Obstructive sleep apnea (OSA) is one of the most common respiratory disorders. It is characterized by apnea and hypopnea caused by repeated collapse of the upper airway during sleep. OSA can result in numerous medical sequelae, including cardiovascular and neurocognitive consequences [1].

Continuous positive airway pressure (CPAP) is the primary treatment for most patients with OSA owing to its higher efficacy, regardless of the OSA severity, compared with other treatment options, which range from conservative approaches to invasive interventions. Surgery is reserved for selected patients with anatomical limitations [2]. Although the use of mandibular advancement devices (MAD), an oral appliance (OA) that protrudes the mandible while keeping the airway open and improving pharyngeal collapsibility, is currently viewed as a second-line treatment option in the “one-size-fits-all” paradigm [2,3], awareness of their clinical benefits is increasing as a result of high patient acceptance, low adherence to CPAP, and the challenges associated with surgical therapy [4,5]. Based on a 2018 systemic review comparing the various treatment outcomes of CPAP and MAD, the superior efficacy of CPAP does not necessarily lead to better health outcomes in clinical settings [4].

The American Academy of Sleep Medicine currently recommends diagnosing and managing patients with OSA based on a single metric, the apnea–hypopnea index (AHI). In the present paradigm, the use of MAD is limited to patients with mild and moderate OSA and patients with severe OSA who refused to receive CPAP [2]. Previous studies and case reports suggested that MAD is effective in some patients with severe OSA and is the preferred treatment for patients with OSA of varying severities [6-9]. However, some patients may not respond favorably to this treatment [10]. A previous study demonstrated that the outcomes of MAD therapy are less predictable than those of CPAP [11]. The lower predictability of MAD therapy than CPAP may be attributed to the complex pathophysiology of OSA, including high loop gain, arousal threshold, and airway dilator muscle activity in addition to pharyngeal collapsibility. This is because while MAD improves pharyngeal collapsibility, it leaves other factors unchanged [12]. As a result, studies aimed at efficiently identifying patients who are likely to benefit from MAD therapy to improve its efficacy and personalized medicine have been increasing.

This study investigated the treatment outcomes of MAD therapy in patients with OSA of varying severities. In particular, we sought to identify the predictors of response to MAD therapy based on clinical and sleep parameters. It was hypothesized that the outcomes of MAD therapy vary depending on the OSA severity.

1. Subjects

Consecutive patients who complained of snoring and who were referred for OA therapy due to low adherence to CPAP during sleep were screened at the Department of Orofacial Pain and Oral Medicine, Dankook University Dental Hospital, from 2012 to 2020. Patients who completed OA titration after being diagnosed with OSA by a home sleep apnea test (HSAT) and patients diagnosed with mild OSA by laboratory polysomnography (lab-PSG) were included despite having AHI<5 in the HSAT.

The inclusion criteria for the subjects were an AHI of 5 or greater in the HSAT or lab-PSG, completion of the HSAT before and after MAD treatment, sufficient teeth to support MAD therapy in each arm, and no ongoing neurological disorders, jaw limitation, or pain due to temporomandibular disorders (TMD). The exclusion criterion was the lack of full titration due to noncompliance, TMD, or dental treatment during MAD therapy. Demographic data (age, sex, height, weight, and neck circumference) were collected as part of the initial clinical assessment during the first visit. Body mass index (BMI) was calculated from the patients’ weight and height. The study was approved by the Institutional Review Board of Dankook University Dental Hospital (IRB No. DKUDH IRB 2020-04-004) and conducted in accordance with the principles of the Declaration of Helsinki. Furthermore, informed consent was obtained from the patients.

2. Oral Appliance and Treatment Procedure

Two types of custom-made, two-piece, adjustable MAD were used in this study. One was an anteriorly adjustable MAD device (Dr.Prevent Co.) [13], and the other was SomnoDent (SomnoMed Limited, Australia). The bite registration at a comfortable protrusion, approximately 50%-60% of the maximum protrusion with a minimum of 3-mm interarch space, was fabricated as an initial amount of protrusion. The device was advanced 0.5 to 1 mm with each titration until subjective symptoms (self-reported or witnessed snoring, fatigue, daytime sleepiness, headache) resolved. For optimal titration, HSAT was conducted when self-reported symptoms improved. Titration was continued until maximum comfortable protrusion was achieved. The patients were followed up monthly for 6 months and, after confirmation of good compliance with the MAD, at 6 months and 1 year. The protrusion ratio was calculated as the ratio of the amount of protrusion to the maximum amount of protrusion.

3. Epworth Sleepiness Scale (ESS) for Subjective Assessment

The ESS, a self-administered questionnaire with eight questions, was administered at the time of both the pre- and post-treatment visits [14]. The higher the ESS score, the more severe the patient’s subjective sleepiness in daily life.

4. Home Sleep Apnea Test for Objective Assessment

Four sleep variables were selected and assessed at baseline and after the final titration of MAD therapy using an unattended HSAT device, ApneaLink (ResMed Ltd.): AHI, respiratory disturbance index (RDI), average oxygen saturation (average SpO₂), and minimum oxygen saturation (minimum SpO₂). OSA severity was evaluated using the AHI, and treatment outcomes were determined based on the success criteria that AHI should show a 50% improvement and be lower than 10.

5. Statistical Analysis

The sleep parameter-related data used for statistical analysis was based on HSAT data. Descriptive data were expressed as mean and standard deviation (SD) or median and interquartile range (IQR). The subjects were classified into three groups to compare the MAD therapy outcomes based on the OSA severity: mild (5≤AHI<15), moderate (15≤AHI<30), and severe (AHI≥30). The Wilcoxon signed-rank test was conducted to compare the sleep parameters and ESS scores of the three groups before and after MAD therapy.

The Kruskal–Wallis test was used to compare quantitative data derived from anthropometric data, sleep parameters, and ESS scores among the groups. Based on the success criteria of MAD therapy (reduction in AHI of more than 50% with an AHI of <10 events/h), patients with moderate and severe OSA were further classified into non-responders and responders. The Mann–Whitney U test was used to compare the treatment outcomes of the responders and non-responders. If there were significant factors resulting in different treatment outcomes for MAD therapy, Cohen’s d effect sizes were calculated to compare the responders and non-responders in terms of these factors. Statistical calculations were performed using IBM SPSS Statistics for Windows, Version 23.0 (IBM Co.). Probability levels of p<0.05 were considered significant.

Among the 47 patients who received MAD therapy, 11 refused to participate in the study, and 3 did not undergo HSAT following OA therapy. During follow-up, two patients did not visit the hospital and two patients discontinued MAD therapy while undergoing dental prosthesis replacement. Therefore, only 29 patients were included in the study, of whom 10 were treated with an anteriorly adjustable MAD device and 19 with SomnoDent. Subjective preferences between the two appliances have not been reported.

As presented in Table 1, the mean (SD) age of the patients was 41.1 (15.4) years, and majority of them were men (82.7%). For all patients, the mean (SD) BMI and neck circumference were 25.5 (2.9) and 39.4 (1.7), respectively. Of the 29 patients, 7 were diagnosed with severe OSA (AHI≥30) and 6 with moderate OSA (15≤AHI<30). Sixteen patients were diagnosed with mild OSA (AHI<15) based on HSAT or lab-PSG data. The median protrusion ratio of all patients was 77.8% (IQR, 75.0-83.3). Protrusion ratio (%) was not statistically different among three groups (p=0.252, Mann–Whitney U test).

Table 2 and Fig. 1 present the treatment outcomes, including four objective sleep parameters and a subjective measure of daytime sleepiness, both before and after MAD therapy for all patients and subgroups based on the OSA severity. Except for the average SpO₂, all sleep-related variables exhibited statistically significant improvements (p<0.001 for AHI, RDI, and ESS and p=0.004 for minimum SpO₂). In patients with severe OSA, subgroup analyses revealed that all sleep-related parameters exhibited statistically significant improvements after MAD therapy (p=0.018 for AHI from 49.0 to 9.0, p=0.018 for RDI from 53.0 to 13.0, p=0.024 for average SpO₂ from 91.0 to 95.0, p=0.018 for minimum SpO₂ from 79.0 to 87.0, p=0.017 for ESS from 11.0 to 7.0).

In patients with moderate OSA who received MAD therapy, significant differences were observed in sleep parameters (p=0.027 for AHI from 23.2 to 5.0, p=0.027 for RDI from 25.7 to 7.5, p=0.027 for minimum SpO₂ from 81.5 to 91.0, p=0.039 for ESS from 12.0 to 5.5), except for average SpO₂ (p=0.102). Compared with moderate and severe OSA cases, mild OSA cases did not exhibit significant differences in AHI and minimum SpO₂, but there were significant improvements in RDI (p=0.046) and ESS (p=0.006).

Table 3 presents a comparison of the clinical characteristics and outcomes of MAD therapy based on the OSA severity. The sex ratio and mean neck circumference did not differ among the three groups (p=0.276 and p=0.134, respectively). However, patients with mild OSA had a comparatively lower mean age than those with moderate and severe OSA (p=0.004). Patients with severe OSA had the highest mean BMI among the three groups, but the difference was not statistically significant (p=0.073). Before MAD therapy, the AHI (p<0.001), RDI (p<0.001), and ESS (p=0.006) increased in the order of mild, moderate, and severe OSA. The average (p=0.038) and minimum SpO₂ (p=0.001) levels decreased in the order of mild, moderate, and severe OSA. The severe OSA group had the highest decrease in the AHI and RDI and increase in the average and minimum SpO₂ levels among the groups; it also had a median (IQR) residual AHI after MAD therapy of 9 (3.0-15.0), which was higher than those of mild and moderate OSA groups (p=0.045). The severe OSA group had a median (IQR) residual minimum SpO₂ of 87.0 (84.0-89.0), which was not significantly lower than those of the mild and moderate OSA groups (p=0.092). No significant difference was observed in the median ESS scores of the three groups (p=0.618) after MAD therapy.

Of 13 patients with moderate and severe OSA, 10 and 3 were classified as responders and non-responders, respectively, as presented in Table 4 and Fig. 2. No statistical difference was observed in the protrusion ratio (%) between the responders and non-responders (p=0.342, Mann–Whitney U test). Furthermore, there were no significant differences between the two groups as regards sex ratio (p=0.400), age (p=0.811), BMI (p=0.217), neck circumference (p=0.217), and ESS (p=0.811). However, the mean baseline values of the average (p=0.049) and minimum (p=0.007) SpO₂ were found to be significantly lower in non-responders than in responders. The mean (SD) AHI of the non-responders was 50.0 (13.1), which was significantly higher (p=0.049) than that of the responders, which was 32.2 (12.0). The mean RDI followed a similar trend to the AHI, although the difference was not statistically significant (p=0.077). Cohen’s d for AHI, average SpO₂, and minimum SpO₂ were 4.0613 (95% confidence interval [CI]=2.0361, 6.0866), 1.6504 (95% CI=0.2126, 3.0881), and 1.4605 (95% CI=0.0534, 2.8675), respectively, for the responders and non-responders.

The clinical practice guideline, issued jointly by the American Academy of Sleep Medicine and American Academy of Dental Sleep Medicine, recommends considering OA therapy for adult patients with primary snoring without OSA and those with OSA who are intolerant to CPAP or prefer alternative therapy [15]. Although CPAP is more effective than MAD in reducing AHI, there is increasing evidence that patients have better tolerance to MAD than to CPAP, with a self-reported compliance rate of around 80% [16]. Therefore, current evidence suggests that both OA and CPAP are equally important in the treatment of patients with mild to moderate OSA and even severe OSA. However, OA is suggested to be a potential first-line option despite the higher efficacy of CPAP in reducing AHI [16,17]. According to the clinical practice guideline published in 2015 [15], OA reduces AHI, RDI, and daytime sleepiness and modestly improves minimum SpO₂ in adult patients with OSA with the moderate level of evidence. Consistent with the findings of previous studies [15,18], significant improvements were observed in patients with mild OSA after MAD therapy, including reductions in RDI (p=0.046) and ESS (p=0.006), as presented in Table 2 and Fig. 1. Interestingly, 1 of 16 patients with mild OSA exhibited an increase in AHI (AHI=27 events/h) after titration of MAD than at baseline (AHI=12 events/h). This patient was 50 years old, had obesity (BMI=28.7 kg/m²), and had no medical history. His weight did not change during the follow-up period. This outlier case suggests that a diagnosis of mild OSA based on AHI does not necessarily ensure the efficacy of MAD therapy.

In patients with moderate OSA, significant reductions were observed in AHI (p=0.027), RDI (p=0.027), and ESS (p=0.006), alongside significant increases in minimum SpO₂ (p=0.039). Notably, patients with severe OSA reported significant improvements in all sleep-related parameters, including significant decreases in AHI (p=0.018) and RDI (p=0.018), significant increases in average SpO₂ (p=0.024) and minimum SpO₂ (p=0.018), and a significant decrease in ESS (p=0.017).

Comparison of the clinical characteristics and treatment outcomes of OA therapy based on the OSA severity revealed a positive correlation between OSA severity and patient age (p=0.004) (Table 3). This finding is consistent with that of a previous study that examined polygraphic data in 1090 patients with OSA and indicated that the older the patient, the higher the risk for pharyngeal collapsibility [19]. The median BMI of the patients in the severe OSA group was higher than those in the mild and moderate OSA groups, but the difference was not statistically significant (p=0.073). No differences were observed in the sex ratio or neck circumference between the groups. As expected, the differences in sleep-related parameters between pre- and post-MAD therapy followed the order of severe, moderate, and mild OSA groups. The present study showed successful outcomes in 83.3% and 71.4% of moderate and severe OSA cases, respectively, based on the previously stated success criteria (significant reduction in AHI of more than 50% and AHI less than 10 events/h).

Ramar’s study reported that the efficacy of OA is limited to 60%-70% of patients due to the interindividual variability of the treatment outcomes [15]. Identification of patients who are likely to benefit from OA treatment is currently a challenge for clinicians in this context. To examine the relevant features that distinguish responders from non-responders to MAD therapy, a subgroup analysis was conducted between them based on the AHI criteria. Non-responders had a higher AHI (p=0.049) and lower average and minimum SpO₂ (p=0.049 and p=0.007, respectively) than responders (Table 4 and Fig. 2). The results suggest that higher AHI and lower nocturnal SpO₂ are associated with less effective therapy for MAD. Considering the high efficacy of MAD (71.4%) in the group with severe OSA categorized by AHI≥30, a high AHI score does not ensure the efficacy of MAD therapy. Cohen’s d effect size indicated that the minimum SpO₂ had a greater predictive power than the AHI and average SpO₂ in determining responders and non-responders. Thus, these results suggest that the minimum SpO₂, which is considered a secondary measure, has a stronger predictive power than AHI, which is regarded as a primary measure, and average SpO₂, which is another oximetric measure, given the limited sample size of this study.

Although the AHI is still the most commonly used diagnostic metric for OSA, its predictability in clinical practice may not be adequate owing to its oversimplified single index, requiring a critical evaluation of its role as a primary biomarker for the diagnosis and management of OSA [20-25]. Previous studies have reported that OSA severity, graded by the AHI score, was not correlated with physiological consequences (e.g., cardiovascular events and related morbidity or mortality), CPAP titration pressure, and daytime sleepiness [20-22].

In this study, Spearman’s rho calculated via correlation analysis was found to be high (p=−0.666, p<0.001) between the AHI and minimum SpO₂.

In fact, the AHI and minimum SpO₂ are closely related but are not the same parameter. The AHI is a quantifiable measure of the frequency of apnea–hypopnea events, whereas the minimum SpO₂ is a qualitative biomarker that indicates physiological consequences associated with AHI events. The AHI system has several inherent shortcomings. Hypoxic and nonhypoxic events, which are associated with different health outcomes, are not distinguished by the AHI [26]. It does not account for the duration of apnea and hypopnea as well as body position during these two events [23,27-29]. Furthermore, it gives equal weight to apnea and hypopnea, whereas apnea might have greater physiological consequences than hypopnea through severe hypoxemia and increased autonomic responses [23].

Prior research suggested possible prognostic markers as predictors of MAD efficacy [10,30]. Direct visual verification of pharyngeal widening via videoendoscopy has been suggested as a potential predictor of good efficacy of MAD [30]. A 2022 systemic review examined phenotypic differences between responders and non-responders to MAD therapy and suggested that anatomic characteristics, sleep study findings, and high CPAP pressure are associated with low MAD efficacy [10]. Consistent with the findings of our study, a 2022 systemic review found that a high minimum SpO₂ during sleep can serve as a biomarker for the prediction of responders to MAD therapy [10].

Based on the results of the present study, consistent with those of previous ones, it seems adequate to consider the minimum SpO₂ level as a biomarker of nocturnal hypoxemia in addition to the AHI as a quantifiable respiratory cessation parameter for predicting MAD efficacy for the treatment of OSA. This can facilitate the screening and selection of patients for OA therapy. Although a statistically significant difference was observed in the mean (SD) between non-responders (71.3%, SD=1.1) and responders (81.4%, SD=2.7) in the present study, a clear-cut minimum SpO₂ level that can predict the efficacy of OA treatment is uncertain. Interestingly, a recent study suggested the usefulness of minimum SpO₂ as another success criterion of AHI in the treatment of OSA with MAD [31]. This study reported that there were three different phenotypes in terms of treatment outcomes for MAD, including a group with improvement in AHI only, other groups with improvement in minimum SpO₂ only, greater than 4% SpO₂, and a group with congruent improvement. The cutoff value of pretreatment minimum SpO₂ was 86.25% with a positive predictive value of 89.47%.

The success criteria to divide responders and non-responders to MAD discussed in this study were based on the AHI. The mean (SD) improvement of the minimum SpO₂ in the three non-responders in this study was 15.3% (4.72). Based on the AHI, all the non-responders showed improvement of the minimum SpO₂ greater than 4%. These results suggest that MAD treatment is beneficial for responders based on the minimum SpO₂, even if they are non-responders based on AHI. Given the detrimental effects of oxygen desaturation during sleep on long-term health outcomes [32], defining the success criterion for MAD treatment should not focus on a single metric of AHI. This is an important issue, and further investigation using a large sample size would provide diagnostic and prognostic information for OSA treatment with MAD therapy.

Future studies should explore various parameters associated with hypoxic burden beyond the minimum SpO₂. Several sleep parameters, such as arousal intensity, oxygen desaturation rate, heart rate variability, and apnea–hypopnea event duration, have been proposed as physiological surrogates beyond AHI [33-37]. Polysomnographic endotypes, such as ventilatory instability and high loop gain of non-responders to MAD therapy, were observed in a previous clinical trial [38]. Identifying and understanding the relevant factors would facilitate the development of individualized MAD treatment approaches for OSA. Knowledge of the underlying mechanism of MAD therapy needs to be continuously updated.

As a preliminary study, the main limitation of the current study with limited sample size should be mentioned. The lack of statistical power due to the small sample size of the non-responders would lead to overestimation of significance and false-positive results. Thus, the results of this study should be interpreted with caution, although previous studies have reported results similar to ours [10]. To strengthen our findings, further validation studies using a larger sample size of nonresponders are warranted. The possible effect on treatment outcomes related to the heterogeneity of two different types of MAD should also be addressed. The present study did not compare the treatment outcomes between the two appliances. However, considering the lack of a robust effect of OA design on efficacy of MAD therapy [39], the possible effects of different MAD designs would be minimal. Despite the drawbacks of this study, which was based on a small dataset, the findings encourage further research that considers various sleep-related metrics other than AHI in predicting the efficacy of MAD.

In conclusion, the results of the present study suggest that the MAD is effective in majority of patients with OSA of varying severities. The success of MAD therapy does not seem to depend solely on AHI severity. In addition to AHI, minimum SpO₂ level may serve as a prognostic marker for the efficacy of MAD treatment in clinical dental practice.

Hye Kyoung Kim, Editor-in-Chief of the Journal of Oral Medicine and Pain, was not involved in the editorial evaluation or decision to publish this article. Mee Eun Kim serves on the editorial board of the Journal of Oral Medicine and pain and had no role in the editorial evaluation or decision to publish this article. Except for this, no other potential conflicts of interest relevant to this article have been reported.

The datasets used in the current study are available from the corresponding author upon reasonable request.

None.

Conceptualization: HKK, MEK. Investigation: HKK. Methodology: HKK. Supervision: HKK, MEK. Data analysis and interpretation: HKK, MEK. Writing of original draft: HKK. Writing, review-editing: HKK, MEK.