Impact of Exercise Training on Obstructive Sleep Apnea: A Systematic Review and Meta-analysis of Randomized Trials
Randomized controlled trials have highlighted the importance of exercise training on the severity of obstructive sleep apnea (OSA). The primary objective of this meta-analysis was to analyze the data from these studies on the severity of OSA indices as measured by formal polysomnography. The secondary objective was to evaluate the effect of exercise on oxygen desaturation index, body mass index and several metabolic parameters associated with OSA. Two independent reviewers and an experienced librarian searched MEDLINE, EMBASE, COCHRANE, and Scopus and SPORTDiscus databases from January 1, 2000, to October 21, 2021 to identify randomized controlled trials that evaluated the impact of exercise training on OSA severity. Pre-exercise and postexercise training data on the primary and secondary objectives were extracted. A total of 11 randomized controlled trials (533 study participants) were identified based on the inclusion criteria. The mean pre-exercise and postexercise reduction in apnea hypopnea index was −6.94 event per hour (95% confidence interval [CI]: −10.74 to −3.13, I2 = 84%). The mean changes in oxygen desaturation index and body mass index were −5.32 events per hour (95% CI: −9.65 to −0.99, I2 = 73%) and −1.21 kg·m−2 (95% CI: −1.80 to −0.62, I2 = 49%) respectively. In adult patients with OSA, exercise training was associated with statistically significant reduction in OSA severity.ABSTRACT
Background
Methods
Results
Conclusion
INTRODUCTION
Obstructive sleep apnea (OSA) is a concerning diagnosis defined as the presence of at least 5 episodes of upper airway closure with each hour of sleep, and each episode halting airflow for at least 10 seconds (1). These episodes are known as apnea and can result in a decrease in oxygen saturation, hypercapnia, elevated oxidative stress, and increased sympathetic activity, which have been implicated as part of the pathophysiologic processes, and have deleterious short-term and long-term consequences (1–3). These include but are not limited to diurnal hypertension, arrhythmia, stroke, pulmonary hypertension, heart failure, chronic sleep deprivation, chronic fatigue, decreased executive function, personality changes, and an elevated risk for sudden cardiac death (1,2). The number of apnea episodes measured per hour during a sleep evaluation is known as the apnea hypopnea index (AHI).
Several lifestyle-related risk factors are known to pre-dispose to the development of OSA (2,4–6). These are thought to directly or indirectly result in complete or partial collapse of the pharynx causing enlargement of the surrounding soft tissues through hypertrophy, inflammation, and/or edema, which are mostly pronounced in the lateral pharyngeal wall resulting in a reduction of the genioglossus muscle activity (7,8).
The diagnosis and severity measures of OSA are based on laboratory polysomnography (PSG) with several indices measured that estimate the severity of the disease burden. Mild, moderate, and severe OSA are diagnosed as AHI ≥5, ≥15 and ≥30, respectively (4,9). According to prior studies, treatment options including continuous positive airway pressure (10), bilevel-positive airway pressure (11), oral appliances including mandibular advancement devices (11–13), surgical options (including bariatrics, maxillomandibular adjustments, and hypoglossal nerve stimulation) (11,13,14), weight loss via dietary modification (15), and pharmacotherapy (16) have reduced the severity of OSA. However, their impact on AHI reduction appears to be temporary and variable. This is based on the outcomes of studies (15,17–21) that have evaluated their long-term effectiveness.
The role of exercise training in reducing AHI and improving other PSG indices has been growing in the recent literature (9,22,23). The mechanisms behind the beneficial impact of exercise on OSA indices are yet to be elucidated (24). Previous systematic reviews that evaluated the impact of exercise on OSA indices have either included studies other than RCTs (9,24–26), with exercise duration as short as 3 weeks (27), or excluded OSA participants with comorbidities such as heart failure and stroke (9,25). The primary aim of this meta-analysis is to investigate the impact of exercise training on AHI as measured by formal PSG in randomized controlled studies. The secondary aim is to evaluate the effect of exercise on oxygen desaturation index (ODI),body mass index (BMI), and risk factors associated with OSA.
METHODS
Data Sources and Searches
Two authors (AO and FM) and a qualified librarian independently searched MEDLINE, EMBASE, COCHRANE, and Scopus and SPORTDiscus databases from January 1, 2000, to October 21, 2021. The keywords and Medical Subject Heading search terms used were exercise, therapy, physical, rehabilitation, sleep, apnea, obstructive, sleep disorders, and sleep breathing disorders. In addition, different combinations of these keywords were applied in the database search (Supplemental Figure S1). To ensure thoroughness, we manually searched the reference lists of the articles and previously published systematic reviews and meta-analyses on exercise and OSA to identify references not included in the automated search. The search was restricted to articles published in English language. We adhered to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guideline (28) (table 4; Supplemental Figure S1). The PRISMA flow chart and the reasons for study exclusion are provided in Figure 1.
Citation: Journal of Clinical Exercise Physiology 11, 3; 10.31189/2165-6193-11.3.80
Selection of Studies
Eligible studies were randomized controlled trials (RCTs) that included OSA patients that underwent a structured supervised or unsupervised exercise training as the treatment group and a control group that did not participate in the exercise training but was encouraged to maintain their baseline physical activity level. RCTs were selected if they reported preintervention and postintervention changes in AHI, ODI, and BMI, included a minimum of 20 participants with OSA diagnosed by PSG with AHI of at least 5, and age > 18 years. The total exercise duration must be a minimum of 3 months, at least 3 exercise sessions a week, each session lasting ≥30 minutes, and comprised of aerobic exercises and/or resistance training. Studies not meeting these criteria, or used dental sleep devices, oropharyngeal muscle training, or surgery were excluded. Disagreements on article inclusion were discussed among authors and resolved by consensus.
Outcomes and Data Extraction
The primary outcome was to compare postintervention changes in AHI between the exercise group and the control group. The secondary outcome was to compare the postintervention changes in ODI and BMI between both groups. A prespecified subgroup analysis was performed for diabetes mellitus (DM), hypertension, waist circumference, age, gender and smoking. Data extracted from the selected RCTs included author, publication year, mean age of the participants, number of subjects, study duration, frequency and duration of exercise, type of exercise, and preintervention and postintervention PSG indices (AHI, ODI), BMI, DM, hypertension, gender, age, smoking, waist circumference. The accuracy of the extracted data was assessed and confirmed by authors SU and MM.
Quality Assessment
The methodologic quality of the RCTs were assessed using the tool for assessment of study quality and reporting in exercise (TESTEX) scale. It comprises 12 criteria with a maximum of 15 points: 5 points for study design and 10 points for reporting (29) (Table 1). Studies with a score of 12 to 15 points are of high quality, those with 7 to 11 points are good quality, and those with ≤6 points are low quality (29).
Data Analysis
Data analysis was performed by NY using the RevMan (version 5.3; Copenhagen: The Nordic Cochrane Center, The Cochrane Collaboration, Copenhagen, Denmark). The forest plots of the above outcomes were visually represented after pooling the mean differences using the random-effects model. Heterogeneity was assessed with the I2 test. Publication bias was assessed with the funnel plot and standard error effect sizes. We used the literature recommended threshold of statistical significance which is P < 0.05 for outcomes and P < 0.10 for heterogeneity.
RESULTS
Results of Search
We retrieved a total of 3,536 articles from the initial search (663 from MEDLINE, 1,175 from EMBASE, 1,425 from CENTRAL, 273 from Scopus and SPORTDiscus). Of these, 83 were RCTs. After excluding 13 duplicates, we analyzed the title and method section of the remaining 70 articles, and 59 of these articles were excluded for failing to meet the priori inclusion criteria. Therefore, the final analysis comprised of 11 articles with 533 participants (30–40) (Table 2).
Characteristics of Included Studies
All studies were randomized, included preintervention and postintervention AHI measures with formal PSG, and had exercise of >3 months. Five studies assessed changes in ODI (31–33,35,36), and 7 (22,30,31,33,35,37,38) reported changes in BMI. Patients had a mean age of 51 years, 69% were males. The exercise group had moderate to severe OSA with a mean AHI of 32 ± 14 events per hour with an average BMI of 29 ± 3 kg·m−2. The mean ODI is 28 ± 15 events per hour. The control group also had moderate to severe OSA with a mean AHI of 31 ± 12 events per hour, the mean ODI was 12 events per hour with a range from 0 to 15 events per hour and an average BMI of 24.7 ± 12.5 kg·m−2. The exercise programs were supervised in 9 of the studies (22,30,31,33–36,38,39). The mean exercise duration was 5 months, with an average of 3 sessions per week, each lasting an average of 60 minutes. Aerobic exercises included running/walking on the treadmill and elliptical/cycle ergometer (30,31,33–37). Resistance training (22,33,39,40) included shoulder press, latissimus dorsi pulldown, chest press, bicep/triceps curls, leg extension/curl, isometric handgrips, and isokinetic dynamometry. Additionally, the studies reported improvement in daytime excessive sleepiness either via the Epworth sleepiness scale or Pittsburgh sleep quality index (30,32,34,36,38). The funnel plot analysis was insignificant for publication bias (Supplemental Figure S1).
Outcome Measures
In the exercise group, the mean change in AHI compared to baseline is −8.7 ± 5.5 events per hour. The mean change in AHI compared to baseline in the control group is 0.5 ± 6.4 events per hour. The mean change in ODI in the exercise group is −5.6 ± 6.8 events per hour, and 0.6 ± 8.4 events per hour in the control group. The mean change in BMI in the exercise group is −1.3 ± 1.3 kg·m−2, and −0.3 ± 0.6 kg·m−2 in the control group. Exercise treatment was associated with a statistically significant reduction in the AHI (mean difference compared to control of −8.2 events per hour, 95% CI: −11.9 to −4.6, I2 = 84%). The ODI was significantly lower in the exercise group (mean difference compared to control of −5.3 events per hour, 95% CI: −9.7 to −1.0, I2 = 73%). In addition, the exercise also led to a decrease in BMI that was statistically significant (mean difference compared to control of −1.2 kg·m−2, 95% CI: −1.8 to −0.6, I2 = 49%) (Figure 2A–C). A subgroup analysis based on the risk factors associated with OSA (including age, smoking, alcohol use, DM, and hypertension) demonstrated no significant reduction in AHI with exercise training (Figure 3A–G).
Citation: Journal of Clinical Exercise Physiology 11, 3; 10.31189/2165-6193-11.3.80
Citation: Journal of Clinical Exercise Physiology 11, 3; 10.31189/2165-6193-11.3.80
Citation: Journal of Clinical Exercise Physiology 11, 3; 10.31189/2165-6193-11.3.80
Citation: Journal of Clinical Exercise Physiology 11, 3; 10.31189/2165-6193-11.3.80
DISCUSSION
In this analysis, we evaluated the impact of exercise training on PSG-derived OSA indices and BMI by comparing the mean pre-exercise and postexercise parameters in a cohort of 533 participants from 11 studies. The outcomes showed a significant reduction in AHI, ODI, and BMI (by 22%, 19%, and 4%, respectively) in OSA patients that underwent exercise training.
In contrast to previous meta-analyses (9,24,26,27), our meta-analysis included only RCTs, which provide the best level of evidence that support an intervention. Importantly, this study had a higher effect size (Cohen d: 3.1 vs. 1.6, 2.8, 1.2, and 2.0, respectively). Additionally, with a longer average duration of exercise, the statistical power was enough to highlight a significant reduction in BMI. Furthermore, via the subgroup analysis, this study demonstrates that the improvement of OSA with exercise therapy extends across risk factors associated with OSA such as age, DM, and BMI. This paper represents the most contemporary literature to support the important benefits of exercise as an important, validated, inexpensive, and portable intervention in addition to continuous positive airway pressure for OSA patients. The magnitude of AHI and ODI reduction noted in this meta-analysis is lower compared to that of previous meta-analyses and observational studies that have evaluated other interventions besides exercise training. These include dietary weight loss (AHI: −14.3, 95% CI: −23.5, −5.1) (40), oral appliances (AHI: −12.1, 95% CI: −9.7, −14.3) (41), surgical weight loss (pre-AHI: 54.7, 95% CI: 49, 60.3; post-AHI: 15.8, 95% CI: 12.6, 19.0) (42), (ODI: 43.6 to 18.3) (43), and upper airway surgery (ODI: 12.3 ± 9.8 to 4.0 ± 4.2) (AHI: 21.4 ± 13.8 to 5.8 ± 7.2) (44,45). This difference in magnitude might be due to the combined effect of exercise training with these interventions. The subjects that underwent these interventions had severe OSA as similar to this analysis. However, these interventions (except for surgical weight loss) had an identical impact on OSA severity with exercise training (i.e., a reduction from severe/moderate to moderate/mild OSA). Hence, future RCTs are needed to evaluate and compare the impact of these interventions to exercise training. It is not well established if the reduction in AHI and ODI by exercise training is sustainable in the long term, especially if exercise ceases (46). It is known that some of the other interventions (21,47,48) do not have a long-lasting impact on OSA indices.
The reduction in BMI noted in this study is not clinically significant. It might be due to the inclusion of RCTs in which the intervention groups were subjected to both dieting and exercise training (38,39). In support of this, Araghi et al (26) noted a decrease in pooled mean BMI that was not significant with exercise training after adjusting for diet as a cointervention. Hence, our findings highlight that the improvement in AHI and ODI is independent of weight loss. This is because it implies that exercise as a single intervention may be sufficient to improve OSA severity and in fact may impact long-term compliance (40). In addition, improvement in AHI also led to clinically significant reduction in both systolic and diastolic blood pressure without a significant change in BMI (27).
The mechanism involved in the reduction of AHI and ODI by exercise training possibly includes the effect of exercise on reduction of pharyngeal adipose volume, or simply strengthening the pharyngeal muscles (49), nocturnal rostral fluid shift that reduces airway edema (51) and oxidative stress (51,52). The independence of its impact on BMI reduction might be explained by the selective deposition of adipose tissue in pharyngeal airway in OSA patients compared to BMI matched controls without OSA (53). In support of this, changes in pharyngeal adipose deposition have been noted on cervical magnetic resonance imaging and facial computed tomography have been correlated with the severity of OSA (54,55). Furthermore, the independence of BMI raises the question of whether neck adipose tissue has a higher metabolism than that of the abdomen and hence reduces at a faster rate to exercise than does abdominal tissue.
The ideal or minimal duration of exercise needed to significantly reduce the OSA indices in the long term remains unknown (46). Our results are similar to those of other studies that have evaluated different exercise duration and form (aerobic and/or resistant training) (9,24,27,38,55,56). Therefore, comparing the long-term impact of aerobic exercises and resistance training on OSA indices might be the aim of future studies. Additionally, studies are needed to evaluate the optimal time of the day for exercise training that would significantly improve OSA indices. Although this study sought to include both supervised and unsupervised exercise programs in its analysis, only 1 of the included studies (35) used unsupervised exercise. Hence, it underrepresents the impact of unsupervised exercise on OSA severity. In studies that included unsupervised exercise as an intervention, the reduction in OSA severity was not statistically significant (57). Therefore compared to supervised exercise, the literature currently lacks robust evidence supporting the efficacy of unsupervised exercise.
Limitations
This meta-analysis was limited by the inclusion and exclusion criteria described above. Furthermore, this study has a significant in-between trial heterogeneity observed in terms of AHI and ODI. While this might be related to the differences in exercise protocols, training atmosphere, and use of diet in some studies, it highlights the need for a patient-level meta-analysis. Lastly, this study did not address the long-term effects of exercise on OSA severity after completion of the exercise protocols. For Figure 4b: controlling for gender, DM2.
CONCLUSION
This study demonstrated a statistically significant reduction in AHI and ODI observed with exercise training. The improvement in these OSA indices is independent of weight loss. Future studies are needed to determine if the improvement in OSA indices observed with exercise training is sustainable over the long term.
PRISMA flowchart of the study selection.
Forest plots highlighting statistical significance improvement in apnea-hypopnea index (AHI), oxygen desaturation index (ODI), and body mass index (BMI).
Subgroup analysis based on the risk factors associated with obstructive sleep apnea.
Continued.
Continued.
Contributor Notes