Abstract
Study Objectives:
Obstructive sleep apnea (OSA) is a multilevel problematic disease. Major septal deviation (SD) can lead to severe nasal congestion, which, in turn, can lead to sleep apnea. Although SD seems to be related to OSA, very few studies have quantitatively examined this relationship. In this study, we investigate this using a 9-year large-scale cohort study.
Methods:
The SD group was selected out of 1 million individuals randomly extracted by the National Health Insurance Service. The non-SD group was obtained through propensity score matching considering several variables. The primary end point was OSA diagnosis.
Results:
The study (SD) group included 11,238 individuals and the non-SD group (control group) included 22,476 persons. The overall hazard ratio for OSA in the SD group was 4.39 (95% confidence interval [CI]: 3.56–5.42). In subgroup analysis, the hazard ratio for OSA of male individuals was 3.77 (95% CI: 2.83–5.03), high economic status was 1.27 (95% CI: 1.05–1.56), metropolitan area was 1.31 (95% CI: 1.07–1.62), young age was 0.79 (95% CI: 0.64–0.98), hypertension was 1.00 (95% CI: 0.37–2.7), and diabetes mellitus was 2.44 (95% CI: 1.15–5.21). In the SD group, the hazard ratio for OSA after septoplasty was 0.71 (95% CI: 0.54–0.94).
Conclusions:
From long-term follow-up, the prevalence of OSA was 4.39 times higher in the SD group compared with the control group. This phenomenon was more pronounced with increasing body mass index and decreased significantly after septoplasty.
Citation:
Yeom SW, Chung SK, Lee EJ, et al. Association between septal deviation and OSA diagnoses: a nationwide 9-year follow-up cohort study. J Clin Sleep Med. 2021;17(10):2099–2106.
Keywords: septal deviation, obstructive sleep apnea, national health insurance service, cohort study
BRIEF SUMMARY
Current Knowledge/Study Rationale: Although septal deviation seems to be related to obstructive sleep apnea, very few studies have examined this relationship quantitatively.
Study Impact: Our big-data study with long-term follow-up has shown that the prevalence of obstructive sleep apnea in patients with septal deviation was 4.39 times higher than in patients with no septal deviation. This phenomenon was more pronounced with increasing body mass index and decreased significantly after septoplasty.
INTRODUCTION
Obstructive sleep apnea (OSA) is a multilevel problematic disease.1 There is general agreement that the most successful method of controlling OSA is continuous positive airway pressure (CPAP) therapy. However, since CPAP therapy is known to be associated with low compliance, multilevel surgery has emerged as a means to solve the multilevel problem.2 As the nose is the first and major route through which air passes on its way to the lungs, nasal surgery is the basis of this multilevel surgery. In patients with low CPAP adherence and nasal obstruction, nasal surgery is known to increase CPAP compliance and adherence.3
Major septal deviation (SD) can lead to severe nasal congestion, which can lead to sleep apnea.4 It is rare to report that nasal surgery alone objectively improves sleep apnea, and studies have described self-reported improvement in symptoms of OSA.5,6
Taken together, SD seems to be associated with the increased incidence of OSA. So, to what extent is septal deviation really related to OSA? Our research began with this question and aimed to investigate it using 9 years of big-data cohort research. We hypothesized that SD has a quantitative relationship with OSA, and that quantitatively investigated demographic factors, obesity, and septoplasty influence this relationship.
METHODS
Database and type of study
The National Health Insurance Service–National Sample Cohort consists of about 1 million people, 2% of the total 50 million people in Korea, whose data on age, sex, residential area, and economic status were extracted randomly from the database. This retrospective cohort study contains basic demographic variables, such as the patient’s age and sex, and complex information such as the patient’s visit date, diagnosis code, treatment history, medications, and insurance claims.
Study population with SD: SD group
The definition of the experimental cohort was SD for the preceding disease with the diagnostic code J342 (based on the International Classification of Diseases, 10th revision [ICD-10], diagnosis codes) diagnosed between 2002 and 2004. The definition of target disease for OSA is ICD code G473. There were 3 exclusion criteria: (1) patients diagnosed with SD during the period 2005–2013, (2) patients with records for both SD and OSA in whom OSA was diagnosed earlier or at the same time as the SD diagnosis, and (3) patients under 20 years of age.
The starting point for the SD group was the date of SD diagnosis, and the end point was set as the date of first diagnosis of OSA if there was a record of OSA diagnosis or December 31, 2013, if there was no record of OSA diagnosis.
Control (non-SD) group
From the SD cohort defined above, the control group was calculated by 1:2 propensity score (PS) matching considering 7 independent variables. We performed PS matching using a “greedy nearest neighbor” algorithm with a 1:1 ratio. Confirmation of successful PS matching was or was not judged by the presence or absence of major imbalances in the figure or standardized mean differences in each group.
All independent variables were divided into 2 categories and were as follows: (1) Male or female; (2) old or young separated by age 50; (3) metropolitan city or country; (4) from the 10th quintile for economic status, the upper 30% is high economic status and the rest are low economic status; (5) history/no history of hypertension; (6) history/no history of diabetes mellitus; (7) history/no history of chronic kidney disease. The definition of history (categories 5–7) is given in Table S1 (149.5KB, pdf) in the supplemental material.
The starting point for the control group was the date of first visit to hospital between 2002 and 2004 and the end point was set as the date of first diagnosis of OSA if there was a record of OSA diagnosis or December 31, 2013, if there was no record of OSA diagnosis.
Outcome variables and statistical analysis
Data analysis was conducted between June and August 2020. Two data analysts (J.S.K., S.W.Y.) independently conducted the data analysis, and the 2 conferred on any discrepancy. The hazard ratio (HR) in the Cox proportional-hazards model was calculated by taking the difference between the end point and start point. When considering the relationship between 1 independent variable and the dependent variable, the adjusted HR was obtained by considering all other variables, and the unadjusted HR was calculated without considering other variables. The relative risk or risk ratio was calculated as the ratio of the probability of an outcome in the SD group to the probability of an outcome in the non-SD group. The cumulative hazard ratio was obtained through survival analysis, and R 3.5.3 statistical program (R Foundation for Statistical Computing, Vienna, Austria) was used to analyze the results.
Subgroup analysis
In subgroup analysis of the SD group, we examined sex, economic status, age, residential area, and underlying diseases. We also evaluated body mass index (BMI) and septoplasty as moderator variables. BMI classification followed World Health Organization standards.7
RESULTS
Of the 1 million cohort samples, a total of 33,714 individuals (11,238 in the SD group and 22,476 in the non-SD group) were followed up for a total of 9 years from January 2005 to December 2013.
Validation of the PS matching
Table 1 shows the chi-square test results obtained for PS matching between the experimental group (SD) and the control group (non-SD) for a total of 7 independent variables including SD. All P values exceeded .05 and were close to 1, so there was no significant difference between the 2 groups. From Figure 1, the distributions of the SD and non-SD groups are similar to each other, so PS matching was well executed.
Table 1.
Demography of SD and non-SD groups.
| Variable | Control (Non-SD) Group (n = 22,476) | Study (SD) Group (n = 11,238) | Chi-Square χ2 | P |
|---|---|---|---|---|
| Sex | .0007 | .9778 | ||
| Male | 14,027 | 7,011 | ||
| Female | 8,449 | 4,227 | ||
| Age | .0021 | .9627 | ||
| Young (< 50 y) | 16,021 | 8,014 | ||
| Old (> 50 y) | 6,455 | 3,224 | ||
| Residential area | .0004 | .9834 | ||
| Metropolitan | 6,994 | 3,495 | ||
| Rural | 15,482 | 7,743 | ||
| Economic status | 0 | .9969 | ||
| Low (< 70%) | 9,628 | 4,815 | ||
| High (> 70%) | 12,848 | 6,423 | ||
| HTN | .0006 | .9795 | ||
| Yes | 171 | 99 | ||
| No | 22,305 | 11,139 | ||
| DM | .96085 | .327 | ||
| Yes | 128 | 63 | ||
| No | 11,175 | 22,348 | ||
| CKD | .9609 | .327 | ||
| Yes | 12 | 10 | ||
| No | 22,464 | 11,228 |
CKD = chronic kidney disease, DM = diabetes mellitus, HTN = hypertension, SD = septal deviation.
Figure 1. Validation of propensity score matching.
DM = diabetes mellitus, HTN = hypertension, SD = septal deviation.
Incidence rate and HR for OSA
Table 2 shows the relationship between the 7 independent variables for OSA and 3 statistics: 10,000 person years (incidence rate), unadjusted HR, and adjusted HR. The incidence rate (incidence per 10,000 person years) is a measure of the frequency with which a disease occurs over a specified time period.8,9 Figure 2 and Table 2 show that the HR for OSA in the SD group was 4.39 (95% confidence interval [CI]: 3.56–5.42), ie, the prevalence of OSA was 4.39 times higher in the SD group compared with the control group.
Table 2.
Incidence rate and adjusted and unadjusted HRs of each group.
| Variable | Study Group Total | Number of Cases | Incidence per 10,000-Person Years | HR Adjusted (95% CI) | HR Unadjusted (95% CI) |
|---|---|---|---|---|---|
| Group | |||||
| Non-SD | 22,476 | 131 | 5.12 | 1 | 1 |
| SD | 11,238 | 262 | 22.41 | 4.39 (3.56–5.42) | 4.39 (3.55–5.42) |
| Sex | |||||
| Female | 12,676 | 54 | 3.81 | 1 | 1 |
| Male | 21,038 | 339 | 14.69 | 3.77 (2.83–5.03) | 3.85 (2.89–5.13) |
| Economic status | |||||
| Low (< 70%) | 18,391 | 180 | 8.86 | 1 | 1 |
| High (> 70%) | 15,323 | 213 | 12.56 | 1.27 (1.05–1.56) | 1.42 (1.16–1.73) |
| Residential area | |||||
| Rural | 23,225 | 247 | 9.61 | 1 | 1 |
| Metropolitan | 10,489 | 146 | 12.62 | 1.31 (1.07–1.62) | 1.30 (1.06–1.59) |
| Age | |||||
| Old (> 50 y) | 9,679 | 140 | 13.01 | 1 | |
| Young (< 50 y) | 24,035 | 253 | 9.54 | 0.79 (0.64–0.98) | 0.73 (0.60–0.9) |
| HTN | |||||
| No | 33,444 | 389 | 10.52 | 1 | 1 |
| Yes | 270 | 4 | 13.24 | 1.00 (0.37–2.7) | 1.26 (0.47–3.38) |
| DM | |||||
| No | 33,523 | 386 | 10.41 | 1 | 1 |
| Yes | 191 | 7 | 33.10 | 2.44 (1.15–5.21) | 3.19 (1.51–6.74) |
| CKD | |||||
| No | 33,692 | 393 | 10.55 | ||
| Yes | 22 | 0 | 0 |
CI = confidence interval, CKD = chronic kidney disease, DM = diabetes mellitus, HR = hazard ratio, HTN = hypertension, SD = septal deviation.
Figure 2. Overall cumulative hazard ratio for obstructive sleep apnea in the SD group.
SD = septal deviation.
Subgroup analysis
In the subgroup for demographic factors, the HR was 3.77 (95% CI: 2.83–5.03) in the subgroup of male patients compared with female patients, 1.27 (95% CI: 1.05–1.56) in the subgroup of patients with high economic status compared with those with low economic status, 1.31 (95% CI: 1.07–1.62) in the subgroup of patients living in cities compared with those living in rural areas, and 2.44 (95% CI: 1.15–5.21) for the subgroup of patients with diabetes mellitus compared with those with no diabetes mellitus. In addition, HR was relatively low at 0.79 (95% CI: 0.64–0.98) in the subgroup of young people compared with old people (Figure 3).
Figure 3. Cumulative hazard ratio plot for subgroups: sex, economic status, age, and residential area.
Figure 4 is a forest plot that shows the HRs of each variable. In the subgroups excluding hypertension, there was a statistically significant difference in HRs between the subgroups.
Figure 4. Forest plot of cumulative hazard ratio for each factor: SD, sex, economic status, age, residential area, and underlying disease (DM, HTN).
DM = diabetes mellitus, HTN = hypertension, RA = residential area, SD = septal deviation, SES = socioeconomic status.
In the subgroup for BMI, the relative risk value of OSA in the SD group was much higher than that of OSA in the non-SD group, and this finding showed a sharper contrast as the BMI increased (Table 3).
Table 3.
Relationship between BMI and relative risk of OSA with and without a history of SD.
| Classification | BMI (kg/m2) | Relative Risk of OSA without History of SD | Relative Risk of OSA with History of SD |
|---|---|---|---|
| Underweight | < 18.5 | 0.55 (0.44–0.7) | 8.99 (0.73–111.06) |
| Normal | 18.5–24.9 | 1.00 (0.79–1.28) | 16.38 (1.33–202.27) |
| Preobese | 25.0–29.9 | 2.00 (1.58–2.54) | 32.65 (2.64–403.04) |
| Class I obesity | 30.0–34.9 | 3.65 (2.88–4.63) | 59.46 (4.82–734.03) |
| Class II obesity | 35.0–39.9 | 6.64 (5.24–8.43) | 108.28 (8.77–1336.83) |
| Class III obesity | ≥ 40.0 | 12.11 (9.55–15.36) | 197.2 (15.97–2434.66) |
BMI = body mass index, OSA = obstructive sleep apnea, SD = septal deviation.
In the subgroup for septoplasty, we selected 1841 patients who received septoplasty and 3,682 patients in a 1:2 PS-matched cohort who did not receive septoplasty among the 11,238 patients in the SD group from 2002 to 2004 (Table 4). All standardized mean differences were below .05, which means there were no major imbalances in each group. As a result of follow-up until 2013, in the SD group, the HR for OSA after septoplasty was 0.71 (95% CI: 0.54–0.94) (Figure 5A), and the cumulative incidence rates for those with septoplasty and those with no septoplasty are shown in Figure 5B.
Table 4.
For 1 million patients with a history of SD between 2002 and 2004, PS matching for cohort receiving septoplasty and cohort not receiving septoplasty among the SD group.
| Entire Cohort (n = 11,238) | PS-Matched Cohort | |||
|---|---|---|---|---|
| With Septoplasty Surgery (n = 1,841) | Without Septoplasty Surgery (n = 3,682) | SMD | ||
| Sex = male, n (%) | 7,011 (62.4) | 1,418 (77.0) | 2,827 (76.8) | 0.006 |
| Age = youth, n (%) | 8,014 (71.3) | 1,412 (76.7) | 2,858 (77.6) | 0.022 |
| RA = metro, n (%) | 3,495 (31.1) | 584 (31.7) | 1,199 (32.6) | 0.018 |
| OSA = yes, n (%) | 162 (1.4) | 23 (1.2) | 64 (1.7) | |
OSA = obstructive sleep apnea, PS = propensity score, RA = residential area, SD = septal deviation, SMD = standardized mean difference.
Figure 5. HRs and incidence rates for septoplasty.
(A) Forest plot of cumulative HRs for septoplasty in septal deviation group and other factors: sex, residential area, age. (B) Cumulative incidence rates for those with septoplasty and those with no septoplasty. C.I. = confidence interval, HR = hazard ratio, RA = residential area.
DISCUSSION
The prevalence of OSA was significantly higher in the SD group compared with the control group. This result was more pronounced as the BMI increased and decreased significantly after septoplasty.
The nose is the first and major route that air passes through on its way to the lungs. Laminar airflow is necessary in both nostrils to benefit from wearing a positive pressure device. Nasal congestion is a factor that adversely affects air passage, and surgery to improve nasal congestion (for example, septoplasty) lowers nasal resistance and increases CPAP adherence.10,11 Nakata et al12 reported that reducing nasal resistance directly improves quality of life and helps relieve daytime sleepiness. Kim et al13 reported that nasal surgery reduced the respiratory disturbance index and the apnea-hypopnea index (AHI) by 20%, respectively. Through a meta-analysis of 102 patients in 9 studies, Verse and Pirsig14 reported that there was a success rate of about 20% in OSA alone. In a meta-analysis consisting of 13 papers from 1999 to 2009, Li et al15 also found that the AHI value improved from 35.2 ± 22.6 to 33.5 ± 23.8 per hour after nasal surgery (P = .69). They reported that there was an impressive outcome in the visual analog scale from a questionnaire survey compared with the improvement in the AHI.
In this study, the prevalence of OSA was 4.39 times higher in the SD group compared with the control group (Figure 2). In the SD group, those who received septoplasty were 29% less likely to be diagnosed with OSA in the observed time period (Figure 5A, Figure 5B). In subgroup analysis, the HR for OSA was lower in younger patients with SD compared with older patients (HR: 0.79 [0.64–0.98]) (Figure 3). At a young age, airway resistance is relatively low, so the rate of OSA is lower.16,17
Other factors increasing the prevalence of OSA include male sex (HR: 3.77 [2.83–5.03)], living in large cities (HR: 1.31 [1.07–1.62]), and higher socioeconomic status (HR: 1.27 [1.05–1.56]). That OSA is more common in men than in women is a consistently proven effect in several studies.18
It is worth noting that the higher the socioeconomic status, the higher the prevalence of OSA diagnosed in the SD patient group. In this study, in terms of income, the top 30% were considered to have a high socioeconomic status, and the prevalence of OSA was 1.27 (1.05–1.56) times higher in this subgroup compared with the bottom 70%. In Korean society, since 2002, the income of the top 10% has risen sharply from 37.8% of the total population to 43.3% in 2016.19 This income disparity leads to inequality in the opportunity to go to a hospital. Therefore, patients with symptoms of SD and with lower economic status (the bottom 70% in terms of income) would be less likely to go to the hospital for treatment of OSA compared with high income patients, as this complaint is not covered by insurance funds. It is thought that this is the reason why OSA diagnosis was higher in the SD subgroup with higher income.
The greater prevalence of OSA in metropolitan areas can be interpreted in a similar way. The population density of Korea is high with 26,316 persons/km2 in the case of Seoul, while the area with the lowest population density (Kangwon Province) has 19 persons/km2, so there is a large gap between regions in terms of population density. In the case of large cities such as Seoul, medical facilities are also densely concentrated, and in the case of rural areas, there is a much lower density of medical facilities.20 Therefore, it is judged that the number of times each individual visits a clinic dealing with sleep disorder to receive medical treatment, especially for diseases such as OSA, decreases in rural areas, and this has resulted in a difference in the incidence of OSA between these regions.
PS matching has the advantage of being able to analyze data from larger populations compared with randomized controlled trials, and to deal more effectively with regions that are more clinically unconstrained thus reducing bias. However, studies using PS matching of real world registry data also have disadvantages in that they often rely on administrative datasets and they treat a broader, less selective patient population.21 Our study has shown that the distributions of the SD and non-SD groups were very similar so PS matching was well executed, thus the results of our study are reliable (Figure 1). Our study also has the advantage in that PS matching was performed on a wide range of variables: age, sex, residential area, income level, and underlying diseases such as hypertension, diabetes, and kidney failure.
This study has the limitation that several inherent biases may be present, as follows: (1) patients with a diagnosis of SD would more likely have gone to otolaryngologists for breathing complaints and therefore were more likely to have been offered a sleep study and (2) the study has inherent limitations because of its retrospective nature.
Despite these limitations, this study has the following intrinsic strengths: (1) it was a large population-based study using detailed national health records; (2) PS matching was conducted to compensate for the shortcomings of its retrospective nature and to regularize comparisons as much as possible; and (3) it evaluated the relationship between SD and OSA, with subgroup analysis of patients receiving SD treatment (septoplasty).
In conclusion, our study with long-term follow-up has shown that the prevalence of OSA in patients with septal deviation was 4.39 times higher than in patients with no septal deviation. This phenomenon was more pronounced with increasing BMI and decreased significantly after septoplasty. Clinicians should pay close attention to detailed sleep-disordered breathing histories, if available, and offer diagnostic sleep testing, if clinically indicated, to patients presenting with septal deviation.
DISCLOSURE STATEMENT
All authors have seen and approved the final manuscript. This study was funded by the Biomedical Research Institute at Jeonbuk National University Hospital. The authors report no conflicts of interest.
ACKNOWLEDGMENTS
This study used National Health Insurance Service–National Sample Cohort data (REQ000042830-001), made by the National Health Insurance Service. Author contributions: S.W.Y., S.K.C., E.J.L., M.G.K., D.H.K., M.H.L., Y.N.Y., C.M.L., J.S.K. contributed to the study design, protocol and study materials. S.W.Y., S.K.C., E.J.L., M.G.K., D.H.K., J.S.K. collected study data. S.W.Y., S.K.C., E.J.L., M.G.K., D.H.K., J.S.K., S.J.N. designed the statistical plan and data analysis. S.W.Y., J.S.K. performed the statistical analysis. S.W.Y., S.K.C., E.J.L., M.G.K., D.H.K., J.S.K. wrote the first draft of the manuscript. All authors contributed to interpretation of the data and revision of the manuscript.
ABBREVIATIONS
- BMI,
body mass index
- CI,
confidence interval
- CPAP,
continuous positive airway pressure
- HR,
hazard ratio
- OSA,
obstructive sleep apnea
- PS,
propensity score
- SD,
septal deviation
REFERENCES
- 1. Emara TA , Khazbak AO , Mohammed O , Elgaml M , Zidan A , Hosny SM . Changes in serum leptin level after multilevel surgery in patients with obstructive sleep apnea . Laryngoscope. 2020. ; 131 : E665 – E670 . [DOI] [PubMed] [Google Scholar]
- 2. Riley RW , Powell NB , Guilleminault C . Inferior sagittal osteotomy of the mandible with hyoid myotomy-suspension: a new procedure for obstructive sleep apnea . Otolaryngol Head Neck Surg. 1986. ; 94 ( 5 ): 589 – 593 . [DOI] [PubMed] [Google Scholar]
- 3. Georgalas C . The role of the nose in snoring and obstructive sleep apnoea: an update . Eur Arch Otorhinolaryngol. 2011. ; 268 ( 9 ): 1365 – 1373 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Olsen KD , Kern EB , Westbrook PR . Sleep and breathing disturbance secondary to nasal obstruction . Otolaryngol Head Neck Surg. 1981. ; 89 ( 5 ): 804 – 810 . [DOI] [PubMed] [Google Scholar]
- 5. Poirier J , George C , Rotenberg B . The effect of nasal surgery on nasal continuous positive airway pressure compliance . Laryngoscope. 2014. ; 124 ( 1 ): 317 – 319 . [DOI] [PubMed] [Google Scholar]
- 6. Nakata S , Noda A , Yagi H , et al . Nasal resistance for determinant factor of nasal surgery in CPAP failure patients with obstructive sleep apnea syndrome . Rhinology. 2005. ; 43 ( 4 ): 296 – 299 . [PubMed] [Google Scholar]
- 7. Weir CB , Jan A . BMI Classification Percentile and Cut Off Points. Jun 29, 2021. Treasure Island, FL: : StatPearls Publishing; ; 2021. . https://www.ncbi.nlm.nih.gov/books/NBK541070/. Accessed June 11, 2021. [PubMed] [Google Scholar]
- 8. Hargrave M . What does the incidence rate measure? Investopedia. Retrieved 2019-09-30. https://www.investopedia.com/terms/i/incidence-rate.asp. Accessed June 11, 2021.
- 9. Monson RR . Occupational Epidemiology. 2nd ed. Boca Raton, FL: : CRC Press; ; 1990. : 27 . ISBN 978-0-8493-4927-0. [Google Scholar]
- 10. Wakayama T , Suzuki M , Tanuma T . Effect of nasal obstruction on continuous positive airway pressure treatment: Computational fluid dynamics analyses . PLoS One. 2016. ; 11 ( 3 ): e0150951 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Kempfle JS , BuSaba NY , Dobrowski JM , Westover MB , Bianchi MT . A cost-effectiveness analysis of nasal surgery to increase continuous positive airway pressure adherence in sleep apnea patients with nasal obstruction . Laryngoscope. 2017. ; 127 ( 4 ): 977 – 983 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Nakata S , Noda A , Yasuma F , et al . Effects of nasal surgery on sleep quality in obstructive sleep apnea syndrome with nasal obstruction . Am J Rhinol. 2008. ; 22 ( 1 ): 59 – 63 . [DOI] [PubMed] [Google Scholar]
- 13. Kim ST , Choi JH , Jeon HG , Cha HE , Kim DY , Chung YS . Polysomnographic effects of nasal surgery for snoring and obstructive sleep apnea . Acta Otolaryngol. 2004. ; 124 ( 3 ): 297 – 300 . [DOI] [PubMed] [Google Scholar]
- 14. Verse T , Pirsig W . Impact of impaired nasal breathing on sleep-disordered breathing . Sleep Breath. 2003. ; 7 ( 2 ): 63 – 76 . [DOI] [PubMed] [Google Scholar]
- 15. Li HY , Wang PC , Chen YP , Lee LA , Fang TJ , Lin HC . Critical appraisal and meta-analysis of nasal surgery for obstructive sleep apnea . Am J Rhinol Allergy. 2011. ; 25 ( 1 ): 45 – 49 . [DOI] [PubMed] [Google Scholar]
- 16. Gao F , Li YR , Xu W , et al . Upper airway morphological changes in obstructive sleep apnoea: effect of age on pharyngeal anatomy . J Laryngol Otol. 2020. ; 134 ( 4 ): 354 – 361 . [DOI] [PubMed] [Google Scholar]
- 17. Le TB , Moghaddam MG , Woodson BT , Garcia GJM . Airflow limitation in a collapsible model of the human pharynx: physical mechanisms studied with fluid-structure interaction simulations and experiments . Physiol Rep. 2019. ; 7 ( 10 ): e14099 . [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Quintana-Gallego E , Carmona-Bernal C , Capote F , et al . Gender differences in obstructive sleep apnea syndrome: a clinical study of 1166 patients . Respir Med. 2004. ; 98 ( 10 ): 984 – 989 . [DOI] [PubMed] [Google Scholar]
- 19. World Inequality Database . Income Inequality, Korea, 1996– 2012. . https://wid.world/country/korea . Accessed June 11, 2021.
- 20. Jun MJ , Kim JI , Kwon JH , Jeong JE . The effects of high-density suburban development on commuter mode choices in Seoul, Korea . Cities. 2013. ; 31 : 230 238 . [Google Scholar]
- 21. Atkins BZ , Aldea GS . The paradox between randomized controlled trials and propensity score-matched. real-world data: Moving from dissonance to dialog? J Thorac Cardiovasc Surg. 2018. ; 156 ( 3 ): 1026 1027 . [DOI] [PubMed] [Google Scholar]