Nonapnea sleep disorders are associated with subsequent ischemic stroke risk: a nationwide, population-based, retrospective cohort study.

Original Article

Nonapnea sleep disorders are associated with subsequent

ischemic

stroke risk: a nationwide, population-based, retrospective

cohort study

Wei-Shih Huang

, Chon-Haw Tsai

, Cheng-Li Lin

, Fung-Chang Sung

,

Yen-Jung Chang

,

Chia-Hung Kao

Stroke results in a sudden loss of neurologic function caused by disrupting the blood flow to the brain and is the leading cause of death globally [1,2]. In Taiwan, the major type of stroke is ischemic stroke [2,3], which is similar to that in the rest of the world. Stroke also is a leading cause of disability in adults [4]. Identification of risk factors for stroke is important for the primary and secondary prevention of stroke.

Sleep disorders constitute some of the most unrecognized modifiable risk factors for stroke [5,6]. The field of sleep disorders has

strongly focused on obstructive sleep apnea (OSA), which relates to an increased risk for stroke and cardiovascular disease (CVD). OSA is a treatable form of sleep-disordered breathing, in which

the upper airway repeatedly closes during sleep. OSA remarkable increases the risk for stroke or death from any cause, and the increase

is independent of other risk factors including hypertension

[7]. Several studies have shown a prevalence of OSA among patients with stroke that exceeds 60% [8–11], compared with 4% in

the middle-aged adult population [12].

However, studies investigating the involvement of nonapnea sleep disorder (NSD) in increasing subsequent ischemic stroke risk are scant. Of the studies conducted thus far, habitual sleep patterns have never been found to be important neurobehavioral determinants in the risk for ischemic stroke in postmenopausal women

[13]. A meta-analysis of 15 prospective population-based studies showed that short sleep duration (66 h/night) and long sleep duration (P9 h/night) are predictors of ischemic stroke [14]. Identifying and describing the relationship of NSD and stroke is

The objective of our nationwide, population-based, retrospective cohort study was to assess the association between NSD and the risk for acute ischemic stroke in Taiwan. We hypothesized that NSD was associated with the development of acute ischemic stroke. The study data were derived from Taiwan’s National Health Insurance Research Database (NHIRD).

2. Materials and methods 2.1. Data sources

Taiwan launched a single-payer National Health Insurance program on March 1, 1995. Virtually, the entire Taiwanese population (23 million individuals) is enrolled in this program. The database of this program, the NHIRD, contains registration files and original claim data for reimbursement. Large computerized databases derived from this system (http://nhird.nhri.org.tw/en/index.htm) are

provided to scientists in Taiwan for research purposes. From the original dataset of 23 million individuals, we randomly selected 1 million individuals with an identical sex and age distribution in relation to the entire Taiwanese population [15]. The data were retrospectively collected from 1996 to 2010 and included the registry of medical facilities, details of inpatients’ orders, ambulatory care, dental services, and prescriptions. The data files can be linked to scrambled identification to protect individual privacy. Previous studies showed the accuracy and high validity of diagnoses in theNHIRD[16,17]. The International Classification of Diseases, ninth revision, Clinical Modification (ICD-9-CM) was used to identify the diagnosis of disease.

The study was approved by the Ethics Review Board of China Medical University (CMU-REC-101-012).

2.2. Study participants

In our longitudinal cohort study, we selected patients with newly diagnosed NSD (ICD-9-CM codes, 307.4 and 780.5), except for sleep apnea syndrome (ICD-9-CM codes, 780.51, 780.53, and 780.57) between January 1, 1997 and December 31, 2001. We excluded patients who had any type of stroke (ICD-9-CM codes, 430–

438) before the index date. We conducted a systematic random sampling design to select a comparison cohort from rest of the insured population that was free from sleep disorders and stroke and

that was frequency matched by age (every 5 years), sex, and the year of the index date. The comparison to case ratio was 2 to 1.

Fig. 1 displays a flowchart diagram explaining the numbers of individuals at each stage of the study.

2.3. Outcome definition

Using the National Health Insurance inpatients files, we selected the study participants who were diagnosed with ischemic stroke (ICD-9-CM codes, 433–438) from January 1, 1997 to December 31, 2010. All study participants were followed up to measure the incidence of ischemic stroke until December 31st, 2010, or until being

censored due to death; until withdrawal from the insurance program; or until they were lost to follow-up. Moreover, the baseline

comorbidity history for each participant was identified, including hypertension (ICD-9-CM codes, 401–405), diabetes mellitus (DM) (ICD-9-CM code, 250), coronary artery disease (CAD) (ICD-9-CM codes 410–414), congestive heart failure (CHF) (ICD-9-CM codes, 398.91, 402, 404.01, 404.03, 404.10, 404.11, 404.13, and 404.9),

peripheral arterial disease (PAD) (ICD-9-CM codes, 440.2, 440.3, 444.2, and 444.81), atrial fibrillation (ICD-9-CM codes 427.31 and

427.32), and hyperlipidemia (ICD-9-CM code, 272). 2.4. Statistical analysis

The SAS statistical package (version 9.2 for Windows; SAS Institute, Inc., Cary, NC, USA) was used to manage and analyze the data.

The distributions of categorical demographic variables and comorbidities were compared between NSD and the comparison cohort.

The differences among categorical variables were analyzed using a

v

tested using t tests. Likewise, the incidence densities rate by demographic variables and comorbidities were calculated for each cohort.

We calculated hazard ratio (HR) and 95% confidence

intervals (CI) for each variable to assess the NSD cohort to comparison cohort rate ratios for ischemic stroke. Cox proportional hazards regression analysis was used to assess the effects of NSD on the risk for ischemic stroke, adjusting for covariates related to NSD. Comorbidity-specific analyses and Kaplan–Meier survival analysis with a log-rank test were conducted. Sleep disorder–specific analyses of incidence rates and HRs of ischemic stroke also

were conducted. Statistical significance was accepted at

a

3. Results

In total, we analyzed data collected from 94,160 participants as a comparison cohort and 47,080 participants as an NSD cohort, with the diagnosis date as the index date. The average follow-up duration was 9.61 years for the NSD cohort and 9.42 years for the comparison cohort. In our study, most participants were women and were aged 35–50 years (Table 1). The age range of cohorts was 20.0–101.7 years. NSD patients were more likely to have hypertension, DM, CAD, and CHF than the comparison group (P < .0001).

3.2. End points

Among NSD and comparison cohorts, the incidence density

rates of ischemic stroke were 8.87 and 6.00 per 1000 individualyears, respectively (Table 2). The overall risk for ischemic stroke

was higher in the NSD cohort than in the comparison cohort, with an adjusted HR of 1.19 (95% CI, 1.14–1.24). 3.3. Subgroup analyses Stratified by sex, the incidence density rates were 11.6 and 7.44 per 1000 individual-years in men and women with NSD as well as

a 49% and a 48% increase in ischemic stroke risk compared to comparison cohorts, respectively. However, our study showed a

1.35-fold significantly higher risk for developing ischemic stroke in men compared to women (95% CI, 1.30–1.41). Stratified by age, the incidence density rate was highest in the age group that was older than 65 years (27.1/1000 individual-years) in the NSD cohort. The age-specific analysis showed that patients with NSD aged 635 years had the highest risk for ischemic stroke compared to the comparison cohort (HR, 4.32 [95% CI, 2.75–6.79]). However, the adjusted HR was 31.2 (95% CIs, 25.2–38.6) for elderly patients compared with participants aged 635 years.

The comorbidity-specific analysis showed that NSD was associated with the risk for ischemic stroke in the noncomorbid participants (Table 3). The multivariate Cox regression model analysis showed that all comorbidities, including hypertension, DM, CAD, CHF, PAD, atrial fibrillation, and hyperlipidemia, were significant factors associated with the increased possibility of ischemic stroke. We further examined the association between the risk for ischemic stroke and subgroups of NSD, and we found that the adjusted HR was 1.19 for insomnia, 1.87 for hypersomnia, and 1.16 for other sleep disorders (Table 4). In addition, results showed a 1.35-fold

(95% CI, 1.30–1.41) higher risk for developing ischemic stroke in men compared to women. The adjusted HR was 31.2 (95% CI, 25.2–38.6) for elderly patients compared with participants aged 635 years.

The Kaplan–Meier survival analysis showed that the incidence of ischemic stroke was 3.22% higher in the NSD cohort than in the comparison cohort (log-rank; P < .0001; Fig. 2).

4. Discussion

Our findings suggested that the NSD cohort had a 1.19-fold

higher risk for subsequent ischemic stroke than the comparison cohort, after adjusting sex, age, and comorbidities. Comorbidities

including hypertension, DM, CAD, CHF, PAD, atrial fibrillation, and hyperlipidemia were significant factors associated with the occurrence of ischemic stroke. Our study also showed a greater risk for ischemic stroke among the men and the elderly. We further examined the association between the risk for ischemic stroke and subgroups of NSD, and the adjusted HR was 1.19 for insomnia, 1.87 for hypersomnia, and 1.16 for other sleep disorders (Table 4). Sleep disorders are highly prevalent in patients with stroke and in those at risk for stroke. Insomnia is characterized by a complaint of difficulty falling asleep, staying asleep, early morning awakening, or by nonrestorative sleep, and it is the most common sleep disorder; up to 22% of middle-aged and older Americans have reported chronic insomnia [18,19]. It has been reported that insomnia is associated with all-cause mortality (odds ratio, 1.7 [95% CI, 1.2– 2.5]) and cardiovascular death (odds ratio, 1.8 [95% CI, 1.0–3.1]) [20]. Increased sympathetic activity and hypothalamic–pituitary– adrenal axis activation leading to increased cortisol levels are proposed mechanisms for the effect of insomnia on vascular disease

[21]. The rates of ischemic stroke are lowest during nocturnal sleep and then increase after awakening [22]. The alteration of the immune system, particularly after awakening; abnormalities in the

autonomic nervous system through the hyperactivation of the sympathetic response; and the hypothalamic–hypophyseal axis might

all contribute to the link between insomnia and subsequent ischemic stroke [23,24]. Nakazaki et al. [25] discovered that insomnia and short sleep duration were associated with atherosclerosis risk. Atherosclerosis and incident hypertension might predispose patients

to suddenly reduced or occluded blood flow to the brain,

which might explain why patients with insomnia had a higher risk for developing ischemic stroke in our study (adjusted HR, 1.19). A prospective study of participants with a low risk for OSA and an optimal body mass index revealed a 4-fold greater risk for stroke in participants who had less than 6 h of sleep per night, compared to their counterparts who had 7–8 h of sleep per night [6]. Another study found that short sleep duration (<7.5 h) was independently associated with the risk for stroke and silent cerebral infarcts [26]. Additional evidence came from a Nurses’ Health Study, which indicated that nurses experienced a 4% increased risk for ischemic stroke for every 5 years of rotating shift work [27]. On the other hand, a U-shaped relation between sleep duration and vascular outcomes has been described. Short sleep (66 h of sleep) was associated with DM, obesity, incident hypertension, and CAD; and long sleep duration (P9 h of sleep) was related to

stroke and CAD [14,28–32]. A meta-analysis of 15 prospective population-based studies concluded that both short sleep duration and

long sleep duration were predictors of ischemic stroke [14]. Short sleep has been shown to impair insulin sensitivity, increase sympathetic tone and cortisol levels, and alter inflammatory markers

[31,32]. However, the mechanisms by which long sleep is associated to cardiovascular comorbidities are not well-understood.

There is a continuing controversy on whether or not long sleep duration is a risk factor for CVD or if it is only a marker of underlying poor health. Findings from the Northern Manhattan Study

suggested that daytime sleepiness was an independent risk factor for stroke and other vascular events [33]. In our study, we also found a relationship trend between the hypersomnia and the risk for ischemic stroke (adjusted HR, 1.87; Table 4).

Sleep-related movement disorders and their impact on incident stroke have received much less attention than OSA and sleep duration; however, there is mounting evidence that suggests that these

disorders also may contribute to elevated stroke risk. Restless legs syndrome (RLS) is one of the most common sleep disorders, with population-based surveys estimating its frequency between 5% and 10% of the general population in North America and Europe [34–38]. The pathophysiology of RLS implicates deficiencies in central

dopaminergic transmission, which is vital in modulating spinal excitability [39]. Periodic limb movements (PLMs) occur in up to 80% of patients with RLS and are the cause of sleep fragmentation [39]. These repetitive stereotypical limb movements often are described as ‘‘jerks’’ by bed partners and can cause brief arousals or

full awakenings from sleep. Evidence of causative roles for RLS and PLMs in ischemic stroke is lacking; however, RLS and PLMs have been associated with increased risk for hypertension and

CVD in some epidemiologic studies after controlling for well-established vascular risk factors [34,38]. In a sample of 861 RLS patients,

those with PLMs of >30 an hour had a 2-fold increased risk for hypertension [40]. Thus RLS and associated PLMs may mediate the risk for CVD, hypertension, and incident stroke through repeated nocturnal sympathetic arousals associated with blood pressure and heart rates surges [41]. However, because of the coding system used in Taiwan is ICD-9-CM instead of ICSD, we were unable to distinguish those patients with sleep-related movement

disorders from patients with insomnia in our study.

A significant association existed between NSD and comorbidities, such as hypertension, DM, CAD, and CHF in our study, which is consistent with findings of previous studies [25,42,43]. Moreover, most patients in the NSD cohort in our study were women.

Several studies have found a female preponderance of insomnia, and a meta-analysis by Zhang and Wing [44] confirmed an excess of insomnia risk in women [45,46]. However, our study found that men with NSD had a higher risk for developing ischemic stroke than women after adjusting for age and comorbidities. The NSD to reference HR of ischemic stroke risk was higher in younger adults than in the middle-aged and older adults in our study. However, the risk for subsequent ischemic stroke increased with age after adjusting for sex and comorbidities, as the older adults had a higher proportion of comorbidities interacting with NSD to develop ischemic stroke. This epiphenomenon can be explained by the fact that ischemic stroke risk increased when the

patients had other comorbidities in addition to NSD. However, the association between NSD and the risk for ischemic stroke only reduced if individual comorbidity was accounted for. This phenomenon can be explained by effect modification.

Our study had some limitations. First, the NHIRD does not provide detailed patient information, such as smoking habits, alcohol consumption, body mass index, physical activity, socioeconomic

status, family history, and medication use, which are all potentially confounding factors for our study. Second, unlike randomized trials,

a retrospective cohort study design is subject to biases related to confounder adjustments. Despite the fact that our meticulous

study design adequately controlled for confounding factors, the potential bias caused by unmeasured or unknown confounders could

be a key limitation.

To the best of our knowledge, our population-based retrospective cohort study is the first study to provide epidemiologic data on the association between NSD and subsequent ischemic stroke in Taiwanese population. Taiwanese individuals exhibit similar characteristics of stroke compared to the rest of the world population [2,3]; however, the comparison in characteristics of sleep disorders between Taiwanese and global population is inadequate.

Therefore, researchers should be cautious about the generalizability of our study results. Although the study cannot establish causality, our results suggest that clinical and research

collaborations between ischemic stroke and sleep specialists should be encouraged to improve knowledge, prevention strategies, and clinical outcomes of patients with ischemic stroke. Sleep disorders screening through questionnaires could become standard procedure in stroke clinics.

In summary, we found that patients with NSD were at increased risk for developing ischemic stroke compared to those without diagnosed sleep disorders. Because the number of patients with NSD is progressively increasing, enhancing sleep disorder management may be crucial in preventing subsequent ischemic stroke. It is

vital to provide education strategies for healthcare professionals and the public regarding the importance of sleep and its association with ischemic stroke.