and with total stroke in subgroups of structured interview and non-Asian countries. Additionally, we found a slightly decreased risk of IS among 8
ACCEPTED MANUSCRIPT short sleepers. For piecewise linear analyses, we also found a positive linear dose–response association among long sleepers. Every 1-h increase in sleep duration led to an increased risk of 13% and 12% of total stroke and fatal stroke, respectively.
RI PT
Long sleep duration might be a predictor for future stroke.
4.1 Comparison with previous studies
Several epidemiological studies have explored this association. A meta-analysis by Ge showed significantly increased risk of stroke incidence and mortality at either end of
SC
the distribution of sleep duration in both cohort and cross-sectional studies; and RR for long sleep is stronger than that for short sleep. Their subgroup analysis also
M AN U
showed that only long sleep duration was a statistical stroke risk in both sexes and in Asians [19]. Those results were in accordance with ours. Leng identified prolonged sleep as a potentially useful marker of increased stroke risk in an apparently healthy aging population [11]. In addition, some cross-sectional studies found the same results. Akinseys using the data of the National Health Interview Survey (2004–2013) found
TE D
that the prevalence of stroke in diabetes was 16.1% in long sleepers, higher than that at 8.3% in normative sleepers (p<0.01), but only males showed a continued association between long sleep and stroke when compared to females, which was
EP
different to our findings [1]. Some studies such as Westerlund et al. [15] found no harmful influence of sleep duration on stroke incidence. Our study on fatal stroke did not show a significant result for short sleep duration.
AC C
Studies on this association were relatively scarce. The Singapore Chinese Health Study, a population-based cohort of 63,257 Chinese adults, found that both short and long sleep durations were associated with increased risk of stroke mortality [35]. However, another three other studies, conducted in Japan [37], Los Angeles/Hawaii [36], and China [39], each comprising more than 95,000 participants, showed the same outcome as ours—only long sleep duration was significantly associated with a higher risk of stroke mortality in both sexes. A possible explanation is that the bulk of the population in the long sleep duration group consisted of elderly people, who experience more unhealthy status and have more comorbidities with advancing age; 9
ACCEPTED MANUSCRIPT thus they generally have a higher chance of suffering from a more severe stroke. There were five studies reporting the association between the subtypes of stroke and sleep duration in our meta-analysis, and four supplied eligible data that we included in our subgroup analysis. An increased risk of IS was found among long
RI PT
sleepers in three studies, which was borderline significant in ours. Additionally, we observed a weak decreased risk of IS incidence among people sleeping less than 7 h, though it was not significant. It was somewhat consistent with the study by Kawachi et al.; and they also found a significant risk reduction of hemorrhagic stroke mortality
SC
for less than 7 h of sleep which was contrary with ours. We suggested that the effect of sleep duration on IS is different from that on HS, but the relevant data were few. More
M AN U
researches need to be conducted to examine this association.
4.2 Potential mechanisms under this association
The underlying mechanisms are not fully understood. Sleep deprivation has been widely demonstrated to be linked to elevated ghrelin levels, reduced leptin levels [40],
TE D
disrupted lipid metabolism [41], high blood glucose [6], et cetera. Covassin revealed that compared to normal sleep, sleep restriction increased nocturnal blood pressure and attenuated percentage of blood pressure dipping [42]. The risk for long sleep
EP
duration was unclear; some researchers believe that excessive sleep might increase some inflammatory biomarkers, such as C-reactive protein and interleukin-6 [43]. The China Health and Nutrition Survey, selecting individuals from 228 communities
AC C
found that shorter (<6 h) or longer (>10 h) sleep durations were associated with higher risks of abnormal lipid profiles in female [41]. Long sleep duration was also reported to be associated with carotid artery atherosclerosis [44], atrial fibrillation [45,46], white matter hyperintensity volume [47], and left ventricular mass [48], which might increase an individual’s risk of stroke.
4.3 Limitations and strengths This study has some limitations. The assessment of sleep duration in all included studies was self-reported, and thus could lead to an overestimation of the true sleep 10
ACCEPTED MANUSCRIPT duration because the subjects might consider awake time in bed as a part of sleep time, though a study has indicated that sleep duration assessed by self-reporting is highly related to that by actigraph or polysomnography [49]. In addition, habitual sleep durations were collected from a single question, and subjects might change their sleep
RI PT
patterns during their follow-up periods. However, Leng’s study indicated that the results for combined sleep duration of the baseline and follow-up were in line with their results on a single measure of sleep at baseline [11]. We used five knots in our restricted cubic functions taking into consideration our relatively large sample size
SC
and at least 4 categories of sleep duration available in most studies. Although Stone suggested that five knots may provide enough flexibility for a reasonable number of
M AN U
degrees of freedom, we cannot deny that the variance of the spline estimator and the risk of overfitting increase with knots, especially in the analysis of subtype stroke with relatively small sample size [27]. Finally, several studies have reported that sleep quality was associated with the risk of stroke [3], but only three studies in our meta-analysis included presence of snoring or sleep disorders as confounders
TE D
[10,13,15]. Two of the three studies showed a continued association between long sleep duration and stroke after snore status was adjusted, though the results by von Ruestern et al. became nonsignificant after taking body mass index (BMI), waist-to-hip ratio and history of high blood lipid levels into the model based on the
EP
snore-adjusted model [10,13]3. The third study did not find an association between sleep duration and stroke either before or after snore adjustment [15], but the
AC C
proportion of people with snoring habits or sleep disorders was higher in the short sleep duration group in all three studies. Perhaps worse sleep quality is related to shorter sleep duration rather than long sleep duration. However, the sleep disorders in long sleepers should not be ignored, and studies should take these into account when they assess the association between sleep duration and stroke incidence. Our study has two major strengths. Firstly, we evaluated the potential nonlinear and linear association between sleep duration and stroke incidence using a restricted cubic spline and a piecewise linear function, respectively. Previous meta-analyses on this association were mostly based on a two-category model, losing much information on 11
ACCEPTED MANUSCRIPT different categories of exposure and having lower statistical power. Secondly, the individual studies included were prospective and all subjects had no stroke history at baseline, so the causal association was somewhat rigorous.
RI PT
5. Conclusion In conclusion, this dose–response meta-analysis showed a J-shaped association between sleep duration and total stroke incidence. Long sleep duration was served as a higher risk than short sleep duration. Additionally, we found a positive linear trend
SC
among long sleepers that, compared to 7 h sleep duration, every 1-h additional increment in sleep duration led to an increase of 13% in the risk of total stroke. Long
need further investigation.
Conflicts of interest
M AN U
sleep duration might be a useful predictor of stroke risk. The underlying mechanisms
Acknowledgements
TE D
The authors have no conflicts of interest to declare.
References 1.
EP
We thank Mr Guangxiao Li for his guidance.
Akinseye OA, Ojike NI, Akinseye LI, Dhandapany PS, Pandi-Perumal SR.
AC C
Association of sleep duration with stroke in diabetic patients: Analysis of the national health interview survey. Journal of stroke and cerebrovascular diseases 2016;25:650–5
2.
Feigin VL, Krishnamurthi RV, Parmar P, Norrving B, Mensah GA, Bennett DA, et al. Update on the global burden of ischemic and hemorrhagic stroke in 1990–2013: The gbd 2013 study. Neuroepidemiology. 2015;45:161–76
3.
Chien KL, Chen PC, Hsu HC, Su TC, Sung FC, Chen MF, et al. Habitual sleep duration and insomnia and the risk of cardiovascular events and all-cause death: Report from a community-based cohort. Sleep. 2010;33:177–84 12
ACCEPTED MANUSCRIPT 4.
Theorell-Haglow J, Berglund L, Berne C, Lindberg E. Both habitual short sleepers and long sleepers are at greater risk of obesity: A population-based 10-year follow-up in women. Sleep medicine. 2014;15:1204–11
5.
Lou P, Zhang P, Zhang L, Chen P, Chang G, Zhang N, et al. Effects of sleep
RI PT
duration and sleep quality on prevalence of type 2 diabetes mellitus: A 5-year follow-up study in china. Diabetes research and clinical practice. 2015;109:178–84 6.
Kim JY, Yadav D, Ahn SV, Koh SB, Park JT, Yoon J, et al. A prospective study
SC
of total sleep duration and incident metabolic syndrome: The arirang study. Sleep medicine. 2015;16:1511–15
Wang T, Lu J, Wang W, Mu Y, Zhao J, Liu C, et al. Sleep duration and snoring
M AN U
7.
associate with hypertension and glycaemic control in patients with diabetes. Diabetic medicine 2015;32:1001–7 8.
Fang J, Wheaton AG, Ayala C. Sleep duration and history of stroke among adults from the USA. Journal of sleep research. 2014;23:531–7 Liu Y, Wheaton AG, Chapman DP, Croft JB. Sleep duration and chronic
TE D
9.
diseases among u.S. Adults age 45 years and older: Evidence from the 2010 behavioral risk factor surveillance system. Sleep. 2013;36:1421–7 von Ruesten A, Weikert C, Fietze I, Boeing H. Association of sleep duration
EP
10.
with chronic diseases in the european prospective investigation into cancer and nutrition (epic)-potsdam study. PloS one. 2012;7:e30972 Leng Y, Cappuccio FP, Wainwright NW, Surtees PG, Luben R, Brayne C, et al.
AC C
11.
Sleep duration and risk of fatal and nonfatal stroke: A prospective study and meta-analysis. Neurology. 2015;84:1072–9
12.
Chen JC, Brunner RL, Ren H, Wassertheil-Smoller S, Larson JC, Levine DW, et al. Sleep duration and risk of ischemic stroke in postmenopausal women. Stroke; a journal of cerebral circulation. 2008;39:3185–92
13.
Song Q, Liu X, Zhou W, Wang L, Zheng X, Wang X, et al. Long sleep duration and risk of ischemic stroke and hemorrhagic stroke: The kailuan prospective study. Scientific reports. 2016;6:33664 13
ACCEPTED MANUSCRIPT 14.
Helbig AK, Stockl D, Heier M, Ladwig KH, Meisinger C. Symptoms of insomnia and sleep duration and their association with incident strokes: Findings from the population-based monica/kora augsburg cohort study. PloS one. 2015;10:e0134480 Westerlund A, Bellocco R, Sundstrom J, Adami HO, Akerstedt T, Trolle
RI PT
15.
Lagerros Y. Sleep characteristics and cardiovascular events in a large swedish cohort. European journal of epidemiology. 2013;28:463–73 16.
Hamazaki Y, Morikawa Y, Nakamura K, Sakurai M, Miura K, Ishizaki M, et al.
SC
The effects of sleep duration on the incidence of cardiovascular events among middle-aged male workers in japan. Scandinavian journal of work,
17.
M AN U
environment & health. 2011;37:411–17
Amagai Y, Ishikawa S, Gotoh T, Kayaba K, Nakamura Y, Kajii E. Sleep duration and incidence of cardiovascular events in a japanese population: The jichi medical school cohort study. Journal of Epidemiology. 2010;20:106–10
18.
Amagai Y, Ishikawa S, Gotoh T, Doi Y, Kayaba K, Nakamura Y, et al. Sleep
TE D
duration and mortality in japan: The jichi medical school cohort study. Journal of epidemiology / Japan Epidemiological Association. 2004;14:124–8 19.
Ge B, Guo X. Short and long sleep durations are both associated with
EP
increased risk of stroke: A meta-analysis of observational studies. International journal of stroke : official journal of the International Stroke Society. 2015;10:177–84 Cappuccio FP, Cooper D, D’Elia L, Strazzullo P, Miller MA. Sleep duration
AC C
20.
predicts cardiovascular outcomes: A systematic review and meta-analysis of prospective studies. European heart journal. 2011;32:1484–92
21.
Li W, Wang D, Cao S, Yin X, Gong Y, Gan Y, et al. Sleep duration and risk of stroke events and stroke mortality: A systematic review and meta-analysis of prospective
cohort
studies.
International
journal
of
cardiology.
2016;223:870–6 22.
Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of observational studies in epidemiology - a proposal for 14
ACCEPTED MANUSCRIPT reporting. Jama-J Am Med Assoc. 2000;283:2008–12 23.
Wells GA, Shea B, O'Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies
in
meta-analyses.
2011.
Available
at:
RI PT
http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. accessed in Sep 2014. 24.
DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled clinical trials. 1986;7:177–88
Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. Bmj. 2003;327:557–60
Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis
M AN U
26.
SC
25.
detected by a simple, graphical test. Bmj. 1997;315:629–34 27.
Durrleman S, Simon R. Flexible regression models with cubic splines. Statistics in medicine. 1989;8:551–61
28.
Orsini N, Li R, Wolk A, Khudyakov P, Spiegelman D. Meta-analysis for linear
TE D
and nonlinear dose–response relations: Examples, an evaluation of approximations,
and
software.
American
journal
of
epidemiology.
2012;175:66–73
Greenland S, Longnecker MP. Methods for trend estimation from summarized
EP
29.
dose–response data, with applications to meta-analysis. American journal of epidemiology. 1992;135:1301–9 Bekkering GE, Harris RJ, Thomas S, Mayer AM, Beynon R, Ness AR, et al.
AC C
30.
How much of the data published in observational studies of the association between diet and prostate or bladder cancer is usable for meta-analysis? American journal of epidemiology. 2008;167:1017–26
31.
Gould WW. Linear splines and piecewise linear functions. Stata Tech Bull 1993;15:13e7.
32.
Yun Y, Zhang C, Weng H, Kuang X, Gong T, Lu Y, et al. How to Conduct Dose–response meta-analysis by using linear relation and piecewise linear regression
model.
Chinese
Journal 15
of
Evidence-based
Medicine.
ACCEPTED MANUSCRIPT 2016;16;111–4 33.
Berlin JA, Longnecker MP, Greenland S. Meta-analysis of epidemiologic dose–response data. Epidemiology. 1993;4:218–28
34.
Kawachi T, Wada K, Nakamura K, Tsuji M, Tamura T, Konishi K, et al. Sleep
RI PT
duration and the risk of mortality from stroke in japan: The takayama cohort study. Journal of epidemiology / Japan Epidemiological Association. 2016;26:123–30 35.
Pan A, De Silva DA, Yuan JM, Koh WP. Sleep duration and risk of stroke
SC
mortality among chinese adults: Singapore chinese health study. Stroke; a journal of cerebral circulation. 2014;45:1620–5
Kim Y, Wilkens LR, Schembre SM, Henderson BE, Kolonel LN, Goodman
M AN U
36.
MT. Insufficient and excessive amounts of sleep increase the risk of premature death from cardiovascular and other diseases: The multiethnic cohort study. Preventive medicine. 2013;57:377–85 37.
Ikehara. Association of sleep duration with mortality from cardiovascular
38.
TE D
disease and other causes for japanese men and women: The jacc study. 2009 Qureshi AI, Giles WH, Croft JB, Bliwise DL. Habitual sleep patterns and risk for stroke and coronary heart disease: A 10-year follow-up from nhanes i.
39.
EP
Neurology. 1997;48:904–911
Cai H, Shu XO, Xiang YB, Yang G, Li H, Ji BT, et al. Sleep duration and mortality: A prospective study of 113 138 middle-aged and elderly chinese
AC C
men and women. Sleep. 2015;38:529–36
40.
Taheri S, Lin L, Austin D, Young T, Mignot E. Short sleep duration is
associated with reduced leptin, elevated ghrelin, and increased body mass index. PLoS medicine. 2004;1:e62
41.
Zhan Y, Chen R, Yu J. Sleep duration and abnormal serum lipids: The china health and nutrition survey. Sleep medicine. 2014;15:833–39
42.
Covassin N, Bukartyk J, Sahakyan K, Svatikova A, Calvin A, St Louis EK, et al. Experimental sleep restriction increases nocturnal blood pressure and attenuates blood pressure dipping in healthy individuals. Journal of the 16
ACCEPTED MANUSCRIPT American College of Cardiology. 2015;65:A1352 43.
Prather AA, Vogelzangs N, Penninx BW. Sleep duration, insomnia, and markers of systemic inflammation: Results from the netherlands study of depression
and
anxiety
(nesda).
Journal
of
psychiatric
research.
44.
RI PT
2015;60:95–102 Abe T, Aoki T, Yata S, Okada M. Sleep duration is significantly associated with carotid artery atherosclerosis incidence in a japanese population. Atherosclerosis. 2011;217:509–13
Khawaja O, Sarwar A, Albert CM, Gaziano JM, Djousse L. Sleep duration and
SC
45.
risk of atrial fibrillation (from the physicians' health study). The American
46.
M AN U
journal of cardiology. 2013;111:547–51
Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: The framingham study. Stroke; a journal of cerebral circulation. 1991;22:983–88
47.
Ramos AR, Dong C, Rundek T, Elkind MS, Boden-Albala B, Sacco RL, et al.
TE D
Sleep duration is associated with white matter hyperintensity volume in older adults: The northern manhattan study. Journal of sleep research. 2014;23:524–30
Ramos AR, Jin Z, Rundek T, Russo C, Homma S, Elkind MS, et al. Relation
EP
48.
between long sleep and left ventricular mass (from a multiethnic elderly cohort). The American journal of cardiology. 2013;112:599–603 Signal TL, Gale J, Gander PH. Sleep measurement in flight crew: Comparing
AC C
49.
actigraphic and subjective estimates to polysomnography. Aviation, space, and environmental medicine. 2005;76:1058–63
17
ACCEPTED MANUSCRIPT Fig. 1. Flow chart showing the study selection process. Fig. 2. Forest plots of the association between shortest, longest sleep duration sleep duration and total stroke. (a) Shortest sleep duration; (b) longest sleep duration. I2, heterogeneity index; RR, risk ratios; I2, heterogeneity index; m and w represent
RI PT
different cohorts with separate data in a single study. Fig. 3. The nonlinear association between sleep duration and total stroke incidence. Fig. 4. The nonlinear association between sleep duration and nonfatal stroke incidence.
AC C
EP
TE D
M AN U
SC
Fig. 5. The nonlinear association between sleep duration and fatal stroke incidence.
18
ACCEPTED MANUSCRIPT
Song [13]
year 2016
Country
China
Sex
Both
Sample
Follow-up
size
years
95,023
7.9
Age at
Assessment of
baseline (range/mean)
sleep duration
Covariates in fully adjusted model
SC
Author
Publication
RI PT
Table 1. The characteristics of studies included in this meta-analysis.
18–98/51.2
Structured interview
Age, sex, BMI, SBP, DBP, total cholesterol
NOS (0–9 points) 9
Kawachi
2016
Japan
Both
27,896
14.2
2015
Great Britain
Both
9692
9.5
AC C
Leng [11]
35–97/NA
EP
[34]
TE D
M AN U
level, FBG, high sensitive C-reactive protein, smoking status, drinking consumption, physical activity, history of MI or atrial fibrillation, family history of stroke, snoring status, marital status, education level, family per member monthly income, hypotensive drug use, lipid-lowering drug use, hypoglycemic drug use, stratified by hospitals
Self-administered questionnaires
age, sex, BMI, smoking status, alcohol
7
consumption, physical activity score, histories of hypertension or diabetes, education years, marital status.
42–81/61.6
Structured interview
Age, sex, BMI, SBP, DBP, cholesterol level, smoking, alcohol consumption, physical activity, histories of diabetes or MI, family history of stroke, depression, education, marital status, social class, hypnotic drug use, hypnotic drug use, hypertension drug use
19
8
ACCEPTED MANUSCRIPT
Helbig [14]
2015
Germany
Both
12,131
14
25–74/NA
Structured interview
Age, BMI, dyslipidemia, current smoking
7
RI PT
activity, alcohol consumption, physical activity, histories of hypertension and diabetes, education, survey,
Pan [35]
2014
China
Both
63,257
14.7
45–74/NA
Structured interview
Age, sex, BMI, years of smoking, dose of
7
SC
smoking, alcohol consumption, histories of hypertension, diabetes, stroke, CHD, or cancer,
Westerlund
2013
Sweden
Both
41,192
M AN U
education, dialect, energy intake, dietary
13.2
NA/52.8
2013
USA
Both
135,685
12.9
45–75/NA
Ruesten [10]
2012
Europe
Both
AC C
von
23,620
7.8
polyunsaturated fatty acids Age, sex, BMI, lipid disturbance, smoking,
8
alcohol consumption, physical activity, histories of diabetes or hypertension, depressive symptoms, snoring, education, employment status, work schedule, self-rated health
Self-administered questionnaires
Age, sex, BMI, smoking history, alcohol
6
consumption, physical activity, hours spent daily watching television, histories of
EP
Kim [36]
questionnaires
TE D
[15]
Self-administered
intakes of vegetables, fruits, fiber,
hypertension or diabetes, education, marital status, ethnicity, energy intake
35–65/NA
Structured interview
Age, sex, smoking status, alcohol consumption, walking, cycling, sports, sleeping disorders, education, employment status
20
8
ACCEPTED MANUSCRIPT
Hamazaki
2011
Japan
Male
2282
14
35–54/43.7
Age, BMI, type of job, total cholesterol level,
questionnaires
mean blood pressure, HbA1c, smoking status,
6
RI PT
[16]
Self-administered
alcohol consumption, physical activity, working hours, mental workload, medication for hypertension, diabetes, or
Amagai
2010
USA
Both
11,367
10.7
18–90/
[17]
Structured interview
male, 55.1;
Japan
Both
98,634
M AN U
2009
14.3
40–79/NA
Self-administered
1997
USA
Both
[38]
7844
10
TE D
questionnaires
Qureshi
Age, BMI, SBP, total cholesterol level,
7
smoking, alcohol consumption.
female, 55.3 Ikehara [37]
SC
hypercholesterolemia.
31–74/NA
Structured interview
Age, BMI, smoking, alcohol consumption,
7
hours of exercise, hours of walking, histories of hypertension or diabetes, depressive symptoms, perceived mental stress, education, employment, frequency of fresh fish intake Age, sex, BMI, SBP, cholesterol level,
7
smoking, history of diabetes, education, race
AC C
Assessment Scale; SBP, systolic blood pressure.
EP
BMI, body mass index; CHD, coronary heart disease; DBP, diastolic blood pressure; FBG, fast blood glucose; MI, myocardial infarction; NA, not available; NOS, Newcastle-Ottawa Quality
21
ACCEPTED MANUSCRIPT
RI PT
Table 2. The nonlinear dose–response association between sleep duration and stroke incidence.
Stroke risk (RR (95% CI)) 5 h
6 h
7 h
1.17 (0.99–1.38)
1.17 (1.00–1.37)
1.10 (1.00–1.21)
1
—
1.43 (1.13–1.80)
1.30 (1.09–1.54)
1.08 (0.91–1.29)
1.09 (0.92–1.28)
1.05 (0.95–1.16)
ischemic stroke
0.92 (0.70–1.21)
0.91 (0.70–1.19)
0.94 (0.82–1.07)
hemorrhagic stroke
1.02 (0.71–1.47)
1.01 (0.71–1.45)
Male
1.03 (0.80–1.33)
Female
Total stroke
8 h
9 h
10 h
1.17 (1.07–1.28)
1.45 (1.23–1.70)
1.64 (1.40–1.92)
1.08 (0.98–1.20)
1.51 (1.15–1.99)
—
1.15 (1.05–1.26)
1.43 (1.19–1.71)
1.65 (1.36–2.00)
1
1.09 (0.94–1.28)
1.31 (0.98–1.76)
1.79 (1.39–2.32)
0.98 (0.82–1.17)
1
1.18 (0.96–1.46)
1.33 (0.91–1.94)
1.18 (0.82–1.68)
1.02 (0.81–1.30)
1.01 (0.87–1.17)
1
1.10 (0.97–1.24)
1.30 (1.02–1.65)
1.58 (1.24–2.01)
1.07 (0.86–1.32)
1.07 (0.86–1.33)
1.04 (0.90–1.19)
1
1.19 (1.07–1.34)
1.46 (1.15–1.85)
1.46 (1.16–1.83)
1.11 (0.88–1.42)
1.11 (0.88–1.39)
1.06 (0.92–1.21)
1
1.19 (1.03–1.36)
1.49 (1.19–1.86)
1.76 (1.43–2.15)
—
1.28 (1.06–1.54)
1.19 (1.04–1.36)
1
1.16 (1.06–1.28)
1.49 (1.20–1.85)
—
—
1.25 (1.02–1.54)
1.18 (1.03–1.35)
1
1.10 (1.02–1.19)
1.50 (1.20–1.87)
—
1.14 (0.86–1.50)
1.13 (0.86–1.50)
1.07 (0.90–1.27)
1
1.20 (1.00–1.46)
1.49 (1.16–1.91)
1.78 (1.44–2.20)
1.08 (0.99–1.18)
1
1.14 (1.06–1.22)
1.41 (1.21–1.63)
1.60 (1.37–1.87)
nonfatal stroke fatal stroke a
Stroke subtype
Asian countries Non-Asian countries
Structured interview Self-administered questionnaires a
Others
Excluding the study of Hamazaki [16] CI: confidence interval; RR: risk ratio. a
AC C
Sleep duration assessment
a
TE D
Locationa
EP
Sex
a
1.14 (0.98–1.31)
1.14 (0.99–1.30)
1 1
M AN U
Stroke outcome
SC
4 h
Risk of total stroke.
22
ACCEPTED MANUSCRIPT Table 3 The piecewise linear dose–response association between sleep duration and stroke incidence Stroke risk (RR (95% CI)) Sleep less than 7 ha
Sleep more than 7 hb
1.07(1.00,1.14)
1.13 (1.07, 1.20)
Nonfatal stroke
1.08(0.95,1.23)
1.08 (0.98, 1.18)
Fatal stroke
1.05(0.97,1.15)
1.12 (1.04, 1.21)
—
—
1.04(0.90–1.19)
1.10 (0.96–1.26)
1.11(0.99, 1.24)
1.15 (1.04, 1.27)
Total stroke
Stroke subtype
c
Ischemic stroke Hemorrhagic stroke
Male Female Location
SC
Sex
c
0.99 (0.89, 1.10)
1.13 (1.03, 1.24)
c
1.10 (1.00, 1.21)
Non-Asian countries Exposure assessment
1.03 (0.94 ,1.13)
1.08 (1.00, 1.16)
1.07 (0.96, 1.19)
1.09 (1.00, 1.18)
1.07 (0.99, 1.16)
1.21 (1.12, 1.31)
1.06 (1.00, 1.13)
1.13 (1.07, 1.19)
c
Structured interview Self-administered Questionnaires Othersc
TE D
Excluded the study of Hamazaki [16]
1.19 (1.09, 1.30)
M AN U
Asian countries
RR, risk ratio; CI, confidence interval. a
RI PT
Stroke outcomes
RRs were corresponded to every 1-h decrease of sleep duration. RRs were corresponded to every 1-h increase of sleep duration.
c
Risk of total stroke.
AC C
EP
b
23
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT Highlights ● Restricted cubic spline functions were conducted to examine the nonlinear dose–response association between sleep duration and stroke risk.
association between sleep duration and stroke risk.
RI PT
● Piecewise linear functions were conducted to examine the linear dose–response
AC C
EP
TE D
M AN U
higher risk of fatal and nonfatal stroke incidence.
SC
● Compared to people who sleep for 7 h, people who sleep for more than 7 h have a
S1389-9457(16)30329-X
DOI:
10.1016/j.sleep.2016.12.012
Reference:
SLEEP 3263
To appear in:
Sleep Medicine
Received Date: 8 September 2016 Revised Date:
17 November 2016
Accepted Date: 5 December 2016
Please cite this article as: He Q, Sun H, Wu X, Zhang P, Dai H, Ai C, Shi J, Sleep duration and risk of stroke: a dose–response meta-analysis of prospective cohort studies, Sleep Medicine (2017), doi: 10.1016/j.sleep.2016.12.012. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT Sleep duration and risk of stroke: a dose–response meta-analysis of prospective cohort studies
Qiao He a, b, Hao Sun a, b, Xiaomei Wu a, b, Peng Zhang a, b, Huixu Dai a, b, Cong Ai a, b,
a
RI PT
Jingpu Shi a, b, *
Department of Clinical Epidemiology, Institute of Cardiovascular Diseases and
Center of Evidence Based Medicine, The First Affiliated Hospital, China Medical
Center of Evidence Based Medicine, Liaoning Province & China Medical University,
Shenyang, China *
Corresponding author.
M AN U
b
SC
University, Shenyang, China
Department of Clinical Epidemiology,
Institute of
Cardiovascular Diseases and Center of Evidence Based Medicine, The First Affiliated Hospital, China Medical University, No.155 Nanjing Bei Street, Heping District,
TE D
Shenyang, Liaoning 110001, China.
AC C
EP
E-mail address: [email protected] (J. Shi).
1
ACCEPTED MANUSCRIPT ABSTRACT Objectives: Suboptimal sleep duration has been considered to increase the risk of stroke incidence. Thus we aimed to conduct a dose–response meta-analysis to examine the association between sleep duration and stroke incidence.
RI PT
Methods: We searched PubMed, Web of science and the Cochrane Library to identify all prospective studies evaluating the association of sleep duration and nonfatal and/or fatal stroke incidence. Then, restricted cubic spline functions and piecewise linear functions were used to evaluate the nonlinear and linear dose–response association
SC
between them.
Results: We included a total of 16 prospective studies enrolling 528,653 participants
M AN U
with 12,193 stroke events. Nonlinear dose–response meta-analysis showed a J-shaped association between sleep duration and total stroke with the lowest risk observed with sleeping for 7 h. Considering people sleeping for 7 h as reference, long sleepers had a higher predicted risk of total stroke than short sleepers (the pooled risk ratios (95% confidence intervals): 4 h: 1.17 (0.99–1.38); 5 h: 1.17 (1.00–1.37); 6 h: 1.10
TE D
(1.00–1.21); 8 h: 1.17 (1.07–1.28); 9 h: 1.45 (1.23–1.70); 10 h: 1.64 (1.4–1.92); pnonlinearity<0.001). Short sleep durations were only significantly associated with nonfatal stroke and with total stroke in the subgroups of structured interview and
EP
non-Asian countries. Additionally, we found a slightly decreased risk of ischemic stroke among short sleepers. For piecewise linear trends, compared to 7 h, every 1-h increment of sleep duration led to an increase of 13% (the pooled risk ratios (95%
AC C
confidence intervals): 1.13 (1.07–1.20); p<0.001) in risk of total stroke. Conclusion: Both in nonlinear and piecewise linear dose–response meta-analyses, long sleep duration significantly increased the risk of stroke incidence.
Keywords Stroke Sleep duration Meta-analysis Prospective studies 2
ACCEPTED MANUSCRIPT 1. Introduction Stroke is the second-leading cause of death and a leading cause of disability worldwide [1]. The Global Burden of Disease (GBD) 2013 study showed that, in 2013, there were globally almost 25.7 million stroke survivors; and over the
RI PT
1990–2013 period, there was a statistically significant increase in the absolute number of DALYs due to ischemic stroke (IS), and deaths, survivors and incident events of as a result of both IS and hemorrhagic stroke [2]. Therefore, it is important to look for the risk factors of stroke, especially those that can be
SC
ameliorated by modified lifestyles.
Sleep, an important role in human life, takes up approximately one-third of our
M AN U
lifetime. An optimal sleep duration (7–8 h) predicts fewer deaths [3]. In recent decades, suboptimal sleep duration was widely demonstrated to be associated with bad health outcomes such as obesity, diabetes, metabolic syndrome and hypertension which are risk factors of stroke [4–7]. Thus, accumulating epidemiological studies have begun to investigate whether habitual sleep duration is related to stroke. Several
TE D
primary studies have put forward a curvilinear relationship between sleep duration and stroke [8,9], where both short and long sleep duration led to an increase in the risk of stroke. But discrepancies exist in the literatures that support this association. Some
EP
conclude that either short or long sleep duration is associated with stroke [10–13], while others found that a null association exists between them [14–18]. Four meta-analyses found that both short and long sleep duration were significant
AC C
predictors of stroke [11,19–21], but most of them used a traditional two-category model, which led to massive loss of information for sleep durations in primary studies divided into more than two categories. This we conducted a dose–response meta-analysis based on prospective studies to explore the association between sleep duration and stroke incidence.
2. Materials and Methods We performed this study following the guidelines recommended by the meta-analysis of observational studies in epidemiology (MOOSE) checklist [22]. 3
ACCEPTED MANUSCRIPT 2.1 Data source and searches We searched PubMed, Web of Science and the Cochrane Library. PubMed search terms were (((stroke) or cerebral infarction) or cerebrovascular accident) and sleep duration. Similar search terms were used for Web of Science and the Cochrane
RI PT
Library (for detailed search queries, see Table S1). The search included all relevant studies published before 3 November 2016 with no language limitations. We also screened the reference lists of relevant review articles and included studies for
SC
additional information.
2.2 Study selection
M AN U
We included studies investigating the association of between sleep duration and nonfatal and/or fatal stroke incidence that met the following criteria: (1) the study design was prospective; (2) the population were adults (age > 18 years); (3) the mean or median follow-up duration was more than 2 years; (4) there were available effect estimates (risk ratio (RR), hazard ratio (HR)), and 95% confidence intervals (CIs) for
TE D
at least three quantitative categories of sleep duration; (5) subjects had no stroke history at the baseline. Furthermore, if multiple publications were available for a study,
EP
we included the report with the longest follow-up.
2.3 Data extraction and quality assessment Two of our authors extracted the following data independently and in duplicate, using
AC C
a pre-defined standardized data-extraction form: name of first authors, publication year, country of origin of the population studied, study design, follow-up period, participant characteristics (sex and age), exposure and outcome assessment, sleep duration categories, sample size (numbers of participants and incident cases ) in each category, covariates adjusted in the multivariable analysis, and effect size (RRs or HRs) with 95% CI for all categories of sleep duration. When studies had several adjusted models, we extracted those that had the maximum extent of adjustment for potentially confounding variables. The differences between these were resolved 4
ACCEPTED MANUSCRIPT through discussion and consensus with another of our authors. We used ‘hours’ as the common scale of sleep duration. Quality assessment was performed according to the Newcastle-Ottawa Quality Assessment Scale (NOS), which is a validated scale for non-randomized studies in
RI PT
meta-analyses. The NOS tool contains nine items, four for selection of participants and measurement of exposure, two for comparability of cohorts on the basis of the design or analysis, and three for assessment of outcomes and adequacy of follow-up, and each item is assigned with a star if a study meets the criteria of the item [23]. We
SC
considered a study to be of high quality if it scored more than 6.
M AN U
2.4 Statistical analysis
Firstly, we separately summarized the RRs for the shortest and longest sleep duration in each included study compared to its own reference category to explore possible association between sleep duration and total stroke incidence using the random effects meta-analysis [24]. Heterogeneity was evaluated by I2 and Cochran’s Q [25], and it
TE D
was classified as low (I2 < 25%), moderate (25% ≤ I2 < 50%) or high (≥ 50%) Potential publication bias was evaluated by funnel plots and by Egger’s test and Begg’s test [26].
EP
For dose–response meta-analysis, considering the log relative risk as an independent variable and the exposure level as a dependent variable, we first adopted restricted cubic splines with five knots at 1st, 25th, 50th, 75th, and 99th percentiles of
AC C
exposure distribution in the dose–response meta-regression model to evaluate potential nonlinear trends between sleep duration and total stroke, nonfatal stroke, and fatal stroke [27]. Generalized least-square method was used to estimate the parameters [28,29]. In each of the included studies, we assigned the reported median or mean sleep duration of each category as the category sleep duration. When a study reported only the range of sleep duration for a category, we used the average value of the lower and upper bounds of that category. When the shortest or the longest category was open-ended, we assumed that the open-ended interval length had the same length as the adjacent interval. Additionally, the number of stroke cases and participants (or 5
ACCEPTED MANUSCRIPT person-years) of each category of sleep duration was required, and these numbers were inferred based on total number of cases and the reported risk estimates if the study didn’t report them [30]. The p-value for nonlinearity was calculated by testing the null hypothesis that coefficients of the second, third and fourth spline were zero
RI PT
[29]. Then, we set 7 h as the reference category to predict RRs and 95% CI of stroke under different sleep durations [10]. Finally, if a U-, J- or S-shape or their reverse association was observed, we treated the slope as two piecewise with the cut point of 7 h to show the linear trends, respectively [31,32].
SC
If the study only reported information by sex, we considered them as two cohorts. All the regression coefficient and results were pooled in a random-effects model to
M AN U
account for heterogeneity among studies [33]. At the same time, we conducted subgroup analyses of nonlinearity and linearity based on stroke subtype, sex, location, the assessment of sleep duration, excluding studies that enrolled special populations because they were considered to be potential residual confounders. Analyses were performed with STATA version 12.0 (StataCorp, College Station, TX) and all tests
3. Results
TE D
were two-sided with a significance level of 0.05.
EP
3.1 Literature search and study characteristics A detailed description of how studies were selected is presented in Fig. 1. Briefly, from the preliminary literature search, a total of 825 titles and abstracts were
AC C
identified. After excluding 786 irrelevant records, we read 39 full-text articles for further assessment. Eventually, we identified 12 articles that met all criteria for this meta-analysis, which enrolled a total of 528,653 participants with 12,193 stroke events [10,11,13–17,34–38]. The characteristics and quality score of the individual studies are shown in Table 1. Among 12 included studies, four only reported information on separate sexes [14,17,36,37]. Thus, we eventually had 16 cohorts. The mean or median follow-up duration ranged from 7.8 to 14.7 years. The sample size ranged from 2282 to 98,634. The study population was the general population except in one study that enrolled 6
ACCEPTED MANUSCRIPT middle-aged male workers in a light metal factory of Japan [16]. The age at the entry of population ranged from 25 to 98. Two studies reported total stroke, nonfatal and fatal stroke [11,14], four fatal stroke [34–37], three nonfatal stroke [10,15,17] and three combined [13,16,38]. Four studies also supplied information about different
RI PT
stroke types [11,13,34,37]. The exposure of sleep duration in all studies was self-reported either by structured interview or self-administered questionnaires.
3.2 Shortest and longest sleep duration and the risk of total stroke
SC
The original reference category in each included study was slightly different, as were the shortest and longest sleep duration. The reference duration of five studies are
M AN U
at 7 h, two at 6–8 h, five at 7–8 h. For shortest sleep duration, six studies were set to ≤6 h, four to ≤5 h and 1 to ≤4 h. The pooled RR for the shortest sleep duration for total stroke incidence was 1.10 (95% CI: 0.97–1.24; Fig. 2) with moderate heterogeneity (I2 = 49.2%, p = 0.014). No significant publication bias was found (Begg’s test p = 0.44, Egger’s test p = 0.48). For longest sleep duration, five were set to ≥8 h, five
TE D
to ≥9 h and one to ≥10 h, and the corresponding pooled RR was 1.37 (95% CI: 1.23–1.54; Fig. 2) with significant heterogeneity (I2 = 55.7%, p = 0.004) and no
EP
publication bias exsited (Begg’s test p = 1.00, Egger’s test p = 0.62).
3.3 The nonlinear association between sleep duration and stroke Figures 3–5 show the nonlinear curves between self-reported sleep duration and
AC C
total stroke, nonfatal stroke, and fatal stroke, respectively. A J-shaped association was found between sleep duration and total stroke incidence with lowest risk observed at sleeping for 7 h. When we considered 7 h as the reference category, the risk of total stroke was higher among long sleepers than that among short sleepers. The pooled RRs (95% CI) were as follows: 4 h: 1.17 (0.99–1.38); 5 h: 1.17 (1.00–1.37); 6 h: 1.10 (1.00–1.21); 8 h: 1.17 (1.07–1.28); 9 h: 1.45 (1.23–1.70); 10 h: 1.64 (1.4–1.92) (pnonlinearity<0.001, Table 2). For nonfatal stroke, a U-shaped association including five cohorts was observed with borderline significance of nonlinearity, and the predict RRs were as follows: 5 h: 1.43 (1.13–1.80); 6 h: 1.30 (1.09–1.54); 8 h: 1.08 (0.98–1.20); 7
ACCEPTED MANUSCRIPT 9 h: 1.51 (1.15–1.99) (pnonlinearity = 0.05, Table 2). The trend of fatal stroke was similar to that of total stroke and we only found a significantly increased risk among long sleepers (4 h: 1.08 (0.91–1.29); 5 h: 1.09 (0.92–1.28); 6 h: 1.05 (0.95–1.16); 8 h: 1.15 (1.05–1.26); 9 h: 1.43 (1.19–1.71); 10 h: 1.65 (1.36–2.00); pnonlinearity<0.001, Table 2).
RI PT
The nonlinear analyses of subgroups are shown in Table 2. J-shaped associations were found in most subgroups, and the risk for short sleepers did not reach significance
except
in
the
subgroups
of
non-Asian
countries
with
the
structured-interview. Additionally, we observed a slightly protective effect of IS
SC
among short sleepers compared to people sleeping for 7 h, though it was nonsignificant (4 h: 0.92 (0.70–1.21); 5 h: 0.91 (0.70–1.19); 6 h: 0.94 (0.82–1.07)).
M AN U
After excluding the study by Hamazaki et al. [16], which enrolled a special population (workers), the results compared to that of overall analysis attenuated slightly.
3.4 The linear association between sleep duration and strokes
All nonlinear analyses found a J-shaped or U-shaped association between sleep
TE D
duration and stroke except the IS one, so we treated the slope as two piecewise with the cut point of 7 h to show the linear trends, respectively. Among people who slept more than 7 h, every 1 h of sleep duration increment led to an increase of 13% (RR
EP
(95% CI) = 1.13 (1.07–1.20); p<0.001) in risk of total stroke and 12% (RR (95% CI) = 1.12 (1.04–1.21)) of fatal stroke. Sleeping less than 7 h did not show a statistically significant risk of each stroke outcome. The risk by sleeping more than
AC C
7 h were also significant in most of subgroups (Table 3).
4. Discussion
In this dose–response meta-analysis of 16 prospective studies, we found a J-shaped
association between sleep duration and total stroke incidence, and long sleepers had a significantly higher risk of total, nonfatal, and fatal stroke incidence compared to people sleeping for 7 h. Short sleep durations were only significantly related with nonfatal stroke, and with total stroke in subgroups of structured interview and non-Asian countries. Additionally, we found a slightly decreased risk of IS among 8
ACCEPTED MANUSCRIPT short sleepers. For piecewise linear analyses, we also found a positive linear dose–response association among long sleepers. Every 1-h increase in sleep duration led to an increased risk of 13% and 12% of total stroke and fatal stroke, respectively.
RI PT
Long sleep duration might be a predictor for future stroke.
4.1 Comparison with previous studies
Several epidemiological studies have explored this association. A meta-analysis by Ge showed significantly increased risk of stroke incidence and mortality at either end of
SC
the distribution of sleep duration in both cohort and cross-sectional studies; and RR for long sleep is stronger than that for short sleep. Their subgroup analysis also
M AN U
showed that only long sleep duration was a statistical stroke risk in both sexes and in Asians [19]. Those results were in accordance with ours. Leng identified prolonged sleep as a potentially useful marker of increased stroke risk in an apparently healthy aging population [11]. In addition, some cross-sectional studies found the same results. Akinseys using the data of the National Health Interview Survey (2004–2013) found
TE D
that the prevalence of stroke in diabetes was 16.1% in long sleepers, higher than that at 8.3% in normative sleepers (p<0.01), but only males showed a continued association between long sleep and stroke when compared to females, which was
EP
different to our findings [1]. Some studies such as Westerlund et al. [15] found no harmful influence of sleep duration on stroke incidence. Our study on fatal stroke did not show a significant result for short sleep duration.
AC C
Studies on this association were relatively scarce. The Singapore Chinese Health Study, a population-based cohort of 63,257 Chinese adults, found that both short and long sleep durations were associated with increased risk of stroke mortality [35]. However, another three other studies, conducted in Japan [37], Los Angeles/Hawaii [36], and China [39], each comprising more than 95,000 participants, showed the same outcome as ours—only long sleep duration was significantly associated with a higher risk of stroke mortality in both sexes. A possible explanation is that the bulk of the population in the long sleep duration group consisted of elderly people, who experience more unhealthy status and have more comorbidities with advancing age; 9
ACCEPTED MANUSCRIPT thus they generally have a higher chance of suffering from a more severe stroke. There were five studies reporting the association between the subtypes of stroke and sleep duration in our meta-analysis, and four supplied eligible data that we included in our subgroup analysis. An increased risk of IS was found among long
RI PT
sleepers in three studies, which was borderline significant in ours. Additionally, we observed a weak decreased risk of IS incidence among people sleeping less than 7 h, though it was not significant. It was somewhat consistent with the study by Kawachi et al.; and they also found a significant risk reduction of hemorrhagic stroke mortality
SC
for less than 7 h of sleep which was contrary with ours. We suggested that the effect of sleep duration on IS is different from that on HS, but the relevant data were few. More
M AN U
researches need to be conducted to examine this association.
4.2 Potential mechanisms under this association
The underlying mechanisms are not fully understood. Sleep deprivation has been widely demonstrated to be linked to elevated ghrelin levels, reduced leptin levels [40],
TE D
disrupted lipid metabolism [41], high blood glucose [6], et cetera. Covassin revealed that compared to normal sleep, sleep restriction increased nocturnal blood pressure and attenuated percentage of blood pressure dipping [42]. The risk for long sleep
EP
duration was unclear; some researchers believe that excessive sleep might increase some inflammatory biomarkers, such as C-reactive protein and interleukin-6 [43]. The China Health and Nutrition Survey, selecting individuals from 228 communities
AC C
found that shorter (<6 h) or longer (>10 h) sleep durations were associated with higher risks of abnormal lipid profiles in female [41]. Long sleep duration was also reported to be associated with carotid artery atherosclerosis [44], atrial fibrillation [45,46], white matter hyperintensity volume [47], and left ventricular mass [48], which might increase an individual’s risk of stroke.
4.3 Limitations and strengths This study has some limitations. The assessment of sleep duration in all included studies was self-reported, and thus could lead to an overestimation of the true sleep 10
ACCEPTED MANUSCRIPT duration because the subjects might consider awake time in bed as a part of sleep time, though a study has indicated that sleep duration assessed by self-reporting is highly related to that by actigraph or polysomnography [49]. In addition, habitual sleep durations were collected from a single question, and subjects might change their sleep
RI PT
patterns during their follow-up periods. However, Leng’s study indicated that the results for combined sleep duration of the baseline and follow-up were in line with their results on a single measure of sleep at baseline [11]. We used five knots in our restricted cubic functions taking into consideration our relatively large sample size
SC
and at least 4 categories of sleep duration available in most studies. Although Stone suggested that five knots may provide enough flexibility for a reasonable number of
M AN U
degrees of freedom, we cannot deny that the variance of the spline estimator and the risk of overfitting increase with knots, especially in the analysis of subtype stroke with relatively small sample size [27]. Finally, several studies have reported that sleep quality was associated with the risk of stroke [3], but only three studies in our meta-analysis included presence of snoring or sleep disorders as confounders
TE D
[10,13,15]. Two of the three studies showed a continued association between long sleep duration and stroke after snore status was adjusted, though the results by von Ruestern et al. became nonsignificant after taking body mass index (BMI), waist-to-hip ratio and history of high blood lipid levels into the model based on the
EP
snore-adjusted model [10,13]3. The third study did not find an association between sleep duration and stroke either before or after snore adjustment [15], but the
AC C
proportion of people with snoring habits or sleep disorders was higher in the short sleep duration group in all three studies. Perhaps worse sleep quality is related to shorter sleep duration rather than long sleep duration. However, the sleep disorders in long sleepers should not be ignored, and studies should take these into account when they assess the association between sleep duration and stroke incidence. Our study has two major strengths. Firstly, we evaluated the potential nonlinear and linear association between sleep duration and stroke incidence using a restricted cubic spline and a piecewise linear function, respectively. Previous meta-analyses on this association were mostly based on a two-category model, losing much information on 11
ACCEPTED MANUSCRIPT different categories of exposure and having lower statistical power. Secondly, the individual studies included were prospective and all subjects had no stroke history at baseline, so the causal association was somewhat rigorous.
RI PT
5. Conclusion In conclusion, this dose–response meta-analysis showed a J-shaped association between sleep duration and total stroke incidence. Long sleep duration was served as a higher risk than short sleep duration. Additionally, we found a positive linear trend
SC
among long sleepers that, compared to 7 h sleep duration, every 1-h additional increment in sleep duration led to an increase of 13% in the risk of total stroke. Long
need further investigation.
Conflicts of interest
M AN U
sleep duration might be a useful predictor of stroke risk. The underlying mechanisms
Acknowledgements
TE D
The authors have no conflicts of interest to declare.
References 1.
EP
We thank Mr Guangxiao Li for his guidance.
Akinseye OA, Ojike NI, Akinseye LI, Dhandapany PS, Pandi-Perumal SR.
AC C
Association of sleep duration with stroke in diabetic patients: Analysis of the national health interview survey. Journal of stroke and cerebrovascular diseases 2016;25:650–5
2.
Feigin VL, Krishnamurthi RV, Parmar P, Norrving B, Mensah GA, Bennett DA, et al. Update on the global burden of ischemic and hemorrhagic stroke in 1990–2013: The gbd 2013 study. Neuroepidemiology. 2015;45:161–76
3.
Chien KL, Chen PC, Hsu HC, Su TC, Sung FC, Chen MF, et al. Habitual sleep duration and insomnia and the risk of cardiovascular events and all-cause death: Report from a community-based cohort. Sleep. 2010;33:177–84 12
ACCEPTED MANUSCRIPT 4.
Theorell-Haglow J, Berglund L, Berne C, Lindberg E. Both habitual short sleepers and long sleepers are at greater risk of obesity: A population-based 10-year follow-up in women. Sleep medicine. 2014;15:1204–11
5.
Lou P, Zhang P, Zhang L, Chen P, Chang G, Zhang N, et al. Effects of sleep
RI PT
duration and sleep quality on prevalence of type 2 diabetes mellitus: A 5-year follow-up study in china. Diabetes research and clinical practice. 2015;109:178–84 6.
Kim JY, Yadav D, Ahn SV, Koh SB, Park JT, Yoon J, et al. A prospective study
SC
of total sleep duration and incident metabolic syndrome: The arirang study. Sleep medicine. 2015;16:1511–15
Wang T, Lu J, Wang W, Mu Y, Zhao J, Liu C, et al. Sleep duration and snoring
M AN U
7.
associate with hypertension and glycaemic control in patients with diabetes. Diabetic medicine 2015;32:1001–7 8.
Fang J, Wheaton AG, Ayala C. Sleep duration and history of stroke among adults from the USA. Journal of sleep research. 2014;23:531–7 Liu Y, Wheaton AG, Chapman DP, Croft JB. Sleep duration and chronic
TE D
9.
diseases among u.S. Adults age 45 years and older: Evidence from the 2010 behavioral risk factor surveillance system. Sleep. 2013;36:1421–7 von Ruesten A, Weikert C, Fietze I, Boeing H. Association of sleep duration
EP
10.
with chronic diseases in the european prospective investigation into cancer and nutrition (epic)-potsdam study. PloS one. 2012;7:e30972 Leng Y, Cappuccio FP, Wainwright NW, Surtees PG, Luben R, Brayne C, et al.
AC C
11.
Sleep duration and risk of fatal and nonfatal stroke: A prospective study and meta-analysis. Neurology. 2015;84:1072–9
12.
Chen JC, Brunner RL, Ren H, Wassertheil-Smoller S, Larson JC, Levine DW, et al. Sleep duration and risk of ischemic stroke in postmenopausal women. Stroke; a journal of cerebral circulation. 2008;39:3185–92
13.
Song Q, Liu X, Zhou W, Wang L, Zheng X, Wang X, et al. Long sleep duration and risk of ischemic stroke and hemorrhagic stroke: The kailuan prospective study. Scientific reports. 2016;6:33664 13
ACCEPTED MANUSCRIPT 14.
Helbig AK, Stockl D, Heier M, Ladwig KH, Meisinger C. Symptoms of insomnia and sleep duration and their association with incident strokes: Findings from the population-based monica/kora augsburg cohort study. PloS one. 2015;10:e0134480 Westerlund A, Bellocco R, Sundstrom J, Adami HO, Akerstedt T, Trolle
RI PT
15.
Lagerros Y. Sleep characteristics and cardiovascular events in a large swedish cohort. European journal of epidemiology. 2013;28:463–73 16.
Hamazaki Y, Morikawa Y, Nakamura K, Sakurai M, Miura K, Ishizaki M, et al.
SC
The effects of sleep duration on the incidence of cardiovascular events among middle-aged male workers in japan. Scandinavian journal of work,
17.
M AN U
environment & health. 2011;37:411–17
Amagai Y, Ishikawa S, Gotoh T, Kayaba K, Nakamura Y, Kajii E. Sleep duration and incidence of cardiovascular events in a japanese population: The jichi medical school cohort study. Journal of Epidemiology. 2010;20:106–10
18.
Amagai Y, Ishikawa S, Gotoh T, Doi Y, Kayaba K, Nakamura Y, et al. Sleep
TE D
duration and mortality in japan: The jichi medical school cohort study. Journal of epidemiology / Japan Epidemiological Association. 2004;14:124–8 19.
Ge B, Guo X. Short and long sleep durations are both associated with
EP
increased risk of stroke: A meta-analysis of observational studies. International journal of stroke : official journal of the International Stroke Society. 2015;10:177–84 Cappuccio FP, Cooper D, D’Elia L, Strazzullo P, Miller MA. Sleep duration
AC C
20.
predicts cardiovascular outcomes: A systematic review and meta-analysis of prospective studies. European heart journal. 2011;32:1484–92
21.
Li W, Wang D, Cao S, Yin X, Gong Y, Gan Y, et al. Sleep duration and risk of stroke events and stroke mortality: A systematic review and meta-analysis of prospective
cohort
studies.
International
journal
of
cardiology.
2016;223:870–6 22.
Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of observational studies in epidemiology - a proposal for 14
ACCEPTED MANUSCRIPT reporting. Jama-J Am Med Assoc. 2000;283:2008–12 23.
Wells GA, Shea B, O'Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies
in
meta-analyses.
2011.
Available
at:
RI PT
http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. accessed in Sep 2014. 24.
DerSimonian R, Laird N. Meta-analysis in clinical trials. Controlled clinical trials. 1986;7:177–88
Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. Bmj. 2003;327:557–60
Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis
M AN U
26.
SC
25.
detected by a simple, graphical test. Bmj. 1997;315:629–34 27.
Durrleman S, Simon R. Flexible regression models with cubic splines. Statistics in medicine. 1989;8:551–61
28.
Orsini N, Li R, Wolk A, Khudyakov P, Spiegelman D. Meta-analysis for linear
TE D
and nonlinear dose–response relations: Examples, an evaluation of approximations,
and
software.
American
journal
of
epidemiology.
2012;175:66–73
Greenland S, Longnecker MP. Methods for trend estimation from summarized
EP
29.
dose–response data, with applications to meta-analysis. American journal of epidemiology. 1992;135:1301–9 Bekkering GE, Harris RJ, Thomas S, Mayer AM, Beynon R, Ness AR, et al.
AC C
30.
How much of the data published in observational studies of the association between diet and prostate or bladder cancer is usable for meta-analysis? American journal of epidemiology. 2008;167:1017–26
31.
Gould WW. Linear splines and piecewise linear functions. Stata Tech Bull 1993;15:13e7.
32.
Yun Y, Zhang C, Weng H, Kuang X, Gong T, Lu Y, et al. How to Conduct Dose–response meta-analysis by using linear relation and piecewise linear regression
model.
Chinese
Journal 15
of
Evidence-based
Medicine.
ACCEPTED MANUSCRIPT 2016;16;111–4 33.
Berlin JA, Longnecker MP, Greenland S. Meta-analysis of epidemiologic dose–response data. Epidemiology. 1993;4:218–28
34.
Kawachi T, Wada K, Nakamura K, Tsuji M, Tamura T, Konishi K, et al. Sleep
RI PT
duration and the risk of mortality from stroke in japan: The takayama cohort study. Journal of epidemiology / Japan Epidemiological Association. 2016;26:123–30 35.
Pan A, De Silva DA, Yuan JM, Koh WP. Sleep duration and risk of stroke
SC
mortality among chinese adults: Singapore chinese health study. Stroke; a journal of cerebral circulation. 2014;45:1620–5
Kim Y, Wilkens LR, Schembre SM, Henderson BE, Kolonel LN, Goodman
M AN U
36.
MT. Insufficient and excessive amounts of sleep increase the risk of premature death from cardiovascular and other diseases: The multiethnic cohort study. Preventive medicine. 2013;57:377–85 37.
Ikehara. Association of sleep duration with mortality from cardiovascular
38.
TE D
disease and other causes for japanese men and women: The jacc study. 2009 Qureshi AI, Giles WH, Croft JB, Bliwise DL. Habitual sleep patterns and risk for stroke and coronary heart disease: A 10-year follow-up from nhanes i.
39.
EP
Neurology. 1997;48:904–911
Cai H, Shu XO, Xiang YB, Yang G, Li H, Ji BT, et al. Sleep duration and mortality: A prospective study of 113 138 middle-aged and elderly chinese
AC C
men and women. Sleep. 2015;38:529–36
40.
Taheri S, Lin L, Austin D, Young T, Mignot E. Short sleep duration is
associated with reduced leptin, elevated ghrelin, and increased body mass index. PLoS medicine. 2004;1:e62
41.
Zhan Y, Chen R, Yu J. Sleep duration and abnormal serum lipids: The china health and nutrition survey. Sleep medicine. 2014;15:833–39
42.
Covassin N, Bukartyk J, Sahakyan K, Svatikova A, Calvin A, St Louis EK, et al. Experimental sleep restriction increases nocturnal blood pressure and attenuates blood pressure dipping in healthy individuals. Journal of the 16
ACCEPTED MANUSCRIPT American College of Cardiology. 2015;65:A1352 43.
Prather AA, Vogelzangs N, Penninx BW. Sleep duration, insomnia, and markers of systemic inflammation: Results from the netherlands study of depression
and
anxiety
(nesda).
Journal
of
psychiatric
research.
44.
RI PT
2015;60:95–102 Abe T, Aoki T, Yata S, Okada M. Sleep duration is significantly associated with carotid artery atherosclerosis incidence in a japanese population. Atherosclerosis. 2011;217:509–13
Khawaja O, Sarwar A, Albert CM, Gaziano JM, Djousse L. Sleep duration and
SC
45.
risk of atrial fibrillation (from the physicians' health study). The American
46.
M AN U
journal of cardiology. 2013;111:547–51
Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: The framingham study. Stroke; a journal of cerebral circulation. 1991;22:983–88
47.
Ramos AR, Dong C, Rundek T, Elkind MS, Boden-Albala B, Sacco RL, et al.
TE D
Sleep duration is associated with white matter hyperintensity volume in older adults: The northern manhattan study. Journal of sleep research. 2014;23:524–30
Ramos AR, Jin Z, Rundek T, Russo C, Homma S, Elkind MS, et al. Relation
EP
48.
between long sleep and left ventricular mass (from a multiethnic elderly cohort). The American journal of cardiology. 2013;112:599–603 Signal TL, Gale J, Gander PH. Sleep measurement in flight crew: Comparing
AC C
49.
actigraphic and subjective estimates to polysomnography. Aviation, space, and environmental medicine. 2005;76:1058–63
17
ACCEPTED MANUSCRIPT Fig. 1. Flow chart showing the study selection process. Fig. 2. Forest plots of the association between shortest, longest sleep duration sleep duration and total stroke. (a) Shortest sleep duration; (b) longest sleep duration. I2, heterogeneity index; RR, risk ratios; I2, heterogeneity index; m and w represent
RI PT
different cohorts with separate data in a single study. Fig. 3. The nonlinear association between sleep duration and total stroke incidence. Fig. 4. The nonlinear association between sleep duration and nonfatal stroke incidence.
AC C
EP
TE D
M AN U
SC
Fig. 5. The nonlinear association between sleep duration and fatal stroke incidence.
18
ACCEPTED MANUSCRIPT
Song [13]
year 2016
Country
China
Sex
Both
Sample
Follow-up
size
years
95,023
7.9
Age at
Assessment of
baseline (range/mean)
sleep duration
Covariates in fully adjusted model
SC
Author
Publication
RI PT
Table 1. The characteristics of studies included in this meta-analysis.
18–98/51.2
Structured interview
Age, sex, BMI, SBP, DBP, total cholesterol
NOS (0–9 points) 9
Kawachi
2016
Japan
Both
27,896
14.2
2015
Great Britain
Both
9692
9.5
AC C
Leng [11]
35–97/NA
EP
[34]
TE D
M AN U
level, FBG, high sensitive C-reactive protein, smoking status, drinking consumption, physical activity, history of MI or atrial fibrillation, family history of stroke, snoring status, marital status, education level, family per member monthly income, hypotensive drug use, lipid-lowering drug use, hypoglycemic drug use, stratified by hospitals
Self-administered questionnaires
age, sex, BMI, smoking status, alcohol
7
consumption, physical activity score, histories of hypertension or diabetes, education years, marital status.
42–81/61.6
Structured interview
Age, sex, BMI, SBP, DBP, cholesterol level, smoking, alcohol consumption, physical activity, histories of diabetes or MI, family history of stroke, depression, education, marital status, social class, hypnotic drug use, hypnotic drug use, hypertension drug use
19
8
ACCEPTED MANUSCRIPT
Helbig [14]
2015
Germany
Both
12,131
14
25–74/NA
Structured interview
Age, BMI, dyslipidemia, current smoking
7
RI PT
activity, alcohol consumption, physical activity, histories of hypertension and diabetes, education, survey,
Pan [35]
2014
China
Both
63,257
14.7
45–74/NA
Structured interview
Age, sex, BMI, years of smoking, dose of
7
SC
smoking, alcohol consumption, histories of hypertension, diabetes, stroke, CHD, or cancer,
Westerlund
2013
Sweden
Both
41,192
M AN U
education, dialect, energy intake, dietary
13.2
NA/52.8
2013
USA
Both
135,685
12.9
45–75/NA
Ruesten [10]
2012
Europe
Both
AC C
von
23,620
7.8
polyunsaturated fatty acids Age, sex, BMI, lipid disturbance, smoking,
8
alcohol consumption, physical activity, histories of diabetes or hypertension, depressive symptoms, snoring, education, employment status, work schedule, self-rated health
Self-administered questionnaires
Age, sex, BMI, smoking history, alcohol
6
consumption, physical activity, hours spent daily watching television, histories of
EP
Kim [36]
questionnaires
TE D
[15]
Self-administered
intakes of vegetables, fruits, fiber,
hypertension or diabetes, education, marital status, ethnicity, energy intake
35–65/NA
Structured interview
Age, sex, smoking status, alcohol consumption, walking, cycling, sports, sleeping disorders, education, employment status
20
8
ACCEPTED MANUSCRIPT
Hamazaki
2011
Japan
Male
2282
14
35–54/43.7
Age, BMI, type of job, total cholesterol level,
questionnaires
mean blood pressure, HbA1c, smoking status,
6
RI PT
[16]
Self-administered
alcohol consumption, physical activity, working hours, mental workload, medication for hypertension, diabetes, or
Amagai
2010
USA
Both
11,367
10.7
18–90/
[17]
Structured interview
male, 55.1;
Japan
Both
98,634
M AN U
2009
14.3
40–79/NA
Self-administered
1997
USA
Both
[38]
7844
10
TE D
questionnaires
Qureshi
Age, BMI, SBP, total cholesterol level,
7
smoking, alcohol consumption.
female, 55.3 Ikehara [37]
SC
hypercholesterolemia.
31–74/NA
Structured interview
Age, BMI, smoking, alcohol consumption,
7
hours of exercise, hours of walking, histories of hypertension or diabetes, depressive symptoms, perceived mental stress, education, employment, frequency of fresh fish intake Age, sex, BMI, SBP, cholesterol level,
7
smoking, history of diabetes, education, race
AC C
Assessment Scale; SBP, systolic blood pressure.
EP
BMI, body mass index; CHD, coronary heart disease; DBP, diastolic blood pressure; FBG, fast blood glucose; MI, myocardial infarction; NA, not available; NOS, Newcastle-Ottawa Quality
21
ACCEPTED MANUSCRIPT
RI PT
Table 2. The nonlinear dose–response association between sleep duration and stroke incidence.
Stroke risk (RR (95% CI)) 5 h
6 h
7 h
1.17 (0.99–1.38)
1.17 (1.00–1.37)
1.10 (1.00–1.21)
1
—
1.43 (1.13–1.80)
1.30 (1.09–1.54)
1.08 (0.91–1.29)
1.09 (0.92–1.28)
1.05 (0.95–1.16)
ischemic stroke
0.92 (0.70–1.21)
0.91 (0.70–1.19)
0.94 (0.82–1.07)
hemorrhagic stroke
1.02 (0.71–1.47)
1.01 (0.71–1.45)
Male
1.03 (0.80–1.33)
Female
Total stroke
8 h
9 h
10 h
1.17 (1.07–1.28)
1.45 (1.23–1.70)
1.64 (1.40–1.92)
1.08 (0.98–1.20)
1.51 (1.15–1.99)
—
1.15 (1.05–1.26)
1.43 (1.19–1.71)
1.65 (1.36–2.00)
1
1.09 (0.94–1.28)
1.31 (0.98–1.76)
1.79 (1.39–2.32)
0.98 (0.82–1.17)
1
1.18 (0.96–1.46)
1.33 (0.91–1.94)
1.18 (0.82–1.68)
1.02 (0.81–1.30)
1.01 (0.87–1.17)
1
1.10 (0.97–1.24)
1.30 (1.02–1.65)
1.58 (1.24–2.01)
1.07 (0.86–1.32)
1.07 (0.86–1.33)
1.04 (0.90–1.19)
1
1.19 (1.07–1.34)
1.46 (1.15–1.85)
1.46 (1.16–1.83)
1.11 (0.88–1.42)
1.11 (0.88–1.39)
1.06 (0.92–1.21)
1
1.19 (1.03–1.36)
1.49 (1.19–1.86)
1.76 (1.43–2.15)
—
1.28 (1.06–1.54)
1.19 (1.04–1.36)
1
1.16 (1.06–1.28)
1.49 (1.20–1.85)
—
—
1.25 (1.02–1.54)
1.18 (1.03–1.35)
1
1.10 (1.02–1.19)
1.50 (1.20–1.87)
—
1.14 (0.86–1.50)
1.13 (0.86–1.50)
1.07 (0.90–1.27)
1
1.20 (1.00–1.46)
1.49 (1.16–1.91)
1.78 (1.44–2.20)
1.08 (0.99–1.18)
1
1.14 (1.06–1.22)
1.41 (1.21–1.63)
1.60 (1.37–1.87)
nonfatal stroke fatal stroke a
Stroke subtype
Asian countries Non-Asian countries
Structured interview Self-administered questionnaires a
Others
Excluding the study of Hamazaki [16] CI: confidence interval; RR: risk ratio. a
AC C
Sleep duration assessment
a
TE D
Locationa
EP
Sex
a
1.14 (0.98–1.31)
1.14 (0.99–1.30)
1 1
M AN U
Stroke outcome
SC
4 h
Risk of total stroke.
22
ACCEPTED MANUSCRIPT Table 3 The piecewise linear dose–response association between sleep duration and stroke incidence Stroke risk (RR (95% CI)) Sleep less than 7 ha
Sleep more than 7 hb
1.07(1.00,1.14)
1.13 (1.07, 1.20)
Nonfatal stroke
1.08(0.95,1.23)
1.08 (0.98, 1.18)
Fatal stroke
1.05(0.97,1.15)
1.12 (1.04, 1.21)
—
—
1.04(0.90–1.19)
1.10 (0.96–1.26)
1.11(0.99, 1.24)
1.15 (1.04, 1.27)
Total stroke
Stroke subtype
c
Ischemic stroke Hemorrhagic stroke
Male Female Location
SC
Sex
c
0.99 (0.89, 1.10)
1.13 (1.03, 1.24)
c
1.10 (1.00, 1.21)
Non-Asian countries Exposure assessment
1.03 (0.94 ,1.13)
1.08 (1.00, 1.16)
1.07 (0.96, 1.19)
1.09 (1.00, 1.18)
1.07 (0.99, 1.16)
1.21 (1.12, 1.31)
1.06 (1.00, 1.13)
1.13 (1.07, 1.19)
c
Structured interview Self-administered Questionnaires Othersc
TE D
Excluded the study of Hamazaki [16]
1.19 (1.09, 1.30)
M AN U
Asian countries
RR, risk ratio; CI, confidence interval. a
RI PT
Stroke outcomes
RRs were corresponded to every 1-h decrease of sleep duration. RRs were corresponded to every 1-h increase of sleep duration.
c
Risk of total stroke.
AC C
EP
b
23
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT Highlights ● Restricted cubic spline functions were conducted to examine the nonlinear dose–response association between sleep duration and stroke risk.
association between sleep duration and stroke risk.
RI PT
● Piecewise linear functions were conducted to examine the linear dose–response
AC C
EP
TE D
M AN U
higher risk of fatal and nonfatal stroke incidence.
SC
● Compared to people who sleep for 7 h, people who sleep for more than 7 h have a