Introduction
Divergence from the recommended 7–9 h of sleep duration has been associated with cardiovascular disease and mortality, with the suggestion of a J-shaped relationship. Reference Yazdanpanah, Homayounfar, Khademi, Zarei, Shahidi and Farjam1–Reference Lao, Liu and Deng4 Recently, an association between abnormal sleep duration and stroke incidence has been reported, especially excess sleep >9 h. Reference Chen, Brunner and Ren5–Reference Titova, Michaëlsson and Larsson9
However, the results are inconsistent, with some studies and meta-analyses reporting no association between sleep duration and stroke risk. Reference Amagai, Ishikawa, Gotoh, Kayaba, Nakamura and Kajii10,Reference Westerlund, Bellocco, Sundström, Adami, Åkerstedt and Trolle Lagerros11 In addition, older individuals have lower recommended and self-reported sleep duration, Reference Chaput, Dutil and Sampasa-Kanyinga12 and whether age modifies the association of abnormal sleep duration with stroke is not well understood despite the potential importance of age for the targeting of preventive efforts. Lastly, sleep duration is also associated with all-cause mortality, Reference Cappuccio, D’Elia, Strazzullo and Miller13,Reference Gallicchio and Kalesan14 which may bias estimates of stroke risk.
We sought to determine the association between sleep duration and stroke risk in a large Canadian population-based cohort, evaluate differences with age, and account for the competing risk of death.
Methods
Study Sample – Canadian Community Health Survey
The Canadian Community Health Survey (CCHS) is an annual cross-sectional survey, representing 97% of the Canadian household population aged 12 years and older. 15 The CCHS collects information about health status, health determinants, and health care utilization of the household population. The 3% of the general population that is excluded from the survey target population includes those living on Indigenous reserves, those living in foster care, full-time members of the Canadian Armed Forced, the institutionalized population, and the remote Région du Nunavik and Région des Terres-Cries-de-la-Baie-James.
Administrative Linkages
Linkages were performed by Statistics Canada and included the Canadian Institutes of Health Discharge Abstract Database (CIHI-DAD) for hospitalizations, the National Ambulatory Care Reporting System (NACRS) for emergency department visits, and the Canadian Vital Statistics Database (CVSD) for deaths. 16,Reference Sanmartin, Decady and Trudeau17 The linkage subsample contained 85% of total CCHS respondents who agreed to have their responses linked to administrative records. We used CIHI-DAD, NACRS, and CVSD to determine stroke events and deaths after the survey response until December 31, 2017. All participants had a minimum of 1 year of full follow-up for events and were censored if alive on December 31, 2017.
Cohort
We used CCHS years 2000, 2007, 2008, and 2011–2016 to capture data on sleep duration and other baseline covariates. Sleep duration was an optional module in all years except 2000, meaning that it could be selected by some but not all provinces and territories. We excluded those under the age of 40 and those with a history of heart disease (self-report), cancer (self-report), or prior stroke (either administrative data diagnosis or self-report).
Exposure
Our main exposure was self-reported sleep duration. Participants were asked the number of hours spent sleeping per night. Time spent in daytime napping was not captured. We categorized sleep duration as <4 h, 4–6 h, 7–9 h, and ≥10 h/day, with excess sleep duration defined as ≥10 h/day, as the Canadian guidelines recommend 7–9 h of sleep per night for individuals <65 years of age. Reference Ross, Chaput and Giangregorio18
Covariates
Covariates were obtained from the CCHS, including age, sex, rural residence, and self-report of ethnicity, education level, total household income, marital status, body mass index, current smoking status, hypertension, diabetes, heart disease, cancer, arthritis, chronic obstructive pulmonary disease (COPD), migraine, and asthma. Categorization of the variables can be seen in Table 1. Due to known biases in self-report of body mass index, we employed a correction developed by the CCHS. Reference Shields, Gorber, Janssen and Tremblay19
Table 1: Weighted percentage of baseline characteristics
Outcome
Our primary outcome was acute stroke (ischemic stroke or intracerebral hemorrhage [ICH]) at any time in follow-up, using International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Canada (ICD-10-CA) codes (ischemic stroke: I63.x, I64.x, H34.1; ICH: I61.x) and ICD-9 codes (ischemic stroke: 434.01, 434.11, 434.91, 436; ICH: 431). Our secondary outcome was death during follow-up.
Analysis
We first used Cox proportional regression models to obtain the hazard of stroke by category of sleep duration, censoring people at death or end of follow-up. We assessed for the presence of modification of the sleep duration-stroke association by age (<70 years and ≥70 years) and sex. As there was evidence of modification by age but not sex, we created age-specific stratified models.
In our simple model, we adjusted for sex only. In our vascular model, we adjusted for sex, body mass index, hypertension, diabetes, and smoking. In our full model, we adjusted for the above in addition to income quartile, rural residence, alcohol consumption, education, ethnicity, marital status, and other comorbidities (migraine, arthritis, COPD, and asthma). We evaluated the proportional hazards assumption by assessing the significance (p < 0.05) of an interaction term of sleep duration category and follow-up time.
We conducted three sensitivity analyses. First, we additionally adjusted for symptoms of depression (2 or more weeks of feeling sad, blue, or depressed in the past 12 months) in the sub-set of individuals in which this variable was available. Second, we changed the upper category of sleep duration to ≥9 h. Third, we evaluated the risk of ischemic stroke only; there were insufficient events for an analysis of ICH only.
Lastly, we conducted competing risk regression to confirm that the risk of stroke was elevated while accounting for the competing risk of death. We obtained subdistribution hazard ratios and generated cumulative incidence functions for the risk of stroke over follow-up time. We adjusted for the same factors as in previous models.
Analyses were done in the Prairie Regional Data Centre at the University of Calgary using Stata 16.0 (College Station, TX, USA). Threshold of significance for p-values was <0.05. Under Tri-Council guidelines, this analysis did not require approval by a research ethics board.
Results
Our total cohort comprised 82,795 people with self-reported sleep duration and without prior heart disease, cancer, or stroke. During a median follow-up time of 9.1 years (interquartile range [IQR] 3.2–16.6 years), there was a total of 1705 stroke events (88.3% ischemic). Among those with incident stroke, the median time from survey response until stroke was 7.5 years (IQR 3.8–12.3).
Baseline characteristics of the cohort weighted to the Canadian population are shown in Table 1. The proportion of individuals <70 years was 86% and the proportion of females was 53.2%. The majority of people had 7–9 h of sleep duration (55.8%), with a small proportion having ≥10 h of sleep duration (1.2%).
Weighted percentages of stroke events in follow-up are shown in Supplemental Table 1 and stratified by subgroup. The greatest increases in stroke rate occurred when comparing 7–9 h of sleep with ≥10 h of sleep in those <70 years of age (1.5% to 5.9%) and in females (2.0% to 6.9%; Supplementary Table 1).
In the simple Cox regression model, there was a significant interaction between sleep duration ≥10 h and age (p = 0.030), but not sex (p = 0.27). There was no evidence of violation of the proportional hazards assumption.
Among individuals <70, excess sleep was associated with higher long-term risk of stroke in the simple model (adjusted hazard ratio [aHR] 3.15, 95% CI 1.74–5.73), vascular model (aHR 2.77, 95% CI 1.35–5.70), and full model (aHR 2.26, 95% CI 1.02–5.00; Figure 1A–C). There was no association between excess sleep and risk of stroke in those ≥70 years (Figure 1). The results were consistent in a sensitivity analysis adjusting for symptoms of depression. The association did not persist when the highest category of sleep duration was changed to ≥9 h (Supplementary Table 2). When analyzing ischemic stroke only, there was a higher risk in the simple and vascular models, but the association was attenuated in the full model (Supplementary Table 3).
Figure 1: Association between categories of sleep duration and risk of stroke, stratified by age. Hazard ratios with 95% CIs are from Cox regression models (A, simple model; B, vascular model; C, full model), and subdistribution hazard ratios are from competing risk analysis (D, full model). There is a higher hazard of stroke in those <70 years old who report 10+ hours of sleep per night.
After accounting for the competing risk of death, there remained a significantly elevated risk and higher cumulative incidence of stroke in those with excess sleep and age <70 years (Figure 1D and Figure 2).
Figure 2: Cumulative incidence functions over years of follow-up for categories of sleep duration, stratified by age. Those ≥70 years have lower maximal follow-up time due to higher mortality. The greatest elevation in cumulative incidence of stroke occurs in individuals <70 years who report 10+ hours of sleep per night. Estimates obtained from fully adjusted competing risk regression models.
Discussion
We demonstrate that excess sleep duration ≥10 h is associated with increased long-term risk of stroke among individuals younger than 70 years old, and this risk remains elevated after accounting for the competing risk of death.
Our results are consistent with prior studies showing an increased risk of stroke with excess sleep time. Reference Chen, Brunner and Ren5–Reference Helbig, Stöckl, Heier, Ladwig and Meisinger8,Reference Song, Howard and Howard20,Reference Leng, Cappuccio and Wainwright21 Some studies have suggested a J-shaped relationship, with low amounts of sleep also increasing risk. Reference Li, Wang and Cao22 While the point estimate of stroke risk for the category of <4 h sleep was elevated among individuals <70, it did not reach significance in our study. The mechanism of stroke with excess sleep and stroke is unknown. Long duration of sleep is associated with inflammatory biomarkers, Reference Patel, Zhu and Storfer-Isser23,Reference Dowd, Goldman and Weinstein24 vascular risk factors, Reference Yang, Yang and He25–Reference Grandner, Mullington, Hashmi, Redeker, Watson and Morgenthaler27 carotid intimal thickness, Reference Abe, Aoki, Yata and Okada28 and white matter hyperintensities. Reference Ramos, Dong and Rundek29 In addition, excess sleeping may be a surrogate for higher sedentary time, depression, sleep apnea, or low socioeconomic status, all of which have been linked to stroke and cardiovascular risk. Reference Dong, Zhang, Tong and Qin30–Reference Biswas, Oh and Faulkner33 Our findings are consistent with guidelines recommending 7–9 h of sleep per night to promote optimal health. Reference Ross, Chaput and Giangregorio18,Reference Chaput, Dutil and Featherstone34
The reason for the observed age modification is unclear. Multiple studies have reported associations between sleep and hypertension that are attenuated in the elderly. Reference van den Berg, Tulen and Neven35–Reference Kim and Jo37 Similarly, among individuals <65 years, both short and long duration sleep was associated with excess mortality, but not in individuals 65 years and older. Reference Åkerstedt, Ghilotti, Grotta, Bellavia, Lagerros and Bellocco38 These finding are also in line with prior studies showing the associations between vascular risk factors such as obesity or diabetes and adverse outcomes are also greater at younger age. Reference Rawshani, Rawshani and Franzén39,Reference Peter, Mayer, Concin and Nagel40 Total sleep duration is significantly shorter in elderly, and the overall dynamics and homeostasis of sleep differ compared to middle-aged and younger subjects. Reference Campbell and Murphy41 Retirement also improves sleep quality, which may mitigate the impact of excess sleep duration. Reference Myllyntausta and Stenholm42 However, our results are in contrast to two studies from China showing the association of excess sleep and stroke was stronger at older age, Reference Zhou, Yu and Yang6,Reference Leng, Cappuccio and Wainwright21 and may be due to variability in the population studied, amount of follow-up, or other study differences. We had a smaller number of individuals who were ≥70 years old and shorter follow-up time due to higher mortality, which may have limited our ability to detect an association in this age group.
There were some limitations to this study. First, we relied on self-reported sleep duration in the CCHS. Self-reported sleep duration overestimates sleep duration compared to objective measurements; therefore, the true threshold for increased risk may be lower than reported here. Reference Jackson, Patel, Jackson, Lutsey and Redline43 Second, we relied on administrative data to measure stroke events, although the codes used have high validity in Canada. Reference Porter, Mondor, Kapral, Fang and Hall44 Third, we had no assessment of sleep quality, sleep apnea, or mid-day napping. Lastly, there is the potential for residual confounding from many vascular and socioeconomic factors which may be related to both sleep duration and stroke risk.
In conclusion, excess sleep duration ≥10 h was associated with elevated stroke risk in those <70 years of age. The results support current guidelines for 7–9 h of sleep per night and suggest the need for further research on determining mechanisms of sleep-related cardio- and cerebrovascular disease.
Supplementary Material
To view supplementary material for this article, please visit https://doi.org/10.1017/cjn.2021.242
Acknowledgements
The analysis was conducted at the Prairie Regional Research Data Centre (RDC), which is part of the Canadian Research Data Centre Network. The services and activities provided by the Canadian Research Data Centre Network are made possible by the financial or in-kind support of the Social Sciences and Humanities Research Council, the Canadian Institutes of Health Research, the Canadian Foundation for Innovation, Statistics Canada, and participating universities whose support is gratefully acknowledged. The views expressed in this article do not necessarily represent those of the Canadian Research Data Centre Network or that of its partners. We thank Stephanie Cantlay for her assistance in the Prairie RDC.
Funding
RAJ is supported by a Fellowship Grant from the Canadian Institutes of Health Research (Funding reference number MFE-164702).
Disclosures
EES reports royalties from UpToDate, consulting fees from Alnylam, Biogen, and Javelin. The other authors report no disclosures.
Statement of Authorship
RJ was involved with conception, data analysis and interpretation, and drafting the article; SP, JW, and ES were involved with interpretation of results and critical revision of the article.
