Oral appliance therapy versus nasal continuous positive airway pressure in obstructive sleep apnea: a randomized, placebo-controlled trial on psychological distress

Obstructive sleep apnea (OSA) is characterized by recurrent obstruction of the upper airway, often resulting in oxygen desaturation and arousal from sleep [1]. Excessive daytime sleepiness, snoring, and reduction in cognitive functions are common symptoms of this condition [1]. OSA patients may also report symptoms of psychological distress, such as depression and anxiety [2]. Beebe and Gozal [3] suggested that both intermittent hypoxia and sleep disruption induce dysfunction of the prefrontal regions of the brain cortex, which may predispose to psychological distress.

Although continuous positive airway pressure (CPAP) has been proposed as the most effective treatment for severe OSA patients, nowadays mandibular advancement devices (MADs) are considered as a primary treatment option in mild and moderate OSA patients and in patients who do not tolerate CPAP [4]. The rationale behind the efficacy of MADs is that advancement of the mandible and tongue improves upper airway patency during sleep by enlarging the upper airway and by decreasing upper airway collapsibility [5, 6].

Barnes et al. [7] compared the effects of MAD treatment with CPAP on mood disorders and depression in a randomized placebo-controlled crossover trial and found no significant differences between these two therapies in their improvement of these disorders. Similar results were found by Engleman et al. [8] in a randomized crossover trial in which the effects of CPAP and MAD treatment on anxiety and depression symptoms were compared. To our best knowledge, no randomized placebo-controlled trials have been performed comparing the effects of an objectively titrated MAD and CPAP on symptoms of psychological distress. To enable an unbiased comparison between those treatment modalities, both treatments should be titrated objectively. Further, the crossover design of previous studies may have a risk of carry-over effects. The primary aim of this randomized placebo-controlled trial was, therefore, to compare the effects of an objectively titrated MAD with those of nasal CPAP (nCPAP) and an intraoral placebo appliance on symptoms of psychological distress in a parallel design. The hypothesis was that there is no significant difference between objectively titrated MAD and nCPAP therapies in improving psychological distress symptoms in mild/moderate OSA patients. To control for possible placebo effects, an intraoral placebo device served as a passive control condition for both active treatment modalities. It was hypothesized that the intraoral placebo appliance would not significantly improve psychological distress symptoms in mild/moderate OSA patients. Following the hypothesis of Beebe and Gozal [3], we also hypothesized that a significant correlation between the amount of psychological distress and the apnea-hypopnea index (AHI) values in the three therapy groups would occur. Therefore, the secondary aims of this trial were (1) to determine the relation between the amount of psychological distress and the AHI values at baseline in the three therapy groups and (2) to determine the relation between the amount of psychological distress at baseline and the change of AHI over time in the three therapy groups.

A total of 64 patients were enrolled in the study and were randomized at the start of the RCT as shown in Fig. 1 [9]. Three patients in the nCPAP group terminated the treatment before evaluation, because they experienced more side effects than benefits out of their treatment. One patient in the placebo group terminated the treatment, because of private reasons unrelated to the study. Another patient in the placebo group did not receive the placebo treatment, because of an urgent medical condition that occurred after the allocation. Two other patients, one in the nCPAP group and another in the MAD group, could not be reached after the random allocation and could thus not be evaluated. Hence, 57 patients completed the entire study protocol.

The patient characteristics at baseline are presented in Table 2. BMI was the only baseline characteristic that differed between the three therapy groups (F = 5.170; P = 0.008). LSD analyses revealed that the MAD group had a significantly lower BMI than the placebo and nCPAP groups (P = 0.002 and 0.006, respectively) [9]. The mean (±SD) baseline values of the different dimensions of the SCL-90-R of the three therapy groups, of the drop-outs, and of the normal values of the Dutch healthy population are also shown in Table 2. The baseline values of the different dimensions did not differ significantly between the three therapy groups (P = 0.305–0.987; Table 2). Further, the baseline values of the SCL-90-R of the drop-outs were not different from those of the therapy groups either (P = 0.348–0.997). The three groups showed higher average values of psychological distress at baseline than the reported normal values for the Dutch population in the dimensions “somatization,” “insufficiency of thinking and acting,” “anxiety,” “hostility,” “depression,” “sleeping problems,” and “global severity index” (T = 6357–2.566; P = 0.000–0.013; Table 2).

The mean (±SD) baseline values of the different dimensions of the SCL-90-R of the three therapy groups who completed the entire study protocol as well as the changes in these variables from baseline to therapy evaluation are shown in Table 3. As a result of missing values in the different dimensions of the SCL-90-R per therapy group, the number of observations used in the per-protocol analyses varied per dimension (see Table 3). The MAD group showed significant improvements over time in the dimensions “somatization,” “insufficiency of thinking and acting,” “agoraphobia,” “anxiety,” “sleeping problems,” and the “global severity index” (F = 4.01–15.47, P = 0.048–0.000, Table 3). These improvements in symptoms were, however, not significantly different from the improvements in symptoms observed in the nCPAP and placebo groups (P = 0.374–0.953). The intention-to-treat analysis showed similar results as the per-protocol analyses: the MAD group showed significant improvements over time in the dimensions “somatization,” “insufficiency of thinking and acting,” “agoraphobia,” “anxiety,” “sleeping problems,” and the “global severity index” as well (F = 4.01–16.34, P = 0.025–0.000), while these improvements were not significantly different from those observed in the nCPAP and placebo groups (P = 0.175–0.950).

The MAD group had used their appliance 90.6% (SD, 13.3) of the nights; the nCPAP group 82.9% (SD, 27.2) of the nights; and the placebo group 93.9% (SD, 15.7) of the nights. No significant group differences in compliance were found (F = 1.518, P = 0.228) [9].

There was no significant correlation between the baseline AHI value and the baseline values of the different dimensions of the SCL-90-R in the three groups (P = 0.121–0.888). A significant correlation was found between the baseline values of the “global severity index” and the changes in AHI values (∆AHI) between baseline and therapy evaluation in both the MAD and nCPAP group (P = 0.025). Patients with higher values of the “global severity index” at baseline showed less reduction in the AHI than patients with lower values of this index at baseline (Fig. 2). In the placebo group, there was no significant correlation between “∆AHI” and the baseline values of “global severity index” (P = 0.615).

Both MAD and nCPAP showed significant improvements of symptoms of psychological distress after 6 months of treatment. However, these significant improvements were not different from those observed in the placebo group.

In randomized clinical trials, there are often problems of noncompliance, where the patient does not adhere to the assigned treatment or does not complete questionnaires as we also observed in this study. Typically, this leads to estimates that can potentially be biased when the probability of a missing value is related to the characteristics of the patients. Further, missing data can also lead to a reduction of statistical power [19]. To overcome this problem, linear mixed model analysis can be used. The major strengths of mixed models are their ability to accommodate missing data points often encountered in longitudinal datasets and to generate valid study results [20]. Therefore, we used linear mixed model analyses in this study.

The population in the present study showed higher average values of psychological distress at baseline than the reported normal values for the Dutch population. The relationship between OSA and psychiatric disorders, especially depression, has already been studied for decades [21]. Pillar and Lavie [22] reported in their male population that neither the presence nor the severity of OSA was associated with depression or anxiety. On the other hand, recent evidence has confirmed important connections between OSA and psychiatric disorders. Psychiatric co-morbidity in OSA patients was examined in a large retrospective chart review of more than 100,000 veterans. A significantly higher prevalence of numerous psychiatric disorders, including depression and anxiety, was found in OSA patients as compared to non-OSA patients [23]. Others reported depression symptoms in 17–41% of OSA patients [24, 25]. Harris et al. [2] suggested that direct treatment of depression in OSA patients might improve acceptance of therapy, reduce sleepiness and fatigue, and improve quality of life, but that intervention trials are needed to answer this question. Although the causal relationship between symptoms of psychological distress and OSA has not been determined yet, a higher prevalence of these symptoms in OSA patient seems to be a consistent finding, which corresponds with our results.

All patients who completed the trial showed relatively high compliance rates of approx. 90% (i.e., the percentage of nights per week usage). This relatively high compliance may be explained by the fact that during the study period the patients frequently visited ACTA to be interviewed about the frequency of wearing. This regular contact with the examiner could have motivated the patients to use their appliances frequently [9]. Although self-reported compliance has been suggested to overestimate the actual use of MADs, covert compliance monitoring has shown excellent agreement between subjective and objective compliance [26, 27].

The MAD effects on the OSA condition have been compared with those of CPAP in several randomized clinical trials [4]. Although in most previous crossover studies MADs were considered less effective in reducing the AHI value than CPAP in mild-to-moderate OSA patients, similar improvements in subjective outcomes, such as excessive daytime sleepiness and quality of life, were found [8, 28‐30]. Further, it should be noted that these studies also indicated that, in general, patients find MADs as a more acceptable treatment compared to CPAP. In recent RCTs with a parallel design, no significant differences between MAD and nCPAP were reported in the subjective outcomes [9, 31]. Aarab et al. [9] found no significant difference in efficacy between MAD and CPAP in mild-to-moderate cases. Although Doff et al. [31] showed that CPAP was more effective in lowering the AHI than MAD in a group of mild to severe OSA patients, they found no significant differences between both treatments in the proportions of successful treatments. A recent meta-analysis showed that CPAP is more effective in lowering AHI than MAD in moderate-to-severe OSA patients; however, the superiority of CPAP over MAD is hypothesized to be less in mild cases [32]. Further, a recent crossover study by Phillips et al. [33] showed that important health outcomes were similar after 1 month of optimal MAD and CPAP treatment in patients with moderate-to-severe OSA. Thus, the outcomes of our study are in line with previous findings wherein both MAD and nCPAP show comparable treatment results in a group of mild-to-moderate OSA patients.

Beebe and Gozal [3] suggested that both intermittent hypoxia and sleep disruption induce dysfunction of the prefrontal regions of the brain cortex, which may predispose to mood disorders. Following this hypothesis, at baseline, we suspected a significant correlation between the amount of psychological distress and the AHI values. However, we did not find this correlation. On the other hand, OSA patients with higher values of the psychological distress at baseline showed less reduction in the AHI than patients with lower values of this index at baseline. The nature of this association is unclear, but this finding suggests that the level of psychological distress at the start of the treatment may play a significant role in the treatment outcome.

In this study, the significant improvements in symptoms of psychological distress in the MAD and nCPAP groups were not better than those observed in the placebo group. This is in line with our previous findings wherein we reported significant improvements in the Epworth sleepiness scale (ESS) and the Short-Form General Health Survey (SF-36) in all three groups without any differences in effects between the three therapy groups [9]. These placebo effects on excessive daytime sleepiness were also shown in a recent study, in which the effects of MAD were compared with an intraoral placebo splint in mild-to-moderate OSA patients [34]. Power calculation was performed for the primary outcome variable of this randomized placebo-controlled trial, viz., the AHI [9]. No power calculations were performed for the secondary outcome variables (viz., ESS, SF-36, and SCL-90-R). Therefore, our sample size per therapy group may have not been sufficient to find a significant difference between the therapy groups in the change of the different dimensions of SCL-90-R. However, our findings correspond with many previous, well-designed studies [7, 28, 35, 36] in which it was also reported that most of their OSA patients obtained a significant benefit in neuropsychological function and mood from their placebo treatment compared to MAD and CPAP treatments. These observed improvements in symptoms of psychological distress may be due to extensive attention given to the patients during the entire protocol, to a change in lifestyle as a result of the information given to the patients at baseline, and/or to a placebo response. Further, the high initial values of the SLC-90-R scores at baseline result in a higher possibility of decreases in these scores over time. The tendency of high values to return towards an individual’s more typical average state is known as “regression to the mean.” Stepnowsky et al. [37] reported in a recent study that baseline emotional distress predicted the drop in AHI in response to placebo treatment. Highly distressed patients showed a greater placebo response with a 34% drop in AHI. Although we could not confirm this in our study, all these findings together support the importance of including a placebo treatment in a randomized controlled trial design to determine unbiased treatment effects.

Within the limits of this study, it can be concluded that there is no significant difference between MAD, nCPAP, and an intraoral placebo appliance in their beneficial effects on symptoms of psychological distress in mild-to-moderate OSA patients. Further, this study suggests that the level of psychological distress at the start of the treatment may play a significant role in the treatment outcome of MAD and nCPAP in a group of mild-to-moderate OSA patients.