Comparison of six assessment tools to screen for obstructive sleep apnea in patients with hypertension

Abstract Background Obstructive sleep apnea (OSA) is often accompanied by other complications, especially hypertension. Hypothesis The purpose of this study is to compare the application value of six tools in the screening of OSA in patients with hypertension. Compared with other questionnaires, we hypothesized that Berlin performed better in screening hypertensive patients suspected of OSA. Methods In this study, we collected the basic data and polysomnography (PSG) data of patients diagnosed with hypertension who underwent PSG at the Sleep Medicine Center of the First Affiliated Hospital of Guangzhou Medical University from April 2012 to March 2021. The sensitivity, specificity, positive predictive value, negative predictive value, area under the curv (AUC) and diagnostic odds ratio (DOR) of the six screening tools were then calculated, and their correlation with the sleep apnea hypopnea index (AHI) analyzed. Results There were 303 males (303/398, 76.1%) out of 398 hypertension patients suspected of OSA. The area under the curve of the Berlin questionnaire's receiver operating characteristic (ROC) curve reached 0.753 (95%CI: 0.707–0.794). When the AHI was 5, 15 and 30 times/h as the cut‐off points, the sensitivity and negative predictive value of Berlin were the highest at 0.947 and 0.630, 0.970 and 0.851, and 0.988 and 0.957 respectively, while the specificity and positive predictive value of the Epworth Sleepiness Scale (ESS) were the highest at 0.696 and 0.729, 0.750 and 0.887, and 0.674 and 0.575 respectively. The DOR value of the Berlin questionnaire could reach 18.333 when the AHI cut‐off point was 30 times/h. Berlin had the largest rank correlation coefficient with AHI at 0.466. Conclusion The Berlin questionnaire can be considered a priority for the screening and stratifying of hypertensive patients suspected of OSA.

tifying OSA. 9 The STOP questionnaire is a self-report designed by anesthesiologists and sleep experts based on the Berlin questionnaire and literature reviews. 10 The STOP-Bang questionnaire has become a widely used tool in OSA testing. It was initially used to screen surgical patients, but later proved to be a high-quality method with good consistency in identifying the severity of OSA in patients. 9,11 The Berlin questionnaire was developed by a group of respiratory and primary care doctors in Germany in 1996 through consensus, and has become a qualitative diagnostic tool for OSA that is widely used in the world. 12 In the design of some of these scales, hypertension is considered a risk factor for OSA. It has also been reported that OSA can affect the fluctuation of nocturnal blood pressure. 4 At present, these screening tools are used to screen OSA, but the specificity, sensitivity, predictive value, receiver operating characteristic (ROC) curve analysis and diagnostic odds ratio (DOR) of each screening tool are different.
Thus, we compared the application value of these six screening tools in patients with hypertension suspected of OSA in order to find a more suitable scale. We hypothesize that the screening ability of a scale with a hypertension option will be relatively high in patients with hypertension.  2. No-Apnea 7 : The score is 0-9 points, including two variables: ① Neck circumference (NC) < 37.0 is 0 points, 37.0-39.9 is 1 point, 40.0-42.9 is 3 points, NC ≥43.0 is 6 points; ② Age < 35 is 0 points, 35-44 is 1 point, 45-54 is 2 points, ≥ 55 years old is 3 points. If the No-Apnea score is ≥3 points, it indicates that the patient is at high risk of OSA.
3. ESS 9 : Including 8 questions, the subjects are asked to evaluate the degree of dozing in a specific scenario during the day; 0 is no dozing and 1, 2, and 3 are light, moderate and severe dozing. The total score is 24 points. If the ESS score is ≥9 points, there is daytime sleepiness. 10 : Including 4 problems, namely snoring, fatigue, observed apnea and high blood pressure. Answer with "yes" or "no"; "yes" is 1 point, "no" is 0 points. If the score of the 4 questions is >2, it indicates that the patient is at high risk of OSA.

STOP
5. STOP-Bang 9,11 : On the basis of the STOP scale, add "bang", namely B [body mass index (BMI) > 35 kg/m 2 ], A (age > 50 years old), N (neck circumference > 40 cm), G (male). Answer with "yes" or "no"; "yes" is 1 point, "no" is 0 points. If the STOP-Bang score is ≥3 points, it indicates that the patient is at high risk of OSA.
6. Berlin 12 : There are 11 problems in 3 groups: ① severity of snoring; ② daytime sleepiness; ③ high blood pressure or obesity. Each group is evaluated as negative or positive after calculating the score. If two or more of the 3 groups are positive, the patient is considered to have a high risk of OSA (high-risk group). If only one or none of the 3 groups is positive, the patient is considered to have a low risk of apnea (low-risk group).

| Polysomnography (PSG)
PSG monitoring was mainly used to diagnose sleep disordered breathing.
Recording indicators include electroencephalogram, electrooculogram, mandibular electromyography, oral and nasal airflow and respiratory movement, electrocardiogram, blood oxygen saturation, snoring, limb movement, body position and other parameters. We used an Alice 5 polysomnograph made by the Philips Wellcome Company to record continuously and synchronously for at least 7 h. After automatic analysis, the original parameters were manually reviewed and corrected, and For the measurement data, a one-way analysis of variance test was used for normal distribution, a multi-group independent sample rank test (Kruskal-Wallis one-way ANOVA (k) (w) multiple comparison) was used for skewed distribution data, and a chi-square test or Fisher's exact probability method was used to count the data. The diagnostic results of the scales and PSG were calculated in the form of a four-grid table for the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LR+), and negative likelihood ratio (LR-) of each scale, and reported with respective 95% confidence intervals (CI). The ROC curve was analyzed using MedCalc software to evaluate the diagnostic value of the five scales for OSA, and calculate the p value for comparing the area under the ROC curve of single and multiple indicators.
The correlation between the six scales and AHI was analyzed using the bivariate correlation method. Pearson correlation analysis was used for the normal distribution data, and Spearman correlation analysis was used for the non-normal distribution data. p < .05 was defined as statistically significant. Note: p, p value; p1 is the comparison of the four groups together; p2 is the comparison between the normal group and the mild OSA group; p3 is the comparison between the normal group and the moderate OSA group; p4 is the comparison between the normal group and the severe OSA group; p5 is the comparison between mild OSA group and moderate OSA group; p6 is the comparison between the mild OSA group and the severe OSA group; p7 is the comparison between the moderate OSA group and the severe OSA group.

| Predictive value of six scales
The areas under the curve of the ROC of the six scales were compared with AHI of 5, 10, 15, 20, 25, and 30 times/h as the cut-off points 5 (Figure 1)

| Correlation analysis between six scales and AHI
According to the correlation analysis of the six scales and AHI, it is clear that because the scores of the six scales were non-normally dis- Among them, the p value of no-apnea was .397 with no statistically significant difference, and the p values of the other five scales were less than .001 with statistically significant differences (Table 3).

| DISCUSSION
OSA is the most common sleep disordered breathing disease, as well as a disease with a high incidence and low diagnosis rate. According to the relevant research reports, the prevalence of OSA is rising rapidly, which may be due to the combined effect of various factors such as the obesity epidemic, increased sensitivity of sleep study recording sensors and reduced stringency after updated scoring standards, 12 which is also related to the various high-risk factors of the disease. In our study, 322 out of 398 hypertensive patients suspected of OSA were diagnosed. The proportion of males was much higher than that of females, which is consistent with the results of previous epidemiological studies. 3,14 It has been reported in the literature that the patient's neck circumference, waist circumference and BMI value affect the severity of OSA, 15 which is also consistent with our research results. However, our research shows that the differences between OSA and smoking, drinking and age were not statistically significant. In addition, related studies have shown that neck circumference, waist circumference and BMI are also risk factors for hypertension. 16 It can be seen that high-risk factors such as gender, neck circumference, waist circumference and BMI value will affect the identification and severity distribution of OSA patients with hypertension on the screening form.
The purpose of this study was to compare the application value of six screening tools in the screening of OSA in patients with hypertension.  The ideal screening tool should have high sensitivity and specificity at the same cut-off value, but this is a very rare case. However, the sensitivity and specificity of the screening model are usually inversely correlated, and high sensitivity is often gained at the expense of specificity. 7,11 For diseases such as OSA combined with hypertension, it may be more important that the screening test has high sensitivity so as not to miss OSA patients rather than having high specificity. 17 The Berlin questionnaire classifies patients as high-risk or low-risk based on self-reports of snoring, daytime sleepiness, hypertension and obesity, with different sensitivity and specificity in different studies. 18  breathing, but also in mild cases. 24 In our study, NoSAS performed poorly, and was worse than STOP, STOP-Bang and Berlin in identifying hypertensive patients suspected of OSA. Therefore, as a new screening tool, NoSAS requires further verification.
In our study, in order to further compare the accuracy of the application value of the six screening tools, we conducted an analysis of the diagnostic odds ratio and its correlation with AHI. The DOR represents the best single-point estimate of the ROC curve, which importantly has nothing to do with epidemics but provides a decision-making tool for doctors to distinguish between healthy and unhealthy patients. 13 The greater the DOR, the better the accuracy of the diagnosis. 13

| Limitations
First, this was a retrospective single-center study. Generally speaking, the use of retrospective analysis to verify the predictive value of different screening tools is not as ideal as prospective research. However, the indicators that needed to be used in the research were included in the questionnaire survey of every patient undergoing PSG examination at our center, which basically solves the limitations of retrospective analysis. Second, the sample size was relatively small, predominantly male and aged around 50, which may have restricted our research results, and further study is required to evaluate effectiveness in younger people. Third, in general, patients who come to the sleep laboratory due to main complaints such as snoring, sleepiness, hypertension or apnea are usually suspected of OSA, and patients with hypertension may obtain higher results in the selected scale compared with non-hypertensive individuals, which virtually increases the probability of diagnosis. Fourth, this study may be affected by regional distribution (Asian population), so it needs to be further verified in other environments.

| CONCLUSIONS
In conclusion, for screening hypertensive patients suspected of OSA, although there are many test contents in Berlin, relatively speaking, it is still better than the other five screening tools, and can be considered a priority for the screening and stratification of hypertensive patients suspected of OSA. The screening of STOP-Bang is better than that of STOP, NoSAS and ESS, and it can also be used to screen for hypertensive patients suspected of OSA. NoSAS is simple to use but requires further certification as a new screening tool. Finally, ESS is not effective and no-apnea may not be suitable in screening for hypertensive patients suspected of OSA.

ACKNOWLEDGMENTS
We also would like to thank Xiaoxia Yao from No.