Interaction between obstructive sleep apnea and short sleep duration on insulin resistance: a large-scale study

Objectives Both short sleep duration and obstructive sleep apnea (OSA) seem to be associated with insulin resistance. We aimed to explore whether short sleep duration modifies the relationship between OSA and insulin resistance. Methods Participants were consecutively enrolled from our sleep center during the period from 2007 to 2017. The index of homeostasis model assessment insulin resistance (HOMA-IR) was calculated from insulin and glucose. Sleep duration was determined by standard polysomnography. The associations between sleep duration and insulin resistance were estimated by logistic regression analyses. Results A total of 5447 participants (4507 OSA and 940 primary snorers) were included in the study. OSA was independently correlated with insulin resistance after adjusting for all potential confounders (OR, 1.319; 95% CI, 1.088–1.599), but not short sleep duration. In stratified analysis by sleep duration, compared with primary snorers, in the OSA group only extremely short sleep duration (< 5 h) was significantly associated with insulin resistance after adjusting for all covariates (OR, 2.229; 95% CI, 1.283–3.874). Rapid eye movement predominant OSA was significantly associated with insulin resistance (OR = 1.355, 95% CI: 1.019–1.802) after adjustment for confounding factors including age, sex and body mass index. Conclusions OSA, but not short sleep duration, was independently associated with insulin resistance. It is worth noting that OSA combined with extremely short sleep duration showed a greater detrimental effect than OSA itself with regard to insulin resistance.


Introduction
Obstructive sleep apnea (OSA), which is the most common form of sleep-disordered breathing, affects about 34% of men and 17% of women in middle to older age groups [1]. A recent meta-analysis estimated that 936 million adults worldwide aged 30-69 years have mild to severe OSA. In parallel, the prevalence of insufficient sleep duration due to bedtime restriction, short-term sleep fragmentation, and sleep deprivation is also increasing in modern societies. Both OSA and short sleep duration have attracted a great deal of attention, not only as they are common globally, but also as they are considered to be independent risk factors for cognitive, metabolic, and cardiovascular problems [2][3][4][5][6].
One study showed that short sleep duration is independently associated with hypertension in OSA patients [7]. However, another study indicated that OSA, but not short sleep duration, was independently associated with obesity, hypertension, and dyslipidemia [8]. In patients with prediabetes, self-reported short sleep duration was associated with adverse measures of glycemia [9]. However, after considering the effect of OSA, sleep duration was not significantly associated with an abnormal fasting glucose level [10]. Considering these inconsistent results, studies on OSA should account for sleep duration and vice versa.
Several cross-sectional and longitudinal studies showed that OSA was independently correlated with insulin resistance [11,12]. Insulin resistance may also mediate the relationships of common cardiovascular risk factors with OSA. Similarly, short sleep duration is also correlated with lower insulin sensitivity [13] and increased insulin resistance, even after adjusting for adiposity [14]. Recently, insulin resistance, characterized by a decreased cellular response to insulin, was shown to have a significant impact on various adverse clinical outcomes, including hypertension, diabetes mellitus, nonalcoholic fatty liver disease, cardiovascular diseases, and even kidney diseases [15]. However, few recent studies have explored whether OSA and short sleep duration are independently correlated with insulin resistance. One study showed that subjects in the shortest sleep duration group (< 5 h) had 14.5% higher fasting insulin level and 16.3% higher homeostasis model assessment (HOMA)-β; these associations disappeared after adjustment for body mass index (BMI) and sleep apnea [16]. However, this previous study did not assess OSA by standard polysomnography (PSG) and included no objective measure of sleep duration. In addition, these associations have not been explored in Han Chinese populations.
We used a large-scale hospital-based cross-sectional dataset, including 5447 participants, to examine 1) whether objectively measured short sleep duration and OSA are independently associated with insulin resistance, and 2) whether the presence of OSA modulates the association between sleep duration and insulin resistance.

Methods
This cross-sectional study enrolled consecutive patients referred to the sleep center of Shanghai Jiao Tong University Affiliated Sixth People's Hospital. The study procedure was approved by the Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People's Hospital in accordance with the Declaration of Helsinki (Trial registration number: ChiCTR1900025714). Informed consent was obtained from all subjects.

Participants
All potential participants completed a comprehensive questionnaire regarding their general health status, including smoking and alcohol use (never, past, or current), and medication use. Smoking were defined as current smokers or subjects who smoked cigarettes before. Non-smoking were defined as subjects who had never smoked cigarettes. The exclusion criteria were as follows: 1) aged < 18 years; 2) previous treatment for OSA; 3) absence of full-lead PSG data; 4) chronic diseases, such as hepatic, pulmonary, or cardiac disease; 5) comorbid sleep disorders, such as insomnia, central sleep apnea (CSA), upper airway resistance syndrome (UARS), restless leg syndrome (RLS), and narcolepsy; 6) current use of anxiolytics, hypnotics, antidepressants, or other antipsychotics; and 7) had anti-diabetes drugs.

Clinical and biochemical measurements
Height, weight, waist circumference (WC), neck circumference (NC), and hip circumference (HC) were measured as described previously, without shoes and while wearing lightweight clothing [17]. BMI was calculated as weight in kilograms divided by height in meters squared. After a rest period of about 15 min, the systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured using a standard mercury sphygmomanometer. All of the above indices were measured twice and mean values of the two measurements were calculated.
Fasting venous blood was collected from each participant at 07:00. The lipid profile [total cholesterol (TC), triglycerides (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), apolipoprotein A-I (ApoA-I), apolipoprotein B (ApoB), apolipoprotein E (ApoE), and lipoprotein(a) (Lp(a))] was measured in the clinical laboratory of our hospital. Fasting serum glucose was determined using an H-7600 autoanalyzer (Hitachi, Tokyo, Japan), and serum insulin was determined by an immunoradiological method. The index of homeostasis model assessment insulin resistance (HOMA-IR) was calculated as fasting insulin (μU/ mL) × fasting glucose (mmol/L)/22.5 [18]. Insulin resistance was defined as HOMA-IR ≥ 2.5 [19]. Hypertension was defined as diagnosis provided by the referring physician, and by the use of cardiovascular medications that were specifically used as antihypertensive medications. Diabetes was defined as the use of antidiabetic medication and fasting plasma glucose ≥7.0 mmol/L. Dyslipidemia was defined as use of lipid-lowering medication or serum TC level > 200 mg/dL, LDL-C level > 130 mg/dL, HDL-C level < 39.8 mg/dL, or TG level > 150 mg/dL, according to the diagnostic criteria of the National Cholesterol Education Program Adult Treatment Panel III (NCEP ATP III) [20].

Sleep evaluation
All participants completed the Chinese version of the Epworth Sleepiness Scale (ESS) to evaluate daytime sleepiness; the total ESS scores range from 0 to 24 [21]. An ESS score > 10 was defined as excessive daytime sleepiness (EDS) [22].
All participants underwent overnight standard PSG in a sound-attenuated, light-and temperature-controlled room in our sleep center. Participants were asked to arrive at least 2 h before the sleep study to familiarize themselves with the sleep environment. Bedtime was adjusted according to participants' habitual sleep time, with the recording time ranging from 21:00-22:00 to 06:00-07:00.

Statistical analysis
All statistical analyses were performed using SPSS software (version 22.0; IBM SPSS Statistics, IBM Corp., Armonk, NY). Normally distributed continuous data are shown as means ± SD and were analyzed using one-way analysis of variance (ANOVA) and the independentsamples t test. Skewed data are presented as interquartile range (IQR) and were compared by the Mann-Whitney U test. Categorical data are presented as percentages and were analyzed using the χ 2 test. P-values for linear trends across groups were calculated using the polynomial linear trend test.
Independent associations of OSA and sleep duration with insulin resistance were assessed by logistic regression. The covariates adjusted for in our study included age, gender, BMI, smoking, alcohol consumption, hypertension, hyperlipidemia, time in bed, time in sleep stage N3 (%), ESS, AHI (or TST), waist hip ratio, diabetes mellitus, ApoA-I, ApoB, ApoE, and Lp(a). The adjusted odds ratio (ORs) and 95% confidence intervals (CIs) for insulin resistance were calculated according to OSA status and sleep duration. Finally, the joint effects of OSA and sleep duration were examined using primary snorers with a sleep duration of 7-8 h as the reference group. In all analyses, P < 0.05 was taken to indicate statistical significance.

Results
A total of 5447 participants (4507 OSA and 940 primary snorers) were included in this study; their demographic and sleep characteristics by OSA severity are presented in Table 1. Patients with OSA were predominantly male, more obese, older, and had poorer metabolic and sleep profiles. Table 2 presents the demographic characteristics and sleep parameters by sleep duration. Patients with shorter sleep duration were older, less obese, had  lower AHI, and lower likelihood of insulin resistance ( Table 2). In whole cohort, compared with primary snorers (non-OSA), OSA was independently correlated with insulin resistance after adjusting for all potential confounders (OR, 1.319; 95% CI, 1.088-1.599) ( Table 3). Taking a sleep duration of 7-8 h (optimal sleep time) as the reference, 5-6 h and < 5 h of sleep were associated with an increased risk of insulin resistance (OR, 1.127; 95% CI, 0.921-1.379; and OR, 1.167; 95% CI, 0.927-1.417, respectively) after adjusting for all potential confounders, but not significantly (Table 3).
In stratified analysis by OSA severity, we performed logistic regression analyses to evaluate the correlation between insulin resistance and sleep duration in each subgroup (in mild, moderate and severe OSA). When we use primary snorers (non-OSA) as a reference, no significant differences were found in each OSA group (Table 5).
In order to explore the effect of interaction between OSA and sleep duration on insulin resistance, multiplicative interaction was evaluated with logistic regressions. After adjusting for age, sex, BMI, smoking, alcohol use, hypertension, hyperlipidemia, time in bed, S3, ESS and waist hip ratio, an interaction effect of OSA and extremely short sleep duration (< 5 h) on insulin resistance was found (OR, 1.467; 95% CI, 1.018-2.116), taking primary snorers with a sleep time of 7-8 h as the reference group (Table 7). This interaction disappeared after additional adjustment for diabetes mellitus, ApoA-I, ApoB, ApoE, and Lp(a).
In supplementary Table 1, REM predominant OSA was significantly associated with insulin resistance (OR =  Besides, in multiple logistic regression models, we found almost near all variables listed were significantly associated with insulin resistance (supplementary Table 2).

Discussion
In this large-scale cross-sectional study, only OSA was significantly correlated with insulin resistance even after adjustment for covariates, including sleep duration. OSA and sleep time < 5 h had an interaction effect on insulin resistance.
The relationship between OSA and insulin resistance has been well documented. AHI was independently correlated with insulin resistance even after adjusting for covariates [25]. Our previous study showed that intermittent hypoxia was independently associated with hyperglycemia, hyperinsulinemia, and insulin resistance [11]. In the Sleep Heart Health Study (SHHS study), sleep fragmentation, another important characteristic of OSA, was also shown to be associated with insulin resistance [26]. Consistent with the above studies, we also found that OSA was independently correlated with insulin resistance, even after controlling for sleep duration. Potential mechanisms linking OSA and insulin resistance include sympathetic nerve excitability, oxidative stress, inflammation, and hypothalamic-pituitary-adrenal (HPA) axis dysfunction [27].
The conclusions of previous studies exploring the correlation between insulin resistance and sleep duration were inconsistent. One study showed that, in comparison to the reference group with a sleep time of 6-8 h, sleep times of < 6 and > 8 h tended to be correlated with lower rates of insulin resistance [28]. Interestingly, short sleep duration was reported to be correlated with insulin resistance in women but not in men [14]. The mechanism by which insulin resistance induces short sleep duration was suggested to involve activation of the HPA axis, resulting in inhibition of pancreatic function and, finally, reduced glucose tolerance and insulin resistance [29]. Other clinical studies reported contradictory findings. One study indicated that neither short nor long sleep duration was correlated with insulin resistance in older non-diabetic patients [30]. Another study showed that long, but not short, sleep duration was associated with insulin resistance in non-diabetic subjects with or without sleep disorders, such as sleep apnea and insomnia [31]. Elsewhere, a long sleep duration was not significantly correlated with HOMA-IR [14]. These discrepancies among studies may have been due to differences in ethnicity, the characteristics of the enrolled participants, and gender ratios, as well to insufficient sample sizes and inadequate adjustment for confounders. In the present study, we used data from a large Han Chinese population and found that OSA was independently associated with insulin resistance, but not sleep duration. All of the participants were enrolled from our sleep center and most had severe OSA. Thus, OSA may have masked the small effect of sleep duration on insulin resistance. Interestingly, the risk of insulin resistance in those with OSA and extremely short sleep duration (< 5 h) was comparable to that of primary snorers with the same  sleep duration. These observations suggested that OSA combined with extremely short sleep duration had a greater detrimental effect on insulin resistance than OSA alone. Improving extremely short sleep duration may also improve insulin resistance in OSA. REM sleep accounts for about 25% of the TST and is associated with distinct physiologic alterations. Few previous studies explored the associations of insulin resistance with respiratory events and sleep duration during REM sleep period in OSA. In our study, we preliminary found that REM predominant OSA was associated with insulin resistance to some extent. However, prospective studies with larger sample size are warranted to verify our findings.
The main finding of our study is that OSA is independently associated with insulin resistance rather than short sleep duration. Continuous positive airway pressure (CPAP) is the first-line treatment of OSA and it could improve insulin resistance through abolishing disturbed respiratory events during sleep. Besides, we also found that REM predominant OSA was associated with insulin resistance to some extent. Thus, clinicians should pay more attention to increase CPAP usage time during REM sleep alleviate insulin resistance of OSA.
Despite the large sample size, our study had several limitations. First, as this was a cross-sectional study, it was not possible to determine the causal relationships of short sleep duration, OSA, and insulin resistance. Second, our sample was composed predominantly of male patients, who typically had OSA. This imbalance may have led to bias in the prediction models. Third, as this was a hospital-based study, the generalizability of our conclusions to asymptomatic patients, such as those residing in the community, is unclear. Fourth, though we carefully excluded participants who had medication which could affect sleep or insulin resistance, however, other drugs such as anti-hypertension and lipid-lowering drugs might also potentially influence the reported associations. Fifth, sleep duration was determined based on one overnight PSG session, but night-to-night variability and first-night effects cannot be excluded. In contrast, actigraphy performed using a simple portable can record sleep duration continuously for 7 days. Although PSG was not the optimal choice, good agreement between actigraphy and PSG with respect to the monitoring of sleep duration has been established in OSA [32].

Conclusion
In conclusion, the results of the present study suggested that OSA, but not short sleep duration, is correlated with insulin resistance. It is worth noting that short and extremely short sleep durations (< 5 h) combined with OSA have a more detrimental effect than OSA alone with regard to insulin resistance. Further prospective cohort studies are warranted to determine the causal relationships between OSA, short sleep duration, and insulin resistance, and to ascertain whether improving sleep time in turn improves insulin resistance in patients with OSA.