Healthy Sleep Pattern, Metabolic Diseases, and Risk of Stroke: The Kailuan Cohort Study

Introduction

Evidence has demonstrated that sleep complaints, including short or long sleep duration, insomnia, frequent snoring, and excessive daytime sleepiness, have an impact on the risk of stroke.^1,2 This could be due to their effects on sympathetic nervous system activity,³ glucose homeostasis,^4,5 and vascular structure and function.⁶ These sleep behaviors do not occur in isolation, and modifications in one sleep behavior usually lead to compensatory changes in other sleep behaviors.⁷ Therefore, combining various sleep behaviors into sleep patterns could reflect real-world sleep conditions. However, data on relationship between sleep pattern and risk of stroke are limited and mainly conducted in Western populations.^8–10 Moreover, the association between sleep complaints and stroke subtypes has not been well-studied,^11,12 and few studies were conducted in Asian populations.

In addition, the associations between individual sleep complaints and metabolic diseases, including obesity, diabetes, and hypertension, have been well-established.^13,14 Furthermore, these metabolic diseases are also risk factors for stroke^15–17 and thus may be involved in the association between sleep and stroke risk. To date, no data have explored to which extent the association between sleep pattern and stroke risk could be interpreted by these metabolic abnormalities.

To address these research gaps, we used data from the Kailuan study, a large community-based cohort in China, to examine the association of a healthy sleep score (combination of major sleep behaviors) with the risk of total stroke and subtypes of stroke. We also examined the role of metabolic diseases in the relationship between sleep pattern and risk of stroke.

Methods

Participants

In this analysis, participants were derived from the Kailuan study, an ongoing cohort from 2006–2007 with 101,510 Chinese adults aged 18–98 years recruited. The study design has been described in detail in previous studies.^18–20 Participants were recruited from Kailuan general hospital and 10 hospitals responsible for the healthcare of the community. All participants completed questionnaire assessments (ie socioeconomic status, lifestyles, and history of diseases), clinical, and laboratory examinations. The same assessments were conducted every two years afterwards. In 2012, information on sleep complaints was collected among 12,990 participants;^21,22 thus, the survey in 2012 was regarded as the baseline survey in this study.

We excluded those with cancer and cardiovascular disease (n=504) at baseline, without completed sleep data (n=265) or covariates (n=370), resulting in 11,851 participants for the present analysis. The study was approved by the Ethics Committee of the Kailuan General Hospital (approval number: 2006-05), and all participants provided written informed consent.

Assessment of Sleep Complaints

Sleep complaints, including insomnia, daytime sleepiness, sleep duration, and snoring, were collected via questionnaires. The detailed information has been described in previous studies.^23,24 Insomnia was assessed using the Chinese version of the Athens Insomnia Scale (AIS), comprising eight items evaluating insomnia symptoms over the past three months. Each item was scored from 0 to 3 (0=no event, 1= mild, 2= moderate, and 3= severe), with a total AIS score of ≥6 indicating insomnia. Daytime sleepiness was assessed via the Chinese version of the Epworth Sleepiness Scale (ESS) with eight items. Each item scored 0–3, representing the probability of falling asleep while engaged in specific daily life situations, with a total ESS score ≥10 indicating excessive daytime sleepiness. Sleep duration was collected via the question “usual total hours of actual sleep during the night over the past one month”. Snoring was collected by self-reported snoring frequency as never/rare, occasional, or frequent during the last month.

Definition of Healthy Sleep Score

We then constructed a healthy sleep score using the above four sleep behaviors. According to the previous studies⁸ and the recommendation of the American Academy of Sleep Medicine and Sleep Research Society,²⁵ we defined low-risk sleep factors as follows: no excessive daytime sleepiness, no insomnia, sleep 7–8 h per day, and no frequent snoring. Each low-risk sleep factor was assigned a score of 1 and 0 otherwise. Then, all component scores were summed to obtain a healthy sleep score ranging from 0 to 4, with higher scores indicating a healthier sleep pattern. The healthy sleep score was categorized into three groups according to the data distribution: healthy sleep score≤2, healthy sleep score of 3, and healthy sleep score of 4.

Assessment of Outcome-Incident Stroke Cases

The outcome was the first occurrence of stroke, including ischemic stroke and hemorrhagic stroke, according to the World Health Organization criteria.²⁶ The detailed adjudication of incident stroke in the Kailuan study was described previously.^22,27 In brief, participants were followed up by face-to-face interviews at biennial routine medical examinations until 31 December, 2020. The outcome was further confirmed by checking discharge summaries from the 11 hospitals and medical records from medical insurance. Outcome information was obtained directly by checking death certificates from provincial vital statistics offices, discharge summaries, and medical records for the participants without face-to-face follow-ups.

Assessment of Metabolic Conditions and Other Related Factors

The weight and height of each participant were measured by trained study staff. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared. Obesity was defined as BMI≥28kg/m², aligning with the Chinese criteria.²⁸ The blood pressure (BP) was taken using a calibrated mercury sphygmomanometer, with at least two readings after 5 minutes of rest; the average value was used for hypertension diagnosis. Following the Seventh Report of the Joint National Committee recommendation,²⁹ hypertension was defined as BP measurement ≥140/90 mmHg or the self-reported use of antihypertensive medication. Fasting (8–12h) blood samples were collected in the morning and then stored at −80°C at the central laboratory of Kailuan general hospital. An auto-analyzer (Hitachi 747; Hitachi, Tokyo) was used to measure fasting blood glucose (FBG) with the hexokinase/glucose-6-phosphate dehydrogenase method. The coefficient of variation using blind quality control specimens was <2.0%.³⁰ Diabetes was defined as an FBG value ≥7.0 mmol/L, use of hypoglycemic drugs, or self-reported physician-diagnosed diabetes.³⁰ We also combined obesity, diabetes, and hypertension, and defined a composite index with values ranging from 0 to 3, representing the range from no disease to the presence of the three metabolic diseases.

Information on age, sex, education, average income per month, occupation, alcohol consumption, smoking, and physical activity were also obtained via a questionnaire by well-trained health workers.²³ Serum creatinine was measured using the sarcosine oxidase assay method (creatinine kit; BioSino Bio-technology and Science Inc., Beijing, China), with ≤6% intra- and inter-assay variable coefficients. The estimated glomerular filtration rate (eGFR) was computed according to the chronic kidney disease (CKD) Epidemiology Collaboration equation.²³ Proteinuria was assessed using the dry-chemistry method (H12-MA test strips; Changchun Dirui Medical Technology Co., Ltd., Changchun, China). In this study, participants with eGFR <60mL/min per 1.73 m² or proteinuria ≥2+ (>300mg/dl) were considered to have CKD.²³

Statistical Analysis

Baseline characteristics were described across healthy sleep score, and differences among groups were tested by one-way ANOVA, Kruskal–Wallis test, or chi-square tests where appropriate. Person years were calculated from baseline until the date of stroke diagnosis, death, or 31 December 2020, whichever occurred first. Cox proportional hazards regression was used to estimate the hazard ratios (HRs) and 95% confidence intervals (CIs) of total stroke and stroke subtypes (ischemic and hemorrhagic stroke). The linear trend in risk across healthy sleep pattern groups was tested by using the median value for each group and treating them as continuous variables.

Based on existing literature, we adjusted for age and sex in Model 1; occupation (coal miners, other blue miners, white miners), education (illiterate and primary, middle school, high school and above), average income per month (≤1000, 1000–3000, >3000RMB), alcohol consumption (never and past, current and light drink, current and moderate and above), smoking (never, past, current), physical activity (no, occasional, frequent), use of sleep medication (no, yes) and CKD (no, yes) were further adjusted in Model 2. Obesity (no, yes), hypertension (no, yes), and diabetes (no, yes) were further adjusted in Model 3 to examine the influence of these metabolic diseases on the association between healthy sleep score and stroke. We also analyzed the association of individual low-risk sleep factors with total stroke and its subtypes.

To investigate the consistency of our findings across different demographics, subgroup analyses were performed based on age (<60 and ≥60 years), sex (men and women), education (illiterate and primary, middle school, high school and above), average income per month (≤1000, 1000–3000, >3000RMB), physical activity (no, occasional, frequent), alcohol consumption (never and past, current and light drink, current and moderate and above), smoking status (never, past, current), and CKD (no, yes). An interaction analysis between healthy sleep score and metabolic diseases was conducted using likelihood ratio tests.

Mediation effects of metabolic diseases on the association between healthy sleep score and stroke risk were evaluated using mediation package in R. Indirect, direct, and total effects for each mediator were computed via combining the mediator and outcome models with the adjustment of all the covariates in Model 2. The joint associations were also assessed. The participants were classified into nine groups according to number of healthy sleep factors (≤2, 3, 4) and the number of metabolic diseases (0, 1, and ≥2), with number of healthy sleep factors≤2 and metabolic diseases≥2 as the reference group.

We performed several sensitivity analyses to test the robustness of our findings. First, we excluded incident cases during the first two years of follow-up, considering the possibility of reverse causality. Second, we excluded participants with CKD. Third, we excluded coal miners (41.3%), considering the generalization of results.

All analyses were conducted by R (version 4.2.0), and a two-sided P value<0.05 was considered significant.

Results

Characteristics of the Study Population

Among the 11,851 participants (mean age: 53.8±10.4 years; 82.4% of men), 1468 (12.4%) participants had ≤2 healthy sleep factors, 6349 (53.6%) had 4 healthy sleep factors (Table 1). Participants with higher healthy sleep score were younger, more likely to be women, White-collar workers, and less likely to consume alcohol or smoke, and had a lower proportion of obesity, hypertension, and diabetes compared to those with lower healthy sleep score (Table 1).

Table 1 Baseline Characteristics Across Healthy Sleep Score

Association Between Healthy Sleep Pattern and Stroke Risk

During a mean follow-up period of 7.7±1.0 years, 504 incident cases of stroke were identified. We observed a higher healthy sleep score was associated with a lower risk of stroke in a dose-response manner. The adjusted HR of stroke was 0.86 (95% CI: 0.70, 0.94) for each healthy sleep score increment based on Model 2 (Table 2). Compared to healthy sleep score≤2, the adjusted HR of stroke was 0.71 (95% CI: 0.55, 0.91) for healthy sleep score of 4. The association (HR: 0.75, 95% CI: 0.56, 0.96) was slightly attenuated when further adjusting for obesity, diabetes, and hypertension. The associations were also observed in stroke subtype, with the estimated effect on hemorrhagic stroke (each healthy sleep score increment, HR: 0.62, 95% CI: 0.46, 0.83) were more pronounced than ischemic stroke (each healthy sleep score increment, HR: 0.88, 95% CI: 0.78, 0.99; Table 2).

Table 2 The Association Between Healthy Sleep Score and Risk of Stroke

Association Between Low-Risk Sleep Factors and Stroke Risk

The analyses of individual low-risk sleep factors showed that no excessive daytime sleepiness (HR: 0.60, 95% CI: 0.37, 0.97) and no frequent snoring (HR: 0.69, 95% CI: 0.56, 0.86) were associated with lower risk of total stroke (Table 3). The two sleep factors were also associated with stroke subtypes, with the estimated effect on hemorrhagic stroke (no excessive daytime sleepiness, HR: 0.28, 95% CI: 0.10, 0.79; no frequent snoring, HR: 0.46, 95% CI: 0.25, 0.83) were more pronounced compared to ischemic stroke (no excessive daytime sleepiness, HR: 0.64, 95% CI: 0.38, 1.05; no frequent snoring, HR: 0.73, 95% CI: 0.58, 0.92).

Table 3 The Association Between Low-Risk Sleep Factors and Risk of Stroke

Subgroup Analysis

The association between healthy sleep score and stroke risk remained consistent in participants aged ≥60, men, those with low income, no physical activity, no alcohol drinking and smoking, and without CKD, but the interaction test did not reach statistical significance (P for interaction>0.05 for all; Figure S1).

Mediation Analysis and Joint Analysis

In the mediation analysis, the association between healthy sleep score and stroke risk was mediated by obesity, diabetes, and hypertension with the proportion of mediation effect ranging from 4.4% to16.1%. Collectively, the three mediators explained 21.9% (95% CI: 17.2, 26.5) of the association between healthy sleep score and stroke risk (Table 4). The joint analysis showed that compared with participants with number of healthy sleep factors≤2 and metabolic diseases≥2, participants with 4 healthy sleep factors and no metabolic diseases had a substantially lower risk of stroke (HR: 0.27, 95% CI: 0.16, 0.44; Figure 1).

Table 4 Association of Healthy Sleep Score with Total Stroke Mediated by Metabolic Diseases

Figure 1 The joint association of healthy sleep score and metabolic diseases with stroke. Model was adjusted for age, sex (men or women), education (illiterate and primary, middle school, high school or college and above), average income per month (≤1000, 1000–3000, >3000RMB), occupation (coalminers, other blue collars, white collars), physical activity (no, occasion, frequent), alcohol drinking (never and past, light drinking, moderate-heavy drinking), smoking (never, past, current), use of sleep medication (no or yes), and CKD (no, yes).

In the sensitivity analyses, the results were generally robust when excluding incident cases occurring during the first two years of follow-up, coal miners, patients with CKD (Table S1).

Discussion

In this large-scale community-based study conducted in China, we found that higher healthy sleep score (combination of no insomnia, no excessive daytime sleepiness, sleep 7–8h/d, and no frequent snoring), was associated with lower total stroke and its subtypes risk. For each number increment in low-risk sleep factor, there was an 14% lower risk of total stroke. In addition, the metabolic disease, including obesity, diabetes, and hypertension, collectively explained 21.9% of the association between healthy sleep score and stroke. In addition, individuals with 4 healthy sleep factors and no metabolic diseases had approximately 70% lower risk of stroke than those with number of healthy sleep factors≤2 and metabolic diseases≥2.

Our study contributes to the literature regarding the influence of sleep patterns on stroke risk, which were mainly conducted in Western populations. A study using data from the UK Biobank creating a healthy sleep score using chronotype, duration, insomnia, snoring, and excessive daytime sleepiness, also found that participants with higher sleep pattern scores (healthy sleep pattern) had a 34% (95% CI: 25%, 42%) lower risk of stroke.⁸ A recent retrospective case-control study also found a graded association between the accumulation of sleep disturbance symptoms (included extreme sleep duration, sleep onset latency, waking more than once, napping for >1h, unplanned napping and presence or uncertainty surrounding snorting, snorting or gasping and breathing cessation or choking) and stroke risk.¹⁰ Compared with participants with 0–1 sleep disturbance symptoms, participants with an increased number of sleep disturbance symptoms had a graded increase in stroke risk (2–3: OR: 1.63, 95% CI: 1.36–1.96; 4–5: OR: 3.08, 95% CI: 2.49–3.80; >5: OR: 5.38, 95% CI: 4.03–7.18).¹⁰ Our findings, together with these studies, suggest that the comprehensive index constructed using multi-dimensional sleep parameters could be useful to identify high-risk populations for stroke event.

The association between individual sleep complaints and stroke has been widely discussed. However, the conclusions were still controversial. Our study found that participants without excessive daytime sleepiness or frequent snoring were associated with a lower risk of stroke, while we did not observe significant associations of insomnia or sleep duration with risk of stroke, which were found in other studies.^31,32 Nevertheless, this also provides evidence to support the notion that using a sleep pattern may help us find a stable estimate of the sleep-stroke relationship.

Few studies have specifically explored the relationship between sleep and stroke subtypes. The Dongfeng-Tongji cohort, using a single question to assess sleep quality over the past six months, found a significant association between poor sleep quality and increased risk of ischemic and hemorrhagic strokes.¹¹ Similarly, a large retrospective case-control study found a dose-response association of sleep disturbance symptoms with ischemic and hemorrhagic stroke.¹⁰ Consistent with these studies, our study also found an association of healthy sleep score with ischemic and hemorrhagic stroke. Notably, our study, together with the above-mentioned studies, observed the effect of sleep on hemorrhagic stroke was more pronounced than ischemic stroke. The pathogenesis of stroke may provide some insight. The increased risk of hemorrhagic stroke induced by sleep disturbances may be largely attributed to its direct impact on blood pressure, as hypertension is a major risk factor for cerebral hemorrhage.³³ However, due to the small sample size of incident hemorrhagic stroke cases, we cannot exclude the possibility of chance findings, and the results should be interpreted with caution.

Our mediation analyses contribute to better understanding the lower risk of stroke associated with healthy sleep pattern. Our data showed that the association between healthy sleep pattern and stroke may be explained by metabolic disease, including obesity, hypertension, and diabetes. These metabolic conditions are well-established risk factors of stroke.^17,34,35 Sleep complaints were also reported to be associated with these chronic conditions.^13,36 Experimental studies demonstrated that poor sleep quality might induce leptin production, insulin resistance, and increased sympathetic nervous system activity, which could result in the aforementioned metabolic abnormalities.^{3–5,37–39} However, the moderate mediation proportion (21.9%) in this study indicated other crucial factors need to be identified. Nevertheless, in joint associations, we found that the individuals with 4 healthy sleep factors and no metabolic diseases had approximately 70% lower risk of stroke. Our findings support the importance of maintaining healthy sleep pattern and metabolic health simultaneously in stroke prevention.

Several limitations in this study should be noted. Firstly, the incident case number of hemorrhagic stroke were small, and future studies with a larger sample size are required to replicate our findings. Secondly, information on chronic conditions was collected at baseline, which may affect the inference of causal association. Thirdly, we evaluated the association between baseline sleep patterns and future stroke risk without accounting for changes in sleep habits, potentially leading to non-differential misclassification.⁴⁰ Additionally, sleep complaints were self-reported via questionnaire, introducing the possibility of measurement errors. Fourthly, because most of the participants from the Kailuan study were men (82.4%), they cannot be viewed as a representative sample of the general Chinese population. However, studying such a population could reduce residual confounding due to diverse socioeconomic factors and lifestyle patterns. Finally, residual confounding cannot be completely ruled out because of the limitation of observational study design, despite comprehensive adjustment for the potential confounders.

Conclusion

In this large cohort study in China, our findings suggested that adherence to a healthy sleep pattern, including no insomnia, no excessive daytime sleepiness, sleep 7–8h/d, and no frequent snoring, was associated with a lower risk of stroke. In addition, our data showed significant mediation effect of metabolic diseases on the association between healthy sleep pattern and stroke risk. Our findings highlight the importance of maintaining both healthy sleep pattern and normal metabolism to reduce the risk of stroke.