Abstract
Objective:
This study examined the association between obstructive sleep apnea (OSA) risk and hearing difficulty and evaluated whether oral health problems mediate this relationship in a nationally representative Korean population.
Methods:
We analyzed 15,163 adults aged ≥40 years from the Korea National Health and Nutrition Examination Survey (KNHANES) 2019–2023 datasets. OSA risk was defined using the snoring, tiredness, observed apneas, pressure, body mass index, age, neck circumference, and gender (STOP-BANG) questionnaire (low risk: 0–2; high risk: ≥3). Oral health problems were operationalized as the summed score of self-reported chewing difficulty and speaking difficulty (two 5-point items; range 2–10). Hearing difficulty was assessed using the KNHANES 4-point self-reported hearing scale (1 = no difficulty to 4 = cannot hear at all). Mediation analyses were conducted using complex-sample hierarchical regression with design-weighted estimation, adjusting for survey year, age, sex, education, marital status, income, smoking, alcohol consumption, metabolic/cardiovascular disease comorbidities, occupational noise exposure, and tinnitus.
Results:
Adults with high-OSA risk had higher weighted mean hearing difficulty scores than the low-risk group (1.23 vs. 1.16). In the adjusted model, OSA risk was associated with hearing difficulty [total effect β = 0.031, 95% confidence interval (CI): 0.013–0.048, P < 0.001]. Oral problems were associated with OSA risk (β = 0.111, 95% CI: 0.049–0.174, P < 0.001) and hearing difficulty (β = 0.014, 95% CI: 0.000–0.027, p = 0.047). The indirect effect was statistically significant (β = 0.001, 95% CI: 0.001–0.002). The direct effect of OSA risk on hearing difficulty remained significant (β = 0.019, 95% CI: 0.011–0.047, P < 0.001).
Conclusions:
The risk of OSA based on STOP-BANG was associated with increased hearing difficulty among Korean adults, and oral health problems statistically accounted for a small proportion of this cross-sectional association. These results highlight the potential value of jointly considering sleep-related symptoms, oral function, and hearing outcomes in population health research.
Keywords: Surveys and questionnaires, hearing, obstructive sleep apnea, oral health, mediation analysis
Yuanyuan Xing and Yu-Rin Kim are co-first authors. These authors contributed equally to this work.
KEY MESSAGES
-
(1)
Nationally representative data from Korean adults aged ≥40 years indicated an association between screen-detected obstructive sleep apnea (OSA) risk and poor oral function, a high prevalence of hearing difficulties, and a great burden of comorbidities.
-
(2)
A composite measure of chewing and speaking problems statistically accounted for part of the association between OSA risk and hearing difficulty in a cross-sectional decomposition analysis.
-
(3)
Findings highlight the interconnected nature of sleep-related symptoms, oral health, and auditory outcomes.
-
(4)
All results are hypothesis-generating and require confirmation in longitudinal or experimental studies.
Introduction
Obstructive sleep apnea (OSA) is a common sleep-related breathing disorder that involves recurrent upper-airway obstruction during sleep, accompanied by intermittent hypoxemia and sleep fragmentation. These disturbances affect 9% to 38% of adults globally and impose a substantial public health burden, contributing to cardiovascular, metabolic, and neurocognitive morbidity.[1,2,3,4] Clinical sleep testing is resource intensive, so many surveillance systems, including the Korea National Health and Nutrition Examination Survey (KNHANES), use validated questionnaires to estimate population-level OSA risk.
Since 2019, KNHANES has included selected items from the snoring, tiredness, observed apnea, pressure, body mass index (BMI), age, neck circumference, and gender (STOP-BANG) questionnaire, allowing researchers to assess the risk of OSA per STOP-BANG item rather than confirmed clinical OSA.[5] In the present study, OSA risk was defined using these survey items from the 2019 to 2023 data, with complex sampling weights, strata, and primary sampling units (PSUs) applied to ensure national representativeness.
Biologically plausible mechanisms suggest that OSA may be associated with auditory dysfunction through pathways involving intermittent hypoxemia, vascular compromise, and oxidative stress, which may influence oxygen-dependent cochlear structures.[6] Epidemiologic findings, however, remain heterogeneous. Some studies report associations between OSA severity and impaired hearing, whereas other works, particularly those using self-reported hearing difficulty or limited adjustment for confounding, find no association.[7,8,9,10] A small clinical study noted no improvement in hearing following continuous positive airway pressure therapy, although its limited sample size, short follow-up, and lack of age adjustment limit interpretability.[11] Variability across hearing measures (self-reported difficulty vs. pure-tone audiometry), study designs (cross-sectional vs. longitudinal), and confounder control (age, lifetime noise exposure, metabolic disease, ototoxic medications) likely contribute to these mixed results.[7,8,9,10,11]
Oral health problems, such as chewing and speaking difficulties, may also play a role in this relationship. These oral function impairments reflect localized oral disease burden and broad functional limitations influencing nutrition, communication, and quality of life.[12,13,14] Potential links between OSA risk and poor oral status have been suggested, including mouth breathing–related xerostomia and inflammatory processes shared between intermittent hypoxemia and periodontal pathology.[15,16,17,18] In KNHANES, chewing and speaking difficulties are self-reported measures collected in the same survey cycles as STOP-BANG items and hearing difficulty and used to evaluate their potential mediating role.[19]
This cross-sectional study uses nationally representative KNHANES data (2019–2023) to examine the association between STOP-BANG–based OSA risk and self-reported hearing difficulty and determine whether oral function problems mediate this association. The primary exposure was STOP-BANG–defined OSA risk, the primary outcome was self-reported hearing difficulty, and the mediator was a composite of chewing and speaking difficulty. Analyses accounted for key confounders, including age, sex, socioeconomic status, smoking, alcohol use, metabolic and cardiovascular comorbidities, tinnitus, and occupational noise exposure. Given the cross-sectional design, mediation analyses reflect statistical decomposition rather than causal mediation.
MATERIALS AND METHODS
Study Design and Data Source
We used data from the KNHANES, cycles 2019–2023. KNHANES is administered annually by the Korea Disease Control and Prevention Agency (KDCA) to monitor population health, chronic disease burden, and health behavior. All data were anonymized before public release, and written informed consent was obtained by KDCA. KNHANES employed a nationally representative, multistage, stratified cluster sampling design. This study used publicly available, de-identified data from 2019 to 2023.[19]
The 2019 to 2023 cycle (KNHANES VIII–IX) included the full STOP-BANG instrument for adults aged ≥40 years.[10] Participants who completed the health interview, health examination, and oral examination were eligible. All analyses incorporated survey weights, strata and PSU, and multiyear pooling followed KNHANES guidelines, including weight scaling across survey years. Individuals missing any STOP-BANG item, oral function variables or hearing difficulty outcomes were excluded. The final analytic sample included 15,163 adults aged ≥40 years [Figure 1].
Figure 1.
Study design.
OSA = obstructive sleep apnea, STOP-BANG = snoring, tiredness, observed apneas, pressure, body mass index, age, neck circumference, and gender
Study Participants
STOP-BANG–based Obstructive Sleep Apnea Risk Assessment
OSA risk was assessed using the complete eight STOP-BANG items: loud snoring, daytime tiredness, observed apnea, high blood pressure, BMI >30 kg/m2, age ≥50 years, neck circumference >40 cm, and male sex. Individuals with scores ≤2 were classified as low-OSA risk, and those with scores ≥3 as high-OSA risk, consistent with validated population screening criteria.[5]
Participant Flow
Among individuals with complete STOP-BANG data, 13,999 had valid oral function and hearing difficulty measures. After excluding 20,590 participants with missing oral-health questionnaire or hearing test data, 15,163 participants remained, comprising 7246 low-risk and 7917 high-risk individuals. The final analytic sample size differed from the full KNHANES adult sample because of item-level missingness in oral function and hearing difficulty measures. We confirmed that the weighted distributions of age, sex, and region in the analytic sample were highly consistent with those of the overall KNHANES 2019–2023 population. All analyses incorporated KNHANES complex sampling weights, strata, and PSU, and population representativeness was preserved irrespective of unweighted sample counts [[Figure 1].
Variables
Demographic Characteristics and Comorbidity Covariates
Demographic variables included age (continuous), sex, education, household income (quintiles), marital status, smoking (never/former/current), and alcohol consumption (<1 per month, ≥1 per month). Comorbidities (hypertension, diabetes mellitus, dyslipidemia, allergic rhinitis, sinusitis, otitis media) were based on self-reported physician diagnoses, consistent with KNHANES methodology.[19]
Primary Outcome: Hearing Difficulty
Self-reported hearing difficulty without using a hearing aid was measured using a four-level item (1 = no difficulty to 4 = cannot hear at all), with higher values indicating worse hearing. For the main analysis, hearing difficulty was modeled as an ordinal outcome using survey-weighted ordinal logistic regression after verifying the proportional-odds assumption.[19]
Hearing-related Covariates
Additional hearing-related variables included use of hearing devices, occupational noise exposure (≥3 years in loud environments), tinnitus experience within 12-month or persistent tinnitus (>6 months), and audiometric hearing loss (covariate only). Hearing loss was defined as the objective audiometric outcome based on pure-tone thresholds. Pure-tone thresholds (500–4000 Hz) were obtained in a sound-attenuated booth using calibrated equipment that met KDCA standards. Moderate-to-severe audiometric hearing loss was defined as ≥41 dB in the worse ear using WHO/KNHANES criteria.[19]
Oral Health–related Variables
Oral health variables included dental pain within the past year, dental caries treatment/endodontic treatment, oral surgery treatment, and prosthetic treatment. Additional measures included toothbrushing frequency (0–5 points) and subjective oral health (5-point scale). Chewing difficulty and speaking difficulty (each 1–5) were summed to create an oral function composite score (range 2–10), defined as oral health problems. Internal consistency was high (Cronbach’s α = 0.74), supporting its use as a continuous mediator.[20]
Bias Considerations
Potential bias sources included selection/nonresponse bias, measurement error (self-reported STOP-BANG items and hearing difficulty), residual confounding (lifetime noise exposure, ototoxic medications, socioeconomic factors), exposure-induced mediator–outcome confounding, and cross-sectional limitations that preclude causal interpretation.
Coding and Quantitative Variables
Coding was standardized (higher hearing difficulty scores = worse hearing; higher chewing/speaking = worse oral function; higher subjective oral health = better oral health). No z-standardization was applied before creating composite scores.
Statistical Methods
Data analyses were conducted using IBM SPSS Statistics Version 29.0 (IBM Corp., Armonk, NY, USA). Figures were created using Microsoft® PowerPoint® 2016 (Microsoft Corporation, Redmond, WA, USA; version 16.0). Survey weights, stratification, and cluster variables were applied in all models. Comparisons of characteristics used complex-sample Rao–Scott χ 2 tests and complex-sample t-tests. Mediation analysis followed the Baron and Kenny[21] three-step framework by using complex-sample hierarchical regression models. The Sobel test was performed to evaluate the significance of the indirect (mediated) effect.[22] Covariates included survey year, age, sex, education, income, marital status, smoking, alcohol use, hypertension, dyslipidemia, diabetes, allergic rhinitis, sinusitis, otitis media, tinnitus, and occupational noise exposure. Variance inflation factors (all <10) indicated no multicollinearity. Total effects (TEs), direct effects (DEs), indirect effects (IE), and proportion-mediated (PM) effects were estimated with 95% confidence intervals (CIs). For sensitivity analysis, mediation was repeated using PROCESS Macro Model 4 with 5000 replicate-weight bootstrap samples and 95% CIs.[23] All frequencies were reported as weighted estimates. Missing data were addressed via survey-compatible multiple imputation, with sampling weights reapplied post-imputation according to KDCA guidelines.[19] A two-sided P-value <0.05 was considered statistically significant.
RESULTS
Sociodemographic Characteristics and Comorbidities by Obstructive Sleep Apnea Risk
Table 1 presents the survey-weighted sociodemographic characteristics of adults classified as low-OSA risk (n = 7246) and high-OSA risk (n = 7917). In terms of gender distribution, all participants in the low-OSA risk group were female, whereas the high-OSA risk group consisted of 89.0% males and 11.0% females. The mean age was 57.41 years in the low-OSA risk group and 58.54 years in the high-OSA risk group. Regarding marital status, 3.3% of the low-OSA risk group were single and 96.7% were married, whereas 7.4% of the high-OSA risk group were single and 92.6% were married. With respect to educational attainment, individuals in the low-OSA risk group most commonly had a high-school education (35.3%), whereas those in the high-OSA risk group most frequently had a college degree or higher (41.4%). For household income, the highest proportion in both groups fell within the high-income category (low-OSA risk: 24.2%; high-OSA risk: 24.7%). About 64.7% of individuals in the low-OSA risk group reported drinking less than once per month, whereas 61.3% of the high-OSA risk group consumed alcohol at least once per month. For smoking status, 90.9% of the low-OSA risk group were never smokers, whereas the high-OSA risk group comprised 44.5% former smokers and 27.6% current smokers. In terms of comorbid conditions, the low-OSA risk group reported hypertension (24%), diabetes (9.9%), and allergic rhinitis (16.7%), whereas the high-OSA risk group showed higher proportions of hypertension (37.3%) and diabetes (16.9%) but lower prevalence of allergic rhinitis (11.9%).
Table 1.
Sociodemographic Characteristics of Subjects by Snoring, Tiredness, Observed Apnea, Pressure, Body Mass Index, Age, Neck Circumference, and Gender-based Obstructive Sleep Apnea Risk (Weighted %).
| Characteristics | Division | Low-OSA risk (n = 7246) | High-OSA risk (n = 7917) | χ 2 | P |
|---|---|---|---|---|---|
| *Years | 2019 (n = 1831) | 961 (11.9) | 870 (10.5) | 17.552 | <0.001 |
| 2020 (n = 3057) | 1531 (23.5) | 1526 (21.7) | |||
| 2021 (n = 3242) | 1628 (23.1) | 1614 (21.8) | |||
| 2022 (n = 3267) | 1431 (20.8) | 1836 (23.4) | |||
| 2023 (n = 3766) | 1695 (20.8) | 2071 (22.6) | |||
| *Gender | Male (n = 6746) | 0 (0.0) | 6746 (89.0) | 12,001.355 | <0.001 |
| Female (n = 8417) | 7246 (100.0) | 1171 (11.0) | |||
| †Age | 57.41±0.19 | 58.54±0.16 | <0.001 | ||
| *Marriage status | Single (n = 702) | 229 (3.3) | 473 (7.4) | 133.611 | <0.001 |
| Married (n = 14461) | 7017 (96.7) | 7444 (92.6) | |||
| *Education level | ≤Elementary school (n = 3478) | 1881 (20.3) | 1597 (15.0) | 81.761 | <0.001 |
| Middle school (n = 1839) | 848 (10.4) | 991 (10.6) | |||
| High school (n = 4728) | 2335 (35.3) | 2393 (32.9) | |||
| ≥College (n = 4773) | 2167 (33.9) | 2606 (41.4) | |||
| *Income level | Low (n = 2745) | 1234 (13.5) | 1511 (14.1) | 3.542 | 0.009 |
| Low-middle (n = 3200) | 1552 (19.4) | 1648 (17.9) | |||
| Middle (n = 2934) | 1407 (20.1) | 1527 (20.2) | |||
| Middle-high (n = 3073) | 1487 (22.9) | 1586 (23.1) | |||
| High (n = 3178) | 1549 (24.2) | 1629 (24.7) | |||
| *Alcohol consumption | <1 per month (n = 8180) | 4821 (64.7) | 3359 (38.7) | 1983.304 | <0.001 |
| ≥1 per month (n = 6918) | 2411 (35.3) | 4507 (61.3) | |||
| *Smoking | Never smoker (n = 8997) | 6585 (90.9) | 2412 (27.9) | 3315.946 | <0.001 |
| Former smoker (n = 3859) | 399 (5.7) | 3460 (44.5) | |||
| Current smoker (n = 2230) | 247 (3.4) | 1983 (27.6) | |||
| *Comorbidities | Hypertension (yes, %) | 1978 (24.0) | 3373 (37.3) | 378.411 | <0.001 |
| Dyslipidemia (yes, %) | 2261 (28.4) | 2513 (29.4) | 3.293 | 0.071 | |
| Diabetes (yes, %) | 805 (9.9) | 1482 (16.9) | 423.355 | <0.001 | |
| Allergic rhinitis (yes, %) | 1110 (16.7) | 825 (11.9) | 142.421 | <0.001 | |
| Sinusitis (yes, %) | 452 (6.4) | 462 (6.5) | 0.008 | 0.929 | |
| Otitis media (yes, %) | 362 (5.2) | 403 (5.4) | 0.815 | 0.368 |
OSA = obstructive sleep apnea, STOP-BANG = snoring, tiredness, observed apneas, pressure, body mass index, age, neck circumference, and gender count for categorical variables reflect valid responses only; therefore, category totals may not sum to the full analytic sample size. Missing values in covariates were retained in multivariable analyses using a missing-indicator approach, consistent with KNHANES analytic guidelines; * By complex sample Rao–Scott χ 2 test; †By complex sample linear regression analysis (Reference group; high-OSA risk), P < 0.05.
Hearing-related Characteristics According to Obstructive Sleep Apnea Risk
Weighted distributions for hearing-related variables are shown in Table 2. Use of hearing devices was 1.5 times more common among individuals at high-OSA risk (11.9%) compared with those at low risk (8.0%) [odds ratio (OR):1.559, 95% CI: 1.339–1.814, P < 0.001). Occupational noise exposure was more than twice as common among individuals at high-OSA risk (21.9%) compared with those at low risk (10.6%) (OR: 2.361, 95% CI: 2.235–2.493, P < 0.001). Reported tinnitus was also more prevalent in the high-risk group (11.9% vs. 7.5%, OR: 1.675, 95% CI: 1.562–1.797, P < 0.001). Audiometric hearing loss, used as a covariate rather than outcomes, also differed significantly. Weighted prevalence of mild hearing loss was higher in the high-risk group (31.0%) compared with the low-risk group (21.1%) (OR: 1.686, 95% CI: 1.583–1.795, P < 0.001). Weighted prevalence of moderate-to-severe hearing loss (≥41 dB in the worse ear) was higher in the high-risk group (18.7%) compared with the low-risk group (10.6%) (OR: 1.939, 95% CI: 1.824–2.062, P < 0.001). Self-reported hearing difficulty, which was modeled as a 4-level ordinal outcome (1 = no difficulty to 4 = cannot hear at all), displayed a significantly higher weighted mean in the high-OSA risk group (1.23) than in the low-risk group (1.16), with β = 0.069, 95% CI: 0.028–0.081, and P < 0.001.
Table 2.
Hearing-related Characteristics according to Snoring, Tiredness, Observed Apnea, Pressure, Body Mass Index, Age, Neck Circumference, and Gender −based Obstructive Sleep Apnea Risk (Weighted %).
| Characteristics | Low-OSA Risk (n = 7246) | High-OSA Risk (n = 7917) | OR, β (95% CI) | P |
|---|---|---|---|---|
| *Use of hearing devices (self-report) (used, %) | 100 (8.0) | 249 (11.9) | 1.559 (1.339–1.814) | <0.001 |
| *Occupational noise exposure (self-report) (yes, %) | 822 (10.6) | 1736 (21.9) | 2.361 (2.235–2.493) | <0.001 |
| *Tinnitus experience (self-report) (yes, %) | 613 (7.5) | 1060 (11.9) | 1.675 (1.562–1.797) | <0.001 |
| *Persistent tinnitus (self-report) (yes, %) | 484 (79.3) | 911 (85.6) | 1.548 (1.210–1.980) | <0.001 |
| *Mild hearing loss (clinical) (yes, %) | 1593 (21.1) | 2547 (31.0) | 1.686 (1.583–1.795) | <0.001 |
| *Moderate to severe hearing loss (clinical) (yes, %) | 870 (10.6) | 1684 (18.7) | 1.939 (1.824–2.062) | <0.001 |
| †Hearing difficulties (self-report)‡(Range: 1–4) | 1.16 ± 0.00 | 1.23 ± 0.01 | 0.069 (0.058–0.081 | <0.001 |
CI = confidence interval, OR = odds ratio, OSA = obstructive sleep apnea, STOP-BANG = snoring, tiredness, observed apneas, pressure, body mass index, age, neck circumference, and gender; * By complex sample Rao–Scott χ 2 test; †By complex sample linear regression analysis, P < 0.05; ‡Higher score means worse status.
Oral Health-related Characteristics According to Obstructive Sleep Apnea Risk
Table 3 summarizes the oral health characteristics. Participants at high-OSA risk consistently exhibited worse oral-health profiles, including higher weighted prevalence of dental pain (24.2%), caries treatment (20.4%), endodontic therapy (24.0%), periodontal treatment (22.9%), oral surgery (19.8%), and prosthetic treatment (35.6%) (all P < 0.05). Weighted mean tooth-brushing frequency was lower, and subjective oral-health ratings were poorer among individuals with elevated OSA risk.
Table 3.
Oral Health-related Characteristics according to Snoring, Tiredness, Observed Apnea, Pressure, Body Mass Index, Age, Neck Circumference, and Gender-based Obstructive Sleep Apnea Risk (Weighted % or Mean ± Standard Error).
| Division | Low-OSA Risk (n = 7246) | High-OSA Risk (n = 7917) | OR, β(95% CI) | P |
|---|---|---|---|---|
| *Dental pain experience (self-report) (yes, %) | 1508 (19.2) | 2045 (24.2) | 1.344 (1.269–1.423) | <0.001 |
| *Oral examination (clinical) (yes, %) | 3053(43.6) | 3109 (40.9) | 0.894 (0.851–0.940) | <0.001 |
| *Dental caries treatment (clinical) (yes, %) | 794 (18.6) | 918 (20.4) | 1.121 (1.047–1.200) | 0.001 |
| *Endodontic therapy (clinical) (yes, %) | 735 (16.5) | 1145 (24.0) | 1.600 (1.489–1.718) | <0.001 |
| *Periodontal treatment (clinical) (yes, %) | 904 (20.9) | 1130 (22.9) | 1.124 (1.042–1.213) | 0.003 |
| *Oral surgery treatment (clinical) (yes, %) | 665 (14.4) | 955 (19.8) | 1.465 (1.351–1.588) | <0.001 |
| *Prosthetic treatment (clinical) (yes, %) | 1400 (29.9) | 1899 (35.6) | 1.297 (1.225–1.373) | <0.001 |
| †Daily toothbrushing frequency (self-report) | 2.91 ± 0.01 | 2.63 ± 0.01 | −0.278(−0.301–−0.256) | <0.001 |
| †Self-rated oral health | 2.79 ± 0.01 | 2.70 ± 0.01 | −0.093(−0.112–−0.073) | <0.001 |
| †Chewing function problem ‡(range: 1–5) (self-report) | 2.21 ± 0.01 | 2.39 ± 0.01 | 0.179(0.153–0.205) | <0.001 |
| †Speaking function problem ‡(range: 1–5) (self-report) | 1.52 ± 0.01 | 1.64 ± 0.01 | 0.122(0.101–0.143) | <0.001 |
| †Oral problems (chewing + speaking difficulty; ‡range 2–10) (self-report) | 3.73 ± 0.02 | 4.03 ± 0.02 | 0.301(0.260–0.343) | <0.001 |
By complex sample Rao–Scott χ2 test; †By complex sample linear regression analysis. P < 0.05; ‡Higher score means worse status; CI = confidence interval, OR = odds ratio, OSA = obstructive sleep apnea, STOP-BANG = snoring, tiredness, observed apneas, pressure, body mass index, age, neck circumference, and gender
The oral problem composite score, which was derived from self-reported chewing difficulty and speaking difficulty (range 2–10; higher = worse function), was significantly higher among the high-risk group (4.03) than among the low-risk group (3.73; β = 0.301, 95% CI: 0.260–0.343, P < 0.001).
Mediation Analysis: Oral Problems as a Mediator of the Association between Obstructive Sleep Apnea Risk and Hearing Difficulty
Survey-weighted mediation analyses assessed whether oral problems statistically accounted for part of the association between OSA risk and hearing difficulty. Linear models were evaluated for approximate normality of residuals because hearing difficulty is ordinal; hence, diagnostics supported their suitability for the mediation framework.
Table 4 presents the results of the mediation analysis examining whether oral problems mediate the association between STOP-BANG OSA risk score and self-reported hearing difficulties. In Model 1 (unadjusted), OSA risk was significantly associated with both oral problems (β = 0.188, 95% CI: 0.168–0.208, P < 0.001) and hearing difficulties (β = 0.044, 95% CI: 0.039–0.048, P < 0.001). When oral problems were added to the model, the coefficient for OSA risk on hearing difficulties was attenuated (β = 0.037, P < 0.001), while oral problems showed a significant positive association with hearing difficulties (β = 0.032, 95% CI: 0.029–0.036, P < 0.001). These findings indicate a partial mediation effect. In Model 1, the explanatory power (R 2) was 0.016 at Stage 1, 0.015 at Stage 2, and 0.031 at Stage 3. The Sobel test confirmed that the IE was statistically significant (Z = 13.364, P < 0.001).
Table 4.
Mediation of the Association between Obstructive Sleep Apnea Risk and Hearing Difficulty by Oral Problems.
| Step | Model 1 | β | 95% CI | t | P |
|---|---|---|---|---|---|
| 1 | OSA risk → oral problems | 0.188 | (0.168–0.208) | 18.469 | <0.001 |
| 2 | OSA risk → hearing difficulties | 0.044 | (0.039– 0.048) | 18.614 | <0.001 |
| 3 | OSA risk → hearing difficulties | 0.037 | (0.033–0.042) | 16.181 | <0.001 |
| Oral problems → hearing difficulties | 0.032 | (0.029–0.036) | 20.102 | <0.001 | |
| Step | Model 2 | β | 95% CI | t | P |
| 1 | OSA risk → oral problems | 0.111 | (0.049–0.174) | 3.516 | <0.001 |
| 2 | OSA risk → hearing difficulties | 0.031 | (0.013–0.048) | 3.471 | <0.001 |
| 3 | OSA risk → hearing difficulties | 0.029 | (0.011–0.047) | 3.270 | <0.001 |
| Oral problems → hearing difficulties | 0.014 | (0.000–0.027) | 2.009 | 0.047 |
By complex-sample hierarchical regression analysis. P < 0.05. Model 1; unadjusted, Model 2; adjusted survey year, sex, age, marital status, income, education, smoking, alcohol consumption, hypertension, dyslipidemia, diabetes, allergic rhinitis, sinusitis, otitis media, occupational noise, and tinnitus. Model 1; Step 1’s R 2 = 0.016, Step 2’s R 2 = 0.015, Step 3’s R 2 = 0.031. Model 2; Step 1’s R 2 = 0.149, Step 2’s R 2 = 0.294, Step 3’s R 2 = 0.295; CI = confidence interval, OSA = obstructive sleep apnea
Model 2 was adjusted for survey year, sex, age, marital status, income, education, smoking, alcohol consumption, comorbidities (hypertension, dyslipidemia, diabetes, allergic rhinitis, sinusitis, and otitis media), occupational noise exposure, and tinnitus. In the adjusted model, OSA risk remained significantly associated with oral problems (β = 0.111, 95% CI: 0.049–0.174, P < 0.001) and hearing difficulties (β = 0.031, 95% CI: 0.013–0.048, P < 0.001). In the fully adjusted mediation model, the association between OSA risk and hearing difficulties was further reduced (β = 0.029, P < 0.001), and oral problems continued to have a small but statistically significant IE on hearing difficulties (β = 0.014, 95% CI: 0.000–0.027, P = 0.047). In Model 2, the explanatory power (R2) was 0.149 at Stage 1, 0.294 at Stage 2, and 0.295 at Stage 3. The Sobel test supported the statistical significance of the IE (Z = 1.989, P = 0.047). These findings indicate that oral problems account for a modest but statistically significant proportion of the association between OSA risk and hearing difficulties, demonstrating a consistent partial mediation pattern across unadjusted and adjusted models. The Sobel tests in both models supported the significance of the indirect pathways.
Sensitivity Analysis
PROCESS Macro Model 4 (5000 bootstrap samples) was used as an exploratory sensitivity analysis. The PROCESS assumes simple random sampling, so the resulting estimates were not design-consistent and were therefore interpreted only as supplementary. The bootstrap-derived IEs remained statistically significant in the unadjusted model (β = 0.007; 95% CI: BootLLCI = 0.006, BootULCI = 0.008) and the adjusted model (β = 0.001; 95% CI: BootLLCI = 0.001, BootULCI = 0.002). These findings support the robustness of the observed mediation effect, whereas appropriately acknowledging that the results reflect statistical rather than causal inference [Tables 5 and 6 and Figure 2].
Table 5.
Mediating Effect of Oral Problems on Obstructive Sleep Apnea Risk and Hearing Difficulties.
| Variables | Oral Problems | Hearing Difficulties | ||||||
|---|---|---|---|---|---|---|---|---|
|
|
|
|
||||||
| Model 1 | β | SE | T | P | β | SE | t | P |
| OSA risk | 0.207 | 0.012 | 17.039 | <0.001 | 0.042 | 0.003 | 13.533 | <0.001 |
| Oral problems | 0.035 | 0.002 | 17.496 | <0.001 | ||||
| F (P) | 290.331 (<0.001) | 280.869 (<0.001) | ||||||
| R 2 | 0.132 | 0.183 | ||||||
| Model 2 | β | SE | T | P | β | SE | t | P |
| OSA risk | 0.085 | 0.016 | 5.475 | <0.001 | 0.019 | 0.003 | 5.372 | <0.001 |
| Oral problems | 0.015 | 0.002 | 8.005 | <0.001 | ||||
| F (P) | 116.343 (<0.001) | 325.840 (<0.001) | ||||||
| R 2 | 0.353 | 0.544 | ||||||
Analysis of mediating effects using PROCESS macro model 4 with adjusted survey year, sex, age, marital status, income, education, smoking, alcohol consumption, hypertension, dyslipidemia, diabetes, allergic rhinitis, sinusitis, otitis media, occupational noise, and tinnitus, P < 0.05; OSA = obstructive sleep apnea, SE = standard error
Table 6.
Significance Test on the Indirect Effect of Oral Problems.
| Variables | Model | Effect | BootSE | 95% CI | |
|---|---|---|---|---|---|
|
|
|||||
| BootLLCI | BootULCI | ||||
| Oral problems | 1 | 0.007 | 0.001 | 0.006 | 0.008 |
| 2 | 0.001 | 0.000 | 0.001 | 0.002 | |
Bootstrap samples = 5000; CI = confidence interval
Figure 2.
Mediating effect of oral problems on OSA risk and hearing difficulties.
Notes: Model 1; unadjusted, Model 2; adjusted survey year, sex, age, marital status, income, education, smoking, alcohol consumption, hypertension, dyslipidemia, diabetes, allergic rhinitis, sinusitis, otitis media, occupational noise, and tinnitus. *** P < 0.001. OSA = obstructive sleep apnea
DISCUSSION
Using nationally representative KNHANES data from 2019 to 2023, this study examined the survey-weighted associations among STOP-BANG–defined OSA risk, oral function problems, and self-reported hearing difficulties in Korean adults aged ≥40 years. High-OSA risk was associated with a greater burden of cardiometabolic comorbidities, poorer oral function scores, and more severe hearing difficulty compared with low-OSA risk. Mediation analyses further indicated that oral function problems accounted for a small portion of the cross-sectional association between OSA risk and hearing difficulty. However, the study is cross-sectional and relies on self-reported measures. In this regard, the findings reflect statistical decomposition of associations rather than causal pathways and should be interpreted as exploratory and hypothesis-generating.
Descriptive results showed that participants at high-OSA risk had higher prevalence of hypertension, diabetes, tinnitus, occupational noise exposure, mild and moderate-to-severe hearing loss, and hearing difficulties as well as poorer oral function indicators, including chewing and speaking difficulties. These patterns may be influenced by underlying sociodemographic differences in age, sex, socioeconomic status, and comorbidity burden. After applying survey weights, strata, and PSU and adjusting for relevant covariates, the associations among OSA risk, oral problems, and hearing difficulties remained statistically significant. Hence, the observed relationships could not be fully explained by demographic or health-related confounding data.
Interpretation of STOP-BANG also requires caution. Although widely used as a screening tool, STOP-BANG does not diagnose OSA or indicate clinical severity, and misclassification relative to polysomnography is expected.[5] Thus, references to “OSA risk” should not be interpreted as a confirmed OSA diagnosis. The binary threshold used in this analysis (≥3) is consistent with previous population-based validations but does not necessarily reflect disease staging. The distribution of risk factors in the high-OSA risk group (male sex, older age, higher adiposity, and smoking history) is consistent with prior epidemiologic studies.[24,25,26] The higher burden of cardiometabolic and inflammatory conditions, such as hypertension, dyslipidemia, diabetes, allergic rhinitis, sinusitis, and otitis media, aligns with profiles commonly associated with sleep-disordered breathing,[27,28] although such associations do not imply causation.High-OSA risk was also associated with greater hearing difficulty, tinnitus, occupational noise exposure, and a higher prevalence of audiometric hearing loss. These results are consistent with studies reporting associations between OSA severity and high-frequency hearing loss,[29] although other research using 2019 to 2020 Korean national data found no significant associations after adjusting for age and sex.[10] By incorporating a larger time span (2019–2023) and applying full survey-weighted analyses, the present findings demonstrated consistent associations between STOP-BANG–defined OSA risk and hearing-related outcomes, indicating robust patterns across subjective and objective measures. Nonetheless, crude differences in moderate-to-severe hearing loss may still reflect the different age structures of the OSA risk groups because hearing impairment increases steeply with age and cumulative exposure effects.
Oral health findings further highlighted notable burdens among individuals with high-OSA risk. These participants reported dental pain, recent dental caries or endodontic treatment, periodontal or restorative treatment, and impaired chewing and speaking ability. Although consistent with prior observations linking sleep-disordered breathing to xerostomia, periodontal problems, and impaired oral function,[30,31] the cross-sectional design precludes determining whether OSA symptoms precede or follow oral deterioration. The mediator used in this study, which is a composite score combining chewing and speaking difficulty (range 2–10), showed adequate internal consistency (Cronbach’s α = 0.74), yet remains vulnerable to measurement error due to self-reporting. In addition, OSA symptoms, oral dysfunction, and hearing problems may share upstream determinants, raising the possibility of exposure-induced mediator–outcome confounding, which cannot be fully addressed in this design. The mediation findings therefore represent statistical partitioning of associations, not evidence of biological mediation.
This study benefits from a large, nationally representative sample and the correct application of KNHANES complex survey design elements, including weights, strata, and PSUs, enabling population-level inference. By integrating sleep-related symptoms, oral function indicators, and hearing outcomes, the study provides a multisystem perspective rarely addressed in prior research. However, several limitations should be noted. Hearing difficulty was self-reported, thereby increasing the potential for misclassification compared with objective audiometric data. OSA risk was assessed using a screening tool rather than polysomnography, which may increase the likelihood of exposure misclassification. Residual confounding from lifetime noise exposure, ototoxic medications, genetic factors, and environmental influences may remain despite covariate adjustment. Multiple comparisons across numerous outcomes raise the possibility of type I error. Although sensitivity analyses employed PROCESS Macro Model 4 with replicate-weight bootstrapping, this approach assumes simple random sampling, and fully design-consistent mediation methods would strengthen future causal inferences. The resulting risk groups exhibited extreme sex imbalance (high-risk: mostly men; low-risk: mostly women) because STOP-BANG scores are closely related to sex. While we adjusted for sex in the regression models, covariate adjustment may be insufficient under such near-complete separation, and residual confounding cannot be ruled out. This may affect internal validity and restrict the generalizability of our findings. Finally, the cross-sectional nature of the data prevents establishing temporal relationships among OSA risk, oral dysfunction, and hearing difficulty, and some covariates contained missing values even after excluding individuals with missing primary variables, potentially contributing to residual bias.
Future research should incorporate longitudinal designs with repeated audiometric assessments, polysomnography-confirmed OSA diagnoses and objective periodontal evaluations to clarify temporal ordering and mechanisms. Design-consistent causal mediation methods and effect-modification analyses by sex, age, and noise exposure are needed. Interventional studies exploring whether improvements in oral health can modify hearing-related outcomes among individuals at elevated OSA risk should be pursued, although such hypotheses require rigorous experimental testing. Overall, while this study identifies meaningful associations among STOP-BANG–defined OSA risk, oral function problems and hearing difficulty in a national population, these findings remain exploratory, and causal interpretations must await confirmation through longitudinal or experimental research.
CONCLUSIONS
Using nationally representative data from Korean adults aged ≥40 years, this study identified survey-weighted associations between STOP-BANG–based OSA risk and a range of sociodemographic characteristics, comorbidities, oral health problems, and hearing difficulties. Oral problems, which were measured as a composite of chewing and speaking difficulties, accounted for a portion of the cross-sectional association between OSA risk and hearing difficulty in a statistical decomposition; however, these findings do not establish temporal ordering or a biological mediation pathway. The exposure reflects screen-detected OSA risk rather than clinically diagnosed OSA, so the results should not be interpreted as differences in disease severity.
Our findings suggest that adults with elevated OSA risk exhibit adverse oral and auditory profiles, supporting the value of an integrated, multisystem perspective on sleep-related health. Nevertheless, any implications for screening or intervention should be considered hypothesis-generating because causal relationships cannot be inferred from cross-sectional data. Limitations, including potential misclassification in self-reported hearing difficulty, uncontrolled lifetime noise or ototoxic exposures, and multiple testing, underscore the need for caution in interpretation.
Future research should use design-consistent causal mediation methods, longitudinal follow-up with validated periodontal and audiometric measures and pre-registered analyses to confirm these associations and determine whether oral health may serve as a potential target for mitigating hearing-related outcomes among individuals at risk for OSA. These findings apply to the Korean adult population aged ≥40 years, and external validity to other populations remains uncertain.
Availability of Data and Materials
The data presented in this study are available on request from the corresponding author.
Author Contributions
Conceptualization, Y.R.K. and Y.X.; data curation, S.R.K.; formal analysis, Y.R.K. and Y.X.; methodology, Y.R.K.; resources, Y.X. and S.R.K.; supervision, S.R.K. and Y.R.K.; validation, Y.R.K. and Y.X.; writing—original draft, S.R.K., Y.X., and Y.R.K.; writing—review and editing, S.R.K., Y.X., and Y.R.K. All authors have read and agreed to the published version of the manuscript.
Ethics Approval and Consent to Participate
The KNHANES is a nationwide statutory survey conducted to assess health behaviors, the prevalence of chronic diseases, and dietary and nutritional intake under Article 16 of the National Health Promotion Act. It is designated as an official government statistic (Approval No. 117002) in accordance with Article 17 of the Statistics Act. This study utilized data from the 2019 to 2023 KNHANES. Ethical approval was obtained in consideration of the collection of human biological materials and the provision of deidentified data to third parties. The study protocol was reviewed and approved by the IRB of the KCDC (IRB Nos. 2018-01-03-C-A, 2018-01-03-2C-A, 2018-01-03-3C-A, 2018-01-03-4C-A, and 2022-11-16-R-A). All KNHANES data are anonymized to prevent the identification of individual participants in compliance with the Personal Information Protection Act and the Statistics Act. In addition, informed consent for participation and the use of human-derived materials was obtained from all participants by the KCDC prior to data collection.
Conflicts of Interest
The authors have no conflicts of interest relevant to this article to disclose.
Acknowledgment
We thank the Korea Disease Control and Prevention Agency (KDCA) for granting access to the Korea National Health and Nutrition Examination Survey (KNHANES) 2019–2023 datasets.
Funding Statement
This research received no external funding.
REFERENCES
- 1.de Araujo Dantas AB, Goncalves FM, Martins AA, et al. Worldwide prevalence and associated risk factors of obstructive sleep apnea: a meta-analysis and meta-regression. Sleep Breath. 2023;27:2083–109. doi: 10.1007/s11325-023-02810-7. [DOI] [PubMed] [Google Scholar]
- 2.Ghavami T, Kazeminia M, Ahmadi N, Rajati F. Global prevalence of obstructive sleep apnea in the elderly and related factors: a systematic review and meta-analysis study. J PeriAnesth Nurs. 2023;38:865–75. doi: 10.1016/j.jopan.2023.01.018. [DOI] [PubMed] [Google Scholar]
- 3.Landry SA, Beatty C, Thomson LD, et al. A review of supine position related obstructive sleep apnea: classification, epidemiology, pathogenesis and treatment. Sleep Med Rev. 2023;72:101847. doi: 10.1016/j.smrv.2023.101847. [DOI] [PubMed] [Google Scholar]
- 4.Salari N, Khazaie H, Abolfathi M, et al. The effect of obstructive sleep apnea on the increased risk of cardiovascular disease: a systematic review and meta-analysis. Neurol Sci. 2022;43:219–31. doi: 10.1007/s10072-021-05765-3. [DOI] [PubMed] [Google Scholar]
- 5.Hwang M, Nagappa M, Guluzade N, Saripella A, Englesakis M, Chung F. Validation of the STOP-BANG questionnaire as a preoperative screening tool for obstructive sleep apnea: a systematic review and meta-analysis. BMC Anesthesiol. 2022;22:366. doi: 10.1186/s12871-022-01912-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Seo YJ, Ju HM, Lee SH. Damage of inner ear sensory hair cells via mitochondrial loss in a murine model of sleep apnea with chronic intermittent hypoxia. Sleep. 2017:40. doi: 10.1093/sleep/zsx106. [DOI] [PubMed] [Google Scholar]
- 7.Locke BW, Lee JJ, Sundar KM. OSA and chronic respiratory disease: mechanisms and epidemiology. Int J Environ Res Public Health. 2022;19:5473. doi: 10.3390/ijerph19095473. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Lu CT, Lee LA, Lee GS, Li HY. Obstructive sleep apnea and auditory dysfunction—does snoring sound play a role? Diagnostics. 2022;12:2374. doi: 10.3390/diagnostics12102374. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Lisan Q, van Sloten T, Climie RE. Sleep apnoea is associated with hearing impairment: the Paris prospective study 3. Clin Otolaryngol. 2020;45:681–6. doi: 10.1111/coa.13557. [DOI] [PubMed] [Google Scholar]
- 10.Lee JM, Lee HJ. Is sleep apnea truly associated with hearing loss? A nationwide, population-based study with STOP-BANG questionnaire. Front Public Health. 2023;11:1170470. doi: 10.3389/fpubh.2023.1170470. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Farber MM, Silva LAF, Lorenzi-Filho G, Matas CG. Effects of continuous positive airway pressure on the degree of sleepiness, functional quality of sleep, and electrophysiological hearing responses in individuals with obstructive sleep apnea. Sleep Sci. 2025;18:e281–91. doi: 10.1055/s-0045-1809060. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Northridge ME, Kumar A, Kaur R. Disparities in access to oral health care. Annu Rev Public Health. 2020;41:513–35. doi: 10.1146/annurev-publhealth-040119-094318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Yamano T, Nishi K, Omori F, Wada K, Naito T. Association between oral health and swallowing function in the elderly. Clin Interv Aging. 2023:343–51. doi: 10.2147/CIA.S400032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Qamar Z, Alghamdi AMS, Haydarah NKB, et al. Impact of temporomandibular disorders on oral health‐related quality of life: a systematic review and meta‐analysis. J Oral Rehabil. 2023;50:706–14. doi: 10.1111/joor.13472. [DOI] [PubMed] [Google Scholar]
- 15.Zhu J, Yuan X, Zhang Y, Wei F, Hou Y, Zhang Y. A meta-analysis on the association between obstructive sleep apnea and periodontitis. Sleep Breath. 2023;27:641–9. doi: 10.1007/s11325-022-02668-1. [DOI] [PubMed] [Google Scholar]
- 16.Zhang C, Shen Y, Liping F, Ma J, Wang GF. The role of dry mouth in screening sleep apnea. Postgrad Med J. 2021;97:294–8. doi: 10.1136/postgradmedj-2020-137619. [DOI] [PubMed] [Google Scholar]
- 17.Pau MC, Zinellu A, Mangoni AA, et al. Evaluation of inflammation and oxidative stress markers in patients with Obstructive Sleep Apnea (OSA) J Clin Med. 2023;12:3935. doi: 10.3390/jcm12123935. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Cardoso EM, Reis C, Manzanares-Cespedes MC. Chronic periodontitis, inflammatory cytokines, and interrelationship with other chronic diseases. Postgraduate Med. 2018;130:98–104. doi: 10.1080/00325481.2018.1396876. [DOI] [PubMed] [Google Scholar]
- 19. https://knhanes.kdca.go.kr/knhanes/postSendPage.do?url=%2FdataAnlsGd%2FutztnGd.do&postparam=%7B%22menuId%22%3A%20%22100310021001%22%7D [Google Scholar]
- 20.Im I. Chewing and speaking difficulties, quality of life and depression in elderly Koreans: a survey of the Korea National Health and Nutrition Examination from 2016 ∼ 2018. Speech Lang Hearing. 2020;29:121–30. [Google Scholar]
- 21.Baron RM, Kenny DA. The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol. 1986;51:1173–82. doi: 10.1037//0022-3514.51.6.1173. [DOI] [PubMed] [Google Scholar]
- 22.Sobel ME. Asymptotic confidence intervals for indirect effects in structural equation models. Sociol Methodol. 1982;13:290–312. [Google Scholar]
- 23.Hayes AF. Introduction to Mediation, Moderation, and Conditional Process Analysis. A Regression-Based Approach. 3rd ed. New York, NY: Guilford Publications; 2022. [Google Scholar]
- 24.Tete B, Albdewi MA, Nkodila A, et al. Prevalence of risk and factors associated with obstructive sleep apnea-hypopnea syndrome in an adult population in Kinshasa, Democratic Republic of Congo. J Sleep Res. 2023;32:e13637. doi: 10.1111/jsr.13637. [DOI] [PubMed] [Google Scholar]
- 25.Feltner C, Wallace IF, Aymes S, et al. Screening for obstructive sleep apnea in adults: updated evidence report and systematic review for the US Preventive Services Task Force. JAMA. 2022;328:1951–71. doi: 10.1001/jama.2022.18357. [DOI] [PubMed] [Google Scholar]
- 26.Gottlieb DJ. Screening for obstructive sleep apnea in adults. JAMA. 2022;328:1908–10. doi: 10.1001/jama.2022.20670. [DOI] [PubMed] [Google Scholar]
- 27.Redline S, Azarbarzin A, Peker Y. Obstructive sleep apnea heterogeneity and cardiovascular disease. Nat Rev Cardiol. 2023;20:560–73. doi: 10.1038/s41569-023-00846-6. [DOI] [PubMed] [Google Scholar]
- 28.Cottone C, Im E, Clausen S, Carr M. Is there an increased risk of otitis media in children with obstructive sleep apnea? Int J Pediatr Otorhinolaryngol. 2024;182:112014. doi: 10.1016/j.ijporl.2024.112014. [DOI] [PubMed] [Google Scholar]
- 29.Kasemsuk N, Chayopasakul V, Banhiran W. Obstructive sleep apnea and sensorineural hearing loss: a systematic review and meta-analysis. Otolaryngol Head Neck Surg. 2023;169:201–9. doi: 10.1177/01945998221120777. [DOI] [PubMed] [Google Scholar]
- 30.Arango Jimenez N, Morales Vera DZ, Latorre Uriza C, Velosa-Porras J, Téllez Corral MA, Escobar Arregocés FM. Relationship of obstructive sleep apnea with periodontal condition and its local and systemic risk factors. Clin Oral Investig. 2023;27:2823–32. doi: 10.1007/s00784-023-04869-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Kim SR, Son M, Kim YR. Risk of chronic periodontitis in patients with obstructive sleep apnea in Korea: a nationwide retrospective cohort study. Epidemiol Health. 2023;45:e2023032. doi: 10.4178/epih.e2023032. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data presented in this study are available on request from the corresponding author.