Efficacy of mandibular advancement device in the treatment of obstructive sleep apnea syndrome: A randomized controlled crossover clinical trial

Background Evaluation of the efficacy and safety of a mandibular advancement device (MAD) (KlearwayTM) in the treatment of mild-to-moderate obstructive sleep apnea and chronic roncopathy. Material and Methods A randomized, placebo-controlled, double blinded, and crossover clinical trial was conducted. Placebo device (PD) defined as a splint in the centric occlusion that did not induce a mandibular advancement served as a control. The mandible was advanced to the maximum tolerable distance or to a minimum of 65% of the maximum protrusion. After each sequence of treatment, patients were assessed by questionnaires, conventional polysomnography, and objective measurement of snoring at the patient’s own home. Results Forty two patients participated in the study and 38 completed the study. Patients mean age was 46 ±9 years and the 79% were males. The mean mandibular advancement was 8.6 ±2.8 mm. Patients used the MAD and the PD for 6.4 +2.4 hours and 6.2 +2.0 hours, respectively. Secondary effects (mostly mild) occurred in the 85.7% and the 86.8% of the users of MAD and PD, respectively. The MAD induced a decrease in the apnea-hypopnea index (AHI) from 15.3 +10.2 to 11.9 +15.5. The 50% reduction in the AHI was achieved in the 46.2% and the 18.4% of the patients treated with MAD and PD, respectively. The use of the MAD induced a reduction in the AHI by 3.4 +15.9 while the PD induced an increase by 10.6 +26.1. The subjective evaluation of the roncopathy indicated an improvement by the MAD and an increase in the perceptive quality of sleep. However, the objective evaluation of the roncopathy did not show significant improvements. Conclusions The use of MAD is efficient to reduce the AHI and improve subjectively the roncopathy. MAD could be considered in the treatment of mild-to-moderate OSA and chronic roncopathy. Key words:Obstructive sleep apnea (OSA), mandibular advance device, treatment, efficacy, clinical assay.


Introduction
The obstructive slee� a�nea (OSA) is a �revalent disease that affects in its severe form 2-8% of the general population. It is estimated that more than 20% of the �o�ulation has an a�nea�hy�o�nea index (AHI) value that is ≥ 5 (1,2). Clinical symptoms of OSA include somnolence, neuropsychiatric and cardio-respiratory disorders. These com�lications result from anatomical and functional alterations of the u��er airway where repetitive episodes of obstruction during sleep provoke oxihemoglobin desaturation and tem�oral arouses that lead to a none-reparative sleep (3). Many studies have established OSA as a risk factor for arterial hy�ertension and traffic accidents (2�4�6). The presence of OSA has also been related to cardiovascular and cerebrovascular complications (5,7,8). Even more, higher mortality has been re�orted among �atients with OSA (9,10). In spite of all these consequences, OSA is still not adequately managed. Only 10% of the population with OSA are diagnosed and treated (3�11, 12). The administration of a continuous �ositive airway �ressure (CPAP) is an efficient� cost�effective treatment of OSA (3�13�14). There is a total agreement to indicate this treatment in �atients with AHI > 30 who suffer from com�lications related to the OSA (3). However� the indication of the CPAP in �atients with mild�to�moderate OSA (AHI = 5-29) is not that clear (3,14,15). The mandibular advancement device (MAD) is considered as an alternative to the CPAP (3). This device� also designed to alleviate and treat roncopathy, provokes the protrusion of the mandible in order to elevate and advance both the hyoid bone and the tongue. By this, the volume of the u��er airway is increased making less likely the �assage narrowing or colla�se (16). Several studies have recommended the treatment of mild-tomoderate OSA by MAD and have also indicated its use in �atients with severe OSA intolerable to the CPAP (3�17�18). However� there is still a need for randomi�ed controlled clinical trials to establish with more �recision the indications and the efficacy of the MAD in the treatment of OSA. In this study, a randomized controlled crossover clinical trial has been �erformed to evaluate the efficacy and safety of mandibular advancement device in the treatment of �atients with mild�to�moderate OSA. A �lacebo device (PD) has been used as a control.

* Patients
Randomized, placebo-controlled, double blinded, and crossover clinical trial was conducted at the Interdisciplinary Unit of Sleep Disorders of Alava University Hos�ital. The study �rotocol and informed consent� in full accordance with the ethical �rinci�les of the Declaration of Helsinki of 1975� as revisited in 2000� were a�proved by the ethical committee of University Hospital of Álava (Vitoria-Gasteiz, Spain). Patients were consecutively selected from adult subjects referred due to a clinical suspicion of OSA. Patients from both sexes were eligible to �artici�ate in this study and were selected according to the following inclusion criteria: -Age higher than 18 years, -Presence of chronic snoring. A patient is considered as chronic snorer if his/her bedmate/roommate reported to snore more than 5 days �er week and this is corroborated by a respiratory polygraphy performed in the patient's own home. The result of the res�iratory �olygraphy should indicate the presence of snoring during at least 30% of the nocturnal period.
-Confirmed diagnosis of mild�to�moderate OSA (5 ≤ AHI < 30) by polysomnography (PSG). -Have a roommate or bedmate to submit information. Patients were excluded according to the following exclusion criteria: -High-risk professions and/or controlling dangerous machines.
-Moderate or severe somnolence during day time.
-Coronary cardiopathy, acute vascular disease (less than three months), chronic and severe obstructive pulmonary disease� and chronic treatment with theo�hyllines. -Tem�oro�mandibular joint �roblems or �eriodontitis. -Mandibular protrusion capacity less than 6 mm and/or less than 10 teeth in each jaw. � Severe cognitive disorders and/or �atients whose an� Severe cognitive disorders and/or �atients whose answers to the questionnaires will be altered by chronic and severe diseases.
-Pregnancy (since the third month of pregnancy to 3 months after birth delivery). A total of 118 �atients were screened� of whom 76 were not eligible: 62 did not meet the inclusion criteria or had one or more of the exclusion criteria� and 7 declined to participate (Fig. 1).
Forty�two �atients were randomly assigned to receive two �ossible sequences of treatment (MAD or �lace� ent (MAD or placebo device (PD) according to a computer-generated randomi�ation schedule. To achieve a double�blinded study, professionals not related to the interventions and o�aque�sealed envelo�es were ado�ted. The dentists and ENT surgeons� res�onsible of device fabrication� ada�tation� and following u� of the �atients� had no contact during the clinical trial with the �rofessionals who �erformed the slee� study. All the controls and the results of the tests were code identified and were only known by a one �rofessional who did not �artici�ate in the clinical trial. Initially� each �atient were subjected to a �eriod of 2 weeks without any treatment� followed by 4 weeks of adaptation and standardization of the device (MAD or PD)� and 12 weeks of treatment (Fig. 1). Once this �eriod was finished� �atients were switched to use the other device following the same �rotocol as described above (Fig. 1). Mandibular advancement device (MAD): The commercial device KlearwayTM (University of British Colum-bia� Vancouver� Canada) was used. The fabrication of the device was made on model casts of both jaws and was ada�ted to the �atient's mouth by a dentist with the objective to achieve a sufficient and tolerable mandibular advancement� being at least 65% of the maximum �rotrusion ca�acity of the mandible. This �hase may need more than one visit to the dentist and had a period of 4 weeks at maximum.
Placebo device (PD): The �lacebo device was the same KlearwayTM device but in centric occlusion and did not �rovoke mandibular advancement. The dentist assured the absence of mandibular advancement and alteration to the TMJ �osition. The reference �oint was jaw �osition at the TMJ level in rest as measured by ce�halometry. The PD ada�tation may need more than one visit to the dentist and had a �eriod of 4 weeks at maximum. * Primary outcome The �rimary outcome was the a�nea hy�o�nea index (AHI) that was measured by a conventional �olysom-nogra�hy (PSG). The PSG study was reali�ed at the Interdisciplinary Unit of Sleep Disorders of Alava University Hos�ital. The PSG study was �erformed before the study and after the 12 weeks of the treatment with each device (MAD and PD). All slee� studies were �erformed with PSG (Alice 3 Healthdyne system) according to the standard parameters of electroencephalogram (C3-A2, C4-A1), electrooculo gram, submentonian and tibial electromy-ogram� electrocardiogram (modified V2)� res�iratory effort (thorax and abdominal resistance)� air flow (nasal and oral thermistor)� Oxygen saturation (cutaneous �ulsioximetry with a finger �robe (Palco laboratories P�340) and snore micro�hone. The PSG study was man� and snore micro�hone. The PSG study was manually interpreted in periods of 30 seconds according to the criteria of the American Academy of Sleep Studies (19) and following the guidelines of the S�anish Society of Pneumology and Thoracic Surgery (SEPAR) (3). The minimum time of recording was 6 hours and the minimum time of slee� was 180 minutes. The following definition of the res�iratory variables were used: � A�nea: The com�lete sto� (> 90%) of the naso�oral airflow during a minimum of 10 seconds. The a�nea was then classified as obstructive if it was accom�anied by thoracic and abdominal effort, central if this effort was absent and mixed if both situations occurred in one single apnea. � Hy�o�nea: A dro� in the res�iratory signal between 30% and 90%� accom�anied by a dro� in oxygen saturation ≥ 3% and/or arousal. * Secondary outcomes: Sleep characteristics: the total sleep time, the time of the �artial �hases of slee� (N1� N2 and N3) and the REM �hase were calculated from the PSG study. The slee� fragmentation was measured by the arousal index. Oxygen saturation: Cutaneous �ulsioximetry with a finger �robe (Palco laboratories P�340) was used to measure this variable during the performance of the PSG study. This variable was described by the mean oxygen saturation� minimum oxygen saturation and the �ercentage of time s�ent at SaO2 below 90% (CT90). Snoring: On one hand both patient and bedmate/roommate were asked to evaluate the �atient's ronco�athy by answering a questionnaire after finishing the treatment with each device. The question was How is your snoring? The �ossible answers were 1. Have increased a lot� 2. Have increased slightly, 3. Have not changed, 4. Have decreased slightly, and 5. Have decreased a lot. On the other hand, validated respiratory polygraphy system (Mesam IV) was used for the objective evaluation of the snoring habit (2). This system had the ca�acity to save information of 18 hours of recording. The snoring was measured by a protected microphone placed at the yugulum. The micro�hone filtered the sounds between 50 y 800 cycles/second as snoring sounds occurred in this range. If the volume of the sound is higher than 50% of the total volume� the sound was then identified as snore. If the sound exceeded a threshold of > 1.1 mV and 1000 cycles/second� it was then considered as strong snore. The data was �rocessed by a �ersonal com�uter and could be printed. The system �ermitted an automatic or manual reading of the recordings in �eriods of 10 minutes. The study with Mesam IV was �erformed before treatment and after the 12 weeks of treatment with each device (MAD and PD). The snoring was described by measuring the intensity, number of snores per hour of recording, percentage of time of recording with evidence of snoring and its relationshi� with body �ostures during slee�. Somnolence: the E�worth scale was used to measure the somnolence. The questionnaire also included other questions: Comparing to the situation before treatment how are you during the day? Com�aring to the situation before treatment how is your humor? The roommate was s�ecifically asked about his slee� in com�arison to the he/she had before treating the �atient. The answer for all these questions was selected from the following o�tions: 1. Much worse; 2. Slightly worse; 3. No changes; 4. Slightly better; and 5. Much better. Treatment com�liance: it was evaluated by calculating the time of device use. This time was obtained from the �atient´s declaration and in a consensus with the roommate. If there was no consensus� the �atent's declaration was considered as valid. Since the beginning of the study, patients had been phone called once per month to evaluate the treatment progress and patients adherence to the treatment. The use of the device was determined by the number of nights �er week and the average number of hours of use per night since the last control. The average time of use of the device �er hour was estimated by the multiplication of both variables and then the result was divided by 7. However� if the �atient had used the device for different time between weeks� an average was calculated �er each week and then the results were used to calculate the mean. It was considered a good adherence to the treatment if the mean time of use was ≥ 4 hours �er night as described by the study of Ferguson et al. (20). Complications: the nature, onset, duration, severity, and the outcomes of all adverse events� as well as any association of an adverse event related to the device (MAD and PD) were assessed and documented. In order to evaluate the safety �rofile of the treatments� all com-�lications and/or adverse events were recorded with an accountability scale. * Sample size calculation The sam�le si�e was determined by taking in consideration the crossover design of the study. The objective was to be able to detect the minimum effect of the device on the variables of �ercentage of �atients with AHI > 5� the distribution of AHI, the distribution of number of snores per hour of sleep and the percentage of patients with snoring. The calculation was �erformed considering α value of 0.05� statistical �ower (1�β) of 90% and bilateral hypothesis test. A sam�le si�e of 40 subjects would result in a statistical �ower higher than 90%. In the case of AHI� a sample size of 40 patients permitted detecting a minimum difference of 9.5 in agreement with the data re�orted by Clark et al. (21) and Ferguson et al. (20) giving a standard deviation of 18 and the application of paired tstudent test. In the case of number of snores per hour of sleep objectively measured in bed, a sample size of 40 patients permit detecting a minimum difference of 12.4 giving a standard deviation of 24.1.

* Statistical analysis
The t�student test was used to com�are the characteristics between grou�s. Paired t�student and analysis of variance were used for the com�arison of continu-ous variables of �aired sam�les. McNemar's test was applied to see if the devise use had am effect over the AHI. Linear and multi�le regression analysis were used for the crossover design and to examine the relationshi� between variables. The evaluation of the results was performed considering the potential effect of the period and the results were ex�ressed for intention to treat and for every protocol. Figure 1 describes the study flow chart. During a recruitment �eriod of 6 months� 112 �atients were evaluated. Sixty�three �atients were excluded as they did not complete the inclusion criteria or had one or more of the exclusion criteria. Seven �atients refused to �artici-�ate in the study. Forty�two �atients �artici�ated in the study and 38 patients (90.5%) completed all phases of the study. Four �atients (9.5%) (2 men and 2 women) abandoned the study due to intolerance and/or secondary effects of the device. Table 1 shows the �atients characteristics at the dif� 1 shows the �atients characteristics at the dif� shows the �atients characteristics at the different �hases of the study (ada�tation� treatment with MAD and treatment with PD). Patients had a mean age of 46 ±9 years and 33 (79%) were males. There was no statistically significant difference in relation to the anthropometric characteristics and the clinical variables between the different �hases of the study. However� a discrete increase in the mean body mass index was detected between the ada�tation and �lacebo �hases. This increase was from 27.7 ±3.2 to 29.3 ±9.2 (p < 0.05). The MAD achieved a mandibular �rotrusion of 8.6 ±2.8 mm. It was notorious the decrease in alcohol consum�tion between the ada�tation and treatment �hases for both study arms (p < 0.05). Table 2 shows the results of PSG at the different �hases of the study. Neither MAD nor PD modified significantly the slee� duration� intensity or efficacy. Moreo-ver� the com�arison between the ada�tation �hase and the treatment phase in both study arms (MAD and PD) showed the absence of statistically significant differences in the total sleep time, the time of the partial phases of slee� (N1� N2 and N3) and the REM �hase. The slee� fragmentation� measured by the arousal index� was not reduced by the treatment with MAD. However� it was significantly increased by the treatment with PD. The analysis of the res�iratory events showed that MAD reduced the AHI from 15.3 +10.2 to 11.9 +15.5. This reduction was not statistically significant (p = 0.196). Whereas, the PD incremented the AHI to 25.6 + 26.0 and this increase was statistically significant (p = 0.016). Worth to mention� the use of MAD significantly reduced most of the res�iratory events when com�ared to the PD as shown in when the variables of mean oxygen saturation� minimum oxygen saturation and CT90 were com�ared between the ada�tation �eriod and device�using �hases. The �eriod effect was also evaluated by com�aring the mean difference in the AHI of the patients initially treated with PD and the �atients initially treated with MAD. The statistical analysis showed no significant differences. Additionally, the carryover effect and the interaction between the treatment and the �eriod were also evaluated using the mean AHI after the use of MAD and PD. The results also showed the absence of a significant differences (data not shown). Table 3 describes the evolution of the roncopathy for the arms of the study. The evaluation was �erformed subjectively (questionnaire) and objectively (Mesam IV).

Results
The use of the MAD resulted in a significant reduction in the perception of snoring and increases the perceived quality of slee� when evaluated by both the �atient and the bedmate/roommate. However� the objective evaluation showed no significant changes in the snoring characteristics.
The com�liance with the use of device was 6.4 + 2.4 hours for the MAD and 6.2 + 2.0 for the PD (p > 0.05). The 87.1% and the 76.3% of the �atients used the MAD and PD res�ectively for more than 5 hours/night ( Table  3).
The secondary effects from the use of both s�lints are shown in   Table 4 also showed that the secondary effects tended to be higher when the MAD was used. Figure 2 shows the com�arison of the mean AHI between the basal �hase with the PD and MAD �hases. This com�arison revealed a significant increase in the AHI by the PD (p = 0.017) and a reduction in the AHI, although not statistically significant� was observed in the MAD �hase. When both s�lints were com�ared� the use of MAD reduced significantly the AHI (p = 0.000). Figure 3 shows the results of the mean change in the AHI. An increment in the AHI by 10.6 + 26.1 was caused by the PD while a reduction by 3.4 + 15.9 was caused by the MAD. The differences between the PD and the MAD were statistically significant ( p = 0.000).

Discusion
The �rinci�al finding of this study is that mandibular advancement device, in comparison to a PD, has produced a significant enhancement in all the �arameters that measure the respiratory events. As a secondary out-come� the MAD has reduced significantly the chronic ronco�athy when evaluated by the bedmate/roommate. However� this reduction was not significant when the snoring was objectively measured. In the last years several review studies have been �ublished to evaluate the use of the MAD in the treatment of OSA (17)(18)(19)(20)(21)(22). All these studies have concluded that this device has a positive effect in reducing the AHI but is inferior to the CPAP treatment (22). For that, there is no doubt that MAD is a reasonable alternative to the CPAP in �atients with mild�to�moderate grade of OSA and in �atients with severe OSA intolerant to the CPAP (22).
Although there are numerous clinical trials on the efficacy of MAD in �atients with OSA� some issues remain unresolved. It is difficult to �redict the ty�e of �atients who could maximally benefit from the use of MAD by having the highest tolerance to the splint and the least number of secondary effects (22).
Herein� when com�ared to PD� the MAD has significantly im�roved all res�iratory �arameters. The device reduced the absolute number of apnea and hypopnea as well as the AHI. The decrease in the AHI by more than 50% was achieved in almost half of the �atients treated with MAD. AHI < 10 and AHI < 5 were achieved for the 57.9% and the 31.6% of �atients treated with MAD and PD, respectively. Similar results have been reported by other controlled clinical trials (20�21) confirming the efficiency of MAD in the treatment of mild�to�severe OSA.
The selection of the ty�e of the mandibular advancement device has been evaluated by several studies. Similar results have been reported by clinical trials that use two��iece design of the s�lint and trials that use mono-block designs (23). Recently, adjustable MAD (805 �atients) and fixed (203 �atients) have been com-�ared in a retros�ective study. The im�rovement in the OSA was better for the adjustable device than the fixed device (56.8% Vs 47.0%) (24).
Herein� an adjustable MAD was used to achieve the maximum and tolerable mandibular advancement (a minimum of the 65% of the maximum mandibular �rotrusion). This was ada�ted to reduce the number and/ or intensity of the adverse events, minimize the injury to the temporomandibular joint and obtain better outcomes. A dose-response effect has been reported for the MAD, the more advancement of the mandible better is the response (25). In this study, the mean mandibular advancement was 8.8 + 2.8 mm.
Although it has been re�orted that when 50% of the maximum �rotrusion is achieved� there is a tendency for the secondary event to increase. In our study, 65% of the maximum �rotrusion has been achieved and the �ercentage of secondary effects has been similar to other studies (20,22). One of the consequences of OSA is the loss of sleep architecture; lesser �ro�ortion of �rofound slee� (N3) and REM� and increase in the arousal index as ex�ression of slee� fragmentation (3). The findings of this study    indicated the absence of differences in the sleep charac-teristics� which could be related to the mild�to�moderate grade of OSA. Singh et al. (26) have found significant reduction in the arousals in OSA �atients treated with MAD but the patients in that study had more severe sleep fragmentation (50.8 + 31.0) than our study. Regarding the oxygen saturation� the mild�to�moderate severity of OSA could also contribute to the absence of significant effect of the MAD (27). Some studies have reported that the MAD produces a reduction in the values of arterial tension and improvements in the excessive daytime slee�iness (22). In our study� the MAD did not significantly alter these �arameters nor the PD. Possibly, the severity of OSA (mild to moderate) could contribute to this finding.
With respect to the roncopathy, different studies have re�orted relevant im�rovements when MAD is used (28). The subjective evaluation of the snores has indicated a relevant im�rovement (about 90%). However� the objective evaluation has only shown mild and not significant im�rovements. This discre�ancy could be attributed to the differences in the tolerance to the snores. During the objective evaluation, the system evaluates the �resence/absence of snore but not its intensity. This means that a snore may persist but the reduction in its intensity would make it tolerable to the �atient and the roommate/bedmate. However� Adriana et al. (27) have found improvements in the objective evaluation of the snores. This could be attributed to the fact that the basal situation of the patients in the study by Adriana et al. (27) was higher than the �atients in our study and thus had more room for im�rovement. The snores/hour index was 40.4 and 24.5� res�ectively. The secondary effects due to the use of the MAD were frequent. These effects were �resent in the 86% and 87% of the �atients treated with MAD and PD� res�ectively. Although there were no significant differences between the grou�s� these secondary effects tend to be more severe in the MAD group.
Most of the secondary effects were mild to moderate. They were a subjective com�laints of hy�ersalivation� mouth dryness, gingival pain, dental pain, lingual pain, or TMJ �ain. This is in agreement of the findings of �revious studies (22). However� severe secondary effects� referred to as an irreversible alteration of the occlusion, occurred in 5 �atients due to the use of MAD. There was no association between the subjective com�laints of the patients and the severe secondary effects. Data from the literature could suggest that irreversible occlusal alterations are related to the time of use of the MAD (21). This suggestion could not be evaluated in the study as the follow�u� time was limited to 12 weeks in each arm of the trial.
One of the aspects related to the use of the MAD is the relative inca�acity to �redict which �atients would benefit from the use of MAD and which not. It seems there is an individual variation in the response to a treatment with MAD (22). Some variables have been observed to be associated with the best res�onse to a treatment with MAD. Of these variables are lower severity of the AHI� �ostural OSA� lower age� feminine gender� and lower obesity (22). Due to the relatively small sample size, inherent to a clinical trial� it was im�ossible to identify variables with a ca�acity to �redict the res�onse to the MAD. It is �ossible that the new techniques of image evaluation of the u��er airway �lay a role in identifying �atients that res�ond well to a treatment with MAD. This should be considered as �riority in the clinical research in this field (29). Even is more intriguing the fact that in some �atients if treated with MAD the AHI is worsened. The findings of this study indicated that 10.3% of the �atients with MAD had their AHI increased by 50%, in comparison to the 31.6% in the case of the PD. This has been also observed in other studies (30). The etiology of such a res�onse to the MAD is not totally understood and has been related to anatomical changes particularly in the position of the hyoid bone. This would make em�hasis in the im�ortance of performing a sleep study to monitor the outcomes of a treatment with MAD. This randomi�ed clinical trial has recruited 38 �atients� a sample size that has been adequate to meet the study objectives. However� this sam�le si�e did not allow for patients subgrouping to identify factors that could be associated with a good res�onse to the treatment with MAD. The criterion of maximum mandibular advance-ment� that was limited to a minimum of 65% of the maximum mandibular �rotrusion� was arbitrarily established. It is possible that lesser mandibular advancement could obtain similar results but with fewer secondary effects. Future research would test this hy�othesis and establish a criteria to maximi�e the benefits and minimize the secondary effects. An important limitation to this study is that the used PD has not been a truly placebo (the results of the slee� study were affected by the PD). This limitation should be considered when analy�ing the results of this clinical trial.
Within the limitation of this study, the use of MAD is efficient to reduce the AHI y im�rove subjectively the ronco�athy. Thus� MAD could be considered in the treatment of mild-to-moderate OSA and chronic roncopathy. Further studies should evaluate the hypothesis of obtaining similar outcomes at minimal secondary effects when shorter mandibular advancement is achieved. Studies with larger sam�le si�e are still needed to establish the �atient �rofile that could be associated with a good res�onse to a treatment with MAD.