INTRODUCTION
Obstructive sleep apnea (OSA) is characterized by the contraction of the upper airway muscles during sleep. The breath passes less than normal or cannot pass, arousing the brain to resume normal breathing, resulting in an inability to sleep properly. As a consequence, patients often suffer from excessive drowsiness during the day, irritability, attention deficit hyperactivity disorder, poor work performance, and increased risk of traffic accidents [1].
OSA is often associated with obesity [2]. Body mass index (BMI) is related to total sleep time, with obesity or overweight being associated with less sleep compared to non-obese individuals [3]. The prevalence of sleep apnea in individuals with obesity undergoing bariatric surgery is 77% [4]. Additionally, the prevalence of OSA in the severely obese (BMI 35–39.9 kg/m2), morbidly obese (BMI 40–40.9 kg/m2), and super-obese (BMI 50–59.9 kg/m2) was 71%, 74%, and 77 %, respectively [5].
The diagnosis of OSA was made on the basis of clinical symptoms and risks. Most patients snore and exhibit underlying conditions that put them at high risk for OSA, such as obesity, diabetes, hypertension, heart disease, or a family history of apnea, among others [6]. OSA can be assessed using questionnaires designed to detect sleep-disordered breathing, particularly the Epworth Sleepiness Scale (ESS), STOP-BANG, and Berlin Questionnaire (BQ) [7-9]. These questionnaires can be used to predict a high probability of OSA before performing polysomnography (PSG), which is a standard method for diagnosis. However, the three screening tools exhibit varying accuracy in detecting OSA in different settings [10].
Data on the accuracy of ESS, STOP-BANG, and BQ for predicting OSA among individuals with morbid obesity have been limited. The accuracy of questionnaires in detecting OSA in individuals with non-morbid obesity may differ from those with morbid obesity, depending on the cutoff values used for each questionnaire. We hypothesized that the accuracy of questionnaires varies between morbidly obese and non-obese populations. Therefore, we conducted this study to determine the optimal cutoff values for detecting OSA in individuals with morbid obesity.
METHODS
Study design and participants
A prospective cross-sectional study was conducted at Thammasat University Hospital in Thailand between October 2020 and December 2021. Subjects aged 18 years or older with morbid obesity and suspected of OSA were included. Subjects who were unable to communicate, those who were pregnant, and individuals with a previous diagnosis of OSA were excluded.
Demographic data, including age, sex, height, weight, BMI, neck circumference (NC), comorbidities, and responses to sleep screening questionnaires (ESS, STOP-BANG, and BQ), were recorded before PSG. All questionnaires were completed by the subjects themselves at the outpatient clinics. Subsequently, the subjects proceeded to a sleep lab. PSG data, including apnea-hypopnea index (AHI), lowest peripheral oxygen saturation (SpO2), 3% oxygen desaturation index (ODI), and periodic limb movements of sleep (PLMS) index, were also recorded.
Ethical approval was obtained from the Human Research Ethics Committee of Thammasat University Hospital, Thailand (IRB No. 012/2563, COA No.012/2563, Date of approval: 29 December 2020), in full compliance with international guidelines such as the Declaration of Helsinki, the Belmont Report, CIOMS Guidelines, and the International Conference on Harmonisation–Good Clinical Practice (ICH-GCP). All methods were performed in accordance with these guidelines and regulations. All participants provided written informed consent. This study was registered with Thaiclinicaltrials.org with the number TCTR20210331004.
Obesity and morbid obesity
The normal BMI range for Asians, including Thais, is 18.5–22.9 kg/m2 [11]. BMI values of 23.0–24.9 kg/m2 and ≥25 kg/m2 are classified as overweight and obesity, respectively.
In this study, morbid obesity was defined according to the guideline from the Consensus Statement from the Asia-Pacific Bariatric Surgeons Group in 2005: BMI ≥37 kg/m2 or obesity with a BMI ≥32 kg/m2 with diabetes or two obesity-related comorbidities, such as hypertension, hyperlipidemia, and heart disease [12].
Polysomnography
A type 1 PSG, whether a full-night or split-night test, is considered the gold standard for diagnosing OSA [6]. It involves recording physiological changes during sleep, including electroencephalogram, electrooculogram, airflow/nasal pressure, electrocardiogram, electromyogram, snore detector, body position, thoracic and abdominal measurements, oxygen saturation, and video monitoring [6,13].
Sleep and respiratory assessments were performed according to the American Academy of Sleep Medicine (AASM) guidelines [13]. Apnea was scored when there was a ≥90% reduction in airflow for ≥10 seconds. Hypopnea was scored if there was a ≥30% reduction in airflow for ≥10 seconds, associated with a ≥3% decrease in the oxygen saturation or an arousal. The severity of OSA was classified based on AHI values: AHI <5 indicated no OSA; AHI 5–14.99 indicated mild OSA; AHI 15–29.99 indicated moderate OSA; and AHI ≥30 indicated severe OSA.
Questionnaires
The ESS questionnaire includes 8 questions (with a total score of 24) in various situations to screen for daytime sleepiness, including 1) sitting and reading, 2) watching TV, 3) sitting in a public place such as a meeting or theater, 4) riding as a passenger in a car for an hour without a break, 5) lying down to rest in the afternoon if circumstances permit, 6) sitting and talking to someone, 7) sitting quietly after lunch without alcohol, and 8) sitting in a car stuck in traffic for a few minutes. A total ESS score >11 is classified as high risk for OSA [7].
The STOP-BANG questionnaire includes 8 questions (with a total score of 8), covering snoring, tiredness, observed apnea, hypertension, BMI, age, neck circumference, and gender. A total STOP-BANG score >3 is classified as high risk for OSA [8].
The BQ consists of three categories, including snoring, daytime sleepiness, and hypertension/BMI, with a total of 10 questions (resulting in a total score of 9). A positive BQ total score of ≥2 categories is classified as high-risk for OSA. The questionnaire was scored according to its designated scoring method [9].
Statistical analysis
In a previous study [14], STOP-BANG demonstrated 62% sensitivity and 63% specificity in detecting moderate to severe OSA among obese patients undergoing bariatric surgery. We hypothesized that our study would yield diagnostic accuracy similar to that of the previous study [14]. The sample size was calculated for sensitivity and specificity using 80% power, a 5% type I error, and a 10% precision. Thus, the calculated sample size was 189.
Descriptive statistics are presented as number (%) and mean±standard deviation. The chi-squared test was used to compare categorical variables between the non-moderate to severe OSA group and the moderate to severe OSA group. Student’s t-test was used to compare continuous variables between the two groups. Pearson correlation was used to determine the correlation between AHI and each questionnaire score. To determine the set of variables associated with AHI, we used a linear regression model with AHI designated as the dependent variable. All independent variables—BMI, NC, lowest SpO2, as well as ESS, STOP-BANG, and BQ scores—were entered into the regression model, followed by backward selection using a p-value cutoff of 0.1. We report the regression coefficients, their 95% confidence interval (CI), and corresponding p-values. Variables with a p-value <0.05 were considered statistically associated with AHI. Using the regression coefficients and the intercept, predicted AHI for a patient could be calculated from the following equation:
Predicted PIFR = Intercept + V1β1 + V2β2 + … + Viβi,
where V represents the covariate, β the regression coefficient, and i the number of variables.
The receiver operating characteristic (ROC) curve was used to determine the optimal cutoff value for each questionnaire score to predict moderate to severe OSA. The area under the ROC curve (AUC) was presented with the diagnostic ability to distinguish between two groups. A two-sided p-value <0.05 was considered statistically significant. Statistical analyses were performed using SPSS version 25.0 software (IBM Corp.).
RESULTS
Participants
A total of 676 subjects were screened consecutively, of whom 471 were excluded, resulting in the inclusion of 205 subjects in the study (Figure 1). The mean age was 43.2 (±14.3) years, with 40.5% of the subjects being male. The average BMI was 40.9±5.3 kg/m2. Common comorbidities included hypertension (68.3%), dyslipidemia (39.0%), and diabetes (27.8%). The mean ESS score was 8.5±5.1, the STOP-BANG score was 5.0±1.4, and the BQ score was 2.2±0.7 (Table 1). PSG data revealed an AHI of 59.4±39.2 events/h and the lowest SpO2 of 76.1%±14.7%. The distribution of mild, moderate, and severe OSA was 9.8%, 14.6%, and 73.2%, respectively (Table 2).
Diagnostic accuracy of three questionnaires
The best cutoff values of ESS, STOP-BANG, and BQ for detecting moderate to severe OSA were 6.5, 4.5, and 1.5, with sensitivity levels of 61.7%, 67.2%, and 88.9%, respectively, and specificity levels of 52.0%, 36.0%, and 36.0%, respectively (Table 3). The AUC for ESS, STOP-BANG, and BQ was 0.624, 0.656, and 0.649, respectively (Figure 2). STOP-BANG had the highest AUC (0.656, 95% CI: 0.551–0.750) (Figure 2 and Table 3).
Correlation between AHI and questionnaire scores
In patients with morbid obesity, AHI showed significant correlations with ESS (r=0.411), STOP-BANG (r=0.283), and BQ (r=0.373) (all p<0.001).
Linear regression analysis showed the correlation between AHI and questionnaire scores adjusted by BMI, NC, and lowest SpO2. The equation for predicting AHI, derived from the linear regression model, is as follows: Predicted AHI (events/h) = -12.041 + (12.636×Q scores) + (1.227×BMI) + (1.622×NC) - (0.991×lowest SpO2) (Figure 3).
DISCUSSION
This prospective cross-sectional study assesses the diagnostic accuracy of ESS, STOP-BANG, and BQ in predicting moderate to severe OSA in patients with morbid obesity. Our findings indicate that BQ exhibited the highest sensitivity, while ESS demonstrated the highest specificity for prediction. BQ score, BMI, NC, and lowest SpO2 were identified as independent factors associated with AHI. These factors were identified as independent variables influencing the optimal continuous positive airway pressure in Thai patients with OSA, as demonstrated by Saiphoklang et al. [15]. The prevalence of OSA in Thailand was reported as 15.4% of men and 6.3% of women in a study conducted by Neruntarat and Chantapant [16]. In the Thai population, the prevalence of OSA has increased from 13.0% in men and 23.2% in women with a BMI of 25 kg/m2 or more in 1991 to 22.4% and 34.3% in 2004, respectively [17].
Obesity is associated with fat deposition in the body, particularly around the neck, leading to an increased NC and greater severity of OSA [18,19]. Obese individuals with shorter sleep duration experience twice as many subjective sleep problems compared to non-obese individuals [20]. Inadequate sleep at night and daytime sleepiness predispose individuals to weight gain [21]. BMI is associated with the severity of sleep apnea and hypopnea. A study by Leppänen et al. [22] found that individuals with an average BMI of 28.8 kg/m2 had an average AHI of 8.4 events/h, classifying them as having mild OSA. Those with an average BMI of 30.1 kg/m2 had an average AHI of 19.8 events/h, categorizing them as having moderate OSA. Individuals with an average BMI of 33.3 kg/m2 had an average AHI of 44.9 events/h, indicating OSA. Obesity is associated with poor sleep quantity and quality; thus, weight reduction can ameliorate sleep problems [23,24].
Our study revealed that BQ exhibited the highest sensitivity at 88.9%, consistent with the findings of Amra et al. [25] in Persian patients and Zhang et al. [26] in patients with cerebrovascular disease with an average BMI of 25.1 kg/m2, where the highest sensitivities were 86.4% and 78.8%, respectively. ESS demonstrated the highest specificity at 52%, consistent with findings from the study by Miller et al. [27] in OSA patients, as well as with those from the study by Zhang et al. [26] in patients with cerebrovascular disease, where specificities were reported as 87% and 100%, respectively. The lower specificity of our ESS, compared to the two previous studies, may be attributed to a lower proportion of obese subjects without disease or with mild disease. This results in a lower true negative value and less significant differences in each questionnaire score for distinguishing moderate to severe OSA from other cases. Furthermore, the highest AUC values of STOP-BANG in our study suggest that this questionnaire is reliable and clinically useful for making diagnostic decisions to distinguish moderate to severe OSA from non-moderate to severe OSA.
In addition, our study showed different results compared to a retrospective study by Glazer et al. [14], which was conducted on 266 obese patients with an average BMI of 49.2 kg/m2 who underwent bariatric surgery. Their findings indicated that STOP-BANG had the highest sensitivity (62.6%), and BQ had the highest specificity (71.1%) for detecting moderate to severe OSA. In contrast, our study found that BQ exhibited the highest sensitivity (88.9%), while ESS demonstrated the highest specificity (52.0%) for prediction. This discrepancy in results might be attributed to differences in the studied populations. Our subjects had lower BMI (40.9 vs. 49.2 kg/m2), a higher AHI (59.4 vs. 23.0 events/h), and well as differing racial backgrounds (Asian vs. Caucasian). These parameters might influence the accuracy of the questionnaires in detecting OSA.
This study has certain limitations. Firstly, there was a lower proportion of obese subjects without disease or with mild disease, leading to a lower true negative value and less significant differences in each questionnaire score. Therefore, the AUC indicates a test that is just satisfactory. Secondly, the study was conducted in a single research center in Thailand; thus, the results might not be applicable to other ethnicities or countries. A larger multicenter study is needed to validate these questionnaires for predicting moderate to severe OSA, with the aim of prioritizing obese patients for undergoing PSG.
In conclusion, BQ has the highest sensitivity, while ESS exhibits the highest specificity for predicting moderate to severe OSA in patients with morbid obesity. These questionnaires could be applied as predictive tools with optimal cut-point levels for screening OSA in individuals with morbid obesity.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
