NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Balk EM, Chung M, Chan JA, et al. Future Research Needs for Treatment of Obstructive Sleep Apnea: Identification of Future Research Needs From Comparative Effectiveness Review No. 32 [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2012 Feb. (Future Research Needs Papers, No. 12.)
Future Research Needs for Treatment of Obstructive Sleep Apnea: Identification of Future Research Needs From Comparative Effectiveness Review No. 32 [Internet].
Show detailsComposition of Stakeholder Panel
Through our multipronged approach to compiling the stakeholder panel, we enlisted a total of 27 stakeholders, of whom 22 were on the treatment panel. Table 2 summarizes the number and types of stakeholders on the panel.
Table 2
Stakeholders panel for treatment of sleep apnea.
Stakeholder Participation and Required Methods Modifications
As noted above, this was the first instance of the Tufts Evidence-based Practice Center (EPC) producing a Future Research Needs (FRN) document with a large number of stakeholders. As such, the process was to a large extent a test of the success of the specific protocol we envisaged. Here we describe the substantive changes that we had to make to the protocol methods.
Overall, the use of the Sharepoint Web site was only moderately successful. The majority of stakeholders were slow to sign on to the Web site, were slow to begin reviewing the ongoing discussions, participated only minimally in discussions, and did not offer new FRN topics. Furthermore, a small number of participants were not able to log in to the Sharepoint Web site because of technical difficulties.
There were 30 FRN topics discussed and prioritized; 14 of these were added by the EPC based on the Comparative Effectiveness Review (CER) Future Research section; 7 were added online (or via email to the EPC) by stakeholders; 9 were added by the EPC after a teleconference with the consumer advocates (see below).
The discussion period was started 12 days after we received signed disclosure of interest forms from almost all the stakeholders (with the exception of the patient stakeholder who we were delayed in recruiting). All participants were contacted 5 days before the online discussion board was made available. We informed them of the impending start of the online discussions as confirmed that their email addresses would be available to them during the course of the discussion. The original plan was for the stakeholder discussion to last 2 weeks; in reality, the discussion period was extended to 27 days to allow further stakeholder involvement. Six of 22 stakeholders signed in to the Web site within 2 days. The median time until stakeholders signed in was 4.5 days. Six stakeholders did not log in to the online discussion site at all. The last stakeholder signed in on day 23 of the discussion. Across topics, discussions were started by only five stakeholders; nine stakeholders participated in the online discussions (added any comments). Not including comments or questions added by the EPC, for the 30 topics discussed online, the median number of comments by stakeholders was 2, ranging from 0 (for 11 topics) to 7. After the discussion period ended, 17 of 22 stakeholders participated in topic nomination (which was conducted by email instead of on the Web site, as originally planned; see below).
To improve participation, after 1 week of discussion, we invited all stakeholders (from both panels) to participate in teleconferences where the EPC reviewed the materials that had been sent to them by email and that were available on the Web site, including the goals of the project, the stakeholders’ responsibilities, the main criteria for selecting and discussing FRN topics, and a review of how to use the Sharepoint Web site. Two calls were scheduled, for which 13 of 27 stakeholders joined (from both diagnosis and treatment panels). To increase participation, we also sent numerous email reminders, offered to have stakeholders email us their comments that we would add to the discussion board, and also answered phone queries on technical issues related to the Web site. Furthermore, because several stakeholders could not easily access the Web site (primarily because of overseas travel), we compiled plain text versions of all the FRN topics and discussions, which we placed directly into the text of emails (not as attachments). This was emailed to all stakeholders.
Near the true end of the discussion period, we noted that the consumer stakeholders (the patient, the patient advocate, and the representative from the transportation industry) had not participated in the discussion. Upon reviewing the discussions to date, we thought that they might have been too scientifically technical for the lay stakeholders. We contacted them directly, had our suspicions confirmed, and organized a separate teleconference for the three of them (with the EPC). This 1.5-hour teleconference produced nine new topics. The EPC organized the discussion into FRN topics, summarized the separate discussions, sent the summary to the lay stakeholders, and then uploaded the topics and discussions to the Web site. A summary of the call was also emailed to all stakeholders. No stakeholders added further to these topic discussions.
After the discussion period was closed, the EPC decided that it was not worthwhile to attempt to use the Web site to have stakeholders nominate the topics. Instead this was done by email. The final list of topics and the text of the discussions were emailed to stakeholders. The order of the topics was randomized once, instead of separate randomization for each stakeholder, since the randomization process was too time consuming. (We chose to maintain links between the list of topics and the topic discussions, rather than manually reorder each stakeholder list.) To prevent procrastination and further delay, we asked stakeholders to nominate the topics within 4 days, by Friday, with the expectation that responses would be in by Monday. Eight of 22 stakeholders sent in the topic nominations by Monday. It was likely that some of the delay in response was due to a weather-related phenomenon (Hurricane Irene) so we extended the timeline and sent out a reminder email after 1 week. Two of remaining 14 stakeholders sent in their nominations after receiving the email. Subsequently, followup phone calls were made to the remaining stakeholders to solicit their nominations. The final list of nominated topics was received the following Monday. Topics were nominated by 17 of 22 stakeholders. Stakeholders were asked to nominate up to 10 topics; they nominated between 4 and 10 topics each.
Research Needs
Based on the methods described above, we organized the FRN topics into three categories (see Methods section): High priority FRN topics; Second-tier priority FRN topics; and Other FRN topics. Topics that did not meet minimum requirements for inclusion are not presented. The FRN topics are as listed in Table 3.
Table 3
List of future research needs topics.
High-Priority Future Research Needs Topics
High-Priority Future Research Needs Topic 1
What is the impact of treatment of sleep-disordered breathing on major long-term clinical outcomes, including mortality, cardiovascular disease, and diabetes?
a. What are long-term outcomes of mandibular advancement devices (MAD) treatment?
Background
This FRN topic represents two topics that were nominated and discussed separately by the stakeholders. After nomination of topics by the stakeholder panel, the topics were combined into one overarching topic with different components.
One of the main research needs identified by the CER on treatment of sleep apnea was the use of clinical outcomes in comparative studies. Only 3 of 190 studies reported clinical outcomes. However, nonrandomized comparative studies of various treatment modalities were not reviewed for an assessment of their impact on outcomes. It was recommended that future research should focus on comparative studies with long-term followup and clinical outcomes. This was also an opinion echoed by the stakeholders on the FRN panel.
Stakeholder Discussion
The first nominated topic suggests that clinical outcomes should be the outcome of interest in comparisons of various treatments for sleep apnea, and the second topic focused on long-term outcomes for mandibular advancement devices (MAD). For assessing long-term clinical outcomes, the stakeholders’ suggested study design was a database analysis from a prospective cohort study like the Framingham Heart Study or Nurses Health Study. For assessing clinical benefits from use of MAD, a randomized trial or comparative study was the suggested study design with MAD serving as the intervention arms and continuous positive airway pressure (CPAP) or other oral devices serving as the comparator arm.
Proposed Study Designs
To address the overarching FRN topic and its subcomponents, different study designs are reasonable and would address different aspects of this topic. In assessment of comparative effectiveness, the two overarching comparisons of interest would be autotitrating (and other nonfixed) CPAP versus fixed CPAP and CPAP versus MAD. Sham or no treatment would not be a practical or ethical treatment option for a long-term study given the known effect of CPAP and MAD on sleep measures and symptoms. The clinical outcomes of interest are mortality, cardiovascular disease, and diabetes, including surrogate or intermediate markers of the latter two conditions.
Randomized Controlled Trials
Value of Study Design
Only one of 89 trials studied clinical outcomes. Any future trials of CPAP or MAD should study clinical outcomes. However, the resources required for the necessary long-term clinical outcomes would be great and loss to followup would be a major concern. Given the poor compliance of many patients with obstructive sleep apnea (OSA) treatments, it would also be likely that over the long term, many patients would cross from one treatment to another (and back again), complicating any analyses of these trials.
Resource Use, Size, and Duration
The primary analysis of interest in such a study would be the comparative effectiveness of various treatments on mortality, cardiovascular disease, and diabetes. These could result in the use of large resources in terms of cost, time and effort, and the decision to use these resources for conducting trials has to be balanced against the value of information that can be gained from studying clinical outcomes. While a study of mortality would require large sample sizes and long periods of followup, cardiovascular disease and diabetes could be followed up using both surrogate markers as well as hard clinical outcomes. For example, for the comparison between CPAP and MAD, the analysis of interest could be the effectiveness of CPAP and MAD on measures of glucose metabolism and insulin insensitivity, as markers of impaired glucose tolerance and incipient diabetes. The number of patients that would need to be enrolled in a trial would depend on the relative effect of the interventions and the event rate for the clinical outcome in the comparator arm. Regardless of the relative difference in effects between groups, trials using mortality as an outcome would need a sample size in the thousands, while using surrogate outcomes would need smaller sample sizes.
To evaluate sample size for trials, we used the data from trials between MAD and CPAP to conduct a power calculation using standard formulae for a two -sided equivalence test. Given the lack of any data to suggest an expected relative effect of treatment on a composite cardiovascular outcome, we assumed that it would be equal to the effect seen in trials for improvement in apnea-hypopnea index (AHI)—greater than 50 percent reduction in the AHI to less than 5 events/hr). We set the range of relative risk reduction that was seen in the report, which ranged from 0.6 to 0.9. From the CER, we also obtained the event rates of outcomes in the control arm (in this case, the control arm was CPAP) which was 0.7 to 0.8. The required total sample size (1:1 ratio in a two-arm randomized controlled trial [RCT]) for each value of relative risk and the control rate is presented in Figure 4.
Figure 4
Sample size calculations for an RCT of MAD versus CPAP with varying estimates of risk ratios and control rates for combined cardiovascular outcomes.
The trial duration and followup for mortality as an outcome would be measured in years, while trials focusing on surrogate outcomes for cardiovascular disease and diabetes can be relatively short term, for example 6 to 12 months.
Ability To Recruit
Thousands of patients have already been recruited into OSA treatment trials. Patients are usually sufficiently symptomatic that they are likely to be interested in treatment. There are numerous treatment options, without clear guidance as to who should receive which treatment. However, these trials have used sleepiness and laboratory indicators like sleepiness scores and AHI as outcomes. When considering clinical outcomes, there is a potential for a large loss in followup. This would necessitate a much larger recruitment population, with an extended recruitment period, which may present as a resource constraint.
Ethical Issues
The primary ethical issue (beyond the standard ethical issues in conducting a randomized trial) pertains to whether there is equipoise. An argument can be made that another trial that evaluates only CPAP or MAD for the same sleep-related outcomes that the existing 89 trials have evaluated may be unethical since the value of the treatment is known and the trial would not change the conclusions from the body of evidence. However, if a new trial truly addresses a question that is inadequately addressed, such as how these treatments would affect objective clinical outcomes and to what degree, then there is adequate equipoise.
Post Hoc Analyses of Existing Observational Studies
Value of Study Design
A second-tier option for using existing data to evaluate the effects of different patient characteristics would be to perform post hoc regressions or subgroup analyses within observational cohort studies like the Framingham Heart Study, Wisconsin Sleep Study or Nurses Health Study. The CER did not include observational study designs when evaluating the comparative effects of different treatments. It is not clear whether the information on treatments for sleep apnea is captured in adequate detail in these studies. It is known that the outcomes of interest, that is, mortality, cardiovascular disease and diabetes, have been well documents in the study populations. However, if the required information is available, a retrospective analysis of impact of sleep apnea treatments would be invaluable in providing insight. However, such post hoc analyses may be susceptible to type I error (falsely significant associations) due to multiple testing Nevertheless, they would form a good basis to determine which treatments and outcomes could be further investigated a priori in future trials. This analysis could be performed either by the original study investigators or other researchers with access to the study data.
Resource Use, Size, and Duration
Post hoc analyses of existing data can be done quickly with modest resources.
Ethical Issues
There should be no ethical issues involved in reanalyzing existing study data, provided that consent has been obtained from study participants to use their anonymized data for future studies.
Observational Studies
Value of Study Design
Though less informative than randomized trials for the evaluation of comparative effectiveness, observational studies would be less resource intensive in assessing impact of various treatments. Case-control studies of patients treated for OSA, with or without a clinical outcome (e.g., a myocardial infarction), could also address this question, though they would be subject to all the biases and other problems of any case-control study. Efforts to mimic a random assignment in prospective observational studies, like propensity score matching, could play a part in a priori design of an observational study as well as in the post hoc analysis of the data from these studies.
Resource Use, Size, and Duration
Retrospective data could be gathered quickly and easily by chart review or similar approaches. Prospective cohorts would take somewhat more time and effort, but would still be less resource-intensive than a trial. Since outcomes generally change rapidly (within a few weeks or months) after the start of treatment, data could be collected fairly rapidly, whether retrospective or prospective. The only exception to this rule occurs when mortality is the outcomes, as the timeframe is then extended to years. A retrospective study is more practical when this is the case, even though it is susceptible to significant bias. The sample size will depend on the estimated relative effect of the treatment, and the event rate in the comparator arm These studies are better suited for mortality outcomes as it is less resource intensive to follow up patients for a period of a few years.
Ability To Recruit
Patient recruitment should be straightforward and relatively simple, as there would be little added burden for them by entering a study, beyond a formalized assessment of symptoms, other outcomes of interest, and patient characteristics, all of which are reasonable parts of normal patient care.
Ethical Issues
There should be no substantive ethical issues involved in these observational studies.
High-Priority Future Research Needs Topic 2
Cost-effectiveness analysis of a management strategy (diagnosis of symptomatic or high-risk patients through treatments of patients diagnosed with OSA), specifically for patients with mild to moderate disease severity
a. Research on CPAP devices that are both economical as well as clinically effective
Background
This FRN topic represents two topics that were nominated and discussed separately by the stakeholders. After nomination of topics by the stakeholder panel, the topics were combined into one overarching topic.
The topic as proposed covers both diagnosis and treatment strategies for a complete cost-effectiveness analysis. Since diagnosis and treatment FRN topics are being dealt with separately, here we focus primarily on the treatment strategy component of the cost-effectiveness analysis. A discussion of a cost-effectiveness analysis of treatment strategies can be found in the companion report.3 This FRN is distinguished from FRN 3 (below) on three dimensions: (1) it assesses the expected value (not just benefit) of selected treatments for OSA; (2) it is designed to address these questions for patients with mild to moderate disease, an understudied group; and (3) it proposes to assess expected value from both the societal and patient perspectives. Out-of-pocket expenses for patients with mild to moderate disease were of considerable concern to patients in our stakeholder group. This FRN seeks to address that concern.
It would likely be of value for future cost-effectiveness analyses to incorporate both phases of OSA management (diagnosis and treatment). Cost-effectiveness analysis allows the comparison of different interventions on similar benefit, cost and utility scales. If clinical trial data are available on the treatments and populations of interest, benefit and cost estimates have both internal and external validity.
Stakeholder Discussion
The objective of this research need is to establish better evidence about the costs and benefits of alternative management strategies for individual patients with mild to moderate disease severity. Ultimately, the goal is to develop evidence about care management from efficient testing high-risk individuals, accurate and low-cost diagnosis of OSA, through effective and low-cost treatment of patients. Issues that should be considered in a cost-effectiveness model include the patient-related outcomes, including measures of functional status such as productivity (both absenteeism and presenteeism—working in spite of illness, with resulting poor work performance), quality of life, and work safety. The feasibility of such an analysis will depend on combining clinical trial results with health insurance data on health care utilization, employer data on absenteeism, and potentially auto insurance data on motor vehicle accidents in a large cohort of patients with OSA. Other issues that should be considered are the patients’ costs for using OSA treatments. An example of such a cost, which would not be included in most analyses, was the cost to a long-haul truck driver who requires extra battery backup since a CPAP device can drain the truck’s battery overnight.
Proposed Study Designs
Systematic Review
Cost-effectiveness analyses were not addressed by the CER. Conducting a systematic review may be the first step to ascertain the level of existing evidence.
Cost–Benefit Analysis
On the basis of the best evidence identified in the CER, a quality-adjusted cost-benefit analysis is recommended, comparing the incremental costs and benefits of different treatments to each other and to a base case of no active treatment. Benefit, utility, and cost estimates may be derived from previous clinical trial data, where available, and from observational data where trial data are not available. These estimates should include not only the standard clinical outcomes, but also work-related, accident, and quality of life outcomes. In the absence of cost estimates, charges may be derived from administrative data and adjusted cost-to charge ratios. Out-of-pocket patient costs should also be included.
Analytic Approach
Preference should be given to a Markov-chain, discreet events analysis. The proper outcome measure is quality-adjusted life-years (QALYs) gained. Future QALYs and costs should be discounted over the follow-up period. Adjustment should be made for major outcomes and adverse events, using rate and utility estimates. Probabilistic sensitivity analyses should be conducted on estimated outcome rates, utilities and cost estimates. A societal perspective should be assumed in the main analysis. The patient perspective should be assumed in a subanalysis, given high out-of-pocket costs for devices and other treatments, substantial concerns about adherence (due to inconvenience and discomfort of treatment), and the effects of OSA on work and lifestyle.
Interventions
Treatment strategies should be tested both with and without the involvement of a sleep medicine specialist, where possible. Phased treatment combinations should also be considered. A short list of treatment strategies to be compared could include
- No treatment
- Fixed CPAP
- Fixed CPAP followed by alternative devices if fixed CPAP fails
- Initial use of alternative CPAP devices (including autotitrating CPAP)
- CPAP with interventions to improve compliance
- CPAP followed by MAD, if necessary
- MAD alone
- MAD with interventions to improve compliance
- MAD followed by CPAP, if necessary
- Bariatric surgery, when indicated
- Surgery, when indicated
Numerous other treatments and combinations of treatments are also used, including strategies that base initial treatment choice on patient and disease characteristics.
Unfortunately, as the CER highlights, the evidence for the effect of most of these treatments on any patient-related outcomes is of low strength or insufficient. Thus, the sensitivity analyses may be of greatest value, by describing how effective an intervention would have to be to be a cost-effective option for patients with OSA.
Resource Use, Size and Duration
Because a cost-effectiveness analysis can draw from previously collected data, the cost, size and duration of such studies can be limited. However, as discussed by the stakeholders, it may be challenging to gather all the relevant data.
Ethical, Legal and Social Issues
No new data collection is proposed and therefore the direct risk to patients is minimal.
High-Priority Future Research Needs Topic 3
Comparative studies of different sleep apnea treatments based on patient characteristics
a. Analyses of CPAP stratified by disease severity
b. Analyses of non-CPAP treatments stratified by disease severity
c. Comparison of alternative treatments for patients who do not tolerate CPAP
Background
This FRN topic represents four topics that were nominated and discussed separately by the stakeholders. After nomination of topics by the stakeholder panel, the topics were combined into one overarching topic with different components.
Key Question 5 in the CER asked about the comparative effect of different treatments. The two subquestions asked about whether the comparative effect varies by different presenting characteristics. Specifically:
- Does the comparative effect of treatments vary based on presenting patient characteristics, severity of obstructive sleep apnea, or other pretreatment factors? Are any of these characteristics or factors predictive of treatment success?
- Characteristics: Age, sex, race, weight, bed partner, airway, other physical characteristics, and specific comorbidities
- Obstructive sleep apnea severity or characteristics: Baseline questionnaire (and similar tools) results, formal testing results (including hypoxemia levels), baseline quality of life, positional dependency
- Other: specific symptoms
- Does the comparative effect of treatments vary based on the definitions of obstructive sleep apnea used by study investigators?
For all comparisons between treatments, the CER concluded that the strength of evidence is insufficient to determine which patients might benefit most from treatment in general (i.e., compared to no treatment) or from which treatment (from direct comparisons of different treatments). Trials of sleep apnea treatments (CPAP, surgery, MAD, and drugs) do not compare the effectiveness of these treatments in different patient subgroups that are based on obesity, sex, and other patient characteristics.
Stakeholder Discussion
The overarching theme of the discussion in the CER and among the stakeholders was that there is a need for subgroup analyses to help clinicians base treatment decisions on baseline characteristics to maximize early effective, tolerable treatment, to better match treatments to specific patient types, and to minimize costly trial and error. Two major categories of studies could be performed to address this question: (1) comparisons of treatments specifically within narrow subpopulations of patients, or (2) subgroup analyses of large comparative studies. The discussants agreed that it would be a better use of research resources to conduct large RCTs that recruit a wide range of patients and then to perform prespecified subgroup analyses. Specific baseline patient characteristics (or subgroups) of interest among stakeholders included: race, sex, obesity level (body mass index), age group, disease severity, morphometrics (e.g., measurements of the jaw), and interactions within these categories (e.g., obese vs. nonobese in blacks and whites separately).
A different approach may be warranted for evaluation of effectiveness of CPAP (versus no treatment) for disease severity. It was noted that trials of CPAP for severe OSA have already been done and convincingly show improvement in short-term and sleep outcomes. Further trials are not necessary to show the value of CPAP (compared to no treatment) for this group of patients. In contrast, additional studies are needed to assess the short- and long-term value of CPAP in patients with mild to moderate OSA severity. This caveat, though, does not hold for comparisons of non-CPAP interventions to control, where further trials are necessary for all groups of patients. However, the CER found that MAD was effective at improving symptoms and sleep study measures, primarily in patients with moderate to severe OSA (as defined by AHI). Furthermore, trials comparing different active interventions should include a wide range of patients based on OSA severity to allow evaluation of the relative effectiveness of different treatments across the range of disease severity.
Also of interest, are future studies to delineate the most effective treatment in the subgroup of patients who have not tolerated CPAP. To date, the large majority of trials either explicitly or implicitly included only patients newly diagnosed with OSA. Thus, the current evidence may not be fully applicable to patients who have tried and failed CPAP. Related questions of interest include comparisons of different types of CPAP (e.g., of different alternatives to fixed CPAP), evaluation of different adjunctive treatments (e.g., humidification), and comparisons of different non-CPAP treatments. In particular, it was noted that oxygen via nasal prongs, as a standalone treatment, is commonly prescribed by physicians for patients who do not tolerate CPAP, but this treatment has not been adequately tested. Of note, oxygen alone as a treatment was not evaluated in the CER.
Proposed Study Designs
To address the overarching FRN topic and its subcomponents, different study designs are reasonable and would address different aspects of this topic.
Patient-Level Meta-analysis
Value of Study Design
For comparisons of several of the treatments (particularly CPAP, autotitrating CPAP, and MAD), there were adequate trials to allow meta-analysis. However, meaningful conclusions about the relative effects of treatments on different subgroups of patients could not be ascertained from these meta-analyses because of the risk of ecological fallacy—erroneous conclusions based on ascribing the study mean estimate (e.g., mean AHI) to all participants in a study. But the large body of existing studies could be used to evaluate subgroup effects, without the risk of ecological fallacy, if patient-level meta-analyses could be conducted.
Briefly, typical meta-analysis combines the mean or overall estimates of effects from different studies. In contrast, patient-level meta-analysis uses the same patient-level data available to the original study investigators from multiple trials, and combines these. This approach allows unbiased regressions and subgroup analyses, as could be conducted in any individual trial. However, acquiring the data is the usual main obstacle to conducting patient-level meta-analyses.
The outcomes of interest (for this and all study designs discussed here) will be those outcomes that have been commonly obtained and which clinicians use to determine treatment effectiveness. These will primarily include AHI (both continuous and dichotomized [e.g., improvement by a certain amount or improvement to below a certain threshold]), other sleep measures, and symptoms including the Epworth Sleepiness Scale. Other outcomes of interest will be compliance and adverse events or discomfort.
Resource Use, Size, and Duration
If a researcher, funder, or other entity can convince the principal investigators of the largest, best CPAP and MAD trials to share their patient-level data for the purpose of meta-analysis, most of the patient characteristics could be quickly evaluated, with modest resources, regarding their association with effects.
Ethical Issues
Meta-analysis uses existing data. So long as the principal investigators agree to the use of the original data, there should be no ethical issues.
Post Hoc Analyses of Existing Trials
Value of Study Design
A second tier option for using existing data to evaluate the effects of different patient characteristics would be to perform post hoc regressions or subgroup analyses within already-published trials. This could be performed either by the original study investigators or other researchers with access to the study data. To maximize accuracy and validity, such analyses should be done in multiple large, high-quality trials. However, such post hoc analyses may be susceptible to both type I error (falsely significant associations) due to multiple testing or type II error (falsely nonsignificant associations) due to lack of statistical power. Nevertheless, they would form a good basis to determine which patient characteristics could be further investigated a priori in future trials.
Resource Use, Size, and Duration
Post hoc analyses of existing data can be done quickly with modest resources.
Ethical Issues
There should be no ethical issues involved in reanalyzing existing study data.
Randomized Controlled Trials
Value of Study Design
The CER found 43 trials comparing CPAP to no or sham treatment, 21 trials comparing autotitrating CPAP with fixed CPAP, 10 trials comparing MAD to no or sham treatment, and 10 trials comparing CPAP to MAD. However, only 5 of these 84 trials were rated good quality. Regardless, it is unlikely that any new trial addressing these comparisons would substantially change the current conclusions about the effect and relative effect of CPAP, autotitrating CPAP, and MAD on measures of sleep and sleep quality. Thus the primary purpose of future studies should be to address this FRN topic, namely which group of patients would benefit most (or least) from treatment. Ideally, the a priori choice of patient characteristics would be based on existing patient-level meta-analyses or post hoc analyses of existing trials.
Of note, the EPC agrees with the stakeholders that it would be less desirable to conduct separate trials in narrow subgroups of patients. This would require indirect comparisons across studies where one has to assume that the differences in study findings are due to a single, explicit difference in population (e.g., obese vs. nonobese), even though it may be very likely that differences are due to multiple other, often subtle, differences between studies. Direct comparisons are always more reliable than indirect comparisons.
Resource Use, Size, and Duration
The primary analysis of interest in such a study would be whether the interaction term (between the main effect and the predictor variable) is statistically significant. Using as an example the comparison between CPAP and MAD, the goal would not be to demonstrate again that both CPAP and MAD are effective, but CPAP is superior. Instead, the analysis of interest would be whether the relative effectiveness of CPAP and MAD are different in different subpopulations of patients (e.g., baseline AHI <30 events/hr versus ≥30 events/hr) or analyses of the effect based on the predictor variable on a continuous scale (e.g., body mass index). The number of patients that would need to be enrolled in a trial to find a significant interaction would depend on the distribution of the predictor in the study sample and the strength of the interaction. Studies with a small minority of patients in the predictor category (e.g., if 10 percent were obese) would require a larger strength of interaction (e.g., a larger difference in the effect between the obese and nonobese) than studies that were more evenly balanced. Regardless of the actual distribution of study subjects and the relative difference in effects between groups, such trials would need to be substantially larger than an equivalent trial evaluating only the main effect.
Focusing on intermediate outcomes, such as AHI or measures of sleepiness, trials can be relatively short term, for example 2 to 3 months.
Ability to Recruit
Thousands of patients have already been recruited into OSA treatment trials. Patients are usually sufficiently symptomatic that they are likely to be interested in treatment. There are numerous treatment options, without clear guidance as to who should receive which treatment. For all these reasons, it is likely that there will not be major obstacles in recruitment to future trials.
Ethical Issues
The primary ethical issue (beyond the standard ethical issues in conducting a randomized trial) pertains to whether there is equipoise. An argument can be made that another trial that evaluates only CPAP or MAD for the same sleep-related outcomes that the existing 84 trials have evaluated may be unethical since the value of the treatment is known and the trial would not change the conclusions from the body of evidence. However, if a new trial truly addresses a question that is inadequately addressed, such as which patients would not benefit from treatment, or which treatment would work best in which patients, then there is adequate equipoise.
Observational Studies
Value of Study Design
Though less informative than randomized trials for the evaluation of comparative effectiveness, analyses of observational studies could be fruitful to evaluate predictors of effective treatment. Either prospective or retrospective cohort (single treatment group) data could be evaluated, using multivariable regression, to determine which patient characteristics (e.g., baseline body mass index) are associated with more successful treatment (e.g., clinically significant improvement in sleepiness). Case-control studies of patients treated for OSA, with or without a clinical outcome (e.g., a myocardial infarction), could also address this question, though they would be subject to all the biases and other problems of any case-control study. For any study design, the analyses that would be clinically useful would be predictors of outcomes based on baseline characteristics (e.g., relative risk of outcome in obese vs. nonobese), not vice versa (e.g., the average weight of successfully treated vs. untreated patients). The latter analyses are commonly reported, but do not provide useful insights for clinicians making treatment decisions.
Resource Use, Size, and Duration
Retrospective data could be gathered quickly and easily by chart review or similar approaches. Prospective cohorts would take somewhat more time and effort, but would still be less resource-intensive than a trial. Since symptoms generally change rapidly (within a few weeks) after the start of treatment, data could be collected fairly rapidly, whether retrospective or prospective. The sample size will depend on the number of regression covariates and the event rate (for dichotomous outcomes).
Ability To Recruit
Patient recruitment should be straightforward and relatively simple, as there would be little added burden for them by entering a study, beyond a formalized assessment of symptoms, other outcomes of interest, and patient characteristics, all of which are reasonable parts of normal patient care.
Ethical Issues
There should be no substantive ethical issues involved in these observational studies.
High-Priority Future Research Needs Topic 4
Trials to improve compliance with CPAP, MAD, and other treatments, particularly evaluating cognitive therapy approaches
Background
Key Question 7 in the CER asked about the efficacy of interventions to improve compliance with device use. Specifically:
What is the effect of interventions to improve compliance with device (positive airway pressure, oral appliances, positional therapy) use on clinical and intermediate outcomes?
The CER concluded that there was a low strength of evidence that some specific adjunct interventions may improve CPAP compliance compared with usual care. None of the included trials reported results on clinical outcomes. The CER identified 18 trials that evaluated a wide variety of interventions to improve CPAP compliance, but did not find trials of MAD or other treatments of sleep apnea. The 18 trials of CPAP were mostly applicable to patients initiating CPAP with severe disease (AHI >30 events/hr who were obese (body mass index >30 kg/m2). They showed inconsistent effects across a wide variety of interventions. No general type of intervention was found to be more promising than others. However, compared with usual care, several interventions were shown to significantly increase hours of CPAP use per night in some studies. These included intensive support or literature (designed for patient education), cognitive behavioral therapy (given to patients and their partners), telemonitoring, and a habit-promoting audio-based intervention. However, these trials generally had small sample sizes (<40 patients in each arm) with less than 1 year of followup.
Stakeholder Discussion
Particular interventions discussed by the stakeholders included intensive patient education, social support programs engaging spouse or other family support, cognitive behavioral therapy, feedback from daily adherence monitoring, and peer support in a Web-based fashion. Stakeholders discussed that such interventions have been found to be effective in other domains such as weight control, drug addiction, and medication-taking behavior, but have not been adequately tested for OSA treatment. These types of interventions were discussed as being of greater importance than further modifications of CPAP and other devices.
Proposed Research Design
RCTs of multicomponent or multidimensional behavioral interventions aimed to improve long-term compliance with CPAP, MAD, and other treatments of sleep apnea.
Value of Study Design
A well-done RCT will produce the most convincing results, and if patient eligibility criteria and study setting are realistic, should be fairly applicable to the majority of patients. Several study design considerations need to be addressed before a RCT can be properly designed and carried out. These considerations are described as follows:
- Use established metrics to measure compliance. Commonly used metrics in the published literature include the continuous outcome hours of increased use of the device and the dichotomized outcome of at least 4 hours of use per night for more than 70 or 80 percent of the nights per week. However, it remains unclear how to interpret the clinical significance of these metrics. The relationships between improvements in compliance (using these measures) and clinical outcomes or quality of life have not been established.
- Use objective measures of compliance. Due to the nature of behavioral interventions, blinding of patients to the interventions cannot be achieved. Therefore, it is important to use objective measures of compliance that do not rely on self- or provider-reported data. An example of an objective measure is automatically recorded usage data from built-in memory chips in CPAP devices.
- Use behavioral change models and theories for designing interventions. Behavioral change models are multicomponent and multidimensional, involving patients, family members, peers, care providers, and changes to the patient’s physical environment. Many interventions to promote health-related behaviors have utilized behavioral change models, and have been shown to be effective in changing patients’ behaviors. For example, cognitive behavioral therapy (e.g., focusing on coping skills or environmental manipulation) was shown to be effective in improving compliance with CPAP treatment.5 All behavioral interventions should be piloted among the targeted patient population. Ideally, qualitative studies, such as focus groups, should be conducted to assess patients’ preferences or feedback and barriers to the behavioral interventions as part of program evaluation.
- Conduct program evaluation during trial period. Program evaluation—a systematic method for collecting, analyzing and utilizing information regarding delivery of intervention program—is essential to identify barriers to either the implementation or effectiveness of the behavioral interventions.
Resource Use, Size, and Duration
The primary analysis of interest in such a RCT would be whether the compliance with CPAP, MAD, or other treatments is statistically significantly improved compared with usually care (no behavioral intervention). We conducted a power calculation using standard formulae for a two-sided equivalence test for mean difference in hours increase in use (as a continuous measure for compliance) between two independent samples. We set the range of equivalence (the range between the smallest mean difference that would be clinically significant and the largest difference that could be reasonably expected) from 0.5 hours to 8 hours of use. From the CER, we obtained the range of mean difference (from 1 hour to 3 hours of use) and the range of standard deviation of mean differences for the power calculation. The required total sample sizes (assuming a 1:1 ratio in a two-arm RCT) for pairs of assumptions using three mean differences and three standard deviation pair are presented in Figure 5. It should be noted that, most trials in the CER found a mean difference of less than 2 hours increase in CPAP use. Some trials in the CER found a significant improvement in compliance with CPAP use within 1 month, but other trials found that it took much longer (6 to 9 months). Regardless, it would be of greater interest to show long-term (several years) effects on compliance, since in most cases, treatment can be expected to be life-long.
Figure 5
Sample size calculation for a RCT of behavioral interventions to improve compliance with CPAP, MAD, or other treatments, compared with usual care. hr = hours (of use per night), N = number of trial participants (total), SD = standard deviation
Ability To Recruit
Poor compliance with either CPAP or MAD is very common among patients with OSA. The patient advocates on the stakeholder panel described a demand by patients for better training, education, and other means to improve compliance. Therefore, it should be relatively easy to recruit patients with OSA who are being treated with devices into a trial.
Ethical Issues
There should be no substantive ethical issues involved beyond the standard ethical issues in conducting a randomized trial. The clinical value of interventions designed to improve compliance remains unclear.
High-Priority Future Research Needs Topic 5
What is the association between sleep apnea severity and long-term clinical outcomes?
Background
The CER reported strong evidence from four studies (three of which were of good quality) demonstrating that baseline AHI >30 events/hr is an independent predictor of all-cause mortality over several years of followup. Evidence for the association of baseline AHI and other long-term clinical outcomes, however, is lacking. To address this research gap, the CER recommended patient-level meta-analyses of existing data from available large cohorts of individuals who had sleep testing, such as the Sleep Heart Health Study6 and the Wisconsin Sleep Cohort Study.7 However, other research study options could address the topic.
Stakeholder Discussion
Stakeholders on the FRN panel echoed the need for evidence on the association of baseline AHI and other long-term clinical outcomes—particularly to inform design of studies to assess long-term efficacy of treatment (e.g., early intervention to treat snoring in individuals with low baseline AHI). Stakeholders recommended prospective natural history studies designed to determine whether patient characteristics (e.g., age, sex, race/ethnicity), in addition to disease severity, play a role in determining long-term clinical outcomes.
Proposed Research Designs
To address the overarching FRN topic and its subcomponents, different study designs are reasonable and would address different aspects of this topic.
Patient-Level Meta-analyses
Value of Study Design
Standard meta-analysis of existing studies is not feasible, as discussed in the CER, because of the large degree of clinical heterogeneity in the existing studies, both in terms of their participant eligibility criteria and their analytic methods (e.g., adjusting for different variables). However, patient-level meta-analysis could overcome these problems, by incorporating the differences into a regression analysis and performing a unitary analysis. Existing studies of individuals receiving sleep testing—such as the Sleep Heart Health Study and the Wisconsin Sleep Cohort Study—can be assessed to determine whether their data include a baseline measure of sleep apnea severity (e.g., AHI), and relevant demographic and clinical variables to allow for patient-level meta-analysis. In general, this type of meta-analysis would combine original patient-level data from multiple studies, allowing for unbiased regressions and subgroup analyses. Meta-regression could be used to evaluate associations between sleep apnea severity, and other patient characteristics, with long-term clinical outcomes. Typically, acquiring data is the main obstacle to conducting patient-level meta-analyses. Clinical outcomes of interest would include incident clinical events, quality of life, and psychological or neurocognitive measures.
Resource Use, Size, and Duration
If appropriate data exist and are available, modest resources would be needed to evaluate associations between sleep apnea severity, in combination with other relevant patient characteristics, and long-term clinical outcomes.
Ethical Issues
Meta-analysis uses existing data. With permission to use original study data, there should be no ethical issues.
Post Hoc Analyses of Existing Studies
Value of Study Design
A second-tier option for using existing data to investigate potential associations between sleep apnea severity and long-term clinical outcomes would be to perform post hoc regressions or subgroup analyses within already-published cohort studies. This could be performed either by the original study investigators or other researchers with access to the study data. To maximize accuracy and validity, such analyses should be done separately in multiple large, high-quality studies. The primary reasons this approach would be a second-tier option to patient-level meta-analysis are that they would be less well powered to show an association and they would be intrinsically less generalizable than an analysis of combined cohorts. However, they may be more feasible since they would not require collaboration of multiple sets of study investigators. Such post hoc analyses may be susceptible to both type I error (falsely significant associations) due to multiple testing or type II error (falsely nonsignificant associations) due to lack of statistical power. Nevertheless, they would form a good basis to determine which demographic and clinical variables could be further investigated a priori in future studies.
Resource Use, Size, and Duration
Post hoc analyses of existing data can be done quickly with modest resources.
Ethical Issues
There should be no ethical issues involved in reanalyzing existing study data.
Prospective Natural History Study
Value of Study Design
Prospective natural history studies can be conducted to investigate potential associations between sleep apnea severity and long-term clinical outcomes, particularly, if existing studies are inappropriate for post hoc or meta-analyses, or they lack data on relevant patient characteristics or outcomes of interest. Studies would require assessment of baseline sleep apnea severity in a specified patient population and documentation of all relevant demographic and clinical variables over time.
Resource Use, Size, and Duration
Moderate resources may be required to recruit and follow a large cohort over a long timeframe.
Ability To Recruit
Patient recruitment should be straightforward and relatively simple, as there would be little added burden for individuals entering a study beyond AHI measurements and documentation of relevant demographic and clinical variables—all of which are components of normal patient care. However, there is potential for large data loss due to long-term followup.
Ethical Issues
Few ethical issues are likely to occur as the study involves components of normal patient care.
Second-Tier Future Research Needs Topics
Second-Tier Future Research Needs Topic 6
What are the barriers to, and predictors of, compliance with different treatments?
Background
Key Question 6 of the CER asked about predictors of treatment compliance. Based on a total of five large cohort studies, there is a moderate strength of evidence that more severe OSA as measured by higher AHI is associated with greater compliance with CPAP use, there is a moderate strength of evidence that a higher measure of sleepiness is also associated with improved compliance, and the strength of evidence is insufficient regarding potential predictors of compliance with MAD.
Stakeholder Discussion
This FRN topic represents two topics that were nominated and discussed separately by the stakeholders. After nomination of topics by the stakeholder panel, the topics were combined into one overarching topic. The original topics were: Studies to predict adherence to CPAP, surgery and MAD; and What are the barriers to compliance?
The original rationale for the first topic was that high quality studies are necessary to determine which factors predict adherence with CPAP and MAD, to which predict successful outcomes after surgery. One stakeholder argued that the topic was moot since treatment must be started and CPAP is superior and thus the clear first choice. Another stakeholder commented that it is very difficult to predict adherence behavior but that it may understanding which factors could influence adherence may allow for designs of better interventions to promote adherence.
The discussion about the more general question of what are the barriers to compliance described several possible barriers, including cognitive impairment due to sleepiness (prior to treatment), a lack of followup care, the value of rapid training on the use of the equipment, the importance of reimbursement for patient education, and a need to demonstrate return-on-investment for interventions that improve compliance. It was stated that diabetes education is reimbursed, but sleep apnea reimbursement is not.
Second-Tier Future Research Needs Topic 7
Direct comparison of compliance rates with different interventions and incorporation of compliance into an overall comparison of effective treatment
Background
Key Question 7 of the CER asked about treatments to improve compliance. Based on 18 trials to improve CPAP compliance, there is a low strength of evidence that some specific adjunct interventions may improve CPAP compliance among overweight patients with more severe OSA who are initiating CPAP treatment. However, studies are heterogeneous and no general type of intervention (e.g., education) was more promising than others.
Stakeholder Discussion
The stakeholder discussion revolved around the need to improve compliance with CPAP to allow for more effective treatment among patients outside the research study setting. It was noted that truck drivers in particular have problems with compliance due to the difficulties of incorporating CPAP into their routine, including the need to sleep in the cab of their trucks without backup power generation. It was put forth that the key reason for poor compliance is lack of proper patient education and aftercare, and that physicians do not have adequate training in educating patients about sleep apnea management. It was stated that nonahderence has been stable at 50 to 60 percent for the past two decades. There was discussion about how cognitive behavioral therapy needs to be researched further as a tool to improve compliance. Trials are needed to compare cognitive behavioral therapy to usual care and different interventions to each other.
Second-Tier Future Research Needs Topic 8
Trials to evaluate weight-loss programs as an adjunctive treatment for sleep apnea
a. What is the value of bariatric surgery for treatment of sleep apnea?
Background
Key Question 5 of the CER included analyses of weight-loss programs and bariatric surgery. Based on a three trials, there is a low strength of evidence to show that some intensive weight loss programs are effective treatment for OSA in obese patients. A single nonrandomized study provided insufficient evidence regarding bariatric surgery.
Stakeholder Discussion
This FRN topic represents two topics—on weight loss programs and bariatric surgery—that were nominated and discussed separately by the stakeholders. After nomination of topics by the stakeholder panel, the topics were combined into one overarching topic. The rationale regarding weight-loss programs noted that obesity is a risk factor for sleep apnea and that the incremental value of adding weight-loss programs to usual care should be examined in RCTs. It was noted that weight-loss alone has been shown to have a modest effect on AHI, so coupling weight loss with usual care could be synergistic, but that the current evidence is inadequate. Of interest were how practitioners can help patients to improve successful weight loss and which factors predict beneficial effect.
It was also noted that bariatric surgery has been shown to greatly improve AHI in cohorts not selected for sleep apnea, but that the effect specifically in OSA patients needs to be demonstrated. It was discussed that single-treatment cohort studies may be sufficient to demonstrate an effect and that patients would likely not be willing to be randomized to no surgery. The expected effect is on AHI and sleepiness is expected to be large enough to mitigate the need for a trial. One stakeholder discussed how it would be important to categorize patients based on the cause of the OSA, since obesity alone may not be the primary cause in many patients. One group would be those with the known physical anomalies including airway size, tongue, jaw, soft palate and nasal constriction and obesity. The other patient group would be those who are obese only and have no known tongue, jaw, soft palate and nasal constriction issues.
Second-Tier Future Research Needs Topic 9
What is the value of including a sleep medicine specialist in the management of the patient with obstructive sleep apnea?
Background
The CER identified no studies that addressed this topic.
Stakeholder Discussion
The discussion consisted primarily of the rationale. It was noted that patients with OSA are increasingly being treated by primary care physicians. However, most of the research has been done in highly specialized sleep centers with patients under the supervision of a sleep medicine specialist. Thus, it remains unknown how effective interventions may be without the sleep medicine specialist involved. Studies comparing treatment management with and without sleep medicine specialists are of interest (primary care physician alone, primary care physician and specialist together, and specialist alone). Outcomes of interest include adherence, quality of life, cost, and cardiovascular outcomes.
Other Future Research Needs Topics
Other Future Research Needs Topic 10
What is consumer willingness to pay for treatment, to identify consumer preferences for strategies to treat sleep apnea?
Other Future Research Needs Topic 11
What are the financial barriers to access to treatment?
Other Future Research Needs Topic 12
Role of surgery for treatment of OSA
a. Comparison of surgery versus CPAP
b. Role of orthognathic surgery (corrective jaw surgery)
c. Comparison of genio-tubercle advancement versus dental devices
Other Future Research Needs Topic 13
Evaluation of postoperative CPAP for all patients with OSA or at high risk of OSA undergoing any surgery with sedation
Other Future Research Needs Topic 14
Trials comparing CPAP versus pharmaceutical interventions
Other Future Research Needs Topic 15
Trials comparing different CPAP masks
Other Future Research Needs Topic 16
Trials comparing CPAP versus oropharyngeal exercises
Other Future Research Needs Topic 17
Trials comparing different degrees of mandibular advancement
Other Future Research Needs Topic 18
Studies of factors influencing therapist decisions concerning CPAP mask choice
Other Future Research Needs Topic 19
Research into how to maintain patients in sleep apnea studies (where dropout rates are unacceptably high)
This topic was not considered to be high priority by any stakeholder.
- Results - Future Research Needs for Treatment of Obstructive Sleep ApneaResults - Future Research Needs for Treatment of Obstructive Sleep Apnea
Your browsing activity is empty.
Activity recording is turned off.
See more...