The Trial of Oral Mandibular Advancement Devices for Obstructive sleep apnoea–hypopnoea

In an adequately powered and efficiently designed randomised study, TOMADO showed that, for patients with mild to moderate OSAH, the rate of apnoea/hypopnoea events per hour for each of the three non-adjustable MADs studied was significantly lower than with no treatment. Although the effect on AHI compared with no treatment increased with sophistication of the MAD, the between-device differences were small and not statistically significant at commonly used thresholds. Arbitrarily defined response to treatment was achieved in just under half the patients when using the SP2 and bMAD and in approximately one-fifth when using no treatment, compared with baseline measurements. The likelihood of response was most closely related to BMI at baseline and during the study. Results for 4% ODI, which is more commonly used in clinical practice in the UK, were very similar to those for AHI.

The effects of MAD on ESS score mirrored those for apnoea/hypopnoea events per hour, with the SP2 and bMAD having a greater effect than the SP1.

Although the trial treatments were administered over a short time period, some evidence of compliance with treatment emerged. This indicated that one reason for the poorer performance of the SP1 may be lower compliance, as shown by the shorter duration of use per night and greater likelihood of discontinuation during the treatment period. The SP1 was also less likely to be chosen as the preferred device by those patients who completed the trial. Similarly, partners of the trial patients reported improvements in snoring during MAD use, with the SP1 having a weaker effect than SP2 and bMAD.

The relationship between MAD treatment and sleepiness-related functioning and QoL (FOSQ and SAQLI) showed a similar pattern to that for AHI and ESS outcomes, with significant effects for all MADs compared with no treatment, and the SP1 performing less well than SP2 and bMAD. Detailed examination of the questionnaires suggested small improvements across all dimensions, but that scales measuring the effect of sleepiness on activities (FOSQ, SAQLI), general productivity (FOSQ) and symptoms (SAQLI) were particularly affected by MAD treatment. General HRQoL measures were largely insensitive to MAD treatment, with the exception that the SP2 was associated with significantly higher SF-6D QALYs compared with control (accounting for baseline differences).

A range of secondary outcome measurements were taken and the general messages from these outcomes were consistent with those of the a priori stated primary outcome (AHI) and main secondary outcome (ESS score). Although these secondary outcomes were useful indicators of patient compliance and QoL and give a more complete picture of the effects of MADs, they should not be overinterpreted or be used in combination as an indicator of the strength of the effects of different MADs.

There were few SAEs during the study period and, out of the four SAEs reported by four patients, three were short-term, cardiac-related events. Two were considered possibly related to OSAH and one was considered possibly related to OSAH or MADs because the patient was on MAD treatment at the time. Almost all patients reported at least one minor AE, with mouth discomfort and excess salivation being the main problems.

The short treatment period meant little opportunity to observe an effect of MADs on RTAs or CVEs, which are the desired longer-term implications of control of EDS. However, patients did report improvements in sleepiness during driving and fewer interruptions to journeys during MAD use.

The trial-based cost-effectiveness analysis was also limited by the short treatment period, but the improvements in HRQoL for all MADs compared with no treatment meant that all were cost-effective at a WTP of £20,000 per QALY. The SP2 was the most cost-effective MAD up to a WTP per QALY of £39,800.

Thus, based on TOMADO, while all MADs have significant benefits and few harms compared with no treatment, the SP2 appears to achieve more benefits than the SP1 and almost all the benefits of the more sophisticated bMAD. However, it achieves these benefits at a lower cost than the bMAD and, so, it can be recommended on cost-effectiveness grounds.

Meta-analysis

Cochrane reviews from Lim et al.⁵¹ and Giles et al.,³³ and a meta-analysis from McDaid et al.,⁸ were updated for the major outcomes that were included in TOMADO. The systematic review identified 12 studies including 629 patients comparing MADs with CM, 13 studies including 746 patients comparing MADs with CPAP and 52 studies including 5400 patients comparing CPAP with CM, all of which had an AHI or ESS score as one of the study end points. Study participants were predominantly middle-aged men who were overweight or obese. Trials including CPAP were generally conducted in patients with more severe OSAH according to AHI than trials of MADs with CM. CM included a range of treatments including sham devices, sham CPAP, placebo tablets, lifestyle advice and no treatment. Although we included only randomised trials, quality was variable, with many trials having fewer than 50 patients and treatment periods were generally short. Both parallel-group and crossover trials have been used.

Heterogeneity between studies, assessed by the I² statistic, was variable and often unreliable as a result of the small number of studies available. Some heterogeneity could be explained, particularly by baseline severity, but there remained unexplained heterogeneity. Although random-effects methods were used, the validity of combining trials in formal meta-analysis is questionable and cautious interpretation is required. Partly for this reason a network meta-analysis including all trials was not attempted.

Both MADs and CPAP resulted in significant improvements in AHI, with the greatest benefit evident in trials of CPAP against CM. However, the reduction in AHI was strongly related to baseline AHI, which is natural since a higher baseline allows greater scope for an absolute decrease. In head-to-head trials of MADs against CPAP, the performances of the two treatments were more similar and there were no head-to-head trials in patients with mild-range AHI.

Excessive daytime sleepiness assessed by the ESS is less variable than AHI so most trial populations were classed as having moderate baseline EDS. The differences between the effects of MADs and CPAP on subjective daytime sleepiness assessed by ESS were smaller and not significant in head-to-head comparisons. The estimated effects on EDS were strongly related to baseline severity and, to a lesser extent, baseline AHI. When trials of similar baseline characteristics were compared, there was little difference between the effects of MADs and CPAP on post-treatment ESS score when assessed against CM, and this is reinforced by the results from head-to-head trials. Treatment effects appeared to be stronger in trials with short duration of treatment, possibly reflecting a tailing-off of compliance over time.⁴⁵

The meta-analysis did not provide much insight into the effect of treatment on daytime BP above previous meta-analyses. There was a large amount of heterogeneity in the methodology used for assessing surrogates of cardiovascular outcomes. Our meta-analysis focused on daytime SBP and DBP because it was included as the primary marker of hypertension in TOMADO and because it is used in the Framingham equation that provides input into the long-term economic model. There was a small effect of both CPAP and MADs on SBP compared with CM, with little to choose between the two.

Few trials, apart from TOMADO, have contributed to the literature on HRQoL so that it was difficult to draw reliable conclusions. In common with TOMADO, there was evidence for a significant improvement in HRQoL as a result of these treatments in the meta-analysis, but the size of the effects is unlikely to be clinically important. The paucity of information did not allow more detailed analysis of published HRQoL. Given the demonstrated clinical effectiveness of both CPAP and MAD, further trial-based studies of HRQoL are unlikely to be conducted, but observational studies to supplement existing trial data would be useful.

Cost-effectiveness

In order to assess the effect of CPAP and MADs on long-term outcomes, including cardiovascular hazards and RTAs, we reviewed and updated an economic model provided by the University of York Centre for Health Economics, developed for McDaid et al.⁸ The model inputs were adapted to better represent patients with mild to moderate OSAH and updated to reflect new research since the original model was developed. Systematic searches of published literature were undertaken to update model inputs related to CHD and stroke risk, RTA rates, HRQoL and costs. In addition, data from TOMADO were used for device-specific costs and to create a more precise mapping function between ESS score and utility measures (both EQ-5D-3L and SF-6D) that would also be more applicable to patients with mild to moderate OSAH.

In the base case, using the SP2 as the ‘standard’ device, MADs were found to be more costly and more effective than CM in patients with mild to moderate OSAH, with an estimated ICER of £6687 per QALY. Compared with MADs, CPAP was more costly and more effective, with an estimated ICER of £15,367. While it was clear that both of these treatments were better than CM, there was substantial uncertainty in the choice, with probabilities of being cost-effective at a WTP of £20,000 per QALY of 47% for MADs and 52% for CPAP. Corresponding figures at a WTP of £30,000 per QALY are 45% for MADs and 55% for CPAP.

The results were sensitive to a number of parameter inputs. If the average lifespan of the SP2 is increased from 12 months to 18 months, the ICER for CPAP compared with MADs becomes £44,066, which is more than traditionally accepted WTP thresholds. Additionally, choice of device has an important effect on the economic decision, with the ICER for the SP1 compared with CM being £1552, and for the bMAD compared with CM being £13,836. The ICER for CPAP compared with the SP1 is high at £89,182, but CPAP is both cheaper and more effective than the bMAD. Using device-specific inputs for treatment effects further confirms the superiority of the SP2 as the most cost-effective treatment for patients with mild to moderate OSAH, although substantial decision uncertainty remains. Differential compliance rate for CPAP also reduces its cost-effectiveness so that MADs become both less costly and more effective if compliance to CPAP is of the order of 90% of SP2.

Strengths

The TOMADO study was a relatively large and rigorously conducted RCT, with robust and precisely estimated treatment effects. To our knowledge, TOMADO is the first trial of MADs in mild to moderate OSAH with both clinical, patient-centred and cost-effectiveness outcomes. The interpretation of results is clear and consistent among different outcome measures. In contrast with most other published randomised trials, TOMADO included a detailed study of HRQoL. This showed consistency with clinical outcomes and highlighted the effects of MADs on activity, general productivity and symptoms. Although these effects might be described as modest, it is remarkable that they can be observed after a short period of treatment. TOMADO fed into updated meta-analyses that offered stronger insights into the relative effectiveness of MADs and CPAP in patients with OSAH. In addition, the effects of baseline severity have been highlighted and used to explain some of the differences in effects between MADs and CPAP. TOMADO also fed into an updated model of the long-term cost-effectiveness of MADs and CPAP devices which was adapted, for the first time, to mild to moderate OSAH.

The study was applicable to general sleep practice as it recruited participants who had been referred from primary care to the sleep clinic at Papworth Hospital. The SP1 and SP2 devices used in the study are available in many countries and are similar to other thermoplastic and semi-bespoke MADs on the market. Although the bMAD was fitted and manufactured by a hospital maxillofacial laboratory, it was done using skills, materials and facilities common to dental sleep services.

Limitations

Women, younger patients and patients with a BMI in the normal range were under-represented in the patients included in TOMADO and other trials so that results may not be generalisable to these populations.

In evaluating three non-adjustable MADs representing a range of sophistication and cost, TOMADO could not also include an adjustable MAD (aMAD). These are increasingly recommended,^202,203 but are often more costly. They allow gradual titration of mandibular advancement according to tolerance and efficacy. This may give larger treatment effects by achieving ultimately greater jaw protrusion without lowering compliance, but whether or not aMADs are more effective than non-adjustable MADs remains unproven.

The aim for the bMAD was at least 50% maximal protrusion, but in practice this value was often lower; and similar to that achieved independently by patients with the other devices. This reflects the pragmatic nature of TOMADO, making its findings more applicable to the wider NHS. AHI effects have been shown to be proportional to mandibular protrusion.^204–206 Mean (SD) protrusion in this trial was between 52.5% (27.8%) of maximal advancement with the SP2 and 63.4% (22.6%) of maximal advancement with the SP1. These figures are lower than reported in other studies,^69,74,76,83 many of which used an aMAD.^{23,52,68,72,75,79,81,84} Nevertheless, although heterogeneity limits comparison, many of these trials did not report greater AHI effects than TOMADO.^23,74,76,83 Furthermore, TOMADO showed no association between protrusion and AHI effects, adding to existing evidence that greater protrusion may be no more effective in milder OSAH. For example, Tegelberg et al.²⁰⁷ compared patients with mild to moderate OSAH who had devices at 50% and 75% maximal protrusion and found no difference in AHI effects. Quality studies comparing aMADs to non-adjustable devices are lacking. A small, non-randomised trial compared an aMAD with a thermoplastic MAD and found a modest difference in AHI favouring the adjustable device.²⁰⁸ However, the sample size was small and the differing costs of the two devices (paid for by the patients) probably influenced device selection. A retrospective study of 805 patients demonstrated a small but statistically significant difference in AHI between an adjustable and non-adjustable device (7.6 vs. 10.0, respectively), but did not show a significant difference in ESS score or tolerability.²⁰⁹ This study was also limited by device selection which was non-randomised. Other studies which have featured both adjustable and non-adjustable MADs have reduced the likelihood of finding real-life effect size differences by using similar or identical protrusions for both devices. Therefore, whether or not adjustability improves MAD effectiveness in OSAH remains uncertain and requires rigorous RCT evaluation.

In the meta-analysis and because of the economic decision analysis, all MADs were considered as a single treatment modality. There were too few studies to allow robust subgroup analyses and so we were unable to identify the more modest differences in effects between different MADs. It has been suggested that future meta-analyses distinguish between trials of non-adjustable MADs and those using aMAD.^209,210 From the 2006 Cochrane analysis, considering only trials comparing CPAP with aMADs moved the sleepiness (ESS score) effect size in favour of MADs, but not significantly.³³ We considered performing a similar subgroup analysis when updating the meta-analysis. However, device adjustability could not always be determined,^80,82 and the potential advantage of titratable advancement was sometimes negated by the use of uniform aMAD protrusion.⁸³ Classifying trials as fixed MADs and aMADs, in order to perform separate meta-analyses, is not straightforward. For example, one trial with relatively weak treatment effects that has previously been excluded from aMAD reviews²² used two non-adjustable MADs, but the authors reported near-maximal (80%) jaw protrusion and performed ‘pseudotitration’ by adapting devices to optimise comfort and benefits. For these reasons we did not include a meta-analysis based on MAD adjustability.

Three separate meta-analyses were conducted comparing MADs with CM, MADs with CPAP and CPAP with CM. A more sophisticated analysis would have been to combine the studies into a network meta-analysis, thereby adding strength to all comparisons and better aligning the studies. However, these analyses rely on the associative law, which was unlikely to be true in this case, given the greater severity of OSAH in populations undergoing trials of CPAP. Furthermore, the heterogeneity observed between studies in Chapter 3 suggested that combining results within and across different treatments may not be sensible. The likely implication of doing separate meta-analyses is a loss of some precision in the results.

Conservative management encompassed a wide range of control treatments so that their influence on the trial-based treatment effects was difficult to estimate with any precision.

In our systematic review we used the Jadad score as a measure of study quality in order to be consistent with previous reviews.^8,33,51 This is a rather insensitive tool and did not provide substantial insight into the relative quality of different studies. It did, however, provide a broad structure for summarising design features reported in existing clinical trials.

The use of data in the model still reflects the lack of robust sources in some important areas. RTA risk after treatment with MADs is still inferred based on ESS score treatment effects compared with CPAP, rather than using direct data. While the link between OSAH, hypertension and cardiovascular outcomes may be increasingly understood, the treatment effects in the model are still based on short-term BP data rather than long-term CVD outcomes. Given the similarity in ESS score, treatment effect pooled from the meta-analysis, the importance of compliance and determining how prolonged the effects are, is clear but there remains a lack of good data to reflect this for MADs. Only crude sensitivity analyses were able to explore the effect this has on cost-effectiveness.

Implications for service

• CPAP remains the most clinically effective and cost-effective treatment for patients with moderate to severe OSAH based on reduction in AHI. For patients who are intolerant of CPAP, treatment with a MAD is also effective compared with no treatment.
• CPAP and MADs are equally effective treatment options for patients with mild to moderate OSAH and there is little to choose between them in terms of clinical effectiveness and cost-effectiveness, although this is contingent on similar compliance rates.
• Of the three MADs investigated, the semi-bespoke SP2 (or an equivalent MAD) is the most cost-effective treatment in the short term and should be used as the first-choice device, with the custom-made bMAD reserved for patients who have difficulties producing the SP2 mould, whose dental eligibility is more marginal or for whom other obstacles to using the SP2 may be overcome by dental intervention with a bMAD.

Implications for research priorities

• Head-to-head pragmatic clinical effectiveness and cost-effectiveness comparisons of adjustable and non-adjustable MADs, across the entire range of OSAH severity, are still required.
• Head-to-head comparisons of CPAP and MADs in milder OSAH, would reduce the uncertainty surrounding the current guideline stance that CPAP should be reserved as second-line treatment in this patient group.
• There is increasing evidence to suggest that the similar effects of CPAP and MADs on EDS may be as a result of differential adherence to treatment. However, there is limited information on this beyond short-term trials. Medium- to long-term compliance with MADs and CPAP should be monitored and reported. Such work would be strengthened by emerging tools to objectively monitor MAD compliance, which would benefit from further clinical and additional economic evaluation in their own right. Observational studies of HRQoL over time to supplement existing trial data would be useful to understand the treatment outcomes of greatest relevance to patients. In particular, it would be useful to know more about the durability of devices.
• Further data on longer-term risk of CVD and its risk factors would reduce model uncertainty and improve the precision of estimates of clinical effectiveness and cost-effectiveness.