Begin main content
CADTH Document Viewer

Interventions for the Treatment of Obstructive Sleep Apnea in Adults: A Health Technology Assessment

March 2017

Executive Summary

Issue

Obstructive sleep apnea (OSA) affects as many as one in four adults in Canada. Untreated OSA is associated with increased risk of motor vehicle crashes, various chronic diseases, and death. A wide range of treatment options exist, but selecting the most appropriate therapy remains a challenge across Canadian jurisdictions.

Objectives

The aim of this health technology assessment (HTA) was to assess the clinical effectiveness, cost-effectiveness, patient perspectives and experiences, ethical issues, implementation issues, and environmental impacts of positive airway pressure (PAP) devices, expiratory positive airway pressure (EPAP) valves, oral appliances (OAs), surgery, and lifestyle modifications for the treatment of OSA in adults.

Clinical Evidence

Methods: A systematic review of the literature was conducted, using MEDLINE, Embase, the Cochrane Database of Systematic Reviews, DARE, Cochrane Central, and PubMed, for an overview of systematic reviews (SRs), meta-analyses (MAs), and HTAs, supplemented by a review of primary studies for areas with gaps, where data were lacking. In total, 33 SRs and 41 primary studies were included in the overview and review, respectively, on adults with OSA who were treated with PAP devices, EPAP valves, OAs, surgery, and lifestyle interventions and assessed on various outcomes, with excessive daytime sleepiness (EDS) as the primary outcome.

Results: The majority of relevant studies identified in the published literature pertained to continuous PAP (CPAP) and mandibular adjustment devices (MADs). The results indicated that CPAP, MADS, EPAP, tongue-retaining devices (TRDs), maxillomandibular advancement (MMA), genial tubercle advancement (GTA), weight-loss programs, and positional therapy were all effective at reducing EDS, commonly measured by Epworth Sleepiness Scale, compared with inactive controls or pre-treatment. Effect sizes were similar across the interventions, except for patients with severe cases of OSA, who may benefit more from CPAP than from MADs, but this may not be clinically significant. Based on the analysis using OSA severity as the outcome, commonly measured by the Apnea-Hypopnea Index (AHI), effect sizes varied across the interventions, with CPAP showing the largest effect. Among a select subgroup of OSA patients for whom surgery may be appropriate, MMA with or without GTA may substantially improve EDS and OSA severity. The majority of studies on MMA and GTA, however, were in highly selected patients in uncontrolled pre-and-post studies with sample sizes of fewer than 10 patients. The results, therefore, must be interpreted with great caution, especially given the invasiveness of the procedures and potential adverse events. Limited evidence was found on non-PAP or non-OA interventions and other outcomes, such as blood pressure, cardiovascular events (CVEs), quality of life, and mortality. The 33 SRs and 41 primary studies were assessed to generally be of high quality, using accepted quality assessment tools, but concerns were identified regarding the study eligibility criteria for the SRs and small samples and uncontrolled pre-and-post study designs for the primary studies. The primary studies included in the 33 SRs ranged widely in their quality.

Economic Evidence

Methods: A Markov cohort model was constructed in order to evaluate the cost-effectiveness of various treatment strategies in adult patients diagnosed with OSA (i.e., 76.5% males, 55 years of age) over a patient's lifetime from a Canadian health care payer perspective. The effect of treatment in terms of change in AHI and blood pressure was determined from the clinical review and was translated to changes in the risk of CVEs and motor vehicle accidents (MVAs) in the economic model. The primary outcome was cost per quality-adjusted life-years (QALYs) gained, in 2016 Canadian dollars. The base-case analysis compared a "no-treatment" strategy against PAP therapy, MADs, and surgery (i.e., MMA with or without GTA). A separate scenario analysis was conducted on obese or overweight patients, in which weight loss would be a suitable treatment strategy.

Results: Cost-effectiveness of treatment strategies for OSA was found to be dependent on a patient's baseline disease severity, as measured by AHI (i.e., lower AHI equates to less severe OSA). In patients presenting with mild baseline severity (AHI < 15), no treatment would be cost-effective if the willingness-to-pay was less than $175,000/QALY. In patients with a baseline AHI of 15, PAP therapy would be cost-effective if the willingness-to-pay was between $8,058/QALY and $9,276/QALY, and thereafter, at a willingness-to-pay greater than $9,276/QALY, MAD would be the most likely cost-effective intervention. In patients with severe AHI (AHI ≥ 30), both PAP therapy and surgery are potentially cost-effective strategies. Specifically, PAP therapy was cost-effective in the lower range of severe OSA (e.g., incremental cost-utility ratio [ICUR] at a baseline AHI of 30 = $7,420) while surgery was cost-effective in the higher range of severe OSA (e.g., ICUR at a baseline AHI of 60 = $17,125). Absolute gains in QALYs from treatment were found to follow a unimodal distribution and were a function of disease severity. Those with mild or more severe OSA had lower gains in QALYs, whereas the largest gains were observed in patients whose baseline severity reduced from severe (AHI ≥ 30) or moderate (15 < AHI < 30) to mild-to-moderate OSA (AHI < 30) or mild OSA (AHI < 15), respectively, due to its impact on subsequent morbidity and mortality risks. Incremental costs were largely driven by the costs of treatment and long-term maintenance costs, given the longer life expectancies of patients on treatment. It is important to note that the estimates on the clinical effectiveness of surgery were taken from the clinical review and, thus, caution is required when interpreting the cost-effectiveness results of surgery. Not all patients will be suitable for surgery and, in such instance, the economic analysis suggests that MAD would be cost-effective for moderate OSA if willingness-to-pay was $7,984/QALY while PAP therapy would be cost-effective for severe OSA if willingness-to-pay was $7,470. The model was found to be most sensitive to changes in treatment adherence.

Patient Perspectives and Experience Evidence

Methods: An SR and thematic synthesis of the literature relevant to the research question on patient experience and perspectives was conducted. Patient experience information was identified by searching the following databases: MEDLINE (1946-), Embase (1974-), and PsycINFO (1967-) via Ovid; CINAHL (1981-) via EBSCO; and PubMed. Studies were eligible if they presented the patient or non-clinical caregiver experience. Qualitative studies, surveys, studies with mixed methodology, or systematic reviews of descriptive studies were eligible. A maximum variation approach was used to identify articles for inclusion in the thematic synthesis from a list of eligible articles. A thematic synthesis was conducted, comprising three stages: coding, developing descriptive themes, and developing analytic themes.

Results: Thirty-two studies, of moderate to good quality, were included in the thematic synthesis, the coding and analysis of which led to two analytic themes. The first theme states that a range of characteristics and factors influence whether people seek and initiate OSA treatment. The second analytic theme states that interventions for OSA require adaptation to daily routines and relationships; some patients are able to integrate these interventions into their lives and experience benefits, while others are unable to do so.

Ethical Issues

Methods: A review of the normative bioethics literature was conducted to identify literature relevant to the identification and analysis of the potential ethical issues on interventions for OSA (i.e., articles that explicitly and specifically raise ethical issues). A CADTH Information Specialist performed targeted literature searches in MEDLINE, PubMed, and CINAHL from database inception to March 2016. Key terms for ethics concepts and related terms were used and combined with search terms for OSA. The search was limited to English- or French-language literature. The selection of relevant literature occurred in two stages. In the first stage, the title and abstracts of citations were independently screened for relevance by two reviewers. Articles were categorized as "potentially relevant" or "not relevant" based on whether ethical issues were explicitly mentioned. In the second stage, full-text reports that pertained to OSA treatments in adults and suggested implicit ethical issues were identified.

Results: The literature search yielded 1,268 unique citations, none of which passed the first stage of screening because no articles on OSA treatment were found that explicitly mentioned ethical issues. However, in the second stage the reviewers selected 142 potentially relevant articles that raised implicit ethical issues. Ethical issues relating to OSA were explored according to six key values that emerged from the literature review. This includes the duties to respect individual autonomy, maximize benefits and minimize harm for patients, maximize benefits and minimize harms for others affected by OSA, maximize benefits and minimize harms for populations, distribute benefits and burdens of health care resources fairly, and steward scarce resources.

Implementation Issues

Methods: A narrative literature review was conducted to identify some of the implementation issues associated with the different interventions for the treatment of OSA in adults. Citations arising from the literature searches conducted to address the clinical, economic, patient perspectives and experience, and ethical issues were screened independently in duplicate for information related to implementation issues. Issues identified from relevant studies are organized by OSA intervention (i.e., PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications) and further categorized by the level where the issue arises: individual, team, organization, or system or policy. This information was summarized narratively.

Results: From the 29 included studies, one of the biggest implementation issues identified for OSA treatment is the difficulty in accessing sleep specialists and specialized sleep labs. Home-based portable diagnostic devices and treatment titration options with telehealth-based support are suggested solutions. Most of the implementation evidence focuses on CPAP devices. Barriers to CPAP use include cost and lack of funding as well as patient discomfort or use problems. Suggested CPAP supports include patient education, training, and support as well as providers and centres that are accredited for the treatment of OSA. Barriers to treatment with OAs include lack of physician knowledge and awareness, anatomical and dental health requirements, and the need for regular re-evaluations. Multidisciplinary sleep clinics that include medical and dental professionals are suggested supports. Little evidence on implementation issues for OSA surgery or lifestyle interventions was found.

Environmental Impact

Methods: Citations arising from the clinical literature search were screened for information relating to environmental considerations associated with OSA.

Results: One narrative review article was identified regarding the environmental implications associated with OSA. The review article briefly examined the environmental considerations of the CPAP unit, including manufacturers adopting green shipping and production methods and creating products that are more energy efficient and more recyclable.

Conclusions

Clinical data show that across all OSA severities, various treatment interventions for OSA were found to significantly improve sleepiness, but CPAP had the largest effect on improving OSA severity. Treatment of patients with moderate-to-severe OSA appears to be a cost-effective use of resources under a willingness-to-pay threshold of $50,000/QALY.

Nevertheless, for any non-surgical therapy, patient adherence was considered key in achieving treatment success and cost-effectiveness. Relevant patient factors were highly individualized and contextual and the factors that influence whether patients seek treatment and how they experience treatment will differ for each individual. It may be that patients who are symptomatic, have a supportive partner, experience resolution of their symptoms when using CPAP or an OA, and experience few or mild side effects may be more likely to be compliant with these interventions. Patients appear to make a trade-off between the discomfort of CPAP and OAs, and the perceived benefits of using these devices. If patients find these interventions acceptable, they experience a period of problem-solving and adaptation to integrate these interventions into their lives. Information needs were expressed during the patient experience, from diagnosis and throughout treatment. Support from peers, health care professionals, and bed partners was also important, although some patients did not feel supported in using interventions for OSA.

From an ethics perspective, interventions for OSA have been shown to offer benefit to OSA patients and to reduce overall costs, and so appear to live up to the values of conferring benefit at a population level and stewarding scarce resources. In light of the significant personal and public harms that undiagnosed OSA can cause, further consideration of screening protocols and public education is warranted. In addition, our duties to distribute benefits and burdens fairly require that we are attentive to the accessibility of testing and diagnostic services, recognizing any sociocultural factors (e.g., gender, ethnicity, socioeconomic class) that may unjustly affect this access. Optimizing interventions for OSA that reflect individuals' individual contexts and abilities appear most likely to maximize adherence, thus leading to benefits at the individual and the population level. OSA treatment should be provided through an ongoing partnership between health care provider and patient, rather than through discrete events of diagnosis, decision, and intervention.

The review of implementation issues further highlighted the difficulties in accessing sleep specialists and laboratories as being critical to initiating treatment of OSA, as well as the benefits of multidisciplinary sleep clinics. A single review was found to recommend environmental considerations for CPAP, such as energy-efficient and recyclable products and green shipping and production methods. Therefore, for the treatment of OSA, in addition to clinical and cost-effectiveness evidence, patient, provider, supplier, and system readiness for the various interventions will need to be considered.

 

Abbreviations

A1C glycated hemoglobin
AE adverse event
AF atrial fibrillation
AHI Apnea–Hypopnea Index
AHRQ Agency for Healthcare Research and Quality
AI arousal index
AMSTAR Assessment of Multiple Systematic Reviews
APAP autotitrating positive airway pressure
BDI Beck Depression Inventory
BiPAP bilevel positive airway pressure
BMI body mass index
BP blood pressure
BSI Brief Symptom Inventory
BSI-A Brief Symptom Inventory–Anxiety Subscale
BSI-D Brief Symptom Inventory–Depression Subscale
CBD cerebrovascular disease
CDSR Cochrane Database of Systematic Reviews
CEAC cost-effectiveness acceptability curve
CENTRAL Cochrane Central Register of Controlled Trials
CES-D Center for Epidemiological Studies Depression Scale
CI confidence interval
CONSORT Consolidated Standards of Reporting Trials
CPAP continuous positive airway pressure
CV cardiovascular
CVD cardiovascular disease
CVE cardiovascular event
DARE Database of Abstracts of Reviews of Effects
DBP diastolic blood pressure
DI desaturation index
EDS excessive daytime sleepiness
EPAP expiratory positive airway pressure
EPHPP Effective Public Health Practice Project
EQ-5D-3L EuroQol-5 dimensions–3 levels
ESS Epworth Sleepiness Scale
FOSQ Functional Outcomes of Sleep Questionnaire
GHQ General Health Questionnaire
GP genioplasty
GQL glaucoma quality of life
GRADE Grading of Recommendations Assessment, Development and Evaluation
GTA genial tubercle advancement
HADS-A Hospital Anxiety and Depression Scale–Anxiety Subscale
HADS-D Hospital Anxiety and Depression Scale–Depression Subscale
HAM-D Hamilton Rating Scale for Depression
HDL high-density lipoprotein
HF heart failure
HTA health technology assessment
ICUR incremental cost-utility ratio
ISPOR International Society for Pharmacoeconomics and Outcomes Research
LOCF last observation carried forward
MA meta-analysis
MAD mandibular advancement device
MADRS Montgomery–Åsberg Depression Rating Scale
MAS mandibular advancement splint
MD mean difference
MI myocardial infarction
MeSH Medical Subject Heading
MMA maxillomandibular advancement
MMPI Minnesota Multiphasic Personality Inventory
MMPI-Pt Minnesota Multiphasic Personality Inventory–Psychasthenia Subscale
MR meta-regression
MVA motor vehicle accident
NA not available
NHP Nottingham Health Profile
NMA network meta-analysis
NICE National Institute for Health and Care Excellence
NR not reported
OA oral appliance
ODI oxygen desaturation index
OR odds ratio
OSA obstructive sleep apnea
PAP positive airway pressure
POMS-D Profile of Mood States–Depression Subscale
POMS-T Profile of Mood States–Tension and Anxiety Subscale
PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses
PSG polysomnography
PTCA percutaneous transluminal coronary angiography
QA quality assessment
QALY quality-adjusted life-year
QoL quality of life
QSQ Quebec Sleep Questionnaire
RCT randomized controlled trial
RDI respiratory disturbance index
RoB risk of bias
RoBANS Risk of Bias Assessment Tool for Nonrandomized Studies
ROBIS Risk of Bias in Systematic Reviews
RR relative risk
SAHS sleep apnea/hypopnea syndrome
SAQLI Calgary Sleep Apnea Quality of Life Index
SBP systolic blood pressure
SD standard deviation
SDS Zung Self-Rating Depression Scale
SF Short Form Health Survey
SHHS Sleep Heart Health Study
SPT sleep position trainer
SR systematic review
SSS Stanford Sleepiness Scale
STAI State-Trait Anxiety Inventory
TAS Tension Anxiety Scale
Tot-C total cholesterol
TRD tongue-retaining device
TSD tongue-stabilizing device
UMACL University of Wales Mood Adjective Checklist
UPPP uvulopalatopharyngoplasty
UWIST University of Wales Institute of Science and Technology
WHO-5 World Health Organization–Five Well-Being Index
WHOQOL World Health Organization Quality of Life questionnaire
WSCS Wisconsin Sleep Cohort Study

 

Protocol Amendments

Section Amendment
Clinical Review For selection criteria, success or cure rate was added to the list of outcomes.
For quality assessment, a fourth item from the AMSTAR checklist was added to the list of additional criteria.
For quality assessment, the network meta-analyses were assessed, using the ISPOR questionnaire, in addition to the ROBIS tool.
For analysis and synthesis, quality assessment of primary studies included in the systematic reviews was presented in the quality assessment section of this report, instead of the summary of study characteristics.
All sections Throughout the report, EPAP was separated from PAP devices.

 

1. Introduction

1.1. Background

Obstructive sleep apnea (OSA) is a common disorder that is characterized by narrowing and collapse of the upper airway during sleep and is associated with arousals and awakenings. A 2009 Canadian survey reported that 3% of Canadian adults were diagnosed with sleep apnea, making it as common as hypertension or diabetes.1,2 Cross-sectional and longitudinal data would suggest that the prevalence of OSA is 15% in males and 5% in females.3,4 The prevalence may be even higher, with more than one in four adults presenting symptoms and factors associated with having or developing OSA.1,5 Its major symptoms include snoring, unrefreshing sleep, excessive daytime sleepiness (EDS), lack of concentration, impaired memory and lower quality of life (QoL).5,6 Aging, the male sex, and obesity are its main risk factors.7,8 Untreated OSA is associated with motor vehicle accidents (MVAs), cardiovascular disease (CVD), stroke, hypertension, diabetes, cognitive dysfunction, and all-cause mortality.1,5,9-11 Untreated OSA is also a known surgical risk and can give rise to cardiovascular (CV) and pulmonary complications.12

Treatment of OSA includes a wide range of options.13 Positive airway pressure (PAP) devices and various oral appliances (OAs) splint the airways open to facilitate airflow. Continuous positive airway pressure (CPAP) forces air into the upper airways to prevent soft tissues from collapsing and is considered the gold standard for the treatment of OSA.13-15 Other PAP technologies, such as autotitrating PAP (APAP) and bilevel PAP (BiPAP), may be offered to patients with specific needs.13,14 An alternative treatment for OSA is nasal expiratory positive airway pressure (EPAP) valves, which are disposable devices that use a patient's own breathing to create positive end-expiratory pressure that prevents obstructive breathing.16 OAs, including mandibular advancement devices (MADs), also known as mandibular advancement splints (MASs) or dental devices, and tongue-retaining devices (TRDs), can be offered as an alternative to CPAP.14,17,18 For patients with mild or asymptomatic OSA, lifestyle interventions, such as exercise programs, diet changes, and positional therapies, may be proposed before proceeding to other interventions.19 Surgeries may be indicated for patients with a defined anatomical obstruction or morbid obesity, or as alternatives in cases where other interventions cannot be considered or have failed.20,21 Upper airway surgeries such as uvulopalatopharyngoplasty (UPPP) aim to facilitate air flow by remodelling soft tissue structure. In contrast, surgical maxillomandibular advancement (MMA) permanently pulls the lower jaw forward to create more space and prevent airway collapse.22,23 Genial tubercle advancement (GTA) is a surgical intervention that removes bone tissue from the chin and pulls the base of the tongue forward to create more airway space, and can be performed in conjunction with MMA or other surgeries to potentially improve therapeutic success.24,25

Therapy selection is based on an assessment of the patient by lab-based polysomnography (PSG) or home-based portable monitors.26 The goal is to determine the presence and severity of OSA by measuring the number of apnea or hypopnea events per hour, which constitutes the Apnea-Hypopnea Index (AHI), as well as the blood oxygen levels and other cardiorespiratory indicators.14 The AHI correlates with the risk of various CV outcomes, including hypertension, as well as all-cause mortality,5,8,10,26,27 and can also be used to determine the effectiveness of interventions intended to treat OSA. As a general rule, the therapeutic effect size is proportional to the severity of OSA.9,10 Other diagnostic measurements include the respiratory disturbance index (RDI) and time spent at oxygen saturation (SpO2) < 90%,7 as well as the Epworth Sleepiness Scale (ESS)14 and the oxygen desaturation index.28 Despite the positive outlook for CPAP, positioning it as the gold standard for treating OSA, between 29% and 83% of patients ultimately fail to comply with regular device use,29-31 which limits its impact. In fact, low CPAP adherence is associated with significantly higher mortality.2 Because of technical challenges, adherence with MADs is not as well documented but is regarded as being superior to CPAP.28,32 Therefore, patient adherence may be a factor in therapy selection and effectiveness. In that respect, surgical interventions may be attractive solutions for circumventing the issue of nightly adherence. However, they are invasive procedures for which evidence of effectiveness and safety is less convincing.14 Overall, while OSA interventions have shown effectiveness at reducing AHI, sleepiness, and some CV measures, such as blood pressure, no large randomized controlled trial (RCT) has yet demonstrated benefits on cardiovascular events (CVEs) or mortality.33

Across jurisdictions, OSA is associated with a substantial economic and societal burden.7,8,34 A cross-sectional study35 in the United States (US) found that, in the year prior to the diagnosis of OSA, the mean annual medical cost per patient was $2,720 for OSA cases, versus $1,384 for age- and sex-matched controls. In the US, OSA was established as the cause of 800,000 MVAs in 2000, for a total of 1,400 deaths and a cost of $15.9 billion to society.8,36 In Australia, the total impact of managing sleep disorders - including direct hospital and non-hospital costs, as well as the costs of associated conditions, such as stroke, heart disease, depression, and accidents - was estimated at $818 million in 2010.37 In Canada, the Assistive Devices Program within Ontario's Ministry of Health and Long-Term Care received approximately 28,000 applications for CPAP in 2008.38 Although no trend information is available, assuming new devices are required each year, extrapolating these figures to the entire Canadian OSA population would result in roughly 72,400 new devices each year. At a cost of approximately $2,000 for CPAP or MADs,38,39 a total of $145 million per year would be incurred as direct expenses.

Currently, public coverage for treatment of OSA varies widely across Canadian jurisdictions, which translates into differences in access. Ontario, Saskatchewan, Newfoundland and Labrador, Manitoba, Yukon, and some federal programs for special populations, including military personnel, support CPAP therapy for OSA patients, by either leasing equipment or reimbursing part of the acquisition cost.2 Further, criteria for patient selection and monitoring using CPAP, as well as supply agreements for lease or reimbursement, vary across these jurisdictions.2 With regard to dental devices, no provincial programs reimburse their cost, while some federal programs will do so for eligible patients.

Given the broad range of therapeutic approaches and the diversity of clinical presentations influenced by OSA severity, symptoms, and comorbidities, the major common issue in Canadian jurisdictions is the challenge of selecting the most appropriate therapy for OSA patients with different clinical profiles and treatment histories.

1.2. Policy Question

What is the optimal use of PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults?

 

2. Objectives

The aim of this health technology assessment (HTA) was to inform the policy question through an assessment of the clinical effectiveness, cost-effectiveness, patient perspectives and experiences, ethical issues, implementation issues and environmental impacts of PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults.

2.1. Research Questions

This HTA addressed the following research questions:

  1. What are the clinical effectiveness, comparative clinical effectiveness, and safety of PAP devices, EPAP valves, oral appliances, surgical interventions, and lifestyle modifications for the treatment of OSA in adults?

    1a. What are the clinical effectiveness, comparative clinical effectiveness, and safety of PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of adult patients with different OSA severity (i.e., mild, moderate, severe)?

    1b. What are the clinical effectiveness, comparative clinical effectiveness, and safety of interventions for the treatment of adult OSA patients with or without comorbidities (e.g., obesity, hypertension, diabetes)?

  1. What is the cost-effectiveness of PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults?
  1. What are the experiences and perspectives of adult patients, their family members, and their caregivers regarding PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults?
  1. What ethical issues are raised by providing PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults? How should these issues be addressed?
  1. What are some of the implementation issues associated with PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults?
  1. What are some potential environmental impacts associated with PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults?

 

3. Clinical Review

This section addressed Research Question 1: What are the clinical effectiveness, comparative clinical effectiveness, and safety of PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults? This section also addressed the following two sub-questions: 1a. What are the clinical effectiveness, comparative clinical effectiveness, and safety of PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of adult patients with different OSA severity (i.e., mild, moderate, severe)? 1b. What are the clinical effectiveness, comparative clinical effectiveness, and safety of interventions for the treatment of adult OSA patients with or without comorbidities (e.g., obesity, hypertension, diabetes)?

3.1. Methods

3.1.1. Study Design

An overview of SRs, meta-analyses (MAs), and HTAs available in the literature on the clinical effectiveness, comparative clinical effectiveness, and safety of interventions for the treatment of OSA in adults was conducted, considering the large volume of literature currently available. Where no published SRs, MAs, or HTAs on any given intervention-comparator combination or certain intervention-comparator-outcome combinations of interest were identified, an SR of primary studies was conducted.

A protocol for the overview of reviews (CRD42016036348)40 was written a priori and followed throughout the review process. A protocol for the review of primary studies was written a priori, as an addendum to the protocol for the overview40 to address gaps within the published SR, MA, and HTA literature, and followed throughout the review process. Any changes to the protocol were identified, and reasons for the changes were provided, throughout this report and in the Protocol Amendments table (see Protocol Amendments).

3.1.2. Literature Search Strategy

The literature search was performed by an information specialist, using a search strategy peer-reviewed according to the PRESS checklist - an evidence-based checklist for the peer review of electronic search strategies.41

Published literature for the original search for SRs, MAs, and HTAs was identified by searching the following bibliographic databases: MEDLINE (1946-), with in-process records and daily updates, via Ovid; Embase (1974-) via Ovid; the Cochrane Database of Systematic Reviews (CDSR) and the Database of Abstracts of Reviews of Effects (DARE) via Ovid; and PubMed. Published literature for the supplemental search for primary studies was identified by searching the following bibliographic databases: MEDLINE (1946-) with in-process records and daily updates via Ovid; Embase (1974-) via Ovid; The Cochrane Central Register of Controlled Trials (CENTRAL) via Ovid; and PubMed. The search strategy comprised both controlled vocabulary, such as the National Library of Medicine's Medical Subject Heading (MeSH) terms, and keywords. The main search concepts for the original search for SRs, MAs, and HTAs were sleep apnea and sleep-disordered breathing. The main search concepts for the supplemental search for primary studies were sleep apnea, sleep-disordered breathing, PAP devices, EPAP, MADs, TRDs, MMA, GTA, and lifestyle modifications.

For the original search, methodological filters were applied to limit retrieval to SRs, MAs, HTAs, network meta-analyses (NMAs), overviews of reviews, and guidelines. Retrieval was limited to documents published since January 1, 2011, considering the large volume of literature currently available on clinical effectiveness. For the supplemental search, no methodological filters were applied to limit retrieval. For this search, retrieval was limited to documents published since January 1, 2006. Both searches were limited to English- or French-language publications. Conference abstracts were excluded from the search results. Detailed strategies for all searches can be found in Appendix 1.

The original search was completed on February 26. The supplemental search was completed on May 13, 2016. Regular alerts were established to update the searches until the publication of the final report. Regular search updates were performed on databases that do not provide alert services. Studies identified in the alerts and that met the selection criteria of the review were incorporated into the analysis if they were identified prior to the completion of the stakeholder feedback. Any studies that were identified from the external peer-reviewer phase until the publication of the report were described briefly in the discussion, with a focus on comparing the results of these new studies to the results of the analysis conducted for this report.

Grey literature (literature that is not commercially published) was identified by searching the Grey Matters checklist42 (https://www.cadth.ca/grey-matters), which includes the websites of HTA agencies, clinical guideline repositories, SR repositories, economics-related resources, patient-related groups, and professional associations. Google and other Internet search engines were used to search for additional Web-based materials. These searches were supplemented by reviewing the bibliographies of key papers and through contacts with appropriate experts and industry. For the supplemental search, an additional search of clinical trial registries was undertaken to retrieve study data from completed trials. More information on the grey literature search strategy can be found in Appendix 1.

3.1.3. Selection Criteria

Overview of Reviews

Inclusion Criteria

The inclusion criteria for the overview of reviews can be found in Table 1. For the overview, the selection criteria apply to the criteria used by the potentially relevant SRs, MAs, and HTAs in identifying primary studies to include. In addition to the list of outcomes pre-specified in the protocol,40 success or cure rate was identified during the review process as a relevant outcome. An amendment was made to the protocol40 to add it to the list of outcomes.

Table 1: Inclusion Criteria for the Overview of Reviews

Population

  • Adults (i.e., aged ≥ 18 yearsa), diagnosed with any severity of OSA (either treatment-naive or previously treated), as measured objectively by PSG or portable monitoring (Type I to Type IV sleep monitors)b

    Subgroups:

    • With or without comorbidities, except heart failure or strokec
    • OSA severity (i.e., mild, moderate, or severe, assessed by baseline AHI, ODI, or RDI)
    • EDS (i.e., mild, moderate, or severe, assessed by ESS)
    • Sex (i.e., male or female)
    • Age (e.g., < 50 years or ≥ 50 years)
    • BMI (e.g., < 30 kg/m2 or ≥ 30 kg/m2)
    • Adherence (e.g., < 4 hours/night or ≥ 4 hours/night for CPAP or OAs)
    • Treatment duration (e.g., ≤ 12 weeks or > 12 weeks)

Intervention

  • PAP devices as follows:
    • A/Bi/CPAP
  • EPAP
  • OAs as follows:
    • MADd
    • TRD
  • Surgical interventions as follows:
    • MMA
    • GTA
  • Lifestyle modificationse as follows:
    • Exercise program
    • Diet or weight-loss program
    • Positional therapyf
  • Combination therapy (i.e., combinations of two or more interventions in scope)

Comparator

  • Inactive controls (e.g., pre-treatment,g oral placebo, sham therapy, or supportive care)
  • Active controls (i.e., other interventions in scope)

Outcome

Primary outcome

  • EDS (assessed by ESS)h

Secondary outcomes

  • OSA severity (assessed by AHI,i ODI, or RDI)
  • Success or cure rate (assessed by AHIi and defined to depict large reductions in AHI or low post-treatment levels of AHI)
  • BP (e.g., daytime, morning, or 24-hour, measured in office or home)
  • Type 2 diabetes mellitus (i.e., incidence or markers of diabetes in diabetic populations [e.g., A1C, insulin resistance])
  • CVEs (i.e., hypertension, AF, or MI)
  • CBEs (i.e., stroke)
  • Accidents (i.e., occupational or motor vehicle)
  • Cognitive function (e.g., memory or concentration, assessed using standardized scales)
  • Psychological function (i.e., depression or anxiety, assessed using standardized scales)
  • Health-related QoL (assessed using standardized scales)
  • Mortality
  • AEs (i.e., any types, including surgical complications, harms, and treatment withdrawal due to AEs)
  • Adherencei (e.g., proportions of patients adhering to treatment)
  • Snoring (assessed using standardized scales)
  • Fatigue (assessed using standardized scales)
  • Change in facial aesthetics (for MMA only)

Study Design

  • SRs, MAs, and HTAs

Time Frame

  • Publications within the last 5 years (i.e., between January 2011 and March 2016)

A1C = glycated hemoglobin; AE = adverse event; AF = atrial fibrillation; AHI = Apnea-Hypopnea Index; APAP = autotitrating positive airway pressure; BiPAP = bilevel positive airway pressure; BMI = body mass index; BP = blood pressure; CBE = cerebrovascular event; CPAP = continuous positive airway pressure; CVE = cardiovascular event; EDS = excessive daytime sleepiness; EPAP = expiratory positive airway pressure; ESS = Epworth Sleepiness Scale; GTA = genial tubercle advancement; HF = heart failure; HTA = health technology assessment; MA = meta-analysis; MAD = mandibular advancement device; MI = myocardial infarction; MMA = maxillomandibular advancement; OA = oral appliance; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; PAP = positive airway pressure; PSG = polysomnography; QoL = quality of life; RDI = respiratory disturbance index; SR = systematic review; TRD = tongue-retaining device.
a Studies that included participants aged < 18 years old would be included if ≥ 80% were adults aged ≥ 18 years.
b Studies that did not identify criteria for diagnosing OSA would still be included. Studies that included non-OSA would be included if ≥ 80% were diagnosed with any severity of OSA or if data for participants with OSA were presented separately.
c Studies of patients with HF or stroke would be excluded because central sleep apnea might occur with those conditions.
d Only personalized MADs, and not over-the-counter, non-personalized devices, were included. If it was unclear from the study report whether in cases where the device was personalized, the study was included.
e Lifestyle interventions, including clinician-directed or -prescribed programs, were considered as interventions, while advice was considered as inactive control.
f Positional therapy prevents patients from sleeping in the supine position (e.g., by attaching a tennis ball onto the back of patients' pyjamas).
g Pre-treatment refers to baseline in a pre-and-post study design, where patients serve as their own controls.
h EDS severity, based on ESS scores, was defined as follows: normal or mild from 0 to 9; moderate from 10 to 15; and severe from 16 to 24.
i OSA severity, based on AHI, was defined as follows: normal < 5; mild from 5 to < 15; moderate from 15 to < 30; and severe from and > 30.
j Adherence was used as an umbrella term, depicting levels of both adherence (e.g., the number of hours/night of device use) and continuation (e.g., rates of device use over time).

There was no restriction regarding the therapy duration or length of follow-up. To be included, SRs and MAs had to have the term "systematic review" or "meta-analysis" in the title or elsewhere in the text; include a detailed description of comprehensive selection criteria and search methods (i.e., as described in the Assessment of Multiple Systematic Reviews [AMSTAR] checklist item #3, with at least two electronic sources having been searched, with adequate reporting of years searched, databases used, and keywords or MeSH terms used and, where feasible, the search strategy provided); assess the quality, or risk of bias, of included studies; and synthesize the findings quantitatively or qualitatively.43 For SRs that did not conduct quality assessment of included studies, they were included only if they had relevant outcomes or subgroups that were not present in any of the other SRs included in this report. In this case, quality assessment of primary studies was conducted de novo in duplicate. To be included, HTAs had to comprise all of the aforementioned elements of SRs, with or without an economic analysis. Only the clinical portion of HTAs was used in the clinical review.

Exclusion Criteria

Studies were excluded if they did not meet the inclusion criteria outlined in Table 1 or if they were duplicate publications. Multiple publications of the same study were excluded, unless they provided additional outcomes of interest. Older SRs (based on publication year) identified in the literature search results were excluded if all the included studies were included in newer SRs included in this report. However, two or more SRs with overlapping primary studies were included if they reported different outcomes or identical outcomes but in different subgroups of interest. The degree of overlap was judged by building a matrix of included studies in the SRs and reported within the Results section of this report.

Review of Primary Studies

Inclusion Criteria

The inclusion criteria for the review of primary studies can be found in Table 2. For the review, which was designed to focus on identified gaps within the published SR, MA, and HTA literature, only certain combinations of interventions, comparators, and outcomes were of interest, as identified in Table 2. In addition to the list of outcomes pre-specified in the protocol,40 success or cure rate was identified during the review process as a relevant outcome. An amendment was made to the protocol40 to add it to the list of outcomes.

Table 2: Inclusion Criteria for the Review of Primary Studies

Population

  • Adults (i.e., aged ≥ 18 yearsa), diagnosed with any severity of OSA (either treatment-naive or previously treated), as measured objectively by PSG or portable monitoring (Type I to Type IV sleep monitors)b

    Subgroups:

    • With or without comorbidities, except heart failure or strokec
    • OSA severity (i.e., mild, moderate, or severe, assessed by baseline AHI, ODI, or RDI)
    • EDS (i.e., mild, moderate, or severe, assessed by ESS)
    • Sex (i.e., male or female)
    • Age (e.g., < 50 years or ≥ 50 years)
    • BMI (e.g., < 30 kg/m2 or ≥ 30 kg/m2)
    • Adherence (e.g., < 4 hours/night or ≥ 4 hours/night for CPAP or OAs)
    • Treatment duration (e.g., ≤ 12 weeks or > 12 weeks)

Intervention, Comparator, and Outcome Combinations

Intervention and comparator combinations:

  • EPAP versus active comparators, including:
    • PAP devices (i.e., APAP, BiPAP, or CPAP)
    • OAs (i.e., MADd or TRD)
    • Surgery (i.e., MMA or GTA)
    • Lifestyle modificationse (i.e., exercise, diet, or weight-loss program or positional therapyf)
  • MADd versus active comparators, including:
    • Other OAs (i.e., TRD)
    • Lifestyle modificationse (i.e., exercise, diet, or weight-loss program or positional therapyf)
  • TRD versus inactive and active comparators, including:
    • Inactive controls (e.g., pre-treatment,g oral placebo, sham therapy, or supportive care)
    • PAP devices (i.e., APAP, BiPAP, or CPAP)
    • Surgery (i.e., MMA or GTA)
    • Lifestyle modificationse (i.e., exercise, diet, or weight-loss program or positional therapyf)
  • MMA versus active comparators, including:
    • PAP devices (i.e., APAP, BiPAP, or CPAP)
    • OAs (i.e., MADd or TRD)
    • Lifestyle modificationse (i.e., exercise, diet, or weight-loss program or positional therapyf)
  • GTA versus inactive and active comparators, including:
    • Inactive controls (e.g., pre-treatment,g oral placebo, sham therapy, or supportive care)
    • PAP devices (i.e., APAP, BiPAP, or CPAP)
    • OAs (i.e., MADd or TRD)
    • Other surgery (i.e., MMA)
    • Lifestyle modificationse (i.e., exercise, diet, or weight-loss program or positional therapyf)
  • Positional therapy versus inactive comparators, including:
    • Inactive controls (e.g., pre-treatment,g oral placebo, sham therapy, or supportive care)
  • Combination therapy (i.e., combinations of two or more PAP devices, EPAP, OAs, surgery, or lifestyle modificationse interventions) versus:
    • Inactive controls (e.g., pre-treatment, oral placebo, sham therapy, or supportive care)
    • Active comparators (i.e., PAP devices, EPAP, OAs, surgery, and lifestyle modificationse)

For the following outcomes:

Primary outcome

  • EDS (assessed by ESS)h

Secondary outcomes

  • OSA severity (assessed by AHI,i ODI, or RDI)
  • Success or cure rate (assessed by AHIi and defined to depict large reductions in AHI or low post-treatment levels of AHI)
  • BP (e.g., daytime, morning, or 24-hour, measured in office or home)
  • Type 2 diabetes mellitus (i.e., incidence or markers of diabetes in diabetic populations [e.g., A1C, insulin resistance])
  • CVEs (i.e., hypertension, AF, or MI)
  • CBEs (i.e., stroke)
  • Accidents (i.e., occupational or motor vehicle)
  • Cognitive function (e.g., memory or concentration, assessed using standardized scales)
  • Psychological function (i.e., depression or anxiety, assessed using standardized scales)
  • Health-related QoL (assessed using standardized scales)
  • Mortality
  • AEs (i.e., any types, including surgical complications, harms, and treatment withdrawal due to AEs)
  • Adherencej (e.g., proportions of patients adhering to treatment)
  • Snoring (assessed using standardized scales)
  • Fatigue (assessed using standardized scales)
  • Change in facial aesthetics (for MMA only)

Intervention and comparator combinations:

  • MMA versus inactive comparators, including:
    • Inactive controls (e.g., pre-treatment,g oral placebo, sham therapy, or supportive care)

For the following outcomes:

  • Mortality
  • Change in facial aesthetics

Study Design

For intervention and comparator combinations that involve an active comparator, studies of the following designs were considered for inclusion:

  • RCTs
  • Non-randomized controlled studies (i.e., controlled clinical trials, cohort studies, case-control studies, and controlled before-and-after studies)

In addition, for intervention and comparator combinations that include an inactive comparator, studies of the following design were additionally considered for inclusion:

  • Uncontrolled pre-and-post studies

Time Frame

  • Publications within the last 10 years (i.e., between January 2006 and May 2016)k

A1C = glycated hemoglobin; AE = adverse event; AF = atrial fibrillation; AHI = Apnea-Hypopnea Index; APAP = autotitrating positive airway pressure; BiPAP = bilevel positive airway pressure; BMI = body mass index; BP = blood pressure; CBE = cerebrovascular event; CPAP = continuous positive airway pressure; CVE = cardiovascular event; EDS = excessive daytime sleepiness; EPAP = expiratory positive airway pressure; ESS = Epworth Sleepiness Scale; GTA = genial tubercle advancement; HF = heart failure; MAD = mandibular advancement device; MI = myocardial infarction; MMA = maxillomandibular advancement; OA = oral appliance; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; PAP = positive airway pressure; PSG = polysomnography; QoL = quality of life; RCT = randomized controlled trial; RDI = respiratory disturbance index; TRD = tongue-retaining device.
a Studies that included participants aged < 18 years old would be included if ≥ 80% were adults aged ≥ 18 years or if data for participants aged ≥18 years were presented separately.
b Studies that did not identify criteria for diagnosing OSA would still be included. Studies that included non-OSA would be included if ≥ 80% were diagnosed with any severity of OSA or if data for participants with OSA were presented separately.
c Studies of patients with HF or stroke would be excluded because central sleep apnea might occur with those conditions.
d Only personalized MADs, and not over-the-counter, non-personalized devices, would be included. If it was unclear from the study report whether in cases the device was personalized, the study would be included.
e Lifestyle interventions, including clinician-directed or -prescribed programs, would be considered as interventions, while advice would be considered as inactive control.
f Positional therapy prevents patients from sleeping in the supine position (e.g., by attaching a tennis ball onto the back of patients' pyjamas).
g Pre-treatment refers to baseline in a pre-and-post study design, where patients serve as their own controls.
h EDS severity, based on ESS scores, was defined as follows: normal or mild from 0 to 9; moderate from 10 to 15; and severe from 16 to 24.
i OSA severity, based on AHI, was defined as follows: normal < 5; mild from 5 to < 15; moderate from 15 to < 30; and severe from and > 30.
j Adherence was used as an umbrella term, depicting levels of both adherence (e.g., the number of hours/night of device use) and continuation (e.g., rates of device use over time).
k The date limit of 10 years was established in consultation with clinical experts, based on their understanding that this limit would capture studies relevant to current clinical practice.

Exclusion Criteria

Studies were excluded if they did not meet the selection criteria outlined in Table 2, if they were case series or case reports, or if they were duplicate publications. Multiple publications of the same study were excluded, unless they provided additional information on the outcomes of interest. There was no restriction regarding the therapy duration or length of follow-up. Studies were excluded if they were not published in English or French.

3.1.4. Screening and Selection of Studies

Two reviewers independently screened the titles and abstracts of all citations retrieved from the literature search, followed by an independent review of the full-text articles, based on the pre-determined selection criteria outlined in Table 1 for the overview of reviews or Table 2 for the review of primary studies. The two reviewers then compared their included and excluded studies from their full-text review and resolved any disagreements through discussion until consensus was reached, involving a third reviewer when necessary.

The study selection process was presented in a Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart43 and can be found in Appendix 2. A list of included studies can be found in Appendix 3. A list of excluded studies, with reasons for exclusion after full-text review, can be found in Appendix 4.

3.1.5. Data Extraction Strategy

Standardized data extraction forms were designed a priori to document and tabulate all relevant information from included studies.

For the overview of reviews, relevant information included both descriptive data and results reported in all included SRs, such as the numbers and types of primary studies included in the SRs; sample size, age, OSA severity, and comorbidities of the study populations included in the SRs; types of interventions and controls, as well as the therapy durations and lengths of follow-up, included in the SRs; types of outcomes included in the SRs; types of subgroup analyses, meta-regression, and MAs, if conducted in the SRs; results (i.e., either narrative description or pooled effect sizes for outcomes meta-analyzed) and conclusions regarding the outcomes of interest (Table 1) and comorbidities reported in the SRs; and types of tools used and results reported regarding quality assessment, if conducted in the SRs.

For the review of primary studies, relevant information included both descriptive data and results reported in all included studies; for example, participant characteristics, types of interventions and controls, therapy duration, length of follow-up, outcomes, results, and subgroup analyses.

Data were extracted by one reviewer and checked for accuracy by a second reviewer. Disagreements were resolved through discussion until consensus was reached, involving a third reviewer when necessary.

3.1.6. Quality Assessment Strategy

Overview of Reviews

The Risk of Bias in Systematic Reviews (ROBIS) tool,44 designed to assess the risk of bias in SRs of RCTs and non-randomized studies, with its 21 questions across four domains, was used as the primary instrument for evaluating the included SRs. Each question was answered as "yes," "probably yes," "probably no," "no," or "no information," with "yes" indicating very low concerns and "no" indicating very high concerns about potential bias.

Four additional criteria (i.e., inclusion of grey literature, provision of a list of included studies and a list of excluded studies, and declaration of conflicts of interest) from the AMSTAR checklist,45 designed to assess the quality of SRs of RCTs, were used to supplement the methodological assessment of the included SRs on items that are not included in the ROBIS tool. While the protocol40 had listed three, and not four, additional criteria, during the review process, an additional item from the AMSTAR checklist45 that was not included in the ROBIS tool was identified (i.e., provision of a list of included studies), and an amendment was made to the protocol40 to include the fourth item. Each question was answered as "yes," "no," "unclear," or "not applicable", with "yes" indicating high quality and "no" indicating low quality.

Because some of the included SRs conducted NMAs, an amendment was made to the protocol40 to also assess the relevance and credibility of the NMAs, using a questionnaire developed by the International Society for Pharmacoeconomics and Outcomes Research (ISPOR),46 with its 26 questions across six domains. Each question was answered "yes," "no," or "cannot answer" to evaluate the level of confidence in the results.

For the one included SR47 that did not conduct quality assessment of included studies but was included in this report as the only report on accidents for outcomes, the Risk of Bias Assessment Tool for Nonrandomized Studies (RoBANS) tool,48 designed to assess the risk of bias in all study designs, including pre-and-post, with its eight domains, was used to evaluate the included primary studies. Each domain was rated as "low," "high," or "unclear," with "low" indicating low concerns and "high" indicating high concerns about potential risk.

A list of the questions from the ROBIS tool, AMSTAR checklist, ISPOR questionnaire, and RoBANS tool can be found in Appendix 5. One reviewer independently assessed the included studies, and another reviewer checked the assessments for accuracy. Disagreements were resolved through discussion, involving a third reviewer when necessary. Although the results of the methodological assessments were not used to exclude the included SRs, the conclusions and discussion of this report focused on the findings of the SRs of higher quality.

Review of Primary Studies

RCTs were assessed using the Cochrane Risk of Bias tool,49 with its seven questions. Each question was answered "yes," "no," or "unclear," with "yes" indicating low concerns and "high" indicating high concerns about potential bias. Non-randomized studies were assessed using the RoBANS tool,48 with its eight domains. Each domain was rated "low," "high," or "unclear," with "low" indicating low concerns and "high" indicating high concerns about potential risk.

A list of the questions from the Cochrane and RoBANS tools can be found in Appendix 5. One reviewer independently assessed the included studies, and another reviewer checked the assessments for accuracy. Disagreements were resolved through discussion, involving a third reviewer when necessary. Although the results of the methodological assessments were not used to exclude the included primary studies, the conclusions and discussion of this report focused on the findings of the primary studies of higher quality.

3.1.7. Data Analysis and Synthesis Strategy

Description of Study Characteristics and Findings

For the overview of reviews, a summary of SR characteristics, including the total number of SRs, as well as the number of primary studies and patients included in each SR, by population, intervention, comparator, outcomes, and study design (PICOS) elements; years of publication, and countries of development and findings was provided in the form of tables and a narrative summary. While the protocol40 had also listed quality assessment of primary studies included in the SRs as part of this summary, an amendment was made to the protocol40 to present it in the quality assessment section of this report, together with the quality assessment of the included SRs. In cases where more than one SR was included for a given intervention, comparator, and outcome of interest, any overlap of included studies among the included SRs was described and presented, by preparing a matrix of included studies in the SRs. A narrative synthesis of the results of included SRs was conducted.

For the review of primary studies, a summary of study characteristics, including the total number of studies by PICOS elements, years of publication, and countries of development, was provided in the form of tables and a narrative summary. The feasibility of conducting MAs of the included primary studies was explored but deemed inappropriate, because of clinical heterogeneity related to the study design and length of follow-up of the included studies. Instead, a narrative synthesis of the results of included primary studies was conducted.

The findings were grouped based on interventions and comparators, and an overview of the outcomes was synthesized narratively, highlighting any trends across studies, subgroup-specific findings, and short-term versus long-term effects. Active and direct comparisons between interventions were reported as such, and no formal testing was conducted to indirectly compare interventions that were not assessed in a head-to-head study. The three PAP devices (i.e., APAP, BiPAP, and CPAP) were considered as one group, because evidence has shown that these devices were similar in adherence and effectiveness.5 For OAs, MADs and TRDs were considered separately. For lifestyle modifications, the interventions were considered separately, unless they had been combined in included SRs, in which case the interventions were considered in combination. No re-synthesis of the findings from primary studies was conducted. Results were presented as reported in the studies, including a summary estimate and confidence intervals, measures of heterogeneity, and numbers of studies and participants contributing to each estimate, as available. Tables were developed to present results by outcome and to accompany the narrative summary, to ensure consistency of presented information across all included studies and to facilitate comparisons by the reader. For each outcome of interest, analysis was conducted for the overall study population and also for each subgroup listed in Tables 1 and 2.

Description of Quality Assessments

For the overview of reviews, a narrative summary of the quality assessment of primary studies included in the SRs was provided. Specifically, a table was developed to summarize the descriptions and overall ratings presented in each included SR, accompanied by a narrative summary of the quality assessment of each included SR. Specifically, a table was developed to present the answers to the questions of the ROBIS tool, using colour codes, where green and light green indicated "yes" and "probably yes," respectively; red and light red indicated "no" and "probably no," respectively; and yellow indicated "no information" for each question. The table also presented the answers to the select questions of the AMSTAR checklist, with colour codes, where green, red, and yellow indicated "yes," "no," and "unclear," respectively, for each question. The answers to the questions of the ISPOR questionnaire were not colour coded, as "yes" and "no" did not always correspond to high and low confidence, respectively. Instead, a table was developed to summarize the assessments.

For the review of primary studies, a narrative summary of the results of the risk of bias assessment for each included study was provided. Specifically, a table was developed to present the answers to the questions of the Cochrane Risk of Bias tool, using colour codes, where green, red, and yellow indicated "yes," "no," and "unclear," respectively, for each question. Another table was developed to present the answers to the questions of the RoBANS tool, using colour codes, where green, red, and yellow indicated "low," "high," and "unclear" risk, respectively, for each domain.

A narrative description of the strengths and limitations of the included studies was also presented in the main text of the report to provide the reader with a holistic, qualitative overview of the literature.

3.2. Results

3.2.1. Quantity of Research Available

The literature search for the overview of reviews yielded 1,087 citations. Upon screening titles and abstracts, 125 potentially relevant articles were retrieved for full-text review. Twenty-three reports were retrieved from other sources (i.e., grey literature, handsearch, and search alerts). Of the 148 potentially relevant articles, 33 SRs5,10,19,24,47,50-77 were included in the overview.

The literature search for the review of primary studies yielded 2,207 citations. Upon screening titles and abstracts, 109 potentially relevant articles were retrieved for full-text review. Six reports were retrieved from other sources (i.e., grey literature, handsearch, and search alerts). Of the 115 potentially relevant articles, 41 primary studies78-118 were included in the review.

The study selection process is outlined in a PRISMA flowchart (Appendix 2). The lists of included and excluded studies are provided in Appendix 3 and Appendix 4, respectively.

3.2.2. Study Characteristics

Overview of Reviews

The characteristics of the included SRs, with respect to the review methods used; numbers, types, and publication years of primary studies included; populations, interventions, comparators, and outcomes studied, and subgroup or meta-regression analyses conducted are summarized in Appendix 6.

Study Dates, Locations, Funding, and Design

The 33 included SRs were published between 2011 and 2016. Twelve SRs were published in 2016,24,50-60 11 in 2015,61-71 five in 2014,10,72-75 three in 2013,19,76,77 and two in 2011.5,47

Based on the location of the corresponding authors, 11 SRs were conducted in China,52,53,57,64-66,69-72,77 seven in the US,5,24,50,55,60,67,68 four in the United Kingdom (UK),19,59,61,76 two in Canada,54,75 two in Italy,10,51 two in Switzerland,62,63 one in Australia,73 one in Greece,47 one in Japan,74 one in South Korea,56 and one in Spain.58

Seven SRs51,60,61,67,70,73,77 did not report their sources of funding. Ten SRs10,24,53-55,57,58,64,72,75 reported that they did not receive any funding. The remaining 16 SRs5,19,47,50,52,56,59,62,63,65,66,68,69,71,74,76 reported that they received funding from non-industry sources.

All SRs, except one,58 conducted MAs for at least one outcome of interest. Two SRs62,63 conducted both pairwise MAs and NMAs for two outcomes of interest (i.e., EDS and blood pressure). One SR55 conducted NMAs for two outcomes of interest (i.e., EDS and OSA severity).

The number of databases and time frames searched to identify primary studies varied across the SRs. Although one SR47 did not perform any quality assessment of its included primary studies, it was included in the overview because its outcome (i.e., road traffic accidents) was of interest to this report and had not been included in any of the other SRs. The quality assessment of the primary studies included in that SR47 was performed by two reviewers (KT and JK), engaging a third reviewer (KS) to resolve any disagreements.

Eighteen SRs5,10,19,51,53,55,57,59,62,63,65,66,70,72-75,77 included RCTs only, with numbers of RCTs ranging from three72 to 80.55 One SR52 included 11 cohort studies only. The remaining 14 SRs24,47,50,54,56,58,60,61,64,67-69,71,76 included both RCTs and observational studies, with numbers of included studies ranging from six64 to 45.60

Populations

The total number of patients in the SRs that included RCTs only varied from 225 from one RCT to 7,88255 from 80 RCTs. The total number of patients in the SR52 that included cohort studies only was 3,112,644 from 11 studies. The total number of patients in the SRs that included both RCTs and observational studies varied from 128 from six studies64 to 60,186 from eight studies.56 All but six SRs58,60,61,66,69,70 reported mean age ranges of patients across the included primary studies, with the lowest mean age at 32 years50 and the highest at 78 years.52,62,75 All but 10 SRs19,24,51,55,58,61,69-71,76 reported that the patients across the included primary studies were predominantly male, with mean proportions varying from 61%57 to 99%.50

All but three SRs52,64,67 reported including primary studies on patients with mixed levels of OSA severity. One SR72 included studies on patients with mild-to-moderate OSA, with the mean AHI ranging from 13 to 23 events/hour. Seventeen SRs5,10,19,50,54-56,58,59,61-63,66,68,75-77 included studies on patients with mild-to-severe OSA, with the lowest mean AHI at 3.5 events/hour50 and the highest at 90 events/hour61). Eleven SRs24,47,51,53,57,65,69-71,73,74 included studies on patients with moderate-to-severe OSA, with the lowest mean AHI at 13 events/hour24 and the highest at 88 events/hour.24 One SR60 included studies on patients with severe OSA, with the mean AHI at 57 events/hour. Three SRs52,64,67 did not specify OSA severity.

All but nine SRs24,52,55,56,69-71,74,77 included patients who were overweight or obese, as classified by the body mass index (BMI). One SR75 included studies on normal-to-obese patients, with the mean BMI ranging from 24.7 kg/m2 to 43 kg/m2. Nineteen SRs5,10,19,47,50,51,53,54,57-59,61-63,66-68,73,76 included studies on overweight-to-obese patients, with the mean BMI ranging from 25 kg/m2 62,67 to 55 kg/m2.76 Four SRs60,64,65,72 included studies on obese patients, with the mean BMI ranging from 30 kg/m2 72 to 43 kg/m2.64 Seven SRs24,52,55,56,69-71,74,77 did not report BMI from their included studies. Other comorbidities reported by the SRs included hypertension65 or resistant hypertension,57,65 type 2 diabetes,64 and nocturia.70

Comparisons, Outcomes, and Subgroups

Twelve major groups of comparisons between the interventions and comparators of interest were identified from the included SRs. Table 3 presents the breakdown of the included SRs into different types of comparisons, highlighting any gaps, where no SRs were found. Table 4 summarizes the outcomes and subgroup or meta-regression analyses reported by the SRs and presents them by comparison. Detailed information can be found in Appendix 7 and Appendix 8.


Table 3: Number of Included Systematic Reviews for Each Comparison

Interventions

Comparators

Inactive Control

A/Bi/CPAP

EPAP

OA (Mixed)

MAD

TRD

Surgery (Mixed)

MMA

GTA

Lifestyle (Mixed)

Weight Loss

Positional Therapy

A/Bi/CPAP

195,10,47,52-57,59,62-67,69,70,75 (3 NMAs55,62,63)

-

x

354,74,77

55,55,59,62,63 (3 NMAs55,62,63)

x

x

x

x

NAa

319,54,55 (1 NMA55)

25,72

EPAP

168

-

-

x

x

x

x

x

x

x

x

x

OA (mixed)

271,74

-

-

-

x

x

x

x

x

x

x

x

MAD

85,51,55,58,59,62,63,75 (3 NMAs55,62,63)

-

-

-

-

15b

x

x

x

x

155c (1 NMA55)

x

TRD

x

-

-

-

-

-

x

x

x

x

x

x

Surgery (mixed)

x

-

-

-

-

-

-

x

x

x

x

x

MMA

160

-

-

-

-

-

-

-

x

x

x

x

GTA

124

-

-

-

-

-

-

-

-

x

x

x

Lifestyle (mixed)

x

-

-

-

-

-

-

-

-

-

x

x

Weight loss

75,19,50,55,61,73,76 (1 NMA55)

-

-

-

-

-

-

-

-

-

155d (1 NMA55)

x

Positional therapy

x

-

-

-

-

-

-

-

-

-

-

-

Combination therapy

15e

                 

119f

x

APAP = autotitrating positive airway pressure; BiPAP = bilevel positive airway pressure; CPAP = continuous positive airway pressure; EPAP = expiratory positive airway pressure; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; NA = not applicable; NMA = network meta-analysis; OA = oral appliance; SR = systematic review; TRD = tongue-retaining device.
Note: Dark grey cells, marked with -, identify duplicate combinations. Light grey cells, marked with x, identify gaps, where no systematic reviews were found.
a Because literature on individual interventions was found, there was no need for literature on mixed interventions.
b This SR was from 2011 and was therefore updated by the review of primary studies.
c This SR was identified after the literature search for the review of primary studies to address gaps within the published SR literature had been conducted.
d This SR was on diet programs versus exercise programs.
e This SR was on TRDs plus positional therapy versus inactive controls.
f The SR was on CPAP plus diet programs versus diet programs.


Table 4: Summary of Outcomes and Subgroup or Meta-Regression Analyses in

Included Systematic Reviews by Comparison

Comparisons

(Number of Reviews [Number of NMAs])

Outcomes

Subgroup or Meta-Regression Analyses

(Outcomes)

1. A/B/CPAP versus inactive controls (19 SRs [3 NMAs])

ESS, AHI, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence

Missing: success rate, snoring, fatigue

  • Baseline ESS (ESS, AHI, BP)
  • Baseline AHI (ESS, AHI, BP, psychological functions, mortality)
  • Baseline BP (BP)
  • Hypertension (BP, CVEs)
  • Diabetes (CVEs)
  • Baseline depression (psychological functions)
  • Age (ESS, BP, CVEs)
  • Sex (BP, CVEs)
  • Baseline BMI (BP, CVEs)
  • Adherence level (ESS, BP, CVEs, diabetes, psychological functions)
  • Study duration (ESS, AHI, BP, CVEs, stroke, psychological functions, mortality)

2. EPAP versus inactive controls (1 SR)

ESS, AHI, snoring

Missing: success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, fatigue

None

3. OA versus inactive controls (10 SRs [3 NMAs])

ESS, AHI, success rate, BP, cognitive functions, psychological functions, QoL, AEs, adherence, snoring

Missing: diabetes, CVEs, stroke, accidents, mortality, fatigue

  • Baseline ESS (ESS, AHI)
  • Baseline AHI (ESS, AHI)
  • Study duration (ESS, AHI)

4. Surgery versus inactive controls (2 SRs)

ESS, AHI, success rate

Missing: BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue, facial aesthetics

  • Baseline AHI (AHI, success rate)

5. Lifestyle interventions versus inactive controls (7 SRs [1 NMA])

ESS, AHI, success rate, BP, diabetes, AEs

Missing: CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, adherence, snoring, fatigue

  • Baseline AHI (AHI)
  • Change in BMI (AHI)
  • Study duration (AHI)

6. CPAP versus OAs (8 SRs [3 NMAs])

ESS, AHI, success rate, BP, cognitive functions, psychological functions, QoL, AEs, adherence

Missing: diabetes, CVEs, stroke, accidents, mortality, snoring, fatigue

  • Baseline ESS (ESS, AHI)
  • Baseline AHI (ESS, AHI, success rate)
  • Study duration (ESS, AHI)

7. CPAP versus lifestyle interventions (5 SRs [1 NMA])

ESS, AHI, cognitive functions, psychological functions, QoL

Missing: success rate, BP, diabetes, CVEs, stroke, accidents, mortality, AEs, adherence, snoring, fatigue

None

8. MADs versus TRDs (1 SR)

AHI

Missing: ESS, success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

9. MADs versus lifestyle interventions (1 SR [1 NMA])

ESS, AHI

Missing: success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

10. Diet versus exercise (1 SR [1 NMA])

ESS, AHI

Missing: success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

11. TRDs plus positional therapy versus inactive controls (1 SR)

AHI

Missing: ESS, success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

12. CPAP plus diet versus diet (1 SR)

ESS, QoL

Missing: AHI, success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, mortality, AEs, adherence, snoring, fatigue

None

AE = adverse event; AHI = Apnea-Hypopnea Index; APAP = autotitrating positive airway pressure; BiPAP = bilevel positive airway pressure; BMI = body mass index; BP = blood pressure; CPAP = continuous positive airway pressure; CVE = cardiovascular event; EPAP = expiratory positive airway pressure; ESS = Epworth Sleepiness Scale; MAD = mandibular advancement device; NMA = network meta-analysis; OA = oral appliance; QoL = quality of life; SR = systematic review; TRD = tongue-retaining device.

1) Positive airway pressure devices versus inactive controls

Nineteen SRs5,10,47,52-57,59,62-67,69,70,75 included studies that compared PAP devices with inactive controls. All 19 SRs included CPAP as the intervention, and no other PAP devices, except for one SR,5 which also included APAP in its reporting of adverse events. Control groups varied across the SRs from pre-treatment to no treatment, sham devices, placebo, usual care, or conservative therapy (e.g., advice on sleep hygiene or weight loss). All but three SRs52,64,67 reported that patients had mild-to-severe5,10,54-56,59,62,63,66,75 or moderate-to-severe47,53,57,65,69,70 OSA. Patients' comorbidities were poorly reported, although most patients were overweight to obese, as classified by BMI. Two SRs57,65 included studies of patients with hypertension or resistant hypertension, one SR64 included studies of patients with diabetes, and one SR70 included studies of patients with nocturia. Study duration of the included primary studies, reported by all but two SRs,67,69 ranged from one week5,62,63,66,75 to 132 months.56 Loss to follow-up in the included primary studies, reported by two SRs,5,53 ranged from 0%5 to 47%.5 Detailed information can be found in Appendix 6 and Appendix 7.

Outcomes reported by the SRs included ESS,5,53-55,59,62,70 AHI/ODI/RDI,5,54,55,59,70 blood pressure,5,10,53,57,63,65 diabetic outcomes,5,64 CVEs,53,56,67,69 stroke,53,56 road traffic accidents,47 cognitive functions,5,66 psychological functions,54,75 QoL,5,54 mortality,52,53,56,69 adverse events,5 and adherence.5 Detailed information can be found in Table 4.

For ESS, one SR59 conducted subgroup analyses on baseline AHI, baseline ESS, and study duration, and one SR62 conducted meta-regression analyses on baseline AHI, baseline ESS, age, adherence level, and study duration. For AHI, one SR59 conducted subgroup analyses on baseline AHI, baseline ESS, and study duration. For blood pressure, four SRs10,57,63,65 conducted subgroup or meta-regression analyses on baseline AHI,10,57,63,65 baseline ESS,10,57,65 baseline blood pressure,57,63,65 hypertension,10,65 age,10,65 sex,65 baseline BMI,57,65 adherence level,10,57,63,65 and study duration.10,57,63,65 For diabetic outcomes, one SR64 conducted subgroup analyses on adherence levels. For CVEs, one SR56 conducted subgroup analyses on adherence level, one SR53 conducted subgroup analyses on study duration, and one SR67 conducted meta-regression analyses on hypertension, diabetes, age, sex, and baseline BMI. For stroke, one SR53 conducted subgroup analyses on study duration. For psychological function, one SR75 conducted subgroup or meta-regression analyses on baseline AHI, depression, adherence level, and study duration. For mortality, one SR69 conducted subgroup analyses on baseline AHI, and one SR53 conducted subgroup analyses on study duration. No subgroup or meta-regression analyses were conducted for road traffic accidents, cognitive functions, QoL, adverse events, or adherence. Detailed information can be found in Appendix 8.

2) Expiratory positive airway pressure versus inactive controls

One SR68 included studies that compared outcomes before and after EPAP. A total of 920 patients had mild-to-severe OSA and were overweight to obese. Study duration of the included primary studies ranged from one night to 12 months. Loss to follow-up in the included primary studies was not reported. Outcomes reported included ESS, AHI, and snoring. No subgroup or meta-regression analyses were conducted for any outcome. Detailed information can be found in Table 4, Appendix 6, Appendix 7, and Appendix 8.

3) Oral appliances versus inactive controls

Ten SRs5,51,55,58,59,62,63,71,74,75 included studies that compared OAs with inactive controls. Control groups varied across the SRs from pre-treatment to no treatment, sham devices, placebo, or usual care. Patients had mild-to-severe55,58,59,62,63 or moderate-to-severe5,51,71,74,75 OSA. Patients' comorbidities were poorly reported, although most patients were overweight to obese, as classified by BMI. Study duration of the included primary studies, reported by all but one SR,55 ranged from one week5,62,63,71,75 to 84 months.58 Loss to follow-up in the included primary studies, reported by three SRs,5,58,74 ranged from 0%5 to 29%.5 Detailed information can be found in Appendix 6 and Appendix 7.

Outcomes reported by the SRs included ESS,5,55,58,59,62,71,74 AHI/ODI,5,55,58,59,71,74 success rate,51 blood pressure,5,63 cognitive functions,5 psychological functions,75 QoL,5,74 adverse events,5,58 adherence,75 and snoring.58 Detailed information can be found in Table 4.

For ESS and AHI, two SRs59,71 conducted subgroup analyses on baseline ESS,59 baseline AHI,59,71 and study duration.59 No subgroup or meta-regression analyses were conducted for any other outcome. Detailed information can be found in Appendix 8.

4) Surgery versus inactive controls

Two SRs24,60 included studies that compared outcomes before and after genioplasty (GP) or GTA24 or MMA with (i.e., 33.6%) or without (i.e., 66.4%) GTA.60 One SR24 included patients who had moderate-to-severe OSA.24 The other SR60 included patients with severe OSA, as measured by mean AHI, who were obese, as measured by mean BMI. Study duration of the included primary studies, reported by one SR,60 ranged from two months to six months. Loss to follow-up in the included primary studies was not reported. Outcomes reported included ESS,24,60 AHI,24,60 RDI,60 and success and cure rates.60 For AHI and success and cure rates, one SR60 conducted subgroup analyses on baseline AHI. No subgroup or meta-regression analyses were conducted for any other outcome. Detailed information can be found in Table 4, Appendix 6, Appendix 7, and Appendix 8.

5) Lifestyle interventions versus inactive controls

Seven SRs5,19,50,55,61,73,76 included studies that compared lifestyle interventions with inactive controls. Lifestyle interventions included exercise or diet programs. Control groups varied across the SRs from pre-treatment to no treatment, general counselling, placebo, usual diet, or usual care. Patients had mild-to-severe5,19,50,55,73,76 or moderate-to-severe61 OSA. Six SRs5,19,50,61,73,76 reported that patients were overweight to obese, while one SR55 did not report on comorbidities. Study duration of the included primary studies, reported by all but one SR,55 ranged from four weeks76 to 94 months.61 Loss to follow-up in the included primary studies, reported by one SR,5 ranged from 3% to 17%. Detailed information can be found in Appendix 6 and Appendix 7.

Outcomes reported by the SRs included ESS,5,19,50,55,73,76 AHI/ODI5,19,50,50,55,61,73,76 treatment response,5 blood pressure,5 diabetes,5 and adverse events.5 Detailed information can be found in Table 4.

For AHI, one SR76 conducted subgroup or meta-regression analyses on baseline AHI, change in BMI, and study duration. No subgroup or meta-regression analysis was conducted for any other outcome. Detailed information can be found in Appendix 8.

6) Continuous positive airway pressure versus oral appliances

Eight SRs5,54,55,59,62,63,74,77 included studies that compared CPAP with MADs5,55,59,62,63 or undefined OAs.54,74,77 Patients had mild-to-severe,55,62,63,77 moderate,54 or moderate-to-severe5,59,74 OSA. Five SRs5,54,59,62,63 reported that patients were overweight to obese, while three SRs55,74,77 did not report on comorbidities. Study duration of the included primary studies, reported by all but one SR,55 ranged from one week62,63 to 157 weeks.62,63 Loss to follow-up in the included primary studies, reported by two SRs,5,74 ranged from 0%5 to 24%.5 Detailed information can be found in Appendix 6 and Appendix 7.

Outcomes reported by the SRs included ESS,5,54,55,59,62,74,77 AHI/ODI,5,54,55,59,74,77 treatment response,5 blood pressure,63,77 cognitive functions,5,77 psychological functions,54,77 QoL,5,54,74,77 adverse events,77 and adherence.5,77 Detailed information can be found in Table 4.

For ESS and AHI, one SR59 conducted subgroup analyses on baseline AHI, baseline ESS, and study duration. For treatment response, one SR5 conducted subgroup analyses on baseline AHI. No subgroup or meta-regression analyses were conducted for any other outcome. Detailed information can be found in Appendix 8.

7) Continuous positive airway pressure versus lifestyle interventions

Five SRs5,19,54,55,72 included studies that compared CPAP with lifestyle interventions. Lifestyle interventions included diet programs,19,55 exercise programs,54,55 and positional therapy.5,72 Patients had mild-to-moderate,72 mild-to-severe,55 moderate,5,54 or moderate-to-severe19 OSA. Four SRs5,19,54,72 reported that patients were overweight,54 overweight to obese,19 or obese,5,72 while one SR55 did not report on comorbidities. Study duration of the included primary studies, reported by all but one SR,55 ranged from three nights72 to 24 months.19 Loss to follow-up in the included primary studies, reported by one SR,5 ranged from 0% to 7%. Detailed information can be found in Appendix 6 and Appendix 7.

Outcomes reported by SRs included ESS,5,19,54,55 AHI/ODI,5,54,55,72 cognitive functions,5 psychological functions,54 and QoL.5,19,54 No subgroup or meta-regression analyses were conducted for any outcome. Detailed information can be found in Table 4 and Appendix 8.

8) Mandibular advancement devices versus tongue-retaining devices

One SR5 included one crossover RCT that compared MADs with TRDs. A total of 22 patients had moderate OSA and were overweight. Study duration of the included primary study was one week for each intervention. Loss to follow-up in the included primary study was 19%. Outcomes reported included AHI. No subgroup or meta-regression analyses were conducted for that outcome. Detailed information can be found in Table 4, Appendix 6, Appendix 7, and Appendix 8.

9) Mandibular advancement devices versus lifestyle interventions

One SR55 included studies that compared MADs with lifestyle interventions. Lifestyle interventions included diet or exercise programs. A total of 7,882 patients had mild-to-severe OSA. Patients' comorbidities, study duration, or loss to follow-up were not reported. Outcomes reported included ESS and AHI. No subgroup or meta-regression analyses were conducted for any outcome. Detailed information can be found in Table 4, Appendix 6, Appendix 7, and Appendix 8.

10) Diet versus exercise

One SR55 included studies that compared diet with exercise programs. A total of 7,882 patients had mild-to-severe OSA. Patients' comorbidities, study duration, or loss to follow-up were not reported. Outcomes reported included ESS and AHI. No subgroup or meta-regression analyses were conducted for any outcome. Detailed information can be found in Table 4, Appendix 6, Appendix 7, and Appendix 8.

11) Tongue-retaining devices plus positional therapy versus inactive controls

One SR5 included one RCT that compared TRD plus posture alarm with no treatment. Patients had mild-to-severe OSA. Patients' comorbidities, study duration, or loss to follow-up were not reported. Outcomes reported included AHI. No subgroup or meta-regression was done for that outcome. Detailed information can be found in Table 4, Appendix 6, Appendix 7, and Appendix 8.

12) Continuous positive airway pressure plus diet versus diet

One SR19 included studies that compared CPAP plus diet programs with diet programs alone. Patients had moderate-to-severe OSA who were overweight to obese. Study duration of the included primary studies ranged from three months to six months. Loss to follow-up in the included primary studies was not reported. Outcomes reported included ESS and quality of life. No subgroup or meta-regression was done for any outcome. Detailed information can be found in Table 4, Appendix 6, Appendix 7, and Appendix 8.

Review of Primary Studies

The characteristics of the included primary studies, regarding the study designs used; populations, interventions, comparators, and outcomes studied; and subgroup analyses conducted are summarized in Appendix 9.

Study Dates, Locations, Funding, and Design

The 41 included primary studies were published between 2006 and 2016. Four studies were published in 2016,78-81 11 in 2015,82-92 five in 2014,93-97 two in 2013,98,99 five in 2012,100-104 five in 2011,105-109 two in 2010,110,111 two in 2009,112,113 two in 2008,114,115 two in 2007,116,117 and one in 2006.118

In line with the location of the corresponding authors, six studies were conducted in the US,79,80,93,96,106,111 six in Netherlands,78,86,88,97,99,104 four in Australia,90,105,109,112 four in Sweden,83,84,107,116 three in France,94,98,113 three in the UK,89,92,95 two in Brazil,100,117 two in Canada,101,115 two in Italy,81,114 two in Turkey,82,91 one in Belgium,87 one in China,103 one in Israel,118 one in Japan,110 one in South Korea,108 one in Switzerland,102 and one in Taiwan.85

Thirteen primary studies79,81,87,89,91,94,95,98,109,111,114,117,118 did not report their sources of funding. Five primary studies78,82,99,103,104 reported that they did not receive any funding. The remaining 23 primary studies80,83-86,88,90,92,93,96,97,100-102,105-108,110,112,113,115,116 reported that they received funding from non-industry sources.

Six studies were RCTs,87,88,90,93,100,115 one of which115 was a crossover study. Five studies80,86,94,107,109 were cohort studies. Thirty studies78,79,81-85,89,91,92,95-99,101-106,108,110-114,116-118 were pre-and-post studies.

Populations

The total number of patients in the primary studies varied from 10110,117 to 198.94 All but six studies79,92,97,100,107,117 reported mean ages of patients, with the lowest mean age at 22 years80 to the highest at 65 years.81 All but five studies79,80,92,101,107 reported that the patients included were predominantly male, with mean proportions varying from 55%111 to 100%.98,100,110,114

All but four studies79,80,89,92 reported including patients at mixed levels of OSA severity, with the lowest mean baseline AHI at 11 events/hour88 and the highest at 64 events/hour.111 Fourteen studies78,81,85,86,88,90,96,97,99,102,104,108,116,118 included positional OSA patients only. All but five studies79,80,92,114,116 reported including patients at mixed levels of obesity, with the lowest mean BMI at 24.8 kg/m2 85 and the highest at 41.7 kg/m2.111

Comparisons, Outcomes, and Subgroups

Eleven major groups of comparisons between the interventions and comparators of interest were identified from the included primary studies. Table 5 presents the breakdown of the included primary studies into different types of comparisons, highlighting any gaps, where no SRs or primary studies were found. Table 6 summarizes the outcomes and subgroup analyses reported by the primary studies and presents them by comparison.


Table 5: Number of Included Primary Studies for Each Comparison

Interventions

Comparators

Inactive Control

A/Bi/CPAP

EPAP

OA (Mixed)

MAD

TRD

Surgery (Mixed)

MMA

GTA

Lifestyle (Mixed)

Weight Loss

Positional Therapy

A/Bi/CPAP

+

-

x

+

+

x

x

x

x

NAa

+

+

EPAP

+

-

-

x

x

x

x

x

x

x

x

x

OA (mixed)

+

-

-

-

x

x

x

x

x

x

x

x

MAD

+

-

-

-

-

1109

x

194

x

x

+

x

TRD

392,113,115

-

-

-

-

-

x

x

x

x

x

x

Surgery (mixed)

NAa

-

-

-

-

-

-

x

x

x

x

x

MMA

389,98,103b

-

-

-

-

-

-

-

x

x

x

x

GTA

+c291,117

-

-

-

-

-

-

-

-

x

x

x

Lifestyle (mixed)

NAa

-

-

-

-

-

-

-

-

-

x

x

Weight loss

+

-

-

-

-

-

-

-

-

-

+

x

Positional therapy

2078,79,81-86,88,90,96,97,99,102,104,105,108,112,116,118

-

-

-

-

-

-

-

-

-

-

-

Combination therapy

680,95,101,110,111,114d

493,100,106,107e

x

x

287,106f

x

x

x

x

x

193g

187h

APAP = autotitrating positive airway pressure; BiPAP = bilevel positive airway pressure; CPAP = continuous positive airway pressure; EPAP = expiratory positive airway pressure; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; NA = not applicable; OA = oral appliance; TRD = tongue-retaining device.
Note: Dark grey cells, marked with -, identify duplicate combinations. Cells, marked with + identify the presence of systematic reviews. Light grey cells, marked with x, identify gaps, where no systematic reviews or primary studies were found.
a Because literature on individual interventions was found, there was no need for literature on mixed interventions.
b These primary studies were on facial aesthetics.
c While GTA versus inactive controls had been identified as an area of gap, a systematic review on the comparison24 was identified from a search alert. Both of the primary studies91,117 had been included in the systematic review,24 which was acknowledged in the presentation of the results.
d Two primary studies were on CPAP plus weight loss, one was on MADs plus TRDs, and three were on MMA plus GTA, all of the pre-versus-post design.
e One primary study was on CPAP plus MADs, one was on CPAP plus exercise, one was on CPAP versus diet, and one was on CPAP plus weight loss, all versus CPAP.
f One primary study was on CPAP plus MADs, and one was on MADs plus positional therapy, both versus MADs.
g One primary study was on CPAP plus weight loss versus weight loss.
h One primary study was on MADs plus positional therapy versus positional therapy.


Table 6: Summary of Outcomes and Subgroup Analyses in

Included Primary Studies by Comparison

Comparisons

(Number of Studies)

Outcomes

Subgroup Analyses

(Outcomes)

1. TRDs versus inactive controls (n = 3)

ESS, AHI, adherence

Missing: success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

2. MMA versus inactive controls (n = 3)

Facial aesthetics

Missing: mortality

None

3. GTA versus inactive controls (n = 2)

ESS, AHI

Missing: success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

4. Positional therapy versus inactive controls (n = 20)

ESS, AHI, success rate, BP, QoL, adherence, snoring

Missing: diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, mortality, AEs, fatigue

  • Baseline AHI (i.e., mild, moderate, or severe)
  • Baseline weight (AHI, adherence, snoring)
  • Positional OSA (AHI)
  • Sleep position (AHI)

5. Combination therapy versus inactive controls (n = 6)

ESS, AHI, mortality, AEs

Missing: success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, AEs, adherence, snoring, fatigue

None

6. MADs versus TRDs (n = 1)

AHI

Missing: ESS, success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

7. MADs versus MMA (n = 1)

AHI

Missing: ESS, success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

8. Combination therapy versus CPAP (n = 4)

ESS, AHI, BP, adherence

Missing: success rate, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, snoring, fatigue

None

9.Combination therapy versus MADs (n = 2)

ESS, AHI

Missing: success rate, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

10. Combination therapy versus weight loss (n=1)

BP

Missing: ESS, AHI, success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

None

11. Combination therapy versus positional therapy (n=1)

AHI

Missing: ESS, success rate, BP, diabetes, CVEs, stroke, accidents, cognitive functions, psychological functions, QoL, mortality, AEs, adherence, snoring, fatigue

  • Sleep position (AHI)

AE = adverse event; AHI = Apnea-Hypopnea Index; APAP = autotitrating positive airway pressure; BiPAP = bilevel positive airway pressure; BMI = body mass index; BP = blood pressure; CPAP = continuous positive airway pressure; CVE = cardiovascular event; EPAP = expiratory positive airway pressure; ESS = Epworth Sleepiness Scale; MAD = mandibular advancement device; OA = oral appliance; QoL = quality of life; TRD = tongue-retaining device.

1) Tongue-retaining devices versus inactive controls

Three pre-and-post studies92,113,115 compared TRDs with pre-treatment. On average, patients had moderate115 or severe113 OSA and were overweight,113,115 as classified by BMI. One study did not report any patient characteristics.92 Study duration, reported by one study,92 was four months. Loss to follow-up, reported by all three studies, ranged from 10%92 to 25%.113 Outcomes reported included ESS,92,113,115 AHI/RDI,113,115 and adherence.92,113 No subgroup analyses were conducted. Detailed information can be found in Table 6 and Appendix 9.

2) Maxillomandibular advancement versus inactive controls

Three pre-and-post studies89,98,103 compared MMA with pre-treatment. On average, patients had severe OSA98,103 and were overweight,89,98,103 as classified by BMI. One study did not report OSA severity.89 Study duration, reported by one study,103 was six months. Loss to follow-up was not reported. Outcomes reported included facial aesthetics. No subgroup analyses were conducted. Detailed information can be found in Table 6 and Appendix 9.

3) Genial tubercle advancement versus inactive controls

Two pre-and-post studies91,117 compared GTA with pre-treatment. Patients had mild-to-severe117 or, on average, severe91 OSA and were obese,91,117 as classified by BMI. Study duration and loss to follow-up were not reported. Outcomes reported included ESS91 and AHI.91,117 No subgroup analyses were conducted. Detailed information can be found in Table 6 and Appendix 9.

While GTA versus inactive controls had been identified as an area of gap, an SR on the comparison24 was identified from a search alert. Both of the primary studies91,117 had been included in the SR,24 which was acknowledged in the presentation of the results.

4) Positional therapy versus inactive controls

Two RCTs,88,90 one cohort study,86 and 17 pre-and-post studies78,79,81-85,96,97,99,102,104,105,108,112,116,118 compared positional therapy with inactive controls. Interventions varied across the studies from mattresses or pillows82-85 to sleep position trainers,78,90,97,99,116 tennis balls,88,102,112,118 neck-positioning devices,79,81,96 waistbands,86 vests,108 vibration devices,104 or supine avoidance devices.78,81,105 Control groups varied across the studies from pre-treatment78,79,81-86,88,96,97,99,102,104,105,108,112,116,118 to supportive care.90 While two studies86,88 had control groups in their RCT88 or cohort study design,86 they compared different types of positional therapy,86,88 which was out of scope for this report. Instead, the pre-and-post data were extracted from the two studies.86,88

Patients had mild,85,86,88,97 moderate,81-83,90,96,99,102,104,105,108,116,118 or severe84,112 OSA, as classified by mean AHI; one study78 included patients with moderate AHI, as classified by median AHI. Patients were overweight,78,81,82,84,86,88,96,97,99,102,104,105,108,112,118 normal weight,85 or obese,90 as classified by mean BMI. One study79 did not report patient characteristics. Two studies83,116 did not report BMI. Study duration, reported by all but four studies,86,88,108,112 ranged from two nights82,84 to 30 months.105 Loss to follow-up, reported by all but three studies,79,82,85 ranged from 0%82,85,104,108,118 to 38%.112

Outcomes reported included ESS,78,82,83,86,88,90,96,97,99,102,116 AHI/ODI/RDI,78,81-86,88,90,96,96,99,102,104,105,108,116,118 success rates,78 blood pressure,90 QoL,78 adherence78,79,83,85,86,88,96,97,99,102,112,118 and snoring.79,85,105 For AHI/ODI, 11 studies78,84-86,88,96,99,102,104,108,118 conducted subgroup analyses on baseline AHI,96 baseline weight,85 positional OSA,84 and sleep position.78,86,88,96,99,102,104,108,118 For adherence and snoring, one study85 conducted subgroup analyses on baseline weight. Detailed information can be found in Table 6 and Appendix 9.

5) Combination therapy versus inactive controls

Six pre-and-post studies80,95,101,110,111,114 compared combination therapy with pre-treatment. Interventions varied across the studies from CPAP plus weight-loss programs,110,111 MADs plus TRDs,101 or MMA plus GTA.80,95,114 On average, patients had severe OSA95,101,110,111,114 and were overweight80,95 or obese,101,110,111 as classified by BMI. One study80 did not report OSA severity. One study114 did not report BMI. Study duration, reported by three studies,101,110,111 ranged from 12 weeks111 to four months.110 Loss to follow-up, reported by all but one study,95 ranged from 0%80,110,111,114 to 6.8%.101 Outcomes reported included ESS95,101,110,111 AHI,95,101,114 mortality,80 and adverse events.80 No subgroup analyses were conducted. Detailed information can be found in Table 6 and Appendix 9.

6) Mandibular advancement devices versus tongue-retaining devices

One cohort study109 compared MADs with TRDs. On average, patients had moderate OSA and were overweight, as classified by BMI. No patient was lost to follow-up. Study duration was not reported. Outcomes reported included AHI. No subgroup analyses were conducted. Detailed information can be found in Table 6 and Appendix 9.

7) Mandibular advancement devices versus maxillomandibular advancement

One cohort study94 compared MADs with MMA. Patients had moderate-to-severe OSA. Of the patients in the MADs and MMA groups, 17.4% and 16.2%, respectively, were obese, as classified by BMI. Study duration and loss to follow-up were not reported. Outcomes reported included AHI. No subgroup analyses were conducted. Detailed information can be found in Table 6 and Appendix 9.

8) Combination therapy versus continuous positive airway pressure

Two RCTs,93,100 one cohort study,107 and one three-intervention pre-and-post study106 compared combination therapy with CPAP. Interventions varied across the studies from CPAP plus MADs106 to CPAP plus diet,107 exercise,100 or weight loss93 programs. Control groups across the studies were CPAP. On average, patients had moderate106 or severe93,100,107 OSA and were overweight100,106 or obese,93,107 as classified by BMI. Study durations, reported by all studies, ranged from eight weeks106 to one year.107 Loss to follow-up, reported by all but one study,107 ranged from 7.1%106 to 31.9%.100 Outcomes reported included ESS,100,106,107 AHI,100,106,107 and blood pressure.93 No subgroup analyses were conducted. Detailed information can be found in Table 6 and Appendix 9.

9) Combination therapy versus mandibular advancement devices

One RCT87 and one three-intervention pre-and-post study106 compared combination therapy with MADs. Interventions varied across the studies from CPAP plus MADs106 to MADs plus positional therapy (i.e., sleep position trainers).87 Control groups across the studies were MADs. On average, patients had moderate OSA and were overweight,87,106 as classified by BMI. Study duration, reported by one study,106 was eight weeks. Loss to follow-up, reported by both studies, ranged from 0%87 to 7.5%.106 Outcomes reported included ESS106 and AHI/ODI.87,106 No subgroup analyses were conducted. Detailed information can be found in Table 6 and Appendix 9.

10) Combination therapy versus weight-loss programs

One RCT93 compared CPAP plus weight-loss programs with weight-loss programs alone. On average, patients had severe OSA and were obese, as classified by BMI. Study duration was 24 weeks, and loss to follow-up was 19.3%. Outcomes reported included blood pressure. No subgroup analyses were conducted. Detailed information can be found in Table 6 and Appendix 9.

11) Combination therapy versus positional therapy

One RCT87 compared MADs plus positional therapy (i.e., sleep position trainers) with positional therapy alone. On average, patients had moderate OSA and were overweight, as classified by BMI. No patient was lost to follow-up. Study duration was not reported. Outcomes reported included AHI/ODI. For AHI, the study conducted subgroup analyses on sleep position. Detailed information can be found in Table 6 and Appendix 9.

3.2.3. Quality Assessment

Overview of Reviews

All of the SRs included in this report, with the exception of one SR,47 assessed the quality of, or risk of bias (RoB) in, their included studies. Seven SRs10,19,50,57,65,66,119 used the Jadad scale. Fourteen SRs51,54-56,62-64,71-77 used the Cochrane RoB tool, whereas two SRs53,70 used select criteria from the Cochrane tool. Other SRs used the Agency for Healthcare Research and Quality (AHRQ) methods guide,5 the Cochrane Effective Practice and Organisation of Care (EPOC) RoB tool,76 the Cochrane Handbook for Systematic Reviews of Interventions,67 the CONSORT criteria,58 the Downs and Black checklist,69 the Effective Public Health Practice Project's (EPHPP's) quality assessment tool,51 the Newcastle-Ottawa scale52 or a modified version of it,61 the National Institute for Health and Care Excellence (NICE) quality assessment tool,24,68 RoBANS,56 a quality control questionnaire,60 and undefined quality criteria.72 Five SRs also used Grading of Recommendations Assessment, Development and Evaluation (GRADE).10,51,54,71,74 For the SR47 with no quality assessment, two reviewers of this report performed quality assessment using RoBANS.

The quality assessment of the included studies conducted by the SRs, as well as descriptions of the different quality assessment tools used, are summarized in Appendix 10. The overall quality of the studies included in each of the SRs was rated as very low to low in one SR,54 low in six SRs,5,24,47,64,71,77 low to moderate in three SRs,5,10,74 moderate in seven SRs,5,59,60,62,67,70,72 moderate to high in three SRs,51,52,58 high in 10 SRs,53,55-57,63,65,66,68,69,73 or mixed in seven SRs.5,19,50,56,61,75,76 Two of the SRs provided multiple ratings, separately for RCTs versus non-RCTs56 or for each intervention-comparator combination.5

The quality assessment of the 33 SRs, conducted with the ROBIS tool and four additional criteria from AMSTAR, is summarized in Appendix 11. Across the SRs, substantial concerns regarding specifications of study eligibility criteria (i.e., Domain 1) were identified, with the rating of moderate or high concerns in 18 SRs, of low concerns in 14 SRs, and of unclear concerns in one SR, where concerns were unclear if relevant information had not been reported. Concerns regarding methods used to identify or select studies (i.e., Domain 2), methods used to collect data and appraise studies (Domain 3), or the synthesis of findings (i.e., Domain 4) were generally low, with the rating (presented for Domains 2, 3, and 4, respectively) of low concerns in 24, 25, and 24 SRs, of moderate concerns in nine, six, and seven SRs, of high concerns in no, two, and two SRs. On a per-SR basis, nine SRs51,54,62,63,67,68,71,75,76 were rated as having low concerns across the four domains, and 12 SRs5,10,19,47,50,53,55-57,59,64,74 as having low concerns across three domains. Three SRs24,60,65,66 were rated as having low-to-moderate concerns across the four domains, and one SR58 as having moderate-to-high concerns across the four domains. All other SRs52,61,69,70,72,73,77 were rated as having mixed levels of concerns across the four domains. Except for one SR,70 with high concerns in two domains, none of the SRs were rated as having high concerns in more than one domain. Of the 33 SRs, one SR included grey literature as an inclusion criterion, and two SRs provided a list of excluded studies. All 33 SRs provided a list of included studies, and 32 SRs disclosed conflicts of interests, noting no concerns.

The quality assessment of the three NMAs,55,62,63 conducted with the ISPOR questionnaire, is summarized in Appendix 12. In terms of relevance (Domain 1), while all three NMAs had relevant populations and applicable contexts, not all relevant interventions and outcomes were included. In terms of credibility (Domain 2), all three NMAs made attempts to identify and include all relevant RCTs, and the included RCTs formed a connected network. Further, no bias was induced by selective outcome reporting. However, while two NMAs62,63 conducted an analysis on study quality, excluding poor-quality studies, and identified systematic differences in treatment effect modifiers, the other NMA55 conducted no such analyses. In terms of analysis (Domain 3), all three NMAs used statistical methods to preserve within-study randomization, evaluated consistency, included both direct and indirect comparisons, and provided a valid rationale for the use of random-effects models. However, while two NMAs62,63 handled inconsistency in the analysis and conducted meta-regression analyses to account for heterogeneity, the other NMA55 did neither. In terms of reporting quality and transparency (Domain 4), two NMAs,62,63 but not the other NMA,55 provided the numbers of RCTs per comparison; reported individual study results, as well as results of both direct and indirect comparisons; and evaluated some patient characteristics on the outcomes. One NMA,55 but not the other two NMAs,62,63 provided rankings, albeit without uncertainty. In terms of interpretation (Domain 5) and conflict of interest (Domain 6), two NMAs62,63 drew appropriate conclusions and declared no conflict of interest, whereas it was not clear for the other NMA55 whether the conclusions for CPAP comparisons were appropriate and how the potential for conflict of interest regarding past funding from OSA device companies was mitigated.

Review of Primary Studies

The quality assessment of the 41 primary studies, conducted with the Cochrane Risk of Bias tool49 on six RCTs87,88,90,93,100,115 or the RoBANS tool48 on 35 non-randomized controlled or pre-and-post studies,78-86,89,91,92,94-99,101-114,116-118 is summarized in Appendix 13 and Appendix 14, respectively.

For the six RCTs,87,88,90,93,100,115 the Cochrane Risk of Bias rating was low risk for two studies88,90 and unclear risk for four studies,87,93,100,115 on a scale of low to high risk, where concerns were unclear if relevant information had not been reported. Across the included studies, sequence generation and allocation concealment were assessed to largely be at unclear risk. Blinding of participants and personnel was mostly not applicable, but blinding of outcome assessment was assessed to be largely low risk. Incomplete outcome and selective reporting were assessed to be largely low risk. Other potential threats to validity remained largely unclear throughout the studies.

For the 35 non-randomized controlled or pre-and-post studies,78-86,89,91,92,94-99,101-114,116-118 the RoBANS rating was low risk for five criteria, mixed for one criterion, high risk for one criterion, and not applicable for one criterion, on a scale of low to high risk. For comparability, exposure measurement, outcome assessment, incomplete outcome data, and selective outcome reporting, 34, 23, 28, 25, and 31 of the 35 studies were rated as having low risk of bias, respectively. For selection of patients, a third of the studies were rated as having low risk of bias, another third was rated as unclear, and the last third was rated as having high risk of bias, where risk was unclear if relevant information had not been reported. For confounding, most of the studies were rated as having high risk of bias. For blinding of assessors, this domain was not applicable in all but one study,94 where risk of bias was unclear. Twenty-eight studies78-86,91,92,94-99,101-110,117 had low risk of bias, two studies111,113 had moderate risk of bias, four studies89,112,116,118 had high risk of bias, and one study114 had unclear risk of bias.

Overall, the 41 studies were assessed to be at a low risk of bias. Further to risk of bias, it is important to note that the majority of the studies included small samples (i.e., fewer than 20 patients) and did not reach statistical significance or did not provide results of statistical tests and may not have been adequately powered. Moreover, 30 of the 41 studies were of the pre-and-post design, with no controls, making it difficult to draw definitive conclusions.

3.2.4. Data Analysis and Synthesis

The findings are presented below, first from the overview of reviews and then from the review of primary studies. For each outcome, a summary of results is provided at the end of each section.

Excessive Daytime Sleepiness

Information on the validity and reliability of EDS, as measured by the ESS, can be found in Appendix 15.

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Seven SRs,5,53-55,59,62,70 including two NMAs,55,62 reported on EDS in adults with mild-to-severe5,54,55,59,62,70 or moderate-to-severe5,53,70 OSA. Five SRs5,53,54,59,62 included overweight-to-obese patients. One SR70 included patients with nocturia only. One SR55 provided no information on comorbidities.

All seven SRs,5,53-55,59,62,70 with sample sizes ranging from 131 patients70 to 7,882 patients55 from two studies70 to 80 studies,55 reported significantly greater reductions in ESS with CPAP, compared with controls5,53-55,59,62 or pre-treatment.54,70 The mean difference in ESS, reported by all seven SRs, ranged from -5.8870 to -0.61,54 with five SRs5,53,55,59,62 reporting mean differences around -2. Study duration of the included primary studies, reported by all seven SRs, ranged from one week5,62 to 157 weeks.62 I2 scores, reported by all six applicable SRs,5,53,54,59,62,70 ranged from 0%54 to 99%70 and were greater than 75% in four SRs.5,53,59,70 The SRs reported the quality of the included studies as very low to low,54 low to moderate,5 moderate,59,62,70 or high53,55 (Appendix 10).

From subgroup or meta-regression analyses, two SRs59,62 reported that the effect of CPAP versus controls on ESS increased with increasing EDS and OSA severity at baseline.59,62 However, while one of the two SRs59 reported that the effect of CPAP versus controls on ESS decreased with longer treatment duration, the other SR62 reported no significant differences in the effect of CPAP versus controls on ESS with varying levels of follow-up duration, CPAP adherence, or age.

Across the seven SRs, 84 primary studies had been included, 34 of which had been included in one SR, 13 in two SRs, 12 in three SRs, 21 in four SRs, and four in five SRs (Appendix 16.1). No two SRs completely overlapped on ESS as the outcome.

The findings of the SRs are summarized in Table 7.

2) Expiratory positive airway pressure versus inactive controls

One SR68 reported on EDS in adults with mild-to-severe OSA who were overweight to obese. The SR,68 with a sample size of 359 patients from five studies, reported significantly greater reductions in ESS with EPAP, compared with pre-treatment. The mean difference in ESS was −2.61. Study duration of the included primary studies ranged from one month to three months. I2 scores ranged from 0% to 18%. The SR68 reported the quality of the included studies as high (Appendix 10). The findings of the SR are summarized in Table 8.

3) Oral appliances versus inactive controls

Seven SRs,5,55,58,59,62,71,74 including two NMAs,55,62 reported on EDS in adults with mild-to-severe55,58,59,62 or moderate-to-severe5,71,74 OSA. Four SRs5,58,59,62 included overweight58 or overweight-to-obese5,59,62 patients. Three SRs55,71,74 provided no information on comorbidities.

Six of the seven SRs,5,55,58,59,62,71 with sample sizes ranging from 397 patients58 to 7,882 patients55 from nine studies58 to 80 studies,55 reported significantly greater reductions in ESS with MADs5,55,58,59,62 or undefined OAs,71 compared with controls5,55,59,62,71,74 or pre-treatment.58 The mean difference in ESS, reported by all six SRs, ranged from -3.6558 to −1.17,5 with five SRs5,55,59,62,71 reporting mean differences around -2. Study duration of the included primary studies, reported by all six SRs, ranged from one week62,71 to 157 weeks.62 I2 scores, reported by all four applicable SRs,5,59,62,71 ranged from 0%5 to 90%71 and were greater than 75% in one SR.71 The SRs reported the quality of the included studies as low,71 moderate,5,59,62 moderate to high,58 or high55 (Appendix 10).

One of the seven SRs,74 with a sample size of 106 patients from three studies, reported no significant differences in change in ESS between undefined OAs and control appliances. Study duration of the included studies ranged from one month to six months. The I2 score was 81%. The SR74 reported the quality of the included studies as low to moderate (Appendix 10).

From subgroup analyses, one SR74 reported that the effect of MADs versus controls on ESS increased with increasing EDS at baseline but decreased with longer treatment durations. Two SRs59,71 reported no significant differences in the effect of MADs59 or undefined OAs71 versus controls on ESS with varying OSA severity at baseline.

Across the seven SRs, 26 primary studies had been included, 15 of which had been included in one SR, two in two SRs, two in three SRs, three in four SRs, three in five SRs, and one in six SRs (Appendix 16.2). No two SRs completely overlapped on ESS as the outcome.

The findings of the SRs are summarized in Table 9.

4) Surgery versus inactive controls

Two SRs24,60 reported on EDS in adults with moderate24 or severe60 OSA who were obese.60 The SRs,24,60 with sample sizes of 20 patients24 to 113 patients60 from two studies24 or an unknown number of studies,60 reported greater reductions in ESS with GP,24 GTA,24 or MMA with or without GTA,60 compared with pre-treatment. The mean difference in ESS was -5.8 with GP and -2.9 with GTA.24 The mean ESS decreased from 13.5 at baseline to 3.2 at follow-up with MMA with or without GTA.60 Study duration of the included primary studies, reported by one SR,60 ranged from two months to six months. I2 scores were not applicable. The SRs24,60 reported the quality of the included studies as low24 or moderate60 (Appendix 10). The findings of the SRs are summarized in Table 10.

5) Lifestyle interventions versus inactive controls

Six SRs,5,19,50,55,73,76 including one NMA,55 reported on EDS in adults with mild-to-severe19,55,76 or moderate-to-severe5,50,73 OSA. Five SRs5,53,54,59,62 included overweight19 or overweight-to-obese5,50,73,76 patients. One SR55 provided no information on comorbidities.

Five of the six SRs,5,50,55,73,76 with sample sizes ranging from 120 patients50 to 7,882 patients55 from two studies5,19,73 to 80 studies,55 reported significantly greater reductions in ESS with diet5,55,73,76 or exercise5,50,55,76 programs, compared with controls5,50,55,73 or pre-treatment.76 The mean difference in ESS, reported by all five SRs, ranged from -45 to -1,5 with three SRs55,73,76 reporting mean differences around -2. Study duration of the included primary studies, reported by all five SRs, ranged from two weeks55 to 144 weeks.55 I2 scores, reported by the two applicable SRs,50,76 ranged from 0%50 to 37%.76 The SRs reported the quality of the included studies as moderate,5 high,55,73 or mixed50,76 (Appendix 10).

Two of the six SRs,19,76 with sample sizes ranging from 114 patients19 to 142 patients76 from two studies19 to three studies,76 reported no significant differences in change in ESS with diet and exercise programs, compared with controls. Study duration of the included primary studies, reported by both SRs, ranged from three months19 to two years.76 I2 scores, reported by both SRs, ranging from 6%76 to 33.5%.19 The SRs19,76 reported the quality of the included studies as mixed (Appendix 10).

Across the six SRs, 15 primary studies had been included, 11 of which had been included in one SR, two in three SRs, one in four SRs, and one in five SRs (Appendix 16.3). Two SRs5,73 completely overlapped on ESS as the outcome. One SR55 included all primary studies included in three other SRs5,19,73 on ESS as the outcome.

The findings of the SRs are summarized in Table 11.

6) Continuous positive airway pressure versus oral appliances

Seven SRs,5,54,55,59,62,74,77 including two NMAs,55,62 reported on EDS in adults with mild-to-severe54,55,62,77 or moderate-to-severe5,59,74 OSA. Four SRs5,54,59,62 included overweight-to-obese patients. Three SRs55,74,77 provided no information on comorbidities.

Six of the seven SRs,5,54,55,59,74,77 with sample sizes ranging from 139 patients54 to 7,882 patients55 from two studies54 to 80 studies,55 reported no significant differences in change in ESS5,54,55,59,74,77 or Stanford Sleepiness Scale (SSS)54 between CPAP and MADs5,55,59 or undefined OAs.54,74,77 Study duration of the included primary studies, reported by all six SRs, ranged from two weeks55 to 144 weeks.55 I2 scores, reported by the five applicable SRs,5,54,59,74,77 ranged from 0%54,74,77 to 89.4%5 and were greater than 75% in two SRs.5,77 The SRs reported the quality of the included studies as very low to low,54 low,77 low to moderate,5,74 moderate,59 or high55 (Appendix 10).

One of the seven SRs,62 with a sample size of 6,873 patients from 67 studies, reported significantly greater reductions in ESS with CPAP, compared with MADs. The mean difference in ESS ranged from -0.9 to -0.8 from network and pairwise comparisons, respectively. Study duration of the included studies ranged from one week to 157 weeks. The I2 score for the pairwise comparison was 67%. The SR62 reported the quality of the included studies as moderate (Appendix 10).

From subgroup analyses, one SR59 reported no significant differences in the effect of CPAP versus MADs on ESS with varying EDS or OSA severity at baseline or study duration. However, severe cases of OSA may benefit more from CPAP than from MADs.

Across the seven SRs, 16 primary studies had been included, five of which had been included in one SR, one in two SRs, two in three SRs, two in four SRs, three in five SRs, and three in six SRs (Appendix 16.4). One SR55 included all primary studies included in four other SRs5,54,62,74 on ESS as the outcome.

The findings of the SRs are summarized in Table 12.

7) Continuous positive airway pressure versus lifestyle interventions

Three SRs,5,54,55 including one NMA,55 reported on EDS in adults with mild-to-severe55 or moderate5,54 OSA. Two SRs5,54 included overweight54 or obese5 patients. One SR55 provided no information on comorbidities.

Two of the three SRs,5,55 with sample sizes ranging from 94 patients5 to 7,882 patients55 from three studies5 to 80 studies,55 reported no significant differences in the effect of CPAP versus diet55 or exercise55 programs or positional therapy.5 Study duration of the included primary studies, reported by both SRs, ranged from two weeks5,55 to 144 weeks.55 I2 scores were not applicable. The SRs reported the quality of the included studies as moderate5 or high55 (Appendix 10).

One of the three SRs,54 with a sample size of 16 patients from one study, reported significantly greater reductions in ESS with CPAP, compared with exercise. The ESS Hedges' g was −0.71. Study duration of the included primary study was 60 days. I2 scores were not applicable. The SR54 reported the quality of the included studies as very low to low (Appendix 10).

Across the three SRs, six primary studies had been included, five of which had been included in one SR, and one in two SRs (Appendix 16.5). Two SRs54,55 completely overlapped on ESS as the outcome.

The findings of the SRs are summarized in Table 13.

8) Mandibular advancement devices versus lifestyle interventions

One SR,55 which included an NMA, reported on EDS in adults with mild-to-severe OSA, with no information on comorbidities. The SR,55 with a sample size of 7,882 patients from 80 studies, reported no significant differences in change in ESS between MADs and diet or exercise programs. Study duration of the included primary studies ranged from two weeks to 144 weeks. I2 scores were not applicable. The SR55 reported the quality of the included studies as high (Appendix 10). The findings of the SR are summarized in Table 14.

9) Diet versus exercise

One SR,55 which included an NMA, reported on EDS in adults with mild-to-severe OSA, with no information on comorbidities. The SR,55 with a sample size of 7,882 patients from 80 studies, reported no significant differences in change in ESS between diet and exercise programs. Study duration of the included primary studies ranged from two weeks to 144 weeks. I2 scores were not applicable. The SR55 reported the quality of the included studies as high (Appendix 10). The findings of the SR are summarized in Table 15.

10) Combination therapy versus active controls

One SR19 reported on EDS in adults with moderate-to-severe OSA who were overweight to obese. The SR,19 with a sample size of 114 patients from two studies, reported significantly greater reductions in ESS with CPAP plus diet programs, compared with diet programs alone. The mean difference in ESS was -3.2. Study duration of the included primary studies ranged from three months to six months. The I2 score was 82.7%. The SR19 reported the quality of the included studies as mixed (Appendix 10). The findings of the SR are summarized in Table 16.

Review of Primary Studies

1) Tongue-retaining devices versus inactive controls

Three studies92,113,115 reported on EDS in adults with moderate115 or severe113 OSA, providing mean AHI. Two studies113,115 included overweight pateints,113,115 providing mean BMI. One study92 provided no information on comorbidities.

All three studies,92,113,115 with sample sizes ranging from 20 patients92 to 84 patients,113 reported reductions in ESS with TRDs, compared with pre-treatment, with two of the studies113,115 reporting statistical significance. The mean difference in ESS, reported by all three studies, ranged from -4.892 to -1.5.115 One study115 reported that using TRDs with suction led to better outcomes than using TRDs without suction. Concerns with the quality of the three studies were assessed to be low,92 moderate,113 or unclear115 (Appendix 13 and Appendix 14).

The findings of the primary studies are summarized in Table 17.

2) Genial tubercle advancement versus inactive controls

One study91 reported on EDS in adults with moderate OSA who were obese, providing mean AHI and BMI. The study,91 with a sample size of 17 patients, reported a significant reduction in ESS with GTA, compared with pre-treatment. The mean difference in ESS was -2.9. However, the study authors cautioned that only individuals with less severe OSA may benefit from this type of surgery. The study91 had been included in, and its findings reported by, the SR on the comparison24 that was later identified from a search alert. Concerns with the quality of the study were assessed to be low91 (Appendix 14). The findings of the primary study are summarized in Table 18.

3) Positional therapy versus inactive controls

Eleven studies78,82,83,86,88,90,96,97,99,102,116 reported on EDS in adults with mild,87,88,97 moderate,78,82,83,96,99,102,116 or severe OSA,84 providing mean78,82-84,87,88,96,102,116 or median AHI.97,99 Nine studies78,86,88,90,96,97,99,102,116 included positional OSA patients only. Ten studies included patients who were overweight,78,82-84,87,88,96,97,99,102 providing mean78,82-84,87,88,96,102 or median BMI.97,99 One study116 provided no information on comorbidities.

Eight of the 11 studies,78,83,86,96,97,99,102,116 with sample sizes ranging from 14 patients83 to 145 patients,97 reported reductions in ESS with positional therapy (i.e., with tennis balls,102 pillows,83 an apparatus designed to prevent sleep in the supine position,78,96,97,99,116 or an apparatus designed to mimic the tennis ball technique86), compared with pre-treatment, with seven of the studies78,83,86,97,99,102,116 reporting statistical significance. The mean difference in ESS, reported by all eight studies, ranged from -497 to -1.6,116 with five studies86,96,99,102,116 reporting mean differences around -2. Concerns with the quality of the eight studies were assessed to be low78,83,86,96,97,99,102 or high116 (Appendix 14).

Three of the 11 studies,82,88,90 with sample sizes ranging from 29 patients82 to 86 patients,90 reported little change in ESS with positional therapy (i.e., with tennis balls,90 pillows,82 or an apparatus designed to prevent sleep in the supine position88), compared with controls90 or pre-treatment.82,88 Concerns with the quality of the 11 studies were assessed to be low82,88,90 (Appendix 13 and Appendix 14).

The findings of the primary studies are summarized in Table 19.

4) Combination therapy versus inactive controls

Four studies95,101,110,111 reported on EDS in adults with severe OSA95,101,110,111 providing mean95,101,110,111 AHI/RDI. moderate-to-severe95,101 or severe110,111 OSA. The studies included overweight95 or obese patients,101,110,111 providing mean95,101,110,111 BMI.

One of the four studies,111 with a sample size of 11 patients, reported non-significant reductions in ESS with CPAP plus weight loss, compared with pre-treatment. The mean difference in ESS was -3. However, another study,110 with a sample size of 10 patients, reported no difference in ESS between CPAP plus weight loss and pre-treatment. The other two studies,95,101 with sample sizes ranging from 41 patients101 to 51 patients,95 reported significant reductions in ESS with TRDs plus MADs101 or MMA plus GP,95 compared with pre-treatment. The mean difference in ESS, reported by both studies, ranged from -995 to -3.1.101 Concerns with the quality of the four studies were assessed to be low95,101,110 or moderate111 (Appendix 14).

The findings of the primary studies are summarized in Table 20.

5) Combination therapy versus active controls

Three studies100,106,107 reported on EDS in adults with moderate106 or severe100,107 OSA, providing mean100,106,107 AHI. The studies included overweight100,106 or obese107 patients, providing mean100,106,107 BMI.

All three studies,100,106,107 with sample sizes ranging from 10 patients106 to 63 patients,107 reported greater reductions in ESS with CPAP plus MADs,106 CPAP plus exercise,100 or CPAP plus diet programs,107 compared with CPAP alone, with two of the studies106,107 reporting statistical significance. The mean difference in ESS, reported by two studies,106,107 ranged from
-2.2 106 to -2.107 One of the three studies,106 with a sample size of 10 patients, also reported a significantly greater reduction in ESS with CPAP plus MADs, compared with MADs alone. Concerns with the quality of the three studies were assessed to be low,106,107 or unclear100 (Appendix 13 and Appendix 14).

The findings of the primary studies are summarized in Table 21.

Summary of Results on Excessive Daytime Sleepiness

For ESS, evidence was found on inactive comparisons with CPAP, EPAP, OAs (i.e., MADs, TRDs, and undefined OAs), surgery (MMA and GTA), lifestyle interventions (i.e., diet, exercise, and positional therapy), and combination therapy (i.e., CPAP plus weight loss, MADs plus TRDs, and MMA plus GTA). Evidence was also found on active comparisons between CPAP and OAs (i.e., MADs and undefined OAs), CPAP and lifestyle interventions (i.e., diet, exercise, and positional therapy), MADs and lifestyle interventions (i.e., diet and exercise), diet and exercise, and combination therapy and other interventions (i.e., CPAP plus MADs versus MADs alone, CPAP plus diet versus CPAP alone, and CPAP plus exercise versus CPAP alone).

Compared with inactive controls or pre-treatment, CPAP, EPAP, MADs, TRDs, undefined OAs, MMA, GTA, diet, exercise, positional therapy, CPAP plus weight loss, MADs plus TRDs, and MMA plus GTA were all effective at reducing EDS, commonly measured by ESS. While there is no commonly accepted clinically important difference for ESS (Appendix 15), effect sizes across the interventions were similar, with mean differences around -2. Mean differences in ESS around -10 were observed with MMA with or without GTA from small, uncontrolled studies on severe cases of OSA.60,95 The findings on EPAP were also from uncontrolled studies.68 The similarity in effect sizes is also reflected in the findings on active comparisons, where no significant differences in ESS scores were found between CPAP and MADs or undefined OAs; CPAP and diet, exercise, or positional therapy; MADs and diet or exercise; and diet and exercise. These findings suggest that the various interventions may be comparable in their effectiveness in improving EDS, although no formal statistical testing was conducted. MMA with or without GTA demonstrated the largest effect size for severe cases OSA, who are eligible for surgery. Some of the findings on CPAP or MADs versus inactive controls and CPAP versus OAs were associated with high heterogeneity.

Combination therapy was more effective at reducing EDS, compared with the interventions alone, specifically for CPAP plus diet programs versus CPAP or diet programs alone, CPAP plus exercise programs versus CPAP alone, and CPAP plus MADs versus MADs alone. Effect sizes were similar, with mean differences in ESS across the comparisons around -2. These findings suggest that the various interventions in combination may have additive effects in their effectiveness in improving EDS.

Patients with nocturia experienced reduced EDS after CPAP. Subgroup and meta-regression analyses suggest that patients with higher EDS at baseline experienced greater effects with CPAP, MADs, or undefined OAs. Patients with more severe OSA at baseline experienced greater effects with CPAP but not with MADs or undefined OAs. Longer study duration was associated with lower effects with CPAP and MADs. Age and CPAP adherence levels were not significantly associated with the effects of CPAP. No subgroup or meta-regression analyses were found on sex or BMI.


Table 7: Summary of Change in ESS From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Guo 201653

CPAP versus inactive controls

ESS was significantly decreased in the CPAP group, compared with the inactive control group.

2,743 patients from 7 RCTs (2010 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 28 to 45 events/hour)
  • Overweight to obese (mean BMI range: 28 to 34 kg/m2)
  • Diabetes (% range: 29% to 38%, where reported)
  • Smoking (% range: 23% to 84%, where reported)
  • 2 to 48 months of study duration
  • ESS MD (95% CI) =
    -1.78
    (-2.31 to
    -1.24); P < 0.00001; I2 = 76%

None

High

Gupta 201654

CPAP pre versus post

The effect size was high for CPAP, compared with pre-treatment.

The effect size was moderate for CPAP, compared with oral placebo.

832 patients from 16 studies (1998 to 2013):

  • Mild-to-severe OSA (mean AHI range: 11 to 71.5 events/hour, where reported)
  • Overweight to obese (mean BMI range: 27.8 to 38 kg/m2, where reported)
  • 11 days to 2 years of study duration
  • ESS Hedges' g (95% CI) = -1.02 (-1.33 to -0.70); P < 0.001; I2 = 0%

None

Very low to low

CPAP versus oral placebo

187 patients from 4 studies (1997 to 2004):

  • Mild-to-severe OSA (mean AHI range: 11 to 43 events/hour)
  • Overweight to obese (mean BMI range: 29.8 to 31.3 kg/m2)
  • 4 weeks to 3 months of study duration
  • ESS Hedges' g (95% CI) = -0.61 (-1.13 to -0.09); P = 0.021; I2 = 0%

None

Very low to low

Iftikhar 201655

CPAP versus inactive controls

ESS was significantly decreased with CPAP, compared with inactive controls.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • ESS MD (95% CI) = −2.43 (-2.95 to -1.92); P = NR

None

High

Sharples 201659

CPAP versus inactive controls

Overall, ESS was significantly decreased in the CPAP group, compared with the inactive control group.

The effect of CPAP on ESS increased with baseline OSA severity, as judged by baseline AHI, and in sleepier patients, as judged by baseline ESS.

Longer treatment durations were associated with decreasing treatment effects.

4,894 patients from 38 RCTs (1997 to 2012):

  • Mild-to-severe OSA (AHI or DI: NR)
  • (for the 6,757 patients included in the SR) Overweight to obese (mean BMI range: 28.3 to 35.1 kg/m2)
  • 3 to 156 weeks of study duration
  • ESS MD (95% CI) = −2.23 (-2.76 to -1.71); P < 0.001; I2 = 83%

Subgroup analysis:

  • Baseline AHI:a
    • Mild: ESS MD (95% CI) = -1.23 (-2.19 to -0.27); P = 0.012; I2 = 59%
    • Moderate: ESS MD (95% CI) = -1.82 (-2.73 to -0.92); P < 0.001; I2 = 60%
    • Severe: ESS MD (95% CI) = -2.64 (-3.44 to -1.84); P < 0.001; I2 = 86%
  • Baseline ESS:b
    • Normal/mild: ESS MD (95% CI) = -0.83 (-1.16 to -0.51); P < 0.001; I2 = 30%
    • Moderate: ESS MD (95% CI) = -2.19 (-2.84 to -1.53); P < 0.001; I2 = 76%
    • Severe: ESS MD (95% CI) = -4.99 (-6.51 to -3.47); P < 0.001; I2 = 46%
  • Treatment duration:
    • 2 to 4 weeks: ESS MD (95% CI) = -2.58 (-3.66 to -1.51); P <0.001; I2 = 75%
    • 5 to 12 weeks: ESS MD (95% CI) = -2.20 (-3.02 to -1.39); P < 0.001; I2 = 68%
    • > 12 weeks: ESS MD (95% CI) = -1.87 (-2.83 to -0.90); P < 0.001; I2 = 93%

Moderate

Bratton 201562

CPAP versus inactive controls

Overall, CPAP was associated with significant reductions in ESS, when compared with inactive controls.

The effect of CPAP was greater in patients with severe OSA.

The effect of CPAP was not different with greater CPAP adherence, longer follow-up, or aging.

Pairwise MA: 6,142 patients from 54 RCTs (1997 to 2015):

  • Mild-to-severe OSA (mean AHI range: 10 to 65 events/hour)
  • Overweight to obese (mean BMI range: 25 to 43 kg/m2)
  • Other comorbidities (e.g., Alzheimer disease, CBD, CVD, HF, hypertension, and resistant hypertension)
  • 1 to 157 weeks of study duration

Pairwise MA:

  • ESS MD (95% CI) = −2.4 (-2.8 to -2.0); P < 0.0001; I2 = 74%

Meta-regression:

  • The effect of CPAP on ESS was greater in sleepier patients (i.e., -0.2 (95% CI = -0.4 to -0.1) for each unit increase in baseline ESS, P = 0.003).
  • There was a trend that the effect of CPAP on ESS was greater in patients with increasing OSA severity (i.e., -0.0 (95% CI = -0.1 to 0.0) for each event/hour increase in baseline AHI, P = 0.05 or -0.1 (95% CI = -0.1 to -0.0) for each dip/hour increase in baseline ODI, P = 0.02).
  • There was no evidence of larger treatment effect of CPAP on ESS with greater CPAP adherence (P = 0.70), longer follow-up (P = 0.50), or aging (P = 0.80).

Moderate

Network MA: 6,873 patients from 67 RCTs (1997 to 2015):

  • Mild-to-severe OSA (mean AHI range: 10 to 65 events/hour)
  • Overweight to obese (mean BMI range: 25 to 43 kg/m2)
  • Other comorbidities (e.g., Alzheimer disease, CBD, CVD, HF, hypertension, and resistant hypertension)
  • 1 to 157 weeks of study duration

Network MA:

  • ESS MD (95% CI) = −2.5 (-2.9 to -2.0); P < 0.0001

Wang 201570

CPAP pre versus post

CPAP was associated with a significant reduction in ESS, when compared with pre-treatment.

131 patients from 2 RCTs (2004, 2014):

  • Mild-to-moderate OSA (mean AHI range: 11.4 to 14.4 events/hour)
  • Nocturia (100%)
  • 1 to 12 months of study duration
  • ESS MD (95% CI) = −5.88 (-6.56 to -5.21); P < 0.00001; I2 = 99%

None

High

Balk 20115

CPAP versus inactive controls

CPAP was associated with a significant reduction in ESS, when compared with inactive controls and when compared with sham CPAP.

1,067 patients from 12 RCTs (1997 to 2010):

  • Mild-to-severe OSA (mean AHI range: 10 to 58 events/hour)
  • Overweight to obese (mean BMI range: 27.3 to 37 kg/m2)
  • 1 to 12 months of study duration
  • ESS MD (95% CI) = -2.37 (-3.23 to -1.51); P < 0.001; I2 = 66%

None

Low to moderate

CPAP versus sham CPAP

904 patients from 16 RCTs (1999 to 2010):

  • Moderate-to-severe OSA (mean AHI range: 22 to 65 events/hour, where reported)
  • Overweight to obese (mean BMI range: 27.3 to 37.8 kg/m2)
  • 1 week to 3 months of study duration
  • ESS MD (95% CI) = −2.50 (-3.54 to -1.45); P < 0.001; I2 = 80.1%

None

Mixed

AHI = Apnea-Hypopnea Index; BMI = body mass index; CBD = cerebrovascular disease; CI = confidence interval; CPAP = continuous positive airway pressure; CVD = cardiovascular disease; DI = desaturation index; ESS = Epworth Sleepiness Scale; HF = heart failure; MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SR = systematic review.
a Baseline AHI: mild, 5 to 14; moderate, 15 to 30; severe, > 30 events/hour.
b Baseline ESS: normal/mild, 0 to 9; moderate, 10 to 15; severe, 16 to 24.

Table 8: Summary of Change in ESS From EPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MA Analyses

Riaz 201568

EPAP pre versus post

EPAP was associated with a significant, but moderate, decrease in ESS post-treatment, when compared with pre-treatment.

359 patients from 5 studies (2009 to 2015):

  • Mild-to-severe OSA (mean AHI range: 14.4 to 43.3 events/hour)
  • Overweight to obese (mean BMI range: 29.4 to 34.9 kg/m2, where reported)
  • 1 to 3 months of study duration, where reported
  • ESS MD (95% CI) = −2.61 (-3.29 to -1.94); P < 0.00001, I2 = 0%
  • ESS SMD (95% CI) =
    -0.52 (-0.71 to -0.33); P < 0.00001, I2 = 18%

None

High

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; EPAP = expiratory positive airway pressure; ESS = Epworth Sleepiness Scale; MA = meta-analysis; MD = mean difference; MR = meta-regression; OSA = obstructive sleep apnea; SMD = standardized mean difference.

Table 9: Summary of Change in ESS From OAs Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Iftikhar 201655

MADs versus inactive controls

ESS was significantly decreased with MADs, compared with inactive controls.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • ESS MD (95% CI) = -2.70 (-3.62 to -1.78); P = NR

None

High

Serra-Torres 201658

MADs pre versus post

MADs were associated with a decrease in ESS post-treatment, when compared with pre-treatment.

397 patients from 9 studies (2005 to 2014):

  • Mild-to severe OSA (mean AHI range: 14 to 45.5 events/hour)
  • Overweight (mean BMI range: 25.9 to 29.2 kg/m2)
  • 1 to 24 months of study duration
  • Mean ESS range:
    • Baseline: 7 to 13
    • Follow-up: 4.2 to 8.5
  • No MA

None

Moderate to high

Sharples 201659

MADs versus inactive controls

Overall, ESS was significantly decreased in the MAD group, compared with the inactive control group.

The effect of MADs on ESS increased in sleepier patients, as judged by baseline ESS.

The effect of MADs on ESS may have been stronger in trials of shorter treatment durations.

Baseline AHI did not appear to have a significant effect on ESS by MADs.

485 patients from 9 RCTs (1997 to 2014):

  • Mild-to-severe OSA (AHI or DI: NR)
  • (for the 6,757 patients included in the SR) Overweight to obese (mean BMI range: 28.3 to 35.1 kg/m2)
  • 4 to 26 weeks of study duration, where reported
  • ESS MD (95% CI) = -1.64 (-2.46 to -0.82); P < 0.001; I2 = 48.2%

Subgroup analysis:

  • Baseline AHI:a
    • Mild (from 1 RCT):
      -2.01 (-2.70 to −1.32); P < 0.001; I2 = NA
    • Moderate: -1.38 (−2.48 to -0.27); P = 0.15; I2 = 42.0%
    • Severe: -2.68 (-5.89 to 0.54); P = 0.103; I2 = 73.0%
  • Baseline ESS:b
  • Moderate: -1.36 (−2.07 to -0.64); P < 0.001; I2 = 55%
  • Severe (from 1 RCT): -8.50 (-13.64 to -3.36); P = 0.001; I2 = NA
  • Treatment duration:
  • 2 to 12 weeks: -1.75 (-2.22 to -1.28); P < 0.001; I2 = 0%
  • >12 weeks: -3.26 (−13.15 to 6.63); P = 0.518; I2 = 90%

Moderate

Pairwise MA: 515 patients from 8 RCTs (2002 to 2014):

  • Mild-to-severe OSA (mean AHI range: 14 to 42 events/hour)
  • Overweight to obese (mean BMI range: 27 to 32 kg/m2)
  • Other comorbidities: NR

4 to 13 weeks of study duration

Pairwise MA:

ESS MD (95% CI) = −1.7 (−2.5 to −1.0); P < 0.0001; I2 = 17%

MADs were associated with a significant reduction in ESS, when compared with inactive controls.

   

None

Moderate

Network MA: 6,873 patients from 67 RCTs (1997 to 2015):

  • Mild-to-severe OSA (mean AHI range: 10 to 65 events/hour)
  • Overweight to obese (mean BMI range: 25 to 43 kg/m2)
  • Other comorbidities (e.g., Alzheimer disease, CBD, CVD, HF, hypertension, and resistant hypertension)
  • 1 to 157 weeks of study duration

Network MA:

  • ESS MD (95% CI) = -1.7 (-2.3 to -1.1); P < 0.0001

None

Zhu 201571

OAs versus inactive controls

OAs were associated with a significant reduction in ESS, when compared with inactive controls.

Baseline OSA severity did not appear to have any significant effect on ESS by OAs.

557 patients from 11 RCTs and 1 cohort study (1997 to 2015):

  • (for the 840 patients included in the SR) Moderate-to-severe OSA (mean AHI ± SD: 23.15 ± 8.18 events/hour [for the OA groups] and 22.55 ± 7.59 events/hour [for the control groups])
  • Comorbidities: NR
  • 1 week to 3 months of study duration
  • ESS MD (95% CI) = -1.76 (-2.57 to -0.94); P < 0.0001; I2 = 90%

Subgroup analysis:

  • Baseline AHI:a
    • Mild-to-moderate: ESS MD (95% CI) = -1.85 (-2.94 to −0.77); P =NR; I2 = NR
    • Mild-to-severe: ESS MD (95% CI) = -1.59 (-2.87 to -0.31); P =NR; I2 = NR

Low

Okuno 201474

OAs versus control appliances

There was no significant difference in change in ESS between OAs and control appliances.

106 patients from 3 RCTs (2005 to 2011):

  • Moderate-to-severe OSA (mean AHI range: 22.1 to 39.1 events/hour [for the OA groups] and 20.1 to 32.6 events/hour [for the control groups])
  • Comorbidities: NR
  • 1 to 6 months of study duration
  • ESS MD (95% CI) = -2.26 (-6.82 to 2.31); P = 0.33; I2 = 81%

None

Low to moderate

Balk 20115

MADs versus inactive controls

MADs were associated with a significant reduction in ESS, when compared with inactive controls or sham OAs.

283 patients from 3 RCTs (2004 to 2008):

  • Moderate-to-severe OSA (mean AHI range: 19-34 events/hour)
  • Overweight to obese (mean BMI range: 27.3 to 31.3 kg/m2); exclusion of patients with heart disease and diabetes
  • 4 weeks to 3 months of study duration
  • ESS MD (95% CI) = -1.17 (-1.74 to −0.61); P = NR; I2 = 45.0%

None

Moderate

MADs versus sham OAs

216 patients from 3 RCTs (1997 to 2005):

  • Moderate-to-severe OSA (mean AHI range: 25 to 36 events/hour)
  • Overweight to obese (mean BMI range: 29 to 32 kg/m2)
  • 2 weeks to 6 weeks of study duration
  • ESS MD (95% CI) = -1.95 (-2.93 to −0.97); P = NR; I2 = 0.0%

None

Moderate

MADs versus inactive controls and sham OAs

499 patients from 6 RCTs (1997 to 2008):

  • Moderate-to-severe OSA (mean AHI range: 19 to 36 events/hour)
  • Overweight to obese (mean BMI range: 27.3 to 32 kg/m2)
  • 2 weeks to 3 months of study duration
  • ESS MD (95% CI) = -1.36 (-1.88 to −0.83); P = NR; I2 = 34.7%

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CBD = cerebrovascular disease; CI = confidence interval; CVD = cardiovascular disease; DI = desaturation index; ESS = Epworth Sleepiness Scale; HF = heart failure; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NA = not applicable; NR = not reported; OA = oral appliances; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SR = systematic review.
a Baseline AHI: mild, 5 to 14; moderate, 15 to 30; severe, > 30 events/hour.
b Baseline ESS: normal/mild, 0 to 9; moderate, 10 to 15; severe, 16 to 24.

Table 10: Summary of Change in ESS From Surgery Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Song 201624

GP or GTA pre versus post

GP and GTA can improve OSA outcomes.

20 patients from 2 studies (2015):

  • Moderate OSA (mean AHI range: 19.9 to 27.5 events/hour)
  • Comorbidities: NR
  • Study duration: NR
  • ESS MD:
    • GP (1 study): -5.8, P < 0.0001
    • GTA (1 study): -2.9, P < 0.0001

None

Low

Zaghi 201660

MMA ± GTA pre versus post

MMA with or without GTA was associated with a significant improvement in EDS, when pre- and post-treatment were compared.

113 patients from an unknown number of studies:

  • (for 455 of the 518 patients included in the SR) Severe OSA (mean AHI: 57.2 events/hour)
  • (for 82 of the 518 patients included in the SR) Obese (mean BMI: 33.8 kg/m2)
  • (for the 45 studies included in the SR) 2 to 6 months of study duration
  • Mean ESS scorea ±SD:
    • Baseline: 13.5 ± 5.2
    • Follow-up: 3.2 ± 3.2
    • P < 0.001
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; EDS = excessive daytime sleepiness; ESS = Epworth Sleepiness Scale; GP = genioplasty; GTA = genial tubercle advancement; MA = meta-analysis; MMA = maxillomandibular advancement; MR = meta-regression; OSA = obstructive sleep apnea; NR = not reported; SD = standard deviation; SR = systematic review.
a ESS scores range from 0 to 24, with lower scores indicating improvement.

Table 11: Summary of Change in ESS From Lifestyle Interventions Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Aiello 201650

Exercise programs versus inactive controls

Exercise programs were associated with having a greater decrease in ESS, compared with inactive controls.

120 patients from 4 RCTs (2009 to 2013):

  • Moderate-to-severe OSA (mean AHI range: 15.4 to 32.2 events/hour)
  • Overweight to obese (mean BMI range: 25.9 to 35.5 kg/m2)
  • 2 to 6 months of study duration (for the 180 patients included in the SR)
  • ESS MD (95% CI) = -1.25 (-2.40 to -0.10); P = NR; I2 = 0%

None

Mixed

Iftikhar 201655

Exercise programs versus inactive controls

ESS was significantly decreased with diet and exercise programs, compared with inactive controls.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • ESS MD (95% CI) = -3.08 (-5.48 to -0.68); P = NR

None

High

Diet programs versus inactive controls

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • Study duration: NR
  • ESS MD (95% CI) = -2.12 (-4.22 to -0.01); P = NR

None

High

Mitchell 201473

Weight-loss programs versus inactive controls

The effect of intensive weight-loss programs on ESS was inconclusive, when compared with inactive controls.

135 patients from 2 RCTs (2009):

  • Moderate-to-severe OSA (AHI threshold: ≥ 15 events/hour) or documented OSA
  • Overweight to obese (BMI range: 28 to 40 kg/m2)
  • 9 weeks to 12 months of study duration
  • Intensive intervention versus conservative intervention as control: change in ESS ± SD =
    -3.1 ± 4 versus -2.1 ± 2.9
  • Intensive intervention versus usual diet as control: change in ESS ± SD = -3 ± 5 versus 1 ± 3
  • No MA

None

High (for the RCTs included in the MA)

Araghi 201376

Weight-loss programs versus inactive controls or pre versus post

Weight-loss programs were associated with an improvement in ESS, when pre- and post-treatment were compared but not when compared with inactive controls.

142 patients from 3 RCTs (2009 to 2011) and 108 patients from 3 before-and-after studies (2009 to 2011):

  • Mild-to-severe OSA (mean AHI range: 10.0 to 43.0 events/hour)
  • Overweight to obese (mean BMI range: 29.7 to 40.0 kg/m2)
  • 1 to 2 years of study duration, where reported
  • Overall effect: ESS MD (95% CI) = -2.01 (-3.22 to -0.79); P = 0.001; I2 = 50%
  • RCTs only: ESS MD (95% CI) = -1.04 (-2.31 to 0.23); P = 0.11; I2 = 6%
  • Before-and-after studies only: ESS MD (95% CI) = -2.87 (-4.30 to -1.44); P < 0.0001; I2 = 37%

None

Mixed

Thomasouli 201319

Weight-loss programs versus inactive controls

There was no significant difference in change in ESS between intensive weight-loss programs and usual care.

114 patients from 2 RCTs (2009 and 2011):

  • Mild-to-moderate OSA (mean AHI range: 9.7 to 28.3 events/hour)
  • Obese (mean BMI range: 33.4 to 34.8 kg/m2)
  • 3 to 12 months of study duration
  • ESS MD (95% CI) = -0.31 (-2.03 to 1.40); P = 0.22; I2 = 33.5%

None

Mixed

Balk 20115

Weight-loss programs versus inactive controls

The effect of intensive weight-loss programs on ESS was inconclusive, when compared with inactive control.

144 patients from 2 RCTs (2009):

  • Moderate-to-severe OSA (mean AHI range: 9 to 37 events/hour)
  • Obese (mean BMI range: 31.4 to 34.8 kg/m2)
  • 9 weeks to 1 year of study duration
  • Intensive intervention versus conservative intervention as control: change in ESS (95% CI) = -1 (-2.7 to 0.7); P = 0.25
  • Intensive intervention versus usual diet as control: change in ESS (95% CI) = -4 (-6 to -2); P < 0.001
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; ESS = Epworth Sleepiness Scale; MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SD = standard deviation; SR = systematic review.

Table 12: Summary of Change in ESS/SSS From CPAP Versus OAs

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP versus OAs

There was no significant difference between CPAP and OAs in their effects on ESS.

139 (included) or 132 (analyzed) patients from 2 RCTs (2004 and 2013):

  • Moderate OSA (mean AHI range: 21.3 to 26.2 events/hour)
  • Overweight to obese (mean BMI range: 27.8 to 31.1 kg/m2)
  • 60 days to 3 months of study duration
  • ESS/SSS Hedges' g (95% CI) = -0.06 (-0.24 to 0.12); P = 0.53; I2 = 0%

None

Very low to low

Iftikhar 201655

CPAP versus MADs

There was no significant difference between CPAP and MADs in their effects on ESS.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • ESS MD (95% CI) = 0.26 (-0.64 to 1.18); P = NR

None

High

Sharples 201659

CPAP versus MADs

Any treatment effect of CPAP over MADs on ESS was small, with clinically significant differences likely only for those with severe baseline AHI.

No significant effects of baseline ESS or study durations on ESS were identified between CPAP and MADs.

675 (included) or 664 (analyzed) patients from 10 RCTs (1997 to 2013):

  • Moderate-to-severe OSA (mean AHI/DI range: NR)
  • (for the 6,757 patients included in the SR) Overweight to obese (mean BMI range: 28.3 to 35.1 kg/m2)
  • 4 to 26 weeks of study duration
  • ESS MD (95% CI) = -0.67 (-1.44 to 0.11); P = 0.093; I2 = 45.2%

Subgroup analysis:

  • Baseline AHI:a
    • Moderate: ESS MD (95% CI) = -0.06 (−0.72 to 0.61); P = 0.86; I2 = 0%
    • Severe: ESS MD (95% CI) = -1.42 (-3.08 to 0.24); P = 0.09; I2 = 68%
  • Baseline ESS:b
    • Moderate: ESS MD (95% CI) = -0.81 (−1.65 to 0.04); P = 0.06; I2 = 49%
  • Study duration:
    • 2 to 12 weeks: ESS MD (95% CI) = -0.82 (-1.73 to 0.09); P = 0.078; I2 = 55%
    • >12 weeks: ESS MD (95% CI) = 0.06 (-1.54 to 1.66); P = 0.94; I2 = 0%

Moderate

Bratton 201562

CPAP versus MADs

CPAP was estimated to further reduce ESS, when compared with MADs.

Pairwise MA: 704 patients from 11 RCTs (1997 to 2014):

  • Moderate-to-severe OSA (mean AHI range: 21 to 42 events/hour)
  • Overweight to obese (mean BMI range: 27 to 33 kg/m2)
  • Other comorbidities: NR
  • 4 to 26 weeks of study duration

Pairwise MA:

  • ESS MD (95% CI) = −0.9 (−1.8 to 0.0); P = 0.06; I2 = 67%

None

Moderate

Network MA: 6,873 patients from 67 RCTs (1997 to 2015):

  • Mild-to-severe OSA (mean AHI range: 10 to 65 events/hour)
  • Overweight to obese (mean BMI range: 25 to 43 kg/m2)
  • Other comorbidities (e.g., Alzheimer disease, CBD, CVD, HF, hypertension, and resistant hypertension)
  • 1 to 157 weeks of study duration

Network MA:

  • ESS MD (95% CI) = −0.8 (−1.4 to −0.1); P = 0.015

None

Okuno 201474

CPAP versus OAs

There was no significant difference between CPAP and OAs in their effects on ESS.

278 (included) or 205 (analyzed) patients from 3 RCTs (2007 to 2011):

  • Moderate-to-severe OSA (mean AHI range: 20.9 to 40.3 events/hour)
  • Comorbidities: NR
  • 8 weeks to 6 months of study duration
  • ESS MD (95% CI) = −1.28 (−2.74 to 0.18); P = 0.09; I2 = 0%

None

Low to moderate

Li 201377

CPAP versus OAs

There was no significant difference between CPAP and OAs in their effects on ESS.

460 patients from 8 RCTs (1997 to 2009):

  • Mild-to-severe OSA (AHI threshold: ≥ 5 events/hour)
  • Comorbidities: NR
  • 8 to 16 weeks of study duration

Crossover trials (from 5 RCTs):

  • ESS MD (95% CI) = −0.74 (−2.17 to 0.69); P = 0.31; I2 = 88%

Parallel-group trials (from 3 RCTs):

  • ESS MD (95% CI) = −1.33 (−2.85 to 0.19); P = 0.09; I2 = 0%

None

Low

Balk 20115

CPAP versus MADs

There was no significant difference between CPAP and MADs in their effects on ESS.

361 patients from 7 RCTs (2002 to 2009):

  • Moderate-to-severe OSA (mean AHI range: 21 to 40 events/hour)
  • Overweight to obese (mean BMI range: 26.7 to 34.1 kg/m2)
  • 1 to 3 months of study duration
  • ESS MD (95% CI) = −1.27 (−2.77 to 0.23); P = 0.098; I2 = 89.4%

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CBD = cerebrovascular disease; CI = confidence interval; CPAP = continuous positive airway pressure; CVD = cardiovascular disease; DI = desaturation index; ESS = Epworth Sleepiness Scale; HF = heart failure; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NR = not reported; OA = oral appliances; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SSS = Stanford Sleepiness Scale.
a Baseline AHI: mild, 5 to 14; moderate, 15 to 30; severe, > 30 events/hour.
b Baseline ESS: normal/mild, 0 to 9; moderate, 10 to 15: severe, 16 to 24.

Table 13: Summary of Change in ESS From CPAP Versus Lifestyle Interventions

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP versus exercise programs

CPAP was moderately more effective than exercise at reducing ESS.

16 patients from 1 RCT (2013):

  • Moderate OSA (mean AHI: 26.2 events/hour)
  • Overweight (mean BMI: 27.8 kg/m2)
  • 60 days of study duration
  • ESS Hedges' g (SE) = -0.71 (0.49)

None

Very low to low

Iftikhar 201655

CPAP versus exercise programs

There was no significant difference between CPAP and diet or exercise in reducing ESS.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • ESS MD (95% CI) = 0.64 (-1.79 to 3.08); P = NR

None

High

CPAP versus diet programs

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • ESS MD (95% CI) = -0.31 (-2.48 to 1.85); P = NR

None

High

Balk 20115

CPAP versus positional therapy (i.e., shoulder-head elevation pillows or devices worn on the back)

There was no significant difference between CPAP and positional therapy in reducing ESS.

94 patients from 3 RCTs (1999 to 2008):

  • Moderate OSA (mean AHI range: 18 to 27 events/hour)
  • Obese (mean BMI range: 30 to 34 kg/m2)
  • 2 weeks to 1 month of study duration
  • ESS MD: no significant differences between CPAP and positional therapy
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; ESS = Epworth Sleepiness Scale; MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SE = standard error; SR = systematic review.

Table 14: Summary of Change in ESS From OAs Versus Lifestyle Interventions

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Iftikhar 201655

MADs versus exercise programs

There was no significant difference between MADs and diet or exercise in reducing ESS.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • ESS MD (95% CI) = 0.37 (-2.15 to 2.90); P = NR

None

High

MADs versus diet programs

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • ESS MD (95% CI) = -0.58 (-2.88 to 1.71); P = NR

None

High

AHI = Apnea-Hypopnea Index; CI = confidence interval; ESS = Epworth Sleepiness Scale; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NR = not reported; OA = oral appliances; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SR = systematic review.

Table 15: Summary of Change in ESS From Diet Versus Exercise

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Iftikhar 201655

Diet programs versus exercise programs

There was no significant difference between diet and exercise in reducing ESS.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 64 RCTs provided ESS:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • ESS MD (95% CI) = 0.96 (-2.23 to 4.15); P = NR

None

High

AHI = Apnea-Hypopnea Index; CI = confidence interval; ESS = Epworth Sleepiness Scale; MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SR = systematic review.

Table 16: Summary of Change in ESS From Combination Therapy Versus Lifestyle Interventions

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Thomasouli 201319

CPAP plus diet programs versus diet programs alone

CPAP plus diet programs was associated with a significant reduction in ESS, compared with diet programs alone.

230 patients from 2 RCTs (1999 and 2001):

  • Moderate-to-severe OSA (mean AHI range: 20 to 56 events/hour)
  • Overweight to obese (mean BMI range: 29 to 32 kg/m2)
  • 3 to 6 months of study duration
  • ESS MD (95% CI) = −3.20 (-6.62 to -0.23); P = NR; I2 = 82.7%

None

Mixed

AHI = Apnea-Hypopnea Index; BMI = body mass index; CPAP = continuous positive airway pressure; ESS = Epworth Sleepiness Scale; MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; OSA = obstructive sleep apnea; RCT = randomized controlled trial.

Table 17: Summary of Change in ESS From TRDs Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Roplekar 201592

TRDs pre versus post

TRDs were associated with a reduction in ESS, although no statistical testing was conducted.

20 patients:

  • Patient characteristics: NR
  • Mean ESS (n = 11):
    • Before: 11.9 (range 3 to 24)
    • After 4 months: 7.1 (range 2 to 13)

None

Lazard 2009113

TRDs pre versus post

TRDs were associated with a significant reduction in ESS.

84 patients:

  • Mean AHI ± SD: 37 ± 19.5 events/hour
  • Mean BMI ± SD: 26 ± 3.8 kg/m2
  • Mean ESS ± SD (n = 24):
    • Before: 9 ± 5.0
    • After: 6 ± 3.7
    • Difference: P < 0.05

None

Dort 2008115

TRDs pre versus post

TRDs with or without suction were both associated with significant reductions in ESS. There was no significant difference between TRDs with suction and TRDs without suction.

38 patients:

  • Mean RDI ± SD: 15.5 ± 17.7 events/hour
  • Mean BMI ± SD: 29.4 ± 5.7 kg/m2
  • Mean ESS ± SD (n = 32):
    • Before: 12.4 ± 4.5
    • After:
      • No suction: 10.3 ± 4.9
      • Suction: 10.9 ± 4.4
    • Difference (before versus after):
      • No suction: P = 0.017
      • Suction: P = 0.025
    • Difference (no suction versus suction): -0.65 (95% CI = -1.8 to 0.47), P = 0.25

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; ESS = Epworth Sleepiness Scale; NR = not reported; OSA = obstructive sleep apnea; RDI = respiratory disturbance index; SD = standard deviation; TRD = tongue-retaining device.

Table 18: Summary of Change in ESS From GTA Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Kuscu 201591

GTA pre versus post

GTA resulted in significant reductions in ESS.

17 patients:

  • Mean AHI ± SD: 27.5 ± 8 events/hour
  • Mean BMI ± SD: 30.2 ± 4 kg/m2
  • Mean ESS ± SD:
    • Before: 7.7 ± 1.6 (range: 5-9)
    • After: 4.8 ± 1.9 (range: 2-8)
    • Difference: P = 0.001

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; ESS = Epworth Sleepiness Scale; GTA = genial tubercle advancement; OSA = obstructive sleep apnea; SD = standard deviation

Table 19: Summary of Change in ESS From Positional Therapy Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Benoist 201678

Positional therapy (i.e., sleep position trainers) pre versus post

Positional therapy with a sleep position trainer significantly improved ESS scores in patients with positional OSA.

33 positional OSA patients:

  • Median AHI: 18.3 (IQR: 13.7 to 24.0) events/hour
  • Mean BMI ± SD: 27.9 ± 2.8 kg/m2
  • Median ESS:
    • Before (n = 33): 10.0 (IQR: 5.5 to 15.0)
    • After 3 months (n = 32): 7.0 (IQR: 5.0 to 12.0)
    • Difference: P = 0.029

None

Afrashi 201582

Positional therapy (i.e., pillows for prone positioning) pre versus post

Pillows for prone positioning saw no significant change in ESS.

29 patients:

  • Mean AHI ± SD: 15.5 ± 6.2 events/hour
  • Mean BMI ± SD: 28.9 ± 3.2 kg/m2
  • Mean ESS ± SD:
    • Before: 7.3 ± 5.1 (range: 1-18)
    • After 2 nights:
      • Respondentsa (n = 15): 6 (range: 3 to 12)
      • Non-respondentsa (n = 14): 6 (range: 2 to 11.25)
    • Difference (respondents versus non-respondents): P = 0.58

None

Bidarian-Moniri 201583

Positional therapy (i.e., mattresses and pillows for prone positioning) pre versus post

Positional therapy with mattresses and pillows for prone positioning significantly improved ESS scores.

14 patients:

  • Mean AHI: 26 events/hour (range: 6 to 53 events/hour)
  • Mean ODI: 21 events/hour (range: 5-51 events/hour)
  • Mean BMI: 26 kg/m2
  • Mean ESS (n = 10):
    • Before: 12
    • After 4 weeks: 9
    • Difference: P = 0.007

None

Eijsvogel 201588

Positional therapy (i.e., tennis balls or sleep position trainers) pre versus post

No significant change in ESS was found between baseline and 1 month after positional therapy.

26 (TBT) or 29 (SPT) positional OSA patients:

  • Mean AHI ± SD: 13.1 ± 9.1 (TBT) or 11.4 ± 4.9 (SPT) events/hour
  • Mean BMI ± SD: 26.8 ± 3.0 (TBT) or 27.6 ± 4.5 (SPT) kg/m2
  • Mean ESS ± SD:
    • TBT patients:
      • Before (n = 26): 7.3 ± 4.2
      • After 1 month (n = 21): 7.8 ± 4.3
    • SPT patients:
      • Before (n = 29): 6.4 ± 3.4
      • After 1 month (n = 27): 6.0 ± 3.6
    • Difference: P = non-significant

None

de Vries 201586

Positional therapy (i.e., commercial devices or self-made constructions) pre versus post

Short-term positional therapy was associated with a significant reduction in ESS and may be an easy and effective method in patients with positional OSA. The commercial device and self-made construction had similar effects.

40 positional OSA patients:

  • Median AHI: 14.5 events/hour (IQR: 10.7 to 19.6 events/hour)
  • Mean BMI ± SD: 28.0 ± 4.1 kg/m2
  • Mean ESS ± SD:
    • Before: 12.2 ± 5.4
    • After: 10.2 ± 5.5
    • Difference: P < 0.01

None

Jackson 201590

Positional therapy (i.e., sleep position modification devices) versus inactive controls

A significant reduction in ESS was found in the active group but not in the control group.

47 (in the intervention group) or 39 (in the control group) positional OSA patients:

  • Mean AHI ± SD: 20.1 ± 8.8 (in the intervention group) or 21.8 ± 10.1 (in the control group) events/hour
  • Mean BMI ± SD: 30.0 ± 5.3 (in the intervention group) or 30.9 ± 7.7 (in the control group) kg/m2
  • Mean ESS ± SD (intervention versus control):
    • Before: 9.9 ± 4.7 versus 10.0 ± 5.9
    • After 4 weeks: 8.1 ± 4.1 versus 9.4 ± 6.6
    • Difference: P < 0.005 versus P = non-significant

None

Levendowski 201496

Positional therapy (i.e., neck position devices) pre versus post

Patients showed non-significant improvements in ESS.

30 positional OSA patients:

  • Mean AHI ± SD: 24.7 ± 14.7 events/hour
  • Mean BMI ± SD: 28 ± 3.4 kg/m2
  • Mean ESS ± SD:
    • Before: 11.3 ± 4.6
    • After 30 nights: 9.5 ± 4.6
    • Difference: P = 0.064

None

van Maneen 201497

Positional therapy (i.e., sleep position trainers) pre versus post

There was a significant decrease in median ESS.

145 positional OSA patients:

  • Median AHI: 11.5 events/hour (IQR: 2.5 to 20.5 events/hour)
  • Median BMI: 27.0 kg/m2 (IQR: 23.0 to 31.0 kg/m2)
  • Median ESS ± IQR (n = 53):
    • Before: 11 ± 6
    • After 1, 3, and 6 months: 9 ± 8, 8 ± 8, 7 ± 6
    • Difference after 1, 3, and 6 months: P < 0.001

None

van Maanen 201399

Positional therapy (i.e., sleep position trainers) pre versus post

Sleep position trainers significantly decreased median ESS.

31 positional OSA patients:

  • Median AHI: 16.4 events/hour (IQR: 6.6 to 29.9 events/hour)
  • Mean BMI ± SD: 27.0 ± 3.7 kg/m2
  • Median ESS:
    • Before: 11 (IQR: 2 to 20)
    • After: 9 (IQR: 0 to 19)
    • Difference: P = 0.004

None

Heinzer 2012102

Positional therapy (i.e., tennis balls) pre versus post

There was a significant decrease in sleepiness after three months of device use.

16 positional OSA patients:

  • Mean AHI ± SD: 26.7 ± 17.5 events/hour
  • Mean BMI ± SD: 25.4 ± 4.1 kg/m2
  • Mean ESS ± SD:
    • Before: 9.4 ± 4.5
    • After 3 months: 6.6 ± 4.7
    • Difference: P = 0.02

None

Loord 2007116

Positional therapy (i.e., the Positioner) pre versus post

The mean ESS score decreased significantly during treatment.

18 positional OSA patients:

  • Mean AHI ± SD: 21.8 ± 12.0 events/hour
  • Mean BMI: NR
  • Mean ESS ± SD:
    • Before: 11.8 ± 4.2
    • After 29 ± 2 nights: 10.2 ± 5.2
    • Difference: P = 0.02

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; ESS = Epworth Sleepiness Scale; IQR = interquartile range; NR = not reported; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; SD = standard deviation; SPT = sleep position trainer; TBT = tennis ball technique.
a Respondents and non-respondents to treatment.

Table 20: Summary of Change in ESS From Combination Therapy Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Islam 201495

MMA plus GTA pre versus post

Maxillomandibular and mandibular advancement surgery had a significant impact on decreasing ESS.

51 patients:

  • Mean AHI ± SD: 42 ± 17 events/hour
  • Mean BMI: 29 ± 3.4 kg/m2
  • Mean ESS ± SD:
    • Before: 14 ± 4
    • After: 5 ± 4
    • Difference: P < 0.001

None

Dort 2012101

MADs plus TRDs pre versus post

The combination of mandibular advancement and tongue retention was an effective treatment for moderate-to-severe OSA by significantly decreasing ESS scores post-treatment.

41 patients:

  • Mean RDI ± SD: 33.5 ± 15.9 events/hour
  • Mean BMI ± SD: 32.2 ± 5.8 kg/m2
  • Mean ESS ± SD:
    • Before: 12.3 ± 5.4
    • After 20 weeks:
      • 6 mm mandibular advancement: 9.2 ± 4.3
      • 8 mm mandibular advancement: 9.0 ± 4.6
    • Difference:
      • 6 mm mandibular advancement: P = 0.001
      • 8 mm mandibular advancement: P = 0.009

None

Fujii 2010110

CPAP plus weight-loss programs pre versus post

Weight-loss programs had no significant impact on ESS scores.

10 patients:

  • Mean AHI ± SD: 59.0 ± 20.9 events/hour
  • Mean BMI ± SD: 30.7 ± 2.5 kg/m2
  • Mean ESS ± SD:
    • Before: 9.1 ± 4.0
    • After 4 months: 9.2 ± 4.4
    • Difference: P = 0.89

None

McDoniel 2010111

CPAP plus weight-loss programs pre versus post

Patients had improved ESS scores, but it was not significant.

11 patients:

  • Mean AHI ± SD: 64.2 ± 28.2 events/hour
  • Mean BMI ± SD: 41.7 ± 6.8 kg/m2
  • Mean ESS ± SD:
    • Before: 10.7 ± 3.2
    • After 12 weeks: 7.7 ± 4.1
    • Difference: P = 0.09

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; CPAP = continuous positive airway pressure; ESS = Epworth Sleepiness Scale; MAD = mandibular advancement device; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; RDI = respiratory disturbance index; SD = standard deviation

Table 21: Summary of Change in ESS From Combination Therapy Versus Active Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Ackel-D'Elia 2012100

CPAP plus exercise programs versus CPAP alone

CPAP plus exercise and CPAP alone were both effective at significantly reducing ESS.

13 (CPAP plus exercise) or 19 (CPAP) patients:

  • Mean AHI ± SD: 40.5 ± 22.9 (CPAP plus exercise) or 42.3 ± 21.6 (CPAP) events/hour
  • Mean BMI ± SD: 28.0 ± 3.1 (CPAP plus exercise) or 28.5 ± 2.2 (CPAP) kg/m2
  • Mean ESS ± SD:
    • Before:
      • CPAP plus exercise: 14.0 ± 4.1
      • CPAP:13.0 ± 4.8
    • After 2 months:
      • (data presented as a graph) CPAP plus exercise more effective than CPAP alone
    • Difference:
      • CPAP plus exercise: P < 0.05
      • CPAP: P < 0.05

None

El-Solh 2011106

CPAP plus MADs versus MADs alone

The combination therapy of CPAP plus MADs and MADs alone were both effective in significantly reducing ESS.

10 patients:

  • Mean AHI ± SD: 23.5 ± 13.4 events/hour
  • Mean BMI ± SD: 26.9 ± 3.2 kg/m2
  • Mean ESS ± SD:
    • Before: 12.7 ± 2.1
    • After 8 weeks:
      • CPAP plus MADs: 7.5 ± 4.1
      • MADs: 9.7 ± 3.1
    • Difference:
      • CPAP plus MADs: P = 0.007
      • MADs: P = 0.04

None

Johansson 2011107

CPAP plus diet programs versus CPAP alone

After CPAP with a weight-loss program and low-energy diet, patients had a significant decrease in ESS.

63 patients:

  • Mean AHI ± SD: 36 ± 15 events/hour
  • BMI range: 30 to 40 kg/m2
  • Mean ESS ± SD:
    • Before: 8 ± 5
    • After 1 year of study duration:
      • CPAP plus very-low-energy diet (0 to 9 weeks): 5 ± 4
      • After full program (0 to 52 weeks): 6 ± 5
    • Difference (MD ± SD, 95% CI):
      • CPAP plus very-low-energy diet (0 to 9 weeks): -2 ± 4, -1 to -3, P < 0.001
      • After full program (0 to52 weeks): −2 ± 3, -1 to -3, P < 0.001

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; ESS = Epworth sleepiness score; CPAP = continuous positive airway pressure; MAD = mandibular advancement device therapy; MD = mean difference; OSA = obstructive sleep apnea; SD = standard deviation.


OSA Severity

Information on the validity and reliability of measurement of OSA severity by AHI can be found in Appendix 15.

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Five SRs,5,54,55,59,70 including one NMA,55 reported on OSA severity in adults with mild-to-severe55,59 or moderate-to-severe5,54,70 OSA. Three SRs5,54,59 included overweight-to-obese patients. One SR70 included patients with nocturia only. One SR55 provided no information on comorbidities.

All five SRs,5,54,55,59,70 with sample sizes ranging from 84 patients54 to 7,882 patients55 from two studies54,70 to 80 studies,55 reported significantly greater reductions in AHI,5,54,55,59,70 ODI,55 or RDI,54 with CPAP, compared with controls5,54,55,59 or pre-treatment.54,70 The mean difference in AHI, reported by four SRs,5,55,59,70 ranged from -46.39 events/hour5 to -19.85 events/hour,5 with three SRs5,55,59 reporting mean differences around -20 events/hour. Study duration of the included primary studies, reported by all five SRs, ranged from one week5 to 156 weeks.59 I2 scores, reported by the four applicable SRs,5,54,59,70 ranged from 0%54 to 98%70 and were greater than 75% in three SRs.5,59,70 The SRs reported the quality of the included studies as very low to low,54 low to moderate,5 moderate,59,70 or high55 (Appendix 10).

From subgroup analyses, one SR59 reported that the effect of CPAP versus controls on AHI increased with increasing EDS and OSA severity at baseline but decreased with longer treatment durations.

Across the five SRs, 66 primary studies had been included, 42 of which had been included in one SR, 15 in two SRs, seven in three SRs, and two in four SRs (Appendix 16.6). No two SRs completely overlapped on AHI as the outcome.

The findings of the SRs are summarized in Table 22.

2) Expiratory positive airway pressure versus inactive controls

One SR68 reported on OSA severity in adults with moderate-to-severe OSA who were overweight to obese. The SR,68 with a sample size of 345 patients from 10 studies, reported significantly greater reductions in AHI and ODI with EPAP, compared with pre-treatment. The mean difference in AHI was -14.78 events/hour. Study duration of the included primary studies ranged from one night to two months. I2 scores ranged from 72% to 80%. The SR68 reported the quality of the included studies as high (Appendix 10). The findings of the SR are summarized in Table 23.

3) Oral appliances versus inactive controls

Six SRs,5,55,58,59,71,74 including one NMA,55 reported on OSA severity in adults with mild-to-severe55,58,59 or moderate-to-severe5,71,74 OSA. Three SRs5,58,59 included overweight-to-obese patients. Three SRs55,71,74 provided no information on comorbidities.

All six SRs,5,55,58,59,71,74 with sample sizes ranging from 106 patients5,74 to 7,882 patients55 from three studies74 to 80 studies,55 reported significantly greater reductions in AHI5,55,58,59,71,74 or ODI55 with MADs5,55,58,59 or undefined OAs,71,74 compared with controls5,55,59,71,74 or pre-treatment.58 The mean difference in AHI, reported by all six SRs, ranged from -15.20 events/hour55 to -7.05 events/hour,74 with four SRs5,58,59,71 reporting mean differences around -10 events/hour. Study duration of the included primary studies, reported by all six SRs, ranged from one week5,74 to 84 months.58 I2 scores, reported by the four applicable SRs,5,59,71,74 ranged from 0%5 to 90%71 and were greater than 75% in one SR.71 The SRs reported the quality of the included studies as low,71 low to moderate,74 moderate,5,59 moderate to high,58 or high55 (Appendix 10).

From subgroup analyses, one SR59 reported that the effect of MADs versus controls on AHI was greater with moderate versus severe ESS at baseline59 and with shorter treatment durations.59 However, two SRs59,71 reported no significant differences in the effect of MADs59 or undefined OAs71 versus controls on AHI with varying levels of OSA severity at baseline.

Across the six SRs, 37 primary studies had been included, 25 of which had been included in one SR, four in two SRs, three in three SRs, four in four SRs, and one in five SRs (Appendix 16.7). Three SRs55,59,71 included all primary studies included in another SR74 on AHI as the outcome.

The findings of the SRs are summarized in Table 24.

4) Surgery versus inactive controls

Two SRs24,60 reported on OSA severity in adults with moderate-to-severe24 or severe60 OSA who were obese.60 The SRs,24,60 with sample sizes of 61 patients24 or 455 patients60 from nine studies24 to 36 studies,60 reported greater reductions in AHI24,60 or RDI60 with GP,24 GTA,24 or MMA with or without GTA,60 compared with pre-treatment. The mean difference in AHI was −7.78 events/hour with GP,24 -11.1 events/hour with GTA,24 and -47.8 events/hour with MMA with or without GTA.60 Study duration of the included primary studies, reported by one SR,60 ranged from two months to six months. I2 scores, reported by both SRs,24,60 ranged from 0%24 to 61.3%.60 The SR60 reported the quality of the included studies as low24 or moderate60 (Appendix 10). From subgroup analyses, one SR60 reported that the effect of MMA with or without GTA versus pre-treatment on AHI increased with increasing OSA severity at baseline. The findings of the SRs are summarized in Table 25.

5) Lifestyle interventions versus inactive controls

Seven SRs5,19,50,55,61,73,76 reported on OSA severity in adults with mild-to-severe OSA. Six SRs5,19,50,61,73,76 included overweight-to-obese patients. One SR55 provided no information on comorbidities.

All seven SRs,5,19,50,55,61,73,76 with sample sizes ranging from 184 patients50 to 7,882 patients55 from three studies5 to 80 studies,55 reported significantly greater reductions in AHI5,19,50,55,61,73,76 and ODI55,73,76 with diet5,19,55,61,73,76 or exercise5,19,50,55,61,76 programs, compared with controls5,19,50,55,73,76 or pre-treatment.61,76 The mean difference in AHI, reported by all seven SRs, ranged from -23 events/hour5 to -0.54 events/hour,50 with four SRs55,61,73,76 reporting mean differences between -17 events/hour and -11 events/hour. Study duration of the included primary studies, reported by all seven SRs, ranged from two weeks55 to 94.3 months.61 I2 scores, reported by the five applicable SRs,19,50,61,73,76 ranged from 5%76 to 92.2%73 and were greater than 75% in three SRs.61,73,76 The SRs reported the quality of the included studies as moderate,5 high,55,73 or mixed19,50,61,76 (Appendix 10).

From subgroup analyses, one SR76 reported that the effect of diet or exercise programs on AHI increased with increasing OSA severity at baseline and was greater with shorter treatment durations in before-and-after, but not controlled, studies. The SR76 also reported that the greatest source of heterogeneity appeared to be studies with higher baseline AHI and also those with greater change in BMI.

Across the seven SRs, 39 primary studies had been included, 24 of which had been included in one SR, nine in two SRs, one in three SRs, four in four SRs, and one in five SRs (Appendix 16.8). One SR73 included all primary studies included in another SR5 on AHI as the outcome.

The findings of the SRs are summarized in Table 26.

6) Combination therapy versus inactive controls

One SR5 reported on OSA severity in adults with mild-to-severe OSA, with no information on comorbidities. The SR,5 with a sample size of 60 patients from one study, reported no significant differences in change in AHI between TRDs plus positional therapy and no treatment. Study duration was not reported. I2 scores were not applicable. The SR5 reported the quality of the included study as low (Appendix 10). The findings of the SR are summarized in Table 27.

7) Continuous positive airway pressure versus oral appliances

Six SRs5,54,55,59,74,77 reported on OSA severity in adults with mild-to-severe,55,77 moderate,54 or moderate-to-severe5,59,74 OSA. Three SRs5,54,59 included overweight-to-obese patients. Three SRs55,74,77 provided no information on comorbidities.

All six SRs,5,54,55,59,74,77 with sample sizes ranging from 139 patients54 to 7,882 patients55 from two studies54 to 80 studies,55 reported significantly greater reductions in AHI5,54,55,59,74,77 and ODI55 with CPAP compared with MADs5,55,59 or undefined OAs.54,74,77 The mean difference in AHI, reported by all six SRs, ranged from -10.06 events/hour55 to -0.9 events/hour,54 with four SRs5,59,74,77 reporting mean differences between -8 events/hour and -5 events/hour. Study duration of the included primary studies, reported by all six SRs,5,54,59,74,77 ranged from two weeks5,55 to 144 weeks.55 I2 scores, reported by the five applicable SRs, ranged from 0%54 to 68%.77 The SRs reported the quality of the included studies as very low to low,54 low,77 low to moderate,5,74 moderate,59 or high55 (Appendix 10).

From subgroup analyses, one SR59 reported no significant differences in the effect of CPAP versus undefined OAs on AHI with varying EDS or OSA severity at baseline or treatment durations.

Across the six SRs, 17 primary studies had been included, five of which had been included in one SR, two in two SRs, three in three SRs, four in four SRs, and three in five SRs (Appendix 16.9). One SR55 included all primary studies included in three other SRs54,74,77 on AHI as the outcome. Another SR59 also included all primary studies included in two other SRs74,77 on AHI as the outcome.

The findings of the SRs are summarized in Table 28.

8) Continuous positive airway pressure versus lifestyle interventions

Four SRs5,54,55,72 reported on OSA severity in adults with mild-to-moderate55,72 or moderate5,54 OSA. Three SRs5,54,72 included overweight54 or obese5,72 patients. One SR55 provided no information on comorbidities.

All four SRs,5,54,55,72 with sample sizes ranging from 16 patients54 and 7,882 patients55 from one study54 to 80 studies,55 reported significantly greater reductions in AHI with CPAP, compared with diet,19,55 exercise,54 or positional therapy.5,72 The mean difference in AHI, reported by all four SRs, ranged from -16 events/hour5 to -2.01 events/hour.54 However, one of the four SRs,55 with a sample size of 7,882 patients from 80 studies, reported no significant differences in the effect of CPAP versus exercise. Study duration of the included primary studies, reported by all four SRs, ranging from three nights72 to 144 weeks.55 The I2 score, reported by the one applicable SR,72 was 70%. The SRs reported the quality of the included studies as very low to low,54 moderate,5,72 or high55 (Appendix 10).

Across the four SRs, five primary studies had been included, two of which had been included in one SR, and three in two SRs (Appendix 16.10). Two SRs54,55 completely overlapped on AHI as the outcome.

The findings of the SRs are summarized in Table 28.

9) Mandibular advancement devices versus tongue-retaining devices

One SR5 reported on OSA severity in adults with moderate OSA who were overweight. The SR,5 with a sample size of 22 patients from one study, reported no significant differences in change in AHI between MADs and TRDs. Study duration of the included primary study was one week. I2 scores were not applicable. The SR5 reported the quality of the included study as moderate (Appendix 10). The findings of the SR are summarized in Table 30.

10) Mandibular advancement devices versus lifestyle interventions

One SR55 reported on OSA severity in adults with mild-to severe OSA, with no information on comorbidities. The SR,55 with a sample size of 7,882 patients from 80 studies, reported no significant differences in change in AHI between MADs and diet or exercise programs. Study duration of the included primary studies ranged from two weeks to 144 weeks. I2 scores were not applicable. The SR55 reported the quality of the included studies as high (Appendix 10). The findings of the SR are summarized in Table 31.

11) Diet versus exercise

One SR55 reported on OSA severity in adults with mild-to severe OSA, with no information on comorbidities. The SR,55 with a sample size of 7,882 patients from 80 studies, reported no significant differences in change in AHI between diet and exercise programs. Study duration of the included primary studies ranged from two weeks to 144 weeks. I2 scores were not applicable. The SR55 reported the quality of the included studies as high (Appendix 10). The findings of the SR are summarized in Table 32.

Review of Primary Studies

1) Tongue-retaining devices versus inactive controls

Two studies113,115 reported on OSA severity in adults with moderate115 or severe113 OSA, providing mean113,115 AHI. The studies included overweight113,115 patients, providing mean113 BMI. Both studies,113,115 with sample sizes ranging from 38 patients115 to 84 patients,113 reported significant reductions in AHI115 or RDI113 with TRDs113 or TRDs with and without suction,115 compared with pre-treatment. The mean difference, reported by both studies, was -24 events/hour in AHI115 or -6.6 events/hour to -2 events/hour in RDI.113 Concerns with the quality of the two studies were assessed to be moderate113 or unclear115 (Appendix 13 and Appendix 14). The findings of the primary studies are summarized in Table 33.

2) Genial tubercle advancement versus inactive controls

Two studies91,117 reported on OSA severity in adults with mild-to-moderate117 OSA, providing AHI ranges, or moderate91 OSA, providing mean AHI. The studies included obese91 patients, providing mean91 AHI, or patients with a BMI under 30 in another study,117 normal-to-overweight117 or overweight-to-obese91 patients. Both studies,91,117 with sample sizes ranging from 10 patients117 to 17 patients,91 reported significant reductions in AHI with GTA, compared with pre-treatment. The mean difference in AHI, reported by both studies, ranged from -10.2 events/hour117 to -7.98 events/hour.91 However, the authors of one study91 cautioned that only individuals with less severe OSA may benefit from this type of surgery. The studies91,117 had been included, and their findings reported, by the SR on the comparison24 that was later identified from a search alert. Concerns with the quality of the two studies were assessed to be low91,117 (Appendix 14). The findings of the primary studies are summarized in Table 34.

3) Positional therapy versus inactive controls

Seventeen studies78,81-86,88,90,96,99,102,104,105,108,116,118 reported on OSA severity in adults with mild,85,86,88 moderate,78,81-83,90,96,99,102,104,105,108,116,118 or severe OSA,84 providing mean81-84,88,90,96,102,104,105,108,116,118 or median AHI.78,85,86,99 Ten studies included patients who were normal weight,85 overweight,78,81-84,86,88,96,99,102,104,105,108,118 or obese,90 providing mean78,81-84,86,88,90,96,102,104,105,108,118 or median BMI.85,99 Twelve studies78,81,85,86,90,96,99,102,104,108,116,118 included positional OSA patients only. Fourteen studies included normal-to-overweight,85,102,108 normal-to-obese,78,81,83,84,90,99,118 or overweight-to-obese78,82,86,88,96,104,105 patients. One study116 provided no information on comorbidities.

Sixteen of the 17 studies,78,81-84,86,88,90,96,99,102,104,105,108,116,118 with sample sizes ranging from 14 patients83 to 86 patients,90 reported significantly greater reductions in AHI,78,81-84,86,88,90,96,99,102,104,105,108,116,118 ODI,78,81 or RDI81 with positional therapy (i.e., with tennis balls,88,90,102,118 pillows,83,84 an apparatus designed to mimic the tennis ball technique,86 or an apparatus designed to avoid sleep in the supine position78,81,82,96,99,104,105,108,116), compared with controls90 or pre-treatment.78,81-84,86,88,96,99,102,104,105,108,116,118 One of the 17 studies,85 with a sample size of 25 patients, also reported a reduction in ODI with positional therapy (i.e., with pillows), compared with pre-treatment, but the reduction was not significant. One of the 17 studies102 reported higher reductions in AHI from tennis balls with shorter duration (i.e., one night) than longer duration (i.e., three months). The mean difference in AHI, reported by all but one study,105 ranged from -29 events/hour118 to -0.7 events/hour,85 with seven SRs82,86,88,90,99,108,116 reporting mean differences between -15 events/hour and -5 events/hour. Concerns with the quality of the 17 studies were assessed to be low78,81-86,88,90,96,99,102,104,105,108 or high116,118 (Appendix 13 and Appendix 14).

From subgroup analyses, two studies85,96 reported that the effect of positional therapy versus pre-treatment on AHI96 or ODI85 was greater with mild or moderate OSA, compared with severe OSA,96 but was indifferent between normal-weight and overweight patients.85 Nine studies78,86,88,96,99,102,104,108,118 conducted subgroup analyses on supine versus non-supine sleep position, seven of which86,88,96,99,102,104,108 reported significant reductions in AHI in the supine sleep position but not in non-supine sleep positions, whereas one reported significant reductions in AHI in both supine and non-supine sleep positions.78 One study84 conducted subgroup analyses on positional versus non-positional OSA and reported reductions in AHI and ODI in patients with positional or non-positional OSA.

The findings of the primary studies are summarized in Table 35.

4) Combination therapy versus inactive controls

Three studies95,101,114 reported on OSA severity in adults with severe95,101 OSA provided with a mean AHI, or with an initial RDI of 35 or greater.114 Two studies included overweight95 or obese101 patients, provided with a mean BMI. One study114 provided no information on comorbidities.

All three studies,95,101,114 with sample sizes ranging from four patients114 to 51 patients,95 reported reductions in AHI95 or RDI114 after MMA plus GP95,114 or a reduction in RDI after MADs plus TRDs,101 compared with pre-treatment, with two studies95,101 reporting statistical significance. The mean difference, reported by all three studies, was -34 events/hour in AHI95 or -25 events/hour114 to -14.3 events/hour101 in RDI. Concerns with the quality of the three studies were assessed to be low95,101 or unclear114 (Appendix 14).

The findings of the primary studies are summarized in Table 36.

5) Mandibular advancement devices versus tongue-retaining devices

One study109 reported on OSA severity in overweight adults with moderate OSA, provided with mean AHI and BMI. The study, with a sample size of 39 patients, reported reductions in AHI with the use of both MADs and TRDs but a greater reduction with TRDs, compared with MADs, with a mean difference of -0.8 events/hour. Concerns with the quality of the study were assessed to be low109 (Appendix 14). The findings of the primary study are summarized in Table 37.

6) Mandibular advancement devices versus maxillomandibular advancement

One study94 reported on OSA severity in adults with moderate-to-severe OSA, with 17.2% of the patients being obese. The study, with a sample size of 198 patients, reported a significantly greater reduction in AHI with MMA, compared with MADs, with a mean difference of -13.59 events/hour. Concerns with the quality of the study were assessed to be low94 (Appendix 14). The findings of the primary study are summarized in Table 38.

7) Combination therapy versus active controls

Four studies87,100,106,107 reported on OSA severity in patients with moderate87,106 or severe,100,107 provided by mean AHI. The studies included overweight patients87,100,106 provided by mean BMI, or obese patients107 provided by a BMI range.

Three of the four studies,87,106,107 with sample sizes ranging from 10 patients106 to 63 patients,107 reported reductions in AHI87,106,107 or ODI87 with CPAP plus MADs106 or CPAP plus diet programs,107 compared with CPAP alone, or with MADs plus positional therapy (i.e., sleep position trainers), compared with MADs or positional therapy alone,87 with two studies106,107 reporting statistical significance. The mean difference in AHI, reported by all three studies, was -21 events/hour107 or -5.5 events/hour,87 with two studies87,106 reporting mean differences between -5 and -7. Concerns with the quality of the three studies were assessed to be low106,107 or unclear87 (Appendix 13 and Appendix 14).

The other study,100 with a sample size of 32 patients, reported similar reductions in AHI for both CPAP plus exercise programs and CPAP alone. Concerns with the quality of the study were assessed to be unclear100 (Appendix 13).

From subgroup analyses, one study107 reported that the effect of CPAP plus diet programs versus CPAP alone on AHI was greater in the supine sleep position, compared with the combination of supine and non-supine positions.

The findings of the primary studies are summarized in Table 39.

Summary of Results on Obstructive Sleep Apnea Severity

For OSA severity, evidence was found on inactive comparisons with CPAP, EPAP, OAs (i.e., MADs, TRDs, and undefined OAs), surgery (MMA and GTA), lifestyle interventions (i.e., diet, exercise, and positional therapy), and combination therapy (i.e., MADs plus TRDs, TRDs plus positional therapy, and MMA plus GTA). Evidence was also found on active comparisons between CPAP and OAs (i.e., MADs and undefined OAs), CPAP and lifestyle interventions (i.e., diet, exercise, and positional therapy), MADs and TRDs, MADs and MMA, MADs and lifestyle interventions (i.e., diet and exercise), diet and exercise, and combination therapy and other interventions (i.e., CPAP plus MADs versus MADs alone, CPAP plus diet programs versus CPAP alone, CPAP plus exercise programs versus CPAP alone, and MADs plus positional therapy versus MADs or positional therapy alone).

Compared with inactive controls or pre-treatment, CPAP, EPAP, MADs, TRDs, undefined OAs, MMA, GTA, diet, exercise, positional therapy, MADs plus TRDs, TRDs plus positional therapy, and MMA plus GTA were all effective at reducing OSA severity, commonly measured by AHI. The minimal clinically important difference for AHI is five events/hour (Appendix 15). While effect sizes varied across the interventions, they tended to be above this threshold. The highest of the mean differences in AHI was around -20 events/hour for CPAP, followed by -15 events/hour for EPAP, -10 events/hour for MADs or undefined OAs, and -8 events/hour to -11 events/hour for GTA. Mean differences in AHI around -48 were observed with MMA with or without GTA from small, uncontrolled studies on severe cases of OSA.60 The findings on EPAP were also from uncontrolled studies.68 Wide ranges of effect sizes were observed for TRDs, diet, exercise, and positional therapy. Some of the differences in effect sizes are reflected in the findings on active comparisons, where significant differences in AHI values were found between CPAP and MADs or undefined OAs and also between CPAP and diet, exercise, or positional therapy. No significant differences in effect sizes were found between MADs and diet or exercise programs and also between diet and exercise programs. Mixed findings were reported for the comparison between MADs and TRDs, where a 2011 SR5 reported comparable effects between the two interventions, while a 2011 study109 reported superior effects of TRDs, compared with MADs. These findings suggest that CPAP, compared with other interventions, may be most effective at improving OSA severity, while the largest effect was observed for MMA with or without GTA for severe cases OSA, who are eligible for surgery. The majority of studies on MMA and GTA, however, were in highly selected patients in uncontrolled pre-and-post studies with sample sizes of fewer than 10 patients. The results, therefore, must be interpreted with great caution, especially given the invasiveness of the procedures and potential adverse events. Some of the findings on CPAP, EPAP, MADs, diet, or exercise versus inactive controls were associated with high heterogeneity.

For combination therapy, CPAP plus MADs compared with MADs alone, CPAP plus diet programs compared with CPAP alone, and MADs plus positional therapy compared with MADs or positional therapy alone were more effective at reducing OSA severity. However, CPAP plus exercise programs, compared with CPAP alone, was not effective at reducing OSA severity. In general, these findings suggest that the various interventions in combination may have additive effects in their effectiveness in improving OSA severity.

Patients with nocturia experienced reduced OSA severity after CPAP. Subgroup and meta-regression analyses suggest that patients with higher EDS at baseline experienced greater effects with CPAP but lower effects with MADs or undefined OAs. Patients with more severe OSA at baseline experienced greater effects with CPAP, MMA, diet, and exercise but lower effects with MADs or undefined OAs and positional therapy. Longer study duration was associated with lower effects with CPAP, MADs, diet, exercise, and positional therapy. Baseline weight was not significantly associated with the effects of positional therapy. No subgroup or meta-regression analyses were found on sex, age, or adherence.


Table 22: Summary of Change in AHI/ODI/RDI From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP pre versus post

There was a large effect size on AHI/RDI in favour of CPAP, compared with pre-treatment and when compared with sham CPAP.

278 patients from 6 studies (1999 to 2013):

  • Moderate-to-severe OSA (mean AHI range: 21.3 to 63.6 events/hour)
  • Overweight to obese (mean BMI range: 27.8 to 37 kg/m2, where reported)
  • 11 days to 6 months of study duration
  • AHI/RDI Hedges' g (95% CI) = -1.48 (-1.68 to -1.28); P < 0.001; I2 = 5.5%

None

Very low to low

CPAP versus sham CPAP

84 patients from 2 studies (1999 to 2007):

  • Severe OSA (mean AHI range: 40.6 to 63.6 events/hour)
  • Obese (mean BMI: 33.1 kg/m2, where reported)
  • 2 to 3 months of study duration
  • AHI/RDI Hedges' g (95% CI) = -1.88 (-2.39 to -1.37); P < 0.001; I2 = 0%

None

Very low to low

Iftikhar 201655

CPAP versus inactive controls

AHI and ODI were significantly decreased with CPAP, compared with inactive controls.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI and 15 RCTs provided ODI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) = -25.27 (-28.52 to -22.03); P = NR
  • ODI MD (95% CI) = -20.40 (-25.19 to -15.62); P = NR

None

High

Sharples 201659

CPAP versus inactive controls

Overall, AHI was significantly decreased in the CPAP group, compared with the inactive control group.

The effect of CPAP on AHI increased with baseline OSA severity, as judged by baseline AHI, and in sleepier patients, as judged by baseline ESS (i.e., moderate versus severe).

Longer treatment durations were associated with decreasing treatment effects.

1,596 patients from 25 RCTs (2001 to 2013):

  • Mild-to-severe OSA (AHI or DI: NR)
  • (for the 6,757 patients included in the SR) Overweight-to-obese (mean BMI range: 28.3 to 35.1 kg/m2)
  • 3 to 156 weeks of study duration
  • AHI MD (95% CI) = -25.37 events/hour (-30.67 to -20.07); P < 0.001; I2 = 96.1%

Subgroup analysis:

  • Baseline AHI:a
    • Mild: AHI MD (95% CI) = -2.40 (-3.67 to -1.13); P < 0.001; I2 = NA
    • Moderate: AHI MD (95% CI) = -13.67 (-16.13 to -11.20); P < 0.001; I2 = 47%
    • Severe: AHI MD (95% CI) = -33.04 (-39.75 to -26.34); P < 0.001; I2 = 90%
  • Baseline ESS:b
    • Normal/mild: AHI MD (95% CI) = -32.50 (−43.55 to -21.45); P < 0.001; I2 = NA
    • Moderate: AHI MD (95% CI) = -17.54 (-22.51 to -12.56); P < 0.001; I2 = 95%
    • Severe: AHI MD (95% CI) = -34.73 (-58.90 to -10.57); P = 0.005; I2 = 95%
  • Study duration:
    • 2 to 4 weeks: AHI MD (95% CI) = -32.90 (−43.78 to -22.02); P < 0.001; I2 = 93%
    • 5 to 12 weeks: AHI MD (95% CI) = -22.34 (−29.84 to -14.85); P < 0.001; I2 = 96%
    • >12 weeks: AHI MD (95% CI) = -14.25 (−19.03 to -9.46); P < 0.001; I2 = 82%

Moderate

Wang 201570

CPAP pre versus post

CPAP was associated with a significant reduction in AHI, when compared with pre-treatment.

85 patients from 2 RCTs (2004, 2015):

  • Moderate-to-severe OSA (mean AHI range: 24.0 to 51.5 events/hour)
  • Nocturia (100%)
  • 1 month of study duration
  • AHI MD (95% CI) = -31.57 (−33.87 to −29.28); P < 0.00001; I2 = 98%

None

Moderate

Balk 20115

CPAP versus inactive controls

CPAP was associated with a significant reduction in AHI, when compared with inactive controls and when compared with sham CPAP.

417 patients from 7 RCTs (2001 to 2007):

  • Moderate-to-severe OSA (mean AHI range: 19 to 46 events/hour)
  • Overweight to obese (mean BMI range: 27.3 to 33.5 kg/m2)
  • 1 to 6 months of study duration
  • AHI MD (95% CI) = -19.85 (-26.06 to -13.65); P < 0.001; I2 = 86%

None

Low to moderate

CPAP versus sham CPAP

311 patients from 8 RCTs (1999 to 2008):

  • Severe OSA (mean AHI range: 35 to 65 events/hour)
  • Overweight to obese (mean BMI range: 27.5 to 33.4 kg/m2)
  • 1 week to 3 months of study duration
  • AHI MD (95% CI) = -46.39 (-56.97 to -35.81); P < 0.001; I2 = 69.6%

None

Mixed

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; DI = desaturation index; ESS = Epworth Sleepiness Scale; MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; RCTs = randomized controlled trials; RDI = respiratory disturbance index; SR = systematic review.
a Baseline AHI: mild, 5 to 14; moderate, 15 to 30; severe, > 30 events/hour.
b Baseline ESS: normal/mild, 0 to 9; moderate, 10 to 15; severe, 16 to 24.

Table 23: Summary of Change in AHI/ODI From EPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Riaz 201568

EPAP pre versus post

EPAP was associated with a significant, and large, decrease in AHI post-treatment, when compared with pre-treatment.

EPAP was associated with a significant, but moderate, decrease in ODI post-treatment, when compared with pre-treatment.

345 patients from 10 studies (2008 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 18.1 to 43.3 events/hour)
  • Overweight to obese (mean BMI range: 28.2 to 34.9 kg/m2, where reported)
  • 1 night to 2 months of study duration, where reported
  • AHI MD (95% CI) = -14.78 (-19.12 to -10.45); P < 0.00001, I2 = 72%
  • AHI SMD (95% CI) = −0.94 (-1.31 to -0.57); P < 0.00001, I2 = 80%

None

High

247 patients from 7 studies (2008 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 18.1 to 43.3 events/hour)
  • Overweight to obese (mean BMI range: 28.2 to 34.9 kg/m2, where reported)
  • 1 night to 2 months of study duration, where reported
  • ODI MD (95% CI) = -7.69 (-11.78 to -3.60); P = 0.0002, I2 = 67%
  • ODI SMD (95% CI) = −0.58 (-0.91 to -0.25); P = 0.0006, I2 = 69%

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; EPAP = expiratory positive airway pressure; MA = meta-analysis; MD = mean difference; MR = meta-regression; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; SMD = standardized mean difference.

Table 24: Summary of Change in AHI/RDI From OAs Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Iftikhar 201655

MADs versus inactive controls

ESS was significantly decreased with MADs, compared with inactive controls.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI and 15 RCTs provided ODI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) =
    -15.20 (-19.50 to -10.91); P = NR
  • ODI MD (95% CI) = -12.58 (-18.84 to -6.32); P = NR

None

High

Serra-Torres 201658

MADs pre versus post

MADs were associated with a decrease in AHI post-treatment, compared with pre-treatment.

1,505 patients from 18 studies (2004 to 2014):

  • Mild-to severe OSA (mean AHI range: 14 to 45.5 events/hour)
  • Overweight to obese (mean BMI range: 25.9 to 32.3 kg/m2, where reported)
  • 0.5 to 84 months of study duration
  • Mean AHI range:
    • Baseline: 14 to 45.5 events/hour
    • Follow-up: 4.1 to 19.6 events/hour
  • Mean AHI % reduction range: 21% to 80%
  • No MA

None

Moderate to high

Sharples 201659

MADs versus inactive controls

Overall, AHI was significantly decreased in the MAD group, compared with the inactive control group.

Baseline AHI and baseline ESS did not appear to have any effect on the reduction of AHI by MADs.

The effect of MADs on AHI was greater in trials of shorter durations.

557 patients from 11 RCTs (1997 to 2014):

  • Mild-to-severe OSA (AHI or DI: NR)
  • (for the 6,757 patients included in the SR) Overweight to obese (mean BMI range: 28.3 to 5.1 kg/m2)
  • 4 to 26 weeks of study duration, where reported
  • AHI MD (95% CI) = -9.29 (-12.28 to -6.30); P < 0.001; I2 = 60.4%

Subgroup analysis:

  • Baseline AHI:a
    • Mild: -7.79 (-16.38 to 0.79); P = 0.075; I2 = 65.1%
    • Moderate: -10.72 (−14.59 to -6.85); P < 0.001; I2 = 52.0%
    • Severe: -7.95 (-15.94 to 0.05); P = 0.051; I2 = 31.6%
  • Baseline ESS:b
    • Moderate: -6.69 (−8.98 to -4.41); P < 0.001; I2 = 35%
    • Severe (from 1 RCT): -2.10 (-12.33 to 8.13); P = 0.687; I2 = NA
  • Treatment duration:
    • 2 to 12 weeks: -9.69 (-13.27 to -6.12); P < 0.001; I2 = 68%
    • >12 weeks: -6.78 (−13.24 to -0.33); P = 0.039; I2 = 23%

Moderate

Zhu 201571

OAs versus inactive controls

OAs were associated with a significant reduction in AHI, when compared with inactive controls.

Baseline OSA severity did not appear to have any significant effect on AHI by OAs.

679 patients from 12 RCTs and 1 cohort study (2002 to 2015):

  • (for the 840 patients included in the SR) Moderate-to-severe OSA (mean AHI ± SD: 23.15 ± 8.18 events/hour [for OA group] and 22.55 ± 7.59 events/hour [for control group])
  • Comorbidities: NR
  • 4 weeks to 6 months of study duration
  • AHI MD (95% CI) = −10.26 (-12.59 to −7.93); P < 0.00001; I2 = 90%

Subgroup analysis:

  • Baseline AHI:a
    • Mild-to-moderate: AHI MD (95% CI) = -8.13 (-10.69 to -5.57); P = NR; I2 = NR
    • Mild-to-severe: AHI MD (95% CI) = -13.54 (-15.75 to -11.36); P = NR; I2 = NR

Low

Okuno 201474

OAs versus control appliances

OAs were associated with a significant reduction in AHI, when compared with control appliances.

106 patients from 3 RCTs (2005 to 2011):

  • Moderate-to-severe OSA (mean AHI range: 22.1 to 39.1 events/hour [for OA group] and 20.1 to 32.6 events/hour [for control group])
  • Comorbidities: NR
  • 1 to 6 months of study duration
  • AHI MD (95% CI) = −7.05 (-12.07 to −2.03); P = 0.006; I2 = 0%

None

Low to moderate

Balk 20115

MADs versus inactive controls

MADs were associated with a significant reduction in AHI, when compared with inactive controls or sham OAs.

331 patients from 4 RCTs (2000 to 2008):

  • Moderate-to-severe OSA (mean AHI range: 19 to 34 events/hour)
  • Overweight to obese (mean BMI range: 27.3 to 31.3 kg/m2); exclusion of patients with heart disease and diabetes
  • 1 to 10 weeks of study duration
  • AHI MD (95% CI) = −11.39 (-15.21 to −7.58); P = NR; I2 = 55.3%

None

Moderate

MADs versus sham OAs

234 patients from 3 RCTs (2001 to 2005):

  • Moderate-to-severe OSA (mean AHI range: 25 to 31 events/hour)
  • Overweight to obese (mean BMI range: 29 to 32 kg/m2)
  • 1 to 6 weeks of study duration
  • AHI MD (95% CI) = −14.04 (-20.06 to −8.02); P = NR; I2 = 0.0%

None

Moderate

MADs versus control and sham OAs

565 patients from 7 RCTs (2000 to 2008):

  • Moderate-to-severe OSA (mean AHI range: 19 to 34 events/hour)
  • Overweight to obese (mean BMI range: 27.3 to 32 kg/m2)
  • 1 to 10 weeks of study duration
  • AHI MD (95% CI) = −11.76 (-14.64 to −8.87); P = NR; I2 = 30.8%

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; DI = desaturation index; ESS = Epworth Sleepiness Scale; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NR = not reported; OA = oral appliance; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; RCTs = randomized controlled trials; RDI = respiratory disturbance index; SD = standard deviation; SR = systematic review.
a Baseline AHI: mild, 5 to 14; moderate, 15 to 30; severe, > 30 events/hour.
b Baseline ESS: normal/mild, 0 to 9; moderate, 10 to 15; severe, 16 to 24.

Table 25: Summary of Change in AHI/RDI From Surgery Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Song 201624

GP or GTA pre versus post

Standard GP and GTA can improve OSA outcomes.

61 patients from 9 studies (1994 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 13.0 to 88.2 events/hour)
  • Comorbidities: NR
  • Study duration: NR
  • AHI % change before and after:
    • Standard GP: -43.8%
    • Modified GP: +37.3%
    • GTA: -45.7%
  • AHI MD (95% CI):
    • GP (from 2 studies): -7.78 events/hour (−9.84 to -5.72); P < 0.00001; I2 = 0%
    • GTA (from 3 studies): -11.1 events/hour (-17.9 to -4.25); P = 0.001; I2 = 0%
  • AHI SMD (95% CI):
    • GP: -2.97 (-4.31 to -1.63); P < 0.0001; I2 = 0%
    • GTA: -0.94 (-1.59 to -0.28); P = 0.005; I2 = 0%

None

Low

Zaghi 201660

MMA ± GTA pre versus post

MMA with or without GTA was associated with a significant decrease on AHI and RDI, when pre- and post-treatment were compared.

The effect of MMA with or without GTA on AHI increased with baseline OSA severity.

455 patients from 36 studies (1986 to 2014):

  • (for 455 of the 518 patients included in the SR) Severe OSA (mean AHI: 57.2 events/hour)
  • (for 82 of the 518 patients included in the SR) Obese (mean BMI: 33.8 kg/m2)
  • (for the 45 studies included in the SR) 2 to 6 months of study duration
  • AHI MD ± SD = -47.8 ± 25.0; P < 0.001; I2 = 61.3%

Subgroup analysis:

  • Baseline AHI:
    • < 30 events/hour: AHI MD ± SD = -14.1 ± 11.6
    • 30 to 60 events/hour: AHI MD = -37.4
    • 60 to 90 events/hour: AHI MD = -61.0
    • ≥ 90 events/hour: AHI MD ± SD = -94.5 ± 23.5

Moderate

68 patients from 11 studies (2001 to 2013):

  • (for 455 of the 518 patients included in the SR) Severe OSA (mean AHI: 57.2 events/hour)
  • (for 82 of the 518 patients included in the SR) Obese (mean BMI: 33.8 kg/m2)
  • (for the 45 studies included in the SR) 2 to 6 months of study duration
  • RDI MD ± SD = -44.4 ± 33.0; P < 0.001; I2 = 41.3%

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; GP = genioplasty; GTA = genial tubercle advancement; MA = meta-analysis; MD = mean difference; MMA = maxillomandibular advancement; MR = meta-regression; NR = not reported; RDI = respiratory disturbance index; OSA = obstructive sleep apnea; SD = standard deviation; SR = systematic review.

Table 26: Summary of Change in AHI/ODI From Lifestyle Interventions Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Aiello 201650

Exercise programs versus inactive controls

Exercise was associated with having a greater decrease in AHI, when compared with inactive control.

184 patients from 7 RCTs (2000 to 2014):

  • Mild-to-severe OSA (mean AHI range: 3.5 to 42.3 events/hour, where reported)
  • Overweight to obese (mean BMI range: 25.9 to 35.5 kg/m2, where reported)
  • 2 to 6 months of study duration (for the 8 RCTs included in the SR)
  • AHI MD (95% CI) = −0.54 (-0.86 to -0.21); P = NR; I2 = 20%
  • AHI decrease of 4 events/hour OR (95% CI) = 77.33 (27.91 to 187.49)
  • AHI decrease of 4 events/hour RR (95% CI) = 7.29 (4.07 to 13.06)

None

Mixed

Iftikhar 201655

Exercise programs versus inactive controls

AHI, but not ODI, was significantly decreased with diet and exercise, compared with inactive controls.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI and 15 RCTs provided ODI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) = −17.23 (-25.82 to −8.64); P = NR
  • ODI MD (95% CI) = −9.99 (-22.01 to 2.02); P =NR

None

High

Diet programs versus inactive controls

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) = −12.27 (-18.79 to −5.75); P = NR

None

High

Ashrafian 201561

Weight-loss programs pre versus post

Non-surgical weight-loss programs were associated with a significant reduction in AHI, when pre- and post-treatment were compared.

825 patients from 20 studies (1987 to 2014):

  • Mild-to-severe OSA (mean AHI range: 10 to 90 events/hour)
  • Overweight to obese (mean BMI range: 29.8 to 54 kg/m2)
  • 1 to 94.3 months of study duration
  • AHI MD (95% CI) = −11.39 (-14.98 to −7.81); P < 0.00001; I2 = 82%

None

Mixed

Mitchell 201473

Weight-loss programs versus inactive controls

Intensive weight-loss programs were associated with a significant reduction in AHI, when compared with inactive controls.

410 patients from 4 RCTs (2008 to 2009):

  • Mild-to-severe OSA (AHI threshold range: ≥ 5 to ≥ 15 events/hour)
  • Overweight to obese (BMI range: 28 to 40 kg/m2)
  • 8 weeks to 12 months of study duration
  • AHI MD (95% CI) = −16.09 (-25.64 to −6.54); P = 0.001; I2 = 92.2%
  • ODI MD (95% CI) = −14.18 (-24.23 to −4.13); P = 0.006; I2 = 83.5%

None

High (for the RCTs included in the MA)

Araghi 201376

Weight-loss programs versus inactive controls or pre versus post

Weight-loss programs were associated with a significant reduction in AHI, when compared with inactive control and when pre- and post-treatment were compared.

There was evidence of greater effects of weight-loss programs on AHI with increasing baseline AHI and shorter treatment durations, when pre- and post-treatment were compared but not when compared with inactive controls.

The greatest source of heterogeneity appeared to be studies with higher baseline AHI and also those with greater change in BMI.

519 patients from 7 RCTs (2006 to 2011) and 256 patients from 9 before-and-after studies (1987 to 2011):

  • Mild-to-severe OSA (mean AHI range: 10.0 to 66.5 events/hour)
  • Overweight to obese (mean BMI range: 26.5 to 54.6 kg/m2)
  • 1 to 2 years of study duration, where reported
  • RCTs only: AHI MD (95% CI) = -6.04 (−11.18 to -0.90); P = 0.02; I2 = 86%
  • Before-and-after studies only:
    • AHI MD (95% CI) = -12.26 (-18.51 to −6.02); P =0.0001; I2 = 88%
    • ODI MD (95% CI) = -18.91 (-23.40 to −14.43); P < 0.00001; I2 = 5%

Subgroup analysis:

  • Baseline AHI:
    • RCTs only:
      • < 15 events/hour: AHI MD (95% CI) = -1.99 (-5.58 to 1.61); I2 = 48%
      • 15 to 25 events/hour: AHI MD (95% CI) = −9.08 (-12.87 to −5.30); I2 = 0%
      • ≥ 25 events/hour: AHI MD (95% CI) = -4.91 (-21.97 to 12.15); I2 = 93%
    • Before-and-after studies only:
      • < 15 events/hour: −4.90 (-6.43 to −3.37); I2 = NA
      • 15 to 25 events/hour: -8.72 (-14.19 to -3.25); I2 = 0%
      • ≥ 25 events/hour: −15.60 (-22.95 to −8.24); I2 = 71%
  • Change in BMI:
    • RCTs only:
      • 0 to 3 kg/m2: -7.12 (-9.14 to -5.10); I2 = 12%
      • 3 to 5 kg/m2: -4.11 (-5.89 to -2.33); I2 = 0%
      • ≥ 5 kg/m2: -13.74 (−35.10 to 7.61); I2 = 95%
    • Before-and-after studies only:
      • 0 to 3 kg/m2: -10.50 (-16.46 to -4.53); I2 = 0%
      • 3 to 5 kg/m2: -15.00 (-25.44 to -4.56); I2 = NA
      • ≥ 5 kg/m2: -12.30 (−20.92 to -3.68); I2 = 92%
  • Treatment duration:
    • RCTs only:
      • ≤ 12 weeks: -8.78 (-15.65 to -1.87); I2 = 88%
      • < 12 weeks: -0.67 (-11.18 to 9.83); I2 = 88%
    • Before-and-after studies only:
      • ≤ 12 weeks: -14.79 (-24.41 to -5.17); I2 = 69%
      • < 12 weeks: -12.66 (-19.85 to -5.47); I2 = 76%

Meta-regression analysis:

  • For every event/hour increase in baseline AHI, change in AHI was additionally reduced by 0.41 event/hour (P = 0.001).
  • There was no significant association between weight loss and change in AHI (P = 0.186).

Mixed

Thomasouli 201319

Weight-loss programs versus inactive controls

Intensive weight-loss programs were associated with a significant reduction in AHI, when compared with usual care.

554 patients from 6 RCTs (2008 to 2012):

  • Mild-to-severe OSA (mean AHI range: 9.7 to 46.2 events/hour)
  • Overweight to obese (mean BMI range: 28.2 to 36.7 kg/m2)
  • 2 to 12 months of study duration
  • AHI MD (95% CI) = −4.55 (-7.12 to -1.98); P =0.041; I2 = 54.4%

None

Mixed

Balk 20115

Weight-loss programs versus inactive controls

Intensive weight-loss programs were associated with a significant reduction in AHI, when compared with inactive controls.

345 patients from 3 RCTs (2009):

  • Mild-to-severe OSA (mean AHI range: 9 to 37 events/hour)
  • Overweight to obese (BMI range: 31.4 to 36.7 kg/m2)
  • (in 1 RCT) 100% diabetes
  • 9 weeks to 1 year of study duration
  • Intensive intervention versus conservative intervention as control: change in AHI (95% CI) = -4.4 (-7.6 to −1.0); P = 0.011
  • Intensive intervention versus usual diet as control: change in AHI (95% CI) = -23 (-30 to -15); P < 0.001
  • Intensive intervention versus diabetes support and education as control: change in AHI (95% CI) = -9.7 (−13.6 to -5.7); P < 0.001
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; MA = meta-analysis; MD = mean difference; MR = meta-regression; NA = not applicable; NR = not reported; ODI = oxygen desaturation index; OR = odds ratio; OSA = obstructive sleep apnea; RCTs = randomized controlled trials; RDI = respiratory disturbance index; RR = risk ratio; SR = systematic review.

Table 27: Summary of Change in AHI From Combination Therapy Versus No Treatment

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

TRDs plus positional therapy (i.e., posture alarm) versus no treatment

There was no significant difference between TRDs plus posture alarm and no treatment in reducing AHI.

60 patients from 1 RCT (1991):

  • Mild-to-severe OSA (AHI threshold: > 12.5 events/hour)
  • Comorbidities: NR
  • Study duration: NR
  • AHI MD: no significant differences between combination therapy and no treatment
  • No MA

None

Low

AHI = Apnea-Hypopnea Index; MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; OSA = obstructive sleep apnea; RCT = randomized controlled trial; TRD = tongue-retaining device.

Table 28: Summary of Change in AHI/ODI From CPAP Versus OAs

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP versus OAs

There was a moderate effect size, favouring CPAP over OAs, for the reduction of AHI.

139 (included) or 132 (analyzed) patients from 2 RCTs (2004 and 2013):

  • Moderate OSA (mean AHI range: 21.3 to 26.2 events/hour)
  • Overweight to obese (mean BMI range: 27.8 to 31.1 kg/m2)
  • 60 days to 3 months of study duration
  • AHI Hedges' g (95% CI) = -0.90 (-1.11 to -0.69); P < 0.001; I2 = 0%

None

Very low to low

Iftikhar 201655

CPAP versus MADs

AHI and ODI were significantly decreased with CPAP, compared with MADs.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI and 15 RCTs provided ODI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) = -10.06 (−14.21 to -5.91); P = NR
  • ODI MD (95% CI) = -7.82 (−13.04 to -2.59); P = NR

None

High

Sharples 201659

CPAP versus MADs

CPAP had a significantly greater effect in reducing AHI, compared with MADs, regardless of baseline OSA severity, baseline ESS, or treatment durations.

746 (included) or 735 (analyzed) patients from 13 RCTs (1996 to 2013):

  • Moderate-to-severe OSA (mean AHI or DI range: NR)
  • (for the 6,757 patients included in the SR) Overweight to obese (mean BMI range: 28.3 to 35.1 kg/m2)
  • 4 to 26 weeks of study duration
  • AHI MD (95% CI) = -7.03 (−8.66 to -5.41); P < 0.001; I2 = 51.9%

Subgroup analysis:

  • Baseline AHI:a
    • Moderate: AHI MD (95% CI) = -7.48 (−9.19 to -5.77); P < 0.001; I2 = 28%
    • Severe: AHI MD (95% CI) = -7.22 (-11.25 to -3.20); P < 0.001; I2 = 74%
  • Baseline ESS:b
    • Moderate: AHI MD (95% CI) = -6.70 (−8.54 to -4.86); P < 0.001; I2 = 57%
  • Treatment duration:
    • 2 to 12 weeks: AHI MD (95% CI) = -7.19 (−9.12 to -5.25); P < 0.001; I2 = 59%
    • > 12 weeks: AHI MD (95% CI) = -6.78 (−10.31 to -3.25); P < 0.001; I2 = 42%

Moderate

Okuno 201474

CPAP versus OAs

CPAP was significantly more effective than OAs in reducing AHI.

278 (included) or 200 (analyzed) patients from 3 RCTs (2007 to 2011):

  • Moderate-to-severe OSA (mean AHI range: 20.9 to 40.3 events/hour)
  • Comorbidities: NR
  • 8 weeks to 6 months of study duration
  • AHI MD (95% CI) = -6.11 (−8.98 to -3.24); P < 0.0001; I2 = 63%

None

Low to moderate

Li 201377

CPAP versus OAs

CPAP was significantly more effective than OAs in reducing AHI.

471 patients from 9 RCTs (1996 to 2011):

  • Mild-to-severe OSA (AHI threshold: ≥ 5 events/hour)
  • Comorbidities: NR
  • 6 to 48 weeks of study duration

Crossover trials (from 6 RCTs):

  • AHI MD (95% CI) = -8.25 (−10.61 to −5.89); P < 0.001; I2 = 68%

Parallel-group trials (from 3 RCTs):

  • AHI MD (95% CI) = -5.96 (−8.51 to -3.40); P < 0.001; I2 = 0%

None

Low

Balk 20115

CPAP versus MADs

CPAP was significantly more effective than OAs in reducing AHI.

393 patients from 9 RCTs (1996 to 2009):

  • Moderate-to-severe OSA (mean AHI range: 18 to 40 events/hour)
  • Overweight to obese (mean BMI range: 26.7 to 34.1 kg/m2)
  • 2 weeks to 4 months of study duration
  • AHI MD (95% CI) = -7.69 (−10.09 to −5.29); P < 0.001; I2 = 60.3%

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; DI = desaturation index; ESS = Epworth Sleepiness Scale; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NR = not reported; OA = oral appliances; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; RCT = randomized controlled trial; RDI = respiratory disturbance index; SR = systematic review.
a Baseline AHI: mild, 5 to 14; moderate, 15 to 30; severe, > 30 events/hour.
b Baseline ESS: normal/mild, 0 to 9; moderate, 10 to 15: severe, 16 to 24.

Table 29: Summary of Change in AHI/ODI From CPAP Versus Lifestyle Interventions

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP versus exercise programs

CPAP was substantially more effective than exercise at reducing AHI.

16 patients from 1 RCT (2013):

  • Moderate OSA (mean AHI range: 26.2 events/hour)
  • Overweight (mean BMI range: 27.8 kg/m2)
  • 60 days of study duration
  • AHI Hedges' g (SE) = -2.01 (0.59)

None

Very low to low

Iftikhar 201655

CPAP versus exercise programs

AHI was significantly decreased with CPAP, compared with diet programs. However, there was no significant difference between CPAP and exercise programs in reducing AHI or ODI.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI and 15 RCTs provided ODI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) = -8.04 (-17.00 to 0.92); P = NR
  • ODI MD (95% CI) = -10.41 (-23.35 to 2.52); P = NR

None

High

CPAP versus diet programs

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) = -13.00 (-20.28 to -5.72); P = NR

None

High

Ha 201472

CPAP versus positional therapy (i.e., backpacks, thoracic anti-spine bands, or the Zzoma positional sleeper)

CPAP was associated with a significant reduction in AHI, compared with positional therapy.

71 patients from 3 RCTs (1999 to 2010):

  • Mild-to-moderate OSA (mean AHI range: 13 to 22.7 events/hour)
  • Obese (mean BMI range: 30 to 31 kg/m2)
  • 3 nights to 9 weeks of study duration
  • AHI MD (95% CI) = -4.28 (-7.83 to -0.72); P = 0.02; I2 = 70%

None

Moderate

Balk 20115

CPAP versus positional therapy (i.e., shoulder-head elevation pillows or devices worn on the back)

CPAP was associated with a significant reduction in AHI, compared with positional therapy.

94 patients from 3 RCTs (1999 to 2008):

  • Moderate OSA (mean AHI range: 18 to 27 events/hour)
  • Obese (mean BMI range: 30 to 34 kg/m2)
  • 2 weeks to 1 month of study duration
  • AHI MD: significant reductions, ranging from 6.1 to 16 events/hour, associated with CPAP compared with positional therapy
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; MA = meta-analysis; MD = mean difference; MR = meta-regression; NA = not applicable; NR = not reported; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; RCTs = randomized controlled trials; RDI = respiratory disturbance index; SE = standard error; SR= systematic review.

Table 30: Summary of Change in AHI From MADs Versus TRDs

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

MADs versus TRDs

There was no significant difference between MADs and TRDs in reducing AHI.

22 patients from 1 RCT (2009):

  • Moderate OSA (mean AHI: 27.0 events/hour)
  • Overweight (mean BMI: 29.3 kg/m2)
  • 1 week of study duration
  • AHI MD (95% CI) = -1 (-9.7 to 7.7); P = NR

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NR = not reported; OSA = obstructive sleep apnea; RCT = randomized controlled trial; TRD = tongue-retaining device.

Table 31: Summary of Change in AHI/ODI From MADs Versus Lifestyle Interventions

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Iftikhar 201655

MADs versus exercise programs

There was no significant difference between MADs and diet or exercise in reducing AHI or ODI.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI and 15 RCTs provided ODI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) = 2.02 (-7.31 to 11.37); P = NR
  • ODI MD (95% CI) = 2.59
    (-16.14 to 10.96); P = NR

None

High

MADs versus diet programs

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) = -2.93
    (-10.73 to 4.87); P = NR

None

High

AHI = Apnea-Hypopnea Index; CI = confidence interval; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NR = not reported; OSA = obstructive sleep apnea; RCTs = randomized controlled trials; ODI = oxygen desaturation index; SR = systematic review.

Table 32: Summary of Change in AHI From Diet Versus Exercise

Study

Patient Characteristics

Pooled Estimates from MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Iftikhar 201655

Diet programs versus exercise programs

There was no significant difference between diet and exercise programs in reducing AHI.

Network MA: 7,882 patients from 80 RCTs (1985 to 2015), where 56 RCTs provided AHI:

  • (for the 7,882 patients included in the SR) Mild-to-severe OSA (mean AHI range: 10 to 66.6 events/hour [for intervention groups] or 9 to 68.1 years [for control groups])
  • Comorbidities: NR
  • 2 to 144 weeks of study duration
  • AHI MD (95% CI) = 4.96 (-5.82 to 15.74); P = NR

None

High

AHI = Apnea-Hypopnea Index; CI = confidence interval; MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; OSA = obstructive sleep apnea; RCTs = randomized controlled trials.

Table 33: Summary of Change in AHI/RDI From TRDs Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Lazard 2009113

TRDs pre versus post

TRDs were associated with a significant reduction in AHI.

84 patients:

  • Mean AHI ± SD: 37 ± 19.5 events/hour
  • Mean BMI ± SD: 26 ± 3.8 kg/m2
  • Mean AHI ± SD (n=55):
    • Before: 38 ± 22.2 events/hour
    • After: 14 ± 13.2 events/hour
    • Difference: P < 0.001

None

Dort 2008115

TRDs pre versus post

TRDs with suction had significant reductions in RDI; TRDs without suction had non-significant reductions in RDI. TRDs with suction had greater reductions in RDI than TRDs without suction.

38 patients:

  • Mean RDI ± SD: 15.5 ± 17.7
  • Mean BMI ± SD: 29.4 ± 5.7 kg/m2
  • Mean RDI ± SD (n = 32):
    • Before: 15.5 ± 17.6 events/hour
    • After:
      • No suction: 13.5 ± 15.4, P = 0.391
      • Suction: 8.9 ± 7.6, P = 0.006
    • Difference:
      • No suction: P = 0.391
      • Suction: P = 0.006
      • Suction versus no suction: 4.9 (95% CI: 0.85 to 8.9), P = 0.019

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; OSA = obstructive sleep apnea; RDI = respiratory disturbance index; SD = standard deviation; TRD = tongue-retaining device.

Table 34: Summary of Change in AHI From GTA Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Kuscu 201591

GTA pre versus post

Genioglossus advancement surgery was associated with a significant reduction in AHI.

17 patients:

  • Mean AHI ± SD: 27.5 ± 8 events/hour
  • Mean BMI ± SD: 30.2 ± 4 kg/m2
  • Mean AHI ± SD:
    • Before: 27.5 ± 8 events/hour (range: 15.6 to 44.3)
    • After: 17.3 ± 12.6 events/hour (range: 4.1 to 40.4)
    • Difference: P = 0.002

None

Santos Junior 2007117

GTA pre versus post

Genioplasty for genioglossus advancement significantly reduced OSA severity and may provide an alternate treatment for non-obese OSA patients with hypopharynx obstruction.

10 patients:

  • AHI range: 5 to 30 events/hour
  • BMI threshold: < 30 kg/m2
  • Mean AHI ± SD:
    • Before: 12.38 ± 4.62 events/hour
    • After: 4.40 ± 5.7 events/hour
    • Difference: P < 0.001

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; GTA = genial tubercle advancement; OSA = obstructive sleep apnea; SD = standard deviation.

Table 35: Summary of Change in AHI/RDI/ODI From Positional Therapy Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Benoist 201678

Positional therapy (i.e., sleep position trainers) pre versus post

Positional therapy with a sleep position trainer significantly improved AHI and ODI in patients with positional OSA.

33 positional OSA patients:

  • Median AHI: 18.3 (IQR: 13.7 to 24.0) events/hour
  • Mean BMI ± SD: 27.9 ± 2.8 kg/m2
  • Median AHI:
    • Before (n = 33): 18.3 (IQR: 13.7 to 24.0) events/hour
    • After 3 months (n = 32): 12.5 (IQR: 4.5 to 21.8) events/hour
    • Difference: P = 0.034
  • Median ODI:
    • Before (n = 33): 21.0 (IQR: 14.7 to 29.2) events/hour
    • After 3 months (n = 32): 12.9 (IQR: 5.5 to 23.3) events/hour
    • Difference: P = 0.011
  • Sleep position:
    • Supine median AHI:
      • Before (n = 33): 43.0 (IQR: 24.2 to 59.6) events/hour
      • After 3 months (n = 32): 32.4 (IQR: 14.2 to 66.2) events/hour
      • Difference: P = 0.023
    • Non-supine median AHI:
      • Before (n = 33): 4.8 (IQR: 2.3 to 8.6) events/hour
      • After 3 months (n = 32): 7.9 (IQR: 3.3 to 16.3) events/hour
      • Difference: P = 0.002

Scarlata 201681

Positional therapy (i.e., neck position devices) pre versus post

The neck position device was effective at significantly improving AHI, ODI, and RDI in patients with positional OSA.

20 positional OSA patients:

  • Mean AHI ± SD: 16.8 ± 9.5 events/hour
  • Mean BMI ± SD: 28.9 ± 4.0 kg/m2
  • Mean AHI ± SD:
    • Before: 16.8 ± 9.5 events/hour
    • After 3 nights: 4.4 ± 5.5 events/hour
    • Difference: P < 0.001
  • Mean ODI ± SD:
    • Before: 13.7 ± 7.5 events/hour
    • After 3 nights: 3.8 ± 5.2 events/hour
    • Difference: P < 0.001
  • Mean RDI ± SD:
    • Before: 20.0 ± 9.5 events/hour
    • After 3 nights: 5.2 ± 5.6 events/hour
    • Difference: P < 0.001

None

Afrashi 201582

Positional therapy (i.e., pillows for prone positioning) pre versus post

A pillow for prone positioning significantly improved AHI and ODI in patients with mild-to-moderate OSA.

29 patients:

  • Mean AHI ± SD: 15.5 ± 6.2 events/hour
  • Mean BMI ± SD: 28.9 ± 3.2 kg/m2
  • Mean AHI ± SD:
    • Before: 15.5 ± 6.2 events/hour
    • After 2 nights: 10.3 ± 12.8 events/hour
    • Difference (95% CI): -5.2 (-0.1 to −10.3) events/hour, P = 0.04
  • Mean ODI (95% CI):
    • Before: 13.0 (10.0 to 16.1) events/hour
    • After 2 nights: 4.0 (1.5 to 7.6) events/hour
    • Difference: -6.7 (−11.8 to -1.4) events/hour, P = 0.04

None

Bidarian-Moniri 201583

Positional therapy (i.e., mattresses and pillows for prone positioning) pre versus post

Positional therapy with a mattress and pillow for prone positioning significantly improved AHI and ODI levels in patients with OSA.

14 patients:

  • Mean AHI: 26 events/hour (range: 6 to 53 events/hour)
  • Mean AHI:
    • Before: 26 events/hour (range 6 to 53)
    • After 4 weeks: 8 events/hour (range 1 to 26)
    • Difference: P < 0.001
  • Mean ODI:
    • Before: 21 events/hour (range 6 to 51)
    • After 4 weeks: 7 events/hour (range 1 to 25)
    • Difference: P < 0.001

None

Bidarian-Moniri 201584

Positional therapy (i.e., mattresses and pillows for prone positioning) pre versus post

Positional therapy with the mattress and a pillow for prone positioning significantly improved AHI and ODI levels in patients with OSA.

27 patients:

  • Mean AHI: 31 events/hour (range: 5 to 93 events/hour)
  • Mean BMI: 28 kg/m2 (range: 23 to 36 kg/m2)
  • Median AHI:
    • Before: 23 events/hour
    • After 2 nights: 7 events/hour
    • Difference: P < 0.001
  • Median ODI:
    • Before: 21 events/hour
    • After 2 nights: 6 events/hour
    • Difference: P < 0.001
  • Positional OSA:
    • Positional OSA:
      • Median AHI:
        • Before: 20 events/hour
        • After 2 nights: 5 events/hour
        • Difference: P < 0.001
      • Median ODI
        • Before: 19 events/hour
        • After 2 nights: 5 events/hour
        • Difference: P < 0.001
    • Non-positional OSA:
      • Median AHI:
        • Before: 45 events/hour
        • After 2 nights: 22 events/hour
        • Difference: NR
      • Median ODI:
        • Before: 22 events/hour
        • After 2 nights: 11 events/hour
        • Difference: NR

Chen 201585

Positional therapy (i.e., head-positioning pillows) pre versus post

The use of a head-positioning pillow had no significant impact on reducing ODI in patients with positional OSA.

25 patients:

  • Median AHI: 7.0 events/hour (IQR: 6.0 to 15.2 events/hour)
  • Median BMI: 24.8 kg/m2 (IQR: 23.1 to 26.4 kg/m2)
  • Median ODI:
    • Before (regular pillow): 4.2 events/hour (IQR: 1.5 to 8.9)
    • After 3 nights (head-positioning pillow): 3.5 events/hour (IQR: 1.6 to 8.5)
    • Difference: P = 0.247
  • Baseline weight:
    • Normal-weight patients (n = 13) median ODI:
      • Before (regular pillow): 4.0 events/hour (IQR: 1.3 to 7.7)
      • After 3 nights (head-positioning pillow): 3.4 events/hour (IQR: 1.0 to 9.1)
      • Difference: P = 0.366
    • Overweight patients (n = 12) median ODI:
      • Before (regular pillow): 4.4 events/hour (IQR: 2.7 to 9.1)
      • After 3 nights (head-positioning pillow): 4.3 (IQR: 2.3 to 8.6)
      • Difference: P = 0.346

de Vries 201586

Positional therapy (i.e., commercial devices or self-made constructions) pre versus post

Commercial devices or self-made constructions significantly reduced median AHI.

40 positional OSA patients:

  • Median AHI: 14.5 events/hour (IQR: 10.7 to 19.6 events/hour)
  • Mean BMI ± SD: 28.0 ± 4.1 kg/m2
  • Median AHI:
    • Before: 14.5 events/hour (IQR: 10.7 to 19.6)
    • After: 5.9 (IQR: 3.1-8.5)
    • Difference: P < 0.001
  • Sleep position:
    • Supine median AHI:
      • Before: 38.0 (IQR: 24.0 to 52.4)
      • After: 8.5 (IQR: 0 to 21.5)
      • Difference: P < 0.001
    • Non-supine median AHI:
      • Before: 3.9 (IQR: 2.2 to 7.1)
      • After: 4.3 (IQR: 1.4 to 8.9)
      • Difference: P = 0.2

Eijsvogel 201588

Positional therapy (i.e., tennis balls or sleep position trainers) pre versus post

Both tennis balls and sleep position trainers significantly reduced AHI, ODI, and RDI in patients with positional OSA.

26 (TBT) or 29 (SPT) positional OSA patients:

  • Mean AHI ± SD: 13.1 ± 9.1 (TBT) or 11.4 ± 4.9 (SPT) events/hour
  • Mean BMI ± SD: 26.8 ± 3.0 (TBT) or 27.6 ± 4.5 (SPT) kg/m2

TBT patients:

  • Mean AHI:
    • Baseline AHI ± SD:
      • TBT: 13.1 ± 9.1 events/hour
      • SPT: 11.4 ± 4.9 events/hour
    • After 1 month AHI (95% CI):
      • TBT: 5.8 (0.2 to 23.1) events/hour
      • SPT: 3.9 (0.4 to 30.8) events/hour
    • Difference:
      • TBT: P < 0.05
      • SPT: P < 0.05
  • Mean ODI:
    • Baseline ODI ± SD:
      • TBT: 10.9 ± 7.7 events/hour
      • SPT: 9.9 ± 5.0 events/hour
    • After 1 month ODI (95% CI):
      • TBT: 5.4 (0.4 to 15.1) events/hour
      • SPT: 4.4 (0.5 to 33.8) events/hour
    • Difference:
      • TBT: P < 0.01
      • SPT: P < 0.05
  • Mean RDI:
    • Baseline RDI ± SD:
      • TBT: 13.3 ± 9.1 events/hour
      • SPT: 11.9 ± 4.6 events/hour
    • After 1 month RDI (95% CI):
      • TBT: 6.0 (0.2 to 14.2) events/hour
      • SPT: 3.9 (0.4 to 30.8) events/hour
    • Difference:
      • TBT: P < 0.01
      • SPT: P < 0.01
  • Sleep position:
    • Supine:
      • Before AHI ± SD:
        • TBT: 37.3 ± 24.0 events/hour
        • SPT: 30.7 ± 15.3 events/hour
      • After 1 month AHI (95% CI):
        • TBT: 0.0 (0.0 to 116) events/hour
        • SPT: 0.0 (0.0 to 64.2) events/hour
      • Difference:
        • TBT: P < 0.01
        • SPT: P < 0.01
    • Non-supine:
      • Before AHI (95% CI):
        • TBT: 3.3 (0.0 to 13.7) events/hour
        • SPT: 3.9 (0.5 to 13.0) events/hour
      • After 1 month:
        • TBT: 5.0 (0.2 to 14.2) events/hour
        • SPT: 3.6 (0.4 to 30.8) events/hour
      • Difference:
        • TBT: not significant
        • SPT: not significant

Jackson 201590

Positional therapy (i.e., sleep position modification devices) versus inactive controls

A sleep position modification device was significantly more effective in reducing AHI, compared with sleep hygiene advice.

47 (sleep position modification device) or 39 (sleep hygiene advice) positional OSA patients:

  • Mean AHI ± SD: 20.1 ± 8.8 (sleep position modification device) or 21.8 ± 10.1 (sleep hygiene advice) events/hour
  • Mean BMI ± SD: 30.0 ± 5.3 (sleep position modification device) or 30.9 ± 7.7 sleep hygiene advice) kg/m2
  • Mean AHI ± SD:
    • Before:
      • Sleep position modification device: 20.1 ± 8.8 events/hour
      • Sleep hygiene advice: 21.8 ± 10.1 events/hour
    • After 4 weeks:
      • Sleep position modification device: 10.8 ± 9.9 events/hour
      • Sleep hygiene advice: 16.8 ± 15.9 events/hour
    • Difference (device versus advice): P = 0.013

None

Levendowski 201496

Positional therapy (i.e., neck position devices) pre versus post

Neck position devices were effective at significantly reducing supine sleep and improving AHI in patients with positional OSA.

30 positional OSA patients:

  • Mean AHI ± SD: 24.7 ± 14.7 events/hour
  • Mean BMI ± SD: 28 ± 3.4 kg/m2
  • Mean AHI ± SD:
    • Before: 24.7 ± 14.7
    • After 30 nights: 7.5 ± 7.7
    • Difference: P < 0.00001
  • Sleep position:
    • Supine:
      • Mean AHI ± SD:
        • Before: 44.9 ± 25.5 events/hour
        • After 30 nights: 4.5 ± 12.7 events/hour
        • Difference: P < 0.00001
    • Non-supine:
      • Mean AHI ± SD:
        • Before: 8.1 ± 7.9 events/hour
        • After 30 nights: 7.1 ± 7.8 events/hour
        • Difference: P = 0.300
  • OSA severity:
    • Mild OSA (n = 11):
      • AHI < 10 and > 50% decrease: 81.8% (n = 9)
      • AHI > 50% reduction: 0% (n = 0)
      • AHI > 35% reduction: 9.1% (n = 1)
      • No response: 9.1% (n = 1)
    • Moderate OSA (n = 10):
      • AHI < 10 and > 50% decrease: 80.0% (n = 8)
      • AHI > 50% reduction: 0% (n = 0)
      • AHI > 35% reduction: 0% (n = 0)
      • No response: 20.0 (n = 2)
    • Severe OSA (n = 9):
      • AHI < 10 and > 50% decrease: 55.6% (n = 5)
      • AHI > 50% reduction: 33.3% (n = 3)
      • AHI > 35% reduction: 11.1% (n = 1)
      • No response: 0% (n = 0)

van Maanen 201399

Positional therapy (i.e., sleep position devices) pre versus post

Sleep position trainers proved to be a significantly effective treatment for patients with positional OSA.

31 positional OSA patients:

  • Median AHI: 16.4 events/hour (IQR: 6.6 to 29.9 events/hour)
  • Mean BMI ± SD: 27.0 ± 3.7 kg/m2
  • Median AHI:
    • Before: 16.4 events/hour (IQR: 6.6 to 29.9)
    • After 29 ± 2 nights: 5.2 events/hour (IQR: 0.5 to 46.5)
    • Difference: P < 0.001
  • Sleep position:
    • Supine median AHI:
      • Before: 35.7 (IQR: 9.3 to 81.0)
      • After 29 ± 2 nights: 0.0 (IQR: 0.0 to 100.7)
      • Difference: P < 0.001
    • Non-supine median AHI:
      • Before: 3.2 (IQR: 0.0 to 16.2)
      • After 29 ± 2 nights: 4.3 (IQR: 0.1 to 48.0)
      • Difference: P = 0.052

Heinzer 2012102

Positional therapy (i.e., tennis balls) pre versus post

After 1 night of device use, there was a significant decrease in AHI. After 3 months of device use, there was no significant decrease in AHI, when compared with the first night of device use.

16 positional OSA patients:

  • Mean AHI ± SD: 26.7 ± 17.5 events/hour
  • Mean BMI ± SD: 25.4 ± 4.1 kg/m2
  • Mean AHI:
    • Before: 26.7 ± 17.5 events/hour
    • After first night with positional device: 6.0 ± 3.4 events/hour
    • After 3 months: 10.3 ± 8.2
    • Difference:
      • Before versus T0: P = 0.0002
      • T0 versus T3: P = 0.58
  • Sleep position:
    • Supine:
      • Mean AHI ± SD:
        • Before: 54.0 ± 21.2 events/hour
        • After first night with positional device (T0): 32.9 ± 19.2 events/hour
        • After 3 months (T3): 39.4 ± 24.6
    • Non-supine:
      • Mean AHI ± SD:
        • Baseline: 5.6 ± 5.0 events/hour
        • After first night with positional device (T0): 5.0 ± 3.6 events/hour
        • After 3 months (T3): 8.7 ± 7.4 events/hour

van Maanen 2012104

Positional therapy (i.e., sleep position modification devices) pre versus post

Patients with the device in the on or off modes experienced significantly more reductions on OSA severity than wearing no device.

30 positional OSA patients:

  • Mean AHI ± SD: 27.7 ± 2.4 events/hour
  • Mean BMI ± SD: 27.7 ± 3.6 kg/m2
  • Mean AHI:
    • Before (i.e., no device): 27.7 ± 2.4 events/hour
    • After:
      • Device attached in OFF mode: 23.5 ± 2.6 events/hour
      • Device attached in ON mode: 12.8 ± 2.2
    • Difference:
      • Before versus device in OFF mode: P = 0.04
      • Before versus device in ON mode: P = 0.00
      • Device in ON versus OFF mode: P = 0.00
  • Sleep position:
    • Supine:
      • Mean AHI ± SD:>
        • Baseline (no device): 59.7 ± 3.6 events/hour
        • Device attached in OFF mode: 45.0 ± 4.8 events/hour
        • Difference between baseline versus device in OFF mode: P = 0.00
        • Device attached in ON mode: 12.5 ± 3.1 events/hour
        • Difference between baseline versus device in ON mode: P = 0.50
        • Difference between ON versus OFF modes: P = 0.00
      • Non-supine:
        • Mean AHI ± SD:
          • Baseline (no device): 6.7 ± 1.2 events/hour
          • Device attached in OFF mode: 13.4 ± 2.7 events/hour
          • Difference between baseline versus device in OFF mode: P = 0.03
          • Device attached in ON mode: 11.2 ± 2.2 events/hour
          • Difference between baseline versus device in ON mode: P = 0.03
          • Difference between ON versus OFF modes: P = 0.49

Bignold 2011105

Positional therapy (i.e., sleep position modification devices) pre versus post

Position recording and supine avoidance devices significantly reduced AHI.

15 patients:

  • Mean AHI ± SD: 24.1 ± 10.5 events/hour
  • Mean BMI ± SD: 28.8 ± 2.5 kg/m2
  • AHI:
    • Difference: 45% reduction with active treatment, P = 0.03

None

Kim 2011108

Positional therapy (i.e., sleep position modification devices) pre versus post

There was a significant improvement in AHI and ODI in patients with positional OSA.

14 positional OSA patients:

  • Mean AHI ± SD: 22.8 ± 9.3 events/hour
  • Mean BMI ± SD: 26.3 ± 3.6 kg/m2
  • Mean AHI:
    • Baseline: 22.8 ± 9.3 events/hour
    • Experimental examination: 9.3 ± 8.3 events/hour
    • Control examination: 13.5 ± 9.0
    • Differences: P < 0.001
  • Mean ODI:
    • Baseline: 19.0 ± 9.9
    • Experimental examination: 8.6 ± 8.2
    • Control examination: 10.4 ± 8.1
    • Differences: P < 0.001
  • Sleep position:
    • Supine:
      • Mean AHI ± SD:
        • Baseline: 33.2 ± 13.9 events/hour
        • Experimental examination: 25.6 ± 22.0 events/hour
        • Control examination: 7.6 ± 13.1
        • Differences: P < 0.001
    • Non-supine:
      • Mean AHI ± SD:
        • Baseline: 5.6 ± 7.0 events/hour
        • Experimental examination: 6.7 ± 7.1 events/hour
        • Control examination: 1.1 ± 7.3
        • Differences: P = 0.585

Loord 2007116

Positional therapy (i.e., the Positioner) pre versus post

The Positioner used in this study significantly reduced OSA symptoms in patients with OSA.

18 positional OSA patients:

  • Mean AHI ± SD: 21.8 ± 12.0 events/hour
  • Mean AHI ± SD:
    • Before: 21.8 ± 12.0 events/hour
    • After 10 months: 14.3 ± 15.2 events/hour
    • Difference: P = 0.02

None

Oksenberg 2006118

Positional therapy (i.e., tennis balls) pre versus post

Most of the patients who used tennis balls had significant reductions in AHI.

78 positional OSA patients:

  • Mean AHI ± SD: 25.5 ± 17.3 events/hour
  • Mean BMI ± SD: 28.1 ± 3.7 kg/m2
  • Mean AHI ± SD:
    • Before (n = 12): 46.5 ± 19.9 events/hour
    • After (n = 12): 17.5 ± 19.4 events/hour
    • Difference: P < 0.002
  • Sleep position:
    • Supine:
      • Mean AHI ± SD:
        • Before (n = 12): 57.0 ± 22.4 events/hour
        • After (n = 12): 44.4 ± 28.7 events/hour
        • Differences: P = NR
    • Non-supine:
      • Mean AHI ± SD:
        • Before (n = 12): 11.6 ± 8.2 events/hour
        • After (n = 12): 13.8 ± 22.0 events/hour
        • Differences: P = NR

AHI = Apnea-Hypopnea Index; AI = Apnea Index; BMI = body mass index; CI = confidence interval; HI = hypopnea index; HPP = head-positioning pillow; IQR = interquartile range; NR = not reported; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; RDI = respiratory disturbance index; SD = standard deviation; SPT = sleep position trainer; TBT = tennis ball technique.

Table 36: Summary of Change in AHI/RDI From Combination Therapy Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Islam 201495

MMA plus GTA pre versus post

Maxillomandibular and mandibular advancement surgery was successful at significantly decreasing AHI values post-surgery.

51 patients:

  • Mean AHI ± SD: 42 ± 17
  • Mean BMI ± SD: 26.4 ± 3.0 kg/m2
  • Mean AHI ± SD:
    • Before: 42 ± 17 events/hour
    • After: 8 ± 7 events/hour
    • Difference: P < 0.001

None

Dort 2012101

MADs plus TRDs pre versus post

The combination of mandibular advancement and tongue retention had a significant decrease in RDI values post-treatment.

41 patients:

  • Mean RDI ± SD: 33.5 ± 15.9 events/hour
  • Mean BMI ± SD: 32.2 ± 5.8 kg/m2
  • Mean RDI (95% CI):
    • Before (n = 41): 33.5 events/hour (28.6 to 38.4)
    • After 20 weeks:
      • 6 mm MADs + TRDs (n = 28): 19.2 events/hour (13.9 to 24.5)
      • 8 mm MADs + TRDs (n = 26): 18.1 events/hour (13.3 to 23.0)
    • Difference:
      • 6 mm MADs + TRDs: P = NR
      • 8 mm MADs + TRDs P = 0.001

None

Bruno 2008114

MMA plus GTA pre versus post

MMA plus modified genioplasty surgery reduced RDI values, but no statistical test was conducted.

4 patients:

  • RDI threshold: > 35 events/hour
  • Mean BMI: NR

RDI thresholds:

  • Before: > 35 events/hour
  • After: < 10 events/hour

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; NR = not reported; OSA = obstructive sleep apnea; RDI = respiratory disturbance index; SD = standard deviation; TRD = tongue-retaining device.

Table 37: Summary of Change in AHI From MADs Versus TRDs

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Sutherland 2011109

MADs versus TRDs

Both MADs and TRDs reduced AHI, but no statistical testing was conducted.

39 patients:

  • Mean AHI ± SD: 26.9 ± 17.1 events/hour
  • Mean BMI ± SD: 29.2 ± 5.5 kg/m2
  • Mean AHI ± SD (events/hour):
    • Baseline:
      • MADs: 26.9 ± 17.1
      • TRDs: 26.8 ± 18.1
    • After:
      • MADs: 12.0 ± 12.6
      • TRDs: 11.0 ± 9.1

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; MAD = mandibular advance device; SD = standard deviation; TRD = tongue-retaining device.

Table 38: Summary of Change in AHI From MADs versus MMA

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Garreau 201494

MADs versus MMA

For patients with moderate-to-severe OSA, MMA surgery was significantly more effective than MAD treatment and may be an alternative for those who refuse or find CPAP unacceptable.

161 (MADs) or 37 (MMA) patients:

  • Moderate-to-severe OSA (46.6% moderate and 53.4% severe [MADs] or 27.0% moderate and 72.1% severe [MMA])
  • 17.4% (MADs) or 16.2% (MMA) obese

MMA patients:

  • Mean AHI ± SD:
    • Before:
      • MADs: 32.45 ± 15.0 events/hour
      • MMA: 42.76 ± 22.72 events/hour
      • Difference: P < 0.001
    • After:
      • MADs: 14.88 ± 11.4 events/hour
      • MMA: 11.59 ± 12.11 events/hour
      • Difference: P = 0.120
    • Change in AHI:
      • MADs: -17.57 ± 14.16 events/hour,
      • MMA: -31.16 ± 21.19 events/hour
      • Difference: P < 0.001

None

AHI = Apnea-Hypopnea Index; CPAP = continuous positive airway pressure; MAD = mandibular advancement device therapy; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; SD = standard deviation.

Table 39: Summary of Change in AHI/ODI From Combination Therapy Versus Active Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Dieltjens 201587

MADs plus positional therapy (i.e., sleep position trainers) versus MADs or positional therapy (i.e., sleep position trainers) alone

A combination of MADs plus a sleep position trainer led to more significant reductions in AHI than either intervention alone.

20 patients:

  • Mean AHI ± SD: 24.6 ± 10.2 events/hour
  • Mean BMI ± SD: 26.4 ± 3.0 kg/m2
  • Mean AHI (IQR):
    • Before:
      • MADs plus SPT: 20.9 events/hour (17.0 to 34.0)
      • MADs: 20.9 events/hour (17.0 to 34.0)
      • SPT: 20.9 events/hour (17.0 to 34.0)
    • After:
      • MADs plus SPT: 5.5 events/hour (3.4 to 7.2)
      • MADs: 11.0 events/hour (6.6 to 14.0)
      • SPT: 12.8 events/hour (3.9 to 17.9)
    • Differences:
      • MADs plus SPT: P < 0.008
      • MADS: P < 0.008
      • SPT: P < 0.008
      • MADS plus SPT, compared with MAD: P < 0.001
      • MADS plus SPT, compared with SPT: P < 0.008
  • Mean ODI (IQR):
    • Before:
      • MADs plus SPT: 7.7 events/hour (6.6 to.5)
      • MADs: 7.7 events/hour (6.6 to 16.5)
      • SPT: 7.7 events/hour (6.6 to 16.5)
    • After:
      • MADs plus SPT: 1.8 events/hour (1.0 to 3.0)
      • MADs: 3.8 events/hour (1.2 to 5.5)
      • SPT: 2.6 events/hour (1.0 to 4.6)
    • Difference:
      • MADs plus SPT: P < 0.008
      • MADS: P < 0.008
      • SPT: P < 0.008
      • MADS plus SPT, compared with MAD: P < 0.001
      • MADS plus SPT, compared with SPT: P = non-significant

None

Ackel-D'Elia 2012100

CPAP plus exercise programs versus CPAP alone

CPAP plus exercise and CPAP both significantly reduced AHI.

13 (CPAP plus exercise) or 19 (CPAP) patients:

  • Mean AHI ± SD: 40.5 ± 22.9 (CPAP plus exercise) or 42.3 ± 21.6 (CPAP) events/hour
  • Mean BMI ± SD: 28.0 ± 3.1 (CPAP plus exercise) or 28.5 ± 2.2 (CPAP) kg/m2
  • Mean AHI ± SD:
    • Before:
      • CPAP plus exercise: 40.5 ± 22.9 events/hour
      • CPAP: 42.3 ± 21.6 events/hour
    • After 2 months:
      • CPAP plus exercise: 5.9 ± 4.4, P < 0.05
      • CPAP: 6.2 ± 4.4
    • Difference:
      • CPAP plus exercise: P < 0.05
      • CPAP: P < 0.05

None

El-Solh 2011106

CPAP plus MADs versus MADs alone

The combination therapy of CPAP plus MADs and MADs alone significantly reduced AHI. The combination of CPAP plus MADs reduced AHI more so than MADs alone.

10 patients:

  • Mean AHI ± SD: 23.5 ± 13.4 events/hour
  • Mean BMI ± SD: 26.9 ± 3.2 kg/m2
  • Mean AHI ± SD:
    • Before: 23.5 ± 13.4
    • After 8 weeks:
      • CPAP plus MADs: 4.8 ± 2.8 events/hour
      • MADs: 11.4 ± 4.8 events/hour
    • Difference:
      • CPAP plus MADs: P < 0.05
      • MADs: P < 0.05

None

Johansson 2011107

CPAP plus diet programs versus CPAP alone

After CPAP plus a weight-loss program, patients with severe OSA experienced greater reductions in AHI, compared with CPAP alone. However, both treatments had significant reductions in AHI.

63 patients:

  • Mean AHI ± SD: 36 ± 15 events/hour
  • BMI range: 30 to 40 kg/m2
  • Mean AHI ± SD:
    • Before: 36 events/hour
    • After 1 year of study duration:
      • CPAP with very-low-energy diet (0 to 9 weeks): 15 ± 9 events/hour
      • After full program (0 to 52 weeks): 19 ± 14 events/hour
    • Difference (MD ± SD, 95% CI):
      • After very-low-energy diet (0 to 9 weeks): −21 ± 16 (95% CI, −17 to −25), P < 0.001
      • After full program (0 to 52 weeks): -17 ± 16 (95% CI, 13 to -21), P < 0.001
  • Mean supine AHI ± SD:
    • Before: 36 events/hour
    • After 1 year of study duration:
      • CPAP with very-low-energy diet (0 to 9 weeks): 25 ± 17 events/hour
      • After full program (0 to 52 weeks): 27 ± 18 events/hour
    • Difference (MD ± SD, 95% CI):
      • After very-low-energy diet (0 to 9 weeks): -23 ± 21 (95% CI, -18 to −29), P < 0.001
      • After full program (0 to 52 weeks): -21 ± 3 (95% CI, -16 to -27), P < 0.001

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; IQR = interquartile range; MAD = mandibular advancement device therapy; MD = mean difference; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; PSG = polysomnography; SD = standard deviation; SPT = sleep position trainer.


Success or Cure Rate

Overview of Reviews

1) Oral appliances versus inactive controls

One SR51 reported on rates of success, defined as ([mean AHI at baseline - mean AHI after treatment] / mean AHI at baseline), in adults with moderate-to-severe OSA who were overweight to obese. The SR,51 with a sample size of 484 patients from 13 studies, reported significantly positive success rates of 0.62 with MADs, compared with pre-treatment. Study duration of the included primary studies ranged from three weeks to one year. No I2 scores were reported. The SR51 reported the quality of the included studies as moderate to high (Appendix 10). The findings of the SR are summarized in Table 40.

2) Surgery versus inactive controls

One SR60 reported on rates of success and cure, defined as greater than 50% reductions in AHI to fewer than 10, 15, or 20 events/hour or AHI levels of fewer than five events/hour, respectively, in adults with severe OSA who were obese. The SR,60 with a sample size of 518 patients from 45 studies, reported significantly positive success rates with MMA with or without GTA, compared with pre-treatment. Overall success rates of achieving AHI values of 10, 15, or 20 events/hour were 63.3%, 80.4%, and 85.5%, respectively. Overall cure rates of achieving AHI values of five events/hour were 38.5%. Study duration of the included primary studies ranged from two months to six months. No I2 scores were reported. The SR60 reported the quality of the included studies as moderate (Appendix 10). From subgroup analyses, the SR60 reported that patients with a higher preoperative AHI are less likely to achieve surgical success and cure with MMA with or without GTA, compared with pre-operation. The findings of the SR are summarized in Table 41.

3) Lifestyle interventions versus inactive controls

One SR5 reported on rates of cure, defined as AHI levels of fewer than five events/hour, in adults with moderate OSA who were obese. The SR,5 with a sample size of 72 patients from one study, reported significantly positive cure rates with very-low-calorie diets, compared with controls, with an odds ratio of 4.17. Study duration of the included primary study was one year. I2 scores were not applicable. The SR5 reported the quality of the included studies as moderate (Appendix 10). The findings of the SR are summarized in Table 42.

4) Continuous positive airway pressure versus oral appliances

One SR5 reported on treatment response in adults with severe OSA who were overweight to obese. The SR,5 with a sample size of 131 patients from two studies, reported that CPAP was significantly more effective at achieving AHI levels of fewer than five events/hour, compared with MADs, with risk differences ranging from 20% to 28.6%. Study duration of the included primary studies ranged from two months to 78 days. No I2 scores were reported. The SR5 reported the quality of the included studies as low to moderate (Appendix 10). From subgroup analyses, the SR5 reported that the effect of CPAP, compared with MADs, was specific to patients with baseline AHI greater than 30 events/hour. The findings of the SR are summarized in Table 43.

Review of Primary Studies

1) Positional therapy versus inactive controls

One study78 reported on treatment response and success rates in adults with moderate OSA, provided by median AHI. The study78 included positional OSA patients only who were overweight, provided by mean BMI. The study,78 with a sample size of 33 patients, reported treatment response, defined as greater than 50% reductions in AHI, in 37.5% of the patients with an apparatus designed to prevent sleep in the supine position. The study78 also reported treatment success, defined as greater than 50% reductions in AHI to fewer than five events/hour, in 31.3% of the patients with the apparatus. Concerns with the quality of the study78 were assessed to be low (Appendix 14). The findings of the primary study are summarized in Table 44.

Summary of Results on Success or Cure Rate

For success or cure rates, evidence was found on inactive comparisons with MADs, MMA, and lifestyle interventions (i.e., weight-loss programs and positional therapy). Evidence was also found on active comparisons between CPAP and MADs. Compared with inactive controls or pre-treatment, MADs, MMA, weight-loss programs, and positional therapy were associated with positive success or cure rates, although success and cure rates were not always identically defined, and it is unclear if these results are clinically important. CPAP was associated with higher cure rates, compared with MADs. Subgroup analyses suggest that patients with more severe OSA at baseline were less likely to achieve success and cure with MMA. No subgroup or meta-regression analyses were found on comorbidities, baseline EDS, sex, age, BMI, adherence, or study duration.


Table 40: Summary of Success Rates From OAs Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Bartolucci 201651

MADs pre versus post

MADs were associated with a significant improvement in AHI, when compared with pre-treatment.

484 patients from single arms in 13 RCTs (2000 to 2010):

  • Moderate-to-severe OSA (mean AHI range: 16.2 to 50.4 events/hour)
  • Overweight to obese (mean BMI range: 25.9 to 32.3 kg/m2)
  • 3 weeks to 1 year of study duration
  • Mean success ratea (95% CI) = 0.62 (0.56 to 0.68)

*

None

Moderate to high

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; MA = meta-analysis; MAD = mandibular advancement device; MR = meta-regression; OSA = obstructive sleep apnea; RCT = randomized controlled trial.
a Defined as ([mean AHI at baseline - mean AHI after treatment] / mean AHI at baseline).

Table 41: Summary of Success and Cure Rates from Surgery Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Zaghi 201660

MMA ± GTA pre versus post

Patients with a higher preoperative AHI are less likely to achieve surgical success and cure from MMA with or without GTA.

518 patients from 45 studies (1986 to 2013):

  • (for 455 of the 518 patients included in the SR) Severe OSA (mean AHI: 57.2 events/hour)
  • (for 82 of the 518 patients included in the SR) Obese (mean BMI: 33.8 kg/m2)
  • (for the 45 studies included in the SR) 2 to 6 months of study duration

Overall curea rate = 38.5% (175/455)

Overall success-10b rate = 63.3% (288/455)

Overall success-15c rate = 80.4% (366/455)

Overall success-20d rate = 85.5% (388/455)

Surgical success (%)

Preoperative AHI in events/hour

< 30 (n = 61)

30- < 60 (n = 192)

60- < 90 (n = 161)

≥90 (n = 41)

Curea

55.7e

45.8e

28.0

19.5

Success-10b

77.0e

72.9e

47.8

58.5

Success-15c

83.6f

88.0f

72.7

70.7

Success-20d

83.6g

91.7g

80.7g

75.6

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; GTA = genial tubercle advancement; OSA = obstructive sleep apnea; MA = meta-analysis; MMA = maxillomandibular advancement; SR = systematic review.
a AHI levels of fewer than 5 events/hour after MMA.
b > 50% reduction in AHI to < 10 events/hour after MMA.
c > 50% reduction in AHI to < 15 events/hour after MMA.
d > 50% reduction in AHI to < 20 events/hour after MMA.
e P < 0.001.
f P = 0.009.
g P = 0.003

Table 42: Summary of Cure Rates From Lifestyle Interventions Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

Weight-loss programs versus inactive controls

Treatment with a very-low-calorie diet was associated with a four-fold increase in the odds of being cured from OSA, when compared with inactive controls.

72 patients from 1 RCT (2009):

  • Moderate OSA (mean AHI = 9 events /hour)
  • Obese (mean BMI = 31.4 kg/m2)
  • 1 year of study duration
  • Curea rate OR (95% CI) = 4.17 (1.41 to 12.34); P = 0.011

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; MA = meta-analysis; MD = mean difference; OR = odds ratio; OSA = obstructive sleep apnea; RCT = randomized controlled trial.
a OSA was considered objectively cured when AHI < 5 events/hour.

Table 43: Summary of Treatment Response From CPAP Versus MADs

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

CPAP versus MADs

CPAP appeared to be more effective than MADs in achieving AHI below 5 events/hour.

131 patients from 2 RCTs (2008 and 2009):

  • Severe OSA (mean AHI range: 34 to 40 events/hour)
  • Overweight to obese (mean BMI range: 26.7 to 33.3 kg/m2)
  • 2 months to 78 days of study duration

From 1 RCT:

  • Complete response:a
    • RD (95% CI) = 28.6% (3.8% to 53%); P = 0.02
  • Partial response:b
    • No significant difference between CPAP and MADs
  • Overall (i.e., combined complete and partial) response:
    • No significant difference between CPAP and MADs

From 1 RCT:

  • Complete response:c
    • RD (95% CI) = 20% (1.9% to -37%); P = 0.02
  • Effective treatment:d
    • No significant difference between CPAP and MAD

Subgroup analysis:

  • Baseline AHI (from 1 RCT):
    • ≤ 30 events/hour:
      • Complete responsee RD (95% CI) = -4% (-25% to 18%); P = NS
      • Effective treatmentf RD (95% CI) = 4% (-25% to 18%); P = NS
    • > 30 events/hour:
      • Complete responsee RD (95% CI) = 43.3% (17% to 62%); P < 0.001
      • Effective treatmentf RD (95% CI) = 1.6% (-6.8% to 37%); P = NS

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; MA = meta-analysis; MAD = mandibular advancement device; MR = meta-regression; NS = not significant; OSA = obstructive sleep apnea; RCTs = randomized controlled trials; RD = risk difference.
a Defined as ≥50% reductions in AHI to < 5 events/hour
b Defined as ≥50% reductions in AHI to > 5 events/hour
c Defined as achieving AHI < 5 events/hour
d Defined as achieving AHI < 5 events/hour or > 50% reductions in AHI to < 20 events/hour.
e Defined as achieving AHI < 5 events/hour
f Defined as achieving AHI < 5 events/hour or > 50% reduction in AHI to < 20 events/hour

Table 44: Summary of Treatment Response and Success Rates From Positional Therapy Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Benoist 201678

Positional therapy (i.e., sleep position trainers) pre versus post

Levels of response and success with positional therapy, using a sleep position trainer, were 37.5% and 31.3%, respectively, in patients with positional OSA.

33 positional OSA patients:

  • Median AHI: 18.3 (IQR: 13.7 to 24.0) events/hour
  • Mean BMI ± SD: 27.9 ± 2.8 kg/m2
  • Treatment response:a
    • 37.5% (i.e., 12 of 32 patients)
  • Treatment success:b
    • 31.3% (i.e., 10 of 32 patients)

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; IQR = interquartile range; OSA = obstructive sleep apnea; SD = standard deviation.
a Defined as > 50% reductions in AHI.
b Defined as > 50% reductions in AHI to < 5 events/hour.


Blood Pressure

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Six SRs,5,10,53,57,63,65 including one NMA,63 reported on blood pressure in adults with mild-to-severe5,10,63 or moderate-to-severe53,57,65 OSA. The SRs5,10,53,57,63,65 included overweight-to-obese5,10,53,57,63 or obese65 patients. Two SRs57,65 included patients with hypertension65 or resistant hypertension only.57,65

Four of the six SRs,10,53,63,65 with sample sizes ranging from 794 patients65 to 4,888 patients63 from seven studies65 to 51 studies,63 reported significantly greater reductions in most10,53,65 or all63 systolic blood pressure (SBP) and diastolic blood pressure (DBP) measures included (e.g., daytime, nighttime, or 24-hour) with CPAP, compared with controls. The mean difference in blood pressure, reported by all four SRs, ranged from -4.39 mm Hg53 to -1.41 mm Hg53 with 14 of the 22 mean differences in SBP and DBP measures around -2 mm Hg. Study duration of the included primary studies, reported by all four SRs, ranged from one week63 to 157 weeks.63 I2 scores, reported by three of the four applicable SRs,53,63,65 ranged from 0%53,65 to 89%65 and were greater than 75% in two SRs.57,65 The SRs reported the quality of the included studies as low to moderate10 or high53,63,65 (Appendix 10).

Two of the six SRs,5,57 with sample sizes ranging from 446 patients57 to 586 patients5 from five studies57 to seven studies,5 reported mixed findings, including significantly greater reductions in some SBP and DBP measures (e.g., 24-hour SBP and DBP57 or unspecified SBP and DBP in a subset of studies5) but not others (e.g., daytime and nighttime SBP and DBP5,57) with CPAP, compared with controls. Study duration of the included primary studies, reported by both studies, ranged from one week5 to 12 months.5 I2 scores, reported by the one applicable study,57 ranged from 0% to 87%. The SRs reported the quality of the included studies as low to moderate5 or high57 (Appendix 10).

Results of subgroup or meta-regression analyses conducted by four SRs10,57,63,65 were also mixed. The effect of CPAP versus controls on blood pressure was reported to decrease,57 increase,10 or not change65 with increasing EDS at baseline; decrease,57 increase,10,65 or not change63 with increasing OSA severity at baseline; decrease,63 increase,65 or not change57 with increasing blood pressure at baseline; decrease57 or increase10,63,65 with greater CPAP adherence; and decrease10,57,63 or increase65 with longer study durations. The effect of CPAP versus controls on blood pressure was also reported to not change with increasing BMI at baseline;57,65 be higher in patients with hypertension versus no hypertension10 and higher in patients with resistant hypertension versus hypertension;65 increase with age;10,65 and not depend on sex.65

Across the six SRs, 53 primary studies had been included, 17 of which had been included in one SR, 17 in two SRs, 14 in three SRs, four in four SRs, and one in five SRs (Appendix 16.11). No two SRs completely overlapped on blood pressure as the outcome.

The findings of the SRs are summarized in Table 45.

2) Oral appliances versus inactive controls

Two SRs,5,63 including one NMA,63 reported on blood pressure in adults with mild-to-severe63 or moderate5 OSA. The SRs5,63 included overweight5 or overweight-to-obese63 patients. Both SRs,5,63 with sample sizes ranging from 146 patients5 to 4,888 patients63 from one study5 to 51 studies,63 reported significantly greater reductions in overall63 or 24-hour5 SBP and DBP with MADs, compared with controls. The mean difference in blood pressure, reported by both SRs, ranged from -2.1 mm Hg63 to -1.5 mm Hg.5 Study duration of the included studies, reported by both SRs, ranged from one week63 to 157 weeks.63 I2 scores, reported by one of the two applicable SRs,63 ranged from 0% to 45%. The SRs reported the quality of the included studies as moderate5 or high63 (Appendix 10).

Across the two SRs, seven primary studies had been included, all of which had been included in one or the other SR, with no overlap between the two SRs (Appendix 16.12).

The findings of the SRs are summarized in Table 46.

3) Lifestyle interventions versus inactive controls

One SR5 reported on blood pressure in adults with mild OSA who were obese. The SR,5 with a sample size of 81 patients from one study, reported no significant differences in change in SBP or DBP with diet and exercise programs, compared with controls. Study duration of the included primary study was one year. I2 scores were not applicable. The SR5 reported the quality of the included studies as moderate (Appendix 10). The findings of the SR are summarized in Table 47.

4) Continuous positive airway pressure versus oral appliances

Two SRs,63,77 including one NMA,63 reported on blood pressure in adults with mild-to-severe OSA. One SR63 included overweight-to-obese patients. The other SR77 provided no information on comorbidities.

Both SRs,63,77 with sample sizes ranging from 128 patients77 to 4,888 patients63 from three studies77 to 51 studies,63 reported no significant differences in change in SBP or DBP between CPAP and MADs63 or undefined OAs,77 except the findings of one RCT that nighttime DBP was significantly lower with OAs, compared with CPAP.77 Study duration of the included primary studies, reported by both SRs, ranged from one week63 to 157 weeks.63 I2 scores, reported by the one applicable SR,63 ranged from 0% to 5%. The SRs reported the quality of the included studies as low77 or high63 (Appendix 10).

Across the two SRs, six primary studies had been included, five of which had been included in one or the other SR, and one in both SRs (Appendix 16.13). The two SRs did not completely overlap on blood pressure as the outcome.

The findings of the SRs are summarized in Table 48.

Review of Primary Studies

1) Positional therapy versus inactive controls

One study90 reported on blood pressure in obese adults with mild OSA, provided by mean AHI and BMI. The study,90 with a sample size of 86 patients, reported marginal differences in blood pressure between positional therapy (i.e., an apparatus designed to avoid sleep in the supine position) and sleep hygiene advice. Concerns with the quality of the study were assessed to be low90 (Appendix 13). The findings of the primary study are summarized in Table 49.

2) Combination therapy versus active controls

One study93 reported on blood pressure in obese adults with severe OSA, provided by mean AHI and BMI. The study,93 with a sample size of 181 patients, reported mixed findings, including significantly greater reductions in SBP with CPAP plus weight loss, compared with either intervention alone, in patients who lost 5% or more of their baseline weight and adhered to CPAP but no significant differences in SBP among the interventions in intention-to-treat analyses. CPAP plus weight loss was more effective with longer study duration (i.e., 24 versus eight weeks), CPAP was less effective with longer study duration (i.e., 24 versus eight weeks), and weight loss was more effective with longer study duration (i.e., 24 versus eight weeks).93 Concerns with the quality of the study were assessed to be unclear93 (Appendix 13). The findings of the primary study are summarized in Table 50.

Summary of Results on Blood Pressure

For blood pressure, evidence was found on inactive comparisons with CPAP, MADs, and lifestyle interventions (i.e., diet, exercise, and positional therapy). Evidence was also found on active comparisons between CPAP and OAs (i.e., MADs and undefined OAs) and combination therapy and other interventions (i.e., CPAP plus weight loss versus CPAP or weight loss alone).

Compared with inactive controls, CPAP and MADs were effective at reducing blood pressure. Effect sizes across the interventions were similar, with mean differences in blood pressure around -2 mm Hg, although it is unclear whether these results are clinically important. The similarity in effect sizes is reflected in the findings on active comparisons, where no significant differences in blood pressure were found between CPAP and MADs or undefined OAs. For diet, exercise, and positional therapy, compared with inactive controls, non-significant reductions in blood pressure were found. These findings suggest that CPAP and MADs or undefined OAs may be comparable in their effectiveness in improving blood pressure, while lifestyle interventions may not significantly improve blood pressure. Some of the findings on CPAP versus inactive controls were associated with high heterogeneity.

Combination therapy was more effective at reducing blood pressure, compared with the interventions alone, specifically for CPAP plus weight loss versus CPAP or weight loss alone, in a subset of patients achieving tangible weight loss and CPAP adherence but not in intention-to-treat analyses. These findings suggest that the various interventions in combination have additive effects in their effectiveness in improving blood pressure in cases of high adherence.

Patients with hypertension or resistant hypertension experienced greater effects with CPAP. Subgroup and meta-regression analyses suggest that older patients experienced greater effects with CPAP. Findings on the role of baseline EDS, OSA severity, and blood pressure, as well as CPAP adherence levels and study duration on the effect of CPAP were mixed and inconclusive. However, longer study duration was associated with greater effects with weight-loss programs. Sex and BMI were not significantly associated with the effects of CPAP.


Table 45: Summary of Change in Blood Pressure From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Guo 201653

CPAP versus inactive controls

CPAP was associated with significantly lowered 24-hour blood pressure (for both SBP and DBP), daytime blood pressure (for DBP only), and nighttime blood pressure (for both SBP and DBP), when compared with inactive controls.

1,171 patients from 7 RCTs (2006 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 24 to 60 events/hour)
  • Overweight to obese (mean BMI range: 28 to 40 kg/m2)
  • Diabetes (% range: 28% to 47%, where reported)
  • Smoking (% range: 12% to 84%, where reported)
  • 3 to 6 months of study duration
  • 24-hour SBP MD (95% CI) = -2.03 mm Hg (-3.64 to -0.42); P = 0.01; I2 = 0%
  • 24-hour DBP MD (95% CI) = -1.79 mm Hg (-2.89 to -0.68); P = 0.001; I2 = 0%
  • Daytime SBP MD (95% CI) = -1.41 mm Hg (-3.80 to 0.97); P = 0.25; I2 = 43%
  • Daytime DBP MD (95% CI) = -1.43 mm Hg (-2.67 to -0.19); P = 0.02; I2 = 0%
  • Nighttime SBP MD (95% CI) = -4.39 mm Hg (-6.85 to -1.93); P = 0.0005; I2 = 34%
  • Nighttime DBP MD (95% CI) = -1.64 mm Hg (-2.88 to -0.40); P = 0.009; I2 = 0%

None

High

Liu 201657

CPAP versus inactive controls

In patients with OSA and resistant hypertension, CPAP was associated with significantly lowered 24-hour blood pressure (for both SBP and DBP) but not daytime blood pressure (for both SBP and DBP) or nighttime blood pressure (for SBP only), when compared with inactive controls.

The effect of CPAP on 24-hour DBP decreased with severe OSA, sleepier patients, obese patients, longer CPAP usage at night, and longer follow-up durations but was not influenced by baseline SBP/DBP.

446 patients from 5 RCTs (2010 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 20 to 52.7 events/hour)
  • Overweight to obese (mean BMI range: 29.8 to 34.1 kg/m2)
  • Resistant hypertension (100%)
  • 3 to 8 months of study duration
  • 24-hour SBP MD (95% CI) = -4.78 mm Hg (-7.95 to -1.61); P = 0.003; I2 = 48%
  • 24-hour DBP MD (95% CI) = -2.95 mm Hg (-5.37 to -0.53); P = 0.02; I2 = 69%
  • Daytime SBP MD (95% CI) = -3.15 mm Hg (-9.20 to 2.89); P = 0.31; I2 = 87%
  • Daytime DBP MD (95% CI) = -2.51 mm Hg (-6.23 to 1.22); P = 0.19; I2 = 84%
  • Nighttime SBP MD (95% CI) = -1.89 mm Hg (-4.14 to 0.35); P = 0.10; I2 = 0%
  • Nighttime DBP MD (95% CI) = -1.53 mm Hg (-3.07 to 0.00); P = 0.05; I2 = 0%

Subgroup analysis:

  • Baseline ESS:
    • < 10: 24-hour DBP MD (95% CI) = -3.24 (-5.21 to -1.28); P = 0.001; I2 = 0%
    • ≥ 10: 24-hour DBP MD (95% CI) = -2.44 (-8.32 to 3.43); P = 0.41; I2 = 89%
  • Baseline AHI:
    • < 30 events/hour: 24-hour DBP MD (95% CI) = -5.3 (-7.29 to -3.31); P < 0.0001; I2 = 0%
    • > 30 events/hour: 24-hour DBP MD (95% CI) = -1.99 (-4.51 to -0.2); P = 0.12; I2 = 52%
  • Baseline SBP/DBP:
    • < 140/90 mm Hg: 24-hour DBP MD (95% CI) = -2.9 (-6.41 to 0.61); P = 0.11; I2 = 0%
    • ≥ 140/90 mm Hg: 24-hour DBP MD (95% CI) = -2.89 (-6.03 to 0.25); P = 0.07; I2 = 79%
  • Baseline BMI:
    • ≤ 32 kg/m2: 24-hour DBP MD (95% CI) = -4.55 (-6.73 to -2.38); P < 0.0001; I2 = 26%
    • > 32 kg/m2: 24-hour DBP MD (95% CI) = -1.51 (-5.51 to 2.5); P = 0.46; I2 = 75%
  • Follow-up duration:
    • ≤ 3 months: 24-hour DBP MD (95% CI) = -3.24 (-5.21 to -1.28); P = 0.001; I2 = 0%
    • > 3 months: 24-hour DBP MD (95% CI) = -2.44 (-8.32 to 3.43); P = 0.41; I2 = 89%
  • CPAP adherence:
    • ≤ 5 hours/night: 24-hour DBP MD (95% CI) = -4.55 (-6.73 to -2.38); P < 0.0001; I2 = 26%
    • > 5 hours/night: 24-hour DBP MD (95% CI) = -1.51 (-5.51 to 2.5); P = 0.46; I2 = 75%

High

Bratton 201563

CPAP versus inactive controls

CPAP was associated with significantly lowered blood pressure (for both SBP and DBP), when compared with inactive controls.

The association of CPAP with reductions in both SBP and DBP was greater in patients using CPAP for longer periods at night or in those with higher baseline blood pressure but less with longer follow-up durations.

Pairwise MA: 4,533 patients from 47 RCTs (1996 to 2015):

  • Mild-to-severe OSA (mean AHI range: 13 to 64 events/hour)
  • Overweight to obese (mean BMI range: 26 to 37 kg/m2)
  • Other comorbidities (e.g., CVD, HF, hypertension, resistant hypertension, and panic disorder)
  • 1 to 157 weeks of study duration

Pairwise MA:

  • Overall SBP MD (95% CI) = -2.6 mm Hg (-3.6 to -1.6); P < 0.001; I2 = 54%
  • Overall DBP MD (95% CI) = -2.1 mm Hg (-2.8 to -1.4); P < 0.001; I2 = 52%

Meta-regression:

  • For every mm Hg of increase in baseline SBP and DBP, SBP and DBP were additionally reduced by 0.2 mm Hg (95% CI = -0.3 to 0; P = 0.04) and 0.2 mm Hg (95% CI = -0.4 to 0; P = 0.01), respectively.
  • For every week of increase in the length of follow-up, SBP and DBP was additionally raised by 0.2 mm Hg (95% CI = 0.1 to 0.3; P = 0.003) and 0.1 mm Hg (95% CI = 0 to 0.2; P = 0.006), respectively.
  • For every 1-hour/night increase in CPAP use, SBP and DBP were additionally reduced by 1.5 mm Hg (95% CI = 0.8 to 2.3; P < 0.001) and 0.9 mm Hg (95% CI = 0.3 to 1.4; P = 0.001), respectively.
  • There was no significant association between OSA severity (i.e., baseline AHI or ODI) and CPAP treatment effect on SBP (P = 0.24) or DBP (P = 0.17).

High

Network MA: 4,888 patients from 51 RCTs (1996 to 2015):

  • Mild-to-severe OSA (mean AHI range: 13 to 64 events/hour)
  • Overweight to obese (mean BMI range: 26 to 37 kg/m2)
  • Other comorbidities (e.g., CVD, HF, hypertension, resistant hypertension, and panic disorder)
  • 1 to 157 weeks of study duration

Network MA:

  • Overall SBP MD (95% CI) = -2.5 mm Hg (-3.5 to -1.5); P < 0.001
  • Overall DBP MD (95% CI) = -2.0 mm Hg (-2.7 to -1.3); P < 0.001

Hu 201565

CPAP versus inactive controls

In patients with OSA and hypertension or resistant hypertension, CPAP was associated with significantly lowered 24-hour blood pressure (for both SBP and DBP), and nighttime blood pressure (for both SBP and DBP), when compared with inactive controls.

The effect of CPAP on 24-hour blood pressure (for both SBP and DBP) was higher in patients with resistant hypertension, compared with those with hypertension.

CPAP adherence, age, and baseline SBP were positively correlated with a reduction in 24-hour DBP, but not SBP, by CPAP.

CPAP adherence, age, OSA severity, study durations, and baseline SBP were positively correlated with reductions in daytime blood pressure (for both SBP and DBP) by CPAP.

794 patients from 7 RCTs (2006 to 2014):

  • Moderate-to-severe OSA (mean AHI range: 28.1 to 58.3 events/hour)
  • Obese (mean BMI range: 30.8 to 35.7 kg/m2)
  • Hypertension (100% in 3 RCTs) or resistant hypertension (100% in 4 RCTs)
  • 1 to 6 months of study duration
  • 24-hour SBP MD (95% CI) = -2.32 mm Hg (-3.65 to -1.00); P = 0.001; I2 = 0%
  • 24-hour DBP MD (95% CI) = -1.98 mm Hg (-2.82 to -1.14); P < 0.001; I2 = 21%
  • Daytime SBP MD (95% CI) = -3.58 mm Hg (-8.04 to 0.89); P = 0.12; I2 = 89%
  • Daytime DBP MD (95% CI) = -2.85 mm Hg (-5.58 to -0.12); P = 0.04; I2 = 88%
  • Nighttime SBP MD (95% CI) = -2.74 mm Hg (-4.26 to -1.23); P < 0.001; I2 = 0%
  • Nighttime DBP MD (95% CI) = -1.75 mm Hg (-2.79 to -0.71); P = 0.001; I2 = 0%

Subgroup analysis:

  • Hypertension:
    • Hypertension:
      • 24-hour SBP MD (95% CI) = -1.81 mm Hg (-3.34 to -0.29); P = 0.02; I2 = NR
      • 24-hour DBP MD (95% CI) = -1.28 mm Hg (-2.28 to -0.27); P = 0.01; I2 = NR
    • Resistant hypertension:
      • 24-hour SBP MD (95% CI) = -3.88 mm Hg (-6.55 to -1.22); P = 0.004; I2 = NR
      • 24-hour DBP MD (95% CI) = -3.65 mm Hg (-5.19 to -2.10); P < 0.001; I2 = NR

Meta-regression:

  • CPAP adherence (P = 0.016), age (P = 0.024), and baseline SBP (P = 0.036) were positively correlated with reductions in 24-hour DBP by CPAP.
  • CPAP adherence (P < 0.001), age (P < 0.05), OSA severity (P < 0.001), study duration (P < 0.001), and baseline SBP (P < 0.001) were positively correlated with reductions in daytime SBP and daytime DBP by CPAP.
  • Sex, BMI, and baseline ESS were not significantly correlated with reductions in 24-hour or daytime SBP or DBP.

High

Fava 201410

CPAP versus inactive controls

CPAP was associated with significantly lowered overall blood pressure (for both SBP and DBP), as well as daytime and nighttime blood pressure (for both SBP and DBP), when compared with inactive controls.

The effect of CPAP in lowering both SBP and DBP was significantly larger in patients with hypertension (versus no hypertension), patients with severe OSA, sleepier patients, older patients, shorter study durations, and higher CPAP adherence.

1,820 patients from 29 RCTs (1996 to 2012):

  • Mild-to-severe OSA (mean AHI range: 12.9 to 63.8 events/hour)
  • Overweight to obese (mean BMI range: 27.2 to 37.0 kg/m2)
  • Hypertension (% range: 0% to 100%, where reported)
  • 2 to 52 weeks of study duration
  • Overall SBP MD (95% CI) = -2.56 mm Hg (-3.68 to -1.44); P < 0.001; I2 = NR
  • Overall DBP MD (95% CI) = -2.00 mm Hg (-2.81 to -1.19); P < 0.001; I2 = NR
  • Daytime SBP MD (95% CI) = -2.19 mm Hg (-3.50 to 0.89); P = 0.001; I2 = NR
  • Daytime DBP MD (95% CI) = -1.89 mm Hg (-3.05 to −0.73); P = 0.001; I2 = NR
  • Nighttime SBP MD (95% CI) = -3.77 mm Hg (-5.33 to −2.21); P < 0.001; I2 = NR
  • Nighttime DBP MD (95% CI) = -1.78 mm Hg (-2.90 to −0.65); P = 0.002; I2 = NR

Subgroup analysis:

  • Hypertension:
    • Yes:
      • SBP WMD ± SE = -1.9 ± 0.7 (P < 0.01)
      • DBP WMD ± SE = -1.6 ± 0.4 (P < 0.001)
    • No:
      • SBP WMD ± SE = -1.0 ± 0.9 (NS)
      • DBP WMD ± SE = -1.3 ± 0.6 (P < 0.05)
  • Baseline ESS:
    • < 10:
      • SBP WMD ± SE = -1.3 ± 1.2 (NS)
      • DBP WMD ± SE = -1.6 ± 0.9 (NS)
    • ≥ 10:
      • SBP WMD ± SE = -2.4 ± 0.6 (P < 0.001)
      • DBP WMD ± SE = -1.9 ± 0.5 (P < 0.001)
  • Baseline AHI:
    • ≤ 30 events/hour:
      • SBP WMD ± SE = -0.5 ± 1.0 (NS)
      • DBP WMD ± SE = -0.4 ± 0.9 (NS)
    • > 30 events/hour:
      • SBP WMD ± SE = -2.6 ± 0.6 (P = 0.001)
      • DBP WMD ± SE = -2.1 ± 0.5 (P < 0.001)
  • Age:
    • < 50 years:
      • SBP WMD ± SE = -2.0 ± 0.7 (P < 0.01)
      • DBP WMD ± SE = -1.7 ± 0.5 (P < 0.001)
    • ≥ 50 years:
      • SBP WMD ± SE = -3.6 ± 1.0 (P < 0.001)
      • DBP WMD ± SE = -2.7 ± 0.8 (P < 0.001)
  • Study duration:
    • ≤ 9 weeks:
      • SBP WMD ± SE = -2.9 ± 0.8 (P < 0.001)
      • DBP WMD ± SE = -2.5 ± 0.6 (P = 0.001)
    • > 9 weeks:
      • SBP WMD ± SE = -2.3 ± 0.9 (P < 0.01)
      • DBP WMD ± SE = -1.6 ± 0.6 (P < 0.01)
  • CPAP adherence:
    • < 4 hours/night:
      • SBP WMD ± SE = -2.0 ± 1.0 (NS)
      • DBP WMD ± SE = -1.9 ± 0.7 (P < 0.05)
    • ≥ 4 hours/night:
      • SBP WMD ± SE = -2.8 ± 0.7 (P < 0.001)
      • DBP WMD ± SE = -2.0 ± 0.5 (P < 0.001)

Meta-regression:

  • For each 10 event/hour increase in baseline AHI, a 1 mm Hg decrease in SBP by CPAP was observed (P < 0.01).

Low to moderate

Balk 20115

CPAP versus inactive controls

No significant differences were found between CPAP and inactive controls in their effect on blood pressure.

586 patients from 7 RCTs (1996 to 2010):

  • Mild-to-severe OSA (mean AHI range: 13 to 65 events/hour)
  • Overweight to obese (mean BMI range: 29.4 to 36 kg/m2)
  • 1 to 12 months of study duration
  • Daytime or nighttime SBP or DBP (from 7 RCTs): no significant differences between CPAP and control in any RCT
  • No MA

None

Low to moderate

CPAP versus sham CPAP

The effect of CPAP on blood pressure was inconsistent across studies, when compared with sham CPAP.

608 patients from 12 RCTs (2001 to 2010):

  • Moderate-to-severe OSA (mean AHI range: 30 to 65 events/hour, where reported)
  • Overweight to obese (mean BMI range: 27.2 to 42.6 kg/m2)
  • 1 week to 3 months of study duration
  • SBP (from 10 RCTs): significant decreases associated with CPAP in 2 RCTs
  • DBP (from 10 RCTs): significant decreases associated with CPAP in 3 RCTs
  • No MA

None

Mixed

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; CVD = cardiovascular disease; DBP = diastolic blood pressure; ESS = Epworth Sleepiness Scale; HF = heart failure; MA = meta-analysis; MAP = mean arterial pressure; MD = mean difference; MR = meta-regression; NR = not reported; NS = not significant; OSA = obstructive sleep apnea; RCTs = randomized controlled trials; SBP = systolic blood pressure; SE = standard error; WMD = weighted mean difference.

Table 46: Summary of Change in Blood Pressure From OAs Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Bratton 201563

MADs versus inactive controls

MADs were associated with significant reductions in blood pressure (for both SBP and DBP), compared with inactive controls.

Pairwise MA: 473 patients from 6 RCTs (2004 to 2014):

  • Mild-to-severe OSA (mean AHI range: 14 to 42 events/hour)
  • Overweight to obese (mean BMI range: 27 to 31 kg/m2)
  • 4 to 13 weeks of study duration

Pairwise MA:

  • Overall SBP MD (95% CI) = -1.9 mm Hg (-3.2 to -0.6); P = 0.004; I2 = 0%
  • Overall DBP MD (95% CI) = -1.1 mm Hg (-2.4 to 0.2); P = 0.11; I2 = 45%

None

High

Network MA: 4,888 patients from 51 RCTs (1996 to 2015)

  • Mild-to-severe OSA (mean AHI range: 13 to 64 events/hour)
  • Overweight to obese (mean BMI range: 26 to 37 kg/m2)
  • Other comorbidities (e.g., CVD, HF, hypertension, resistant hypertension, and panic disorder)
  • 1 to 157 weeks of study duration

Network MA:

  • Overall SBP MD (95% CI) = -2.1 mm Hg (-3.4 to -0.8); P = 0.002
  • Overall DBP MD (95% CI) = -1.9 mm Hg (-3.2 to -0.5); P = 0.008

None

Balk 20115

MADs versus sham OAs

MADs were associated with significant reductions in blood pressure (for both SBP and DBP), compared with sham OAs.

146 patients from 1 RCT (2002):

  • Moderate OSA (mean AHI: 25 events/hour)
  • Overweight (BMI: 29 kg/m2)
  • 4 weeks of study duration
  • 24-hour SBP MD (95% CI) = -1.5 (-3.0 to -0.0); P = 0.05
  • 24-hour DBP MD (95% CI) = -1.6 (-2.5 to -0.6); P = 0.001

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CVD = cardiovascular disease; DBP = diastolic blood pressure HF = heart failure; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; OA = oral appliance; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SBP = systolic blood pressure; SE = standard error.

Table 47: Summary of Change in Blood Pressure From Lifestyle Interventions Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

Weight-loss programs versus inactive controls

No significant differences were found between weight-loss programs and inactive controls for blood pressure (for both SBP and DBP).

81 patients from 1 RCT (2009):

  • Mild OSA (mean AHI = 9 events /hour)
  • Obese (mean BMI = 31.4 kg/m2)
  • 1 year of study duration
  • SBP MD (95% CI) = -0.6 (-8.4 to 7.2); P = 0.88
  • DBP MD (95% CI) = -1.5 (-7.4 to 4.4); P = 0.62

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; DBP = diastolic blood pressure; MA = meta-analysis; MD = mean difference; MR = meta-regression; OSA = obstructive sleep apnea; RCTs = randomized controlled trials; SBP = systolic blood pressure.

Table 48: Summary of Change in Blood Pressure From CPAP Versus OAs

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Bratton 201563

CPAP versus MADs

There was no significant difference between CPAP and MADs in their effect on SBP and DBP.

Pairwise MA: 370 patients from 4 RCTs (2004 to 2014):

  • Moderate-to-severe OSA (mean AHI range: 21 to 42 events/hour)
  • Overweight to obese (mean BMI range: 27 to 31 kg/m2)
  • 4 to 13 weeks of study duration

Pairwise MA:

  • Overall SBP MD (95% CI) = 0.3 mm Hg (-1.0 to 1.5); P = 0.68; I2 = 5%
  • Overall DBP MD (95% CI) = 0.2 mm Hg (-0.6 to 0.9); P = 0.68; I2 = 0%

None

High

Network MA: 4,888 patients from 51 RCTs (1996 to 2015):

  • Mild-to-severe OSA (mean AHI range: 13-64 events/hour)
  • Overweight to obese (mean BMI range: 26 to 37 kg/m2)
  • Other comorbidities (e.g., CVD, HF, hypertension, resistant hypertension, and panic disorder)
  • 1 to 157 weeks of study duration

Network MA:

  • Overall SBP MD (95% CI) = -0.5 mm Hg (-2.0 to 1.0); P = 0.55
  • Overall DBP MD (95% CI) = -0.2 mm Hg (-1.6 to 1.3); P = 0.82

None

Li 201377

CPAP versus OAs

The findings on blood pressure were mixed but generally indifferent between CPAP and OA.

128 patients from 3 RCTs (2002 to 2009):

  • Mild-to-severe OSA (AHI threshold: ≥ 5 events/hour)
  • Comorbidities: NR
  • 8 to 10 weeks of study duration
  • 24-hour SBP and DBP (from 1 RCT): no significant difference between CPAP and OAs
  • Nighttime DBP (from 1 RCT): significantly lower with OAs, compared with CPAP (P < 0.05)
  • Undefined blood pressure (from 2 RCTs): no significant difference between CPAP and OAs
  • No MA

None

Low

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; CVD = cardiovascular disease; DBP = diastolic blood pressure; HF = heart failure; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NR = not reported; OA = oral appliances; OSA = obstructive sleep apnea; RCTs = randomized controlled trials; SBP = systolic blood pressure.

Table 49: Summary of Change in Blood Pressure From Positional Therapy Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Jackson 201590

Positional therapy (i.e., sleep position modification devices) versus inactive control

Marginal differences were found in blood pressures in the active group compared with the control group.

47 (in the intervention group) or 39 (in the control group) positional OSA patients:

  • Mean AHI ± SD: 20.1 ± 8.8 (in the intervention group) or 21.8 ± 10.1 (in the control group) events/hour
  • Mean BMI ± SD: 30.0 ± 5.3 (in the intervention group) or 30.9 ± 7.7 (in the control group) kg/m2
  • Systolic BP ± SD:
    • Before:
      • Sleep position modification device: 128.3 ± 15.5 mm Hg
      • Sleep hygiene advice: 127.6 ± 15.4 mm Hg
    • After 4 weeks:
      • Sleep position modification device: 125.7 ± 9.6 mm Hg
      • Sleep hygiene advice: 133.4 ± 15.2 mm Hg
    • Difference:
      • Sleep position modification device: P = non-significant
      • Sleep hygiene advice: P = non-significant
  • Diastolic BP ± SD:
    • Before:
      • Sleep position modification device: 77.6 ± 9.2 mm Hg
      • Sleep hygiene advice: 78.1 ± 10.6 mm Hg
    • After 4 weeks:
      • Sleep position modification device: 75.1 ± 9.2 mm Hg
      • Sleep hygiene advice: 79.4 ± 9.8 mm Hg
    • Difference:
      • Sleep position modification device: P < 0.05
      • Sleep hygiene advice: P = non-significant

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; BP = blood pressure; OSA = obstructive sleep apnea; SD = standard deviation.

Table 50: Summary of Change in Blood Pressure From Combination Therapy Versus Active Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Chirinos 201493

CPAP plus weight-loss programs versus CPAP or weight-loss programs alone

A combination of CPAP and weight loss may result in slight but significant reductions in blood pressure than either intervention alone.

62 (CPAP plus weight loss), 58 (CPAP), or 61 (weight loss) patients:

  • Mean AHI ± SD: 47.1 ± 26.86 (CPAP plus weight loss), 41.2 ± 20.96 (CPAP), or 39.7 ± 20.3 (weight loss) events/hour
  • Mean BMI ± SD: 38.4 ± 6.4 (CPAP plus weight loss), 39.8 ± 7.1 (CPAP), or 38.1 ± 5.8 (weight loss) kg/m2

CPAP plus weight-loss patients:

  • Per cent change in systolic BP (95% CI):
    • CPAP plus weight loss:
      • Modified ITTa at week 8: -5.4 (-8.7 to -2.1) mm Hg
      • Modified ITTa at week 24: -7.8 (-11.4 to -4.3) mm Hg
      • Adherenceb at week 8: -2.5 (-7.1 to 2.1) mm Hg
      • Adherenceb at week 24: -14.1 (-18.7 to -9.5) mm Hg
    • CPAP:
      • Modified ITTa at week 8: -6.6 (-9.8 to -3.4) mm Hg
      • Modified ITTa at week 24: -4.2 (-7.7 to -0.6) mm Hg
      • Adherenceb at week 8: -5.6 (-9 to -2.1) mm Hg
      • Adherenceb at week 24: -3 (-6.5 to 0.5) mm Hg
      • Difference between CPAP and CPAP plus weight loss: P < 0.001
    • Weight loss:
      • Modified ITTa at week 8: -4.5 (-7.9 to -1.1) mm Hg
      • Modified ITTa at week 24: -5.1 (-8.7 to -1.4) mm Hg
      • Adherenceb at week 8: -3.5 (-7.9 to 0.8) mm Hg
      • Adherenceb at week 24: -6.8 (-10.8 to --2.7) mm Hg
      • Difference between weight loss and CPAP plus weight loss: P = 0.02

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; BP = blood pressure; CPAP = continuous positive airway pressure; ITT = intention-to-treat; SD = standard deviation.
a Defined as all patients who were randomly assigned to a study group, with ≥ 1 observation after being randomized.
b Included participants who met minimum requirements for weight loss (i.e., ≥ 5% of baseline weight) and adhered to CPAP therapy (i.e., on average, used ≥ 4 hours/night on at least 70% of the total number of nights).


Diabetes

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Two SRs5,64 reported on diabetes in adults with diabetes and moderate5 or unknown severity64 OSA. Both SRs5,64 included obese patients.

Both SRs,5,64 with sample sizes ranging from 13 patients5 to 128 patients64 from one study5 to six studies,64 reported no significant differences in change in A1C with CPAP, compared with controls5 or pre-treatment.64 Study duration of the included primary studies, reported by both SRs, ranged from four weeks5 to four months.64 The I2 score, reported by the one applicable SR,64 was 0%. One of the two SRs,64 with a sample size of 128 patients from six studies, reported a significant improvement in insulin sensitivity with CPAP, compared with pre-treatment, with a mean difference of 0.33. Study duration of the included primary studies ranged from one month to four months. The I2 score was 86.4%. The SRs reported the quality of the included studies as low64 or low to moderate5 (Appendix 10).

From subgroup analyses, one SR64 reported that the effect of CPAP versus pre-treatment on A1C did not vary with CPAP adherence levels.

Across the two SRs, seven primary studies had been included, all of which had been included in one or the other SR, with no overlap between the two SRs (Appendix 16.14).

The findings of the SRs are summarized in Table 51.

2) Lifestyle interventions versus inactive controls

One SR5 reported on diabetes in adults with diabetes and moderate OSA who were obese. The SR,5 with a sample size of 264 patients from one study, reported significantly greater reductions in A1C with diet and exercise programs, compared with controls, with a mean difference of -0.5. Study duration was one year. I2 scores were not applicable. The SR5 reported the quality of the included studies as moderate (Appendix 10). The findings of the SR are summarized in Table 52.

Review of Primary Studies

No primary studies on any of the comparisons were found that reported on diabetes.

Summary of Results on Diabetic Outcomes

For diabetic outcomes, evidence was found on inactive comparisons with CPAP and lifestyle interventions (i.e., diet and exercise programs). No evidence was found on active comparisons. Compared with inactive controls, diet and exercise, but not CPAP, were effective at reducing A1C levels in patients with diabetes, although it is unclear if these results are clinically important. However, the studies on CPAP were shorter in duration (i.e., up to four months), compared with the study on diet and exercise (i.e., one year). Compared with pre-treatment, CPAP was effective at improving insulin sensitivity in patients with diabetes. However, this finding was associated with high heterogeneity. Subgroup analyses suggest that CPAP adherence levels were not significantly associated with the effects of CPAP. No subgroup or meta-regression analyses were found on comorbidities, baseline EDS or OSA severity, sex, age, BMI, or study duration.


Table 51: Summary of Change in Diabetic Outcomes From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Feng 201564

CPAP pre versus post

CPAP had no significant effect on A1C but significantly improved insulin sensitivity, when compared with pre-treatment.

There was no significant difference in change in A1C with change in CPAP adherence levels.

128 patients from 2 RCTs and 4 observational studies (1994 to 2012):

  • OSA severity: NR
  • Obese (mean BMI range: 33.6 to 42.7 kg/m2)
  • Diabetes (100%)
  • 1 to 4 months of study duration
  • A1C (from 2 RCTs and 4 observational studies) MD (95% CI) = -0.07 (-0.25 to 0.10); P = 0.42; I2 = 0%
  • Insulin sensitivitya (from 1 RCT and 2 observational studies) MD (95% CI) = 0.33 (0.001 to 0.66); P = 0.05; I2 = 86.4%

Subgroup analysis:

  • CPAP adherence:
    • ≤ 5 hours/night: A1C MD (95% CI) = -0.12 (-0.41 to 0.18); P = 0.44; I2 = 0%
    • > 5 hours/night: A1C MD (95% CI) = -0.06 (-0.28 to 0.18); P = 0.64; I2 = 0%

Low

Balk 20115

CPAP versus inactive controls

No significant difference was found between CPAP and inactive controls in their effect on A1C.

13 patients from 1 RCT (2009):

  • Moderate OSA (mean AHI: 28 events/hour)
  • Obese (mean BMI: 31.1 kg/m2)
  • 4 weeks of study duration
  • A1C MD (95% CI) = 0.04 (-0.27 to 0.34)

None

Low to moderate

A1C = glycated hemoglobin; AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; = MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; OR = odds ratio; OSA = obstructive sleep apnea; RCT = randomized controlled trial.
a Measured by the hyperinsulinemic-euglycemic clamp.

Table 52: Summary of Change in Diabetic Outcomes From Lifestyle Interventions Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

Weight-loss programs versus inactive controls

Intensive weight-loss programs were associated with a significant reduction in A1C in patients with type 2 diabetes, when compared with inactive controls.

264 patients from 1 RCT (2009):

  • Moderate OSA (mean AHI = 24 events /hour)
  • Obese (mean BMI = 36.7 kg/m2)
  • Diabetes (100%)
  • 1 year of study duration
  • A1C MD (95% CI) = -0.5 (-0.8 to -0.2); P < 0.001

None

Moderate

A1C = glycated hemoglobin; AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; = MA = meta-analysis; MD = mean difference; MR = meta-regression; OSA = obstructive sleep apnea; RCT = randomized controlled trial.


Cardiovascular Events

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Four SRs53,56,67,69 reported on cardiovascular events (CVEs) in adults with mild-to-severe56,67 or moderate-to-severe53,69 OSA. Two SRs53,67 included overweight-to-obese patients. Two SRs56,67 included patients with previous cardiovascular disease (CVD)56 or atrial fibrillation (AF).67

Three of the four SRs,56,67,69 with sample sizes ranging from 1,247 patients67 to 35,323 patients56 from five studies56 to eight studies,67 reported significantly reduced risk of CVEs,56,69 cardiac disease (including recurrent),56 or recurrent AF,67 with CPAP, compared with no treatment56 or no CPAP.67,69 The relative risk of the events, reported by all three SRs, ranged from 0.4656 to 0.57.69 Study duration of the included primary studies, reported by one of the three SRs,56 ranged from 49 months to 132 months. I2 scores, reported by all three SRs, ranged from 0%56,67 to 42.4%.69 The SRs reported the quality of the included studies as moderate,67 high,69 or high to mixed56 (Appendix 10).

Two of the four SRs,53,56 with sample sizes ranging from 2,669 patients53 to 35,323 patients56 from five studies56 to six studies,53 reported no significant differences in the risk of major adverse cardiac events,53 hypertension and CVEs,56 or myocardial infarction (MI),56 with CPAP, compared with controls53 or no treatment.56 Study duration of the included studies, reported by both SRs, ranged from three months53 to 132 months.56 The I2 score, reported by the one applicable SR,53 was 0%. The SRs reported the quality of the included studies as high53 or high to mixed56 (Appendix 10).

From subgroup or meta-regression analyses, three SRs53,56,67 reported that the effect of CPAP versus controls on reducing the risk of CVEs was greater with higher CPAP adherence,56 older age,67 lower BMI,67 and female patients67 but did not vary with study durations53 or having hypertension or diabetes.67

Across the four SRs, 21 primary studies had been included, 18 of which had been included in one SR, two in two SRs, and one in three SRs (Appendix 16.15). No two SRs completely overlapped on CVEs as the outcome.

The findings of the SRs are summarized in Table 53.

Review of Primary Studies

No primary studies on any of the comparisons were found that reported on CVEs.

Summary of Results on Cardiovascular Events

For CVEs, evidence was found on inactive comparisons with CPAP. No evidence was found on active comparisons. Compared with inactive controls, CPAP was effective at reducing the risk of CVEs, cardiac disease (including recurrent), and recurrent AF, but not that of major adverse cardiac events, hypertension and CVEs, or MI. Subgroup and meta-regression analyses suggest that patients who were male, older, with lower BMI, or with higher CPAP adherence experienced greater effects with CPAP. Having diabetes or hypertension (versus not having those conditions) and study duration were not significantly associated with the effects of CPAP. No subgroup or meta-regression analyses were found on baseline EDS or OSA severity.


Table 53: Summary of Change in Cardiovascular Events From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Guo 201653

CPAP versus inactive controls

CPAP did not significantly reduce the risk of CVEs, when compared with inactive controls.

There was no significant difference in the risk of CVEs between short and long-term follow-up.

2,669 patients from 6 RCTs (2012 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 24 to 42 events/hour)
  • Overweight to obese (mean BMI range: 28 to 40 kg/m2)
  • Diabetes (% range: 28% to 47%, where reported)
  • Smoking (% range: 26% to 84%, where reported)
  • 3 to 60 months of study duration
  • MACEsa OR (95% CI) = 0.84 (0.62 to 1.13); P = 0.25; I2 = 0%

Subgroup analysis:

  • Study duration:
    • < 12 months: MACE OR (95% CI) = 0.68 (0.26 to 1.79); I2 = 29%
    • ≥ 12 months: MACE OR (95% CI) = 0.80 (0.55 to 1.16); I2 = 0%

High

Kim 201656

CPAP versus no treatment

The effect of CPAP on the incidence of CVEs varied across studies and across different sets of outcomes, when compared with no treatment.

There was evidence that greater CPAP adherence might improve the effect of CPAP on reducing the risk of hypertension and CVEs.

35,323 patients from 1 RCT, 3 cohort studies, and 1 administrative database study (2005 to 2014):

  • Mild-to-severe OSA (AHI threshold range: ≥ 5 to ≥ 20 events/hour)
  • Previous CVD (excluded from the RCT and one cohort study but included in the other studies)
  • 48 to 132 months of study duration
  • Hypertension and CVEsb (from 1 RCT) adjusted IDR (95% CI) = 0.81 (0.61 to 1.06)
  • CVEsb (from 3 cohort studies) RR (95% CI) = 0.46 (0.35 to 0.61); P < 0.00001; I2 = 0%
  • Cardiac diseaseb (from 3 cohort studies) RR (95% CI) = 0.54 (0.38 to 0.75); P = 0.0003; I2 = 0%
  • MI (from 1 administrative database study) adjusted IRR (95%) = 0.99 (0.85 to 1.15)

Subgroup analysis:

  • CPAP adherence:
    • ≥ 4 hours/night (from 1 RCT): hypertension and CVEs adjusted IDR (95% CI) = 0.69 (0.50 to 0.94)

High (for the RCT) or mixed (for the non-RCTs)

Qureshi 201567

CPAP versus no CPAP

CPAP was associated with significantly reduced risk of recurrence of AF in patients with OSA and AF, when compared with no CPAP.

The treatment effect of CPAP may be higher for older, slimmer, and female patients.

Having hypertension or diabetes was not significantly associated with the risk of recurrence of AF by CPAP.

1,247 patients from 1 RCT and 7 cohorts (2003 to 2013):

  • Mild-to-severe OSA (% severe OSA range: 10% to 100%)
  • Overweight to obese (mean BMI range: 25 to 35 kg/m2)
  • 100% AF
  • Hypertension (% range: 24% to 70%)
  • Diabetes (% range: 15% to 24%, where reported)
  • Study duration: NR
  • CVEsc RR (95% CI) = 0.56 (0.47 to 0.68); P < 0.001; I2 = 0%

Meta-regression:

  • In univariate analysis, decreasing age (slope = -0.06; P = 0.03), male sex (slope = 0.07; P = 0.01), and increasing BMI (slope = 0.07; P = 0.04) were significantly correlated with increasing risk of recurrence of AF by CPAP. Having hypertension or diabetes was not significantly associated with the risk of recurrence of AF by CPAP.
  • In multivariate analysis, none of the above associations were significant.

Moderate

Wang 201569

CPAP versus no CPAP

CPAP was associated with significantly reduced risk of non-fatal CVEs, compared with no CPAP.

2,255 patients from 6 studies (2005 to 2015):

  • Moderate-to-severe OSA (in 4 studies) or unselected OSA (in 2 studies)
  • Comorbidities: NR
  • Study duration: NR
  • CVEsd OR (95% CI) = 0.57 (0.43 to 0.75); P < 0.0001; I2 = 42.4%

None

Moderate

AF = atrial fibrillation; AHI = Apnea-Hypopnea Index; BMI = body mass index; CABG = coronary artery bypass grafting; CI = confidence interval; CPAP = continuous positive airway pressure; CVD = cardiovascular disease; CVE = cardiovascular event; IDR = incidence density ratio; IRR = incidence rate ratio; MA = meta-analysis; MACE = major adverse cardiac event; MI = myocardial infarction; MR = meta-regression; NR = not reported; OR = odds ratio; OSA = obstructive sleep apnea; PCI = percutaneous coronary intervention; RCT = randomized controlled trial; RR = relative risk.
a Including mortality, non-fatal MI, HF, and non-fatal stroke.
b Included events not specified.
c Including recurrence of AF.
d Non-fatal CVEs, including MI, stroke, CABG, and PCI.


Cerebrovascular Events

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Two SRs53,56 reported on cerebrovascular events (CBEs) in adults with mild-to-severe56 or moderate-to-severe53 OSA. One SR53 included overweight-to-obese patients. The other SR56 included patients with previous CVD.56

One of the two SRs,56 with a sample size of 34,600 patients from four studies, reported significantly reduced risk of stroke (including recurrent), but not of ischemic stroke (IS), with CPAP, compared with controls, with the relative risk of stroke reported as 0.27. Study duration of the included primary studies ranged from 72 months to 132 months. The I2 score was 0%. The SR56 reported the quality of the included studies as high to mixed (Appendix 10).

The other SR,53 with a sample size of 922 patients from four studies, reported no significant differences in the risk of stroke with CPAP, compared with controls. Study duration of the included primary studies ranged from three months to 60 months. The I2 score was 12%. The SR56 reported the quality of the included studies as high (Appendix 10).

From subgroup analyses, one SR53 reported that the effect of CPAP versus controls on the risk of stroke did not vary with study durations.

Across the two SRs, eight primary studies had been included, all of which had been included in one or the other SR, with no overlap between the two SRs (Appendix 16.16).

The findings of the SRs are summarized in Table 54.

Review of Primary Studies

No primary studies on any of the comparisons were found that reported on CBEs.

Summary of Results on Cerebrovascular Events

For CBEs, evidence was found on inactive comparisons with CPAP. No evidence was found on active comparisons. Compared with inactive controls, CPAP was effective at reducing the risk of stroke in patients with previous CVD but not that of stroke in other patients or IS in patients with previous CVD. Subgroup analyses suggest that study duration was not significantly associated with the effects of CPAP. No subgroup or meta-regression analyses were found on baseline EDS or OSA severity, sex, age, BMI, or adherence.


Table 54: Summary of Change in Cerebrovascular Events From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Guo 201653

CPAP versus inactive controls

CPAP did not significantly reduce the risk of stroke, when compared with inactive controls.

There was no significant difference in the risk of stroke between short and long-term follow-up.

922 patients from 4 RCTs (2012 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 24 to 42 events/hour)
  • Overweight to obese (mean BMI range: 28 to 35 kg/m2)
  • Diabetes (% range: 33% to 47%, where reported)
  • Smoking (% range: 26% to 66%, where reported)
  • 3 to 60 months of study duration
  • Stroke OR (95% CI) = 0.56 (0.18 to 1.73); P = 0.32; I2 = 12%

Subgroup analysis:

  • Study duration:
    • < 12 months: stroke OR (95% CI) = 0.36 (0.01 to 8.84); I2 = NR
    • ≥ 12 months: stroke OR (95% CI) = 0.56 (0.18 to 1.73); I2 = 11.8%

High

Kim 201656

CPAP versus no treatment

The effect of CPAP on the risk of stroke was inconsistent across studies and across different outcomes, when compared with no treatment.

34,600 patients from 3 cohort studies and 1 administrative database study (2005 to 2014):

  • Mild-to-severe OSA (AHI threshold range: ≥ 5 to ≥ 15 events/hour)
  • Previous CVD (excluded from one cohort study but included in the other studies)
  • 72 to 132 months of study duration
  • Stroke (from 3 cohort studies) RR (95% CI) = 0.27 (0.14 to 0.53); P < 0.0001; I2 = 0%
  • IS (from 1 administrative database study) adjusted IRR (95%) = 0.99 (0.82 to 1.19)

None

High (for the RCT) or mixed (for the non-RCTs)

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; CVD = cardiovascular disease; IRR = incidence rate ratio; IS = ischemic stroke; MA = meta-analysis; MR = meta-regression; NR = not reported; OR = odds ratio; OSA = obstructive sleep apnea; RCT = randomized controlled trial; RR = relative risk.


Accidents

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

One SR47 reported on MVAs in adults with severe OSA who were overweight to obese. The SR,47 with a sample size of 1,221 patients from 10 studies, reported significant reductions in the risk of real and near-miss accidents with CPAP, compared with pre-treatment, with the odds ratios (ORs) of the accidents reported as 0.21 and 0.09, respectively. Study duration of the included primary studies ranged from six months to 36 months. I2 scores ranged from 26% to 85%. The SR47 reported the quality of the included studies as low (Appendix 10). The findings of the SR are summarized in Table 55.

Review of Primary Studies

No primary studies on any of the comparisons were found that reported on accidents.

Summary of Results on Accidents

For accidents, evidence was found on inactive comparisons with CPAP. No evidence was found on active comparisons. Compared with pre-treatment, CPAP was effective at reducing the risk of real and near-miss accidents. However, the findings on CPAP were from uncontrolled studies of low quality and associated with high heterogeneity.47 No subgroup or meta-regression analyses were found.


Table 55: Summary of Road Traffic Accidents From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Antonopoulos 201147

CPAP pre versus post

CPAP was associated with significant reductions in the risk of real and near-miss road traffic accidents, when compared with pre-treatment.

1,221 patients from 10 studies (1992 to 2007):

  • Severe OSA (mean AHI range: 37.9 to 60 events/hour, where reported)
  • Overweight to obese (mean BMI range: 29.5 to 35.5 kg/m2, where reported)
  • 6 to 36 months of study duration
  • Real accidents (from 10 studies):
    • OR (95% CI) = 0.21 (0.12 to 0.35); I2 = 48%
    • IRR (95% CI) = 0.45 (0.34 to 0.59); I2 = 26%
    • RD (95% CI) = -0.22 (-0.32 to -0.13)
    • NNT (95% CI) = 5 patients (3 to 8)
  • Near- miss accidents (from 5 studies):
    • OR (95% CI) = 0.09 (0.04 to 0.21); I2 = 67%
    • IRR (95% CI) = 0.23 (0.08 to 0.67); I2 = 85%
    • RD (95% CI) = -0.47 (- 0.69 to - 0.25)
    • NNT (95% CI) = 2 patients (1 to 4)

None

Low

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; IRR = incidence rate ratio; MA = meta-analysis; MR = meta-regression; NNT = number needed-to-treat; NR = not reported; OR = odds ratio; OSA = obstructive sleep apnea; RD = risk difference.


Cognitive Functions

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Two SRs5,66 reported on cognitive functions in adults with mild-to-severe OSA who were overweight to obese.

One of the SRs5 with sample sizes ranging from 291 patients to 382 patients from seven studies to eight studies, reported significantly greater increases in various neurocognitive and psychological outcomes5 with CPAP, compared with controls, with no effect sizes reported. Study duration of the included primary studies ranged from 11 days to 12 months. I2 scores were not applicable. The SR reported the quality of the included studies as low to moderate5 (Appendix 10).

The other SR,66 with a sample size of 1,744 patients from 13 studies, reported significantly greater increases in one domain of cognitive function (i.e., vigilance), but not others (i.e., attention, processing speed, working memory, memory, verbal fluency, and visuoconstructive skills) with CPAP, compared with controls. The standardized mean difference in vigilance was -0.12. Study duration of the included primary studies ranged from one week to 24 weeks. I2 scores ranged from 0% to 85%. The SR reported the quality of the included studies as high66 (Appendix 10).

Across the two SRs, 20 primary studies had been included, 12 of which had been included in one SR, and eight in two SRs (Appendix 16.17). The two SRs did not completely overlap on cognitive functions as the outcome.

The findings of the SRs are summarized in Table 56.

2) Oral appliances versus inactive controls

One SR5 reported on cognitive functions in adults with moderate OSA who were overweight. The SR,5 with 146 patients from one study, reported significantly greater increases in speed and vigilance with MADs, compared with controls, with no effect sizes reported. Study duration of the included primary study was four weeks. I2 scores were not applicable. The SR5 reported the quality of the included studies as moderate (Appendix 10). The findings of the SR are summarized in Table 57.

3) Continuous positive airway pressure versus oral appliances

Two SRs5,77 reported on cognitive functions in adults with mild-to-severe77 or severe5 OSA. One SR5 included overweight-to-obese patients. The other SR77 provided no information on comorbidities.

Both SRs,5,77 with sample sizes ranging from 76 patients5 to 221 patients77 from two studies5 to three studies,77 reported no significant differences in change in cognitive functions, including intelligence quotient, executive function, and processing speed, between CPAP and MADs5 or undefined OAs.77 Study duration of the included primary studies, reported by both SRs, ranged from eight weeks5,77 to 12 weeks.77 I2 scores were not applicable. The SRs reported the quality of the included studies as low77 or low to moderate5 (Appendix 10).

Across the two SRs, three primary studies had been included, one of which had been included in one or the other SR, and two in both SRs (Appendix 16.18). One SR77 included all primary studies included in another SR5 on cognitive functions as the outcome.

The findings of the SRs are summarized in Table 58.

4) Continuous positive airway pressure versus lifestyle interventions

One SR5 reported on cognitive functions in adults with moderate OSA who were obese. The SR,5 with a sample size of 26 patients from one study, reported no significant differences in change in cognitive functions, including executive function and processing speed, with CPAP, compared with positional therapy. Study duration of the included primary study was two weeks. I2 scores were not applicable. The SR5 reported the quality of the included studies as moderate (Appendix 10). The findings of the SR are summarized in Table 59.

Review of Primary Studies

No primary studies on any of the comparisons were found that reported on cognitive functions.

Summary of Results on Cognitive Functions

For cognitive functions, evidence was found on inactive comparisons with CPAP and OAs (i.e., MADs and undefined OAs). Evidence was also found on active comparisons between CPAP and MADs or undefined OAs and between CPAP and positional therapy. Findings on CPAP were mixed, where one SR reported improved cognitive functions with CPAP, compared with inactive controls, while another SR reported improvement in one of seven domains. Further, it is unclear if these results are clinically important, and some of the findings were associated with high heterogeneity. Compared with inactive controls, MADs were effective at improving speed and vigilance. No significant differences in cognitive functions were found between CPAP and MADs or undefined OAs and between CPAP and positional therapy. However, the findings on CPAP versus positional therapy were based on one study of 26 patients that was two weeks in duration. No subgroup or meta-regression analyses were found.


Table 56: Summary of Change in Cognitive Functions From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Pan 201566

CPAP versus inactive controls

Out of seven domains of cognitive function, only vigilance was found significantly improved by CPAP, compared with inactive controls.

1,744 patients from 13 RCTs (1994 to 2012):

  • Mild-to-severe OSA (mean AHI range: 10.0 to 55.4 events/hour)
  • Overweight to obese (mean BMI range: 27.8 to 37.1 kg/m2)
  • 1 to 24 weeks of study duration
  • Attention SMD (95% CI) = -0.10 (-0.27 to 0.07); P = 0.24; I2 = 85%
  • Vigilance SMD (95% CI) = - 0.12 (-0.23 to -0.01); P = 0.04; I2 = 11%
  • Processing speed SMD (95% CI) = -0.08 (-0.20 to 0.03); P = 0.16; I2 = 20%
  • Working memory SMD (95% CI) = 0.00 (-0.14 to 0.15); P = 0.95; I2 = 70%
  • Memory SMD (95% CI) = - 0.04 (-0.11 to 0.04); P = 0.30; I2 = 10%
  • Verbal fluency SMD (95% CI) = -0.06 (-0.19 to 0.07); P = 0.34; I2 = 12%
  • Visuoconstructive skills SMD (95% CI) = -0.01 (−0.15 to 0.14); P = 0.92; I2 = 0%

None

High

Balk 20115

CPAP versus inactive controls

Four RCTs reported significant increases in some neurocognitive and psychological functions with CPAP, compared with inactive controls.

382 patients from 8 RCTs (1994 to 2004):

  • Mild-to-severe OSA (mean AHI range: 10 to 43 events/hour)
  • Overweight to obese (mean BMI range: 29.4 to 33 kg/m2)
  • 4 weeks to 12 months of study duration
  • Wide variety of neurocognitive and psychological outcomes (from 8 RCTs): significant increases in cognitive performance, executive function, processing speed, and semantic fluency with CPAP, compared with control, in 4 RCTs
  • No MA

None

Low to moderate

CPAP versus sham CPAP

One RCT reported significant increases in some neurocognitive and psychological functions with CPAP, compared with sham CPAP.

291 patients from 7 RCTs (1999 to 2007):

  • Moderate-to-severe OSA (mean AHI range: 22-68 events/hour)
  • Overweight to obese (mean BMI range: 29 to 42.6 kg/m2)
  • 11 days to 6 weeks of study duration
  • Wide variety of neurocognitive and psychological outcomes (from 7 RCTs): significant increases in digital vigilance with CPAP, compared with sham CPAP, in 1 RCT
  • No MA

None

Mixed

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; MA = meta-analysis; MR = meta-regression; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SMD = standardized mean difference.

Table 57: Summary of Change in Cognitive Functions From OAs Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates from MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

MADs versus sham OAs

MADs were associated with significant improvements in speed and vigilance on the neuropsychological test, compared with sham OAs.

146 patients from 1 RCT (2002):

  • Moderate OSA (mean AHI: 25 events/hour)
  • Overweight (BMI: 29 kg/m2)
  • 4 weeks of study duration
  • Speed and vigilance:a significant improvements in MADs compared with sham OAs (P < 0.001)
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; MA = meta-analysis; MAD = mandibular advancement device; MR = meta-regression; OA = oral appliance; OSA = obstructive sleep apnea; RCT = randomized controlled trial.
a Measured by a choice reaction time task.

Table 58: Summary of Change in Cognitive Functions From CPAP Versus OAs

Study

Patient Characteristics

Pooled Estimates from MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Li 201377

CPAP versus OAs

There was no significant difference between CPAP and OAs for cognitive functions.

221 patients from 3 RCTs (2002 to 2009):

  • Mild-to-severe OSA (AHI threshold: ≥ 5 events/hour)
  • Comorbidities: NR
  • 8 to 12 weeks of study duration
  • No significant differences in cognitive functions between CPAP and OAs, measured by the following:
    • (from 1 RCT) performance IQ decrement score, Trail-Making test B, Steer Clear Performance Test, and Paced Auditory Serial Addition Test 2s
    • (from 1 RCT) Paced Auditory Serial Addition Test 1.2
    • (from 1 RCT) Trail-Making tests A and B
  • No MA

None

Low

Balk 20115

CPAP versus MADs

There were no significant differences between CPAP and MADs for various neurocognitive tests.

76 patients from 2 RCTs (2002 and 2009):

  • Severe OSA (mean AHI range: 31 to 34 events/hour)
  • Overweight (mean BMI: 26.7 kg/m2, where reported)
  • 8 weeks to 2 months of study duration
  • No significant differences between CPAP and MADs in a range of tests of cognitive performance (IQ), executive function (Trail-Making), processing speed (Paced Auditory Addition Test), error making (Oxford sleep resistance), and driving skills (SteerClear)
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CPAP = continuous positive airway pressure; IQ = intelligence quotient; MA = meta-analysis; MAD = mandibular advancement device; MR = meta-regression; NR = not reported; OA = oral appliances; OSA = obstructive sleep apnea; RCT = randomized controlled trial.

Table 59: Summary of Change in Cognitive Functions From CPAP Versus Lifestyle Intervention

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

CPAP versus positional therapy (i.e., devices worn on the back)

There was no significant difference in cognition between CPAP and positional therapy.

26 patients from 1 RCT (1999):

  • Moderate OSA (mean AHI = 18 events /hour)
  • Obese (mean BMI = 30 kg/m2)
  • 2 weeks of study duration
  • No significant differences between CPAP and positional therapy in cognition assessed by various scalesa
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CPAP = continuous positive airway pressure; MA = meta-analysis; MR = meta-regression; OSA = obstructive sleep apnea; RCT = randomized controlled trial.
a Including Wechsler Memory Scale, Purdue Pegboard, Trail-Making Test, Symbol Digit Modalities, Consonant Trigram, or Concentration Endurance Test scores.


Psychological Functions

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Three SRs5,54,75 reported on psychological functions in adults with mild-to-severe OSA. The SRs included normal-to-obese75 or overweight-to-obese5,54 patients.

All three SRs,5,54,75 with sample sizes ranging from 169 patients54 to 1,314 patients75 from four studies54 to 19 studies,75 generally reported significantly greater reductions in depression5,54,75 and anxiety5,54 with CPAP, compared with controls5,54,75 or pre-treatment.54 The depression or anxiety Hedges' g, reported by one of the three SRs,54 ranged from -0.23 to -0.52. Study duration of the included primary studies, reported by all three SRs, ranged from one week75 to two years.54 I2 scores, reported by the two applicable SRs, ranged from 0%54 to 71.3%.75 The SRs reported the quality of the included studies as very low to low,54 moderate,5 or mixed75 (Appendix 10).

From subgroup analyses, one SR75 reported that the effect of CPAP versus controls on depressive symptoms was significantly greater in patients with depression at baseline. However, the SR75 reported that OSA severity at baseline, trial lengths, and CPAP adherence were not significant factors for the effect of CPAP on depression.

Across the three SRs, 36 primary studies had been included, 26 of which had been included in one SR, seven in two SRs, and three in three SRs (Appendix 16.19). No two SRs completely overlapped on psychological functions as the outcome.

The findings of the SRs are summarized in Table 60.

2) Oral appliances versus inactive controls

Two SRs5,75 reported on psychological functions in adults with moderate5 or moderate-to-severe75 OSA. The SRs5,75 included normal-to-obese75 and overweight-to-obese5 patients.

Both SRs,5,75 with sample sizes ranging from 146 patients5 to 418 patients75 from one study5 to five studies,75 reported mixed findings, including significant improvements in some scales (e.g., somatic items on the Beck Depression Inventory [BDI] scale in one study) but not others (e.g., Short Form [36] Health Survey [SF-36] mental health in three studies) with MADs, compared with controls. Study duration of the included primary studies, reported by both SRs, ranged from four weeks5,75 to three months.5 The I2 score, reported by the one applicable SR,75 was 0%. The SRs reported the quality of the included studies as moderate5 or mixed75 (Appendix 10).

Across the two SRs, six primary studies had been included, five of which had been included in one or the other SR, and one in both SRs (Appendix 16.20). The two SRs did not completely overlap on psychological functions as the outcome.

The findings of the SRs are summarized in Table 61.

3) Continuous positive airway pressure versus oral appliances

Two SRs54,77 reported on psychological functions in adults with mild-to-severe77 or moderate54 OSA. One SR54 included overweight-to-obese patients. The other SR77 provided no information on comorbidities.

One of the two SRs,54 with a sample size of 132 patients from two studies, reported mixed findings, including a significant improvement in anxiety but no improvement in depression with CPAP, compared with undefined OAs. Study duration of the included primary studies ranged from 60 days to three months. The I2 score was 0%. The SR54 reported the quality of the included studies as very low to low (Appendix 10).

The other SR,77 with a sample size of 151 patients from two studies, reported no significant differences in change in psychological functions between CPAP and undefined OAs. Study duration of the included studies ranged from eight weeks to 12 weeks. I2 scores were not applicable. The SR77 reported the quality of the included studies as low (Appendix 10).

Across the two SRs, four primary studies had been included, all of which had been included in one or the other SR, with no overlap between the two SRs (Appendix 16.21).

The findings of the SRs are summarized in Table 62.

3) Continuous positive airway pressure versus lifestyle interventions

One SR54 reported on psychological functions in adults with moderate OSA who were overweight. The SR,54 with a sample size of 16 patients from one study, reported significant improvements in anxiety and depression with CPAP, compared with exercise, with the depression and anxiety Hedges' g reported as -0.09 and -0.35, respectively. Study duration of the included primary study was 60 days. I2 scores were not applicable. The SR54 reported the quality of the included studies as very low to low (Appendix 10). The findings of the SR are summarized in Table 63.

Review of Primary Studies

No primary studies on any of the comparisons were found that reported on psychological functions.

Summary of Results on Psychological Functions

For psychological functions, evidence was found on inactive comparisons with CPAP and MADs. Evidence was also found on active comparisons between CPAP and undefined OAs and between CPAP and exercise. Compared with inactive controls or pre-treatment, CPAP was effective at reducing depression and anxiety severity, although it is unclear if these results are clinically important. Findings on MADs were mixed, including improvements in some scales, but not others, of psychological functions. Findings on CPAP versus undefined OAs were mixed, where one SR reported an improvement in anxiety but not in depression, while another SR reported no significant differences between the two interventions. CPAP may be more effective at improving anxiety and depression, compared with exercise. Subgroup analyses suggest that patients with baseline depression, compared with those without depression, experienced greater effects with CPAP. Baseline OSA severity, adherence, and study duration were not significantly associated with the effects of CPAP. No subgroup or meta-regression analyses were found on baseline EDS, sex, age, or BMI.


Table 60: Summary of Change in Psychological Functions From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP pre versus post

Depression and anxiety severity was significantly decreased with CPAP, compared with oral placebo and when pre- and post-treatment were compared.

There was no significant difference between CPAP and sham CPAP in their effects on depression and anxiety severity.

803 patients from 21 studies (1988 to 2013):

  • Mild-to-severe OSA (mean AHI range: 11 to 71.5 events/hour, where reported)
  • Overweight to obese (mean BMI range: 27.8 to 38 kg/m2, where reported)
  • 11 days to 2 years of study duration
  • Depression Hedges' g (95% CI) = - 0.52 (-0.65 to -0.40); P < 0.001; I2 = 4.3%

None

Very low to low

406 patients from 12 studies (1988 to 2013):

  • Moderate-to-severe OSA (mean AHI range: 21.3 to 67 events/hour, where reported)
  • Overweight to obese (mean BMI range: 27.8 to 33.1 kg/m2, where reported)
  • 11 days to 12 months of study duration
  • Anxiety Hedges' g (95% CI) = −0.41 (-0.56 to -0.26); P < 0.001; I2 = 3.6%

None

CPAP versus oral placebo

219 patients from 5 studies (1994 to 2004):

  • Mild-to-severe OSA (mean AHI range: 11 to 43 events/hour)
  • Overweight to obese (mean BMI range: 27.8 to 33 kg/m2)
  • 4 weeks-3 months of study duration
  • Depression Hedges' g (95% CI) = -0.36 (-0.52 to -0.19); P < 0.001; I2 = 0.29%
  • Anxiety Hedges' g (95% CI) = −0.23 (-0.36 to -0.09); P = 0.001; I2 = 0%

None

Very low to low

CPAP versus sham CPAP

169 patients from 4 studies (1999 to 2012):

  • Moderate-to-severe OSA (mean AHI range: 21.6 to 63.6 events/hour)
  • Overweight to obese (mean BMI range: 29.2 to 33.1 kg/m2, where reported)
  • 4 weeks to 3 months of study duration
  • Depression Hedges' g (95% CI) = 0.05 (-0.19 to 0.29); P = 0.69; I2 = 0%
  • Anxiety Hedges' g (95% CI) = 0.07 (-0.17 to 0.32); P = 0.55; I2 = 0%

None

Very low to low

Povitz 201475

CPAP versus inactive controls

CPAP resulted in an improvement in depressive symptoms, compared with inactive controls. However, the overall effect size was small, and there was significant heterogeneity among the included studies.

The effect of CPAP on depressive symptoms was significantly larger in patients with depression at baseline. OSA severity, trial length, and CPAP adherence were not significant factors for the effect of CPAP on depression.

1,314 patients from 19 RCTs (1998 to 2013):

  • Mild-to-severe OSA (mean AHI range: 10 to 65.1 events/hour, where reported)
  • (for the 1,732 patients included in the SR) Normal-to-obese (mean BMI range: 24.7 to 42.5 kg/m2)
  • 1 to 24 weeks of study duration
  • Depressiona SMD (95% CI) = −0.31 (-0.53 to -0.10); P = 0.004; I2 = 71.3%

Subgroup analysis:

  • Baseline depression:
    • Yes: depression SMD (95% CI) = −2.00 (-2.62 to −1.39); I2 = 12%
    • No: depressiona SMD (95% CI) = −0.20 (-0.33 to −0.06); I2 = 30%
  • Baseline AHI:
    • < 30 events/hour: depressiona SMD (95% CI) = -0.46 (-0.85 to -0.06); I2 = 84%
    • ≥ 30 events/hour: depressiona SMD (95% CI) = -0.23 (-0.46 to 0.002); I2 = 48%
  • Trial length:
    • < 4 weeks: depressiona SMD (95% CI) = -0.31 (-0.73 to 0.11); I2 = 58%
    • 4 to 8 weeks: depressiona SMD (95% CI) = -0.39 (-0.81 to 0.04); I2 = 83%
    • > 8 weeks: depressiona SMD (95% CI) = -0.20 (-0.36 to -0.04); I2 = 0%
  • CPAP adherence:
    • < 4 hours/night: depressiona SMD (95% CI) = -0.47 (-0.81 to -0.13); I2 = 80%
    • ≥ 4 hours/night: depressiona SMD (95% CI) = -0.16 (-0.32 to 0.001); I2 = 0%

Mixed

Balk 20115

CPAP versus inactive controls

Four RCTs reported significant increases in some neurocognitive and psychological functions with CPAP, compared with inactive controls.

382 patients from 8 RCTs (1994 to 2004):

  • Mild-to-severe OSA (mean AHI range: 10 to 43 events/hour)
  • Overweight to obese (mean BMI range: 29.4 to 33 kg/m2)
  • 4 weeks to 12 months of study duration
  • Wide variety of neurocognitive and psychological outcomes (from 8 RCTs): significant increases in anxiety and depression scores with CPAP, compared with control, in 4 RCTs
  • No MA

None

Low to moderate

AHI = Apnea-Hypopnea Index; BDI = Beck depression index; BMI = body mass index; BSI = Brief Symptom Inventory; CI = confidence interval; CPAP = continuous positive airway pressure; HADS-D = Hospital Anxiety and Depression Scale-Depression Subscale; MA = meta-analysis; MR = meta-regression; OSA = obstructive sleep apnea; POMS-D = Profile of Mood States-Depression Subscale; RCT = randomized controlled trial; SF-36 = Short Form (36) Health Survey; SMD = standardized mean difference; SR = systematic review.
a Measured by BDI, BSI, HADS-D, POMS-D, or SF-36.

Table 61: Summary of Change in Psychological Functions From OAs Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Povitz 201475

MADs versus inactive controls

There was a significant improvement in depressive symptoms with MADs, compared with inactive controls.

418 patients from 5 RCTs (2004 to 2008):

  • Moderate-to-severe OSA (mean AHI range: 21.3 to 34.7 events/hour)
  • (for the 1,732 patients included in the SR) Normal-to-obese (mean BMI range: 24.7 to 42.5 kg/m2)
  • 4 to 12 weeks of study duration

Depression:

  • BDI or SF-36 Mental Health (from 5 RCTs) SMD (95% CI) = -0.21 (-0.40 to -0.03); P = 0.025; I2 = 0%
  • SF-36 Mental Health (from 3 RCTs) MD (95% CI) = -2.98 (−7.78 to 1.81); P = NR; I2 = 0%
  • BDI (from 2 RCTs) MD (95% CI) = -0.80 (-1.52 to -0.08); P = NR; I2 = 0%

None

Mixed

Balk 20115

MADs versus inactive controls

There was no significant difference between MADs and inactive controls on the BDI scale.

160 patients from 1 RCT (2004)

  • Moderate OSA (mean AHI: 21 events/hour)
  • Obese (mean BMI: 31.1 kg/m2)
  • 3 months of study duration
  • BDI (from 1 RCT): no significant differences between MADs and controls

None

Moderate

MADs versus sham OAs

MADs were associated with significant improvements in somatic items on the BDI scale, compared with sham OAs.

146 patients from 1 RCT (2002):

  • Moderate OSA (mean AHI: 25 events/hour)
  • Overweight (mean BMI: 29 kg/m2)
  • 4 weeks of study duration
  • Somatic items on the BDI scale (P < 0.05)

None

Moderate

AHI = Apnea-Hypopnea Index; BDI = Beck Depression Inventory; BMI = body mass index; CI = confidence interval; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NR = not reported; OA = oral appliance; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SF = Short Form Health Survey; SMD = standardized mean difference; SR = systematic review.

Table 62: Summary of Change in Psychological Functions From CPAP Versus OAs

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP versus OAs

There was a small effect size, favouring CPAP over OAs, for anxiety.

There was no significant difference between CPAP and OAs for depression.

132 patients from 2 RCTs (2004 and 2013):

  • Moderate OSA (mean AHI range: 21.3 to 26.2 events/hour)
  • Overweight to obese (mean BMI range: 27.8 to 31.1 kg/m2)
  • 60 days to 3 months of study duration
  • Anxietya Hedges' g (95% CI) = -0.10 (-0.28 to -0.08); P = 0.27; I2 = 0%
  • Depressiona Hedges' g (95% CI) = -0.11 (-0.29 to 0.72); P = 0.24; I2 = 0%

None

Very low to low

Li 201377

CPAP versus OAs

There was no significant difference between CPAP and OAs for anxiety and depression.

151 patients from 2 RCTs (2002 and 2008) using Hospital Anxiety and Depression Scale:

  • Mild-to-severe OSA (AHI threshold: ≥ 5 events/hour)
  • Comorbidities: NR
  • 8 to 12 weeks of study duration

Hospital Anxiety and Depression Scale (from 2 RCT):

  • No significant differences between CPAP and OAs

None

Low

AHI = Apnea-Hypopnea Index; BDI = Beck depression index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; MA = meta-analysis; MR = meta-regression; NR = not reported; OA = oral appliances; OSA = obstructive sleep apnea; POMS = profile of mood states; RCT = randomized controlled trial.
a Measured by BDI or POMS.

Table 63: Summary of Change in Psychological Functions From CPAP Versus Lifestyle Interventions

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP versus exercise programs

CPAP was marginally more effective than exercise at reducing depression and anxiety.

16 patients from 1 RCT (2013):

  • Moderate OSA (mean AHI: 26.2 events/hour)
  • Overweight (mean BMI: 27.8 kg/m2)
  • 60 days of study duration
  • POMS-depression Hedges' g (SE) = -0.09 (0.48)
  • POMS-tension and anxiety Hedges' g (SE) = -0.35 (0.48)

None

Very low to low

AHI = Apnea-Hypopnea Index; BMI = body mass index; CPAP = continuous positive airway pressure; MA = meta-analysis; MR = meta-regression; OSA = obstructive sleep apnea; POMS = profile of mood states; RCT = randomized controlled trial; SE = standard error.


Quality of Life

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Two SRs5,54 reported on QoL in adults with mild-to-severe OSA who were overweight to obese.

One of the SRs,54 with sample sizes ranging from 219 patients from five controlled studies to 365 patients from nine pre-and-post studies, reported significantly greater increases in QoL with CPAP, compared with oral placebo or pre-treatment. The QoL Hedges' g ranged from 0.44 to 0.49. Study duration of the included primary studies ranged from four weeks to eight months. I2 scores ranged from 4% to 50%. The SR54 reported the quality of the included studies as very low to low (Appendix 10).

The other SR,5 with sample sizes ranging from 347 patients to 618 patients from six studies to 11 studies, reported in general no significant differences in change in QoL between CPAP and inactive controls. Study duration of the included studies ranged from three weeks to six months. I2 scores were not applicable. The SR5 reported the quality of the included studies as low to moderate (Appendix 10).

Across the two SRs, 22 primary studies had been included, 16 of which had been included in one or the other SR, and six in both SRs (Appendix 16.22). The two SRs did not completely overlap on QoL as the outcome.

The findings of the SRs are summarized in Table 64.

2) Oral appliances versus inactive controls

Two SRs5,74 reported on QoL in adults with moderate-to-severe OSA. One SR5 included overweight-to-obese patients. The other SR5 provided no information on comorbidities.

Both SRs,5,74 with sample sizes ranging from 52 patients5 to 227 patients5 from one study5 to two studies,5,74 reported in general no significant differences in change in QoL between OAs74 or MADs,5 compared with controls. Study duration of the included primary studies, reported by both SRs, ranged from four weeks5 to three months.5,74 I2 scores, reported by the one applicable SR,74 ranged from 0% to 65%. The SRs reported the quality of the included studies as low to moderate74 or moderate5 (Appendix 10).

Across the two SRs, four primary studies had been included, three of which had been included in one or the other SR, and one in both SRs (Appendix 16.23). The two SRs did not completely overlap on QoL as the outcome.

The findings of the SRs are summarized in Table 65.

3) Continuous positive airway pressure versus oral appliances

Four SRs5,54,74,77 reported on QoL in adults with mild-to-severe77 or moderate-to-severe5,54,74 OSA. Two SRs5,54 included overweight-to-obese patients. Two SRs74,77 provided no information on comorbidities.

Three of the four SRs,5,74,77 with sample size ranging from 161 patients5 to 376 patients77 from two studies74,77 to seven studies,77 reported mixed findings, including significantly greater increases in some components of QoL74,77 or some scales of QoL,5 but not others, with CPAP, compared with OAs. Study duration of the included studies, reported by all three SRs, ranged from one month5 to 48 weeks.77 I2 scores, reported by the two applicable SRs,74,77 ranged from 30%74 to 86%77 and were greater than 75% in one SR.77 The SRs reported the quality of the included studies as very low77 or low to moderate5,74 (Appendix 10).

The other SR,54 with a sample size of 132 patients from two studies, reported no significant differences in change in QoL between CPAP and OAs. Study duration of the included primary studies ranged from 60 days to three months. The I2 score was 0%. The SR54 reported the quality of the included studies as very low to low (Appendix 10).

Across the four SRs, nine primary studies had been included, five of which had been included in one SR, one in two SRs, and three in three SRs (Appendix 16.24). Two SRs5,77 included all primary studies included in another SR74 on QoL as the outcome.

The findings of the SRs are summarized in Table 66.

4) Continuous positive airway pressure versus lifestyle interventions

Two SRs5,54 reported on QoL in adults with moderate OSA who were overweight54 or obese.5

One of the two SRs,54 with a sample size of 16 patients from one study, reported significantly greater increases in QoL with CPAP, compared with exercise,54 with the QoL Hedges' g of 1.23. Study duration of the included primary study was 60 days. I2 scores were not applicable. The SR54 reported the quality of the included studies as very low to low (Appendix 10).

The other SR,5 with a sample size of 94 patients from three studies, reported no significant differences in the effect of CPAP versus positional therapy. Study duration of the included primary studies ranged from two weeks to one month. I2 scores were not applicable. The SR5 reported the quality of the included studies as moderate (Appendix 10).

Across the two SRs, six primary studies had been included, all of which had been included in one or the other SR, with no overlap between the two SRs (Appendix 16.25).

The findings of the SRs are summarized in Table 67.

5) Combination therapy versus active controls

One SR19 reported on QoL in adults with moderate-to-severe OSA who were overweight to obese. The SR,19 with a sample size of 230 patients from two studies, reported no significant differences in the effect of CPAP plus diet programs, compared with diet programs alone. Study duration of the included primary studies ranged from three months to six months. The I2 score was 0%. The SR19 reported the quality of the included studies as mixed (Appendix 10). The findings of the SR are summarized in Table 68.

Review of Primary Studies

1) Positional therapy versus inactive controls

One study78,81-86,88,90,96,99,102,104,105,108,116,118 reported on QoL in adults with mild-to-moderate OSA. The study78,81,85,86,90,96,99,102,104,108,116,118 included positional OSA patients only who were overweight, provided by mean BMI. The study,78,81-84,86,88,90,96,99,104,105,108,116,118 with a sample size of 33 patients, reported no significant change in QoL with an apparatus designed to prevent sleep in the supine position, compared with pre-treatment. Concerns with the quality of the study78,81-84,86,88,90,96,99,104,105,108,116,118 were assessed to be low (Appendix 14). The findings of the primary study are summarized in Table 69.

Summary of Results on Quality of Life

For QoL, evidence was found on inactive comparisons with CPAP, OAs (i.e., MADs and undefined OAs), and positional therapy. Evidence was also found on active comparisons between CPAP and MADs or undefined OAs and between CPAP and lifestyle interventions (i.e., diet, exercise, and positional therapy). Findings on CPAP were mixed, where one SR reported improved QoL with CPAP, compared with inactive controls, while another SR reported no significant differences between CPAP and inactive controls, although it is unclear if these results are clinically important. No significant differences between MADs, undefined OAs, or positional therapy and inactive controls were reported. Findings on CPAP versus MADs or undefined OAs, as well as those on CPAP versus diet, exercise, and positional therapy, were mixed, and some of the findings were associated with high heterogeneity. No significant differences between CPAP plus diet programs and diet programs alone were reported. No subgroup or meta-regression analyses were found.


Table 64: Summary of Change in Quality of Life From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP pre versus post

CPAP was associated with improved QoL, when pre- and post-treatment were compared and when compared with oral placebo.

365 patients from 9 studies (1998 to 2013):

  • Moderate-to-severe OSA (mean AHI range: 21.3 to 54.5 events/hour)
  • Overweight to obese (mean BMI range: 27.8 to 34.4 kg/m2, where reported)
  • 4 weeks to 8 months of study duration
  • QoLa Hedges' g (95% CI) = 0.44 (0.24 to 0.63); P < 0.001; I2 = 4.32%
  • FOSQ Hedges' g (95% CI) = 0.51 (0.22 to 0.80); P = 0.001; I2 = 49.9%

None

Very low to low

CPAP versus oral placebo

219 patients from 5 studies (1994 to 2004):

  • Mild-to-severe OSA (mean AHI range: 11 to 43 events/hour)
  • Overweight to obese (mean BMI range: 27.8 to 33 kg/m2)
  • 4 weeks to 3 months of study duration
  • QoLa Hedges' g (95% CI) = 0.49 (0.13 to 0.86); P = 0.008; I2 = 34%

None

Very low to low

Balk 20115

CPAP versus inactive controls

In general, no significant differences were found between CPAP and inactive controls in FOSQ.

The effect of CPAP on QoL is uncertain, when compared with inactive controls.

618 patients from 11 RCTs (1994 to 2007):

  • Mild-to-severe OSA (mean AHI range: 13 to 58 events/hour)
  • Overweight to obese (mean BMI range: 27.3 to 33.5 kg/m2)
  • 4 weeks to 6 months of study duration
  • FOSQ (from 4 RCTs): no significant differences between CPAP and control in any RCT
  • QoLb (from 10 RCTs): inconsistent findings across RCTs
  • No MA

None

Low to moderate

CPAP versus sham CPAP

347 patients from 6 RCTs (2001 to 2008):

  • Moderate-to-severe OSA (mean AHI range: 22 to 57 events/hour)
  • Overweight to obese (mean BMI range: 29 to 35.2 kg/m2)
  • 3 weeks to 3 months of study duration
  • FOSQ (from 3 RCTs): no significant differences between CPAP and sham CPAP in any RCT
  • QoLc (from 6 RCTs): significant increases in SF-36 physical and mental health and SAQLI with CPAP, compared with sham CPAP, in 1 RCT
  • No MA

None

Mixed

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; FOSQ = Functional Outcomes of Sleep Questionnaire; GHQ = General Health Questionnaire; GQL = glaucoma quality of life; MA = meta-analysis; MR = meta-regression; NHP = Nottingham Health Profile; NR = not reported; OSA = obstructive sleep apnea; QoL = quality of life; RCT = randomized controlled trial; SAHS = sleep apnea/hypopnea syndrome; SAQLI = the Calgary Sleep Apnea Quality of Life Index; SF = Short Form (36) Health Survey; UMACL Energetic Arousal Score = energetic arousal score of the University of Wales Mood Adjective Checklist, WHO = World Health Organization.
a Measured by SF-36, GQL, SF-12, WHOQOL, NHP-2, GHQ-28, WHO-5, or FOSQ.
b Measured by SF-36, NHP, GHQ-28, UMACL, SAHS-related symptoms questionnaire, or SAQLI.
c Measured by SF-36 or SAQLI.

Table 65: Summary of Change in Quality of Life From OAs Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Okuno 201474

OAs versus control appliances

There was no significant difference in the change in SF-63 General Health, Mental Health, or Vitality between OAs and control appliances.

67 patients from 2 RCTs (2005 and 2008):

  • Moderate-to-severe OSA (mean AHI: 33.8 and 39.1 events/hour [for OA group] and 24 and 32.6 events/hour [for control group])
  • Comorbidities: NR
  • 1 to 3 months of study duration
  • SF-36 General Health MD (95% CI) = 1.34 (-8.16 to 10.85); P = 0.78; I2 = 0%
  • SF-36 Mental Health MD (95% CI) = −3.04 (-9.82 to 3.73); P = 0.38; I2 = 0%
  • SF-36 Vitality MD (95% CI) = 3.76 (−13.77 to 21.29); P = 0.67; I2 = 65%

None

Low to moderate

Balk 20115

MADs versus inactive controls

There was no significant difference in the overall quality of life between MADs and inactive controls.

227 patients from 2 RCTs (2004 and 2007):

  • Moderate OSA (mean AHI range: 19 to 21 events/hour)
  • Overweight to obese (mean BMI range: 27.3 to 31.1 kg/m2)
  • 10 weeks-3 months of study duration
  • SAQLI score (from 1 RCT) MD (95% CI) = 0.7 (0.6 to 0.8); P < 0.001
  • FOSQ social domain outcome (from 1 RCT): no significant differences between MAD and control
  • SAQLI-social interactions treatment-related symptoms or SF-36 mean score and physical and mental components (from 1 RCT): no significant differences between MAD and control
  • No MA

None

Moderate

MADs versus sham OAs

MADs were associated with a significant improvement in SF-36 Vitality domain but not in the other domains, when compared when sham OAs.

52 patients from 1 RCT (2008):

  • Severe OSA (mean AHI: 35 events/hour)
  • Obese (mean BMI: 31 kg/m2)
  • 4 weeks of study duration
  • SF-36:
  • Vitality: net difference = 18.7; P = 0.001
  • Other domains: no significant differences between MADs and sham OAs
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass inventory; CI = confidence interval; FOSQ = Functional Outcomes of Sleep Questionnaire; MA = meta-analysis; MD = mean difference; MR = meta-regression; NR = not reported; OA = oral appliance; OSA = obstructive sleep apnea; RCT = randomized controlled trial; SAQLI = Calgary Sleep Apnea Quality of Life Index; SF = Short Form (36) Health Survey.

Table 66: Summary of Change in Quality of Life From CPAP Versus OAs

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP versus OAs

There was no significant difference between CPAP and OAs for QoL.

132 patients from 2 RCTs (2004 and 2013):

  • Moderate OSA (mean AHI range: 21.3-26.2 events/hour)
  • Overweight to obese (mean BMI range: 27.8 to 31.1 kg/m2)
  • 60 days to 3 months of study duration

SF-36:

  • All subscales Hedges' g (95% CI) = −0.02 (-0.20 to 0.16); P = 0.82; I2 = 0%

None

Very low to low

Okuno 201474

CPAP versus OAs

CPAP was significantly more effective at improving mental health, measured by SF-36, when compared with OAs. However, there were no significant differences between CPAP and OAs at improving general health and vitality, measured by SF-36.

214 (included) or 167 (analyzed) patients from 2 RCTs (2007 and 2008):

  • Moderate-to-severe OSA (mean AHI range: 20.9 to 40.3 events/hour)
  • Comorbidities: NR
  • 8 to 10 weeks of study duration

SF-36:

  • General Health MD (95% CI) = 0.61 (−1.03 to 2.24); P = 0.47; I2 = 30%
  • Mental Health MD (95% CI) = 1.80 (0.42 to 3.17); P = 0.01; I2 = 44%
  • Vitality MD (95% CI) = 2.68 (-5.32 to 10.69); P = 0.51; I2 = 75%

None

Low to moderate

Li 201377

CPAP versus OAs

The findings on quality of life were mixed but generally indifferent between CPAP and OAs.

162 patients from 2 RCTs (2002 and 2004) using FOSQ or 376 patients from 5 RCTs (2002 to 2011) using SF-36:

  • Mild-to-severe OSA (AHI threshold: ≥5 events/hour)
  • Comorbidities: NR
  • 8-48 weeks of study duration

FOSQ:

  • MD (95% CI): 0.43 (-0.54 to 1.41); P = 0.38; I2 = 86%

SF-36:

  • Components analyzed separately (from 1 RCT): significantly higher health transition (P = 0.001) and mental components (P = 0.008) with CPAP versus OAs and no significant difference in physical component between CPAP and OAs
  • Components analyzed or reported together (from 4 RCTs): no significant differences between CPAP and OAs
  • No MA

None

Low

Balk 20115

CPAP versus MADs

The findings on QoL were inconsistent between CPAP and MADs.

161 patients from 3 RCTs (2002 to 2008) using FOSQ or 361 patients from 7 RCTs (2002 to 2009) using various QoL scales:a

  • Moderate-to-severe OSA (mean AHI range: 21 to 40 events/hour)
  • Overweight to obese (mean BMI range: 26.7 to 34.1 kg/m2)
  • 1 to 3 months of study duration

FOSQ:

  • MD (95% CI) = -0.86 (-2.49 to 0.77)

Various QoL scales:a

  • (from 5 RCTs) no significant difference between CPAP and MADs
  • (from 2 RCTs) components of SF-36 favouring CPAP versus MADs
  • (from 1 RCT) greater improvement on SALQI but more treatment-related symptoms with CPAP versus MADs
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BDI = Beck Depression Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; FOSQ = Functional Outcomes of Sleep Questionnaire; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NHP = the Nottingham Health Profile; NR = not reported; OA = oral appliances; OSA = obstructive sleep apnea; QoL = quality of life; RCT = randomized controlled trial; SAQLI = the Calgary Sleep Apnea Quality of Life Index; SF-36 = Short Form (36) Health Survey.
a Including SF-36, Hospital Anxiety and Depression Scale, BDI, SAQLI, NHP, a "General Health" measure, and the Scottish National Sleep Laboratory symptom questionnaire.

Table 67: Summary of Change in Quality of Life From CPAP Versus Lifestyle Interventions

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Gupta 201654

CPAP versus exercise programs

CPAP was substantially more effective than exercise at improving QoL.

16 patients from 1 RCT (2013):

  • Moderate OSA (mean AHI: 26.2 events/hour)
  • Overweight (mean BMI: 27.8 kg/m2)
  • 60 days of study duration
  • SF-36 Hedges' g (SE) = 1.23 (0.52)

None

Very low to low

Balk 20115

CPAP versus positional therapy (i.e., shoulder-head elevation pillows or devices worn on the back)

There was no significant difference in QoL between CPAP and positional therapy.

94 patients from 3 RCTs (1999 to 2008):

  • Moderate OSA (mean AHI range: 18 to 27 events/hour)
  • Obese (mean BMI range: 30 to 34 kg/m2)
  • 2 weeks to 1 month of study duration
  • (from 3 RCTs) no significant differences between CPAP and positional therapy in QoL assessed by various scalesa
  • (from 1 RCT) significantly higher NPH energy subscale with CPAP compared with positional therapy (difference = 1; P = 0.04)
  • No MA

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CPAP = continuous positive airway pressure; FOSQ = Functional Outcomes of Sleep Questionnaire; GHQ = General Health Questionnaire; MA = meta-analysis; MR = meta-regression; NPH = Nottingham Health Profile; OSA = obstructive sleep apnea; QoL = quality of life; RCT = randomized controlled trial; SF-36 = Short Form (36) Health Survey; UWIST = University of Wales Institute of Science and Technology.
a Including SF-36 Mental and Physical Component Summaries, FOSQ, Hospital Anxiety and Depression Scale, UWIST mood adjective checklist, and GHQ.

Table 68: Summary of Change in Quality of Life From Combination Therapy Versus Lifestyle Interventions

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Thomasouli 201319

CPAP plus diet programs versus diet programs alone

There was no significant difference in QoL between CPAP plus diet programs and diet programs alone.

230 patients from 2 RCTs (1999 and 2001):

  • Moderate-to-severe OSA (mean AHI range: 20 to 56 events/hour)
  • Overweight to obese (mean BMI range: 29 to 32 kg/m2)
  • 3 to 6 months of study duration
  • NHP MD (95% CI) = −0.93 (-5.93 to 4.06); P = NR; I2 = 0%

None

Mixed

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; MA = meta-analysis; MD = mean difference; MR = meta-regression; NHP = the Nottingham Health Profile; NR = not reported; OSA = obstructive sleep apnea; QoL = quality of life; RCT = randomized controlled trial.

Table 69: Summary of Change in Quality of Life From Positional Therapy Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Benoist 201678

Positional therapy (i.e., sleep position trainers) pre versus post

Positional therapy with a sleep position trainer did not improve quality of life in patients with positional OSA.

33 positional OSA patients:

  • Median AHI: 18.3 (IQR: 13.7 to 24.0) events/hour
  • Mean BMI ± SD: 27.9 ± 2.8 kg/m2
  • Median FOSQ score:
    • Before (n = 33): 15.8 (IQR: 10.5 to 17.0)
    • After 3 months (n = 32): 16.0 (IQR: 10.8 to 18.2)
    • Difference: P = 0.616

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; FOSQ = Functional Outcomes of Sleep Questionnaire; IQR = interquartile range; OSA = obstructive sleep apnea; SD = standard deviation.


Mortality

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Four SRs52,53,56,69 reported on mortality in adults with mild-to-severe,56 moderate-to-severe,53,69 or unknown severity52 OSA. One SR53 included overweight-to-obese patients. One SR56 included patients with previous CVD. Two SRs52,69 provided no information on comorbidities.

Three of the four SRs,52,56,69 with sample sizes ranging from 1,455 patients56 to 3,112,644 patients52 from three studies56 to 11 studies,52,69 reported significantly greater reductions in the risk of death from stroke,56 cardiac disease,56 and CVEs overall52,56,69 with CPAP, compared with controls. The risk of the events, reported by all three SRs,52,56,69 ranged from 0.37 in hazard ratio (HR),52 0.06 to 0.19 in relative risk,56 and 0.32 in OR.69 One of the SRs52 also reported significantly greater reductions in the risk of death from all causes, with an HR reported as 0.66. Study duration of the included primary studies, reported by one of the three SRs,56 ranged from 72 months to 89 months. I2 scores, reported by all three SRs,52,56,69 ranged from 0%56 to 48%.56 The SRs reported the quality of the included studies as moderate to high,52 high,69 or high to mixed56 (Appendix 10).

The other SR,53 with a sample size of 2,020 patients from four studies, reported no significant differences in the risk of death from all causes with CPAP, compared with controls. Study duration of the included primary studies ranged from six months to 60 months. The I2 score was 0%. The SR53 reported the quality of the included studies as high (Appendix 10).

From subgroup analyses, one SR53 reported no significant differences in the effect of CPAP versus controls on mortality with varying levels of follow-up durations.

Across the four SRs, 19 primary studies had been included, 11 of which had been included in one SR, six in two SRs, and two in three SRs (Appendix 16.26). No two SRs completely overlapped on mortality as the outcome.

The findings of the SRs are summarized in Table 70.

Review of Primary Studies

1) Combination therapy versus inactive controls

One study80 reported on mortality in adults with unknown severity OSA who were overweight, provided by mean BMI. The study,80 with a sample size of 28 patients, concluded that MMA plus genioglossus advancement (GA) may be a safe method for treating OSA, as no mortality was reported. Concerns with the quality of the study were assessed to be low80 (Appendix 14). The findings of the primary study are summarized in Table 71.

Summary of Results on Mortality

For mortality, evidence was found on inactive comparisons with CPAP and combination therapy (i.e., MMA plus GTA). No evidence was found on active comparisons. Compared with inactive controls, CPAP was effective at reducing death from specific causes, including stroke, cardiac disease, and CVEs, in patients with or without previous CVD. Findings on death from all causes were mixed, where one SR reported significant risk reductions with CPAP, compared with inactive controls, whereas another SR reported no differences. For MMA plus GTA, no mortality was reported, either from the surgical procedure or from OSA, although this study included a small number of patients. Subgroup analyses suggest that baseline OSA severity and study duration were not significantly associated with the effects of CPAP. No subgroup or meta-regression analyses were found on baseline EDS, sex, age, BMI, or adherence.


Table 70: Summary of Mortality Outcomes From CPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Fu 201652

CPAP versus inactive controls

All-cause mortality and cardiovascular mortality were significantly lower in CPAP-treated patients than in untreated patients.

3,112,644 patients from 11 cohort studies (2005 to 2015):

  • OSA severity: NR
  • Comorbidities: NR
  • 5 to 10.3 years of study duration
  • All-cause mortality HR (95% CI) = 0.66 (0.59 to 0.73); I2 = 12.7%
  • Cardiovascular mortality HR (95% CI) = 0.37 (0.16 to 0.54); I2 = 5.7%

None

Moderate to high

Guo 201653

CPAP versus inactive controls

There was no difference in the incidence of all-cause mortality between CPAP and inactive controls.

There was no significant difference in the incidence of mortality between short and long-term follow-up.

2,020 patients from 4 RCTs (2012 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 28 to 42 events/hour)
  • Overweight to obese (mean BMI range: 28 to 32 kg/m2)
  • Diabetes (% range: 33% to 38%, where reported)
  • Smoking (% range: 26% to 62%, where reported)
  • 6 to 60 months of study duration
  • All-cause mortality OR (95% CI) = 0.85 (0.35 to 2.06); P = 0.72; I2 = 0%

Subgroup analysis:

  • Study duration:
    • < 12 months: all-cause mortality OR (95% CI) = 0.97 (0.14 to 6.94); I2 = NR

≥ 12 months: all-cause mortality OR (95% CI) = 0.82 (0.31 to 2.22); I2 = 0%

High

Kim 201656

CPAP versus no treatment

CPAP was associated with significantly decreased mortality from stroke, cardiac diseases, or overall CVEs, when compared with no treatment.

1,455 patients from 3 cohort studies (2005 to 2012):

  • Mild-to-severe OSA (AHI threshold range: ≥ 5 to ≥ 15 events/hour)
  • Previous CVD (included in all 3 cohort studies)
  • 72 to 89 months in study duration
  • Mortality from stroke (from 2 cohort studies): RR (95% CI) = 0.06 (0.01 to 0.34); P = 0.001; I2 = 0%
  • Mortality from cardiac disease (from 2 cohort studies): RR (95% CI) = 0.19 (0.09 to 0.40); P < 0.00001; I2 = 0%
  • Mortality from overall CVEs (from 3 cohort studies): RR (95% CI) = 0.19 (0.11 to 0.34); P < 0.00001; I2 = 48%

None

High (for the RCT) or mixed (for the non-RCTs)

Wang 201569

CPAP versus no CPAP

CPAP was associated with significantly decreased mortality from CVEs, when compared with no treatment.

Baseline AHI did not seem to affect the study findings.

4,620 patients from 11 studies (2005 to 2015):

  • Moderate-to-severe OSA (in 5 studies) or unselected OSA (in 6 studies)
  • Comorbidities: NR
  • Study duration: NR
  • Mortality from CVEs:

OR (95% CI) = 0.32 (0.24 to 0.41); P < 0.0001; I2 = 19.9%

Subgroup analysis:

  • OSA severity:
    • Moderate-to-severe: mortality from CVEs: OR (95% CI) = 0.29 (0.18 to 0.47); P < 0.0001; I2 = 0%

High

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; CVD = cardiovascular disease; CVE = cardiovascular event; HR = hazard ratio; MA = meta-analysis; MR = meta-regression; NR = not reported; OR = odds ratio; OSA = obstructive sleep apnea; RCT = randomized controlled trial; RR = risk ratio.

Table 71: Summary of Mortality Outcomes From Combination Therapy Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Passeri 201680

MMA plus GTA pre versus post

MMA plus GTA is a safe method for OSA treatment in terms of mortality, but patients need to be cautioned on potential complications.

28 patients:

  • OSA severity: NR
  • Mean BMI ± SD: 29.6 ± 4.7 kg/m2
  • No incidence of mortality

None

BMI = body mass index; GTA = genial tubercle advancement; MMA = maxillomandibular advancement; NR = not reported; OSA = obstructive sleep apnea; SD = standard deviation.


Adverse Events

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

One SR5 reported on adverse events associated with CPAP use, with no information on OSA severity or comorbidities. The adverse events reported by the SR,5 with a sample size of 368 patients from six studies, included claustrophobia, epistaxis, excessive nasal dryness, excessive pressure, excessive salivation, gums and lip problems, and pressure intolerance. The authors reported that in general, 5% to 15% of patients reported specific adverse events they considered to be a major problem while using CPAP; however, no study reported a severe adverse event that would not resolve quickly upon discontinuing CPAP or that may be amenable to alleviation with ancillary treatments (e.g., humidification). The SR5 reported the quality of the included studies as low to moderate (Appendix 10). The findings of the SR are summarized in Table 72.

2) Oral appliances versus inactive controls

Two SRs5,58 reported on adverse events associated with MAD or undefined OA use. One SR58 included adults with moderate-to-severe OSA who were overweight to obese. The other SR5 provided no information on OSA severity or comorbidities. The adverse events reported by the SRs,5,58 with sample sizes ranging from 168 patients5 to 361 patients58 from five studies,58 included dental crown damage, dry mouth, excessive salivation, mouth muscle and joint discomfort, occlusal changes, and tooth discomfort. The SRs reported the quality of the included studies as moderate5 or moderate to high58 (Appendix 10).

Across the two SRs, 10 primary studies had been included, all of which had been included in one or the other SR, with no overlap between the two SRs (Appendix 16.27).

The findings of the SRs are summarized in Table 73.

3) Lifestyle interventions versus inactive controls

One SR5 reported on adverse events associated with diet programs in adults with severe OSA who were obese. The adverse events reported by the SR,5 with a sample size of 30 patients from one study, included gout and transient elevation of alanine amino transferase. The authors reported that adverse events associated with diet programs were transient and rare. The SR5 reported the quality of the included studies as moderate (Appendix 10). The findings of the SR are summarized in Table 74.

4) Continuous positive airway pressure versus oral appliances

One SR77 reported on adverse events associated with CPAP and undefined OAs in adults with mild-to-severe OSA, with no information on comorbidities. The adverse events associated with CPAP use included eye irritation, nasal congestion, rhinorrhea, a sense of suffocation or pressure on face, stuffy nose, and mask problems. The adverse events associated with OA use included discomfort in mouth, excessive salivation, and sore teeth and jaw muscles. The SR,77 with a sample size of 289 patients from seven studies, reported that although the adverse events associated with CPAP might be more severe than those with undefined OAs, all adverse events for both CPAP and undefined OAs were generally mild. The SR77 reported the quality of the included studies as low (Appendix 10). The findings of the SR are summarized in Table 75.

Review of Primary Studies

1) Combination therapy versus inactive controls

One study80 reported on AEs in overweight patients with unknown OSA severity, provided by mean BMI. All 28 patients in the study reported having complications after MMA plus GTA, including 13.9% of patients having major complications, defined as any complication requiring readmission or operation under general anesthesia. Concerns with the quality of the study were assessed to be low80 (Appendix 14). The findings of the primary study are summarized in Table 76.

Summary of Results on Adverse Events

Lists of adverse events associated with CPAP, OAs (i.e., MADs and undefined OAs), diet, and combination therapy (i.e., MMA plus GTA) were identified. Although common, adverse events associated with both CPAP and OAs were deemed mild and resolvable, even if the adverse events associated with CPAP were more serious than those associated with OAs. Adverse events associated with diet programs were deemed transient and rare. Some of the adverse events associated with MMA plus GTA were deemed major. No subgroup or meta-regression analyses were found.


Table 72: Summary of Adverse Events Associated With PAP Use

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

368 patients from 6 studies (2003 to 2009):

  • 4 weeks to 4 months of study duration

AEs with PAP reported:

  • CPAP (from 1 study):
    • Claustrophobia (1.4%)
    • Pressure intolerance (9.2%)
  • APAP (from 3 studies):
    • Claustrophobia (2.9%)
    • Pressure intolerance (3.6%)
    • Epistaxis (0% for humidified; 9.1% for non-humidified)
  • Nasal CPAP (from 2 studies):
    • Claustrophobia (4.8% to 23%)
    • Excessive pressure (13%)
    • Epistaxis (12%)
    • Excessive nasal dryness (12%)
  • Oral CPAP (from 2 studies):
    • Excessive pressure (19%)
    • Excessive oral dryness (14% to 52%)
    • Major gums/lip problems (9.5% to 14%)
    • Excessive salivation (4.8%)

Other AEs reported: skin irritation, nasal irritation or obstruction, minor aerophagia, abdominal distention, minor chest wall discomfort, and pressure discomfort.

None

Low to moderate or mixed

Generally, 5% to 15% of patients reported specific AEs they considered to be a major problem while using CPAP. However, no study reported a severe AE that would not resolve quickly upon discontinuing CPAP or that may be amenable to alleviation with ancillary treatments (e.g., humidification).

AE = adverse event; APAP = autotitrating positive airway pressure; CPAP = continuous positive airway pressure; MA = meta-analysis; MR = meta-regression; OSA = obstructive sleep apnea; PAP = positive airway pressure.

Table 73: Summary of Adverse Events Associated With OA Use

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Serra-Torres 201658

361 patients from 5 studies (2005 to 2013):

  • Moderate-to-severe OSA (mean AHI range: 26.4 to 45.5 events/hour, where reported)
  • Overweight to obese (mean BMI range: 26.7 to 31.4 kg/m2)
  • 1 to 64 months of study duration

AEs with MADs reported:

  • Muscle discomfort (19% to 56%)
  • Joint discomfort (19% to 69%)
  • Excessive salivation (19% to 44%)
  • Dry mouth (25% to 75%)
  • Occlusal changes (35% to 56%)

None

Moderate to high

The main AEs associated with MAD use were mouth muscle and joint discomfort, excessive salivation, dry mouth, and occlusal changes.

Balk 20115

168 patients from 5 RCTs (1996 to 2008):

  • 4 weeks to 4 years of study duration

AEs with OAs reported:

  • Dental crown damage (6.3%)
  • Tooth or joint discomfort (2.2% to 5.2%)

Other AEs reported: teeth loosening, pressure sensation of the mouth, transient morning mouth and TMJ discomfort or sounds, minor sore teeth or jaw, transient mild mucosal erosions, minor excessive salivation, tooth grinding, and sleep disruption.

None

Moderate

Dental crown damage and tooth or joint discomfort are major AEs associated with OA use.

AE = adverse event; AHI = Apnea-Hypopnea Index; BMI = body mass index; MA = meta-analysis; MAD = mandibular advancement device; MR = meta-regression; OA = oral appliance; OSA = obstructive sleep apnea; RCT = randomized controlled trial; TMJ = temporal mandibular joint.

Table 74: Summary of Adverse Events Associated With Lifestyle Interventions

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Balk 20115

30 patients from 1 RCT (2009):

  • Severe OSA (mean AHI = 37 events /hour)
  • Obese (mean BMI = 34.8 kg/m2)
  • 9 weeks of study duration

AEs with weight-loss programs reported:

  • Gout (3.3%)
  • Transient elevation of alanine amino transferase (6.7%)

Other AEs reported: dizziness, dry lips, and constipation

None

Moderate

AEs associated with a very low-energy diet were transient and rare.

AE = adverse event; AHI = Apnea-Hypopnea Index; BMI = body mass index; MA = meta-analysis; MR = meta-regression; OSA = obstructive sleep apnea; RCT = randomized controlled trial.

Table 75: Summary of Adverse Events Associated With CPAP Versus OA Use

Study

Patient Characteristics

Pooled Estimates from MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Li 201377

289 patients from 7 RCTs (1996 to 2011):

  • Mild-to-severe OSA (AHI threshold: ≥ 5 events/hour)
  • Comorbidities: NR
  • 6 to 48 weeks of study duration

AEs associated with CPAP:

  • Nasal congestion
  • Rhinorrhea
  • Eye irritation
  • Sense of suffocation
  • Sense of pressure on face
  • Stuffy nose
  • Mask problems

AEs associated with OAs:

  • Sore teeth
  • Sore jaw muscles
  • Excessive salivation
  • Early morning discomfort in mouth

1 RCT reported that patients using CPAP reported moderate-to-severe side effects, while those using OAs reported that all side effects were mild and acceptable. 4 RCTs reported that side effects were common but mild and acceptable for both CPAP and OAs.

None

Low

There were various and specific side effects associated with CPAP and OAs.

AE = adverse event; AHI = Apnea-Hypopnea Index; CPAP = continuous positive airway pressure; MA = meta-analysis; MR = meta-regression; NR = not reported; OA = oral appliance; OSA = obstructive sleep apnea; RCT = randomized controlled trial.

Table 76: Summary of Adverse Events Associated with Combination Therapy

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Passeri 201680

28 patients:

  • OSA severity: NR
  • Mean BMI ± SD: 29.6 ± 4.7 kg/m2

AEs associated with MMA plus GTA:

  • Total number of complications: 108
  • 13.9% of the complications were majora
  • 100% (28/28) of patients had complicationsb

None

MMA plus GTA is a safe method for OSA treatment in terms of mortality, but patients need to be cautioned on potential complications.

AE = adverse event; BMI = body mass index; GTA = genial tubercle advancement; MMA = maxillomandibular advancement; NR = not reported; OSA = obstructive sleep apnea; SD = standard deviation.
a Defined as any complication requiring readmission or operation under general anesthesia.
b All major complications in this study were caused by infections.


Adherence

Overview of Reviews

1) Continuous positive airway pressure versus inactive controls

Nine SRs5,10,53,56,57,64-66,75 reported on CPAP adherence in adults with mild-to-severe,10,56,66,75 moderate-to-severe,53,57,65 severe,5 or unknown severity64 OSA. Eight SRs included normal-weight-to-obese,75 overweight-to-obese,10,53,57,66 or obese5,64,65 patients. Three SRs56,64,65 included patients with hypertension or resistant hypertension,65 diabetes,64 or previous CVD.56

Eight of the nine SRs,5,10,53,57,64-66,75 with sample sizes ranging from 118 patients64 to 4,14653 patients from two studies5 to 29 studies,10 reported that the mean CPAP adherence ranged from 2.3 hours/night75 to 6.6 hours/night.75 One of the nine SRs,56 with a sample size of 4,194 patients from four studies, reported that the proportion of patients using CPAP for at least four hours/night ranged from 64.6% to 100% across the included studies. One of the nine SRs,5 with a sample size of 2,160 patients from five studies, reported rates of CPAP discontinuation between 16% at one year and 32% at four years of CPAP use. Study duration of the included primary studies, reported by all nine SRs, ranged from one week66,75 to 80 months.56 I2 scores were not applicable. The SRs reported the quality of the included studies as low,64 low to moderate,5,10 high53,57,65,66 or mixed56,75 (Appendix 10).

The findings of the SRs are summarized in Table 77.

2) Mandibular advancement devices versus inactive controls

One SR75 reported on MAD adherence in adults with mild-to-severe OSA who were normal-weight-to-obese. The SR,75 with a sample size of 325 patients from four studies, reported that the mean MAD adherence ranged from 5.5 hours/night to 7.7 hours/night. Study duration of the included studies ranged from four weeks to 12 weeks. The SR75 reported the quality of the included studies as mixed75 (Appendix 10). The findings of the SRs are summarized in Table 78.

3) Continuous positive airway pressure versus oral appliances

Two SRs5,77 reported on adherence with CPAP versus MADs5 or undefined OAs77 in adults with mild-to-severe77 or moderate-to-severe5 OSA. One SR5 included overweight-to-obese patients. The other SR77 provided no information on comorbidities.

One of the two SRs,5 with a sample size of 28 patients from one study, reported significantly greater adherence with MADs, compared with CPAP, in terms of nightly durations (i.e., seven hours/night versus six hours/night) and numbers of nights (i.e., 98% versus 90%) of the device use. Study duration ranged from one month to three months. I2 scores were not applicable. The SR5 reported the quality of the included studies as low to moderate (Appendix 10).

The other SR,77 with sample sizes ranging from 290 patients to 409 patients from five studies to six studies, reported no significant differences in adherence with CPAP versus MADs, in terms of nightly durations and numbers of nights of the device use, as well as study withdrawals. Study duration ranged from eight weeks to 48 weeks. I2 scores ranged from 22% to 95%. The SR5 reported the quality of the included studies as low (Appendix 10).

Across the two SRs, eight primary studies had been included, all of which had been included in one or the other SR, with no overlap between the two SRs (Appendix 16.28).

The findings of the SRs are summarized in Table 79.

Review of Primary Studies

1) Tongue-retaining devices versus inactive controls

Two studies92,113 reported on adherence in adults with moderate-to-severe113 or unknown severity92 OSA. One study113 included normal-to-overweight patients. The other study92 provided no information on comorbidities. Both studies,92,113 with sample sizes ranging from 20 patients92 to 84 patients,113 reported rates of discontinuation with TRDs, ranging from 35% at four months92 to 48% at five years.113 One of the two studies92 suggested that TRDs may be a possible alternative to MADs. Concerns with the quality of the two studies were assessed to be low,92 or moderate113 (Appendix 14). The findings of the primary studies are summarized in Table 80.

2) Positional therapy versus inactive controls

Twelve studies78,79,83,85,86,88,96,97,99,102,112,118 reported on adherence in adults with mild,85,86,88,97,102 moderate,78,79,96,99,118 or severe OSA,112 providing mean78,88,96,102,112,118 or median AHI.85,86,97,99 Studies included patients who were normal weight85 or overweight78,79,86,88,96,97,99,102,112,118providing mean78,79,86,88,96,102,112,118 or median BMI.85,97,99 One study79 provided no information on comorbidities.

The findings on adherence across the studies, with sample sizes ranging from 14 patients83 to 145 patients,97 were mixed. Eight studies78,79,83,85,86,96,99,102 reported good adherence with neck-positioning devices,79,96 mattresses and pillows,83,85 commercial or self-made bands,86 sleep positioning trainers,78,99 or tennis balls.102 One study97 reported that adherence with a sleep position trainer was higher when measured by a self-reported questionnaire, compared with the device itself. One study88 reported higher adherence with sleep position trainers, compared with tennis balls. Three studies86,112,118 reported high rates of discontinuation with tennis balls112,118 or commercial or self-made bands,86 especially long-term.86,112 The quality of the 12 studies was assessed to be low78,79,83,85,86,88,96,97,99,102 or high112,118 (Appendix 13 and Appendix 14).

From subgroup analyses, one study85 reported 100% adherence in both normal-weight and overweight patients.

The findings of the primary studies are summarized in Table 81.

3) Combination therapy versus inactive controls

One study107 reported on adherence with CPAP plus diet programs in adults with severe OSA who were obese, provided by mean AHI and a BMI range. The study,107 with a sample size of 63 patients, reported that 69.8% of all patients completed the full CPAP plus diet programs. Concerns with the quality of the study were assessed to be low107 (Appendix 14). The findings of the primary studies are summarized in Table 82.

Summary of Results on Adherence

Adherence levels associated with CPAP, OAs (i.e., MADs and TRDs), positional therapy, and combination therapy (i.e., CPAP plus diet) were identified. Evidence was also found on active comparisons between CPAP and MADs or undefined OAs. The mean adherence ranged from 2.3 hours/night to 6.6 hours/night for CPAP and 5.5 hours/night to 7.7 hours/night for MADs. Greater adherence with MADs versus CPAP was also reported in terms of nightly durations (i.e., seven hours/night versus six hours/night) and numbers of nights (i.e., 98% versus 90%) of device use, although it is unclear if these results are clinically important. Discontinuation was reported to be 16% of patients at one year and 32% of patients at four years for CPAP, 35% of patients at four months and 48% of patients at five years for TRDs, and 31.2% at some unknown follow-up time for CPAP plus diet programs. The findings on adherence with positional therapy across different devices were mixed, with no notable trends. No subgroup or meta-regression analyses were found.


Table 77: Summary of Adherence With CPAP

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Guo 201653

4,146 patients from 18 RCTs (2006 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 24 to 60 events/hour)
  • Overweight to obese (mean BMI range: 28 to 40 kg/m2)
  • 22.3% diabetes (% range: 28% to 63%, where reported); 34.7% smoking (% range: 12% to 84%, where reported)
  • 2 to 60 months of study duration
  • Mean CPAP adherence: 5 hours/night (range: 4 to 7 hours/night)

None

High

The mean CPAP adherence was 5 hours/night.

Kim 201656

4,194 patients from 1 RCT and 3 cohort studies (2007 to 2014):

  • Mild-to-severe OSA (AHI threshold range: ≥ 5 to ≥ 20 events/hour)
  • Previous CVD (excluded from the RCT and 1 cohort study but included in the other studies)
  • 48 to 80 months of study duration
  • % CPAP adherence range:a 64.4-100% of patients

None

High (for the RCT) or mixed (for the non-RCTs)

The proportion of patients using CPAP for at least 4 hours/night ranged from 64.6% to 100% across the included studies.

Liu 201657

446 patients from 5 RCTs (2010 to 2015):

  • Moderate-to-severe OSA (mean AHI range: 20 to 52.7 events/hour)
  • Overweight to obese (mean BMI range: 29.8 to 34.1 kg/m2)
  • Resistant hypertension (100%)
  • 3 to 8 months of study duration
  • Mean CPAP adherence range: 4 to 6 hours/night

None

High

The mean CPAP adherence ranged from 4 hours/night to 6 hours/night across the included studies.

Feng 201564

118 patients from 2 RCTs and 3 observational studies (2004 to 2012):

  • OSA severity: NR
  • Obese (mean BMI range: 33.6 to 42.7 kg/m2)
  • Diabetes (100%)
  • 1 to 3 months of study duration
  • Mean CPAP adherence range: 3.6 to 5.8 hours/night

None

Low

The mean CPAP adherence ranged from 3.6 hours/night to 5.8 hours/night across the included studies.

Hu 201565

794 patients from 7 RCTs (2006 to 2014):

  • Moderate-to-severe OSA (mean AHI range: 28.1 to 58.3 events/hour)
  • Obese (mean BMI range: 30.8 to 35.7 kg/m2)
  • Hypertension (100% in 3 RCTs) or resistant hypertension (100% in 4 RCTs)
  • 1 to 6 months of study duration
  • Mean CPAP adherence range: 4.5 to 6.0 hours/night

None

High

The mean CPAP adherence ranged from 4.5 hours/night to 6.0 hours/night across the included studies.

Pan 201566

1,698 patients from 11 RCTs (1994 to 2012):

  • Mild-to-severe OSA (mean AHI range: 10.0 to 55.4 events/hour)
  • Overweight to obese (mean BMI range: 29 to 33.0 kg/m2)
  • 1 to 24 weeks of study duration
  • Mean CPAP adherence range: 2.8 to 5.4 hours/night

None

High

The mean CPAP adherence ranged from 2.8 hours/night to 5.4 hours/night across the included studies.

Fava 201410

1,820 patients from 29 RCTs (1996 to 2012):

  • Mild-to-severe OSA (mean AHI range: 12.9 to 63.8 events/hour)
  • Overweight to obese (mean BMI range: 27.2 to 37.0 kg/m2)
  • Hypertension (% range: 0% to 100%, where reported)
  • 2 to 52 weeks of study duration
  • Mean CPAP adherence range: 3.3 to 6.4 hours/night

None

Low to moderate

The mean CPAP adherence ranged from 3.3 hours/night to 6.4 hours/night across the included studies.

Povitz 201475

1,570 patients from 21 RCTs (1998 to 2013):

  • Mild-to-severe OSA (mean AHI range: 10 to 65.1 events/hour)
  • (for the 1,732 patients included in the SR) Normal-to-obese (mean BMI range: 24.7 to 42.5 kg/m2)
  • Study duration: 1 to 24 weeks
  • Mean CPAP adherence range: 2.3 to 6.6 hours/night

None

Mixed

The mean CPAP adherence ranged from 2.3 hours/night to 6.6 hours/night across the included studies.

Balk 20115

2,160 patients from 1 RCT and 4 cohort studies (1996 to 2009):

  • Severe OSA (mean AHI range: 44 to 70 events/hour)
  • Obese (mean BMI: 30 kg/m2)
  • 3 months to 4 years of study duration
  • Discontinuation of CPAP use (from 2 studies): 16% at 1 year and 32% at 4 years in 1 cohort study; 14% at mean 3.2 years in 1 cohort study
  • CPAP use (from 2 studies): mean 5 hours/night at 3 months in 1 RCT and 1 cohort study

None

Low to moderate or mixed

Discontinuation of CPAP varied between 16% at 1 year and 32% at 4 years of CPAP use.

Each study defined adherence differently.

AHI = Apnea-Hypopnea Index; BMI = body mass index; CPAP = continuous positive airway pressure; CVD = cardiovascular disease; MA = meta-analysis; MR = meta-regression; OSA = obstructive sleep apnea; RCT = randomized controlled trial.
a Defined as usage of CPAP for ≥ 4 hours/night on average.

Table 78: Summary of Adherence With MADs

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Povitz 201475

325 patients4 RCTs (2004 to 2007):

  • Moderate OSA (mean AHI range: 21.3 to 28.9 events/hour)
  • (for the 1,732 patients included in the SR) Normal-to-obese (mean BMI range: 24.7 to 42.5 kg/m2)
  • Study duration: 4 to 12 weeks
  • Mean MAD adherence range: 5.5 to 7.7 hours/night

None

Mixed

The mean MAD adherence ranged from 5.5 hours/night to 7.7 hours/night across the included studies.

AHI = Apnea-Hypopnea Index; BMI = body mass index; MA = meta-analysis; MAD = mandibular adjustment device; MR = meta-regression; OSA = obstructive sleep apnea; RCT = randomized controlled trial.

Table 79: Summary of Adherence With CPAP Versus OAs

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Li 201377

CPAP versus OAs

There was no significant difference between CPAP and OAs for treatment adherence, assessed as hours of use per night, nights of use per week, or study withdrawal.

409 patients from 6 RCTs (2002 to 2008):

  • Mild-to-severe OSA (AHI threshold: ≥ 5 events/hour)
  • Comorbidities: NR
  • 8 to 12 weeks of study duration
  • Hours of use per night:
    • Crossover trials (from 2 RCTs) MD (95% CI) = -1.01 (-2.78 to 0.75); P = 0.26; I2 = 95%
    • Parallel-groups trials (from 2 RCTs) MD (95% CI) = -0.82 (-1.91 to 0.27); P = 0.14; I2 = 93%
  • Number of nights of use per week:
    • Parallel-groups trials (from 2 RCTs) MD (95% CI) = -0.16 (-0.40 to 0.08); P = 0.19; I2 = 54%

None

Low

290 patients from 5 RCTs (2007 to 2011):

  • Mild-to-severe OSA (AHI threshold: ≥ 5 events/hour)
  • Comorbidities: NR
  • 10 to 48 weeks of study duration
  • Study withdrawal:
    • Parallel-groups trials (from 5 RCTs) OR (95% CI) = 0.64 (0.25 to 1.61); P = 0.34; I2 = 22%

None

Balk 20115

CPAP versus MADs

MADs were associated with higher adherence, compared with CPAP.

28 patients from 1 RCT (2009):

  • Moderate-to-severe OSA (mean AHI range: 21 to 40 events/hour)
  • Overweight to obese (mean BMI range: 26.7 to 34.1 kg/m2)
  • 1 to 3 months of study duration
  • Hours of use:
    • 6.0 hours/night for CPAP
    • 7.0 hours/night for MAD
    • P < 0.01
  • Nights of use:
    • 90% for CPAP
    • 98% for MAD
    • P < 0.01

None

Moderate

AHI = Apnea-Hypopnea Index; BMI = body mass index; CI = confidence interval; CPAP = continuous positive airway pressure; MA = meta-analysis; MAD = mandibular advancement device; MD = mean difference; MR = meta-regression; NR = not reported; OA = oral appliance; OSA = obstructive sleep apnea; SR = systematic review.

Table 80: Summary of Adherence With TRDs

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Roplekar 201592

TRDs pre versus post

The level of adherence with TRDs was encouraging.

20 patients:

  • Patient characteristics: NR
  • 65% continuation after 4 months (i.e., 7/20 patients did not use the device)

None

Lazard 2009113

TRDs pre versus post

Adherence with TRDs was shown to be quite good.

84 patients (recruited) or 63 patients (analyzed):

  • Mean AHI ± SD: 37 ± 19.5 events/hour
  • Mean BMI ± SD: 26 ± 3.8 kg/m2
  • 52% continuation after 5 years (i.e., 30/63 patients did not use the device)
    • 79% daily adherencea after 5 years among the 33 users

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; NR = not reported; SD = standard deviation; TRD = tongue-retaining device.
a Defined as > 4 nights/week of use.

Table 81: Summary of Adherence With Positional Therapy

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Benoist 201678

Positional therapy (i.e., sleep position trainers) pre versus post

The sleep position trainer used in this study had high adherence rates during the 3 months of use.

33 positional OSA patients:

  • Median AHI: 18.3 (IQR: 13.7 to 24.0) events/hour
  • Mean BMI ± SD: 27.9 ± 2.8 kg/m2
  • Adherence:
    • Defined as ≥ 4 hours/night and ≥ 5 days/week of usage:
      • 89.0% of patients
    • Average nightly use ± SD per patients:
      • 6.92 ± 0.75 hours/night

None

Levendowski 201679

Positional therapy (i.e., neck position devices) pre versus post

The device used in this study had high adherence rates. Sleep efficiency and loud snoring varied from patient to patient.

135 patients:

  • Patient characteristics: NR
  • Unacceptable adherence:
    • Defined as < 50% 4-hour adherence:
      • 18.4% of patients
    • Defined as < 4 hours average hours of use:
      • 24.3% of patients
  • Marginal adherence:
    • Defined as ≥ 50 and < 70% 4-hour adherence:
      • 22.8% of patients
    • Defined as ≥ 4 and < 5 average hours of use:
      • 15.4% of patients
  • Acceptable adherence:
    • Defined as ≥ 70 and < 80% 4-hour adherence:
      • 22.8% of patients
    • Defined as ≥ 5 and < 6 average hours of use:
      • 25.7% of patients
  • Excellent adherence:
    • Defined as ≥ 80% 4-hour adherence:
      • 36.0% of patients
    • Defined as ≥ 5 hours average hours of use:
      • 34.6% of patients

None

Bidarian-Moniri 201583

Positional therapy (i.e., mattresses and pillows for prone positioning) pre versus post

Positional therapy with the mattress and a pillow for prone positioning improved AHI and ODI levels patients with OSA. Adherence was satisfactory during the course of the study.

14 patients:

  • Mean AHI: 26 events/hour (range: 6 to 53)
  • Mean BMI: 26 kg/m2
  • Adherence:
    • All 14 patients completed the study after 4 weeks, with sleep time of > 4 hours/night with the device

None

Chen 201585

Positional therapy (i.e., head-positioning pillows) pre versus post

The use of the head-positioning pillow in this study was associated with short-term adherence.

25 positional OSA patients:

  • Median AHI: 7.0 events/hour (IQR: 6.0 to 15.2)
  • Median BMI: 24.8 kg/m2 (IQR: 23.1 to 26.4)

Adherence:

  • All patients (n = 25)
    • Regular pillow: Reference
    • Head-positioning pillow: 100% of patients using it every night
  • Baseline weight:
    • Normal-weight patients (n = 13) adherence:
      • Before (regular pillow): Reference
      • After 3 nights (head-positioning pillow): 100
    • Overweight patients (n = 12) adherence:
      • Before (regular pillow): Reference
      • After 3 nights (head-positioning pillow): 100

de Vries 201586

Positional therapy (i.e., commercial devices or self-made constructions) pre versus post

Short-term adherence was good. However, long-term adherence was considered disappointing.

40 positional OSA patients:

  • Mean AHI: 14.5 events/hour (range: 10.7 to 19.6)
  • Mean BMI ± SD: 28.0 ± 4.1 kg/m2
  • Adherence:
    • Short-term: most patients used their device > 7 hours/night, > 6 days/week after < 3 months
    • Long-term: 65% (n = 26) patients stopped using their device after 13 ± 5 months

None

Eijsvogel 201588

Positional therapy (i.e., tennis balls or sleep position trainers) pre versus post

 

26 (TBT) or 29 (SPT) positional OSA patients:

  • Mean AHI ± SD: 13.1 ± 9.1 (TBT) or 11.4 ± 4.9 (SPT) events/hour
  • Mean BMI ± SD: 26.8 ± 3.0 (TBT) or 27.6 ± 4.5 (SPT) kg/m2
  • Effective adherence:
    • TBT: 42.3% (11/16 patients)
    • SPT: 75.9% (22/29 patients)
    • Difference: P = 0.01
  • Median hours used per night:
    • TBT: 4.5 (IQR: 1.1 to 7.0)
    • SPT: 6.5 (IQR: 5.5 to 7.2)
    • Difference: P = 0.078
  • Median percentage of days used:
    • TBT: 77.2 (IQR: 21.2 to 96.6)
    • SPT: 100 (IQR: 79.6 to 100)
    • Difference: P = 0.005
  • Everyday use:
    • TBT: 15.4% (4/26 patients)
    • SPT: 51.7% (15/29 patients)
    • Difference: P = 0.005

None

Adherence was significantly improved with a sleep position trainer, compared with the tennis ball technique.

van Maanen 201497

Positional therapy (i.e., sleep position trainers) pre versus post

Most patients using the sleep position trainer were considered compliant (64.4% of patients using the trainer more than 4 hours per night).

145 positional OSA patients:

  • Median AHI: 11.5 events/hour (IQR: 2.5 to 20.5)
  • Median BMI: 27.0 kg/m2 (IQR: 23.0 to 31.0)
  • Average use:
    • 5.5 hours/night in 106/145 patients
  • Objective adherence (measured by the device as more than 4 hours of use per night over 168 nights):
    • 64.4% (106/145)
  • Subjective adherence (measured by an online questionnaire after 1, 3, and 6 months with the device):
    • After 1 month: 91.8% (110/145 patients)
    • After 3 months: 74.3% (101/145 patients)
    • After 6 months: 59.8% (87/145 patients)

None

Levendowski 201496

Positional therapy (i.e., neck position devices) pre versus post

 

30 positional OSA patients:

  • Mean AHI ± SD: 24.7 ± 14.7 events/hour
  • Mean BMI ± SD: 28 ± 3.4 kg/m2
  • Adherence:
    • Median percentage: 96% (range: 71% to 100%)
    • The device was worn in 99% of treatment nights

None

The positional therapy method used in this study (neck position device) was associated with good adherence in patients with positional OSA.

van Maanen 201399

Positional therapy (i.e., sleep position trainers) pre versus post

 

31 positional OSA patients:

  • Median AHI: 16.4 events/hour (IQR: 6.6 to 29.9 events/hour)
  • Mean BMI ± SD: 27.0 ± 3.7 kg/m2
  • Adherence:
    • Median percentage: 92.7% (range: 62% to 100%)

None

Short-term adherence (1 month) was high using the sleep positioning trainer; however, long-term adherence with the sleep positioning trainer is as yet unknown.

Heinzer 2012102

Positional therapy (i.e., tennis balls) pre versus post

Positional therapy was successful with select patients with an objective adherence of 73.7%

16 positional OSA patients:

  • Mean AHI ± SD: 26.7 ± 17.5 events/hour
  • Mean BMI ± SD: 25.4 ± 4.1 kg/m2
  • Adherence:
    • Median percentage ± SD: 73.7 ± 29.3%
    • 62.5% (n = 10) of the patients used the device > 80% of the nights

None

Bignold 2009112

Positional therapy (i.e., tennis balls) pre versus post

Long-term adherence was very poor with the tennis ball technique among patients with positional OSA; alternative forms of positional therapy are required.

108 patients:

  • Mean AHI ± SD: 32.4 ± 35.2 events/hour
  • Mean BMI ± SD: 28.7 ± 4.2 (67 respondents) or 29.3 ± 8.5 (41 non-respondents) kg/m2
  • Adherence (n = 67):
    • 6% (n = 4) of patients reported still using TBT.
    • 13.4% (n = 9) reported no longer using TBT because they had learned to avoid sleeping in the supine position.
    • 80.6% (n = 54) were no longer using TBT nor avoiding sleep in the supine position. 63% (34 of the 54 patients) reported discomfort as the main reason for no longer using TBT.
    • Other reasons for non-adherence included:
      • Patient claims not to sleep on his/her back
      • Shoulder problems
      • Skin irritation from using TBT
      • Ineffectiveness on a soft mattress
      • Suspicion that TBT would cause back problems

None

Oksenberg 2006118

Positional therapy (i.e., tennis balls) pre versus post

The positional therapy used in this study (tennis ball technique) did not have great long-term adherence.

78 positional OSA patients:

  • Mean AHI ± SD: 25.5 ± 17.3 events/hour
  • Mean BMI ± SD: 28.1 ± 3.7 kg/m2
  • Adherence:
    • 64.1% of patients filled out and returned the questionnaire, of whom 38% were still using TBT after 6 months

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; IQR = interquartile range; NR = not reported; ODI = oxygen desaturation index; OSA = obstructive sleep apnea; SD = standard deviation; SPT = sleep position trainer; TBT = tennis ball technique.

Table 82: Summary of Adherence With Combination Therapy

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Johansson 2011107

CPAP plus diet programs

The level of adherence with the full program was encouraging.

63 patients:

  • Mean AHI ± SD: 36 ± 15 events/hour
  • BMI range: 30 to 40 kg/m2
  • 69.8% of patients completed the full CPAP plus diet programs (i.e., 44/63 patients)

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; CPAP = continuous positive airway pressure; MA = meta-analysis; MR = meta-regression.


Snoring

Overview of Reviews

1) Expiratory positive airway pressure versus inactive controls

One SR68 reported on snoring in adults with moderate OSA who were overweight to obese. The SR,68 with a sample size of 102 patients from three studies, reported uniform reductions in snoring across its three included studies with EPAP, compared with pre-treatment, with no effect sizes reported. Study duration of the included primary studies ranged from one night to 12 months. I2 scores were not applicable. The SR68 reported the quality of the included studies as high (Appendix 10). The findings of the SR are summarized in Table 83.

2) Oral appliances versus inactive controls

One SR58 reported on snoring in adults with mild-to-severe OSA who were overweight. The SR,68 with a sample size of 112 patients from three studies, reported reductions in mean ranges in visual analogue scales for snoring with MADs, compared with pre-treatment, from 3-8 to 2-3. Study duration of the included primary studies ranged from one month to nine months. I2 scores were not applicable. The SR68 reported the quality of the included studies as moderate to high (Appendix 10). The findings of the SR are summarized in Table 84.

Review of Primary Studies

1) Positional therapy versus inactive controls

Three studies79,85,105 reported on snoring in adults with mild85,105 OSA, provided by mean105 or median85 AHI. Two studies included normal weight85 or overweight105 patients, provided by mean105 or median85 BMI. The other study79 provided no information on comorbidities.

Two of the three studies,79,105 with sample sizes ranging from 15 patients105 to 135 patients,79 reported no improvements in snoring with positional therapy (i.e., an apparatus designed to prevent sleep in the supine position79,105), compared with pre-treatment, in most patients79 or in patients with supine-predominant OSA.105 The other study,85 with a sample size of 25 patients, reported that in overweight patients, subjective snoring, measured on a visual analogue scale ranging from 0 to 10, improved significantly after positional therapy (i.e., a head-positioning pillow), but objective snoring, expressed as the number of snoring events/hour measured on an acoustic analytical program, did not; in normal-weight patients, both subjective and objective snoring significantly improved. Concerns with the quality of the three studies were assessed to be low79,85,105 (Appendix 14).

From subgroup analyses, one study85 reported that, while snoring severity improved in both normal-weight and overweight patients with positional therapy, compared with pre-treatment, snoring index improved in normal-weight, and not overweight, patients only.

The findings of the primary studies are summarized in Table 85.

Summary of Results on Snoring

For snoring, evidence was found on inactive comparisons with EPAP, MADs, and positional therapy. No evidence was found on active comparisons. Compared with pre-treatment, EPAP and MADs were effective at reducing snoring. Findings on positional therapy were mixed. Two studies reported no improvement in snoring, and one study reported improvements in both subjective and objective snoring in normal-weight patients but reported improvements in subjective but not objective snoring in overweight patients. All findings were from uncontrolled studies, and it is unclear if these results are clinically important. Subgroup analyses suggest that patients with normal-weight were more likely to experience improvements in snoring, compared with those who were overweight. No subgroup or meta-regression analyses were found on comorbidities, baseline EDS or OSA severity, sex, age, adherence, or study duration.


Table 83: Summary of Change in Snoring From EPAP Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analysis

Riaz 201568

EPAP pre versus post

EPAP was associated with a decrease in snoring, when compared with pre-treatment.

102 patients from 3 studies (2008 to 2015):

  • Moderate OSA (mean AHI range: 15.7 to 24.8 events/hour)
  • Overweight to obese (mean BMI range: 27 to 32.5 kg/m2)
  • 1 night to 12 months of study duration
  • Uniform reductions in snoring in all 3 studies

None

High

AHI = Apnea-Hypopnea Index; BMI = body mass index; EPAP = expiratory positive airway pressure; MA = meta-analysis; MR = meta-regression; OSA = obstructive sleep apnea.

Table 84: Summary of Change in Snoring From OAs Versus Inactive Controls

Study

Patient Characteristics

Pooled Estimates From MAs or Narrative Summary

Quality of Included Studies

Conclusions

Overall

Subgroup or MR Analyses

Serra-Torres 201658

MADs pre versus post

MADs were associated with a decrease in snoring, when compared with pre-treatment.

112 patients from 3 studies (2005 to 2013):

  • Mild-to severe OSA (mean AHI range: 14 to 45.5 events/hour)
  • Overweight (mean BMI range: 27.9 to 29.2 kg/m2)
  • 1 to 9 months of study duration
  • Mean snoring VAS range:
    • Baseline: 3 to 8
    • Follow-up: 2 to 3
  • No MA

None

Moderate to high

AHI = Apnea-Hypopnea Index; BMI = body mass index; MA = meta-analysis; MAD = mandibular advancement device; MR = meta-regression; NA = not applicable; NR = not reported; OA = oral appliance; OSA = obstructive sleep apnea; VAS = visual analogue scale.

Table 85: Summary of Change in Snoring From Positional Therapy Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Levendowski 201679

Positional therapy (i.e., sleep position modification devices) pre versus post

The majority of patients did not find a difference in snoring patterns from the neck device.

135 patients:

  • Patient characteristics: NR
  • Snoring patterns:
    • No difference in 91.3% (105/115) of patients
    • Decrease in 4.3% (5/115) of patients
    • Increase in 4.3% (5/115) of patients

None

Chen 201585

Positional therapy (i.e., head-positioning pillows) pre versus post

The use of the head-positioning pillow provided improvements in snoring severity in overweight patients; a significant improvement was found in both snoring severity and snoring index in normal-weight patients.

25 positional OSA patients:

  • Median AHI: 7.0 events/hour (IQR: 6.0 to 15.2 events/hour)
  • Median BMI: 24.8 kg/m2 (IQR: 23.1 to 26.4 kg/m2)

Snoring severity:

  • All patients (n = 25)
    • Regular pillow: 5.0 (5.0, 7.5)
    • Head-positioning pillow: 4.0 (1.5, 5.0)
    • Difference: P < 0.001

Snoring index:

  • All patients (n = 25)
    • Regular pillow: 218.0 (100.0, 288.5) events/hour
    • Head-positioning pillow: 115.0 (48.0, 260.3) events/hour
    • Difference: P = 0.001
  • Baseline weight:
    • Normal-weight patients (n = 13) snoring severity
      • Before (regular pillow): 7.0 (5.0, 8.0)
      • After 3 nights (head-positioning pillow): 5.0 (1.5, 5.0)
      • Difference: P = 0.007
    • Overweight patients (n = 12) snoring severity:
      • Before (regular pillow): 5.0 (4.3, 6.0)
      • After 3 nights (head-positioning pillow): 4.0 (1.3, 4.0)
      • Difference: P = 0.007
  • Baseline weight:
    • Normal-weight patients (n = 13) snoring index
      • Before (regular pillow): 200.0 (58.5, 256.0) events/ hour
      • After 3 nights (head-positioning pillow): 107.0 (35.0, 204.3) events/hour
      • Difference: P = 0.003
    • Overweight patients (n = 12) snoring index:
      • Before (regular pillow): 244.0 (169.3, 332.0) events/hour
      • After 3 nights (head-positioning pillow): 149.5 (62.5, 288.0)
      • Difference: P = 0.052

Bignold 2011105

Positional therapy (i.e., sleep position modification devices) pre versus post

Positional therapy may only provide a small reduction in snoring and may be only of greatest benefit to patients who sleep alone.

15 patients:

  • Mean AHI ± SD: 24.1 ± 10.5 events/hour
  • Mean BMI ± SD: 28.8 ± 2.5 kg/m2
  • Snoring frequency:
    • Difference: 20% reduction with treatment, no statistical significance
  • Mean snore duration ± SD:
    • Pre-treatment: 1126 ± 169 msec
    • Post-treatment: 1082 ± 96 msec

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; IQR = interquartile range; NR = no reported; OSA = obstructive sleep apnea; SD = standard deviation.


Fatigue

No SRs or primary studies on any of the comparisons were found that reported on fatigue.

Facial Aesthetics (for Maxillomandibular Advancement Only)

Overview of Reviews

No SRs on any of the comparisons were found that reported on facial aesthetics.

Review of Primary Studies

1) Maxillomandibular advancement versus inactive controls

Three studies89,98,103 reported on facial aesthetics in adults with severe,98 moderate-to-severe,98 severe,103 or unknown severity89 OSA, provided by mean AHI. The studies89,98,103 included normal weight-to-overweight patients98 or overweight patients,89 provided by mean BMI. All studies89,98,103 compared MMA with pre-treatment.

All three studies,89,98,103 with sample sizes ranging from 12 patients103 to 51 patients,89 reported improvements in facial aesthetics with MMA, compared with pre-treatment. Two of the three studies98,103 reported that the post-operative profile was given preference by the majority of the panellists who evaluated the patients' profiles. The other study89 reported that a majority of the patients had indicated that their facial appearance had improved after surgery. Concerns with the quality of the three studies were assessed to be low98,103 or high89 (Appendix 14).

The findings of the primary studies are summarized in Table 86.

Summary of Results on Facial Aesthetics

For facial aesthetics, evidence was found on inactive comparisons with MMA. Compared with pre-treatment, MMA was effective at improving facial aesthetics in patients. No subgroup or meta-regression analyses were found.


Table 86: Summary of Change in Facial Aesthetics From MMA Versus Inactive Controls

Study

Patient Characteristics

Effect Estimates

Conclusions

Overall

Subgroup Analyses

Islam 201589

MMA pre versus post

Satisfaction with facial appearance improved after MMA.

51 patients:

  • OSA severity: NR
  • Mean BMI ± SD: 28 ± 3 kg/m2
  • VAS scorea for satisfaction with the facial appearance:
    • Before: 4.1
    • After: 6.5

None

Cohen-Levy 201398

MMA pre versus post

The facial changes created by MMA were interpreted as positive changes.

15 patients:

  • Mean AHI: 50.9 events/hour (range: 19 to 85 events/hour)
  • Mean BMI ± SD: 27.41 ± 3.5 kg/m2
  • Self-assessment questionnaires:
    • 93% (i.e., 14/15) of patients were satisfied with their aesthetic outcome
    • 73% (i.e., 11/15) of patients noted an improvement, while 20% (i.e., 3/15) did not notice or were not concerned with an improvement
  • Panel assessment:
    • 80% (i.e., 12/15) of patients were given preference in their post-operative profile from panellists

None

Liu 2012103

MMA pre versus post

Thoughts on facial aesthetics post MMA were generally positive from patients.

12 patients:

  • Mean AHI ± SD: 60.53 ± 12.66 events/hour
  • Mean BMI ± SD: 27.45 ± 2.14 kg/m2
  • Self-assessment questionnaires:
    • 67% (i.e., 8/12) of patients felt their change in facial appearance was favourable
  • Panel assessment:
    • 92% (i.e., 11/12) of patients were given preference in their post-operative profile from panellists

None

AHI = Apnea-Hypopnea Index; BMI = body mass index; MMA = maxillomandibular advancement; NR = not reported; OSA = obstructive sleep apnea; SD = standard deviation; VAS = visual analogue scale.
a Rated on a 10-point scale, 1 representing the least attractive and 10 representing the most attractive.


3.3. Summary of Clinical Review Results

An overview of SRs, supplemented by a review of primary studies, was conducted on the clinical effectiveness, comparative clinical effectiveness, and safety of PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults. Evidence was found on inactive comparisons with all interventions of interest. The effects of CPAP, MADs, and undefined OAs were most commonly reported. Limited evidence was found on active comparisons. Nevertheless, three NMAs were identified that compared CPAP and MADs or CPAP, MADs, diet, and exercise. In terms of outcomes, EDS and OSA severity were most commonly reported.

For the overview of SRs, the quality of the primary studies included and assessed by the SRs ranged widely, from very low to high and mixed. The risk of bias of the SRs themselves assessed in this report was generally low, with moderate or high concerns raised on one domain of the ROBIS tool (i.e., specifications of study eligibility criteria). Among the three NMAs, one NMA was missing information (e.g., no pairwise results were presented) and analyses (e.g., no handling of inconsistency or conducting subgroup or meta-regression analyses) that were deemed important by the ISPOR questionnaire. For the review of primary studies, the RCTs were at low or unclear risk of bias, and the non-randomized controlled or pre-and-post studies were generally rated as having low risk. However, sample sizes of 20 or fewer patients in 15 of the 41 studies and the pre-and-post design used in 30 of the 41 studies suggest that the findings need to be interpreted with caution.

Based on the analysis of the primary outcome, CPAP, EPAP, MADs, TRDs, undefined OAs, MMA, GTA, diet, exercise, and positional therapy were all effective at reducing EDS, compared with inactive controls or pre-treatment. Effect sizes across the interventions were similar, although it is unclear whether reductions of two ESS score points are clinically significant. In fact, a survey of the literature on what constitutes a clinically significant change in ESS scores did not find any relevant information (Appendix 15). The similarity in effect sizes across the interventions was also reflected in the findings on active comparisons, where no significant differences in ESS scores were found between some interventions, except for severe cases of OSA who may benefit more from CPAP than from MADs. However, based on the analysis using OSA severity as the outcome, effect sizes varied across the interventions, with CPAP showing the largest effect, with mean differences around -20 events/hour. A survey of the literature on what constitutes a clinically significant change in AHI scores found five events/hour to be of significance (Appendix 15). For severe cases with OSA, who are eligible for surgery, MMA with or without GTA may be effective at improving both EDS and OSA severity, although the findings are mostly from small, uncontrolled pre-and-post studies on highly selected patients and warrant caution, considering the invasiveness of the procedure and potential adverse events. The findings on EPAP were also from uncontrolled studies. Some of the findings on ESS (i.e., CPAP or MADs versus inactive controls and CPAP versus OAs) and AHI (i.e., CPAP, EPAP, MADs, diet, or exercise versus inactive controls) were associated with high heterogeneity.

CPAP, MADs, and undefined OAs independently demonstrated similar improvements in blood pressure, although it is unclear whether reductions of 2 mm Hg are clinically significant. Diet, exercise, and positional therapy did not significantly reduce blood pressure. Diet and exercise for one year, but not CPAP for up to four months, were effective at reducing A1C levels in patients with diabetes, although it is unclear whether reductions of 0.5 units are clinically important. While CPAP was effective at improving insulin sensitivity in patients with diabetes, this finding was associated with high heterogeneity. CPAP was effective at reducing the risk of real and near-miss accidents, although the findings were from uncontrolled studies. Findings on the effects of CPAP and MADs on cognitive and psychological functions and QoL were mixed, with generally small effect sizes, if any. CPAP was effective at reducing certain types of death (i.e., caused by stroke, cardiac disease, or CVEs), but findings on all-cause mortality were mixed. No mortality was reported for MMA plus GTA, although this outcome was reported in a single study that included a small number of participants. While long lists of adverse events associated with various interventions were identified, no major adverse events were noted, except for MMA plus GTA, where all patients had complications caused by infections. Adherence was higher with MADs than with CPAP, in terms of both nightly durations and numbers of nights of use, and discontinuation was reported as 16% and 32% of patients after one and four years of use, respectively. From uncontrolled studies, improvements in snoring were reported for EPAP and MADs, but the findings were inconsistent for positional therapy. No evidence on fatigue was identified. Facial aesthetics were deemed more favourable after MMA compared with before. Findings on combination therapy suggest that various interventions in combination may have additive effects in their effectiveness in improving EDS, OSA severity, and blood pressure. Some of the findings on blood pressure (i.e., CPAP versus inactive controls), insulin sensitivity (i.e., CPAP versus inactive controls), accident (i.e., CPAP versus inactive controls), cognitive functions (i.e., CPAP versus inactive controls), and QoL (i.e., CPAP versus OAs) were associated with high heterogeneity.

Evidence on subgroups of interest, including different comorbidities and OSA severity, was found on some of the intervention-comparator-outcome combinations. Patients with hypertension or resistant hypertension experienced greater effects on blood pressure with CPAP. CPAP, diet, and exercise were effective at improving A1C levels or insulin sensitivity in patients with diabetes, although it is unclear whether changes were clinically significant. CPAP was effective at reducing the risk of certain CVEs (e.g., recurrent cardiac disease or AF), but not others (e.g., major adverse cardiac events or MI), and effective at reducing the risk of stroke, but not IS, in patients with previous CVD. Patients with higher EDS at baseline experienced greater effects on EDS with CPAP, MADs, or undefined OAs. Patients with more severe OSA at baseline experienced greater effects on EDS and OSA severity with CPAP, but not with MADs or undefined OAs, and were less likely to achieve success or cure with MMA. Patients with normal weight were more likely to experience improvements in snoring, compared with those who were overweight. Longer study duration was associated with lower effects on EDS and OSA severity with CPAP and MADs. Higher adherence with interventions was associated with higher effects of CPAP and weight loss on blood pressure and higher effects of CPAP on CVEs.

 

4. Economic Evaluation

This section addresses Research Question 2: What is the cost-effectiveness of PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA in adults?

4.1. Review of Economic Studies

A review of the published and grey literature was conducted to identify relevant economic evaluations that assessed therapies for sleep apnea. Thirteen unique economic evaluations were identified that addressed the cost-effectiveness of at least one or more OSA treatments of interest.28,39,120-130 Appendix 18 provides details on each economic evaluation.

The applicability of these studies in addressing the research question of interest in this review is limited. None of the economic evaluations reviewed all therapies of interest but rather several compared one treatment modality against no treatment,39,121,122,125-128 which does not reflect the multiple treatment options that are available for patients with OSA. In addition, only two economic analyses addressed surgical procedures and were focused on either upper airway stimulation128 or palatopharyngeal reconstructive surgery.130 Of particular interest to this review are the surgical interventions involving MMA with or without GTA, for which no economic evaluation has been conducted. Some of the economic models adopted a narrower scope by characterizing only the association between OSA and MVAs39,126 or OSA and cardiovascular morbidity or mortality.121 As such, they do not consider the full range of clinical evidence available to estimate the economic impact of treatments for OSA, nor do they capture the broader health consequences of OSA in the pathogenesis of several major conditions.

Few economic evaluations were conducted with a Canadian setting.39 The analysis by Tan et al.39 compared only the effects of CPAP against no treatment on MVAs, making it difficult to adapt in order to address the research question in this review. Outside of the project's scope, one Canadian economic model131 was identified that explored the economic impact of providing decision aids to patients with OSA. This study evaluated the relationship between decision aids to reduce rates of discontinuation from either CPAP or MADs, and concluded that the cost-effectiveness of patient decision aids will depend on contextual factors. However, it also highlighted the importance of discontinuation as a factor that affects the cost-effectiveness of treatments for OSA.

Overall, existing economic evaluations did not fully address the cost-effectiveness of treatments among different patient populations, with the exception of patients with type 2 diabetes122 and patients intolerant to CPAP.129 This is of particular interest given that the clinical benefit of treatments may vary depending on patient characteristics, resulting in differences in cost-effectiveness.

Given that no Canadian model was identified that would fully address Research Question 2, a de novo economic model was constructed. Identified economic models provided insights in developing the model structure, determining appropriate model assumptions, and as a source for data inputs relating to utilities and disease prognosis.

4.2. Primary Economic Evaluation

4.2.1. Methods

The objective of the economic analysis was to evaluate the cost-effectiveness of various treatments for adult patients with OSA, across different disease severities, within a Canadian health care system over a patient's lifetime. In addition, the cost-effectiveness in specific subgroups was explored, including sex, age, hypertensive patients, smokers, and diabetic patients. A protocol for the economic evaluation was written a priori and followed in the conduct of this review.

1. Type of Analysis

A cost-utility analysis was conducted given the broad set of adverse clinical outcomes associated with OSA. Health outcomes were expressed as quality-adjusted life-years (QALYs), which captures both the mortality and morbidity impacts related to the condition and its treatments. The primary outcome in the economic analysis was the incremental cost per QALY gained (commonly referred to as the incremental cost-utility ratio [ICUR]).

2. Target Populations and Settings

The target population for the economic evaluation was adult Canadians with OSA, aligned with the clinical population identified from the clinical review. The base-case cohort represented patients aged 55 years (76.5% males), none of whom were current smokers or had diabetes. A stratified economic analysis was conducted by disease severity, as defined by the baseline AHI value, into mild (AHI < 15), moderate (15 ≤ AHI < 30) and severe OSA (AHI > 60). Unless specified, it was assumed that, for every analysis, the patients in the cohort comprised only those eligible for all the interventions assessed. It is important to note that not all patients with OSA may be eligible or suitable for all interventions that were being considered. Separate analyses were conducted exploring cases where the patient cohort may not be suitable for a particular intervention by excluding that intervention from the analysis.

The subgroup analyses undertaken included:

  1. Sex: female vs. male
  2. Age
  3. Blood pressure: all patients were hypertensive
  4. Smoking: all patients currently smoked
  5. Diabetes: all patients were diabetic
3. Time Horizon and Discount Rate

OSA is a chronic condition associated with considerable morbidities that manifest over a patient's lifetime. A lifetime time horizon was therefore adopted in which costs and health outcomes were discounted at a rate of 5% per annum as per Canadian guidelines.132 Sensitivity analyses were conducted with a range of discount rates at 0%, 1.5%, and 3%. In addition, sensitivity analyses were further conducted over various modelled time horizons (i.e., two years, seven years, and 10 years).

4. Interventions

Interventions for OSA of interest included C/APAP devices, EPAP, MAD, MMA with or without GTA, and lifestyle modifications. In addition, a no-treatment strategy was modelled, reflecting the natural progression of OSA for untreated patients.

PAP-Based Devices: CPAP and APAP

Feedback from clinical experts suggest that CPAP and APAP devices can be grouped together as a single class of interventions, given similarities in their clinical effectiveness and safety profile. Together, these are referred to as PAP therapy throughout the economic section of this report. The clinical effectiveness was based on a pooled estimate reported in the clinical review while costs were weighted according to the assumption that 19% of patients received CPAP while the remainder received APAP.35 It is important to note that, although APAP and CPAP were grouped together, the cost of testing for these different treatments may differ but was not considered in the economic model, given that this was considered to be outside the scope of this report.

BiPAP was excluded from the analysis given that it is primarily prescribed to a subset of patients with specific needs (e.g., those with cardiopulmonary disorders).13,14

EPAP

Nasal EPAP is a single-use device consisting of a small valve attached externally to each nostril, which act as one-way resistors to permit unobstructed inspiration. Utilizing a patient's own breathing, PAP is created during each expiration to prevent obstruction to breathing. Given that the mechanism of action is different from other PAP therapies, EPAP was considered as a separate strategy.

Oral Appliances

Oral appliances include MAD and TRD. Although the manner in which it prevents or minimizes upper airway obstruction differs among the classes, the clinical review found that the clinical effectiveness between oral devices were similar. Costing analysis between the two devices further suggests that the price and resource consumption for each type of oral appliance is similar. They primarily differ in adherence as, in one study comparing between the two types of oral appliances, the incidence of patients involuntarily removing MAD during the night was 9% compared with 86.4% for TRD.133 As TRD is known to be inferior to MAD, given a similar clinical effectiveness profile and similar costs but lower adherence, MAD was modelled as the oral appliance of interest. Although both over-the-counter and customized oral appliances are commercially available, the former was not considered because patients diagnosed with OSA and seeking treatment from their health care providers would be predominately offered a customized oral appliance based on clinical experts' feedback.

Surgery

Of the numerous surgical interventions available for patients with OSA, oropharyngeal interventions have become increasingly less utilized due to its poor rates of success. According to clinical experts, orthognathic surgery (e.g., MMA with or without GTA) has replaced oropharyngeal surgery and, to reflect current treatment trends, the economic model focused solely on modelling MMA with or without GTA. In practice, the appropriate surgical procedure for OSA depends on the site of the anatomical obstruction and the patient's anatomical features. However, to model a cohort, it was assumed that the majority of patients (90%) would be suitable surgical candidates for MMA, while the remaining would be candidates for GTA. This assumption was based on clinical experts' feedback and the available literature. Among patients receiving MMA, 20% of patients would receive adjunctive GTA.

Lifestyle Modification

Lifestyle modification involves counselling patients with regard to lifestyle advice or behavioural changes in order to reduce and/or to cope with symptoms. Often, the objective of these interventions is to target specific patient characteristics that are also factors that contribute to OSA. As such, lifestyle modifications are suitable interventions for particular subgroups of patients and will be explored in subgroups analysis rather than in the base case. Strategies to modify lifestyle, assessed in this economic evaluation, reflect the interventions studied in the clinical review. This includes interventions for weight loss (via diet or exercise) or changes to sleep position (via positional therapy). In the economic analysis, weight reduction was considered a suitable treatment option in overweight or obese OSA patients, whereas positional therapy was considered a suitable treatment in patients with positional OSA (i.e., defined arbitrary as supine AHI at least twice of non-supine AHI).

Given the results of the clinical review, only single therapies were included in the economic analysis with the exception of surgery (e.g., MMA with or without GTA). A scenario analysis was conducted to explore the economic value of a weight-loss intervention when compared with the interventions included in the base case for obese patients with mild-to-moderate OSA. As the clinical review did not identify any studies that reported on the effects of EPAP or positional therapy on blood pressure, the base case compared PAP therapy, MADs, and MMA with or without GTA, with an exploratory analysis that incorporated EPAP or positional therapy. Table 87 summarizes the set of interventions that was compared in each analysis.

Table 87: Interventions Compared in the Analysis

 

PAP Therapy

Nasal EPAP

MAD

MMA ± GTA

Weight Loss

Positional Therapy

No Treatment

Base case

 

   

Scenario Analysis

 

 

Exploratory

 

EPAP = expiratory PAP; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; PAP = positive airway pressure.

5. Perspective

This analysis was conducted from the perspective of the publicly funded health care system. As such, direct medical costs were captured that included the cost of the medical devices, laboratory tests, emergency visits, in-patient visits, medical services, and physician fees for services covered in provincial fee schedules. Indirect costs, such as productivity losses, out-of-pocket costs, and time lost, were not included. Only medical costs pertaining to an MVA were included. Given the varying coverage for PAP therapy and dental procedures across Canada (Appendix 17), a secondary analysis was undertaken exploring different reimbursement schemes for oral appliances.

6. Decision-Analytic Model

A decision-analytic Markov model was developed to assess the effectiveness and costs of each strategy (i.e., intervention or no treatment). The clinical pathway and decision-analytic model were developed by reviewing existing clinical and economic literature. It was further validated by clinical experts from a range of medical specialities (i.e., respirology, dentistry, maxillofacial surgery) with experience managing patients with OSA.

Figure 1 presents the clinical pathways of OSA, which include the associated conditions (i.e., stroke, MI, MVA, death) that have a suspected causal relationship to OSA and are considered important from both a clinical and a policy-making viewpoint. Although there is literature emerging that suggests a relationship between OSA and other conditions, such as depression or diabetes,5,119 the clinical review reported mixed findings or results applicable to a particular patient subgroup. Given that no clear treatment effect differences were observed, these health states were not captured in the economic model. Appendix 19 presents the decision-analytic model conceptualized from the clinical pathway presented above.

In the base case, all members of the cohort (i.e., 55 years of age, 76.5% males) start in the "at-risk" state regardless of the strategy (i.e., diagnosed with OSA but are event free). They can remain in that state, unless a transition occurs into one of a CVE health state or death. In each one-year cycle of the model, patients could experience a CVE or MVA, die from other causes, or remain event free. MVA was classified into property damage without injury, injury, or death. Patients who survive an MVA return to the health state of the previous cycle. Individuals are assumed to stop driving at the age of 80, based on clinical experts' opinions, or upon experiencing a severe disabling stroke. CV events included MI and stroke (mild-to-moderate or severe) and could be fatal or non-fatal. Patients who survive the CV event progress into either the post-MI or post-stroke state given the increased morbidity and mortality associated with a history of CV disease. From these states, patients can remain until death or until another CVE occurs. Death is the absorbing health state.

Figure 1: Clinical Pathway Diagram Describing the Relationship Between OSA and the Pathogenesis of Cardiovascular, Cerebrovascular Morbidity, and Motor Vehicle Accidents.

Treatment module

AHI = Apnea-Hypopnea Index; MI = myocardial infarction.

Note: Arrows link variables that are related. Green arrows indicate a positive relationship (i.e., all else remaining equal, an increase in the source variable will lead to an increase in the resulting variable), while red arrows suggest a negative relationship. Adherence modulates the relationship between treatment and AHI value and, thus, indirectly affects clinical outcomes.

a In the model, a condition is incorporated whereby patients no longer experience motor vehicle accidents at the age of 80 or upon experiencing a severe disabling stroke.

Surgery differs from other treatments primarily in how AHI is incorporated into the model. The clinical review found that the treatment effects for AHI for non-surgical interventions are typically reported as a continuous measure, while surgery is reported by the following categorical definition: cure (post-operative AHI < 5), response (post-operative AHI ≤ 15 and at least a reduction in AHI by 50%) or failure.60 The outcome for surgery in terms of AHI is categorized into these three states, which affects progression.

The model runs based on yearly transitions. All analyses were conducted using Microsoft Excel 2010.

7. Clinical Inputs

Disease Natural History

Mortality

Mortality due to other causes, by age and sex, was calculated by subtracting MI, stroke, and MVA mortality from general population mortality estimates reported in Statistics Canada lifetables.134

Existing literature suggests that the risk of all-cause mortality is graded in patients with OSA based on disease severity (Table 88).135,136 Specifically, there is a statistically significant increase in all-cause mortality in patients with severe OSA compared with no OSA, even after adjustment for the increased risk in cardiovascular mortality. A similar trend, although not statistically significant, is observed in patients with moderate OSA.137 It is speculated that the increased risk is related to pathophysiological changes associated with OSA. Patients' risk of mortality is adjusted by disease severity using a recently published meta-analysis by Pan et al. (based on 12 prospective cohort studies of 34,382 patients with OSA).137

For patients with history of CVD, the model similarly adjusts for an increased risk of death. The Danish Monitoring Trends and Determinants in Cardiovascular Disease cohort study reported long-term survival following MI and stroke. To reflect the findings from this longitudinal study, a lifelong and a five-year post-event mortality risk was applied for patients with a history of stroke or MI, respectively.138,139

The relative risk of mortality following treatment was determined based on the post-treatment AHI value achieved. A sensitivity analysis was further conducted that assumed that patients adherent on treatment would have their mortality risk return to general population (i.e., non-OSA) rates, regardless of post-treatment AHI value. This reflects the findings of several observational studies that have reported that the mortality risks of patients with moderate-to-severe OSA who tolerated CPAP were found to return to rates similar to those of patients without OSA.140,141

Table 88: Relative Risk of Mortality by Health Condition

Health Condition

Variable Name

(in the Model)

RR (95% CI)

Ref

Severe OSA

rrDeath_SevereOSA

1.601

(1.298 to 1.902)a

Pan, 2016137

Moderate OSA

rrDeath_ModerateOSA

1.178

(0.978 to 1.378)a

Pan, 2016137

Mild OSA

rrDeath_MildOSA

0.945

(0.810 to 1.081)a

Pan, 2016137

History of MI (first 5 yrs)

Males aged 30 to 59, yr 1 to 5

rrDeath_HistoryOfMI_Yr1to5_Male

3.22

(2.65 to 3.87)b

Bronnum-Hansen, 2001139

History of MI (first 5 yrs)

Males ≥ 60 age, yr 1 to 5

rrDeath_HistoryOfMI_Yr1to5_Male_Age60

1.95

(1.69 to 2.24)b

History of MI (first 5 yrs)

Females aged 30 to 59, yr 1 to 5

rrDeath_HistoryOfMI_Yr1to5_Female

4.41

(2.79 to 6.61)b

History of MI (first 5 yrs)

Females ≥ 60 age, yr 1 to 5

rrDeath_HistoryOfMI_Yr1to5_Female_Age60

2.75

(2.18 to 3.44)b

History of stroke (lifelong)

Males aged 25 to 69

rrDeath_HistoryOfStroke_Male

3.01

(2.63 to 3.43)b

Bronnum-Hansen, 2001138

History of stroke (lifelong)

Males ≥ 70 age

rrDeath_HistoryOfStroke_Male_Age70

1.92

(1.68 to 2.18)b

History of stroke (lifelong)

Females aged 25 to 69

rrDeath_HistoryOfStroke_Female

3.52

(2.80 to 4.35)b

History of stroke (lifelong)

Females ≥ 70 age

rrDeath_HistoryOfStroke_Female_Age70

2.05

(1.81 to 2.30)b

CI = confidence interval; MI = myocardial infarction; OSA = obstructive sleep apnea; RR = relative risks; yr = year.
a Measure reported in the publication as a hazard ratio.
b Measure reported in the publication as standardized mortality ratios.

Cardiovascular Outcomes

The model classified patients into either normotensive or hypertensive according to age- and sex-specific incidence and prevalence rates for hypertension based on Canadian sources.142,143 In the Sleep Heart Health Study (SHHS; n = 2,470),144 an elevated risk of developing hypertension was observed among patients with OSA who were normotensive at baseline when compared with those without OSA after a mean follow-up of 5.2 years. This risk of hypertension was proportional to the disease severity and was applied to this model (Table 89).144

Table 89: Model Parameters Pertaining to Incident Cardiovascular Event

Event

Variable Name

(in the Model)

Model Value

(95% CI)

Ref

Baseline data

Incidence rate of hypertension

(Age- and sex-dependent)

CCDSS, 2009145

Prevalence of hypertension

(Age- and sex-dependent)

CCDSS, 2009145

Hypertension

RR hypertension in male patients with mild OSA

rrHypertens_MaleMildOSA

1.19

(0.92 to 1.53)

Calculated from O'Connor, 2009144

RR hypertension in male patients with moderate OSA

rrHypertens_MaleModerateOSA

1.61

(0.76 to 3.40)

RR hypertension in male patients with severe OSA

rrHypertens_MaleSevereOSA

1.65

(0.69 to 3.96)

RR for hypertension in female patients with mild OSA

rrHypertens_FemaleMildOSA

1.37

(0.70 to 2.70)

RR for hypertension in female patients with moderate OSA

rrHypertens_FemaleModerateOSA

1.79

(0.80 to 4.01)

RR for hypertension in female patients with severe OSA

rrHypertens_FemaleSevereOSA

1.90

(0.73 to 4.92)

MI

RR of MI occurrence, severe OSA

rrMI_SevereOSA

1.21

(0.75 to 1.96)

Wang, 2013146

Probability of 28-day mortality

Male

(Age- and sex-dependent)

0.335 to 0.479

Smolina, 2012147

Probability of 28-day mortality

Female

0.331 to 0.506

RR MI is fatal in incident yr:

Males aged 30 to 59

rrDeath_IncidentMI_MaleUnder60

5.84

(4.26 to 7.81)b

Bronnum-Hansen, 2001139

RR MI is fatal in incident yr:

Males ≥ 60 age)

rrDeath_IncidentMI_MaleOver60

4.04

(3.31 to 4.89)b

RR MI is fatal in incident yr:

Females aged 30 to 59)

rrDeath_IncidentMI_FemaleUnder60

16.8

(10.3 to 25.9)b

RR MI is fatal in incident yr:

Females ≥ 60 age

rrDeath_IncidentMI_FemaleOver60

6.22

(8.38 to 4.50)b

RR of MI being fatal, severe OSA

rrDeath_CV_SevereOSA

2.87

(1.17 to 7.51)a

Marin, 20156

Stroke

RR of stroke occurrence, severe OSA

rrStroke_SevereOSA

2.15 (1.42 to 3.24)

Wang, 2013146

Proportion of severe stroke

propSevereStroke

0.12

(0 to 0.36)

Anis, 2006148

Probability of 28-day mortality

Male

(Age- and sex-dependent)

0.181 to 0.338

Bronnum-Hansen, 2001138

Probability of 28-day mortality

Female

0.226 to 0.383

RR stroke is fatal in incident yr:

Males aged 25 69

rrDeath_IncidentStroke_MaleUnder70

4.64

(3.71 to 5.72)b

RR stroke is fatal in incident yr:

Males ≥ 70 age

rrDeath_IncidentStroke_MaleOver70

3.70

(3.15 to 4.32)b

RR stroke is fatal in incident yr:

Females aged 25 69

rrDeath_IncidentStroke_FemaleUnder70

9.27

(6.94 to 12.1)b

RR stroke is fatal in incident yr:

Females ≥ 70 age

rrDeath_IncidentStroke_FemaleOver70

5.18

(4.54 to 5.87)b

RR of stroke being fatal, severe OSA

rrDeath_CV_SevereOSA

2.87

(1.17 to 7.51)a

Marin, 20156

CCDSS = Canadian Chronic Disease Surveillance System; CI = confidence interval; MI = myocardial infarction; OSA = obstructive sleep apnea; RR = relative risk; yr = year.
a Measure reported in the publication as an odds ratio. Given low event risks (< 10%), the odds ratio is assumed to closely approximate relative risk.149
b Measure reported in the publication as standardized mortality ratios.

A principal morbidity associated with OSA is CV complication, as the rates are higher than in the general population.150,151 The clinical review identified only a few studies evaluating the effect of CPAP on stroke and MI, although many studies reported on the effects of treatment on blood pressure changes. To reflect the data available from the clinical review, the Framingham risk equation was used to estimate the probability of stroke or MI given changes to blood pressure.152 A series of published risk equations was estimated separately for men and women using the baseline characteristics shown in Table 90.. Although the equation can be a function of either systolic or diastolic blood pressure, systolic blood pressure was selected in this model given that it has been found to be a more reliable predictor for stroke.152

Table 90: Hypothetical Baseline Characteristics for Patients in the Base-Case Scenario to Estimate Risk of Cardiovascular Events, Based on the Framingham Equation

Inputs

Age

55

SBPa

118

Smoking ( 0 = no; 1 = yes)

0

Total cholesterola

224

HDL cholesterola

43

Diabetes (0 = no; 1 = yes)

0

ECG-LVH

0

Outputs

10-year probability of MI

16.3 (males); 15.9 (females)

10-year probability of stroke

8.1 (males); 6.7 (females)

ECG-LVH = electrocardiographic left ventricular hypertrophy; HDL = high-density lipoprotein; MI = myocardial infarction; SBP = systolic blood pressure.
a Baseline SBP, total cholesterol, HDL were taken from Weatherly.124

It was assumed that OSA patients diagnosed with hypertension would be treated with medication, targeting a return to blood pressure levels at entry into the model. In such a circumstance, the Framingham risk equation would predict that patients with hypertension managed by medication would have similar risks as a patient without hypertension in the model. However, a meta-analysis of patients with OSA reported that, in patients with severe OSA, the risk of stroke (relative risk [RR]: 2.15; 95% CI, 1.42 to 3.24) and MI (RR: 1.21; 95% CI, 0.75, 1.96) increased independent of blood pressure.146 Therefore, in patients with severe OSA, the probability of MI and stroke predicted by the Framingham risk equation was adjusted to reflect this higher risk.

The probability that a CV event was fatal reflects the findings from the Danish Monitoring Trends and Determinants in Cardiovascular Disease cohort study. These data reflect the risk of CV death in a general population. From a long-term observation of CV outcomes in men with OSA, a higher incidence of fatal CV events (defined as fatal MI and stroke) was observed in patients with untreated severe OSA compared with healthy participants (adjusted OR: 2.87; 95% CI, 1.14 to 7.51). The risk of a fatal CV event in patients untreated with mild-to-moderate OSA was found to be at levels similar to a healthy population.

Patient prognosis following a non-fatal CV event was based on several long-term observational studies.147,153-155

Table 91 lists the inputs to the model relating to subsequent CV event risk following an incident event.

Table 91: Model Parameters Pertaining to Prognosis Following Cardiovascular Event

Event

Variable Name

(in the Model)

Model Value

(95% CI)

Ref

RR of Stroke Following Incident MI

Year ≤ 1

rrStroke_Yr1PostMI

3.1

(2 to 4.5) a

Witt, 2005153

1 < Year ≤ 4

rrStroke_SubYrPostMI

1.6

(0.9 to 2.8)a

RR of Stroke Following Incident Stroke

Year 1

rrRecurrentStroke_Yr1PostStroke

15.4

(12.1 to 19)a

Burn, 1994155

Year 2

rrRecurrentStroke_Yr2PostStroke

8.5

(5.6 to 11.8)a

Year 3

rrRecurrentStroke_Yr3PostStroke

6.7

(3.9 to 10.7)a

Year 4

rrRecurrentStroke_Yr4PostStroke

4.5

(2.1 to 8.6)a

Yearly Probability of MI After Incident MI

Year ≤ 1: males

pRecurrentMI_Yr1PostMI_Male

0.056

(0.055 to 0.057)

Smolina, 2012147

Year ≤ 1: females

pRecurrentMI_Yr1PostMI_Female

0.072

(0.071 to 0.074)

1 < Year ≤ 5: males

pRecurrentMI_SubYrPostMI_Male

0.0187

(0.0183, 0.0187)

Converted 4-year probability to yearly probability from Smolina, 2012147

1 < Year ≤ 5: females

pRecurrentMI_SubYrPostMI_Female

0.021

(0.020 to 0.021)

RR of MI Following Incident Stroke

Year ≤ 5

rrMI_PostStroke_Yr5

1.51

(1.12 to 2.01)b

Ducrocq, 2013154

CI = confidence interval; MI = myocardial infarction; RR = relative risk.
a Measure reported in the publication as standardized morbidity ratios.
b Measure reported in the publication as a hazard ratio.

In summary, treatment affects final CV outcomes through its effects on AHI and blood pressure. Alternative assumptions were examined in the sensitivity analyses. For instance, in alignment with the observations of several long-term cohort studies investigating CV risks in untreated OSA, treated OSA, and controls without OSA, it was assumed that, in patients perfectly adherent to PAP-based therapy, the risks of MI, stroke, and incident hypertension would return to general population levels.146,156 As no data were present for the other interventions, a similar assumption was made for those adherent to other therapies.

Motor Vehicle Accidents

The annual probability of MVAs in the general population was based on the findings from the Ontario Road Safety Annual Report (2012).157 MVAs were stratified by the following classes of accidents: property damage without injury, personal injury, or death (Table 92). Because the proportion of OSA in the general population is relatively small, these rates were assumed to apply to a general population without OSA.

Current research has suggested that both OSA severity and EDS lack a clear relationship to road traffic accidents, and are unreliable predictors of collision risk.158,159 Rather, the majority of data on MVAs come from before-and-after studies, in which the rates of MVAs before treatment are compared with the rates of MVA after treatment. The clinical review identified a meta-analysis of such studies by Antonopoulos et al.47 based on 1,221 patients, in which the odds of an MVA in patients on CPAP therapy are 80% less than in the pre-treatment period (OR: 0.21; 95% CI, 0.12 to 0.35). This was assumed to be the increased risk of MVA in untreated patients.47 The MVA rate in OSA patients adherent to treatment was assumed to be comparable to that in the general population, while those non-adherent to treatment would have half the increased risk of MVA. Table 92 summarizes the model parameters pertaining to MVA.

Patients left disabled following a severe stroke event (12.3%) were assumed to no longer be at risk of a MVA. The base-case analysis assumed all patients could drive, with a sensitivity analysis conducted assuming no patients drive, whereby the risk of MVA was removed from the model.

Table 92: Model Parameters Pertaining to MVA Outcomes

Event

Variable Name

(in the Model)

Model Value

(95% CI)

Ref

General

Probability that OSA patient drives

pDriver

1

Assumption

Average age that patients stops driving

sAgeStopDriving

80

Assumption

Relative risk pertaining to OSA

RR for MVA, for untreated OSA

rrMVA_UntreatedOSA

3.59 (2.86 to 4.54)

Tregear, 2010160

RR for MVA, treated OSA

 

1

Assumption

Distribution of MVAs

Property damage only, male

distMVA_PDO_Male

0.74 (0.7 to 0.78)

Ontario Road Safety Annual Report, 2012157

Personal injury, male

distMVA_PI_Male

0.26 (0.23 to 0.28)

Fatal, male

distMVA_Fatal_Male

No distribution assigneda

Property damage only, female

distMVA_PDO_Female

0.71 (0.66 to 0.76)

Personal injury, female

distMVA_PI_Female

0.29 (0.26 to 0.32)

Fatal, female

distMVA_Fatal_Female

No distribution assigneda

CI = confidence interval; MVA = motor vehicle accident; OSA = obstructive sleep apnea.
a Distributions must sum to 1. As such, the value for fatal MVA is based on the subtracting distribution for property damage and personal injury from 1.

Treatment Effect

For non-surgical interventions, treatment effects (reported as mean reduction in AHI and blood pressure) were taken directly from the clinical review and incorporated into the economic model to determine the post-treatment AHI and blood pressure values. This was subsequently taken to calculate event risks. AHI was selected as the clinical outcome in which to model long-term health consequences given the considerable research on the predictive validity of this objective measure of disease severity (Appendix 15). Although an NMA was available for the outcome of AHI that compared no treatment, PAP therapy, oral appliances, and lifestyle modification,161 it was not selected, as the clinical review had found meta-regression and subgroup analyses that suggested that treatment effect on AHI reduction is a function of the baseline disease severity. A meta-analysis that provided naive estimates without incorporating indirect evidence was instead selected, given that it reported treatment effects by baseline disease severity for PAP therapy and oral appliances.55 In comparing the estimates from the NMA161 to the meta-analysis,55 the NMA treatment effect estimates fell within the 95% CI of the meta-analysis estimates for patients with severe OSA. As the inclusion of lifestyle will be conducted as a scenario analysis for obese patients with mild-to-moderate OSA, the treatment effects were taken from meta-analyses of trials that evaluated this patient population specifically.5,76

As discussed previously, surgery differs in how the AHI outcomes are incorporated as, conventionally, studies have reported this outcome categorically based on the achieved post-surgical results. The meta-analysis by Zaghi et al.60 reported treatment response based on patients' baseline AHI value. Treatment response was incorporated into the economic model as follows: patients classified as cured were mapped as a non-OSA population, responders were mapped to an untreated or mild OSA population, and surgical failure was mapped to baseline disease severity. The effect of surgery on blood pressure, a continuous variable, was incorporated into the model in a similar fashion as the other strategies.162 However, given suggestions from clinical experts that these studies on surgery may be biased by recruiting patients with more severe OSA and other factors that would favour a more likely successful outcome, a sensitivity analysis was conducted exploring smaller treatment effects by setting it to the lower bounds of the 95% CI.

Given that no data were available on the effect of EPAP and positional therapy in reducing blood pressure, an exploratory analysis was conducted that assumed a least conservative (i.e., optimistic) scenario whereby the effect on blood pressure would equal the highest non-surgical treatment effect observed (i.e., PAP therapy).

Table 93: Treatment Effect Measures for Patients Remaining on Treatment and Who Are Perfectly Adherent - Base Case and Exploratory Scenario

Analysis

Interventions

Mean Reduction in AHI, vs. Inactive Control (95% CI)

Mean Reduction in SBP, vs. Inactive Control (95% CI)

Base-case analysis

PAP-based therapy55,63

 

Baseline disease severity (AHI)

Mild

Moderate

Severe

Mean reduction in AHI

2.4 (-3.67 to -1.13)

-13.67 (−16.13 to −11.2)

-33.04 (-39.75, −26.34)

-2.5 (-1.5 to -3.5)

MAD55,63

 

Baseline disease severity (AHI)

Mild

Moderate

Severe

Mean reduction in AHI

-7.79 (−16.38 to 0.79)

-10.72 (−14.59 to −6.85)

-7.95 (−15.94 to 0.05)

-2.1 (-0.8 to -3.4)

Surgery60,162

Treatment response

Probability

(based on baseline AHI)

< 30

30 < AHI < 60

60 < AHI < 90

AHI > 90

Cure

(post-surgical AHI < 5)

0.557

0.458

0.28

0.195

Success

(post-surgical AHI < 15 and 50% reduction in baseline AHI)

0.836

0.88

0.727

0.707

-3.5 (-19 to 11)a

Scenario analysis

Weight loss5,76

 

Baseline disease severity (AHI)

Mild

Moderate

Severe

Mean reduction in AHI

-1.99 (-5.58 to 1.61)

-9.08 (−12.87 to -5.30)

-4.91 (−21.97 to 12.15)

-0.6

(-8.4 to 7.2)

Exploratory analysis

EPAP68

-14.78

(-10.45, -19.12)

No data available

Estimate: -2.5 (-1.5 to -3.5)

Positional therapy163

-4.6

No data available

Estimate: -2.5 (-1.5 to -3.5)

AHI = Apnea-Hypopnea Index; CI = confidence interval; EPAP = expiratory PAP; MAD = mandibular advancement device; PAP = positive airway pressure; SBP = systolic blood pressure.
Note: Severity of OSA is defined as: mild (5 ≤ AHI < 15), moderate (15 ≤ AHI < 30) and severe (AHI ≥ 30).
a Assume this estimate applies only to patients who are at least considered "treatment success."

In cases where different studies were combined, the baseline clinical characteristics were compared across studies to ensure similarity in patient population (Appendix 20). Some differences were revealed across studies. For instance, Balk et al.5 reported on the effect of a weight-reduction program in which the inclusion criteria specified for obese patients with mild forms of OSA (AHI: 5 to 15). As a result, weight reduction was conducted as a separate analysis with only mild-to-moderate OSA explored. Similarly, studies on surgical interventions tended to recruit patients with more severe forms of the condition as observed by the baseline AHI value. Given that the study population may differ, sensitivity analysis on the treatment effect estimates were conducted in which lower effectiveness was applied for surgery .

Adherence (Specific to Non-Surgical Interventions)

It is difficult to compare the relative adherence and discontinuation rates of PAP therapy, OA, and/or lifestyle intervention because of differences in the enrolled study population. The majority of studies on PAP therapies were of patients with moderate-to-severe OSA, whereas studies on OA and lifestyle modification considered milder disease severities.123

Discontinuation

With respect to discontinuation, the clinical review observed that studies that directly compared CPAP against OA found no significant difference between the two types of intervention in terms of discontinuation. The literature further suggested that patterns of discontinuation are related to disease severity, given patients' perceptions regarding the balance between benefits and side effects.5,164 As mentioned in the clinical review, the single systematic review that evaluated adherence found that higher baseline AHI was associated with greater treatment continuation with respect to CPAP devices.5 Given these observations, discontinuation rates were assumed to be a function of disease severity rather than treatment. For patients with moderate-to-severe OSA, discontinuation rates were taken from CPAP studies whereas, for patients with mild OSA, the discontinuation rates were taken from OA studies.

The clinical review identified one systematic review with four studies on the rates of CPAP continuation, of which McArdle et al.165 was considered the largest study with complete and appropriate documentation. Compared with the other studies, McArdle et al. had no obvious selection or ascertainment biases. The discontinuation rate, up to four years, for this cohort of Scottish patients with moderate-to-severe OSA (i.e., median age of 50) was incorporated into the model. Discontinuation after four years of treatment was extrapolated through one of two assumptions: i) complete discontinuation, or ii) constant adherence. An additional sensitivity analysis was undertaken using the discontinuation rates calculated by Guest et al.125 that pooled discontinuation rates from several large observational studies. In this study, 74% of patients continued using their device during the first year of treatment and, thereafter, a 3.8% exponential rate of decline per annum was applied.

For mild OSA, discontinuation rates were obtained from a long-term study by Marklund et al.166 in which 450 consecutively recruited patients treated with MAD were followed for an average of 5.4 years. In this study, 72.2% of patients continued treatment in the first year and, by the fifth year, 52.4% of patients remained on treatment. In sensitivity analyses, the discontinuation rates from two other studies with a smaller sample size were considered: Walker-Engstrom et al.,167 who reported lower rates of discontinuation contrasted by the higher drop-off rates from Izci et al.168

As noted, it was modelled that patients who discontinued treatment would return to the levels of CV, stroke, and MVA risks that were associated with no treatment. In addition to the non-treatment-specific discontinuation rates reported above, it was further assumed 10% of patients (4.5% to 32.4%) would initially refuse CPAP and not fill their prescription.130

Adherence

Definitions of (perfect) adherence vary by intervention and are provided in Table 94. Partial adherence captures those patients who remained on treatment but who do not meet the criteria for perfect adherence.

The rates of adherence reported were based on several short-term non-comparative studies. As the studies on adherence come from observational studies of less than a year in duration, the proportion of patients adherent to treatment was applied in the first year and assumed to remain consistent thereafter.

There is a paucity of literature on adherence and its impact on treatment effect. A simplifying assumption was required to model the relationship between adherence and event risks: perfectly adherent patients were assumed to experience the full treatment effect, while those who were partially adherent would receive half the effect. In addition, sensitivity analyses were conducted that varied the magnitude of risk reduced from 0 (i.e., partially adherent users have risks similar to those who have discontinued) to 1 (i.e., partially adherent users have treatment effects similar to perfectly adherent users).

Table 94: Definition of Perfect Adherence

 

Perfect Adherence Definition in Study

Study Duration

Adherence Rate

Ref

PAP therapya

> 4 hours per night

Median: 4 years (IQR: 2.7 to 4.4)

64%

Barbé, 2012169

MAD

> half of each night for > 4 nights/week

1 year

92%

Dieltjens, 2013170

Lifestyle modification

Ability to follow regimen

12 weeks

86%

Dobrosielski, 2015171

Position therapy

≥ 4 h per night and ≥ 5 days of usage

per week

12 weeks

89%

Benoist, 201678

IQR = interquartile range; MAD = mandibular advancement device; PAP = positive airway pressure.
a Given that no literature was identified on the adherence rates for EPAP, the same rates from PAP therapy were applied.

Rate of Relapse (Specific to Surgical Interventions)

Although some of the soft-tissue procedures historically used for OSA provide only temporary relief, most studies on orthognathic surgery report a maintenance of symptom improvement.162,172 One of the longest studies, with a mean follow-up period of 6.6 years, reported no significant change in the AHI score between the short-term (mean AHI: 9.1 ± 7.9) and long-term time intervals (10.9 ± 15.0; P > 0.05) among 16 patients studied.162 As such, for the base case, it was assumed that improvements in outcomes remained stable over time.

A sensitivity analysis was conducted given that a few smaller observational studies have reported on patient relapse. For instance, in a small study (n = 6) in which all patients were initially classified as surgical successful, one patient reported significant relapse with AHI of 43, while two other patients had their AHI > 5 at the eight-year follow-up.173 Given the paucity of literature describing the worsening of symptoms, a clinical expert involved in this review was consulted. It was suggested that, in total, 8% of patients may have worsening of symptoms over their lifetime. Given that the rates of relapse from cure to partial success and from partial success to failure are unknown, a simplifying assumption was made that would be biased against the surgery strategy: the relapse of OSA symptoms was assumed to occur immediately in the first cycle.

In summary, transition probabilities in this model are dependent on age, sex, OSA severity, therapy adherence, and/or disease history (i.e., MI, stroke).

8. Utilities

The health effects of OSA and the impacts of alternative treatments were expressed in terms of QALYs. Baseline Canadian age-specific utility values from a general population, based on the EuroQol 5-Dimensions 3-Levels questionnaire (EQ-5D-3L), were taken from Johnson et al.174 To capture the impact of moderate-to-severe OSA, general population utilities were adjusted by a factor of 0.854 to reflect the lowered QoL associated with this disease.127 No reduction in utility was assumed for patients with mild OSA.

In terms of the health conditions associated with OSA, a patient's age-specific baseline utility values were decreased by a factor of 4.7% for hypertension, 10.8% for mild stroke, 35.8% for MI, and 65.7% for severe stroke (Table 95).175-177 The literature further suggests that patients with a history of MI and stroke have a lowered QoL. As such, post-stroke and post-MI were associated with a reduction of baseline utility by a factor of 17.7% and 15.0%, respectively. Non-fatal MVA was handled differently as it was specifically applied one time, to the relevant cycle period, by subtracting the patients' utility value in that health state by the utility decrement associated with the accident, stratified based on the severity of the vehicular crash.

For joint health states, such as comorbid conditions (e.g., OSA and hypertension), the minimum approach was used to estimate the joint utility value in the base-case analysis. The minimum approach [Ui,j = Min(Ui, Uj)] predicts the value of a joint health state based on selecting the lowest utility value among each individual health state. A structural sensitivity analysis was further conducted using a multiplicative approach (Ui,j = Ui × Uj), which involved multiplying the health state for each relevant condition to determine the joint utility value.

Table 95: EQ-5D Utility Values

Description

Mean

Distribution

References

Moderate-to-severe OSA (multiplicative factor)a

0.854

Beta (α: 43.56; β: 7.44)

Calculated based on Mar, 2003127a

Pietzsch, 2015128

Health condition utility weights (multiplicative)a

Essential hypertension

0.953

Beta (α: 6751.29; β: 333.71)

Calculated based on Sullivan, 2006176a

Mild-to-moderate stroke

0.892

Beta (α: 31.16 β: 13.04)

Calculated based on Golicki, 2015177a

Severe stroke

0.343

Beta (α: 40.55; β: 109.09)

Calculated based on Golicki, 2015177a

MI

0.642

Beta (α: 296.57; β: 62.44)

Calculated based on Little, 2014175a

Post-stroke

0.8228

Beta (α: 2539.11; β: 546.89)

Calculated based on Sullivan, 2006176a

Post-MI

0.8502

Beta (α: 207.46; β: 36.54)

MVA utility decrement (additive)

MVA, full recovery

-0.0146

Beta (α: 3.56; β: 240.44)

Nyman, 2008178

MVA, partial recovery

-0.0238

Beta (α: 5.81; β: 238.19)

MVA, permanent injury

-0.0375

Beta (α: 9.15; β: 234.85)

Treatment-specific utility increment (additive; applied in sensitivity analysis. Assumed to be 0 in the base case)a

PAP therapy

0.04

Gamma (0.79, 0.05)

Chakravorty, 2011179

MAD

0.02

Gamma (0.41, 0.05)

Quinnell, 2014180

Lifestyle modification

0.007

Gamma (0.05, 0.15)

Chakravorty, 2011179

Surgery

0.04

Gamma (1, 0.04)

Assumption

EQ-5D = EuroQol 5-Dimensions questionnaire; MAD = mandibular advancement device; MI = myocardial infarction; MVA = motor vehicle accident; PAP = positive airway pressure; OSA = obstructive sleep apnea.
a The utility weight was calculated by adjusting the reported study value to the Canadian baseline utility weight, according to the mean age of studied patients.

In the base case, treatment affects utility according to the post-treatment AHI response. Studies have reported small treatment-specific differences in the change in utility scores between the baseline and post-treatment period. These utility increments ranged between 0.04 and 0.007 depending on the intervention.179,180 A sensitivity analysis was therefore conducted in which, for patients adherent on treatment, utilities were adjusted by an increment associated with their treatment. Given that no studies were found that elicited patients' preferences by EQ-5D following surgical interventions, several assumptions were necessary. In patients with surgical success, utility values would return to general population values; in patients with surgical response, utility weights would be similar to treated OSA by PAP therapy, while in patients with surgical failure, utility weights would remain as the untreated OSA value.

9. Cost Inputs

Costs in the analysis are described in Table 96, Table 97, Table 98, and Table 99. Based on the perspective of the analysis (i.e., public health care payer), only direct medical costs were considered. Whenever possible, the most current Canadian cost estimates were used. If Canadian costs were unavailable, costs would be estimated from the medical literature from comparable health systems. Conversion of currency was conducted using the Bank of Canada currency converter.181 All cost estimates were adjusted to 2016 Canadian dollars using the health care component of the consumer price index inflation calculator from the Bank of Canada.182

Cost in which the resource had equal utilization across all strategies were omitted from the analysis.

Device Costs

PAP-Based Therapy

As PAP therapies are equally efficacious, these treatments differ primarily on costs. A weighted cost of PAP-based therapy was calculated. For APAP devices, the positive pressure levels applied to the patient change throughout sleep. It is less resource intensive as a titration sleep study is not required because the device automatically adjusts pressures in response to the measured physiological signals. In contrast, CPAP machines require manual titrations, with the majority of patients undergoing a split night in-lab polysomnogram in which sleep apnea is first diagnosed and then a titration test is performed to determine the optimal airway pressures during a single night. In such cases, the cost of titration can be omitted. However, treatment pressure is not always obtained on the first night and, in certain circumstances, a separate titration study may be required on a different day. It was assumed that 10% of patients on CPAP would require an additional titration study to determine their optimal positive pressure levels. In both cases, the device life was assumed to be seven years.

The annual cost of PAP therapy was determined through a micro-costing approach. It captured fixed costs such as the PAP machine alongside the cost of any required refills. Based on the device monographs, it was estimated that yearly replacements would include a new mask, tubing, head gear, and filters.183 Alternative refill requirements were further explored.183 In the base-case analysis, it was assumed that all costs relating to PAP therapy would be covered. Sensitivity analyses were conducted based on a different funding model for the CPAP device (Appendix 17).

Table 96: Price of PAP-Based Therapy

 

Details (Unit Price)

Cost, $

Reference

First-year APAP costs

1 APAP machine ($833), 1 headgear ($118), and 12 filter ($7)

1,039 (Range: 2000)

Individual item costs from Canadian CPAP supply184

First-year CPAP costs

1 CPAP machine ($720), 1 headgear ($118), and 12 filter ($7)

927 (Range: 2000)

Individual item costs from Canadian CPAP supply184

Costs of refills (in subsequent year)

Estimated typical refills: 1 each of new mask/headgear, tubing, and filter

822

Individual costs from Canadian CPAP supply184

Resource use based on183

APAP = automatic positive airway pressure; CPAP = continuous positive airway pressure.

Nasal EPAP

The cost of nasal EPAP was similarly determined by a micro-costing approach. The cost of a 30-day pack of single-use Provent nasal strips from an online source was $69.99 (or $2.33 daily).184

MAD

A macro-costing approach was taken to determine the cost of oral appliances given that the clinical experts involved in this review mentioned that the cost of a customized dental device would entail a package of services that include any required diagnostics tests, the dental appliance itself, and any subsequent visits to make adjustments to the appliance. A 2009 Health Quality Ontario (HQO) report on dental devices estimated that the costs would be approximately $2,000.38 In the base-case analysis, it was assumed that oral appliances required replacement every two years. Based on clinical experts' opinion, the cost of replacements was assumed to be half of the original costs, given that many of the diagnostic tests required would have been completed.

Lifestyle Management

It was assumed that the costs, from the perspective of the Ministry of Health, for a weight-loss program would primarily entail the cost of visits to a registered dietician whose average hourly fee would be $106.185 Changes to eating habit and food consumption were assumed to be out-of-pocket expenses by the patient.

The cost of positional therapy would primarily involve the cost of the intervention (retail online price: 189.95 USD186) with yearly replacement assumed.

Surgery-Related Costs

The cost of surgery included the costs for pre-surgical workup, hospital stay, and post-surgical assessment. Costs were estimated from a variety of sources and are presented in Table 97. When costs were unavailable, the costs of similar procedures were taken with verification by a clinical expert. A study on consecutive patients receiving orthognatic surgeries reported that, in the first year post-operative assessment period, the most common problem was neurosensory injury involving the inferior alveolar nerve, which was mild in 32% of cases, and severe in only 3% of those affected.187 The most serious complication was severe bleeding (> 12%). As the post-surgical complications were relatively minor and resolve with time, the cost of managing complications were omitted from this model.

Table 97: Non-Physician Costs Pertaining to Surgery (2016 Values)

 

Details (if needed)

Cost, $

Quantity

Reference

Pre-Surgical Workup

X-ray

Billing code: X006 "mandible X-ray - 3 views (unilateral or bilateral)"

32

1

Ontario Schedule of Benefits188

Blood tests

 

8

1

Ontario Schedule of Laboratory Fees189

CT scana

Billing code: X400 "CT head - w/out IV contrast"

43

1

Ontario Schedule of Benefits188

Dental cast

Specific to MMA only

1,032

1

Assumption, clinical expert (900 USD, exchange rate: 1 USD = 1.29 CAD)

Hospital Stay (Incl. Nursing, Diagnostic Imaging, Pharmacy, Labs, and Overhead Costs)

MMA

NOTE: Average length of hospital stay was 1.5 days

8,073

 

Ontario Case Costing Initiative190

GTA

Assumed captured in MMA costs above

Post-surgical assessment (to confirm surgery success)

Polysomnography

 

468

1

Ontario Schedule of Benefits188

CT = computed tomography; GTA = genial tubercle advancement; MMA = maxillomandibular advancement.
a The cost of CT reflects solely the cost of physician interpretation and not the cost of the technical component. A sensitivity analysis was conducted with a higher cost associated with CT.191

Physician Fees

Because all patients enter the model diagnosed with OSA, the cost of consultation pertaining to diagnosis was not captured in the model. Pertinent consultation costs that differed across treatment were included, such as visits to the dentist and surgeon for oral appliance and surgery, respectively. Follow-up visits were further captured except in the case of follow-up dental visits, as these are already captured in the cost of the oral appliance itself. Physician costs were obtained from the Ontario Schedule of Benefits for Physician Services188 and are presented in Table 98.

Table 98: Physician Labour Costs

 

Details (If Needed)

Interventions (Related to…)

Cost, $

Quantity

Reference

Consultation Costs

Dentist

 

Oral appliances

175

1 per device

HQO38

One-hour consult with dietician

 

Weight loss

135

1 per year

Dietitians of Canada185

Pre-surgical consultation

 

Surgery

91

1 per procedure

Ontario Schedule of Benefits188

Follow-up visits

Respirologists: (except oral appliances and surgery)

 

PAP therapy

Lifestyle modification

No treatment

80

1 per year

Ontario Schedule of Benefits188

Respirologists: for oral appliances

 

Oral appliance

80

1 every 5 years

Surgeon's post-surgery follow-up (before discharge)

 

Surgery

59

1 per procedure

Surgical follow-up visit

Quantity: 1/week in first 4 weeks; one during months 1-6; one follow-up re. PSG

Surgery

31

6 per procedure

Labour Costs Specific to Surgery (Incl. Anesthesiologist and Surgical Assistant)

Mandibular osteotomies for prognathism, sagittal split

Billing code: Orthognathic surgery #R379 "Le Fort I advancement in one segment"

Incl. 15 units of anesthesia and 6 units of surgical assistant

Surgery

1,229

1

Ontario Schedule of Benefits188

Le Fort I advancement

Billing code: Orthognathic surgery #R518 "mandibular osteotomies for prognathism, sagittal split"

Incl. 20 units of anesthesia and 10 units of surgical assistant

1,224

1

Genioplasty

Used the price of genioplasty as a proxy as per assumption from the clinical expert.

Incl. 10 units of anesthesia and 6 units of surgical assistant

607

1

Total labour cost of the day of surgery (MMA with genioplasty)

1,229 + 1,224+607

= 3,060

Total labour cost of the day of surgery (MMA without genioplasty)

1,229.49 + 1,224.4 = 2,453

MMA = maxillomandibular advancement; PSG = polysomnography.

Health State Costs

Table 99 presents the cost estimates for the clinical events used in this analysis. The initial cost of hypertension, MI, and stroke and the subsequent year cost for ongoing management were obtained from several Canadian sources.148,157

As the analysis is conducted from the perspective of the public payer (i.e., Ministry of Health), only medical costs arising from an MVA would be relevant to capture. These costs were obtained from a publication by the Ontario Ministry of Transportation.157 In the publication, the direct costs of both non-fatal and fatal MVAs were reported and, to map these costs to the stratified categories for MVA (i.e., property damage, personal injury, and fatal), it was assumed that accidents relating to property damage would not lead to any medical costs, whereas the costs from accidents leading to personal injury would reflect non-fatal MVA cost.

Table 99: Cost Estimates for Clinical Events

 

Cost,a $

Distribution

References

Initial

Non-fatal MI

7,760

Gamma (1, 7760)

Anis, 2006148

Fatal MI

6,289

Gamma (1, 6289)

Non-fatal mild-to-moderate stroke

13,327

Gamma (1, 13327)

Non-fatal severe stroke

73,911

Gamma (1, 73911)

Fatal stroke

6,168

Gamma (1, 6168)

Subsequent/ongoing

Hypertension

429

Gamma (1, 429)

Anis, 2006148

MI

1068

Gamma (1, 1068)

Mild-to-moderate stroke

7148

Gamma (1, 7148)

Severe stroke

67,903

Gamma (1, 67903)

Motor vehicle accident (one time)

Non-fatal, personal injury

2113

Gamma (1, 2113)

Voden, 1994157

Fatal

8713

Gamma (1,8713)

MI = myocardial infarction.
a Costs are reported in the year of the study's report but converted to 2016 values by the consumer price index.182

10. Statistical Analysis and Sensitivity Analysis

The ICUR was calculated according to convention and, in most cases, the sequential incremental cost-effectiveness ratio (ICER) was presented, unless otherwise specified. Strategies that were dominated (where there is another strategy that has lower expected costs and higher expected QALYs) or "extendedly dominated" (where at least one possible combination of two treatment strategies exists, which would be less costly and result in higher QALYs) were ruled out.

Sensitivity analyses were conducted to evaluate the degree to which the uncertainty in cost and effectiveness parameters affected the models' findings. The base-case findings for the economic evaluations reflect the probabilistic results based on 5,000 Monte Carlo simulations of the parameters' distributions. For the simulation, probability distributions related to natural history, hazard ratios, resource utilization (costs), and utilities were incorporated into the analysis, adopting standard methods for defining parameter uncertainty. Treatment effect was characterized by normal or beta distributions, transition probabilities and relative risks were characterized by beta and lognormal distributions, utility was characterized by gamma or beta distribution, and costs were characterized by gamma distributions. The probabilistic results characterize the extent to which parameter uncertainty affects the cost-effectiveness of the model. Results of the probabilistic analysis will be presented on a cost-effectiveness acceptability curve (CEAC), whereby interventions on the efficiency frontier will be highlighted. This graph presents the probability that each treatment is optimal given different values of willingness-to-pay for an additional QALY gained.

Sensitivity and scenario analyses were conducted to characterize the impact of different model assumptions and evaluate structural uncertainty. These include:

  • Alternative assumption on mortality: In the process of validation, the model's predicted mortality incidence was found to better align to the observed literature when it was assumed that, despite treatment adherence, patients would have an increased risk of mortality if they remained disease severe (i.e., relative risk of death for severe OSA, irrespective of treatment = 1.601). An alternative assumption was tested in which patients' adherent to treatment would have their mortality risks return to baseline levels.
  • Relative risk of mortality following CVE: In the base-case model, it was assumed that patients with a history of CVD would have an increased risk of mortality. However, other economic models on OSA do not make this assumption. Rather, mortality rates would be the same for those with or without a prior CVE. This assumption was tested in the economic analysis.
  • Effects of surgery on blood pressure: Clinical experts involved in this review have suggested that the treatment estimates for surgery on blood pressure reduction may be an overestimate given that the clinical trial on which this estimate was based recruited patients with a greater disease severity. The mean blood pressure reduction attained by PAP therapy and MAD were both substituted.
  • Effects of surgery on AHI: Similarly, the clinical experts noted a similar concern for the outcome, AHI, given that a selective population may have been recruited in surgical trials, potentially biasing in favour of surgery. The lower bounds of the 95% CI for all surgical outcomes for AHI were simultaneously applied to the model to test when lowered rates were tested.
  • Relapse following surgery: The base-case model assumed surgical outcomes were permanent. However, a few small observational studies have suggested that symptoms of OSA may return over time. Give that the rate of relapse was unknown, the clinical expert consulted in this review provided an estimate of 8%. As the rate was uncertain. the scenario most likely to bias against surgery would be to assume that the relapse occurred within the first year post-surgery.
  • Excluding the impact of MVA: The base case assumes all patients drive and are at risk of vehicular accidents until the age of 80. In a scenario in which patients are non-drivers, the risk of MVAs is removed, along with the costs and utilities associated with MVA.
  • Adherence rate: Scenarios tested on adherence rate included the following: all patients on treatment were assumed to be perfectly adherent (i.e., rate of adherence of 100%), testing the 95% CI of intervention-specific adherence rates in multi-way sensitivity analyses, and testing a lower rate that has been reported in literature (adherence rates = 17%).31 Device lifespan: Different lifespans for PAP therapy and oral appliances were varied across a plausible range.
  • Cost of computed tomography (CT): The full cost of CT to the public player was difficult to determine, as it consists of both the fee billed by physician for the interpretation of the scan and also the operational cost of the CT machine. The base case captures only the publicly reimbursed rate of physician billing and reflects a partial cost of CT. A sensitivity analysis was conducted in which the highest reported cost ($625) from a US laboratory191 was taken as a possible proxy of the full public payer costs for CT imaging.

Interpretation of the economic findings was based on setting a willingness-to-pay threshold of $50,000 per QALY. Sensitivity analyses were conducted to explore situations in which the results would change; i.e., the order in which the treatment strategies were considered cost-effective differed or when the order remained consistent but the AHI cut-off score differed by more than its clinically meaningful difference when compared with the base-case analysis (Appendix 15).

When possible, subgroup analyses were conducted to address whether there was any heterogeneity that could affect the cost-effectiveness results. The clinical review found evidence on some of the pre-specified subgroups of interest on the outcomes of AHI and blood pressure for some of the interventions of interest but not all. As such, in the economic analysis, the subgroup analyses were conducted based on the assumption that subgroups had identical treatment effects but differed by baseline event risks arising from the Framingham risk equation. Subgroups that could be explored in the model included:

  • Sex: male vs. female
  • Age: 30 vs. 70 years
  • Patients who are hypertensive
  • Patients who are current smokers
  • Patients with diabetes.
11. Model Validation

The model structure and data inputs were presented to four Canadian clinical experts in the areas of sleep medicine: respirology, dentistry, and maxillofacial surgery to ensure that the model, parameters, and assumptions reflect clinical practice and the available body of literature (face validity). Internal validity was assessed by ensuring the mathematical calculations were performed correctly and were consistent with the model specification. Logical inconsistencies were assessed by evaluating them under hypothetical and extreme conditions. The model also underwent external peer review.

External validation was further explored by comparing the model predictions to actual outcomes reported in three existing observational cohorts.134,135,141 Multiple validations were performed that crisscrossed different outcomes (i.e., all-cause and CV mortality), interventions (i.e., no treatment vs. treated) and patient populations.

12. Model Assumptions

The base-case economic analysis was conducted under the assumptions shown in Table 100.

Table 100: Assumptions Used to Populate the Economic Models

Assumption

Strategy in Which Applicable

Sensitivity Analysis Description

Patients modelled represent a cohort of adult patients who are representative of the Canadian population (e.g., 76.5% male, aged 55 years).

All

-Sex proportion varied from 100% males to 0% males

-Age varied in the model (30 and 70 years)

In terms of the surgical procedures, it was modelled that 90% of patients would be receiving MMA in which 20% would be performed in adjunct with GTA. The remaining 10% of procedures were GTA.

All

None

(the cost of surgery and treatment effects represents a pooled estimate)

Patients diagnosed with hypertension will be treated with medication, targeting the return to baseline blood pressure levels.

All

None

All patients drive and discontinue this activity at the age of 80, except in patients following an incidence of severe stroke who are assumed unable to drive.

All

Changed the model to reflect non-drivers only

Patients who are perfectly adherent to treatment will have risk of MVAs similar to the general population

Non-surgical interventions

None

Treatment effects on AHI, based on naive treatment comparison against "no treatment," reflect the likely "true" comparative treatment effects.

All treatment interventions

Treatment effect estimates from an existing NMA were applied for all non-surgical interventions

Rates of discontinuation were based on disease severity. Long-term studies on CPAP provided the discontinuation rates for severe OSA patients, while studies on OA provided the discontinuation rates for moderate OSA patients.

Non-surgical interventions

None

After the fourth year of treatment, no patients will discontinue their treatment (i.e., LOCF analysis)

Non-surgical intervention

Assume all patients drop out after 4 years of treatment

Surgical outcomes achieved in the first year post-surgery were assumed to be permanent.

Surgery

Assume 8% of patients will experience relapse and become symptomatic after the first year of surgery

For joint health states, a minimum utility approach was taken to determine the appropriate utility value to apply to that health state.

All

Apply a multiplicative approach to utility determination

Treatment would affect the utility weight depending on the post-treatment disease severity.

All

Apply a treatment-specific utility increments to the model

81% of patients receiving PAP therapy would be prescribed APAP. The remainder would be prescribed CPAP. In addition, for patients on CPAP, 90% of patients would have undergone split night polysomnography.

PAP therapy

None

All devices and regular replacements were assumed to be covered by the Ministry of Health. Yearly replacements for PAP-based treatments included a new mask, tubing, head gear, and filter.

PAP therapy

Different reimbursement structures explored

The device life span for PAP machines was 7 years, while for mandibular advancement devices, it would be 2 years.

PAP therapy, MAD

Lifespan for PAP therapy varied from 5 to 10 years; lifespan for oral appliances varied from 2 to 10 years

For oral appliances, it was assumed that the cost of replacement would be half of the original costs.

MAD

Replacement of oral appliance is the same as the first-time costs

Changes to eating habits and food consumption were assumed to be out-of-pocket expenses by the patient.

Weight Loss

None

AHI = Apnea-Hypopnea Index; APAP = automatic PAP; CPAP = continuous PAP; GTA = genial tubercle advancement; LOCF = last observation carried forward; MAD = mandibular advancement device; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; MVA = motor vehicle accident; PAP = positive airway pressure; RR = relative risk.

4.2.2. Model Results

1. Model Validation

Our model predicted a mortality rate of 14.1% and 42.1% at 10 and 20 years, respectively, under the base case for untreated non-smoker patients with severe OSA, and a mortality rate of 17.6% and 49.3% for patients who are smokers. Figure 3 presents the survival curves at the end of the study follow-up. When the model predictions were compared with the findings from published clinical trials, our model was found to underestimate mortality for untreated severe OSA in both the Wisconsin Sleep Cohort Study (WSCS)134 and Busselton study135 but otherwise, for the other disease severities, the model predictions and study's observations were closely aligned (Figure 2).

Figure 2: Comparison of Model's Predicted and Cohort Studies' Reported Survival Curves.

(A) Wisconsin Sleep Cohort Study134 (B) Busselton Study135

Specifically, the WSCS comes from a considerably different patient population compared with SHHS, upon which the model's CV parameters are based. SHHS recruited an older, geographically and racially diverse sample from a number of "parent" CV cohort studies, whereas WSCS represents a more homogenous population from a common sampled pool of employed individuals. Table 101 reports the incidence rate (per 100 person-years) and proportion of all-cause mortality and CV mortality reported in WSCS and predicted by our model. The incidence rate predicted by the model was slightly higher for all-cause mortality across all disease severities but remained within the reported 95% CI. Given that the estimate was also higher in patients without OSA, this would suggest that there are inherent differences between the general Canadian population (on which the model is based) and the patient population that was studied in the WSCS. The outcome of CV mortality was more aligned between the observed and predicted value.

The model was further validated by comparing the model predictions to a study by Martinez-Garcia et al.141 of elderly, severe OSA patients, which was conducted in Spain. They reported that 34.1% and 16.1% of patients who were untreated and treated by CPAP died over a mean follow-up period of 69 months. Our model led to similar predicted values. For patients on CPAP, the proportion would depend on adherence as values range from 24.2% (normal adherence) to 20.6% (perfect adherence), with clinical experts suggesting that better adherence is observed in Spanish studies, which would be indicative that the lower estimate is more likely accurate.

Table 101: Comparison of Model's Prediction and Wisconsin Sleep Cohort Study Observation on All-Cause and Cardiovascular Mortality

Parameter

WSCS Study

Reported Results

Model Prediction

All-cause mortality

WSCS136,192,193

1,522 men and women

Severe: proportion males, 78%; age, 50, 19% smokers; tot-C, 212; HDL, 39; SBP, 135

Moderate: proportion males, 72%; age, 50; 15% smokers; tot-C, 211; HDL, 42; SBP, 135

Mild: proportion males, 65%; age, 50; 16% smokers; tot-C, 213; HDL, 44; SBP, 130

None: proportion males, 55%; age, 48; 19% smokers; tot-C, 202; HDL, 49; SBP, 120

Mean observation: 13.8 years

Total of 20,963 person-years

Rate per 100 person-years (95% CI)

Severe AHI: 1.46 (0.75 to 2.54)

Moderate AHI: 0.54 (0.20 to 1.18)

Mild AHI: 0.55 (0.32 to 0.90)

None: 0.29 (0.21 to 0.38)

Proportion: %

Severe AHI: 19.1

Moderate AHI: 7.3

Mild AHI: 7.3

None: 4

Rate per 100 person-years (95% CI)

Severe AHI: 1.68

Moderate AHI: 0.82

Mild AHI: 0.66

None: 0.51

Proportion: %

Severe AHI: 18.62

Moderate: 10.40

Mild: 8.52

None: 6.77

Cardiovascular mortality

Rate per 100 person-years (95% CI)

Severe AHI: 0.61 (0.20 to 1.41)

Moderate AHI: 0.18 (0.02 to 0.65)

Mild AHI: 0.21 (0.08 to 0.45)

None: 0.07 (0.04 to 1.30)

Proportion: %

Severe AHI: 7.9

Moderate AHI: 2.4

Mild AHI: 2.7

None: 1

Rate per 100 person-years (95% CI)

Severe: 0.59

Moderate: 0.13

Mild: 0.11

None: 0.062

Proportion: %

Severe: 7.58

Moderate: 1.75

Mild: 1.51

None: 0.85

Incidence of Mortality

Martinez-Garcia, 2012141

Untreated severe:

proportion males, 71.7%; age, 71.9, AHI, 58.6; 47% smokers; tot-C, NR; HDL, NR; SBP, NR

Treated severe: proportion males, 62.2%; age, 70.1; AHI, 52.2; 41% smokers; tot-C, NR; HDL, NR; SBP, NR

Median follow-up: 69 months (49 to 87 months)

Proportion: %

Untreated severe: 34.1%

Treated severe: 16.1%

Proportion: %

Untreated severe: 34.47%

Treated severe: 24.2% (normal adherence)

20.57% (perfect adherence)

AHI = Apnea-Hypopnea Index; HDL = high-density lipoprotein; NR = not reported; SBP = systolic blood pressure; tot-C = total cholesterol; WSCS = Wisconsin Sleep Cohort Study.

2. Base Case

The model was run, varying baseline AHI values between 5 and 60. The incremental costs, QALYs, and ICUR over a lifetime are presented in Table 102 across select AHI values to highlight how cost-effectiveness of interventions for OSA vary by baseline disease severities.

Table 102: Probabilistic Results of Base-Case Scenario Over Cohort's lifetime (5000 simulations)

Intervention

Expected Costs, $

Expected QALYs

Incremental Cost, $

Incremental QALYS

ICUR, Compared With "No Treatment"

($/QALY)

ICUR, Sequential ($/QALY)

Baseline: Mild OSA (AHI = 5)

No treatment

11,598

10.911

-reference-

PAP therapy

18,137

10.947

6,539

0.036

182,261

Ext. Dom

MAD

18,974

10.953

7,375

0.042

175,543

175,543

MMA ± GTA

23,757

10.953

4,783

0.000

289,949

Dominated

Baseline: Moderate OSA (AHI = 15)

No treatment

11,644

9.552

-reference-

PAP therapy

18,152

10.360

6,508

0.808

8,058

8,058

MAD

19,015

10.453

862

0.093

8,183

9,276

MMA ± GTA

23,672

10.340

4,657

-0.113

15,262

Dominated

Baseline: Severe OSA (AHI = 30)

No treatment

8,433

8.917

-reference-

PAP therapy

18,086

10.218

9,653

1.301

7,420

7,420

MAD

19,057

9.383

971

-0.835

22,811

Dominated

MMA ± GTA

23,249

10.120

5,164

-0.097

12,312

Dominated

Baseline: Severe OSA (AHI = 60)

No treatment

8,382

8.908

-reference-

MAD

16,388

8.881

8,006

-0.027

Dominated

Dominated

PAP therapy

16,866

9.101

8,485

0.193

43,899

Ext. Dom

MMA ± GTA

23,042

9.764

14,661

0.856

17,125

17,125

AHI = Apnea-Hypopnea Index; Ext Dom = extendedly dominated; GTA = genial tubercle advancement; ICUR = incremental cost-utility ratio; MAD = mandibular advancement device; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; QALY = quality-adjusted life-year; PAP = positive airway pressure

In general, the treatment strategies were associated with higher expected costs compared with the no-treatment strategy. The higher costs were due to both the costs related to treatment itself but also the costs of long-term maintenance (i.e., post-MI, post-stroke, and hypertension) as the life expectancies tended to be longer for patients on treatment (Table 103). There was partial cost offset given the lower rates of MVAs among patients receiving treatment for OSA (Table 103). Across all disease severities, surgery was associated with the highest expected costs while no treatment was contrarily associated with the lowest expected costs.

The expected QALYs were lower with increasing disease severity across all strategies studied due to increased mortality and morbidity associated with more severe forms of OSA. The magnitude to which QALY changed varied by baseline AHI value.

Table 103: Disaggregate Clinical Outcomes for the Base Case (Deterministic)

 

Average number of events per patient over their lifetime

LYG

MVA

MI

Stroke

Baseline: Mild OSA (AHI = 5)

No treatment

1.998

0.214

0.083

26.12

PAP therapy

1.277

0.211

0.080

26.20

MAD

1.121

0.211

0.080

26.21

MMA ± GTA

1.028

0.209

0.074

25.47

Baseline: Moderate OSA (AHI = 15)

No treatment

1.962

0.202

0.077

25.01

PAP therapy

1.261

0.205

0.078

25.69

MAD

1.107

0.206

0.078

25.77

MMA ± GTA

1.003

0.203

0.072

24.97

Baseline: Severe OSA (AHI = 30)

No treatment

1.800

0.189

0.116

21.25

PAP therapy

1.082

0.201

0.092

24.50

MAD

0.904

0.194

0.085

24.05

MMA ± GTA

1.032

0.200

0.076

24.41

Baseline: Severe OSA (AHI = 60)

No treatment

1.800

0.189

0.116

21.25

PAP therapy

1.054

0.190

0.097

22.91

MAD

0.869

0.185

0.112

21.41

MMA ± GTA

1.242

0.197

0.082

23.61

AHI = Apnea-Hypopnea Index; GTA = genial tubercle advancement; LYG = life-year gained; MAD = mandibular advancement device; MI = myocardial infarction; MMA = maxillomandibular advancement; MVA = motor vehicle accidence; PAP = positive airway pressure.

Table 104: Disaggregate Costs ($) for the Base Case (Deterministic)

 

Device-Specific Costs

Disease-Related Cost (e.g., Professional Visits)

Cardiovascular Event Costs

MVA Event Costs

Long-Term Costs

Baseline: Mild OSA (AHI = 5)

No treatment

0

1,779

3,323

1,911

17,306

PAP therapy

12,942

2,092

3,246

1,131

16,875

MAD

10,863

3,814

3,239

993

16,834

MMA ± GTA

11,799

805

3,163

911

17,568

Baseline: Moderate OSA (AHI = 15)

No treatment

0

1,696

3,140

1,875

17,218

PAP therapy

12,942

2,051

3,160

1,117

16,912

MAD

10,863

3,778

3,164

981

16,805

MMA ± GTA

11,799

769

3,089

888

17,532

Baseline: Severe OSA (AHI = 30)

No treatment

0

1,464

2,754

1,713

12,979

PAP therapy

15,303

1,956

3,023

958

15,347

MAD

12,083

3,905

2,928

801

15,789

MMA ± GTA

11,799

918

3,018

914

17,085

Baseline: Severe OSA (AHI = 60)

No treatment

0

1,464

2,754

1,713

12,979

PAP therapy

13,963

1,829

2,824

934

14,551

MAD

10,619

3,403

2,688

771

12,922

MMA ± GTA

11,799

1,185

2,956

1,100

16,406

AHI = Apnea-Hypopnea Index; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; MVA = motor vehicle accident; OSA = obstructive sleep apnea; PAP = positive airway pressure.

At a willingness-to-pay threshold of $50,000/QALY, no treatment was the most likely cost-effective strategy in patients with mild OSA (probability ranged from 1.000 to 0.999). For moderate OSA, a cut-off was observed as MAD was the most likely cost-effective strategy at the lower AHI range (15 ≤ AHI ≤ 25), but switched to MMA with or without GTA, at the higher AHI range of moderate OSA (26 < AHI < 30). A cut-off was similarly observed for severe OSA, with PAP therapy emerging as the most likely cost-effective strategy up to a baseline AHI value of 32 and switched to MMA with or without GTA at subsequent higher AHI values (Figure 3). It is important to note that interpretation of these thresholds must be tempered by the fact that is partly an artifact emerging from the model structure as, while treatment effect was modelled as a continuous outcome, natural history of the disease was modelled categorically.

Figure 3: The Evolution of the CEAC Across Different Baseline AHI Severity

(A) AHI = 5; (B) AHI = 15; (C) AHI = 25; (D) AHI = 30; (E) AHI = 32; (F) AHI = 40. Black line = no treatment; green line = PAP; purple line = MAD; blue line = surgery

Scenario Omitting PAP Therapy

As the patient perspectives and experiences review highlights, there are certain circumstances (e.g., Canadian forces, individuals unstably housed, patients with physical or sensory impairments or patients with disabilities) whereby patients may not be suitable candidates for PAP therapy. By removing PAP therapy as a possible treatment option, this scenario was found to differ from the base-case findings specifically in patients with severe OSA. The ICER for MMA with or without GTA reduced (e.g., sequential ICER for surgery in patients with a baseline AHI value of 30 = $12,371 per QALY) (Table 105). As such, at a willingness-to-pay threshold of $50,000/QALY, the analysis demonstrated that MMA with or without GTA was the most likely cost-effective strategy for patients with baseline AHI values greater or equal to 26 (figure not shown).

Table 105: Probabilistic Results of Scenario That Excludes PAP Therapy (5,000 Simulations)

Intervention

Expected Costs, $

Expected QALYs

Incremental Cost, $

Incremental QALYs

ICUR, sequential ($/QALY)

Baseline: Mild OSA (AHI = 5)

No treatment

11,547

10.911

-reference-

MAD

18,926

10.953

7,379

0.042

175,553

MMA ± GTA

23,633

10.953

4,706

0.000

Dominated

Baseline: Moderate OSA (AHI = 15)

No treatment

11,663

9.543

-reference-

MAD

19,038

10.450

7,375

0.907

8,134

MMA ± GTA

23,708

10.338

4,670

-0.112

Dominated

Baseline: Severe OSA (AHI = 30)

No treatment

8,328

8.906

-reference-

MAD

19,029

9.369

10,701

0.463

Ext. Dom

MMA ± GTA

23,273

10.114

14,945

1.208

12,371

AHI = Apnea-Hypopnea Index; Ext Dom = extendedly dominated; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; QALY = quality-adjusted life-year; PAP = positive airway pressure.

Scenario Omitting Surgical Interventions

Similarly, not all OSA patients may be clinically eligible for surgery, or there may be settings in Canada in which surgery is not available. By removing MMA with or without GTA as a possible treatment option, the findings were similar to the previous analysis with the exception that, for the baseline AHI values in which surgery was considered the mostly likely cost-effective, this was replaced by PAP therapy (Table 106). Therefore, at a willingness-to-pay threshold of $50,000/QALY, PAP therapy emerged as the most likely cost-effective intervention for patients with a baseline AHI greater or equal to 26 (figure not shown).

Table 106: Probabilistic Results of Scenario That Excludes Surgical Interventions (5,000 Simulations)

Intervention

Expected Costs, $

Expected QALYs

Incremental Cost, $

Incremental QALYs

ICUR, sequential ($/QALY)

Baseline: Mild OSA (AHI = 5)

No treatment

11,673

10.910

-reference-

MAD

18,173

10.946

6,500

0.036

180,397

PAP therapy

19,029

10.952

856

0.006

137,696

Baseline: Moderate OSA (AHI = 15)

No treatment

11,732

9.541

-reference-

PAP therapy

18,217

10.353

6,485

0.812

7,984

MAD

19,120

10,446

903

0.093

9,680

Baseline: Severe OSA (AHI = 30)

No treatment

8,458

8.910

-reference-

PAP therapy

18,206

10.215

9,749

1.305

7,470

MAD

19,163

9.376

957

-0.839

Dominated

AHI = Apnea-Hypopnea Index; MAD = mandibular advancement device; OSA = obstructive sleep apnea; QALY = quality-adjusted life-year; PAP = positive airway pressure.

Different Reimbursement Coverage

There is considerable variation in terms of reimbursement coverage across jurisdictions. For instance, in most publicly funded provincial programs, oral appliances are not covered but represent a patient out-of-pocket expense or are reimbursed by a private third-party insurer. Despite this, by removing the costs of oral appliances and dental consults, the overall findings remained consistent with the exception of mild OSA. The expected costs for MAD reduced across all disease severities as the costs of dental-related expenses shifted from the public payer to the private payers. MAD became the least expensive of the treatments and, in the case of patients with mild OSA, MAD emerged as the dominant strategy (i.e., less costly, more effective) compared with all other strategies that were studied (Table 107). At a willingness-to-pay threshold of $50,000/QALY, MAD was the most likely cost-effective strategy for mild-to-moderate OSA (probability = 0.9974 for AHI 10 and 0.6486 for AHI 25) (figure not shown).

In the patient perspective and experience review and the implementation review, both noted that a barrier to treatment is the funding structure. In particular, our clinical experts have suggested that patients' willingness to be on treatment is influenced by the reimbursement structure. As such, a sensitivity analysis was conducted that also applied the lower bounds of the 95% CI of the MAD adherence rates. The model's findings were found to remain robust to changes in this parameter as MAD remained the most likely cost-effectiveness strategy for mild-to-moderate OSA (Table 107).

Table 107: Probabilistic Results Assuming Oral Appliances Are Not Reimbursed (i.e., Represent Out-of-Pocket Expenses)

Adherence Scenario

Intervention

Expected Costs, $

Expected QALYs

Incremental Cost, $

Incremental QALYs

ICUR, sequential ($/QALY)

Base-case adherence

Baseline: Mild OSA (AHI = 5)

MAD

10,748

10.954

-reference-

No Treatment

11,580

10.912

832

−0.042

Dominated

PAP therapy

18,102

10.948

7,354

−0.006

Dominated

MMA ± GTA

23,749

10.954

13,001

0.000

1.4x109

Baseline: Moderate OSA (AHI = 15)

MAD

10,852

10.458

-reference-

No Treatment

11,654

9.563

802

−0.895

Dominated

PAP therapy

18,118

10.366

7,266

−0.092

Dominated

MMA ± GTA

23,678

10.346

5,561

−0.020

Dominated

Baseline: Severe OSA (AHI = 30)

No treatment

8,400

8.923

-reference-

MAD

9,877

9.384

1,477

0.462

3,199

PAP therapy

18,103

10.221

8,226

0.837

9,830

MMA ± GTA

23,356

10.125

5,253

−0.096

Dominated

Baseline: Severe OSA (AHI = 60)

MAD

7,689

8.884

     

No Treatment

8,329

8.910

640

0.026

Ext. Dom

PAP therapy

16,825

9.108

9,135

0.224

Ext. Dom

MMA ± GTA

22,960

9.770

15,271

0.886

17,235

Lower adherence for MAD

Baseline: Mild OSA (AHI = 5)

MAD

10,800

10.953

-reference-

No treatment

11,608

10.912

808

−0.041

Dominated

PAP therapy

18,107

10.948

7,307

−0.005

Dominated

MMA ± GTA

23,681

10.954

12,881

0.001

10,781,667

Baseline: Mild OSA (AHI = 15)

MAD

10,901

10.449

-reference-

No treatment

11,684

9.555

783

−0.894

Dominated

PAP therapy

18,145

10.359

7,243

−0.090

Dominated

MMA ± GTA

23,700

10.340

5,555

−0.019

Dominated

Baseline: Severe OSA (AHI = 30)

No treatment

8,461

8.910

-reference-

MAD

10,016

9.375

1,555

0.465

3,344

PAP therapy

18,121

10.212

8,105

0.838

9,676

MMA ± GTA

23,405

10.120

5,284

−0.092

Dominated

Baseline: Severe OSA (AHI = 60)

MAD

7,860

8.889

-reference-

No treatment

8,482

8.917

622

0.028

Ext. Dom

PAP therapy

17,028

9.112

9,168

0.223

Ext Dom.

MMA ± GTA

23,181

9.772

15,322

0.883

17,354

AHI = Apnea-Hypopnea Index; Ext Dom = extendedly dominated; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; QALY = quality-adjusted life-year; PAP = positive airway pressure.

Similarly, it has been suggested that the cost of PAP devices can vary. The prices used in the reference case were more similar to the Ontario Assistive Devices coverage program (i.e., in Ontario, the total cost is $1,075, of which the Ministry of Health pays $860 and the remainder represents out-of-pocket costs;194 first-year costs in the model were $1,039 and $927 for APAP and CPAP devices, respectively). Clinical experts involved in this review suggested that the first-year costs of PAP machines vary widely and may be as high as $2,000 in other jurisdictions in Canada. Probabilistic results based on a higher first-year acquisition price for PAP therapy are presented in Table 108. This scenario resulted in higher expected costs for the PAP therapy strategy. Although the ICUR associated with PAP therapy increased to $8,891 per QALY in patients with baseline severe OSA (AHI = 30), the overall economic findings remained consistent.

Table 108: Probabilistic Results Assuming the First-Year Price of PAP Therapy Is C$2,000

Intervention

Expected Costs, $

Expected QALYs

Incremental Cost, $

Incremental QALYs

ICUR, Sequential ($/QALY)

Baseline: Mild OSA (AHI = 5)

No Treatment

11,590

10.914

-reference-

MAD

18,963

10.956

7,373

0.042

175,707

PAP therapy

19,776

10.950

813

-0.006

Dominated

MMA ± GTA

23,677

10.956

4,714

0.001

8,237,240

Baseline: Moderate OSA (AHI = 15)

No Treatment

11,607

9.540

-reference-

MAD

18,987

10.448

7,380

0.909

8,122

PAP therapy

19,808

10.355

822

-0.094

Dominated

MMA ± GTA

23,710

10.335

4,723

-0.114

Dominated

Baseline: Severe OSA (AHI = 30)

No Treatment

8,471

8.927

-reference-

MAD

19,094

9.386

10,624

0.459

Ext. Dom

PAP therapy

19,999

10.224

11,528

1.297

8,891

MMA ± GTA

23,468

10.129

3,469

-0.095

Dominated

Baseline: Severe OSA (AHI = 60)

No Treatment

8,277

8.911

-reference-

MAD

16,265

8.885

7,988

-0.026

Dominated

PAP therapy

18,533

9.106

10,255

0.194

Ext. Dom

MMA ± GTA

22,832

9.771

14,554

0.860

16,932

AHI = Apnea-Hypopnea Index; Ext Dom = extendedly dominated; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; QALY = quality-adjusted life-year; PAP = positive airway pressure

2. Sensitivity Analysis

There are a number of uncertainties about several of the modelling assumptions. Selected results of the sensitivity analysis are shown in Table 109. The majority of the sensitivity analyses had minimal impact on the base-case ICUR or on the interpretation of the cost-effectiveness results, except in patients with moderate baseline disease severity (AHI = 25). Additional sensitivity analysis that had minimal impact on the model are presented in Appendix 21 and include the time horizon (if greater than seven years), variation in adherence rates values, lower treatment effect for surgery, surgical relapse, the type of PAP intervention prescribed (i.e., APAP or CPAP), the average lifespan of devices and higher CT costs.

Selected parameters to which the model was sensitive and their results are presented here:

Discount rate: The reference case assumed a discount rate of 5% as per existing Canadian guidelines.132 With a lowered discount rate, surgery became a more attractive strategy for patients with moderate-to-severe OSA. This trend can be explained because of the fact that the MMA with or without GTA strategy is associated with high upfront treatment costs relating to the one-time surgical procedure, whereas the treatment cost of non-surgical strategies accumulates in the model over time. As such, in circumstances of no discounting, the cost difference between the non-surgical and surgical strategies reduced as future costs were weighed equivalent to present costs. The CEAC suggested that the probability in which MMA with or without GTA would be the most likely cost-effective strategy increases as the discount rates decreased. From a willingness-to-pay threshold of $50,000/QALY, MMA with or without GTA was the most likely cost-effective intervention for patients with baseline AHI values greater or equal to 22 when discounting was set to 0 (figure not shown).

AHI Estimates from NMA: The clinical review identified one NMA of non-surgical interventions for OSA that reported on the outcome of AHI.161 A sensitivity analysis was conducted using the treatment effect estimates reported in that study and incorporating it into the model. Specifically, this study reported an AHI reduction of -25.27 (95% CI, -28.52 to -22.03) and -15.2 (95% CI, -19.5 to -10.91) for PAP therapy and MAD, respectively. Using this NMA assumes that the reduction in AHI by treatment would be consistent across disease severities. In particular, the model was unstable with greater parameter uncertainty toward the higher end of moderate OSA (22 < AHI < 30) as, at a willingness-to-pay threshold of $50,000/QALY, the probability in which any strategy emerged as the most likely cost-effective did not exceed 60%. The model conclusion varied from the base case for patients with moderate and severe OSA. At a willingness-to-pay threshold of $50,000/QALY, the following interventions emerged as the most likely cost-effective: MAD (15 ≤ AHI < 25); CPAP (25 ≤ AHI < 27); MAD (27 ≤ AHI < 31); MMA with or without GTA (AHI ≥ 31) (figure not shown). As noted, caution is required with respect to these cut-offs, given that: i) the cut-off emerges from the model structure with natural history modelled categorically; ii) the competing low probabilities associated with each strategy are in the moderate severity range.

Rate of Discontinuation: There was uncertainty about the long-term rates of treatment discontinuation as the majority of the longitudinal studies had a follow-up period of four to five years. The base case was based on the last-outcome-carried-forward approach given that the patient perspectives and experience review reported that the first six months were the most critical in determining whether patients continued on their treatment. However, when complete discontinuation was assumed after the last observed period, the model was found to overall favour surgery, given that the effects of non-surgical treatment terminated after the last follow-up period. This was found to have an impact on moderate-to-severe OSA, as the ICUR for MMA with or without GTA changed to $41,798 and $19,103/QALY in patients with a baseline AHI of 15 and 30, respectively. The CEAC suggests that, at a willingness-to-pay threshold of $50,000/QALY, MMA with or without GTA was the most likely cost-effective strategy for all patients with moderate-to-severe OSA (i.e., baseline AHI ≥ 15) (figure not shown).

In addition, there was uncertainty regarding the rates of discontinuation, given that the proportion of patients with mild-to-moderate OSA who discontinued after the first year of non-surgical treatment has been reported to range from 88%167 to 18%.168 The model's findings remained mostly robust at the upper and lower bounds of the 95% CI for the discontinuation rates. The sole exception that was observed was under the analysis based on the lower discontinuation rate for moderate OSA in which the cost-effectiveness of treatment was sensitive (i.e., first-year discontinuation rates for non-surgical treatment = 88%; second-year discontinuation rates = 100%). The ICUR of MMA with or without GTA reduced to $17,758/QALY for patients with baseline AHI value of 15 and, similarly, surgery emerged as the most likely cost-effective strategy in patients with moderate-to-severe OSA (figure not shown).

Refusal to fill PAP therapy prescription: The base case assumed 10% of patients would not fill their prescription for a PAP device and their disease progression would be modelled similarly to those who had discontinued treatment (e.g., cost of PAP therapy is not applied to this proportion of patients who refuse to fill their prescription).130 The model was found to be robust to this structural assumption when it was removed (i.e., no patient would refuse filling their PAP prescription; and, rather, all patients would fill their device prescription). In varying the value of this parameter across its reported range (4.5% to 32.4%), the higher refusal rate for PAP therapy was found to affect the patient subgroup with severe OSA. The expected QALYs associated with PAP therapy decreased and, although its ICUR remained mostly unchanged at $7,537/QALY for patients with baseline AHI value of 30, the sequential ICUR for surgery became more favourable at $34,271/QALY given the larger QALY difference between these two strategies. Consequently, at a willingness-to-pay threshold of $50,000/QALY, MMA with or without GTA was the most cost-effective strategy for patients with a baseline AHI greater or equal to 26.

Adherence rate: Modelling adherence is a structural component of the model. In removing this aspect from the model (i.e., by assuming all patients would be perfectly adherent to their non-surgical treatment), the model was found to be sensitive specifically to the subgroup of patients with severe OSA. In particular, this structural assumption was found to affect the cut-off in which PAP therapy was no longer considered the most likely cost-effective intervention as it shifted from the base case reported AHI value of 33 to an AHI value of 48 (figure not shown).

Relationship between adherence and treatment effect: Given the assumption made regarding the treatment response of partially adherent patients (i.e, partially adherent patients would experience half the treatment effects), two extreme scenarios were tested: i) partially adherent patients would have the same treatment effect as perfectly adherent patients and ii) partially adherent patient would have no treatment response (i.e., event risks would return to values similar to those who have discontinued treatment). The model was sensitive to both assumptions.

Under the first assumption whereby partially adherent patients would have complete treatment benefit (i.e., similar to those fully adherent to treatment), the findings differed from the base case in that PAP therapy emerged as the most likely cost-effective intervention at the higher end of moderate OSA. As such, at a willingness-to-pay of $50,000/QALY, PAP therapy was the most likely cost-effective strategy for patients with a baseline AHI value between 26 and 47.

Under the second assumption, whereby partially adherent patients would have no treatment benefit, PAP therapy no longer appeared cost-effective for severe OSA. Rather, MMA with or without GTA emerged as the most likely cost-effective strategy for patients with a baseline AHI value greater or equal to 26.

Multiplicative approach to utility calculation: For joint health states, the base-case model was based upon a minimum approach to estimate the utility value. An alternative is the multiplicative approach, which places lower weights when multiple comorbidities are present. Specific to this model, it would result in lower utility weights for hypertensive patients that experience a clinical outcome. When the multiplicative approach was instead applied to estimate the value of joint health states, the expected utilities were lowered in all cases but with the greatest reduction observed in the no-treatment strategy compared with the active intervention strategies. Furthermore, as the incidence of developing hypertension was lower in patients undergoing surgery, the ICUR for surgery improved to $13,052, $13,928, and $16,058/QALY for patients with baseline AHI values of 15, 30, and 60, respectively. Across moderate-to-severe OSA, MMA with or without GTA was found to be the most likely cost-effective strategy at a willingness-to-pay threshold of $50,000/QALY (figure not shown).

Treatment-specific utility increment: The model was found to be unstable when treatment-specific utility increments were incorporated into the model for those adherent to non-surgical treatment or those who had undergone a surgical procedure. The CEAC results suggest that the probability in which a particular strategy would be considered the most likely cost-effective at a willingness-to-pay threshold of $50,000/QALY was often less than 50% with multiple strategies competing. It is perhaps important to note that this analysis should be considered exploratory and would be suitable only if there is evidence suggesting differential treatment-specific QoL benefit.

Variation in the reimbursement of PAP-related equipment: The model was found to be sensitive to the replacement schedule for PAP-related equipment and accessories. In applying the Medicare yearly max replacement schedule (i.e., four masks, four tubes, filters, and two headgears),183 the yearly cost of refills increased from $822 to $2,686. This resulted in a rise in the expected cost of PAP therapy that was higher than a strategy of MMA with or without GTA. The base-case findings differed as, under this analysis, MMA with or without GTA emerged as the most likely cost-effective treatment for patients with moderate-to-severe OSA, at a willingness-to-pay threshold of $50,000. At a baseline AHI value of 30, the ICUR for MMA with or without GTA was $12,335/QALY and the ICUR for PAP therapy was $101,085/QALY.

Similarly, in some jurisdictions, only the generator unit of the PAP device is reimbursed while the remaining equipment is paid out of pocket by patients.2 A separate analysis was conducted to explore this scenario where the reimbursed cost of PAP therapy comprised only the generator unit. The expected costs of the PAP strategy reduced across all disease severities. The economic findings were found to be sensitive to the reimbursement price for PAP therapy, as the attractiveness of PAP therapy increased to a greater range of baseline OSA severities. In particular, the main difference compared with the base-case findings was that, at a willingness-to-pay threshold of $50,000/QALY, PAP therapy emerged as the most likely cost-effective strategy in patients with mild (i.e., 5 ≤ AHI < 15) and for a wider AHI range of severe OSA (30 ≤ AHI < 46).

Cost of MAD replacement: The base-case model assumed that the price of subsequent oral appliances replacements would be half its first-time cost, based on clinical experts' opinion. This was considered suitable as many of the diagnostic procedures are not required for replacement devices. However, if the cost of replacement devices was assumed to be the same as its first-time cost, MAD was no longer present on the efficiency frontier in patients with moderate OSA as the expected costs for the MAD strategy increased and, rather, PAP therapy replaced MAD as being the most likely cost-effective intervention for moderate OSA.


Table 109: Overview of Sensitivity Analysis Across Select Baseline Disease Severities (Probability in Which Intervention Is Cost-Effective at a Willingness-to-Pay Threshold of $50,000/QALY)

Sensitivity Analysis

Mild, AHI = 5

ICUR ($/QALY)

Moderate, AHI = 15

ICUR ($/QALY)

Severe, AHI = 30

ICUR ($/QALY)

Severe, AHI = 60

ICUR ($/QALY)

Base case

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

Ext. Dom (0.00)

175,543 (0)

Dominated (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

8,058 (0.03)

9,276 (0.93)

Dominated (0.05)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

7,420 (0.79)

Dominated (0.03)

Dominated (0.17)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.01)

Dominated (0)

Ext. Dom (0.00)

17,125 (0.99)

Discount rate, 0%

No Treatment

MMA ± GTA

PAP therapy

MAD

-ref- (0.90)

110,878 (0.10)

Dominated (0.00)

Dominated (0.00)

No Treatment

MMA ± GTA PAP therapy

MAD

-ref- (0)

Ext. Dom (0.27)

Dominated (0.00)

7,276 (0.73)

No Treatment

MMA ± GTA PAP therapy

MAD

-ref- (0)

6,584 (0.47)

33,117 (0.50)

Dominated (0.03)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.00)

Dominated (0)

Ext. Dom (0.00)

8,585 (1.00)

Discount rate, 3%

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

Ext. Dom (0)

154,525 (0)

4,316,789 (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

Ext. Dom (0.02)

7,872 (0.89)

Dominated (0.10)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

7,179 (0.70)

Dominated (0.03)

Dominated (0.26)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.00)

Dominated (0)

Ext. Dom (0.00)

13,197 (1.00)

AHI estimates from NMA

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

Dominated (0)

172,025 (0.00)

Dominated (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

7,954 (0.03)

9,712 (0.93)

Dominated (0.04)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

9,535 (0.13)

Dominated (0.51)

28,630 (0.36)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.01)

Dominated (0.00)

Dominated (0)

17,095 (0.99)

Complete Discontinuation after last observed period

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

167,265 (0.00)

232,813 (0)

609,737 (0)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.00)

8,001 (0.04)

14,402 (0.40)

41,798 (0.56)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

7,582 (0.08)

Dominated (0.00)

19,103 (0.92)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.01)

Ext. Dom (0)

Dominated (0)

16,899 (0.99)

Discontinuation for mild-to-moderate OSA, Izci et al.168 reported rates (higher values)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

188,918 (0.00)

Ext. Dom (0)

312,789 (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.00)

8,645 (0.03)

Ext. Dom* (0.01)

17,758 (0.96)

Not impacted

Discontinuation for mild-to-moderate OSA, Walker et al.167 (lower values)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

Ext. Dom (0)

168,939 (0.00)

Dominated (0)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

Ext. Dom (0.02)

7,849 (0.98)

Dominated (0.00)

Not impacted

Refusal to fill CPAP prescription higher, 32.4%

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

Ext. Dom (0)

175,372 (0.00)

23,179,015 (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

8,026 (0.00)

8,421 (0.95)

Dominated (0.05)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

7,537 (0.36)

Dominated (0.04)

34,271 (0.59)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.01)

Ext. Dom (0.00)

Dominated (0)

17,154 (0.99)

Refusal to fill CPAP prescription lower, 4.5%

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

Ext. Dom (0.00)

175,107 (0)

16,399,477 (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.00)

8,104 (0.19)

11,130 (0.76)

Dominated (0.04)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

7,461 (0.86)

Dominated (0.02)

Dominated (0.12)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.02)

Dominated (0)

Ext. Dom (0.00)

17,108 (0.98)

No refusal to fill CPAP prescription

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

Ext. Dom (0.00)

174,672 (0)

71,573,702 (0)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

8,071 (0.57)

63,566 (0.40)

Dominated (0.03)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0)

Ext. Dom (0.01)

7,516 (0.90)

Dominated (0.09)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.01)

Dominated (0)

Ext. Dom (0.00)

17,092 (0.99)

Perfect adherence, for all treatment strategies

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

155,365 (0.00)

559,444 (0)

Dominated (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.00)

7,788 (0.06)

11,904 (0.90)

Dominated (0.04)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

6,890 (0.92)

Dominated (0.04)

Dominated (0.04)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.01)

Dominated (0)

Ext. Dom (0.01)

17,313 (0.97)

Treatment response of partially adherent patients equals treatment response of those who have discontinued

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1)

Ext. Dom (0)

192,216 (0.00)

1,182,055 (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

Ext. Dom (0)

8,979 (0.84)

Dominated (0.16)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

10,166 (0.14)

Dominated (0.03)

18,118 (0.83)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.01)

Dominated (0)

Dominated (0.00)

16,919 (0.99)

Treatment response of partially adherent patients equals treatment response of perfectly adherent patients

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

Ext. Dom (0.00)

141,785 (0.00)

Dominated (0)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

7,775 (0.03)

10,899 (0.93)

Dominated (0.03)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

6,896 (0.90)

Dominated (0.05)

Dominated (0.05)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.01)

Dominated (0)

Ext. Dom (0.01)

17,135 (0.98)

Utility calculation for joint health states, multiplicative approach

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

157,982 (0.00)

126,360 (0.00)

3,322,371 (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.00)

Ext. Dom (0.01)

Ext. Dom (0.04)

13,052 (0.95)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

10,374 (0.06)

Dominated (0.00)

13,928 (0.94)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.01)

Dominated (0)

Ext. Dom (0.00)

16,078 (0.99)

Treatment-specific utility increments

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.09)

21,659 (0.35)

Dominated (0.15)

34,873 (0.41)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

6,024 (0.34)

Dominated (0.29)

44,443 (0.38)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

6,057 (0.57)

Dominated (0.03)

221,712 (0.40)

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref- (0.00)

Ext. Dom (0.03)

Ext. Dom (0.09)

12,304 (0.88)

Alternative refill strategy for PAP therapy (yearly cost = $2,686)

No Treatment

MAD

MMA ± GTA PAP therapy

-ref- (1.00)

175,881 (0)

45,391,043 (0.00)

Dominated (0)

No Treatment

MAD

MMA ± GTA PAP therapy

-ref- (0)

8,202 (0.96)

Dominated (0.04)

Dominated (0)

No Treatment

MAD

MMA ± GTA PAP therapy

-ref- (0.00)

Ext. Dom (0.04)

12,335 (0.64)

101,085 (0.32)

No Treatment

MAD

MMA ± GTA PAP therapy

-ref- (0.02)

Dominated (0)

17,101 (0.98)

Dominated (0)

Only PAP generator is reimbursed

No treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.19)

20,102 (0.81)

1,176,949 (0)

4,456,111 (0.00)

No treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

832 (0.82)

71,754 (0.16)

Dominated (0.02)

No treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0)

2,207 (0.92)

Dominated (0.01)

Dominated (0.07)

No treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.01)

10,212 (0.02)

Dominated (0)

18,963 (0.97)

Cost of replacement MAD same as the first-time cost

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (1.00)

181,215 (0.00)

Ext. Dom (0)

815,876 (0.00)

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.00)

8,090 (0.55)

56,830 (0.31)

Dominated (0.14)

No Treatment

PAP therapy

MMA ± GTA

MAD

-ref- (0)

7,498 (0.81)

Dominated (0.17)

Dominated (0.01)

No treatment

PAP therapy

MAD

MMA ± GTA

-ref- (0.01)

Ext. Dom (0.00)

Dominated (0)

16,994 (0.99)

AHI = Apnea-Hypopnea Index; Ext Dom = extendedly dominated; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; QALY = quality-adjusted life-year; PAP = positive airway pressure.
Note: Bolded results represent those that deviate significantly from the base-case in terms of interpretation if assuming a cost-effectiveness threshold of $50,000/QALY.


3. Subgroup Analysis

The results of the subgroup analysis are presented in Table 110 with a detailed explanation found below. It is important to note that the clinical review did not identify any differences in the effect of treatment on mean AHI and blood pressure reduction by these subgroups. The cost-effectiveness is rather affected by differences in baseline event risks by these subgroups.

Sex: The clinical review did not identify differences in treatment estimates by sex. Differences exist between males and females given that the baseline event risks are sex-specific. Despite these differences, the cost-effectiveness results remained consistent between sexes. The overall economic findings remained largely unchanged, although the ICUR for MMA with or without GTA was found to be generally lower in the female subgroup, as their longer longevity resulted in greater clinical benefits accrued that surgery, a permanent treatment, offered to their condition. At a willingness-to-pay threshold of $50,000/QALY, the set of interventions most likely to be cost-effective according to increasing levels of disease severity was the same as the base case.

Age: To determine whether the cost-effectiveness results varied according to age, two subgroups were explored: a starting age of 30 and 70 years. Similarly, age had an impact on the baseline rate of clinical events, as patients have lower incidence at an earlier age but the lifetime risks are higher. With a younger age, surgery became a more attractive intervention given the permanence of the effects of surgery and the lack of concern relating to discontinuation and adherence. Regardless, the overall cost-effectiveness findings remained robust regardless of the age cohort.

Smokers: In the base case, it was assumed that all patients were non-smokers. In contrast, if all patients were current smokers, this would affect the risks of CV events predicted by the Framingham risk equations. It was observed that the model results remained robust even when modelling a cohort of patients who were current smokers.

Diabetes: Likewise, another parameter that is predefined in the Framingham risk equation is patients' diabetes status. In the base case, it was assumed that all patients did not have diabetes and a scenario analysis was conducted in which the risk equations were revised to reflect patients with diabetes. Although this would increase the overall baseline risk of CV events, the model results were found to remain robust, as the set of treatments most likely to be cost-effective across different OSA severities remained the same at a willingness-to-pay threshold of $50,000/QALY.

Hypertensive: In the base case, the proportion of OSA patients with hypertension was defined according to general Canadian prevalence rates (male: 33%; female: 33.6%).145 In the model, those who were hypertensive would incur higher costs, arising from the need for additional medications. In addition, hypertension increases the risk of CVEs and their associated costs and QoL impact. A scenario was tested in which all patients had hypertension at baseline. The model's results remained robust with no changes to the set of treatments that would most likely be considered cost-effective across different OSA severities at a willingness-to-pay threshold of $50,000/QALY.


Table 110: Incremental Cost-Utility Results of Subgroups

 

Mild, AHI = 5

ICUR($/QALY)

Moderate, AHI = 15

ICUR ($/QALY)

Severe, AHI = 30

ICUR ($/QALY)

Severe, AHI = 60

ICUR ($/QALY)

Sex

Male

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

Ext. Dom

162,076

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

8,061

10,005

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

7,323

DominatedDominated

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref-

Dominated

Ext. Dom

7,845

Female

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

Ext. Dom

Ext. Dom

191,921

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

8,049

8,703

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

7,769

Dominated

Dominated

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref-

Dominated

Ext. Dom

16,558

Age (years)

30

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

97,568

Dominated

86,737

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

Ext. Dom

6,768

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

7,078

Dominated

Dominated

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref-

Dominated

Ext. Dom

14,694

70

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

Ext. Dom

259,820

406,612

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

8,525

14,405

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

6,972

Dominated

Dominated

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref-

Dominated

Ext. Dom

18,492

Smokers

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

Ext. Dom

149,475

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

8,486

10,176

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

8,143

Dominated

Dominated

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref-

Dominated

Ext. Dom

22,358

Diabetic

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

Ext. Dom

151,484

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

8,393

9,776

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

8,226

Dominated

Dominated

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref-

Dominated

Ext. Dom

24,650

Hypertensives

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

Ext. Dom

275,075

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

10,330

10,847

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

9,947

Dominated

Dominated

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref-

Dominated

Ext. Dom

20,884

AHI = Apnea-Hypopnea Index; Ext Dom = extendedly dominated; GTA = genial tubercle advancement; ICUR = incremental cost-utility ratio; MAD = mandibular advancement device; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; QALY = quality-adjusted life-year; PAP = positive airway pressure.


4. Scenario Analyses

Excluding MVA: The reference case assumed all patients were drivers and, thus, would be at risk of an MVA. In the situation where no patients drive, this would remove the impact of OSA on MVAs. The model was found to be robust to this scenario as the results followed a similar pattern to the base case, with the intervention most likely to be cost-effective a function of baseline disease severity. The AHI cut-off thresholds in which an intervention that was considered most likely cost-effective switched to another remained consistent. Across a willingness-to-pay threshold of $50,000/QALY, the set of strategies most likely to be cost-effective were [AHI < 15] no treatment; [15 ≤ AHI < 26] MAD; [26 ≤ AHI < 30] MMA with or without GTA; [30 ≤ AHI < 33] PAP therapy and [AHI ≥ 32] surgery.

Table 111: Incremental Cost-Utility Results Under the Assumption That Patients Do Not Drive

 

Mild, AHI = 5

ICUR ($/QALY)

Moderate, AHI = 15

ICUR ($/QALY)

Severe, AHI = 30

ICUR ($/QALY)

Severe, AHI = 60

ICUR ($/QALY)

No Drivers

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

344,932

555,222

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

8,006

10,042

Dominated

No Treatment

PAP therapy

MAD

MMA ± GTA

-ref-

7,416

Dominated

Dominated

No Treatment

MAD

PAP therapy

MMA ± GTA

-ref-

Ext. Dom

Ext. Dom

16,631

AHI = Apnea-Hypopnea Index; Ext Dom = extendedly dominated; GTA = genial tubercle advancement; ICUR = incremental cost-utility ratio; MAD = mandibular advancement device; MMA = maxillomandibular advancement; PAP = positive airway pressure; QALY = quality-adjusted life-year.

Overweight or obese patients: As noted, weight loss is a suitable intervention in patients who are overweight or obese. A scenario analysis that compared the base-case strategies to weight loss was conducted. Using treatment effect estimates from studies that recruited overweight or obese patients,5,63,76 the model findings were found to differ from the base case mainly with respect to mild OSA. For patients with a baseline AHI value of 5 to 15, weight loss was the most likely cost-effective strategy at a willingness-to-pay threshold of $50,000/QALY with an ICUR between $19,058/QALY and $727/QALY (Table 112).

Table 112: Probabilistic Results in Overweight or Obese Patients Suitable for Weight-Loss Interventions

Intervention

Expected Costs, $

Expected QALYs

Incremental Cost, $

Incremental QALYs

ICUR, sequential ($/QALY)

Baseline: Mild OSA (AHI = 5)

No treatment

11,625

10.910

-reference-

Weight loss

12,222

10.942

597

0.031

19,058

PAP therapy

18,132

10.946

5,910

0.004

Ext. Dom

MAD

18,988

10.952

6,766

0.010

647,830

MMA ± GTA

23,659

10.953

4,671

0.001

8,707,585

Baseline: Moderate OSA (AHI = 15)

No treatment

11,749

9.550

-reference-

Weight loss

12,394

10.437

645

0.887

727

PAP therapy

18,261

10.358

5,867

-0.080

Dominated

MAD

19,128

10.450

6,734

0.013

524,649

MMA ± GTA

23,881

10.338

4,753

-0.113

Dominated

AHI = Apnea-Hypopnea Index; Ext Dom = extendedly dominated; GTA = genial tubercle advancement; ICUR = incremental cost-utility ratio; MAD = mandibular advancement device; MMA = maxillomandibular advancement; QALY = quality-adjusted life-year; PAP = positive airway pressure.

5. Exploratory Analysis

Given that no information on the effects of EPAP or positional therapy on blood pressure were found in the clinical review, an exploratory analysis was conducted to evaluate the potential cost-effectiveness of these interventions while making assumptions on its potential treatment effect with respect to blood pressure reduction. Two different assumptions were explored, the more conservative being that blood pressure reduction would be similar to that which was observed for MAD (mean change in SBP = 2.1), and the less conservative assuming that the effectiveness of EPAP and positional therapy on blood pressure would be the same as PAP therapy (mean change in SBP = 3.5).

Regardless of which assumption was taken to approximate treatment effect on blood pressure, the model results remained robust, as the set of interventions most likely to be cost-effective at a willingness-to-pay threshold of $50,000/QALY remained the same as the base-case scenario when EPAP or positional therapy were included in the analysis (Table 113). The overall cost-effectiveness conclusions therefore remained the same.

Table 113: Probabilistic Exploratory Analysis to Evaluate the Potential Cost-Effectiveness of EPAP and Positional Therapy

Scenario

Mild, AHI = 10

ICUR ($/QALY)

Moderate, AHI = 25

ICUR ($/QALY)

Severe, AHI = 40

ICUR ($/QALY)

Severe, AHI = 60

ICUR ($/QALY)

EPAP

SBP ~ MAD

No Treatment

PAP therapy

EPAP

MAD

MMA ± GTA

-ref-

Ext. Dom

Ext. Dom

175,281

15,206,081

No Treatment

PAP therapy

EPAP

MAD

MMA ± GTA

-ref-

Ext. Dom

7,894

44,385

Dominated

No Treatment

PAP therapy

MAD

EPAP

MMA ± GTA

-ref-

7,410

Dominated

Dominated

Dominated

No Treatment

EPAP

MAD

PAP therapy

MMA ± GTA

-ref-

Dominated

Dominated

Ext. Dom

17,033

SBP ~ PAP

No Treatment

PAP therapy

EPAP

MAD

MMA ± GTA

-ref-

Ext. Dom

Ext. Dom

118,104

1.3x108

No Treatment

PAP therapy

EPAP

MAD

MMA ± GTA

-ref-

Ext. Dom

7,854

80,301

Dominated

No Treatment

PAP therapy

MAD

EPAP

MMA ± GTA

-ref-

7,480

Dominated

Dominated

Dominated

No Treatment

ePAP

MAD

PAP therapy

MMA ± GTA

-ref-

Dominated

Dominated

Ext. Dom

17,109

Positional Therapy

SBP ~ MAD

No Treatment

Pos therapy

PAP therapy

MAD

MMA ± GTA

-ref-

Dominated

Ext. Dom

175,933

79,958,614

No Treatment

Pos therapy

PAP therapy

MAD

MMA ± GTA

-ref-

2,664

Ext. Dom

14,462

Dominated

No Treatment

Pos therapy

PAP therapy

MAD

MMA ± GTA

-ref-

6,010

8,356

Dominated

Dominated

No Treatment

Pos therapy

MAD

PAP therapy

MMA ± GTA

-ref-

Ext. Dom

Dominated

Ext. Dom

17,110

SBP ~ PAP

No Treatment

Pos therapy

PAP therapy

MAD

MMA ± GTA

-ref-

Dominated

Ext. Dom

176,569

Dominated

No Treatment

Pos therapy

PAP therapy

MAD

MMA ± GTA

-ref-

2,500

Ext. Dom

14,915

Dominated

No Treatment

Pos therapy

PAP therapy

MAD

MMA ± GTA

-ref-

5,917

8,520

Dominated

Dominated

No Treatment

Pos therapy

MAD

PAP therapy

MMA ± GTA

-ref-

Ext. Dom

Dominated

Ext. Dom

17,036

AHI = Apnea-Hypopnea Index; CEAC = cost-effectiveness acceptability curve; Ext Dom = extendedly dominated; GTA = genial tubercle advancement; ICUR = incremental cost-utility ratio; MAD = mandibular advancement device; MMA = maxilomandibular advancement; QALY = quality-adjusted life-year; PAP = positive airway pressure; SBP = systolic blood pressure.

4.3. Summary of Economic Evaluation Results

The economic evaluation presented here is, to our knowledge, the first cost-effectiveness study published to compare all relevant treatment options in Canada and to reflect the implications for long-term costs and QALYs by stratifying baseline OSA severity. We modelled each treatment strategy individually and did not look into a treatment sequence.

The base-case analysis suggested that the cost-effectiveness of treatments for OSA is dependent on the patient's disease severity, as measured by AHI. For mild OSA, the model suggested that no treatment was the most likely cost-effective strategy up to a willingness-to-pay of $175,543/QALY, as the treatments were either extendedly dominated or were associated with a high ICUR value. For moderate OSA, MAD was the most likely cost-effective strategy at the lower range (15 ≤ AHI ≤ 25), and switched to MMA with or without GTA if the baseline AHI value was greater than 25. Likewise, for severe OSA, PAP-based therapy emerged as the most likely cost-effective strategy at the lower rate AHI of 32 and switched to MMA with or without GTA with subsequently higher AHI values (Figure 3). It is important to note that the thresholds must be interpreted with caution, as they are an artifact of the model structure: treatment effect was modelled as a continuous outcome, while natural history of OSA was modelled categorically based on the available literature.

In the economic model, gains in QALYs are achieved with treatment because of their impact in reducing AHI and blood pressure, which had subsequent morbidity and mortality impacts. The absolute gains in QALYs followed a unimodal distribution and were a function of disease severity. Those with mild or extremely severe OSA (AHI~60) had lower gains in QALYs, whereas the largest gains were observed in patients whose baseline severity reduced from severe (AHI ≥ 30) or moderate (15 < AHI < 30) to mild-to-moderate OSA (AHI < 30) or mild OSA (AHI < 15), respectively. Incremental costs were largely driven by the costs of treatment and long-term maintenance costs, given the longer life expectancies of patients on treatment.

Extensive sensitivity and scenario analyses were conducted to test the assumptions and parameters informing the economic model and to explore the heterogeneity in the reimbursement structures across Canadian jurisdictions. In most circumstances, the model conclusions remained robust. It is interesting to note that, when oral appliances were assumed to be out-of-pocket expenses, this strategy appeared on the efficiency frontier for the lower condition severity (i.e., mild OSA), as the expected cost for MAD was less than the no-treatment strategy. At a willingness-to-pay threshold of $50,000/QALY, MAD would be considered the most likely cost-effective intervention from mild-to-moderate OSA as the treatment costs are shifted to patients. Sensitivity analyses further highlighted that the model was most sensitive to parameters relating to discontinuation and adherence, especially in the context of PAP therapy. The findings presented here are aligned to a poster that was presented at the Society for Medical Decision Making and that compared CPAP, oral appliance, and surgery (i.e., uvulopalatopharyngoplasty and MMA) in Canadian patients with OSA. The authors similarly noted that the cost-effectiveness of treatment for OSA is critically dependent on the adherence with CPAP.195 Their model was further sensitive to the effectiveness of oral appliance treatment and the cost of uvulopalatopharyngoplasty surgery, although direct comparison of their model to ours is difficult, given the limited reporting in their current presentation format. Our model was further sensitive to the reimbursement of PAP-related equipment and/or accessories and the cost of replacements for MAD. Although subgroup analyses were conducted, the model's findings were found to not differ by subgroup. This may be partly explained by the fact that differences between subgroups were driven solely by their differences in baseline event risks and not by differences in treatment outcomes.

 

5. Patient Perspectives and Experiences Review

This section addressed Research Question 3: What are the experiences and perspectives of adult patients, their family members, and their caregivers regarding PAP devices, EPAP valves, OAs, surgical interventions, and lifestyle modifications for the treatment of OSA?

5.1. Methods

A systematic review and thematic synthesis of the literature relevant to the research question on patient experiences and perspectives was conducted. The protocol was written a priori and followed throughout the review process. The methods reflect the intention to synthesize results of published studies to address the research question and policy question and yield results that may be useful to decision-makers.

5.1.1. Literature Search Strategy

The literature search was performed by an information specialist, using a search strategy peer-reviewed according to the PRESS checklist - an evidence-based checklist for the peer review of electronic search strategies.41

Patient experiences information was identified by searching the following bibliographic databases: MEDLINE (1946-), with in-process records and daily updates, via Ovid; Embase (1974-) via Ovid; PsycINFO (1967-) via Ovid; CINAHL (1981-) via EBSCO; and PubMed. The search strategy comprised both controlled vocabulary, such as the National Library of Medicine's MeSH terms, and keywords. The main search concepts were sleep apnea, sleep-disordered breathing, and terms related to patient experiences, perspectives, beliefs, and values. No methodological filters were applied to limit retrieval by study design. Retrieval was limited to documents published since January 1, 2006. Results were limited to English- or French-language publications. Conference abstracts were excluded from the search results. The detailed strategy can be found in Appendix 1.

The search was completed on March 3, 2016. Regular alerts were established to update the search until the publication of the final report. Regular search updates were performed on databases that do not provide alert services.

Grey literature (literature that is not commercially published) was identified by searching sources identified in the Grey Matters checklist42 (https://www.cadth.ca/grey-matters), which includes the websites of clinical trial registries, regulatory agencies, Health Technology Assessment agencies, clinical guideline repositories, and professional associations. Google and other Internet search engines were used to search for additional Web-based materials.

5.1.2. Selection Criteria

Eligible reports were those published in English or French of any design that explored or assessed perspectives of adults being treated for OSA, or waiting for treatment for OSA, as well as the perspectives of their partners or other non-clinical caregivers. To be eligible, studies had to explore or assess participants' own perspectives directly. Studies that provided information collected only indirectly - e.g., clinician perspective - were excluded. The following types of publications were excluded: theses and dissertations, data presented in abstract form only, book chapters, editorials, and letters to the editor. Selection criteria are presented in Table 114, characterized following the PICOS elements.

Table 114: Eligibility Criteria

Population

  • Adults (i.e., aged ≥ 18 years),a diagnosed with any severity of OSA (either treatment-naive or previously treated), as measured objectively by PSG or portable monitoring (Type I to Type IV sleep monitors)b
  • Partners, family members, and non-clinical caregivers of adults with OSA

Intervention

  • PAP devices, as follows:
    • APAP
    • BiPAP
    • CPAP
    • EPAP
  • Oral appliances, as follows:
    • MADc
    • Tongue-retaining device
  • Surgical interventions, as follows:d
    • GTA
    • MMA
  • Lifestyle modifications,e as follows:
    • Exercise program
    • Diet or weight-loss program
    • Positional therapy
  • Combination therapy (i.e., combinations of 2 or more interventions in scope)
  • Inactive treatments (e.g., pre-treatment, supportive care)
  • Waiting list
  • No treatment

Comparator

  • Not applicable

Outcomes

  • Perspectives and experiences regarding treatments, including such issues as perspectives and beliefs about interventions; experiences waiting for treatment, with shared decision-making regarding treatment; experiences complying or not complying with treatment; reasons for complying and not complying with treatment; and other issues of importance to patients that emerge in the analysis.f

Study Design

  • Descriptive studies:
    • Qualitative studies
    • Surveys
    • Mixed methods studies
  • SRs of descriptive studies

APAP = autotitrating positive airway pressure; BiPAP = bilevel positive airway pressure; CPAP = continuous positive airway pressure; EPAP = expiratory positive airway pressure; GTA = genial tubercle advancement; MAD = mandibular advancement device; MMA = maxillomandibular advancement; OSA = obstructive sleep apnea; PAP = positive airway pressure; PSG = polysomnography; SR = systematic review.
a Studies that included participants aged < 18 years were included if > 80% were adults aged ≥ 18 years or if data for participants aged > 18 years were presented separately.
b Studies that did not identify criteria for diagnosing OSA were included. Studies that included non-OSA were included if > 80% were diagnosed with any severity of OSA or if data for participants with OSA were presented separately.
c Personalized MADs only were included, and not any over-the-counter, non-personalized devices.
d GTA and MMA surgeries are most frequently performed to treat OSA compared with other surgical procedures. In addition, bariatric surgery is out of scope as it is conceived more as a surgical procedure conducted primarily to produce weight loss, but not to treat OSA. [Major Roch Messier, Canadian Armed Forces, Valcartier Regional Dental Specialty Center, Valcartier, QC, expert opinion: 2016 Jun]. Note that the literature search was conducted prior to the decision to exclude other surgical procedures for OSA.
e Lifestyle interventions including clinician-directed or -prescribed programs were considered as interventions.
f Outcomes of relevance to this research question emerged from the data reported within included study reports. This preliminary list of outcomes is provided to outline issues that were expected to emerge at the outset of this review; however, the final set of outcomes emerged after iterative and careful readings of the data, as reflected in the Results section.

5.1.3. Screening and Selection of Studies

Two reviewers (SG and TR) independently screened the titles and abstracts of all citations retrieved from the literature search, and excluded reports that clearly do not meet the eligibility criteria. The full texts of all potentially relevant reports were ordered for detailed review. Two reviewers (SG and TR) independently reviewed the full-text articles based on the detailed eligibility criteria. Any disagreements among reviewers were resolved through discussion.

5.1.4. Article Sampling

As opposed to quantitative synthesis, which aims for exhaustive sampling, in qualitative and mixed methods synthesis, the principle of saturation drives sampling decisions. In this review, we applied the concept of "conceptual saturation" to developing our sample of included studies. Conceptual saturation refers to the stage when analysis of further evidence provides little in terms of further themes, insights, perspectives, or information.196 In qualitative syntheses, it is not always necessary to include every eligible study in the analysis, if the results of the synthesis are not substantially changed, or enriched, with the analysis of further study reports that contain the same concept.

To develop our sample of included articles for analysis, based on the list of eligible full-text articles, we used a purposeful sampling strategy that applied the technique of maximum variation until conceptual saturation was reached.196 The maximum variation sampling strategy helped ensure that a range of articles representing diverse experiences with all relevant interventions for OSA were included. Articles were first sampled to ensure that each OSA intervention was represented. Further sampling was done to ensure a range of experiences, if and when possible, as all experiences were not represented (e.g., limited experiences regarding surgery were found). For example, within subsets of articles for each included OSA intervention, we included studies that described the experiences of diverse patients, caregivers, or family members in varied contexts, by sampling articles based on their description of experiences by factors that, based on the clinical literature, appear to influence treatment experience. We included such factors as participants' sex and age; rural, remote, or urban settings; OSA severity; duration of treatment; prior experience with treatment; access to treatment; and comorbidities. If no literature was found regarding certain experiences (e.g., experiences of female surgical patients), this was noted as a limitation.

All eligible SRs were included and analyzed, and the maximum variation sampling technique was then applied within a subset of articles relevant to each OSA intervention. To assist with sampling decisions, the full text of each eligible primary study was reviewed and emergent concepts (e.g., benefits to self and bed partner, adherence, comfort, information, and access) were extracted along with data to describe the participant population (e.g., age, sex, OSA severity, prior experience with treatment). See "Coding (Stage 1)" for a description of how emergent concepts were identified. Further, at this stage a preliminary assessment of "information richness" was made, guided by initial impressions of study methodology, reporting quality, and relevance to the research question. Each eligible article was then classified as "good," "moderate," or "poor." A good study provided rich data, often collected through interviews or focus groups; detailed information regarding study methods; and represented a diversity of participant perspectives. For example, the study by Henry et al.197 was considered good as it provided depth and breadth of the patient experience with many supportive quotes. A moderate study lacked details regarding study methods or study participants (e.g., representativeness of the sample population was uncertain) or provided a limited scope of the patient experience (e.g., investigator-driven questionnaire on one aspect of treatment such as comfort). For example, the study by Nolan et al.198 was considered moderate as it lacked rich data, while the study by Rodgers et al.199 lacked details about the study participants. A poor study provided thin data; for example, one question to participants regarding their overall satisfaction with their treatment, and limited detail regarding study methods. Typically, poor studies provided only one or two sentences on the patient experience, such as the included study by Goodday et al., which stated, "Eight patients reported a favorable change after MMA surgery. However, all nine patients considered the surgery a worthwhile experience, and eight would recommend the surgery to others." Within intervention categories, studies classified as "good" were analyzed first, followed by studies that were classified as "moderate" and then "poor."

Sampling of articles related to a particular emergent concept, related to each OSA intervention, ceased when conceptual saturation was reached or if there were no more relevant articles for inclusion. When we suspected saturation had been reached, we reviewed two additional articles to confirm saturation. If saturation was confirmed, no further sampling occurred. If saturation was not confirmed, further articles were sampled and analyzed until reviewers agreed that saturation had been reached.

5.1.5. Data Extraction Strategy

From each eligible article, descriptive data were extracted by one reviewer (either SG or TR) into an a priori developed standardized electronic form (see Appendix 24). Descriptive data included such items as first author, article title, study objectives, participant characteristics, and study design.200 The extracted data were verified by the other reviewer (either SG or TR). Discrepancies were resolved through discussion or referral to a third party if necessary (LW).

Further, result statements from all included (i.e., sampled) articles relevant to the research question were captured for analysis, or coded, using NVivo qualitative data analysis software (QSR International Pty Ltd. Version 11, 2015).201 (For further detail, refer to the Thematic Analysis section.) Result statements are typically presented within the "results" section of a report, and are characterized as data-driven and integrated findings based in participant experiences. Before being coded, each result statement was assessed to ensure it was differentiated from raw data, methods, external data, and researchers' own conclusions and implications.202 The latter were not coded. Only results presented within the main report, but not the abstract, were coded. Data from figures were not used unless data points were explicitly labelled.

5.1.6. Quality Assessment Strategy

One reviewer (either SG or TR) independently assessed the quality of each included study using standardized criteria, depending on the study design. The other reviewer (either SG or TR) verified the assessments. Disagreements were resolved by discussion or referral to a third party (LW). Qualitative studies were assessed using criteria outlined in the Critical Appraisal Skills Programme (CASP) checklist (Appendix 25).203 Likewise, survey studies were assessed using standardized criteria commonly applied to assess the validity and reliability of this approach (Appendix 26).204,205 Systematic reviews were assessed using the JBI Critical Appraisal Checklist for Systematic Reviews and Research Syntheses (Appendix 27).206

The results of the quality assessment process are reported narratively and summarized in a table to highlight the strengths and limitations of each study. Quality assessment was not used as a basis for excluding any studies deemed to be of low quality, although by nature of qualitative data analysis, information-rich and higher-quality studies tend to receive more attention in the analysis as they provide more information relevant to the research and policy questions.

5.1.7. Data Analysis Strategy

Descriptive Analysis

A descriptive analysis of study and patient characteristics was conducted, with the goal of characterizing the set of included studies in terms of important study and patient characteristics (e.g., PICOS, sample size). This involved the calculation of frequencies for relevant categories across studies, summarizing study and patient characteristics in tables as presented in Appendix 32 and Appendix 33, respectively, and developing an accompanying narrative summary.

Thematic Analysis

We conducted a thematic analysis comprising three stages: coding, developing descriptive themes, and developing analytic themes. The analysis was conducted using QSR International's NVivo 11 Software.201

Coding (Stage 1)

The results section of the sampled studies was coded line by line for meaning and content. Coding began with an a priori "start list" of codes developed based on the research questions and emerging concepts from the one included systematic review;207 for example, perceived benefits and concerns, adherence, ineligibility, and access. As coding progressed, other codes not on the start list were added inductively to capture unexpected content. As new codes emerged, all data were recoded to search for further instances of that code. Codes were assigned to results data from all studies regardless of their design in a consistent manner of inductive and iterative coding. Quantitative data were coded in the same manner as the qualitative data, in an approach of qualifying the quantitative data.208

Through a staged coding process, two researchers (SG and TR) coded the first five "good," or information-rich, studies from an alphabetical list of all eligible studies. They independently assigned codes to concepts, ideas, and categories to the results reported within each study report. The two researchers (SG and TR) then compared and discussed their code assignments for the selected studies. This discussion allowed us to organize and reflect upon a wide range of interpretations across the body of research and refine the emerging coding template. Following this discussion, another set of three articles were coded independently, with reviewers subsequently meeting to compare and discuss coding assignments, and refine the coding template accordingly. Another three articles were coded until alignment was reached. After coding was found to align, it proceeded with one researcher as the primary coder, and the other researcher verifying the coding until saturation was reached. At this point, the text assigned to each code was read to assess consistency in interpretation and application, and to determine whether any additional levels of coding were needed.

Descriptive Themes (Stage 2)

In the second stage of the analysis, the codes developed in the prior stage were organized into related areas to construct "descriptive" themes. In this process, two reviewers (SG and TR) independently assessed similarities and differences between codes. New codes were created during this process in order to capture the meaning of groups of initial codes.

Reviewers assessed whether emergent themes were transferable across different study contexts. When they were found to be not transferable, they discussed whether the differences were a result of methods or sample characteristics. By seeking out differences in this way, the range of perspectives held by people became apparent, and subgroups were identified; for example, age, sex, or experience with OSA interventions.202

Once descriptive themes were identified, a draft summary of the results across the studies organized by each theme was written by one reviewer (SG or TR) and subsequently reviewed by a second reviewer (SG or TR). A group discussion took place to review and discuss the emergent themes. The final version was agreed upon by all review team members202 and represents a synthesis that remains close to the original results of the included studies, with minimal interpretation.

Analytic Themes (Stage 3)

During the final stage, the "data-driven" descriptive themes from the prior stage were analyzed through the theoretical structure provided by the policy question to develop "theory-driven" analytic themes in response to the policy question. In this stage, two reviewers (SG and TR) used the descriptive themes to independently infer an answer to the question about the optimal use of treatments for OSA. After each reviewer made these inferences independently, the two reviewers reviewed their results. A group discussion including all team members (SG, TR, and LW) was held to discuss the analytic themes in the context of the policy issue. This cyclical process of theme development resulting from group discussions continued until a set of themes emerged that is inclusive of all of the initial descriptive themes and answers the policy question.202 As in the prior stage, throughout this process, reviewers ensured attention was paid to the transferability of results across different contexts, as a way to determine whether some results might apply only to certain subgroups. Throughout all stages of the analysis, regular meetings between members of the research team took place to discuss emerging results, and analytic ideas. Explicit notes were kept using the memo and annotation features in NVivo to record decisions made regarding coding and theme development, to help demonstrate rigour in the analysis.

5.2. Results

A total of 2,400 citations were identified from the initial electronic database, alerts, and search updates. Of those, 2,102 were deemed ineligible and the full text of the remaining 298 citations was retrieved for eligibility screening. Ninety-three were identified as eligible. Thirty-two studies were ultimately included in the thematic synthesis, following the outlined sampling procedure.

The study selection processes are presented in a PRISMA flow diagram (Appendix 28). A list of included studies is provided in Appendix 29, a list of the remaining eligible studies is provided in Appendix 30, and a list of excluded studies is provided in Appendix 31.

5.2.1. Descriptive Analysis

Study Characteristics

Of the 32 included studies, one systematic review207 was identified, and the remaining 31 included studies were primary studies. Eleven were qualitative descriptions that used interviews as the method of data collection,209-219 one was a qualitative content analysis of a website,220 three were grounded theory studies199,221,222 that also used interviews as the method of data collection, seven were survey studies that used questionnaires to collect data,89,98,112,223-226 three were randomized controlled trials with questionnaire components,198,227,228 one was a retrospective chart review,229 three used a mixed methods approach,197,230,231 and there was one cross-sectional study232 and one cohort study,233 both of which used questionnaires. It should be noted that some of the qualitative description studies used grounded theory techniques, but stopped analysis before a theory was created.

Thirteen papers (40%) were about CPAP.197,207,210,212,213,216,217,219-221,224,230,231 Three studies were about PAP devices in general,199,214,229 while one study was on APAP.198 There were five studies on OAs, all regarding mandibular advancement devices.223,225-228 There was one study that included participants using CPAP or OA.209 Regarding surgery for OSA, there were four studies that looked at MMA.89,98,232,233 For lifestyle modifications, there was one study each on physical activity,215 diet,218 and tennis ball technique.112 Two studies211,222 with untreated patients were also included.

The sample size across the 31 primary studies ranged from nine participants214 to 296 participants.229 The systematic review207 included 22 papers. Fifteen studies included the perspectives of bed partners,197,207,209,211,213,216,217,222-226,228,230,232 although in three of these studies,209,213,230 patients reported on the views of their partners.

Nine countries were represented by the included studies: four studies were from Canada,209,228,232,233 nine from the United States,197,199,212,214,216,224,227,229,231 eight from Sweden,211,213,215,218,219,222,225,230 four from the United Kingdom,89,220,223,226 two studies from both Australia112,217 and New Zealand,207,210 and one study each from France,98 Ireland198 and Taiwan.221 For the systematic review,207 New Zealand was considered the country of origin because it is the country of the first author; however, studies from the following countries are included: Australia, Canada, China, France, Germany, Italy, Scotland, Sweden, and the United States.

All studies were published between 2006 and 2016. Two studies were published in 2006,198,223 one study was published in 2007,211 one study was published in 2008,220 three studies were published in 2009,112,226,228 one study was published in 2010,229 one study was published in 2011,217 six studies were published in 2012,213,215,219,222,224,225 six studies were published in 2013,98,197,209,212,230,231 seven studies were published in 2014,89,199,207,210,218,221,227 one study was published in 2015,214 and three studies were published in 2016.216,232,233

The characteristics of the included studies are summarized in Appendix 32.

Participant Characteristics

Regarding patient characteristics, in 13 studies, the average age of participants was 50 to 59 years.112,198,199,211,216-219,223,225,226,229,230 A range of ages (but no average) was given in three studies,207,221,222 and age was not reported in three studies.213,214,220 In eight studies, the average (or median) age of participants was 40 to 49 years.89,98,197,212,224,227,228,232 In three studies, the average age of participants was 60 to 69 years.209,215,231 In one study, the average age of participants was 30 to 39 years.233 in one study,210 the average ages of participants varied by ethnicity (i.e., Māori patients averaged 39.3 years, Pacific Peoples averaged 43.5 years, and New Zealand Europeans averaged 58.1 years).

As a whole, the female patient and male caregiver experiences were underrepresented in the studies. In all but three of the studies,212,213,222 male participants comprised greater than 50% of participants. It may be that the findings do not provide the full range of experiences for these persons. Ethnicity was not reported in most of the studies, but in the seven studies197,199,210,212,217,224,231 where it was reported, greater than 70% of participants were described as "white," "Caucasian," or "European." One study216 simply reported that the majority of participants were white, but did not provide exact proportions.

Participants in the included studies had varying degrees of OSA severity. More than half of participants at baseline had moderate OSA in four of the studies.216,218,223,225 In 11 studies, more than half of patients at baseline had severe OSA.98,198,209,211-213,221,224,227,230,233 For five studies, OSA severity ranged from mild to severe.197,207,219,222,226 Of these, AHI ranged from 8 to 135 for one study,197 and 19 to 72 for another study.222 Two studies207,219 reported severity as "moderate-to-severe," and one study reported severity as "mild-to-moderate"226 with no further detail provided. Ten studies did not report the severity of OSA.89,112,199,210,214,215,217,228,231,232

Patients had a range of experience with the OSA interventions (excluding surgery, which was a case of having had or not having surgery, although surgical patients may have tried previous interventions). Participants were experienced (i.e., not treatment-naive) with at least one intervention in 27 studies.89,98,197-199,207,209,210,212-221,224-226,228-233 Participants were treatment-naive in two studies.211,227 The views of only non-users (i.e., the views of spouses and not patients) were captured in one study.222 In two studies,112,223 it was uncertain whether participants were treatment-naive, or if they were regular users of the studied intervention.

For those who did report on the user's experience with treatment, the length of use varied. Specifically, the duration of treatment or follow-up was reported in 20 studies.89,98,112,197,198,209,210,212,213,215,219,223-227,229,230,232,233

  • CPAP: two weeks217 to 10 years219
  • OA: three weeks227 to 8.3 years209
  • Surgery follow-up: four weeks232 to 21 months233
  • Positional therapy (tennis ball technique): 2.5 years112
  • Lifestyle modification: 12 months.215

The characteristics of the included study participants are summarized in Appendix 33.

5.2.2. Quality Assessment

Overall, the quality of the included studies was moderate. Some studies had more quality concerns than others, yet the body of literature adds the perspectives and experiences of patients and their partners about treatment for OSA. The choice of study design (e.g., survey, qualitative description) and data collection tools (e.g., questionnaire, interview, focus group) greatly influenced the amount of information and range of experiences reported. It also prescribed how each study should be assessed for quality. For this reason, a portion of the summary of the quality assessment is grouped by study design, although some statements can be made for the studies as a collective group.

The maximum variation approach to article sampling for this review aimed to include a range of perspectives relating to the research question; however, the reviewers were limited to the studies identified and eligible based on the inclusion criteria. Thus, while trying to sample studies that presented the views of both males and females about OSA therapy, as a whole, it was determined that female patients and male caregivers were underrepresented in these studies.

Additionally, race was reported for eight studies197,199,210,212,216,217,224,231 while the remainder did not report on these characteristics. Of those studies that did report on race or ethnicity, the majority of participants were white. As illustrated by these examples, patient demographic information was not well reported by the studies as a whole. As well, CPAP was the most common (40% of papers) intervention studied and fewer papers were available to describe experiences with other OSA interventions.

Qualitative Studies

As a collective, the quality of the qualitative studies was good. All studies clearly reported study objectives and research questions, which were well suited to qualitative inquiry. As well, all of the qualitative studies used some measure to enhance credibility. Many of the studies used verbatim quotes as a means of demonstrating that the findings were rooted in the patient perspective.197,199,209-213,215-219,222,230 The study by Hu221 also used quotes, but it was difficult to determine whether these were reported verbatim for each participant, or were summative across participants. Two of the studies also used member checking to further ensure credibility.210,217 Measures to enhance dependability (e.g., peer review, researcher triangulation, audit trail) were reported by nine studies.211,213,215,217-219,221,222,230 The study that included Māori participants described methods as being culturally appropriate, including conducting focus groups at appropriate locations and following culturally appropriate formats.210

Two of the studies213,230 explicitly stated that they used maximum variation sampling to ensure that a range of perspectives had been captured. It was uncertain whether the remaining studies used a purposeful sampling strategy. Twelve of the studies reported that they sampled until saturation.209,211,213-219,221,222,230 It was uncertain in the remaining studies how sample size was determined. Two of the studies212,214 offered monetary compensation for participation in the study and it may have motivated certain persons to participate. Sampling strategy, typically how eligible patients were identified and recruited, was not well reported for two studies210,211 and the potential for channelling bias by physician recruitment was identified in two other studies (e.g., physicians may have been inclined to recruit certain patients for the study).197,215

Survey Studies

As a group, the survey studies were of moderate quality. A general criticism of survey studies was that based on the study design, survey studies may not capture the full range of patient experiences. For example, one of the studies231 primed patients to negatively reflect on their experience with CPAP, and questions did not allow any positive experiences to be described. Furthermore, with the exception of the study by Butterfield et al.,232 the studies regarding MMA89,98,233 were primarily questionnaires about how patients perceived their appearance after surgery, thereby not allowing participants to describe any other expected or unexpected outcomes. Two of these studies98,232 used previously validated questionnaires. The questionnaire used by Goodday233 had face validity (yes or no questions); however, the validity of the questionnaire by Islam89 is uncertain. This provided a limited perspective on the experience of the MMA patient for the purposes of this review. For two of the other studies, it was uncertain whether the questionnaires used to collect data were sufficiently validated.223,226

Another main criticism for some of these studies is that limited demographic information was provided for the participants, and in 12 studies, it is uncertain how representative these patients are of the population from which they were sampled.89,98,198,223-228,232-234 Sampling strategy was not well reported for nine of the studies.89,98,198,223-227,232 In addition, a few of the studies112,231 used postal surveys, and participants without postal addresses or those with low literacy skills may not have been able to complete the questionnaire, lending these studies to selection bias. An a priori sample size calculation was provided by two of the studies.227,228

Systematic Review

The systematic review207 was well done. The authors clearly reported their objectives and justification for their review. An adequate number and choice of databases was used for the literature search; however, a main criticism of these methods was that no repeatable search strategy was reported (although search concepts were presented). The search strategy for this current review identified 10 eligible studies235-244 that were included in the systematic review by Ward. These 10 studies were not analyzed for the purposes of this review. The stages of synthesis were reported; however, it was uncertain how many researchers took part in data synthesis. It is not clear how coding was approached, and the authors did not use a qualitative software program for the analysis; they used Microsoft Excel. This is not a critical flaw; however, it does pose a limitation in that the researchers would have an added burden of synthesizing information in a platform poorly suited to this type of inquiry. The authors justified the use of the critical appraisal tools used and two researchers took part in this phase, and findings of the critical appraisal were clearly reported.

The quality assessment of the included studies is summarized in Appendix 34.

5.2.3 Thematic Analysis

The following sections explore the results of the thematic analysis. Following the initial phases of coding, data were organized into descriptive themes. In the next phase, the descriptive themes were analyzed through consideration of the research and policy questions to identify "analytic themes." Analytic themes represent the essence of data in direct relation to the research and policy question. For this report, the analytic themes represent the meaning of those experiences and perspectives of OSA patients and their caregivers about interventions that have an impact on their optimal use. Two analytic themes emerged from the data. Figure 4 represents the emergent analytic structure, including analytic themes and their descriptive themes. Table 115 represents the emergent categories, their relationship to the descriptive themes and which ones have results related to each of the categories. In the following section, a descriptive summary of the data in terms of the analytic themes is presented, using results from the descriptive categories as supporting evidence.

Figure 4: Analytic Themes and Related Categories

Table 115: Emergent Data Categories, Descriptive Themes, and Analytic Themes

Analytic Themes

Descriptive Themes

Categories

Interventions Included in Category

Exemplary Quotes

A range of characteristics and factors influence whether people seek and initiate OSA treatment.

Motivation

Risk awareness

Pre-treatment

"I've not noticed myself that I'm snoring, but people have told me so. I've not been aware of having so many breathing interruptions. I thought I was close to death!"230

Disease chronicity

Pre-treatment, CPAP

"I felt like it was a double-edged sword. I felt relief that we were finally going to get something done, but sad that I would have to wear this for the rest of my days."210

Fear

Pre-treatment

I'm nervous about having so many apneas and they say it's not supposed to be good for the brain. It frightens me."222

Partner distress

Pre-treatment

[Spouse] "Sometimes I feel he's not breathing, so I have to nudge him to wake him up. When it got bad, I was pregnant with our 7th child, and I thought, 'OH MY GOD, he's just going to die on me!'…"197

Sense of self

CPAP, diet

"I stepped on the scale and said no that is not me…"218

Expectations and Attitudes

Pre-existing beliefs about treatment

CPAP

"I think it sounds very demanding to learn to live with a CPAP machine; a lot of demands to cope with. On the other hand, I don't want to take the risk of a stroke, or heart attack, so I have to handle it."211

Anticipated benefits

OA, CPAP, physical activity, untreated, diet

"Because there is the benefit to [the bed partner], at least you're not snoring, so you're not disrupting their sleep. Or you're not stopping breathing for 20 seconds, up to a minute, and they're wondering if you're going to gasp and get going again."209

Doubt of ability to comply with physical activity

Physical activity

NA: no direct participant quotes were provided to support this theme.

Negative perception of treatment

CPAP, tennis ball technique

"[Referring to OSA] just another way for the medical establishment to make money."199

Information Needs

Information Needs

PAP, untreated, CPAP

"I think it was just a lot of information at the time and it kind of went right past me, if you know what I mean. It was not the right time… I didn't take it on board."210

Patient Characteristics

Demographics (disabilities, age, relationship status)

CPAP, untreated, OA

"I have a shoulder I can't really go up with much. That makes it hard for me to … get the straps off and set right to put it on my head…"214

Impact on Lifestyle

Physical environment and travel

CPAP and OA, physical activity

"I have travelled 8 to 10 times internationally each year with my [model] machine for the past 5 years. I had never considered it big and heavy until I saw some more modern CPAP units earlier this year."220

Challenges to lifestyle change

Physical activity, diet, CPAP

"I feel like an alcoholic ... I use the food to lower anxiety, you know."218

Cost

Cost

CPAP and OA, PAP, diet

""There was no guarantee that the extended health would cover [the CPAP]. But I was willing to, because it was more effective."209

Interventions for OSA require adaptation to daily routines and relationships. Some patients are able to integrate these interventions into their lives and experience benefits, while others are unable to do so.

Experienced Benefits

Experienced benefits (physical, mental and psychological, social)

CPAP, CPAP, OA, surgery

"I couldn't believe what a difference in how I felt - I've never had such a good night's sleep. I slept!..."197

Side Effects

Side effects

CPAP, OA, surgery, physical activity

"I've been using the CPAP mask for - I think it's about a year, but I'm not positive. I can't stand the thing. I find that it blows all up in around my eyes throughout the night and find my range of motion from my neck is - I'm often stiff in my neck because it holds - makes it so stiff. I do use it."209

Claustrophobia

CPAP

"[With CPAP] I had tremendous panic attacks. Well, they found out in a hurry, because they put it on here and I just - it was horrid. Like, I just ripped it off and I just was panicking and, like, it was awful."209

Comfort of Intervention

Discomfort

OA, CPAP, tennis ball technique

"Sometimes it's just so uncomfortable to put that mask on at night and I wake up in the morning with a big mark on my face. Not that the mark bothers me, but it hurts. It's like sleeping, you know, with something pressing into your face all night. It's not worth it."216

Difficulty with CPAP

CPAP

"If you get up [during the night], you don't want to put it back on. I don't want to wake myself up trying to get that thing threaded through there."216

Difficulty with OA

OA

"…after a while, they break down…Either the wire on the top breaks or it becomes undone on the back or the plastic just deteriorates…"209

Difficulty with TBT

Tennis ball technique

"The tennis ball moved around."112

Impact on Self and Relationships

Perceived attractiveness

Surgery, CPAP, OA

"I didn't want anybody to see me...you know, I looked really stupid; I looked like a Martian."217

"And so, like, I don't know anybody who is in a new relationship who would stick one of those things [the CPAP] to their face, you know. It's just not too appealing. But the oral appliance is not a big deal…Don't even know it's there."209

Relationships

CPAP, surgery, OA

"… I mean, how do you say 'good night' with a big mask over your face and things like that, it's … so, it's a little sad …"219

Partner experience

Untreated, CPAP, OA

"The benefits to me have not been as great as I might have hoped, but my wife loves that mask."210

Presence of Support

Partner support

Untreated, physical activity, CPAP

"For the first couple of weeks I'd help her put it on, and make sure everything was OK, and set the machine."216

Health care professional support

Untreated, physical activity, CPAP

"That I got the possibility to talk to that CPAP nurse made us feel very secure, people you can talk to and tell how you feel. That is what's so important."213

Peer support

CPAP

"Have somebody who's been through it and has used CPAP for a while come in to talk to somebody for 10 or 15 minutes that's about to start the process… can assure somebody that you do get used to it."210

Insufficient support

CPAP, PAP

"So I feel I was in kind of a limbo. And if I should have taken any more responsibility for visiting a sleep doctor or revisiting my doctor, I didn't know that, and my doctors weren't telling me that - unlike any other aspects of health care. Like, when I saw [other specialists], my PCP made sure I knew who they were, that they were good, that I went … she'd get follow-up reports…None of that happened with sleep apnea."199

Information Needs

Information needs

CPAP and OA, diet

"[Speaking about treatment] I didn't know there were any options."199

Adaptation and Problem-Solving

Perseverance and intervention as habit or routine

CPAP, OA, physical activity, diet

"If you're not wearing it all the time, certainly the first night that you put it in, your teeth are sore in the morning. And I find I don't sleep as well, I'm a bit restless that first night, just because…it's a bit uncomfortable if you have to adjust to it being in place."209

Time

CPAP, physical activity, diet

"I'm having trouble keeping regularity ... skipping lunch now and then ... I'm hungry in the mornings, eat a little too much then because it's so very good."218

Motivation for physical activity

Physical activity

NA: no direct participant quotes were provided to support this theme.

Cost

CPAP, OA

"…after a while, they break down…Either the wire on the top breaks or it becomes undone on the back or the plastic just deteriorates…"209

Psychological Impact

Anxiety

Untreated, CPAP, surgery

"I don't go to bed before 12, and sometimes even 1 o'clock. I wait until I'm dead tired. I'm afraid of going to bed and I think this is a good way to cope with my situation."211

Confidence

Physical activity, diet, CPAP

"If I could do it last time, it should be just as easy this time."218

Embarrassment

CPAP and OA

"If there's a tangi [funeral] I'll take it with me. I don't care who's watching. At least I'll wake up feeling good."210

Fear

CPAP

"It was scary at first; it felt like one wouldn't get any air."219

Guilty

Untreated, CPAP

NA

Helplessness

Diet

"... I use to say, if most people ate as I do, no one could be fat ... surely on certain occasions I want a little chocolate ... everybody wants that, I guess ... but not in those amounts, causing me to weigh as much as I do."218

Shame

Untreated, CPAP

"In the beginning, I had the full mask and it was very disconcerting, not only physically but emotionally. I don't know that anyone else experienced this, but I really felt ashamed. I was mad at myself for being in this position and I had to build up time on the machine."216

CPAP = continuous positive airway pressure; NA = not applicable; OA = oral appliance; OSA = obstructive sleep apnea; PAP = positive airway pressure; PCP = personal care physician.

Analytic Theme: A range of characteristics and factors influence whether people seek and initiate OSA treatment.

This first theme captures results that describe the range of characteristics and factors that influence patient experience with OSA interventions, including patient demographics, and expectations and attitudes toward treatment. As described below, experiences with certain interventions may vary based on characteristics of patients. Additionally, the process of becoming diagnosed with OSA can be incredibly overwhelming for some patients. Persons with OSA might not be aware of what options are available to them, or that they even have a choice. All interventions are seen as an inconvenience for patients, at least at the beginning, and patients describe a period of adaptation to find an intervention that works for them. The following section explores the findings related to this first analytic theme.

Motivation

Risk Awareness

Both patients and partners would downplay or be in denial about the seriousness of OSA,197,199,217 which often delayed treatment.199

Similarly, many participants did recognize that being obese was related to their OSA.197,215 However, some also did not recognize, or were in denial about, the severity of their weight and the risks with being obese.197,215,218 This denial, about obesity and OSA, can prevent people from seeing their physicians and seeking treatment. In one study, from the time when patients first recognized a problem, to when they first sought medical help, men averaged five and a half years and women averaged a four-year delay.197

As one female participant stated,

I think I just kind of preferred to live in my imaginary world. It takes a lot for me to go to a doctor. I knew something was wrong; I just chose not to do anything about it. I kept saying, "I can fix this myself; I don't need a doctor, I don't need to bill this to my insurance." It was hard - I didn't want to admit I was snoring as loud as I was, and that I couldn't fix this on my own. I also didn't want to be told that I had completely lied to myself about my weight for the last 10 years. I'm so good at lying to myself - "you don't have a sleeping problem; you just have three kids and you're tired."197

Patients seemed to weigh the risks of untreated OSA with the inconvenience of treatment to make a personal decision to pursue treatment, as illustrated by this quote, "…well, once I'm aware of a condition and I know that there are options or there are no options, and I weighed it up and I take the road that's the best avenue for you, you know it's a considered option…"217

Disease Chronicity

Several studies reported a sense of awareness that patients had about their condition; this was a realization that OSA is a chronic disease, and that treatment may occur over a lifetime. Some patients were uncertain whether OSA was curable, and whether to approach it as a burden or handicap.221 Some were uncertain whether they would have to wear CPAP for the rest of their lives.210,221

Once patients realized that they might have to undergo lifelong treatment, some patients expressed a deep desire to find a cure or for doctors to be able "to just go in and fix it."197 As one man commented about CPAP, "If there were anything that could be done to be free of that machine, I'd do it right now, regardless of cost. I can't imagine being 75 years old, and still stuck on this machine."197 Lifelong treatment can be intimidating, and this was expressed by two patients in one study: "Did I need to wear CPAP equipment to sleep for a lifetime? If I accepted CPAP therapy, would my sleeping problems be cured? The physician said that suspension of treatment would cause ventricular hypertrophy due to sleep apnea. This seemed intimidating. At the end of my life, I was afraid that if I had a stroke, I would become a burden to others. This was concerning."221

Fear

Before treatment, patients expressed feelings of fear related to OSA.211,216,218,230 Patients were afraid of dying in their sleep,211,224 as well as the adverse outcomes of apneas, particularly for patients who had had previous CV problems and feared more events.211 Partners often made patients feel anxious about their condition or felt fear for their partners.211,216,222 One partner stated, "It's probably more this, future problems with his heart or a stroke or something. I'm worried about such things! His driving the car and what is currently happening; I'm scared for that, really scared."222 Some partners became afraid of the patient when they had violent mood swings or aggressive behaviour due to OSA.222

For some patients, this fear of adverse outcomes and the fear that something serious could happen motivated patients to get tested.211,218,230 One participant explained the motivation for testing like this: "I mean, surely if it is a diagnosis - this counts for you, you have got two years left to live if you do not do something drastically now - you would react quite seriously to that. At least I would."218

Patients learning about oxygen desaturations and the anxiety this may cause felt the need for quick and effective treatment.211 One patient stated, "I thought, 'Am I going to die because of this?' I'm afraid that I won't wake up again when I go to bed at night. If my breathing doesn't start again after an apnea, I suppose I'm going to die."211

Partner Distress

Prior to receiving interventions for OSA, partners of patients were distressed by their partners' lack of energy,197 their own disrupted sleep,216 interference with their own daily activities,216 and by witnessing their partner stop breathing during sleep.197,211,216 As one partner described, "It was sometimes hard for me to go to sleep because of the way he was breathing, and if he would stop breathing, then I would have to wake him up sometimes because I didn't think he was gonna ever breathe."216 Patients also recognized their partners' distress; as one commented, "It was my wife who woke me up, and I think that she thought that it was much worse than it was. She got really scared when she heard that I wasn't breathing."211

Sense of Self

Sense of self was related to how patients viewed themselves in general with relation to OSA and treatment, specifically to CPAP and weight loss. A related code of this was Perceived Attractiveness (see second analytic theme), which specifically relates to how attractive patients viewed themselves as being.

With regard to weight loss, some patients had a hard time reconciling their image with their current weight.197,218 As one patient said, "I stepped on the scale and said 'no, that is not me' ... it did not correspond to me. That was kind of what made it tilt and then all there is to do is change it ... no, I could not identify with the numbers."218 This discordant sense of self was motivation for behaviour change.218 Another participant remarked that their recent weight gain had contributed to the problem of their OSA: "I'm new at being heavy. It's a new thing for me. And I don't drink coffee during the day. If I'm sleepy at work, I drink Coke. I drink five or six cans of Coke a day. That's a lot of calories. So the problem kind of feeds on itself."197

For patients with difficulty complying with CPAP, self-image and how they thought others viewed them was a major reason for not using CPAP.217 A patient's self-efficacy was related to CPAP use and influenced a patient's decision to use CPAP based on how they viewed their health and the barriers to use.217

Expectations and Attitudes

Pre-Existing Beliefs About CPAP

The findings of the systematic review were that the pre-existing belief that patients had about CPAP shaped their subsequent experience with CPAP.207 Beliefs that CPAP would improve health, improve relationships, and reduce symptoms of OSA were reinforced when patients experienced benefits; however, those with mild OSA were increasingly likely to have poor experiences with CPAP, such as unresolved excessive sleepiness, which led to a perception that CPAP was a failed cure.207

This finding was confirmed by primary studies, which similarly found that persons viewed CPAP as a way to resolve the negative effects of OSA and formed their attitudes about CPAP.211 The perception that CPAP could relieve symptoms of OSA motivated patients to seek diagnosis and treatment.217 Hearing laypeople or professionals talk about CPAP in a positive or negative way helped shape their attitudes toward treatment.211 Although there was a hope that CPAP would improve their lives, there was also a belief that CPAP could be restrictive and demanding.211

These findings all relate to beliefs patients had about CPAP, and how pre-existing beliefs shaped a user's experience with CPAP. Beliefs about other treatments, such as OAs or surgery, are uncertain and were not identified in the literature review.

Anticipated Benefits

Anticipated benefits refer to instances where patients or partners were expecting positive outcomes of using OSA interventions. In general, persons using an OSA treatment expected to experience a reduction in apneas,209 improved sleep,209,215,230 prevention of heart disease and strokes,209,230 reduced fatigue,209 improved social life,215 improved mental health,215 reduced snoring,209 reduced mortality,230 and benefits for their partners.209

Anticipating these positive outcomes motivated some patients to use CPAP. Patients viewed CPAP as positive when they thought it could rid them of all negative aspects of having OSA.211 As one participant stated, "I was so tired last autumn that I fell asleep standing at work! When I was told I had this sleep apnea and that I would be getting the CPAP, it was, like, aah, a great relief. Feeling so bad really motivated me a lot to participate."230 Another stated, "I had so many apneas that I was afraid; yes, I wanted to spare the heart."230

Patients were also motivated to use CPAP and OAs in order to improve the sleep of their bed partners, primarily due to the reduction of their own apneas and snoring.209 One participant stated, "Because there is the benefit to [the bed partner], at least you're not snoring, so you're not disrupting their sleep. Or you're not stopping breathing for 20 seconds, up to a minute, and they're wondering if you're going to gasp and get going again."209

Regarding physical activity, some people wished to be more active, and felt that engaging in physical activity would improve their general health, social life (as some activities can be social), sleep, and mental well-being, and enhance their self-image, as well as reduce pain.215 An initial weight loss was seen as motivating physical activity.215 However, some patients and partners felt that sufficient weight loss would resolve snoring and OSA symptoms.197 As one participant stated, "I look at it as kind of temporary, you know? Keep losing the weight, and maybe that will help. Dr. R thinks it's a lifetime deal, but I hope it's not. I do know I snore less when I've lost weight."197 This was echoed by a bed partner, who said, "Ideally she'll be able to recondition her body so that it's not necessary for her to have to do this [use a CPAP machine]. But I don't know what kind of outcome is expected. Ideally, she could get back to normal."197 Regarding healthy eating, if participants could anticipate positive outcomes, then they were more likely to view themselves as able to change their behaviour.218

Participants using CPAP, OAs, and physical activity reported an anticipated improvement in overall physical health and mental well-being. Those engaged in physical activity also anticipate an improvement in their social life and their self-image. If persons modifying their diet reported they were able to anticipate positive outcomes, they are more likely to view themselves as capable of changing their behaviour. However, it is uncertain what benefits people anticipate before undergoing surgery or when using positional therapies, as these experiences were not identified in the literature.

Doubt About Ability to Comply with Physical Activity

Persons trying to lose weight by being physically active were often doubtful about their ability to be successful.215 They may have anticipated positive outcomes and knew that it was good for them, but were uncertain whether they could achieve them, as some reported having been unsuccessful in losing weight before.215

Negative Perception of Treatment

Some studies reported participants who had a negative perception of OSA treatments. Regarding OSA and CPAP, one study reported that some participants were skeptical that OSA was a real diagnosis.199 Specifically, one participant thought that OSA was "just another way for the medical establishment to make money" and this was primarily due to CPAP being sold in commercial storefronts; he was uncertain whether OSA was real and needed treatment.199

Some participants describe being anxious about CPAP treatment.217,230 One patient described their feelings of anxiety this way: "I was quite anxious about whether I could do it all by myself - putting the mask together. I think about it now and it's so bloody simple, but at the time it just seemed like too much to take in. What if I didn't get right? How would I travel?..."217 Others did not use CPAP, because of either a lack of motivation to start using it or a lack of motivation to continue using CPAP, because they did not perceive a benefit from the treatment.207,216,217

Participants in one study stated that the shifting bite associated with OAs was a concern for them, and something they would consider when choosing an intervention, however, they would not necessarily choose CPAP instead.209 It was also perceived that OAs could cause gum disease and cavities, and in some cases, OA users were told this by their dentist, but some of this speculation was based on the user's experience.209

One study reported negative expectations with positional therapy. In this study, participants reported their perception that the tennis ball technique would cause back problems.112

In summary, persons with OSA perceive negative outcomes of CPAP, OAs, and tennis ball technique. Anticipating negative outcomes is something that a patient may consider but might not necessarily push them toward another intervention, as demonstrated by the views on OAs and CPAP. Whether persons with OSA anticipate negative outcomes from surgery or lifestyle interventions remains uncertain because of a lack of research on this topic.

Information Needs

The process of being diagnosed with OSA and the initial treatment phases can be an overwhelming time for patients. Some patients need initial information regarding OSA as their knowledge of the condition was limited or non-existent.199 Patients report needing information for both themselves and their partners,211 but that the amount of information received at initial points of care can be overwhelming.210,211 Some mentioned that the timing of information made it difficult to process,210,211,217 and that information should be delivered in steps211 or in a format that is conducive to learning.217 With regard to information delivery, some patients were dissatisfied with how their physician delivered the information, particularly related to their communication style.217

However, some reported that they did not receive enough education or training related to OSA and CPAP.211,213 As one participant said, "The only information I've got is that I will receive a CPAP machine. I want more information about my apneas and sleep disturbances, but I think the information should be broken up and delivered in a couple of steps so you get time to think about it and understand it. It's impossible to understand all the information in one go."211 Another commented, "I thought it was a very advanced machine; I couldn't guess how it would work when I saw these tubes."213 Partners also felt insecure, frustrated, and unable to help patients if they had not been a part of the initiation of treatment and therefore lacked knowledge about treatment.213

Patients sought information about OSA and CPAP from a variety of places. These included:

  • The Internet210,211,217,221
  • Ear, nose, and throat physician221
  • Family and friends210,211,221
  • Media, books, and magazines221
  • CPAP professionals, including salespeople.221

One study found that sources of information varied by race; Māori and Pacific Islanders sought information about OSA and CPAP from family and friends, while New Zealand Europeans sought information from the Internet.210 Māori participants appreciated when information about OSA and CPAP was delivered in a culturally appropriate manner, which in turn furthered their understanding of the condition and treatment.210

Regarding information-seeking on CPAP, one patient said, "Well, there wasn't a lot of personal stuff in there, like people that have actually used machines. So when I was on the 'Net, I was just basically looking at people's experiences with the machines and their own journeys with it, and stuff like that."217

Information about OSA and CPAP was also misunderstood, incomplete, or difficult to understand.211,213,221 Participants also commented that information found on the internet was not always helpful, as it was not tailored to their individual circumstance, and that they wanted to ask questions of persons with professional knowledge.211,221

This section relates to information needs on OSA in general and specific information needs related to CPAP prior to initiation with treatment. Any evidence on the information patients wished they had before surgery, or before using OAs or lifestyle interventions, remains uncertain because of a lack of research on this topic.

Patient Characteristics

Patients With Disabilities

One study214 explored how PAP devices failed to meet the needs of patients with physical or sensory impairments. These patients had a range of conditions, including decreased mobility (e.g., rheumatoid arthritis, osteoarthritis); amputation; diabetic neuropathy; impaired depth perception; Parkinson disease; weakness (e.g., stroke, carpal tunnel syndrome). These patients expressed difficulties they had while using PAP devices, and provided suggestions for change. A few patients recommended modifying the headgear, changing the mechanism for connecting the tubing, and minimizing the number of attachments or adjustments needed. Patients also recommended changes in the location of their machine's control buttons for people with limited mobility or visual impairment. These suggestions included having larger knobs for people with arthritis, placing the controls or displays in the front of the machine, or making them feel different so they were more accessible at bedtime: "I have to turn the light on to find it…Actually, if it wasn't on top, [and instead] was on the front, I'd probably be able to [find it]…But I've been trying to memorize where it is so I can get it by feel." A few participants described ways in which the filter could be designed better. For example, the size of the filter could be increased to reduce the demand for fine motor control, as described by the following patient: "The filter is so small and you have to pinch it on both sides just right. It would be a little bit easier if it was a little bit bigger and easier to get to because it's always in the back and on the bottom and in a corner type…"214 It is not known whether these difficulties reduced adherence or led to the discontinuation of CPAP.

Literature was not available on the experience of people with disabilities with interventions other than CPAP.

Older Veterans and Patient Age

One study231 specifically focused on the challenges older (≥ 60 years) veterans have with CPAP. Similar to the study on patients with disabilities, older veterans had difficulties with putting on the mask, adjusting straps, turning dials, pushing buttons, disconnecting tubing, and removing the water chamber.231 The participants in two other studies had an average age of ≥ 60 years.209,215 These studies were regarding physical activity,215 and both CPAP and OAs.209 It is uncertain how the age of the patient may have influenced their experience with these interventions.

Additionally, participants in the surgical studies89,98,232,233 tended to be younger (i.e., with average ages between 38.6233 and 45.9232 years) than participants in the other studies. It's uncertain how this may affect their experience with surgery. There were no other clear trends regarding the age of participants.

Relationship Status

A patient's life stage, specifically related to relationship status, in some cases will influence their treatment needs and choices. Relationship status was a complex factor that affected many aspects of OSA treatment; thus, it is presented here as an issue when considering treatment, but will also be discussed in the Thematic Analysis section, as it relates to the adoption of OSA interventions. More than half of participants were married or had a bed-sharing partner in 13 studies.197,209,211-216,219,222,224,225,230 In the remaining studies, marital status was not reported.89,98,112,198,199,207,210,217,218,221,223,226-228,231-233

It should be acknowledged that in some cases, having a partner is a catalyst for getting a person tested for OSA. Partners often had a role in diagnosis or recognizing that their partner had a problem and needed treatment.197,221 It is also important to acknowledge the emotional strain that OSA can put on a relationship. Partners and patients often expressed feelings of tension, frustration, anger, and emotional pressure that resulted from both having poor-quality sleep.197,211,216,222 Persons living alone described difficulty in initiating new relationships because of their OSA.211

Patients did consider their bed partners and their relationship before being treated for OSA, and this was related primarily to being aware of the disturbance that snoring caused,197,209,216 but also when it came to choosing a device.209 Before treatment, some patients and partners slept in separate rooms, in an attempt to mitigate the partner's lack of sleep due to the patient's apneas.197 However, some couples felt that there was a stigma with respect to sleeping in separate rooms, because of a partner with OSA, and that bed-sharing was a way to maintain normalcy.197 Participants with untreated OSA described decreased desire for sexual intimacy and decreased sexual activity.197,211,222 Partners also described feeling less like part of a couple, and more like a caregiver for their untreated spouse.197,222

When considering which intervention to use, people may consider how the intervention might affect their sexual relationships.209,217 This is seen in some participants preferring OAs over CPAP, due to their discreetness; as one participant said, "And so, like, I don't know anybody who is in a new relationship who would stick one of those things [the CPAP] to their face, you know. It's just not too appealing. But the oral appliance is not a big deal…Don't even know it's there."209

Impact on Lifestyle

The following section explores some of the considerations a patient may have before undergoing treatment or during treatment initiation.

Physical Environment and Travel

Activities, such as frequent travel, may hinder the use of some interventions. The need for electricity when using CPAP was mentioned by several studies.209,216,219,220,231 For some, this meant that they could not take CPAP camping or hiking,209 and for others it was a problem with respect to power outages,219 or when travelling internationally.220 Participants mentioned the need for proper electrical outlets219 or adaptors.220 For others, CPAP use was difficult because they were unstably housed.231

Travelling with CPAP, specifically on planes, was generally regarded as difficult.209,216 Users preferred CPAP machines that were compact and could travel easily.209 One user stated they would like to have a CPAP that can be used on a plane.209 Another CPAP user commented that when staying in hotels he did not need his humidifier, and suggested tips for travelling with CPAP, including using hard containers to store it, bringing along electrical tape to fix leaks, and always taking CPAP as carry-on.220 One partner recalled the patient's worries about travelling with CPAP, stating, "He had to make a trip out of town and he said, 'I'm not carrying that on the plane and being worried about whether I leave it somewhere. I'm not doing that.'"216

In contrast to CPAP, fewer worries about travel were expressed regarding OAs. Patients using OAs mentioned the need for a hot water supply.209 Some participants suggested alternatives to hot water, such as warming the OA in their armpit.209 The relatively smaller size of OAs compared with CPAP, and the fact that appliances do not require electricity, may appeal to some people with OSA.

There were limited findings for this category regarding lifestyle modifications. The physical environment - more specifically, the weather and place - hindered some patients from engaging in physical activity.215

Challenges to Lifestyle Modifications

Interventions for OSA were seen as a change in lifestyle, especially as it related to prescribed diet and exercise for weight loss. However, several studies reported challenges to these lifestyle changes. For exercise, some patients reported psychological distress as a barrier, as well as making excuses not to exercise.215 Previous negative experiences gaining weight after successful weight loss, or difficulty in currently losing weight, were barriers for patients.215,218

Regarding food and diet changes, patients reported complex relationships with food that prevented them from changing their lifestyle.218 Some used food to cope with feelings of anxiety, as a personal reward, as a substitute for tobacco, or to obtain oral satisfaction.218 As one patient stated, "I feel like an alcoholic… I use the food to lower anxiety, you know."218 Others simply found regulating their food to be challenging, and expressed a wish for a simple way of calculating their energy balance: "It would be easiest if you, in a very simple way, could measure energy in and energy out."218 However, some found it easier to implement diet changes when healthy food was readily accessible, easy to prepare, and enjoyed by the whole family.218

Cost

Cost was a factor in selecting a treatment. Patients perceived both OA and CPAP as expensive.209 Some participants indicated that their choice would be influenced by which treatment was covered by their extended health insurance. Others said that they were willing to pay whatever it costs for the treatment, provided that the treatment is highly effective: "We don't have extended care. So - but as far as health, we would be certainly willing to pay for anything that would help."209 In a choice between an OA or a CPAP, the importance of effectiveness was shown in a participant's comment that he would be willing to pay the cost of the CPAP because he perceived it to be more effective than the OA.209 The cost factor emerged as a strong additional deterrent in those who had initial difficulties with CPAP issues.217

If patients were informed about the features of various CPAP devices - for example, those who had sought information on the Internet or by talking to other CPAP users - then they were aware of types of machines available, as well as different features of each.199 Some were willing to pay out of pocket when the desired CPAP equipment cost more than their insurance would cover.199

For some patient groups, even the partial costs of CPAP would be a barrier when considering starting treatment and long-term use. In one study, a Māori focus group, which included only government-funded patients, discussed financial barriers to long-term treatment in the context of having to replace a device if broken or stolen. Māori patients also reported difficulties arranging transport to appointments.210 Interviews with Taiwanese patients with partial government funding also revealed the cost of CPAP to be a burden.221 In this study, costs were reported as between US$1,050 and $2,800. Several patients described the price as expensive and unaffordable. "The CPAP machine was too expensive. The government reimbursed me US$691. I hoped that the government would grant more money, because the price burdened me."221

Cost was also a factor for patients considering or attempting lifestyle changes. In one study, healthy food alternatives were considered too expensive. Some healthy food "...may contain odd ingredients as well but not too much ... it will be too expensive ... I cannot afford to pay just any price."218

No information was identified on the cost related to surgery.

Analytic Theme: Interventions for OSA require adaptation to daily routines and relationships. Some patients are able to integrate these interventions into their life and experience benefits, while others are unable to do so.

This theme captures results that describe how people learn to adapt to their OSA interventions, or not, and some reasons that compel people to adapt, or cause them to be unable to do so. For example, physical comfort was an important factor to patients, regardless of which intervention was used. Some persons may not find a device that works for them, for whatever reason, and thus choose to have surgery. Additionally, this theme explores experienced benefits and side effects from using OSA interventions. Patients express a trade-off between side effects or physical comfort and experienced benefits. For some, the benefits of using OSA interventions outweighed the discomfort; however, for others, this was not the case.

Other findings were related to how patients viewed themselves in light of OSA and treatment. Part of this was a descriptive theme we called "perceived attractiveness" with CPAP use and after surgery. The majority of patients found their appearance acceptable or satisfactory after surgery, and in some cases, patients found their post-operative appearance to be an improvement. A small number found their post-surgery appearance to be worse, and regretted having the surgery. Related in part to a sense of self, there were a few relevant psychological outcomes related to treatment. For some, these outcomes were positive, like reduced anxiety by using CPAP. However, for others, there were feelings of guilt and shame related to using OSA interventions. The following section explores the findings related to the second analytic theme.

Experienced Benefits

Physical, mental, and social benefits with treatment were frequently mentioned by the included studies. These experienced benefits supported some users to continue treatment and motivated them to overcome problems, adjust to the challenges of treatment, and generally persevere even when treatment was perceived as inconvenient, uncomfortable, or embarrassing.219

Patients described the positive outcomes of using CPAP therapy, and some patients recommended focusing on the positive outcomes of therapy to make the CPAP mask easier to accept.212 For some, the outcomes from using CPAP are profound; as one man stated, "The first day I got that machine, and woke up that morning, it was the clearest day of my life. I'd never seen the world like that."197 Another stated, "I love putting it on every night. I feel night and day differently. My quality of sleep has improved, and my mood and energy during the days. Now in the mornings, it's so much easier to get up."216 These patients - those experiencing substantial benefits - develop a feeling of confidence and dependency on CPAP.221 With increased alertness, patients were able to return to activities that they had not been able to do while untreated.219

However, patients were often said to develop a "love-hate" relationships with their machines.197 Some people reported persistent sleepiness in spite of what they described as good adherence to the prescribed therapy. They reported that they had tried to use the equipment provided and followed directions yet, despite their best efforts, did not feel better after initiation of PAP therapy.199 People who adapted to CPAP experienced the benefits of returned good health, such as the following:

Physical benefits

  • Improved energy levels207,210,216,219,221
  • Improved breathing221
  • Reduced tiredness207,216
  • Reduced other symptoms (e.g., nocturia219 and snoring221)
  • Better sleep197,207,209,213,216,219,221,230

Mental and psychological benefits

  • Better quality of life216,219
  • Sense of well-being219
  • Waking up and feeling refreshed, alert and vibrant197,209,210,213,216,219,221
  • Daytime alertness219,245
  • Revitalized memory221
  • Improved mood209,216,219,221,225,228,232
  • Confidence while driving217,221
  • Better ability to manage stress219
  • Feeling more relaxed219

Social

  • Less inconvenience to family and others221,230
  • Social contacts were made easier: "like being welcome to stay overnight"219
  • Decreased inconvenience to others (i.e., less snoring221)
  • Improved relations with spouses:209,216,219,221,225,230 "my wife is willing to sleep with me"221
  • Bed partner's sleep improved209,216
  • improved work performance221

People using OAs also experienced benefits from the device. In one survey study, almost half of patients (49%) experienced great benefit from using the device, while a smaller proportion experienced moderate benefit (37%) and fewer still experienced no benefit at all (14%).226 Some of the benefits users of OAs experience included:

Physical benefits

  • Improved energy levels223,225
  • Reduced tiredness225,228
  • Better sleep228
  • Improved physical strength225
  • Fewer morning headaches228
  • Improved snoring223,228
  • Improved breathing228
  • Sexual health228

Mental and psychological benefits

  • Sense of well-being and joyfulness225
  • Concentration223,225
  • Increased mental energy225
  • Less irritability228
  • Improved mood225,228
  • Satisfaction with improved health228

Social benefits

  • reduced effort with regards to social interactions225
  • Less inconvenience to family and others228
  • Improved relations with spouses225

People who underwent MMA surgery reported being satisfied after the surgery.98,232,233 One study232 reported 95.5% (n = 21) of patients were satisfied with how MMA managed their OSA symptoms (including quality of sleep; daytime and sexual functioning; physical, mental, and emotional health). For another study,98 100% (n = 15) of participants were satisfied or very satisfied with surgery (all experienced an improved quality of life). Of the nine patients completing both a pre- and post-surgery questionnaire for one study,233 all reported surgery as a worthwhile experience, although eight reported a favourable change after surgery; no further explanation was provided for this difference, although one patient from this study continued to use CPAP after surgery. Surgical patients experienced less daytime sleepiness and less sleepiness while driving.233 After surgery, some patients were able to engage in activities they had previously avoided because of their OSA.232 Patients also reported feeling less annoyed and less likely to become angry after having undergone surgery.232 In general, patients considered the surgery a worthwhile experience, and most would recommend the surgery to others.233 Some of the benefits highlighted by the surgery studies included:

Physical benefits

  • Reduced tiredness232
  • Better sleep232
  • Reduced snoring233
  • Improved physical activity232
  • Improved sexual health232

Mental and psychological benefits

  • Daytime alertness232,233
  • Revitalized memory232
  • Improved concentration98,232
  • Improved decision-making232
  • Improved mood232
  • Better ability to manage stress232
  • Better quality of life98
  • Reduced stress and worry232

Social

  • Improved work performance232
  • Improved interpersonal relationships232
  • Less activity avoidance232

There was some overlap in the benefits experienced by users of CPAP and OAs and for surgical patients. Users of CPAP and OAs, as well as those who underwent surgery, experienced reduced snoring, reduced tiredness, better sleep, and improved mental health. Improved relationships were also reported by CPAP, OA, and surgery studies. CPAP studies and surgery studies both reported improved work performance. An increased confidence while driving was reported by CPAP users.

Side Effects

For this review, the findings regarding side effects do not represent an exhaustive list of side effects experienced by patients on any intervention; rather, this review focused on what patient or partners found important about side effects with regard to their perspectives on treatment. Side effects of different interventions were considered when patients were choosing between interventions.

Several side effects were seen with using OAs. In one study, excess salivation was the most commonly reported side effect, followed by intra-oral soreness, jaw discomfort, and difficulty sleeping due to the device, and difficulty breathing was the least common side effect.223 For a small subset of these patients experiencing each side effect, the side effects were severe enough to prevent them from using their device.223 Pain associated with OAs was prevalent when first using the device, but not reported when users had grown accustomed to it.209

Participants who underwent surgery did not report a change in speech quality, and some reported that their teeth alignment was improved, although there was no improvement in chewing ability.232 Post-operative pain was reported as tolerable, and was not accompanied by increased headache frequency.232

Regarding physical activity, some participants experienced pain or heart sensations while being active, which was a deterrent for being physically active.215 Some participants chose activities that were less painful as a means of mitigating this.215

The findings of the SR207 found that studies regarding CPAP were often negatively framed and that patients often reported side effects from a list of pre-determined outcomes prepared by the investigators, rather than allowing patients to discuss their own experiences. Patients who regarded CPAP negatively were primed to have more severe and more frequent side effects, and that these side effects were more likely to hinder their CPAP use.207 Side effects were reported as a reason for discontinuing CPAP use, although it was noted that patients made trade-offs between the discomfort and side effects experienced with CPAP and the benefits they received from its use, and the exhaustion resulting from untreated OSA.207 Patients who perceived the side effects and discomfort of CPAP to be greater reduced their use of CPAP.207 Side effects were thought of as something a patient tolerates or struggles with, in order to receive the benefits of using CPAP.207

Claustrophobia When Using CPAP

Several studies reported patients feeling claustrophobia when using CPAP209,217,229,231 and that it could be a barrier to CPAP use. Patients reported feeling panicked209 or hungry for air219,229 while using CPAP. One study reported that veterans with post-traumatic stress disorder (PTSD) were more likely to report claustrophobia than veterans without PTSD.229

Side effects experienced by participants varied by intervention. For CPAP, OA, and physical activity, some side effects were a deterrent to being compliant with the intervention.

Comfort of Intervention

Studies on CPAP, OAs, and the tennis ball technique were identified, although it should be noted that the theme Comfort of Intervention encompassed codes on difficulties with the interventions. These themes were not evident for lifestyle modifications or for surgical patients.

Many patients experience discomfort while using interventions for OSA, and comfort with the intervention can change over time. For example, patients using OAs stated that the longer they used the device, the more comfortable they felt with it.209,228 For people using an OA, minimal pain was reported when the device was first used or when it was used infrequently. "If you're not wearing it all the time, certainly the first night that you put it in, your teeth are sore in the morning. And I find I don't sleep as well, I'm a bit restless that first night, just because…it's a bit uncomfortable if you have to adjust to it being in place."209 This was supported by findings from another study, which found that 25% of patients felt comfortable using an OA within a week, with 69% of patients were comfortable within a month.223 However, for some patients, discomfort with the appliances may stop them from using them. This is evidenced by the previously referenced study, in that the remaining 31% of patients stopped using an OA because they felt unable to wear it; follow-up for this study was three months after being provided with the device.223 Patients using the tennis ball technique also stopped using it because it was too uncomfortable - specifically, backache, shoulder ache, skin irritation, and the feeling that the tennis ball would create back problems.112

CPAP was also described as being uncomfortable to use.207,209,216,219,221 Using a CPAP device limited patients' ability to sleep on their side or stomach and limited their nighttime mobility.209,213,221 Patients also complained of air blowing in their faces,209,221 air leaking out of their machine,219 and waking up with a stiff neck.209

Difficulty With CPAP

Related to the physical discomfort of CPAP, another issue affecting patient experience could be characterized as difficulty with the device. Using CPAP is often seen as a hassle, frustration, and inconvenience, and this can frame a patient's experience with the machine. Because of these difficulties with CPAP, patients expressed that it was difficult to persist and stay motivated to use CPAP.239

Some participants experienced frustration and inconvenience related to the functioning of the equipment. Specifically, patients report that the CPAP is noisy, either because of the machine itself or air escaping or moisture in the tube. Patients are also concerned that the noise is disturbing to bed partners,197,209 and some partners did comment on the noise of the machine.213 People mentioned that it is frustrating to have to adjust mask straps, especially if people wake up because of ill-fitting straps and masks.207,216,221

Technical problems with the CPAP device were perceived as hard to manage, and if any part of the CPAP breaks, it can be inconvenient or costly to repair. This situation is expressed as frustration because it is expensive to fix or replace parts and it can be difficult to get the right size.207 Part of the frustration relates to the importance of being able to update equipment, to opt for a better-fitting mask, as equipment wears out or as technology changes.210,239 Inconvenience and frustration were also mentioned in the context of getting insurance to pay for the best-fitting mask.197 Dissatisfaction was noted among patients who could not get their insurance to pay for a machine with a better-fitting mask or straps.197 One study198 had CPAP-experienced users trial APAP devices, with 14 patients (52%) preferring APAP at the end of the study; the authors speculated that the patients may have been biased to CPAP because in studies with CPAP-naive patients, patients find APAP easier to use.

Maintenance and cleaning of CPAP can also be challenging.197,207,209,221,239 As one participant commented, "There is a lot of small, fussy stuff associated with the CPAP machine. Filling the reservoir and washing it, which I don't do every night, but fairly often. It's a chore like flossing; I mean, you know, various other things."209 Several studies mentioned that buying filtered water was also seen as annoying.197,209 People expressed a desire for support or explanations of procedures for cleaning and maintaining of CPAP equipment.221 They complained about the "unavailability" of medical professionals to provide explanations, purposes, and precautions when using CPAP.221

Despite these difficulties, which some describe as a torment or an imposition,207 some people could see using CPAP as a trade-off: "it works, but it's difficult."239 Some described a process to get over the difficulties,221 although some expressed a fear of "flunking" therapy.207 Once these patients feel the benefits of their treatment, they people report feeling dependent on their CPAP machine.207,221 They felt the need to use CPAP and believed it was worth the trouble and difficulties in order to avoid the return of symptoms.219 For these people, there was satisfaction enough to use the CPAP nearly every night, and some accepted it from the onset.219

Difficulty With Oral Appliance

As with CPAP, users of OA expressed difficulties with the device. OA users were disappointed in the durability of the device and the frequency with which the device needed to be repaired or replaced.209 The expense and inconvenience of doing repairs or getting replacement devices was also difficult, as one patient explained: "Well, they break down, you know; after a while, they break down in the, you know - either the wire on the top breaks, or it becomes undone on the back or the plastic just deteriorates and, you know, chunks off."209

Patients described some minor inconvenience associated with cleaning the OA, which was perceived as a nuisance. Patients were also uncertain about the best way to clean the device and which cleaning products were best.209

Difficulty With Tennis Ball Technique

One study reported specific difficulties with the tennis ball technique. Patients felt that the tennis ball moved around too much, that it caused backache, and that it was ineffective on a soft mattress, and some patients felt the treatment was not relevant to them because they did not sleep on their back.112

Perceived Attractiveness

Perceived Attractiveness refers to how a patient feels about their appearance while using CPAP, OAs, or after surgery.

Regarding a patient's view of their aesthetics after surgery, most patients seem to consider their appearance acceptable or satisfactory.89,98,232 In some cases, patients may even consider their post-operative appearance to be an improvement. One study found that 11 out of 15 patients found an improvement in their appearance post-surgery, related to the nose, upper lip, and lower lip;98 this is similar to another study, in which 14 patients (54%) also found an improvement in their appearance.89 Some patients may not notice a difference in appearance, or feel as if their preoperative appearance is just as attractive as their post-operative appearance.89,98

However, some patients may feel worse about their appearance post-surgery, which for some, may mean they regret having the surgery. In one study (n = 15), one patient reported a worse appearance, particularly a worse appearance of their nose and upper lip.98 In this study, five patients also reported a change in their teeth, with two patients reporting their dental situation to be worse and three reporting an improvement.98 One patient who had experienced an improvement in their teeth had orthodontic preparation; however, the rest of the patients who reported an improvement or worse appearance did not have orthodontic preparation.98 In another study, eight patients (31%) of patients felt their post-surgery appearance was worse.89 Of these eight patients, one regretted having the surgery.89

In contrast with the outcomes from surgery, where most participants felt as or more attractive post-surgery, many participants reported feeling unattractive with CPAP197,207,209,216,217 and reported feeling the need to keep their CPAP use a secret.209 Some even felt that their partners might feel differently toward them or view them differently because of their CPAP use.199,216 Patients who were dating, and considering progressing to a more intimate relationship, had concerns about telling their partners about their PAP machine.199 This was expressed as having to "break the news" to a new partner.199 Patients with longer-term relationships still felt less attractive to their partners, but describe more support from their partners.199

There was a certain stigma to using a medical device to aid in sleep,207,216 and using the device was something people described as identifying them as ill.207 Patients also commented on a swollen face and marks on their face from using CPAP, which caused embarrassment and revealed to others that they used CPAP.219 Patients were unable to lessen these marks without compromising the effectiveness of CPAP.219 One woman felt so strongly about her appearance with CPAP that she commented,

I don't like it - it's got these straps that go here and here, kind of like "Hannibal Lecter." It looks really funny. I don't know if it's a personality issue or what, but women aren't supposed to be like that; we're supposed to be dainty when we sleep. So I'll go to sleep at night without it, but then wake up and put it on… Let's face it, that thing is ugly, and putting that big, goofy thing on in front of my husband… those first few nights I would make sure he was asleep first. Then, in the morning, I'll take it off when his alarm goes off. I've even asked him, "what's better, me snoring or you waking up and seeing your wife with this horrible contraption wrapped around her face?"

Partners also recognized that patients were sensitive to their appearance while wearing CPAP.207 One partner commented, "My husband was very uncomfortable wearing that mask when he went to bed. He kind of hid because it was, you know, not attractive. It's an apparatus and it was just a very uncomfortable feeling for him."216

In the one study that included people who had experience with both CPAP and OA, some people reported preferring OAs as a more discreet and more attractive option, which lessened the feeling of embarrassment and ugliness around bed partners.209

Sexual and Intimate Relationships

Some participants found that CPAP enabled them to resume bed-sharing with their partners and improved relationships, whereas for others, the negative impact of CPAP on their intimate and sexual relationships, due to shame or interference of the machine, was a barrier to use.216 Some users of CPAP describe the machine as a hindrance to physical intimacy,197,219 and as being "unsexy."197 As one participant described, "…It's had an impact on our relationship; you've got a frickin' snorkel thing across your marriage bed, and you create a windstorm from the exhaust. That's why it's not so fun. It makes cuddling difficult. If you're snuggling someone, then you're blowing wind on the back of their head, like you're in a hurricane."197 Another commented, "…I mean, how do you say 'good night' with a big mask over your face and things like that, it's… so, it's a little sad…"219 However, although partners may recognize that the patient finds themselves less attractive while wearing CPAP, not all were deterred by this. As one partner stated, "I think somehow in his mind he thinks that this machine interferes to some extent with intimacy. And it's like, you know what, it's a convenient mask to take off."216

Regarding relationships after surgery, one study found that the majority of patients reported an improvement in their relationship with their significant other.232 In this same study, about a third of participants (37%) reported an improvement in their desire for sexual intimacy.232 Another study found that one-fifth of patients slept in separate bedrooms from their partners even after surgery.225

Both people who used CPAP and those who underwent surgery saw improvements in their relationships with bed partners, although users of CPAP also found that the device hindered their intimate relationships.

Partner Experience

The spouses perceived it as important to be a support for their partner and adjusted their lives accordingly.222 Some patients also perceived this as an opportunity to become closer to each other and do activities together, despite the fact that these were steered by the partner's symptoms. "I have made a real effort to get him out for a walk. I enjoy fresh air and getting some exercise; we just can walk round the block with the dog or something."222 The spouses felt responsible for helping their partner to find a solution to their nightly sleep problems. For example, they knew from their own experience that if their partner lay on her/his side, then the problems lessened.222

Often, partners experienced personal and social distress prior to CPAP initiation, and were motivated to help the patient adapt.207 For example, married users problem-solved and incorporated CPAP into their lives quicker than unmarried users.207 Conversely, negative or absent partner support led to poorer experiences with treatment,207 with male spouses reported in one study to be less involved with their partners' CPAP treatment than female spouses.207 For some partners, the positive attitude of the patient toward CPAP was a reason to be less in