https://orcid.org/0000-0002-4136-8069
Abstract
To characterize how mandibular advancement enlarges the upper airway via posterior tongue advancement in people with obstructive sleep apnea (OSA) and whether this is associated with mandibular advancement splint (MAS) treatment outcome.
One-hundred and one untreated people with OSA underwent a 3T magnetic resonance (MRI) scan. Dynamic mid-sagittal posterior tongue and mandible movements during passive jaw advancement were measured with tagged MRI. Upper airway cross-sectional areas were measured with the mandible in a neutral position and advanced to 70% of maximum advancement. Treatment outcome was determined after a minimum of 9 weeks of therapy.
Seventy-one participants completed the study: 33 were responders (AHI<5 or AHI≤10 events/hr with >50% AHI reduction), 11 were partial responders (>50% AHI reduction but AHI>10 events/hr), and 27 nonresponders (AHI reduction<50% and AHI≥10 events/hr). Responders had the greatest naso- and oropharyngeal tongue anterior movement (0.40 ± 0.08 and 0.47 ± 0.13 mm, respectively) and oropharyngeal cross-sectional area enlargement (6.41 ± 2.12%) per millimeter of mandibular advancement. A multivariate model that included tongue movement and percentage of airway enlargement per millimeter of mandibular advancement along with baseline AHI correctly classified 69.2% (5-fold cross-validated 62.5%, n = 39) of participants in response categories when the jaw was advanced in the range that would usually be regarded as sufficient for clinical efficacy (> 4 mm). In comparison, a model using only baseline AHI correctly classified 50.0% of patients (5-fold cross-validated 52.5%, n = 40).
Tongue advancement and upper airway enlargement with mandibular advancement in conjunction with baseline AHI improve treatment response categorization to a satisfactory level (69.2%, 5-fold cross-validated 62.5%).
Prediction of mandibular advancement splint (MAS) treatment outcome for people with obstructive sleep apnea (OSA) is currently unreliable. This study shows that tongue advancement and oropharyngeal enlargement were largest with mandibular advancement for responders, which may help improve MAS response prediction. When the mandible was advanced at least by 4 mm the inclusion of tongue movement and the percentage of upper airway cross-sectional area enlargement per millimeter of mandibular advancement with baseline AHI in a multivariate model correctly classified more patients (69.2%, 5-fold cross-validated 62.5%, n = 39) than a model based only on baseline AHI (50.0%, 5-fold cross-validated 52.5%, n = 40). Dynamic magnetic resonance imaging measures improved MAS treatment prediction to an overall satisfactory level.
Introduction
Obstructive sleep apnea (OSA) is an increasingly common sleep disorder [1,2]. A narrow airway, impaired muscle responsiveness, low arousal threshold, and unstable ventilatory control contribute to disease severity [3,4]. Continuous positive airway pressure is the most efficacious treatment for OSA [5]. However, compliance and acceptance rates can be low [6]. This impedes treatment effectiveness in the real world [7]. Mandibular advancement splint (MAS) therapy is a common second-line treatment [8]. It consists of an oral appliance device worn during sleep to hold the lower jaw in a protruded position to widen the upper airway [9,10], stabilize the pharynx and decrease airway collapsibility [11,12]. Depending on the definition, treatment success rates vary from 36.5% (apnea-hypopnea index (AHI) < 5 events/hr), 52.2% (AHI < 10 events/hr + ≥ 50% AHI reduction), to 63.8% (≥ 50% of AHI reduction) in people with OSA [13,14]. Nonetheless, similar health outcomes to CPAP can be achieved [15], thanks to better adherence with MAS therapy [16].
However, prediction of MAS treatment outcome is unreliable [17] and leads to treatment delays for nonresponders. These delays can be particularly detrimental in patients with moderate to severe OSA [18]. Baseline characteristics such as lower baseline AHI, younger age, female gender, smaller neck circumference, and lower body mass index (BMI) have been associated with better MAS therapy outcomes [19]. However, these variables are poor predictors of MAS treatment outcome [13]. Therefore, the utility of a wide range of more complex methodologies and measurements than baseline clinical characteristics to predict MAS treatment outcomes has also been assessed [20,21]. In particular, it has been reported that patients are less likely to respond to MAS treatment if they have unstable ventilatory control [22], or if they have a pterygomandibular raphe [23], a tendinous connection between the buccinator and superior pharyngeal constrictor. Good predictive accuracy was achieved using measurements collected during an overnight sleep study with multisensory catheters to identify the primary oropharyngeal closure site [24], or with a remotely controlled mandibular device to progressively protrude the bottom jaw until complete elimination of respiratory events [25]. Another study has shown that responders have a larger increase in the nasopharynx cross-sectional area (behind the soft palate) with mandibular advancement than nonresponders [26], using nasopharyngoscopy in awake patients. A model that includes the key endotype traits that contribute to OSA pathogenesis can also provide predictive insight [27]. However, none of the mentioned studies has achieved an excellent prediction level. This is due, at least in part, to incomplete understanding of the variable ways in which MAS therapy changes upper airway structures and function.
Airway enlargement is partially related to the anterior tongue movement, as measured using a tagged magnetic resonance imaging (MRI) technique, that occurs when the mandible is moved forward [10]. This is because the tongue either moved forward to enlarge the upper airway or elongated with minimal impact on the airway size. In addition, elongation of the top section of the tongue at the nasopharyngeal level was positively associated with OSA severity and the subjects with lower OSA severity experienced a greater upper airway enlargement with mandibular advancement than subjects with higher OSA severity [26]. Since lower OSA severity and larger upper airway enlargement are associated with an increased likelihood of positive MAS treatment outcome, how much the airway enlarges and the tongue advances with a mandibular protrusion, as measured by tagged MRI, could offer new mechanistic insight and ultimately improve treatment prediction.
Thus, this study firstly aimed to examine how the upper airway enlarges via posterior tongue advancement using tagged MRI during mandibular advancement in patients with untreated OSA who are candidates for MAS therapy. Second, we aimed to determine how this is related to the MAS treatment outcome. Third, we aimed to assess whether this new combined approach would better predict MAS treatment outcomes than baseline characteristics alone. We hypothesized that larger mandibular protrusion would be associated with greater posterior advancement of the tongue and enlargement of the upper airway, with larger anterior movements of the tongue and upper airway enlargement observed for responders than partial- and nonresponders. Thus, we expected that the amplitude of the tongue advancement and the upper airway enlargement evoked by mandibular advancement would be significant predictors of MAS treatment outcome groups.
Methods
Participants
The study was approved by the South Eastern Sydney Local Health District Human Research Ethics Committee (HREC/14IPOWH/699) and complied with the Declaration of Helsinki (2013), except for registration in a publicly accessible database (clause 35). One hundred and one participants with untreated OSA were recruited from local sleep clinics (76 men, 19–76 years old, body mass index (BMI) 18–51 kg/m2, and neck circumference 31–50 cm). Participants were recommended for MAS treatment by their treating physician after continuous positive airway pressure therapy failure for patients with severe OSA or because it was the preferred treatment option for patients with mild to moderate OSA. Written informed consent was obtained from all participants. Exclusion criteria included failure to meet MRI safety clearance, major comorbid chronic disease (including neurological and respiratory illnesses), use of medications that alter respiratory muscle function or sleep, prior upper airway surgeries or current use of any OSA treatment, and an apnea hypopnea index (AHI) <10 events/hr. Each participant was required to have a baseline polysomnography (PSG) within the two years preceding recruitment to determine baseline OSA severity and not report weight changes since their sleep study. Untreated OSA severity was considered mild if the AHI was > 10 and ≤ 15 events/hr (since only patients with an AHI >10 events/hr were recruited), moderate if the AHI was > 15 and ≤ 30 events/hr, and severe if the AHI was strictly higher than 30 events/hr.
Experimental protocol and MAS therapy
After recruitment, participants underwent an MRI scan before commencing their therapy with a custom-made MAS device (SomnoDent Flex, SomnoMed Pty Ltd, Australia) [18,28]. Participants were asked to wear the device nightly. Once the comfortable maximum jaw protrusion with the splint was reached, the MAS was worn for at least another week before a final in-lab overnight PSG was performed to determine treatment outcomes. At least eight weeks were required between the start of the therapy and the final PSG. PSGs were scored following the American Academy of Sleep Medicine manual (v2.4) [29] across three separate sites (i.e. independent private scorer, Neuroscience Research Australia and Royal North Shore Hospital, Australia). Intraclass correlation coefficients (ICC) were computed to assess the reproducibility of the PSG scoring. Good to high concordance PSG scoring was achieved between the independent scorer with those from the Royal North Shore Hospital (n = 10, ICC = 0.73), and from Neuroscience Research Australia (n = 5, ICC = 0.90). There was also excellent concordance between scoring from Neuroscience Research Australia and Royal North Shore Hospital (n = 13, ICC = 0.99). Where possible, baseline and final PSGs were scored by the same individual.
MAS treatment response was defined by the participant’s AHI measured in the follow-up sleep study with the MAS in situ and the change in AHI from the baseline to final PSGs. Responders had a final AHI of less than or equal to ten events/hr with an AHI reduction of at least 50% or a final AHI ≤ 5 events/hr. Partial-responders had an AHI reduction of at least 50%, with a final AHI > 10 events/hr. Finally, nonresponders achieved an AHI reduction of < 50%. These definitions were chosen because they are commonly used in the literature, and therefore would be beneficial for evaluating our work against previous studies. There is no consensus on treatment success definitions [21].
MR acquisition
Upper airway MR imaging was performed using a clinical 3T scanner (Achieva TX, Philips Medical Systems, Best, The Netherlands) with a 16 channel neurovascular head coil. Participants lay supine and wore an MRI compatible splint for the entire scan (Apnea Guard, Advanced Brain Monitoring, Carlsbad, CA). The custom-made SomnoMed MAS device was not able to be used for scanning due to a metal component. Head position was standardized by aligning the Frankfurt plane perpendicular to the examination table, and the confined space within the head coil prevented forward head movement, as head position can alter upper airway size [30]. A respiratory sensor was also placed around the abdomen to monitor and record breathing cycles. For this study, we used the same MRI scanner and imaging protocol that Brown et al.[10] used to identify the various tongue deformation patterns occurring during mandibular advancement.
First, axial plane anatomical images (Figure 1A) were collected using turbo spin-echo (TSE) imaging in two mandible positions; an “advanced” position set at 70% of the participant’s maximum advancement and a “neutral” position without jaw advancement. Seventy percent of the maximal protrusion was used for the “advanced” position as this is the degree of advancement commonly used at the start of a dentist prescribed treatment and is often an effective position for treatment response. It is also more comfortable to maintain during the scan time than the maximal protrusion. Participants were instructed to keep their mouth closed around the splint and to breathe through their nose. For both axial acquisitions, the parameters were: repetition time/echo time = 4965/ 80 ms; field of view = 192 × 192 mm2, in-plane spatial resolution 0.50 × 0.50 mm2; 50 continuous with a slice thickness of 3 mm; scan duration 4:21 min.
Examples of typical MR images collected. (A) Outline of the upper airway cross-sectional area in white in an axial anatomical image. (B) Outline of the tongue in blue (excluding the geniohyoid) and mandible in orange in the midsagittal plane of a tagged MRI scan. (C) Placement of 5–8 tracking points at the posterior of the tongue and 3 behind the mandible (orange) to track movement during passive jaw advancement. For the tongue, points were separated into nasopharyngeal points (above the base of the soft palate, pink) and oropharyngeal points (green).
Second, mid-sagittal posterior tongue movement during passive mandible advancement was obtained with tagged MRI using an existing modified two-dimensional, complementary spatial modulation of magnetization (CSPAMM) technique (Figure 1B). Passive, and not active, mandible advancement was conducted since the primary mode of action of MAS therapy is via improvement in passive pharyngeal anatomy. Subjects were instructed to hold their breath at a comfortable end-expiratory level while images were acquired. The images were taken at 250 ms intervals over 4 seconds to cover the 1 or 2 seconds required for the mandibular advancement. The passive mandible advancement was achieved by pulling a nylon wire attached to the bottom tray of the temporary MAS, as described previously by Brown et al.[10]. Several practice runs were performed until reproducible movements were obtained. The tagged MRI scans were repeated at least five times. Imaging parameters were: repetition time/echo time = 2.2/ 0.9 ms; field of view = 220 × 220 mm2; in-plane spatial resolution 0.86 × 0.86 mm2; 1 slice of 10 mm thickness; tag spacing 7.21mm; scan duration 4 seconds.
MRI analysis
All data were analyzed blinded to OSA severity and treatment outcome.
Axial anatomical scans.
Two cross-sectional areas of the upper airway were measured from scans with the mandible in both the “neutral” and “advanced” position. The first area was in the nasopharyngeal region, at the narrowest level behind the soft palate. The second was in the oropharyngeal region, just rostral to the tip of the epiglottis. Measurements were obtained by manually outlining the airway lumen using ImageJ (v1.51n NIH, Bethesda, Maryland, USA) and represent averages over the whole respiratory cycle (Figure 1A). The percentage of upper airway size changes with mandible advancement was calculated for both pharyngeal locations.
Tagged MRI.
The dynamic antero-posterior movements of the posterior tongue and mandible during jaw advancement were quantified using harmonic phase methods [31] which has a displacement error of 0.1 pixels [32], implemented in Matlab (v2013b, The MathWorks Inc, MA, USA) (Figure 1B). Movements were measured perpendicular to the participant’s rear pharynx wall. For the tongue, 5 to 8 tracking points (depending on the participant’s anatomy) were manually placed on the posterior section of the tissue, approximately 10 mm anterior to the airway (Figure 1C). The first point was placed halfway down the soft palate and was repeated on each grid line until the tip of the epiglottis (Figure 1C). The points lying anterior to the soft palate were averaged to reflect nasopharyngeal tongue movement. The remaining posterior points (from below the soft palate to the tip of the epiglottis) were averaged to reflect oropharyngeal tongue movement. For the mandible, 3 points were placed in a vertical line just behind the mandible bone (Figure 1C) and averaged to quantify passive mandibular advancement. Positive movements represent anterior movement, and negative displacements represent posterior movements. Three tagged MRI scans were analyzed for each participant. Good consistency was observed between the three tagged MRI scans selected for the nasopharyngeal and oropharyngeal tongue movement and mandibular advancement (n = 67, ICC = 0.785, ICC = 0.81, and ICC = 0.810, respectively).
Statistical analysis
P values less than 0.05 were considered significant. Statistical analysis was performed using SPSS (V24, IBM, Armonk, New York, USA) and GraphPad Prism (V9, GraphPad Software, LLC, San Diego, California, USA). Data are reported as mean ± standard deviation unless stated otherwise. For each participant, three mandibular advancement measurements were obtained: 1—70% of the maximum advancement obtainable at the time of the MRI scan, the Apnea Guard was set at this for the “advanced” anatomical scans; 2—the maximum advancement obtained during the dynamic tagging scans; and 3—the advancement setting of the therapeutic device (SomnoDent Flex) at the final overnight study. These three measurements were compared using a repeated two way ANOVA followed by Tukey’s multiple comparisons test.
Age, BMI, neck circumference, and baseline AHI were compared between participants who completed the study and those who did not using an independent t-test (equal variance not assumed). Baseline characteristics were also compared between MAS treatment response categories using one-way ANOVA followed by Sidak’s multiple comparisons and gender proportion with Fisher’s exact test. Correlations between mandibular advancement, tongue movements, and percentage of changes in cross-sectional area of the pharynx were assessed using partial Pearson correlation to control for age, BMI, gender, and account for multiple movement measurements within the same participants (up to 3 tagged MRI/ participant). Significant correlations were fitted with linear regression to determine how tongue movement and upper airway enlargement varied with mandibular advancement.
Finally, a multinomial logistic regression was performed to assess the utility for MAS treatment response prediction based on the baseline AHI. All participants who achieved mandibular advancement in the range that would usually be regarded as sufficient for clinical efficacy (> 4 mm) were included. Baseline AHI is commonly used to assist treatment prediction [13]. It was chosen as a reference to determine whether our dynamic imaging measures can improve MAS treatment prediction. For this, the model using only baseline AHI was compared to a second model, including baseline AHI and the naso- and oropharyngeal posterior tongue movements/ mm of mandibular advancement and the percentages of change in upper airway cross-sectional area/mm of mandibular advancement. Average dynamic imaging measures were used for each participant. Two log-likelihood ratios were calculated to identify predictors in both models. Parameter estimates are reported as odds ratio with 95% confidence intervals. Robustness of the two models was evaluated using a 5-fold cross-validation. This approach allows testing the ability of the model to predict new MAS responses that were not used in estimating it and assess how well the model would perform in an independent population. Model comparisons across all four levels of mandibular advancement measured (> 1mm to > 4 mm) are reported in the supplementary file. Predictive values of two additional models were also investigated and reported in the supplementary file: the first included 5 baseline characteristics (age, BMI, gender, baseline AHI, and neck circumference), and the second used these 5 baseline characteristics in conjunction with the 4 dynamic imaging measures (i.e. 9 predictors).
This study was designed to conduct a multiple linear regression analysis, with alpha = 0.05, power = 0.9, 3 tested predictors (age, BMI, and gender), for distinguishing the primary outcome (posterior tongue movement) from 0 with an effect size f2 of at least 0.20 [10], using the GPower software. The required sample size calculated was n = 76 (i.e. 102 with 35 % drop-out allowed [33]).
Results
Participant characteristics and clinical outcomes
Posterior tongue and mandible movements during jaw advancement were obtained from 209 tagged MRI scans from 71 participants (71/101, 70%) for whom MAS treatment outcomes were determined. Three were unable to complete the MRI due to claustrophobia and/or nausea, dynamic mandibular advancement was not achieved for 9 participants during the scan, 8 could not tolerate the MAS for treatment, 9 were lost to follow up prior to their final PSG, and 1 withdrew from the study for personal reasons. There were no differences in age, baseline AHI, neck circumference or gender between those who completed the study and those who did not (independent t-test. p > .05), but those who withdrew had a greater BMI (31.6 ± 6.5 vs 28.4 ± 4.6 kg/m2, p = .02). Among these 71 participants, the cohort was largely middle-aged (44.8 ± 11.3 years), overweight (28.4 ± 4.6 kg/m2), and mostly male (19 female). At baseline, 13 subjects had mild OSA, 27 had moderate OSA, and 31 had severe OSA. Thirty-three participants completely responded to MAS (33/71, 46.5%), 11 were partial responders (11/71, 15.5%), and 27 were nonresponders (27/71, 38.0%) (Table 1).
Baseline characteristics for each MAS treatment outcome category
| Responders n = 33 | Partial-responders n = 11 | Nonresponders n = 27 | P values | |
|---|---|---|---|---|
| Number (M:F) | 20:13 | 8:3 | 24:3 | .043† |
| Baseline AHI (events/hr) | 29.8 ± 4.9 | 48.7 ± 18.4 | 27.8 ± 15.6 | <.001‡ |
| Age (years) | 44 ± 13 | 51 ± 9 | 43 ± 10 | .064‡ |
| BMI (kg/m2) | 27.9 ± 4.7 | 29.9 ± 5.2 | 28.5 ± 4.3 | .030‡ |
| Neck circumference (cm) | 38.7 ± 3.3 | 40.4 ± 2.4 | 41.0 ± 3.4 | <.001‡ |
| Responders n = 33 | Partial-responders n = 11 | Nonresponders n = 27 | P values | |
|---|---|---|---|---|
| Number (M:F) | 20:13 | 8:3 | 24:3 | .043† |
| Baseline AHI (events/hr) | 29.8 ± 4.9 | 48.7 ± 18.4 | 27.8 ± 15.6 | <.001‡ |
| Age (years) | 44 ± 13 | 51 ± 9 | 43 ± 10 | .064‡ |
| BMI (kg/m2) | 27.9 ± 4.7 | 29.9 ± 5.2 | 28.5 ± 4.3 | .030‡ |
| Neck circumference (cm) | 38.7 ± 3.3 | 40.4 ± 2.4 | 41.0 ± 3.4 | <.001‡ |
Thirty-three participants completely responded to MAS, 11 were partial responders, and 27 were nonresponders. The proportion of female subjects was the highest in the responder group (39% (13/33) vs 27% (3/11) for partial-responders and 11% (3/27) for nonresponders. Partial-responders had a higher baseline AHI than responders and nonresponders (Sidak, p < .001 for both) and a higher BMI than responders (Sidak, p = .044). Responders had a smaller neck circumference than nonresponders (Sidak, p < .001). Finally, there was no difference in age between treatment outcome groups (p = .064). Continuous variables are presented as mean ± standard deviation. Statistical differences between MAS response categories were assessed using one-way ANOVA (
‡) followed by Sidak’s multiple comparisons or Fisher’s exact test (
†). Abbreviations: (M) male, (F) female, (AHI) Apnea-Hypopnea Index.
Baseline characteristics for each MAS treatment outcome category
| Responders n = 33 | Partial-responders n = 11 | Nonresponders n = 27 | P values | |
|---|---|---|---|---|
| Number (M:F) | 20:13 | 8:3 | 24:3 | .043† |
| Baseline AHI (events/hr) | 29.8 ± 4.9 | 48.7 ± 18.4 | 27.8 ± 15.6 | <.001‡ |
| Age (years) | 44 ± 13 | 51 ± 9 | 43 ± 10 | .064‡ |
| BMI (kg/m2) | 27.9 ± 4.7 | 29.9 ± 5.2 | 28.5 ± 4.3 | .030‡ |
| Neck circumference (cm) | 38.7 ± 3.3 | 40.4 ± 2.4 | 41.0 ± 3.4 | <.001‡ |
| Responders n = 33 | Partial-responders n = 11 | Nonresponders n = 27 | P values | |
|---|---|---|---|---|
| Number (M:F) | 20:13 | 8:3 | 24:3 | .043† |
| Baseline AHI (events/hr) | 29.8 ± 4.9 | 48.7 ± 18.4 | 27.8 ± 15.6 | <.001‡ |
| Age (years) | 44 ± 13 | 51 ± 9 | 43 ± 10 | .064‡ |
| BMI (kg/m2) | 27.9 ± 4.7 | 29.9 ± 5.2 | 28.5 ± 4.3 | .030‡ |
| Neck circumference (cm) | 38.7 ± 3.3 | 40.4 ± 2.4 | 41.0 ± 3.4 | <.001‡ |
Thirty-three participants completely responded to MAS, 11 were partial responders, and 27 were nonresponders. The proportion of female subjects was the highest in the responder group (39% (13/33) vs 27% (3/11) for partial-responders and 11% (3/27) for nonresponders. Partial-responders had a higher baseline AHI than responders and nonresponders (Sidak, p < .001 for both) and a higher BMI than responders (Sidak, p = .044). Responders had a smaller neck circumference than nonresponders (Sidak, p < .001). Finally, there was no difference in age between treatment outcome groups (p = .064). Continuous variables are presented as mean ± standard deviation. Statistical differences between MAS response categories were assessed using one-way ANOVA (
‡) followed by Sidak’s multiple comparisons or Fisher’s exact test (
†). Abbreviations: (M) male, (F) female, (AHI) Apnea-Hypopnea Index.
Mandibular advancement
Mandibular advancement achieved during the tagged MRI (4.5 ± 1.7 mm, n = 71) was not different to the 70% setting used for the static imaging in the advanced mandible position (4.2 ± 1.3 mm, n = 71, Figure 2, Tukey’s comparison, p = .20). However, they were both significantly lower than the mandible advancement recorded during the final overnight study (10.4 ± 1.9 mm, n = 71, Tukey’s comparison, p < .0001 for both) likely because participants continued to adapt over time to the mandibular advancement device.
Box and whiskers (minimum to maximum) of the mandibular advancement used for the final overnight study (named final dental) with the SomnoDent Flex splint, measured using tagged MRI during jaw advancement with the Apnea Guard device and set at 70% of the maximal protrusion to collect advanced MRI anatomical scan also with the Apnea Guard device. Mandibular advancement achieved at the beginning of the study during the MRI scan was less than that recorded at the follow-up polysomnography study.
Mandibular advancement, posterior tongue advancement, and upper airway enlargement
Overall, a larger mandibular advancement was related to the larger anterior movement of the nasopharyngeal and oropharyngeal regions of the tongue (Figure 3A and B). However, during mandible advancement, the tongue moved forward primarily in the oropharyngeal region (mean ± standard error, 0.50 ± 0.08 mm/mm of mandible advancement vs 0.25 ± 0.05 mm/mm of mandible advancement for the nasopharyngeal section) (Table 2). This resulted only in an enlargement of the cross-sectional area of the oropharynx (mean ± standard error, 4.02 ± 1.37 %/mm of mandible advancement). Greater mandibular advancement was weakly associated with a larger increase in the cross-sectional area of the oropharynx (partial Pearson, r = 0.231, p = .001) (Figure 3C).
Amplitude of forward tongue movement and enlargement of the cross-sectional area of the upper airway for each millimetre of mandible advancement for each MAS treatment response group
| Tongue displacements per mm of mandibular advancement (partial Pearson P value) | Overall (n = 209 tagged MRI) | Responders (n = 98 tagged MRI) | Partial-responders (n = 31 tagged MRI) | Nonresponders (n = 80 tagged MRI) |
|---|---|---|---|---|
| Oropharyngeal tongue anterior movement | 0.50 ± 0.08 mm (r = 0.397, p < .001) | 0.47 ± 0.13 mm (r = 0.357, p < .01) | 0.60 ± 0.15 mm (r = 0.62, p = .01) | 0.58 ± 0.13 mm (r = 0.259, p = .027) |
| Nasopharyngeal tongue anterior movement | 0.25 ± 0.05 mm (r = 0.299, p < .001) | 0.40 ± 0.08 mm (r = 0.508, p < .01) | - (p = .38) | 0.26 ± 0.10 mm (r = 0.244, p < .01) |
| Change in oropharynx cross-sectional area | 4.02 ± 1.37 % (r = 0.231, p = .001) | 6.41 ± 2.12% (r = 0.32, p = .001) | - (p = .94) | - (p = .25) |
| Change in nasopharynx cross-sectional area | - (p = .68) | - (p = .32) | - (p = .26) | - (p = .24) |
| Tongue displacements per mm of mandibular advancement (partial Pearson P value) | Overall (n = 209 tagged MRI) | Responders (n = 98 tagged MRI) | Partial-responders (n = 31 tagged MRI) | Nonresponders (n = 80 tagged MRI) |
|---|---|---|---|---|
| Oropharyngeal tongue anterior movement | 0.50 ± 0.08 mm (r = 0.397, p < .001) | 0.47 ± 0.13 mm (r = 0.357, p < .01) | 0.60 ± 0.15 mm (r = 0.62, p = .01) | 0.58 ± 0.13 mm (r = 0.259, p = .027) |
| Nasopharyngeal tongue anterior movement | 0.25 ± 0.05 mm (r = 0.299, p < .001) | 0.40 ± 0.08 mm (r = 0.508, p < .01) | - (p = .38) | 0.26 ± 0.10 mm (r = 0.244, p < .01) |
| Change in oropharynx cross-sectional area | 4.02 ± 1.37 % (r = 0.231, p = .001) | 6.41 ± 2.12% (r = 0.32, p = .001) | - (p = .94) | - (p = .25) |
| Change in nasopharynx cross-sectional area | - (p = .68) | - (p = .32) | - (p = .26) | - (p = .24) |
Mandibular advancement moved the tongue anteriorly to enlarge the oropharynx only for responders. Posterior tongue anterior movement due to mandibular advancement in partial- and nonresponders did not translate into upper airway enlargement. All data are presented as mean ± standard errors. All partial Pearson correlations between variables and mandibular advancement reported in this table between brackets were controlled for age, BMI and gender, and statistically accounted for the multiple measurements made within the same participants.
Amplitude of forward tongue movement and enlargement of the cross-sectional area of the upper airway for each millimetre of mandible advancement for each MAS treatment response group
| Tongue displacements per mm of mandibular advancement (partial Pearson P value) | Overall (n = 209 tagged MRI) | Responders (n = 98 tagged MRI) | Partial-responders (n = 31 tagged MRI) | Nonresponders (n = 80 tagged MRI) |
|---|---|---|---|---|
| Oropharyngeal tongue anterior movement | 0.50 ± 0.08 mm (r = 0.397, p < .001) | 0.47 ± 0.13 mm (r = 0.357, p < .01) | 0.60 ± 0.15 mm (r = 0.62, p = .01) | 0.58 ± 0.13 mm (r = 0.259, p = .027) |
| Nasopharyngeal tongue anterior movement | 0.25 ± 0.05 mm (r = 0.299, p < .001) | 0.40 ± 0.08 mm (r = 0.508, p < .01) | - (p = .38) | 0.26 ± 0.10 mm (r = 0.244, p < .01) |
| Change in oropharynx cross-sectional area | 4.02 ± 1.37 % (r = 0.231, p = .001) | 6.41 ± 2.12% (r = 0.32, p = .001) | - (p = .94) | - (p = .25) |
| Change in nasopharynx cross-sectional area | - (p = .68) | - (p = .32) | - (p = .26) | - (p = .24) |
| Tongue displacements per mm of mandibular advancement (partial Pearson P value) | Overall (n = 209 tagged MRI) | Responders (n = 98 tagged MRI) | Partial-responders (n = 31 tagged MRI) | Nonresponders (n = 80 tagged MRI) |
|---|---|---|---|---|
| Oropharyngeal tongue anterior movement | 0.50 ± 0.08 mm (r = 0.397, p < .001) | 0.47 ± 0.13 mm (r = 0.357, p < .01) | 0.60 ± 0.15 mm (r = 0.62, p = .01) | 0.58 ± 0.13 mm (r = 0.259, p = .027) |
| Nasopharyngeal tongue anterior movement | 0.25 ± 0.05 mm (r = 0.299, p < .001) | 0.40 ± 0.08 mm (r = 0.508, p < .01) | - (p = .38) | 0.26 ± 0.10 mm (r = 0.244, p < .01) |
| Change in oropharynx cross-sectional area | 4.02 ± 1.37 % (r = 0.231, p = .001) | 6.41 ± 2.12% (r = 0.32, p = .001) | - (p = .94) | - (p = .25) |
| Change in nasopharynx cross-sectional area | - (p = .68) | - (p = .32) | - (p = .26) | - (p = .24) |
Mandibular advancement moved the tongue anteriorly to enlarge the oropharynx only for responders. Posterior tongue anterior movement due to mandibular advancement in partial- and nonresponders did not translate into upper airway enlargement. All data are presented as mean ± standard errors. All partial Pearson correlations between variables and mandibular advancement reported in this table between brackets were controlled for age, BMI and gender, and statistically accounted for the multiple measurements made within the same participants.
Across all 209 tagged MRI scans, when controlled for age, BMI and gender, and after statistically accounting for the fact that several measurements were made within the same participants, a larger mandibular advancement was related to the larger anterior movement of the nasopharyngeal and oropharyngeal regions of the tongue (partial Pearson, r = 0.299, p < .001, and r = 0.397, p < .001, respectively) (A and B, respectively). Greater mandibular advancement was also weakly associated with a larger increase in the cross-sectional area of the oropharynx (partial Pearson, r = 0.231, p = .001) (C). When the data were fitted to a linear regression, for each 1 mm advancement of the mandible, the tongue moved forward, primarily in the oropharyngeal region (0.50 ± 0.08 mm vs. 0.25 ± 0.05 mm for the nasopharyngeal section). This resulted in an enlargement of the cross-sectional area of the oropharynx (4.02 ± 1.37%/mm of mandible advancement). Slopes of the linear regression are reported as mean ± standard error.
Relationship to MAS treatment response
Posterior tongue movements.
Typical tagged images and tongue movement observed for each MAS treatment outcome category are presented in Figure 4 and the corresponding 3 videos of the mandibular advancement have been included as supplementary videos. For responders, the nasopharyngeal and oropharyngeal regions of the posterior tongue moved forward with a similar amplitude (~0.5 mm/mm of mandibular advancement) (Table 2). In contrast, for nonresponders, the oropharyngeal tongue advanced more than the nasopharyngeal region (~0.6 mm vs ~0.2 mm/mm of mandibular advancement). For partial-responders, only the oropharyngeal tongue advanced significantly with the mandible (~ 0.6 mm/mm of mandibular advancement).
Typical tagged MRI images and mid-sagittal posterior tongue movement observed during mandibular advancement for an OSA patient from each MAS treatment outcome category. Points and line in black represent the baseline position of the tongue when the mandible was in a neutral position. Points and line in blue indicate the position of the tongue at the end of the mandibular advancement. The posterior tongue’s nasopharyngeal and oropharyngeal regions moved forward for the responder. In contrast, for the partial and nonresponders, minimal advancement of the tongue was observed with mandibular advancement. The corresponding 3 videos of the mandibular advancement have been included as supplementary videos.
For responders and nonresponders, a larger mandibular advancement was associated with a larger nasopharyngeal and oropharyngeal tongue anterior movement, but not for partial-responders (Table 2). For the latter, a larger mandibular advancement was only significantly associated with a larger advancement of the posterior section of the oropharyngeal tongue but not with the nasopharyngeal section.
Upper airway area changes.
Mandibular advancement was associated with an enlargement of the cross-sectional area of the oropharynx only for responders (Table 2). For each advancement of 1 mm of the mandible, the cross-sectional area of the oropharynx enlarged by 6.41 ± 2.12 % (mean ± standard errors). For partial- and nonresponders, mandibular advancement was not associated with a change in the cross-sectional area of the pharynx.
Prediction of MAS treatment response
As the anterior movement of the tongue and the enlargement of the upper airway cross-sectional area with mandible advancement differed between MAS treatment response categories, a multinomial logistic regression analysis was used to determine whether these measures, combined with baseline AHI, could predict MAS response group better than a model that included only the baseline AHI when at least 4 mm of mandible advancement was achieved during the MRI scan. Models using different minimum thresholds for mandibular advancement achieved in the scan (> 1mm to > 4 mm) are reported in the supplementary file, along with cross-validation results.
The first model, using only baseline AHI, (χ 2 = 18.67, p < .001, n = 40) had an overall poor percentage of correct prediction of 50.0%, and this model was 5-fold cross-validated at 52.5% (21/40) (Table 3 and Supplementary Table S1). About 35.7% of responders (5/14), 62.5% of partial-responders (5/8), and 61.1% of nonresponders (11/18) were correctly predicted with the 5-fold cross-validation approach (Table 3 and Supplementary Table S4). The odds of being a nonresponder rather than a responder did not change significantly as baseline AHI increased (0.941 [0.867–1.021], p = .15), whereas the odds of being a partial-responder increased (1.155 [1.031–1.293], p = .013) (Supplementary Table S3).
Five-fold cross-validated predicted vs observed MAS response prediction accuracy when the mandible was advanced by at least 4 mm during the MRI scan
| Observed | Predicted | |||||
|---|---|---|---|---|---|---|
| Model with 1 predictor† | Model with 5 predictors‡ | |||||
| Nonresponders | Partial-responders | Responders | Nonresponders | Partial-responders | Responders | |
| Nonresponders | 11 (61.1%)§ | 2 | 5 | 11 (61.1%)§ | 2 | 5 |
| Partial-responders | 2 | 5 (62.5%) | 1 | 2 | 6 (75.0%) | 0 |
| Responders | 9 | 0 | 5 (35.7%)¶ | 6 | 0 | 8 (57.1%)¶ |
| Observed | Predicted | |||||
|---|---|---|---|---|---|---|
| Model with 1 predictor† | Model with 5 predictors‡ | |||||
| Nonresponders | Partial-responders | Responders | Nonresponders | Partial-responders | Responders | |
| Nonresponders | 11 (61.1%)§ | 2 | 5 | 11 (61.1%)§ | 2 | 5 |
| Partial-responders | 2 | 5 (62.5%) | 1 | 2 | 6 (75.0%) | 0 |
| Responders | 9 | 0 | 5 (35.7%)¶ | 6 | 0 | 8 (57.1%)¶ |
Data are reported as the number of participants “n”, followed by the percentage correctly predicted in brackets. The first model included only baseline AHI, and the second model included the tongue movements per mm of mandible advancement and changes in upper airway cross-sectional area/ mm of mandible advancement with the baseline AHI.
†Baseline AHI. Overall accuracy 50.0%, 5-fold cross-validated at 52.5%.
‡Baseline AHI + naso-and oropharyngeal tongue movement and airway enlargement per millimetre of mandibular advancement. Overall accuracy 69.2%, 5-fold cross-validated at 62.5%.
§Negative predictive value.
¶Positive predictive value. Parameter estimates for both models are reported Supplementary Table S4.
Five-fold cross-validated predicted vs observed MAS response prediction accuracy when the mandible was advanced by at least 4 mm during the MRI scan
| Observed | Predicted | |||||
|---|---|---|---|---|---|---|
| Model with 1 predictor† | Model with 5 predictors‡ | |||||
| Nonresponders | Partial-responders | Responders | Nonresponders | Partial-responders | Responders | |
| Nonresponders | 11 (61.1%)§ | 2 | 5 | 11 (61.1%)§ | 2 | 5 |
| Partial-responders | 2 | 5 (62.5%) | 1 | 2 | 6 (75.0%) | 0 |
| Responders | 9 | 0 | 5 (35.7%)¶ | 6 | 0 | 8 (57.1%)¶ |
| Observed | Predicted | |||||
|---|---|---|---|---|---|---|
| Model with 1 predictor† | Model with 5 predictors‡ | |||||
| Nonresponders | Partial-responders | Responders | Nonresponders | Partial-responders | Responders | |
| Nonresponders | 11 (61.1%)§ | 2 | 5 | 11 (61.1%)§ | 2 | 5 |
| Partial-responders | 2 | 5 (62.5%) | 1 | 2 | 6 (75.0%) | 0 |
| Responders | 9 | 0 | 5 (35.7%)¶ | 6 | 0 | 8 (57.1%)¶ |
Data are reported as the number of participants “n”, followed by the percentage correctly predicted in brackets. The first model included only baseline AHI, and the second model included the tongue movements per mm of mandible advancement and changes in upper airway cross-sectional area/ mm of mandible advancement with the baseline AHI.
†Baseline AHI. Overall accuracy 50.0%, 5-fold cross-validated at 52.5%.
‡Baseline AHI + naso-and oropharyngeal tongue movement and airway enlargement per millimetre of mandibular advancement. Overall accuracy 69.2%, 5-fold cross-validated at 62.5%.
§Negative predictive value.
¶Positive predictive value. Parameter estimates for both models are reported Supplementary Table S4.
The model using the dynamic imaging measures in conjunction with the baseline AHI (χ 2 = 35.17, p < .001, n = 39) had an overall satisfactory prediction accuracy of 69.2 %, 5-fold cross-validated at 62.5% (25/40) (Table 3 and Supplementary Table S1). This model improved the individual prediction accuracy of responders and partial-responders compared to the baseline model: correctly classifying 61.1% of nonresponders (11/18), 75.0% of partial-responders (6/8), and 57.1% of responders (8/14) (Table 3 and Supplementary Table S3). Baseline AHI (p = .011) and oropharyngeal percentage of airway enlargement/mm of mandibular advancement (p = .008) were significant predictors (Supplementary Table S2). Compared to being a responder, the odds of being a nonresponder were lower as the percentages of change in oropharyngeal upper airway cross-sectional area/mm of mandibular advancement increased (OR = 0.828 [0.707–0.970], p = .02), whereas it did not change as baseline AHI increased (0.914 [0.821–1.017], p = .10). The odds of being a partial-responder compared to a responder did not change as the percentages of change in oropharyngeal upper airway cross-sectional area/mm of mandibular advancement and baseline AHI increased (OR = 1.043 [0.862–1.262], p = .67, and OR = 1.148 [0.945–1.395], p = .16, respectively). See Supplementary Table S3 for full model parameter estimates.
Discussion
This study provides a number of novel insights about mandibular advancement splint treatment. Firstly, across all subjects, for each 1 mm of mandibular advancement, the tongue moved forward primarily in the oropharyngeal region, and this resulted in an enlargement of the cross-sectional area of the oropharynx. Secondly, differences were observed across treatment outcome groups in how the tongue moved forward and how the upper airway cross-sectional area enlarged with mandible advancement. Tongue advancement with mandibular protrusion was the greatest for responders (~0.5 mm/mm of mandibular advancement) and resulted in oropharynx cross-sectional area enlargement (~6%/mm of mandibular advancement). Thirdly, the inclusion of tongue movement and the percentage of upper airway cross-sectional area enlargement per millimeter of mandibular advancement in a multivariate model along with baseline AHI correctly classifies more patients than a model based only on baseline AHI. A satisfactory MAS response prediction was achieved for patients who achieved a passive jaw advancement of at least 4 mm during the scan (i.e. 62.5 %, 5-fold cross-validated).
Mandibular advancement, posterior tongue advancement, and upper airway enlargement
Although positive and negative treatment outcomes have both been associated with a larger upper airway enlargement with MAS [9], our results showed that only in responders was mandibular advancement associated with an anterior motion of the oropharyngeal tongue and enlargement of the oropharynx cross-sectional area. For partial- and nonresponders, the oropharynx did not enlarge despite the tongue advancing with mandibular advancement. This was perhaps due to a narrowing of the airway in the lateral dimension, although this was not able to be quantified on the mid-sagittal tagged images in this study. Changes in the midline and lateral dimensions or changes in oropharyngeal airway shape have been previously associated with mandibular advancement [9].
In contrast, within the nasopharynx, the tongue advancement did not translate into airway enlargement for responders. One possible explanation may be that in these patients, the soft palate did not advance sufficiently to enlarge the retropalatal cross sectional area, either because the upper tongue does not move far enough, or because the soft palate does not move with the tongue [34]. Two previous studies have shown that the soft palate remains in contact with the posterior section of the tongue when it advances during inspiration for many people with OSA [35,36]. Pharyngeal dilation (i.e. increased pharyngeal cross-sectional area) was observed when the posterior oblique compartment of the tongue moved forward by at least 1 mm during inspiration. If this is also the case with passive mandibular advancement, the current findings suggest that a jaw protrusion of at least 3 mm would be required for the nasopharyngeal tongue to advance by at least 1 mm, assuming that tongue advancement was not limited by anatomical confinement and that this level of mandibular advancement can be achieved at the start of the therapy by the patients. However, this needs to be verified because soft palate anterior movement was not measured in this study as the tag density on the soft palate was insufficient for reliable quantification. Another possible explanation may be that our cross-sectional area measures don’t reflect the airway enlargement that primarily occurs via an extension of the lateral dimension because our measures represent average measurements across the whole respiratory cycle due to the time required to collect the anatomical scans (~ 4 min).
Nasopharyngeal airway enlargement with mandibular advancement was not observed for partial- and nonresponders, and this appeared to be because the anterior movement of the nasopharyngeal tongue was insufficient. In comparison with responders, the amplitude of tongue advancement was smaller for nonresponders and insignificant for partial-responders. There are two possible explanations for this observation. Firstly, it may be that anatomical confinement of the tongue in the upper oral cavity limited the anterior movement of the nasopharyngeal tongue when the oropharyngeal part of the muscle was pulled forward with mandibular advancement [37]. While mandibular advancement may increase the space available for soft tissue in the oropharyngeal area of the oral cavity [38], this does not necessarily translate into increased volume in the nasopharyngeal region. For those subjects, tongue retaining devices could represent an alternative OSA treatment to MAS device [39], because they specifically pull the tongue forward. However, this needs to be verified, and such devices are less accepted as treatment by patients than MAS [40]. Secondly, lower tongue stiffness in people with OSA [41], perhaps caused by increased tongue fat content [42] and/or the presence of edema [43], could favor tongue elongation instead of tissue advancement with jaw protrusion, but the reasons why this would affect partial- and nonresponders to a greater extent is not yet clear.
Prediction of MAS treatment response
Across all patients, including those who only achieved small amounts of mandibular advancement at baseline, MAS treatment outcome prediction accuracies were good from the model using baseline AHI or using dynamic imaging measurements in conjunction with baseline AHI (67.2% and 64.3%, respectively), but poor when 5-fold cross-validated (54.9% and 53.5%, respectively). These accuracies are in the same range as those reported in previous studies assessing the utility of clinical factors (age and BMI) along with baseline OSA severity (58%) [13] or craniofacial measurements (51%) [44]. However, better prediction accuracies were achieved by Okuno et al. [45] with the use of nasendoscopy in awake subjects (80%), and by Remmers et al. [25] with a remotely controlled mandibular device during sleep (88%). Edwards et al. [22] have also reported an excellent level of classification accuracy (92%), using airway collapsibility and ventilatory control stability (loop gain) to predict MAS treatment outcome
A more complex model using multiple baseline characteristics and dynamic imaging measures, model #4, Supplementary Material) performed very well when participants who were unable to achieve a mandibular advancement that would usually be regarded as sufficient for clinical efficacy (> 4 mm) were excluded (92.3%). However, this more complex model was 5-fold cross-validated to an accuracy of 53.9%, which means that the model was overfitted on the “training data” due to the limited sample size, and would not reliably predict the treatment outcome in an independent population. In comparison, when the number of predictors was decreased to 5 (Baseline AHI and the 4 dynamic imaging measures), the overall accuracy reached a satisfactory level of prediction (69.2%, 5-fold cross-validated 62.5%), for participants with >4mm of mandibular advancement. Although this accuracy is lower than previously reported prediction models, this held up under cross-validation, indicating the model is more likely to perform similarly in an independent cohort.
Comparison of prediction accuracy between studies remains challenging since there is no consensus in the definition of MAS response groups, and different definitions were used in some of the studies mentioned above (e.g. [22]). Also, MAS treatment prediction methods have generally dichotomized responses into responders and nonresponders, whereas we used all three MAS response groups (responder, partial-responder, and nonresponder) in our analysis. This was done because partial-responders may have poorer long-term health benefits than complete responders from MAS therapy due to residual OSA. Being able to predict if patients would be either responders, partial-responders, or nonresponders would represent a significant advancement in treatment management of people with OSA.
Finally, while a simple model using only baseline AHI as a predictor across the whole sample was able to accurately predict responders (positive predictive value = 93.9%) with excellent 5-fold cross-validated prediction accuracy, this simple model performed very poorly in identifying nonresponders (negative predictive value = 11.1%). In contrast, the model using the dynamic imaging measures with baseline AHI classified each of the MAS response groups with a similar accuracy across all levels of mandibular advancement, with a higher overall 5-fold cross-validated prediction accuracy at larger degrees of mandibular advancement. Oropharyngeal tongue movement and upper airway enlargement/ mm of mandibular advancement were significant predictors of treatment outcome when the jaw protrusion was larger than 2 mm. This demonstrates the utility of our dynamic imaging measures when considered in conjunction with baseline AHI, and was expected because the mechanisms of action of the MAS device are thought to be largely anatomical [37], and larger mandibular advancement contributes to larger anatomical effects in this study.
The results emphasize further that baseline characteristics alone are limited for predicting MAS treatment outcomes. More complex measurements need to be used to develop better prediction models. However, this would likely come at the expense of clinical feasibility and practicality in the care of patients with OSA. Although, MRI is currently the only method capable of measuring upper airway cross-sectional area changes and dynamic tongue movement with mandibular advancement, this approach would be impractical unless simpler, cost-effective, accurate alternative imaging modalities can be developed and validated. For example, further investigations of the use of upper airway ultrasound, which can also measure tongue displacements, but cannot provide upper airway cross-sectional area measures [46], may improve clinical accessibility and feasibility.
MAS response groups
MAS treatment outcome was determined using the maximal comfortable jaw protrusion that could be achieved by the patients, following usual practice. However, recent findings [12] suggest that some patients may benefit from a smaller degree of mandibular advancement. In particular, upper airway collapsibility seems to improve in the first half of mandibular advancement for nonresponders compared to responders for whom it occurs toward maximal advancement. That was a small study, and if confirmed, this could imply that some subjects diagnosed as nonresponders in this study may be partial-responders or responders if a smaller mandibular advancement was used, and may explain why the proportion of responders was lower at higher thresholds of mandibular advancement in our study (Supplementary Table S4). It has also been reported that patients with OSA can tolerate more advancement over time [47], consistent with our study, where mandibular advancement at the final polysomnography study was significantly greater than when measured at the start of the study (i.e. during the MRI scan and initial dental visit). This could mean that the counterproductive effect of larger mandibular advancement for some patients at the time of the final polysomnography study may not be present at the start of the therapy when smaller a mandibular advancement is achieved.
Limitations
Despite considerable strengths drawn from the sophisticated imaging technique applied in this study, there are several limitations. Firstly, participants had a lower average BMI than a typical OSA population [48], even though there was no weight limit in the exclusion criteria for recruitment. However, participants were required to undergo an MRI scan in a 60 cm bore scanner, which does not allow some very obese subjects to be scanned comfortably due to the confined space. Also, the heaviest participants may not have been referred for MAS therapy by their physician, based on previous studies that showed that patients with lower BMI are better candidates for MAS therapy [19]. Further investigations, possibly using a wider bore MRI, could determine whether our findings also apply to patients with OSA with higher BMI. Secondly, by design, different MAS devices were used for scanning and clinical treatment, both because the SomnoDent Flex device was not MRI compatible and cannot be passively advanced during scanning, and because we were aiming to explore whether this approach could predict treatment response prior to commencing MAS therapy. The latter requires that an inexpensive MAS be used for imaging rather than a patient needing to have an expensive custom appliance made for testing. However, the OSA treatment efficacy of the Apnea Guard and custom acrylic splints is similar [49,50], and both advance the jaw with similar degrees of mouth opening, although the Apnea Guard occupies a slightly larger volume in the oral cavity because it is not custom-manufactured for each participant. Thirdly, participants were imaged while awake. Therefore, tongue advancement measurements during dynamic protrusion of the mandible may be different to the effect of a static advancement during sleep. Awake measurements are possibly influenced by voluntary and reflex muscle activation [51], although participants were instructed to remain relaxed during the scanning, and several practice runs were done before collecting the data. It is also possible that more complex aspects of the skeletal and soft tissue anatomy beyond those measured in this study (Table 1), and differences between head and jaw position during natural sleep and the postures adopted during the MRI scan might influence the capacity of this technique to predict MAS therapy outcome.
Conclusion
We found that the amplitude of the posterior tongue advancement and oropharynx enlargement during passive mandibular advancement differed between MAS treatment outcome groups, with larger responses in responders than partial- and nonresponders. Tongue advancement and upper airway enlargement with mandibular advancement in conjunction with baseline AHI determine MAS therapy outcomes with a satisfactory accuracy of 69.2% when a mandibular advancement of at least 4 mm was achieved at the start of the therapy. In comparison, a model-based only on baseline AHI correctly classified only 50.0% of patients. It remains to determine whether a sensitivity analysis using the supine derivated AHI to define MAS treatment outcomes would improve the predictive accuracy of the model using the dynamic imaging measures by controlling the effects of body position [13] since the MRI was conducted on supine position. A better understanding of how posterior tongue movement with mandibular advancement impacts MAS treatment outcomes depending on the site of collapse could also provide novel insights on whether or not oral appliances are a suitable option for patients who experience palatal-based collapse. Previous work tends to suggest that palatal collapse reduces the likelihood of successful MAS treatment [24,52–54].
Acknowledgments
The authors thank SomnoMed Pty Ltd for generously providing the SomnoDent Flex MAS device used in this study. They also thank the Neuroscience Research Australia (NeuRA) Imaging Centre, the team from the sleep and breathing clinic at NeuRA and Ms Melanie Madronio from the Centre for Sleep Health and Research at the Royal North Shore Hospital (Australia) for their technical support.
Funding
This research was funded by the National Health & Medical Research Council (NHMRC) of Australia (APP1082364). L.E.B., D.J.E., and J.E.B. are supported by NHMRC Fellowships (APP1077934, APP1116942, and APP1042646, respectively).
Disclosure Statement
Financial Disclosure: PAC has an appointment to an endowed academic Chair at the University of Sydney that was created from ResMed funding. He has received research support from ResMed, SomnoMed, Zephyr Sleep Technologies, and Bayer. He is a consultant/adviser to Zephyr Sleep Technologies, ResMed, SomnoMed, Bayer, and Signifier Medical Technologies. He has received speaker fees from ResMed and Nox Medical. He has a pecuniary interest in SomnoMed related to a previous role in R&D (2004). DJE has a Cooperative Research Centre (CRC)-P grant (Industry partner Oventus Medical) and receives research income from Bayer and Apnimed outside the current study. Somnomed provided the oral appliances used for treatment in this study but have had no role in the design, analysis, or reporting of the research. LJ, KS, JEB, JY, LEB, FLK, PGRB, ABL, KZCG, ECB, JN, PH have none to disclose.
Nonfinancial Disclosure: none for LJ, KS, JEB, JY, LEB, FLK, PGRB, ABL, KZCG, ECB, JN, PH.
Data Availability
Due to conditions of the ethical approvals for this project, the raw data cannot be shared.
References
Comments