2023 Volume 70 Issue 1 Pages 107-119
We aimed to identify the association between obstructive sleep apnea (OSA), insulin resistance (IR), and metabolic syndrome (MetS) in a nationwide population-based sample. A total of 7,900 adults with information on the STOP-Bang score and MetS (3,341 men and 4,469 women) were identified from the dataset of the Korea National Health and Nutrition Examination Survey 2019–2020. The association between OSA, IR, MetS, and its components was estimated using complex sample logistic regression analysis with adjustments for age, body mass index, residence, smoking status, alcohol consumption, household income, education, and the presence of diabetes. Participants with a high OSA score were more likely to have IR (odds ratio [OR] 2.78, 95% confidence interval [CI] 1.96–3.95 in men and OR 2.64, 95% CI 0.55–12.80 in women), MetS (OR 6.05, 95% CI 4.23–8.69 in men and OR 4.20, 95% CI 1.23–15.70 in women), and components of MetS, compared to individuals with a low OSA score. Compared to premenopausal women, postmenopausal women had a more intense association between OSA and IR (OR 1.78, 95% CI 0.13–24.43 for premenopausal women and OR 3.64, 95% CI 0.60–22.28 for postmenopausal women) and MetS (OR 2.58, 95% CI 0.23–29.55 for premenopausal women and OR 5.36, 95% CI 1.42–20.21 for postmenopausal women). OSA was associated with abdominal obesity and hypertension in premenopausal women, while all components of MetS were associated with OSA in postmenopausal women. Further studies are necessary to elucidate the underlying mechanisms of these findings.
OBSTRUCTIVE SLEEP APNEA (OSA) is a sleep-related breathing disorder characterized by repetitive collapse of respiratory passage during sleep [1]. The prevalence of OSA is high and increasing [1], partially explained by the increasing rates of obesity, which is considered a major risk factor for the development of OSA [2]. The pathophysiological response to recurrent pauses during sleep includes intermittent hypoxia, systemic inflammation, metabolic dysregulation, and activation of the sympathetic nervous system, leading to pathological clinical consequences such as cardiovascular diseases [3]. In addition, chronic intermittent hypoxia results in a systemic inflammatory response; therefore, OSA can serve as a substrate for cardiovascular [4], metabolic [5], and neurodegenerative diseases [6].
The apnea-hypopnea index (AHI) is the most widely used tool to establish the diagnosis, evaluate the severity, and estimate the treatment response of OSA [7]. However, calculating AHI generally requires a sleep study, polysomnography (PSG), or sleep laboratory to where the study takes place. In contrast, the snoring, tiredness, observed apnea, high blood pressure, body mass index (BMI), age, neck circumference, and male sex (STOP-Bang) model for OSA can be calculated and can predict the possibility of OSA [8]. It is a reliable, easy-to-use, concise, and effective screening tool with eight dichotomous items related to the clinical features of OSA [9]. The validity of the STOP-Bang model has been evaluated in the general population, and the severity of OSA is robustly correlated with stratification using the STOP-Bang score [10]. The usefulness of this scoring system was further demonstrated in patients undergoing surgery, indicating the necessity of preoperative PSG with a higher STOP-Bang score [11].
Although the STOP-Bang score has been widely regarded as a convenient screening tool for OSA, few studies have examined the association between OSA and components of metabolic abnormalities using this scoring system in the general population. For example, OSA was independently associated with insulin resistance (IR) [12], metabolic syndrome (MetS), and its components [13]; however, these studies did not utilize the STOP-Bang model as a prediction tool for the presence of OSA. One study assessed the relationship between MetS and OSA using the STOP-Bang questionnaire, but the study participants were limited to male drivers [14]. Although a higher STOP-Bang score was associated with a higher probability of type 2 diabetes [15], a direct association between IR and OSA was not evaluated. Considering the increasing prevalence of IR and MetS and the usefulness of simple scoring systems in the real clinical field, it would be useful to investigate the relationship between the STOP-Bang score stratified risk of having OSA and those clinical conditions.
Therefore, the present cross-sectional study aimed to evaluate the association of the STOP-Bang score-based OSA with IR and MetS using the dataset of the Korea National Health and Nutrition Examination Survey (KNHANES) 2019–2020.
We used the dataset from the 2019–2020 KNHANES. The KNHANES is an annual nationwide population-based cross-sectional survey conducted by the Korea Centers for Disease Control and Prevention that was designed as a complex sample survey using a multistage sampling method to represent the general, non-institutionalized Korean population. A total of 20,808 individuals were selected for the survey during the study period, and 15,469 responded (response rate: 74.3%). Participants aged ≥40 years who completed the STOP-Bang questionnaire were selected (n = 8,061, 3,495 men, and 4,566 women). First, participants who had missing values for calculating IR (e.g., fasting insulin and glucose levels) were excluded when estimating the association between STOP-Bang and IR (n = 155). A total of 7,906 adults (3,430 men and 4,476 women) were selected to evaluate the relationship between the STOP-Bang and IR. Second, participants with missing values for defining the presence of MetS (e.g., triglyceride, waist circumference, high-density lipoprotein [HDL]-cholesterol, systolic and diastolic blood pressure, and fasting glucose levels) were excluded when evaluating the association between STOP-Bang and MetS (n = 161). A total of 7,900 adults (3,341 men and 4,469 women) were selected to evaluate the relationship between the STOP-Bang and MetS.
STOP-Bang questionnaireThe STOP-Bang system was originally validated as a screening tool to identify surgical patients at a high risk of OSA. However, this scoring system was useful as a screening tool for OSA not only in surgical patients but also in the sleep clinic population [16], and it presented a high concordance rate with PSG [17]. The STOP-Bang score is well correlated with the AHI obtained from PSG [17], and the usefulness of the scoring model as a prediction tool for OSA has been validated in Koreans [18].
The STOP-Bang scoring system consists of eight dichotomous (1 point for yes/0 point for no) items related to the clinical features of OSA, and the total score ranged from 0 to 8: (1) Do you snore loudly?; (2) Do you often feel sleepy, tired, or fatigued during the day?; (3) Has anyone observed that you stopped breathing during your sleep?; (4) Do you have (or are being treated) hypertension?; (5) BMI >35 kg/m2; (6) age >50 years; (7) neck circumference >40 cm; and (8) are you male? Individuals with a score of 0–2 were classified as low risk for OSA, and 5–8 were classified as high risk. In the case of a STOP-Bang score of 3–4, we classified individuals at intermediate risk for having OSA. However, individuals with a score of 2, male sex, BMI >35 kg/m2, or neck circumference >40 cm were classified as high risk for having moderate to severe OSA [9]. The percentages of the low-, intermediate-, and high-risk OSA groups in the current study according to sex and menopausal status are shown in Fig. 1.
The percentage of individuals with low-, intermediate-, and high-risk of obstructive sleep apnea, according to sex and menopausal status. 1) men, 2) women, 3) premenopausal women, and 4) postmenopausal women.
IR was measured using the homeostatic model assessment for insulin resistance (HOMA-IR) equation. The score was obtained by multiplying the value of serum fasting insulin level (μIU/mL) by fasting glucose level (mg/dL) divided by 405. The cut-off values for IR were 2.2 in men, 2.55 in premenopausal, and 2.03 in postmenopausal women, according to a Korean reference study [19]. Serum insulin levels were measured using an electrochemiluminescence immunoassay with modular E801 (Roche, Germany). Fasting glucose levels were measured using hexokinase UV (Hitachi Automatic Analyzer Labospect 008AS, Hitachi, Japan).
MetS was defined according to the modified Third National Cholesterol Education Program Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (NCEP-ATP III) [20] and the abdominal obesity criteria of the Korean Society for the Study of Obesity [21]. MetS was diagnosed when three or more of the following criteria were present: (a) waist circumference ≥90 cm in men and ≥85 cm in women; (b) triglyceride ≥150 mg/dL; (c) HDL-cholesterol <40 mg/dL in men and <50 mg/dL in women; (d) blood pressure ≥130/85 mmHg or current antihypertensive medications; and (e) fasting glucose level ≥100 mg/dL or current antidiabetic medications. Blood samples were collected by a trained medical technician and transported to the central NEODIN Medical Institute (Seoul, South Korea). Serum triglyceride and HDL-cholesterol levels (mg/dL) were measured using an enzymatic method (Hitachi Automatic Analyzer Labospect 008AS, Hitachi, Japan).
Other variables collection and measurementSystolic and diastolic blood pressure measurements were obtained on 3 consecutive occasions in a relaxed environment. The mean of the second and third measurements was used for the data analysis. Anthropometric measurements were performed on neck and waist circumference and BMI. BMI (kg/m2) was divided into four groups according to the criteria of the Korean Society for the Study of Obesity: underweight (<18.5), normal (18.5–22.9), pre-obese (23.0–24.9), and obese (≥25.0) [21]. Variables related to socioeconomic status were as follows: residence (urban or rural), educational level (middle school or lower, high school, or college or higher), household income (lowest, middle-low, middle-high, or highest), and smoking status (never, former, or current). Smoking status was categorized using the National Health Interview Survey [22]. A standard drink was defined as 10 g of alcohol, according to the World Health Organization. On occasion, high-risk alcohol consumption was defined as seven or more drinks for men and five or more for women [23].
Statistical analysisBecause the KNHANES was produced using a complex, multistage, stratified, and probability-verified sample design, all statistical analyses were performed using complex sample analyses, and sampling weights were applied. The detailed study protocol has been previously reported [24].
Given the sex-based differences in the context of OSA and metabolic diseases [25], the association between OSA, IR, and MetS was evaluated separately by sex. Categorical variables were compared using chi-square or Fisher’s exact tests and presented as weighted percentages with standard error (SE). Continuous variables were compared using Student’s t-test and presented as weighted numbers with SE. In addition, we compared the characteristics of study participants, separated according to menopausal status. The estimated prevalence of IR and MetS was calculated using a complex sample linear regression model, considering the risk of OSA as a continuous variable. Complex sample logistic regression analysis was used to calculate odds ratios (OR) with 95% confidence intervals (CI) for IR, MetS, and its components. We performed unadjusted analysis, age-and BMI-adjusted analysis (model 1), and multivariable-adjusted analysis considering residence, smoking status, degree of alcohol consumption, household income, educational level, and the presence of diabetes as covariates (model 2). Subgroup analysis was performed according to menopause status because the cut-off value of IR differed. Postmenopausal women receiving hormonal therapy were excluded from this analysis. In addition, we analyzed the association between OSA, IR, and MetS, stratified by BMI and neck circumference. All tests were two-tailed, and statistical significance was set at p < 0.05. All statistical analyses were performed using the IBM SPSS for Windows (version 24.0; IBM Corp., Armonk, NY, USA).
Descriptive statistics showed that men were younger and had a higher rate of current and former smokers, high-risk alcohol consumption, high-income level, educational level, diabetes, and pre-obese and obese status than women (Table 1). Regarding metabolic components and laboratory results, waist circumference, triglyceride, fasting glucose, systolic and diastolic blood pressure, insulin, hemoglobin A1c (HbA1c), HOMA-IR, aspartate aminotransferase (AST), alanine aminotransferase (ALT) level, and STOP-Bang scores were higher in men than in women, except for the HDL-cholesterol level. Clinical characteristics were significantly different between pre and postmenopausal women, except for diastolic blood pressure.
| Men (n = 3,495) |
Women (n = 4,566) |
p | Premenopause (n = 1,528) |
Postmenopause (n = 3,034) |
p | |
|---|---|---|---|---|---|---|
| Age (years) | <.001 | <.001 | ||||
| 40–49 | 30.0 (1.1) | 26.7 (1.0) | 69.4 (1.5) | 1.5 (0.3) | ||
| 50–59 | 30.5 (1.0) | 28.3 (0.9) | 23.5 (1.3) | 33.7 (1.2) | ||
| 60–69 | 22.3 (0.9) | 21.7 (0.7) | 4.6 (0.5) | 33.8 (1.1) | ||
| ≥70 | 17.2 (0.8) | 23.3 (0.9) | 2.4 (0.4) | 31.0 (1.1) | ||
| Residence | 0.084 | <.001 | ||||
| Urban | 81.7 (2.1) | 82.8 (1.9) | 87.0 (2.0) | 80.2 (2.2) | ||
| Rural | 18.3 (2.1) | 17.2 (1.9) | 13.0 (2.0) | 19.8 (2.2) | ||
| Smoking | <.001 | <.001 | ||||
| Never | 29.4 (0.8) | 88.9 (0.5) | 89.5 (0.9) | 93.3 (0.6) | ||
| Former | 36.7 (0.8) | 5.4 (0.3) | 5.5 (0.7) | 3.5 (0.4) | ||
| Current | 33.9 (0.8) | 5.7 (0.4) | 5.1 (0.7) | 3.2 (0.4) | ||
| High risk alcohol consumption⁋ | <.001 | <.001 | ||||
| <1/week | 62.9 (0.8) | 85.6 (0.7) | 85.4 (1.1) | 91.9 (0.8) | ||
| ≥1/week | 37.1 (0.8) | 14.4 (0.7) | 14.6 (1.1) | 8.1 (0.8) | ||
| Household income | <.001 | <.001 | ||||
| Lowest | 12.0 (0.6) | 16.0 (0.8) | 7.6 (0.8) | 27.7 (1.3) | ||
| lower middle | 24.4 (0.9) | 24.9 (0.8) | 23.7 (1.3) | 25.0 (1.0) | ||
| higher middle | 29.4 (0.8) | 28.5 (0.8) | 32.4 (1.5) | 23.3 (1.0) | ||
| Highest | 34.2 (1.2) | 30.6 (1.2) | 36.2 (1.7) | 23.9 (1.3) | ||
| Educational level | <.001 | <.001 | ||||
| Middle school or lower | 19.6 (1.0) | 33.7 (1.1) | 8.6 (0.8) | 49.5 (1.4) | ||
| High school | 34.8 (1.1) | 34.3 (0.9) | 36.8 (1.7) | 32.8 (1.2) | ||
| College or more | 45.7 (1.4) | 32.0 (1.2) | 54.6 (1.9) | 17.7 (1.1) | ||
| Type 2 diabetes | <.001 | <.001 | ||||
| Yes | 13.7 (0.7) | 10.9 (0.5) | 93.1 (0.8) | 79.6 (0.7) | ||
| No | 86.3 (0.7) | 89.1 (0.5) | 6.9 (0.8) | 20.4 (0.7) | ||
| BMI (kg/m2) | <.001 | <.001 | ||||
| <18.5 | 2.0 (0.3) | 2.9 (0.3) | 4.1 (0.6) | 2.2 (0.4) | ||
| 18.5–<23.0 | 25.9 (0.9) | 41.7 (0.9) | 46.9 (1.6) | 38.5 (1.1) | ||
| 23.0–<25.0 | 28.2 (0.9) | 23.0 (0.7) | 49.0 (1.6) | 59.3 (1.1) | ||
| ≥25.0 | 43.9 (1.0) | 32.4 (0.8) | ||||
| Metabolic components | ||||||
| WC (cm) | 89.1 (0.2) | 82.3 (0.2) | <.001 | 80.0 (0.3) | 83.7 (0.2) | <.001 |
| Triglyceride (mg/dL) | 168.4 (3.1) | 118.2 (1.4) | <.001 | 110.7 (2.2) | 122.9 (1.8) | <.001 |
| Fasting glucose (mg/dL) | 107.7 (0.6) | 100.9 (0.4) | <.001 | 97.0 (0.7) | 103.4 (0.5) | <.001 |
| HDL-cholesterol (mg/dL) | 48.4 (0.2) | 55.9 (0.2) | <.001 | 57.0 (0.4) | 53.8 (0.3) | <.001 |
| SBP (mmHg) | 122.6 (0.3) | 120.6 (0.3) | <.001 | 114.2 (0.4) | 124.6 (0.4) | <.001 |
| DBP (mmHg) | 78.9 (0.2) | 75.2 (0.2) | <.001 | 75.1 (0.3) | 75.2 (0.2) | 0.915 |
| Insulin (μIU/mL) | 9.41 (0.2) | 8.69 (0.1) | <.001 | 7.98 (0.2) | 9.13 (0.2) | <.001 |
| HbA1c (%) | 5.97 (0.0) | 5.87 (0.0) | <.001 | 5.65 (0.0) | 6.00 (0.02) | <.001 |
| HOMA-IR | 2.61 (0.0) | 2.27 (0.0) | <.001 | 2.01 (0.1) | 2.43 (0.1) | <.001 |
| AST (IU/L) | 27.2 (0.3) | 23.6 (0.2) | <.001 | 21.4 (0.3) | 24.9 (0.2) | <.001 |
| ALT (IU/L) | 28.1 (0.4) | 19.8 (0.2) | <.001 | 18.2 (0.4) | 20.9 (0.3) | <.001 |
| STOP-Bang score | 3.01 (0.0) | 1.54 (0.0) | <.001 | 0.91 (0.0) | 1.94 (0.0) | <.001 |
Data are presented as weighted mean (SE) for continuous variables or weighted percentage (SE) for categorical variables, unless otherwise stated.
⁋ High-risk alcohol consumption was defined as 7 or more drinks for men and 5 or more drinks for women on occasion.
STOP-Bang = the snoring, tiredness, observed apnea, high BP, BMI, age, neck circumference, and male gender, HOMA-IR = homeostatic model assessment for insulin resistance, BMI = body mass index, WC = waist circumference, HDL = high density lipoprotein, SBP = systolic blood pressure, DBP = diastolic blood pressure, HbA1c = glycosylated hemoglobin, AST = aspartate transaminase, ALT = alanine transaminase, SE = standard error.
The prevalence of IR, which was measured using HOMA-IR, increased according to the risk of OSA in men and women, irrespective of menopause status (Table 2). The estimated prevalence of IR in the high-risk OSA group was 61.5%, 65.4%, 69.4%, and 64.5% in men, women, premenopausal, and postmenopausal women, respectively. Both crude, age-and BMI-adjusted, and multivariable-adjusted models revealed a statistically significant trend toward an increasing OR for IR according to the risk of OSA. In addition, as the use of antidiabetic medications could significantly alter glucose and insulin levels, and therefore could affect HOMA-IR, we evaluated this association in participants who did not use antidiabetic drugs (Supplementary Table 1). A significant association between OSA and IR was also observed in the subgroup analysis of men, women, premenopausal- and postmenopausal women.
| Low | Intermediate OR (95% CI) |
High OR (95% CI) |
p trend | |
|---|---|---|---|---|
| Men | ||||
| Prevalence of IR, % (SE) | 32.1 (1.7) | 47.9 (2.2) | 47.7 (1.5) | <.001 |
| Crude | 1 | 1.76 (1.28–2.44) | 4.15 (2.76–6.23) | <.001 |
| Model 1 | 1 | 2.29 (1.58–3.31) | 4.67 (2.94–7.39) | <.001 |
| Model 2 | 1 | 1.57 (1.22–2.01) | 2.78 (1.96–3.95) | <.001 |
| Women | ||||
| Prevalence of IR, % (SE) | 29.8 (0.9) | 55.1 (2.1) | 72.9 (7.5) | <.001 |
| Crude | 1 | 2.99 (2.50–3.57) | 4.47 (1.87–10.69) | <.001 |
| Model 1 | 1 | 1.68 (1.36–2.07) | 2.42 (0.96–6.08) | <.001 |
| Model 2 | 1 | 1.88 (1.40–2.53) | 2.64 (0.55–12.80) | <.001 |
| Premenopause | ||||
| Prevalence of IR, % (SE) | 18.9 (1.3) | 39.6 (5.4) | 83.6 (11.3) | <.001 |
| Crude | 1 | 3.26 (2.10–5.04) | 9.78 (1.47–65.06) | <.001 |
| Model 1 | 1 | 1.62 (0.98–2.69) | 3.04 (0.39–23.72) | 0.037 |
| Model 2 | 1 | 2.25 (1.19–4.27) | 1.78 (0.13–24.43) | 0.019 |
| Postmenopause | ||||
| Prevalence of IR, % (SE) | 38.2 (1.2) | 58.3 (2.1) | 69.3 (9.3) | <.001 |
| Crude | 1 | 2.29 (1.89–2.79) | 2.95 (1.11–7.87) | <.001 |
| Model 1 | 1 | 1.71 (1.37–2.12) | 2.30 (0.85–6.22) | <.001 |
| Model 2 | 1 | 1.83 (1.32–2.52) | 3.64 (0.60–22.28) | <.001 |
Model 1 adjusted for age and BMI.
Model 2 additionally adjusted for residence, smoking, alcohol consumption, household income, education, and the presence of diabetes.
The risk of having OSA was calculated using the STOP-Bang questionnaires.
The IR was estimated using the HOMA-IR.
BMI = body mass index, STOP-Bang = the snoring, tiredness, observed apnea, high BP, BMI, age, neck circumference, and male gender, IR = insulin resistance, HOMA-IR = homeostatic model assessment for insulin resistance, OR = odds ratio, CI = confidence interval, SE = standard error.
The number of metabolic components increased with the severity of the risk of OSA (Fig. 2). For example, in the high-risk OSA group, the estimated number of the metabolic components were 3.08 (95% CI: 2.97–3.20) in men, 2.97 (95% CI: 2.54–3.39) in women, 3.60 (95% CI: 2.18–5.02) in premenopausal, and 2.83 (95% CI: 2.37–3.29) in postmenopausal women, showing p trend <.001 in all estimations. The prevalence of MetS increased significantly with higher STOP-Bang scores in men, women, premenopausal, and postmenopausal women (Table 3). In both the unadjusted and multivariable-adjusted models, the ORs of MetS were significantly increased as the risk of OSA was elevated in men, women, premenopausal, and postmenopausal women. When the participants were stratified by BMI and neck circumference, a significant relationship between OSA and MetS remained (Supplementary Table 2). However, in men, the strength of the association was slightly attenuated in the non-obese group and in individuals with a neck size ≤40 cm.
Estimated number of the components of metabolic syndrome according to the risk of obstructive sleep apnea by sex and menopause status. 1) men, 2) women, 3) premenopausal women, and 4) postmenopausal women.
| Low | Intermediate OR (95% CI) |
High OR (95% CI) |
p trend | |
|---|---|---|---|---|
| Men | ||||
| Prevalence of MetS, % (SE) | 17.4 (2.6) | 41.8 (1.1) | 71.3 (2.4) | <.001 |
| Crude | 1 | 3.40 (2.37–4.88) | 11.75 (7.70–17.94) | <.001 |
| Model 1 | 1 | 3.31 (2.17–5.06) | 9.20 (5.61–15.07) | <.001 |
| Model 2 | 1 | 2.68 (2.12–3.39) | 6.05 (4.23–8.69) | <.001 |
| Women | ||||
| Prevalence of MetS, % (SE) | 24.9 (0.9) | 57.4 (2.1) | 65.5 (10.6) | <.001 |
| Crude | 1 | 4.06 (3.33–4.96) | 5.75 (2.26–14.65) | <.001 |
| Model 1 | 1 | 2.44 (1.94–3.08) | 3.18 (1.24–8.17) | <.001 |
| Model 2 | 1 | 2.63 (1.88–3.67) | 4.20 (1.23–15.70) | <.001 |
| Premenopause | ||||
| Prevalence of MetS, % (SE) | 16.0 (1.1) | 51.9 (4.8) | 77.1 (20.1) | <.001 |
| Crude | 1 | 5.67 (3.69–8.71) | 17.70 (1.84–170.02) | <.001 |
| Model 1 | 1 | 3.10 (1.85–5.19) | 4.89 (0.46–51.55) | <.001 |
| Model 2 | 1 | 3.13 (1.63–6.02) | 2.58 (0.23–29.55) | 0.002 |
| Postmenopause | ||||
| Prevalence of MetS, % (SE) | 31.7 (1.2) | 58.6 (2.3) | 63.0 (12.2) | <.001 |
| Crude | 1 | 3.05 (2.47–3.76) | 3.67 (1.30–10.40) | <.001 |
| Model 1 | 1 | 2.31 (1.84–2.91) | 2.85 (1.03–7.92) | <.001 |
| Model 2 | 1 | 2.62 (1.82–3.78) | 5.36 (1.42–20.21) | <.001 |
Model 1 adjusted for age and BMI.
Model 2 additionally adjusted for residence, smoking, alcohol consumption, household income, education, and the presence of diabetes.
The risk of having OSA was calculated using the STOP-Bang questionnaires.
BMI = body mass index, STOP-Bang = the snoring, tiredness, observed apnea, high BP, BMI, age, neck circumference, and male gender, OR = odds ratio, CI = confidence interval, SE = standard error.
The associations between the risk of OSA and MetS components according to sex are summarized in Table 4. Participants of both sexes with high STOP-Bang scores were more likely to have abdominal obesity, high triglyceride levels, impaired fasting glucose levels, and hypertension, as compared to those with low STOP-Bang scores. In terms of low HDL cholesterol levels, a statistical trend was observed toward an increase in OR by STOP-Bang score.
| Men | Women | |||||||
|---|---|---|---|---|---|---|---|---|
| Low | Intermediate OR (95% CI) |
High OR (95% CI) |
p trend | Low | Intermediate OR (95% CI) |
High OR (95% CI) |
p trend | |
| Abdominal obesity | ||||||||
| Crude | 1 | 3.30 (2.32–4.70) | 11.66 (7.69–17.67) | <.001 | 1 | 3.52 (2.73–4.53) | 6.83 (2.01–23.17) | <.001 |
| Model 1 | 1 | 3.10 (2.04–4.71) | 8.93 (5.43–14.69) | <.001 | 1 | 2.24 (1.54–3.26) | 8.78 (4.00–19.30) | <.001 |
| Model 2 | 1 | 2.30 (1.75–3.03) | 5.25 (3.56–7.75) | <.001 | 1 | 2.02 (1.36–2.99) | 7.24 (3.03–17.30) | <.001 |
| High triglyceride | ||||||||
| Crude | 1 | 1.73 (1.29–2.33) | 3.02 (2.05–4.45) | <.001 | 1 | 1.90 (1.47–2.45) | 6.54 (1.78–24.04) | <.001 |
| Model 1 | 1 | 1.52 (1.08–2.13) | 1.80 (1.18–2.72) | 0.008 | 1 | 1.43 (1.09–1.87) | 5.79 (1.40–24.00) | 0.001 |
| Model 2 | 1 | 1.46 (1.15–1.86) | 1.56 (1.12–2.18) | 0.002 | 1 | 1.41 (1.07–1.86) | 4.77 (1.18–19.40) | 0.002 |
| High fasting glucose | ||||||||
| Crude | 1 | 1.73 (1.29–2.33) | 3.02 (2.05–4.45) | <.001 | 1 | 2.50 (1.96–3.19) | 3.73 (1.10–12.68) | <.001 |
| Model 1 | 1 | 1.31 (0.92–1.89) | 2.04 (1.31–3.19) | 0.001 | 1 | 1.56 (1.19–2.06) | 2.07 (0.67–6.42) | 0.001 |
| Model 2 | 1 | 1.41 (1.13–1.77) | 1.87 (1.29–2.73) | <.001 | 1 | 1.58 (1.18–2.12) | 0.60 (0.14–2.65) | 0.006 |
| Low HDL-cholesterol | ||||||||
| Crude | 1 | 1.90 (1.30–2.78) | 2.66 (1.72–4.12) | <.001 | 1 | 1.67 (1.28–2.17) | 5.08 (1.28–20.25) | <.001 |
| Model 1 | 1 | 1.59 (1.04–2.42) | 2.03 (1.26–3.28) | 0.005 | 1 | 1.19 (0.89–1.58) | 3.38 (0.94–12.19) | 0.085 |
| Model 2 | 1 | 1.08 (0.82–1.42) | 1.48 (1.01–2.17) | 0.073 | 1 | 1.19 (0.89–1.57) | 3.10 (0.89–10.81) | 0.088 |
| High blood pressure | ||||||||
| Crude | 1 | 4.96 (3.57–6.88) | 22.49 (14.15–35.76) | <.001 | 1 | 9.51 (7.03–12.85) | NA (NA–NA) | <.001 |
| Model 1 | 1 | 3.28 (2.22–4.83) | 15.32 (8.93–26.28) | <.001 | 1 | 5.73 (4.21–7.80) | NA (NA–NA) | <.001 |
| Model 2 | 1 | 4.92 (3.86–6.27) | 15.33 (9.87–23.82) | <.001 | 1 | 5.78 (4.21–7.92) | NA (NA–NA) | <.001 |
Model 1 adjusted for age and BMI.
Model 2 additionally adjusted for residence, smoking, alcohol consumption, household income, education, and the presence of diabetes.
The risk of having OSA was calculated using the STOP-Bang questionnaires.
BMI = body mass index, STOP-Bang = the snoring, tiredness, observed apnea, high BP, BMI, age, neck circumference, and male gender, HDL = high-density lipoprotein, OR = odds ratio, CI = confidence interval, SE = standard error.
In subgroup analyses according to menopausal status, the risk of OSA was associated with abdominal obesity and hypertension in premenopausal women (Table 5). In postmenopausal women, the risk of OSA was associated with abdominal obesity, hypertriglyceridemia, impaired fasting glucose, and hypertension, but a statistical trend was only observed with low HDL cholesterol levels (p trend = 0.099).
| Premenopausal women | Postmenopausal women | |||||||
|---|---|---|---|---|---|---|---|---|
| Low | Intermediate OR (95% CI) |
High OR (95% CI) |
p trend | Low | Intermediate OR (95% CI) |
High OR (95% CI) |
p trend | |
| Abdominal obesity | ||||||||
| Crude | 1 | 4.74 (3.02–7.44) | NA (NA–NA) | <.001 | 1 | 2.21 (1.81–2.71) | 2.76 (1.05–7.21) | <.001 |
| Model 1 | 1 | 2.78 (1.36–5.66) | NA (NA–NA) | <.001 | 1 | 1.44 (1.11–1.88) | 1.98 (0.71–5.54) | 0.004 |
| Model 2 | 1 | 2.72 (1.09–6.79) | NA (NA–NA) | 0.014 | 1 | 1.85 (1.20–2.86) | 4.31 (1.49–12.45) | 0.003 |
| High triglyceride | ||||||||
| Crude | 1 | 1.95 (1.25–3.07) | 1.06 (0.11–9.93) | 0.008 | 1 | 1.44 (1.16–1.78) | 3.12 (1.07–9.11) | <.001 |
| Model 1 | 1 | 1.28 (0.79–2.09) | 0.46 (0.04–4.87) | 0.526 | 1 | 1.22 (0.98–1.53) | 2.70 (0.93–7.82) | 0.024 |
| Model 2 | 1 | 1.54 (0.82–2.89) | 1.28 (0.12–13.92) | 0.202 | 1 | 1.42 (1.05–1.93) | 8.11 (1.56–42.23) | 0.003 |
| High fasting glucose | ||||||||
| Crude | 1 | 2.63 (1.76–3.93) | 5.84 (0.85–40.02) | <.001 | 1 | 1.74 (1.45–2.08) | 1.50 (0.55–4.12) | <.001 |
| Model 1 | 1 | 1.50 (0.92–2.43) | 1.74 (0.21–14.13) | 0.095 | 1 | 1.43 (1.17–1.75) | 1.24 (0.47–3.31) | 0.001 |
| Model 2 | 1 | 1.40 (0.76–2.59) | 0.59 (0.06–5.87) | 0.413 | 1 | 1.65 (1.19–2.28) | 0.45 (0.06–3.30) | 0.006 |
| Low HDL-cholesterol | ||||||||
| Crude | 1 | 1.89 (1.27–2.82) | 8.85 (0.93–84.06) | <.001 | 1 | 1.56 (1.26–1.93) | 0.85 (0.27–2.69) | 0.001 |
| Model 1 | 1 | 1.47 (0.92–2.35) | 4.53 (0.43–47.87) | 0.052 | 1 | 1.25 (1.00–1.57) | 0.67 (0.22–2.08) | 0.155 |
| Model 2 | 1 | 1.20 (0.63–2.26) | 2.76 (0.28–26.78) | 0.426 | 1 | 1.22 (0.88–1.68) | 3.42 (0.80–14.53) | 0.099 |
| High blood pressure | ||||||||
| Crude | 1 | 10.76 (7.03–16.47) | NA (NA–NA) | <.001 | 1 | 7.80 (5.88–10.34) | NA (NA–NA) | <.001 |
| Model 1 | 1 | 6.27 (3.99–9.87) | NA (NA–NA) | <.001 | 1 | 6.94 (5.16–9.34) | NA (NA–NA) | <.001 |
| Model 2 | 1 | 6.23 (3.44–11.28) | NA (NA–NA) | <.001 | 1 | 5.95 (4.05–8.75) | NA (NA–NA) | <.001 |
Model 1 adjusted for age and BMI.
Model 2 additionally adjusted for residence, smoking, alcohol consumption, household income, education, and the presence of diabetes.
The risk of having OSA was calculated using the STOP-Bang questionnaires.
BMI = body mass index, STOP-Bang = the snoring, tiredness, observed apnea, high BP, BMI, age, neck circumference, and male gender, HDL = high-density lipoprotein, OR = odds ratio, CI = confidence interval, SE = standard error.
This cross-sectional study analyzed the association of the risk of OSA with IR, MetS, and its components in individuals aged ≥40 years who completed easy-to-use STOP-Bang questionnaires. We observed that the risk of OSA, divided by the STOP-Bang score, was significantly associated with increased ORs of IR and MetS, and these associations were significant in both men and women after adjustment for potential covariates. In premenopausal women, the risk of OSA is associated with abdominal obesity and hypertension. However, in postmenopausal women the risk of OSA was significantly associated with abdominal obesity, hypertriglyceridemia, impaired fasting glucose, and hypertension, but not with low HDL-cholesterol levels. These results indicate that OSA is closely related to metabolic disturbances, such as IR and MetS, although further longitudinal prospective studies on this relationship may provide evidence to demonstrate this.
We found that OSA was associated with IR irrespective of sex and menopausal status. In evaluating the subgroup analysis, we hypothesized that sex and menopause status would be an important confounding factor because men, premenopausal, and postmenopausal women have different insulin sensitivity due to different amounts of visceral fat tissue and endogenous estrogen levels [26]. The association between OSA and IR has been widely investigated. In patients with OSA, the higher the AHI, the higher the IR and HOMA-IR [27]. Another recent study in the United States found that IR was more robustly associated with OSA risk than fasting glucose, which is similar to what was observed in the current study [28]. Short-term intermittent nocturnal hypoxia leads to increased IR [29], and chronic exposure to hypoxia causes a sustained increase in IR, indicating dysregulated glucose metabolism predisposing individuals to IR due to intermittent hypoxia [30]. AHI has been one of the most utilized tools for evaluating OSA in human studies investigating the impact of OSA on IR. However, despite the clinical usefulness of the STOP-Bang model, the association between stratified risk of OSA and IR has rarely been estimated in epidemiologic studies. The relationship between the STOP-Bang score and IR was indirectly studied in patients with type 2 diabetes, showing that nearly half had a STOP-Bang score ≥3 [15].
Given that IR and MetS share several clinical features, it is expected that OSA could similarly play a similar role in the development of MetS. The results of our analysis are in line with those of previous studies, providing additional evidence on the impact of OSA on MetS. For example, patients with OSA have a higher prevalence of MetS than individuals without OSA [31]. A meta-analysis on this relationship showed that numerous studies have consistently reported a positive association between OSA and MetS [32], but enrolled studies diagnosed OSA based only on AHI obtained from PSG, which is generally cost, facility, and time-consuming. The strength of our study is that it revealed the correlation between MetS and OSA using the STOP-Bang score, which can easily be used in a real clinical environment.
Considering that the current study was based on a cross-sectional design, it is unfeasible to determine the direction of the correlation between OSA and MetS. Growing evidence suggests a bidirectional relationship exists between OSA and metabolic abnormalities. For example, a previous cohort study in the United States showed that OSA increased the risk of type 2 diabetes, whereas insulin-treated diabetes could also be associated with the development of OSA [33]. Another recent study indicated that IR is associated with upper airway collapsibility, which may contribute to the development of OSA [34]. Therefore, the association between OSA and metabolic dysregulation seems to be bidirectional, and it remains controversial as to which comes first.
Although several previous studies have reported an independent association between OSA and MetS, it remains controversial which components of MetS are responsible for the relationship between the two conditions [35]. A previous cross-sectional study showed that hypertension, dyslipidemia, and impaired fasting glucose were associated with OSA [13]; however, another study found that abdominal obesity contributes to the increased prevalence of OSA [36]. In a study of 400 female patients, the OR for higher AHI was significant only for abdominal obesity, hypertriglyceridemia, and low HDL-cholesterol levels [37]. Different races, different definitions of MetS, study designs, and cut-off values for categorizing OSA severity might be related to the inconsistency in which components of MetS are attributable [35].
Notably, postmenopausal women showed a higher association between the risk of OSA and MetS than premenopausal women. In addition, while the adjustment for socioeconomic status, smoking, and alcohol consumption attenuated the association in premenopausal women, the association was strengthened in postmenopausal women. In this study, the risk of OSA was linked to all components of MetS in men and postmenopausal women but not in premenopausal women. Multivariate analyses revealed that OSA was not associated with high triglyceride levels, impaired fasting glucose levels, or low HDL cholesterol levels in premenopausal women. The hormonal effect could be related to the observed findings. One study demonstrated that postmenopausal women without hormonal replacement had a prevalence of OSA close to the prevalence in men [38]. Another study showed that estrogen protects against disturbed glucose metabolism [39]. After menopause, metabolic profiles become unfavorable with an increase in triglycerides and a decrease in HDL cholesterol levels [40].
Potential explanations underlying the relationship between OSA, IR, and MetS have been evaluated in animal and human experimental studies. In terms of IR, intermittent hypoxia and sustained oxidative stress could explain the observed findings. A mouse model showed that a 9 h exposure to hypoxia increased IR, causing counter-regulatory activation of glucocorticoid levels [29]. A human study similarly showed that 4 days of exposure to intermittent hypoxia leads to increased oxidative stress, which was measured using deoxyribonucleic acid (DNA), lipid, and protein oxidation [41]. In addition, recurrent arousal and sleep fragmentation may contribute to the development of IR. These events provoke transient cortical and sympathetic nervous system activation during sleep, although it is difficult to determine the individual effects of arousal on the autonomous nervous system. However, a recent human study of 48 healthy middle-aged volunteers found a close correlation between arousal events and muscle sympathetic nerve activity [42].
The data on the causal role of OSA in MetS are still controversial, and several studies have focused on the relationship between IR and MetS. However, IR can account for a further increase in MetS, and the expression of inflammatory markers in OSA has been postulated to link OSA and MetS. For example, in an in vitro model, HeLa cells exposed to intermittent hypoxia were responsible for the selective activation of the pro-inflammatory transcription nuclear factor kappa B (NF-kB), suggesting an association between OSA and cardiometabolic disorders [43]. Adiponectin is another biomarker closely related to the development of metabolic disorders in response to OSA. The serum adiponectin level is reduced in patients with OSA, and OSA could be a potential driver of decreased adiponectin level [44]. This reduction in adiponectin levels contributes to IR [45], and hypoadiponectinemia is closely associated with cardiovascular disease [46].
Our study had several limitations. First, because the analysis was based on data from a cross-sectional survey, causal relationships between OSA, IR, and MetS could be retrieved. As discussed above, a bidirectional relationship may exist. Second, we only used the STOP-Bang questionnaire to assess the risk of OSA. Although several studies have validated the usefulness of this model in predicting the probability and risk of OSA [18], the gold standard for diagnosing OSA is PSG. Third, because the KNHANES did not survey the use of continuous positive airway pressure treatment, the effect of this machine on metabolic abnormalities was not considered. Fourth, because the number of women with a high STOP-Bang score was too small, calculating the OR for abdominal obesity and high blood pressure between high and low STOP-Bang score was not feasible in the subgroup analysis by menopausal status. Fifth, it should be acknowledged that other unmeasured confounding factors, such as physical activity and diet patterns, could alter the observed relationship. Despite these limitations, our study has the strength of evaluating the association between OSA, IR, and MetS in the general population by defining OSA using an easy-to-use STOP-Bang model. Further longitudinal studies are needed to elucidate the causal effects of OSA on the development of IR and MetS.
In conclusion, by analyzing a large sample of Korean adults, we showed that OSA is closely associated with IR, MetS, and its components. The stratified probability of OSA was closely associated with IR and MetS, regardless of sex and menopausal status. Furthermore, we noticed differences in the individual components of MetS related to OSA between pre and postmenopausal women.
The study protocol was approved by the Kosin University Gospel Hospital Institutional Review Board (IRB, No. 2022-05-024). Written informed consent was obtained from all individuals. The study protocol and was conducted in accordance with the principles of the Declaration of Helsinki. All procedures were performed in accordance with the relevant guidelines and regulations.
Availability of data and materialsThe data of the KNHANES is opened to the public, therefore, any researcher can be obtained after request from the website https://knhanes.kdca.go.kr/knhanes/main.do.
Competing interestsThe authors declare no conflicts of interest related to this article.
FundingThis research was funded by a National Research Foundation of Korea (NRF) grant funded by the Korea government (MEST) (No. 2022R1C1C1005458).
Author’s contributionT Kim: Conceptualization, writing–original draft, data curation, formal analysis, investigation, methodology, project administration, supervision, formal analysis, methodology; J Kang: Funding acquisition, writing–review and editing. All authors discussed the results and approved the final version of the manuscript.
AcknowledgementsNone.
| Low | Intermediate | High | p trend | |
|---|---|---|---|---|
| Men | 1 | 1.76 (1.36–2.27) | 3.53 (2.40–5.20) | <.001 |
| Women | 1 | 1.89 (1.39–2.57) | 4.02 (0.55–29.20) | <.001 |
| Premenopause | 1 | 2.34 (1.23–4.46) | 1.75 (0.12–24.73) | 0.017 |
| Postmenopause | 1 | 1.81 (1.30–2.53) | 10.53 (1.08–102.71) | <.001 |
Multivariable model adjusted age, BMI, residence, smoking, alcohol consumption, household income, and education.
The risk of having OSA was calculated using the STOP-Bang questionnaires.
The IR was estimated using the HOMA-IR.
BMI = body mass index, STOP-Bang = the snoring, tiredness, observed apnea, high BP, BMI, age, neck circumference, and male gender, IR = insulin resistance, HOMA-IR = homeostatic model assessment for insulin resistance, OR = odds ratio, CI = confidence interval, SE = standard error.
| Men | ||||
| Low | Intermediate | High | p trend | |
| BMI (kg/m2) | ||||
| ≥25 (n = 1,483) | 1 | 2.34 (1.60–3.42) | 6.18 (3.72–10.26) | <.001 |
| <25 (n = 2,012) | 1 | 2.67 (1.90–3.75) | 4.00 (2.18–7.33) | <.001 |
| Neck circumference (cm) | ||||
| >40 (n = 636) | 1 | 4.27 (1.89–9.65) | 9.73 (3.84–24.66) | <.001 |
| ≤40 (n = 2,859) | 1 | 2.33 (1.80–3.01) | 3.79 (2.34–6.16) | <.001 |
| Women | ||||
| Low | Intermediate | High | p trend | |
| BMI (kg/m2) | ||||
| ≥25 (n = 1,526) | 1 | 3.06 (1.91–4.92) | 4.73 (0.68–33.03) | <.001 |
| <25 (n = 3,040) | 1 | 1.97 (1.28–3.02) | 2.24 (0.38–13.09) | 0.002 |
| Neck circumference (cm) | ||||
| >40 (n = 14) | 1 | NA (NA–NA) | NA (NA–NA) | NA |
| ≤40 (n = 4,552) | 1 | 2.56 (1.86–3.51) | 5.93 (1.52–23.16) | <.001 |
| Premenopausal women | ||||
| Low | Intermediate | High | p trend | |
| BMI (kg/m2) | ||||
| ≥25 (n = 443) | 1 | 3.47 (1.41–8.56) | 1.55 (0.12–20.30) | 0.028 |
| <25 (n = 1,085) | 1 | 2.49 (0.81–7.65) | NA (NA–NA) | 0.11 |
| Neck circumference (cm) | ||||
| >40 (n = 7) | 1 | NA (NA–NA) | NA (NA–NA) | NA |
| ≤40 (n = 1,521) | 1 | 2.55 (1.35–4.80) | 2.02 (0.18–23.19) | 0.009 |
| Postmenopausal women | ||||
| Low | Intermediate | High | p trend | |
| BMI (kg/m2) | ||||
| ≥25 (n = 1,082) | 1 | 3.24 (1.93–5.44) | NA (NA–NA) | <.001 |
| <25 (n = 1,952) | 1 | 1.98 (1.26–3.11) | 2.75 (0.43–17.70) | 0.002 |
| Neck circumference (cm) | ||||
| >40 (n = 7) | 1 | NA (NA–NA) | NA (NA–NA) | NA |
| ≤40 (n = 3,027) | 1 | 2.68 (1.89–3.79) | 9.09 (2.17–37.99) | <.001 |
Multivariable model adjusted age, (BMI), residence, smoking, alcohol consumption, household income, and education.
The risk of having OSA was calculated using the STOP-Bang questionnaires.
BMI = body mass index, STOP-Bang = the snoring, tiredness, observed apnea, BMI, age, neck circumference, and male gender, OR = odds ratio, CI = confidence interval, SE = standard error.
