Effects of aerobic exercise and resistance training on obstructive sleep apnea: a systematic review and meta-analysis

Abstract

Study Objectives:

We investigated the therapeutic effects of exercise in patients with obstructive sleep apnea, aiming to identify the subgroups that benefit the most and determine the optimal exercise protocol.

Methods:

Major databases were searched for randomized controlled trials involving patients with obstructive sleep apnea performing aerobic exercise and/or resistance training. The investigated outcomes included apnea-hypopnea index (AHI), Epworth Sleepiness Scale, body mass index, and peak oxygen consumption during exercise (VO₂peak). The pre- and postintervention unstandardized mean difference (USMD) of these parameters was compared between the exercise and control groups.

Results:

Twelve studies involving 526 patients were included. Exercise training significantly reduced AHI (USMD = −7.08 events/h, 95% confidence interval: −9.98 to −4.17, P < .00001), Epworth Sleepiness Scale (USMD = −2.37, 95% confidence interval: −3.21 to −1.54, P < .00001), and body mass index (USMD = −0.72 kg/m², 95% confidence interval: −1.22 to −0.22, P = .005) and enhanced VO₂peak (USMD = 3.46 ml·kg⁻¹·min⁻¹, 95% confidence interval: 1.20 to 5.71, P = .003). Subgroup analyses revealed that in continuous positive airway pressure–adherent patients, exercise significantly improved VO₂peak but did not reduce AHI and Epworth Sleepiness Scale. A trend was observed that combining resistance training with aerobic exercise resulted in greater AHI reduction and VO₂peak enhancement. Notably, exercise improved AHI, Epworth Sleepiness Scale, body mass index, and VO₂peak regardless of the baseline AHI or body mass index.

Conclusions:

Exercise, including resistance and aerobic training, should be part of treatment for patients with obstructive sleep apnea of all severities, regardless of obesity status, and even for those who are already under continuous positive airway pressure.

Citation:

Lin C-F, Ho N-H, Hsu W-L, Lin C-H, Wang Y-H, Wang Y-P. Effects of aerobic exercise and resistance training on obstructive sleep apnea: a systematic review and meta-analysis. J Clin Sleep Med. 2024;20(11):1839–1849.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Exercise is an essential component of obstructive sleep apnea (OSA) treatment. However, the optimal protocols are unexplored, and the responses in different patient subgroups have not been elucidated.

Study Impact: Our meta-analysis focused on the therapeutic effects of exercise on OSA, with subgroup analyses based on multiple clinically relevant factors. The results of this study may provide valuable information for sleep clinicians, especially when providing exercise prescription to patients of OSA.

INTRODUCTION

Obstructive sleep apnea (OSA) is a sleep disorder characterized by cessation or restriction of airflow because of repetitive upper airway collapse despite respiratory efforts during sleep, eventually causing desaturation and even sleep arousal.¹ OSA has a prevalence of 9–38% worldwide,² and this is gradually increasing over time.³ The most common symptoms of OSA include excessive sleepiness and fatigue, which lead to impaired quality of life, poor work performance, mood disorders, decline of cognitive functions, and increased risk of occupational and traffic accidents.^4–10 Aging, male sex, and high body mass index (BMI) are well-known risk factors of OSA.^1–3 Moreover, many metabolic and cardiovascular diseases are highly associated with OSA, including hypertension, type 2 diabetes mellitus, atrial fibrillation, heart failure, stroke, and death.^11–14

Polysomnography is commonly used to diagnose OSA and determine its severity based on the apnea-hypopnea index (AHI). Current treatments for OSA include exercise, body weight reduction, use of oral appliances, continuous positive airway pressure (CPAP), and various types of surgery.¹ Exercise is widely accepted as a crucial treatment for OSA because it significantly reduces AHI.^15–17 Exercise also improves sleep quality and mood symptoms in patients with OSA.^8,18

Aerobic exercise and resistance training are commonly included in OSA treatment. Other combinations, such as aerobic exercise and respiratory training, have also been proposed.¹⁹ However, the effect of different types of exercise on OSA is poorly understood. Furthermore, the optimal exercise duration and the subgroup of patients that benefit most from this treatment have not been explored. Herein, we systematically reviewed the current literature and conducted a meta-analysis to determine the effects of aerobic exercise and resistance training on patients with OSA.

METHODS

Study designs and search strategy

This study investigated the effects of aerobic exercise and/or resistance training on AHI reduction in adult patients with OSA using a pooled analysis of randomized controlled trials (RCTs). Meta-analysis was conducted in accordance with the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA).²⁰ The study protocol was registered with the PROSPERO database (no. CRD42023423527). We comprehensively searched the PubMed and Embase databases for eligible studies using the following search terms according to Boolean logic: (“Sports” OR “Exercise” OR “Aerobic Exercise” OR “Resistance Training”) AND (“Obstructive Sleep Apnea” OR “Sleep Apnea Syndrome”). The searches comprised free-text and MeSH terms. No restrictions were placed on the publication year or language.

Eligibility criteria

RCTs that (1) were published in English from inception to May 1, 2023; (2) included participants aged ≥ 18 years diagnosed with OSA (AHI ≥ 5 events/h) using polysomnography; (3) implemented exercise intervention involving aerobic exercise and/or resistance training (duration ≥ 30 minutes/time and frequency ≥ 2 times/wk, regardless of the strength and follow-up time) were included. However, studies were excluded if they involved patients who had previously received surgical intervention for OSA, if dietary restriction was included in the intervention, or if the exercise implemented had insufficient duration or frequency.

Selection process and data extraction

First, duplicate studies were removed. Two independent reviewers (C.-F.L. and N.-H.H.) screened the titles and abstracts of the studies and excluded irrelevant ones. Second, the reviewers read the full-text articles and extracted the following data from the eligible studies: study characteristics, participant characteristics, types and duration of exercise, and preintervention and postintervention outcome parameters (AHI, Epworth Sleepiness Scale [ESS], BMI, and peak oxygen consumption during exercise [VO₂peak]). In case of any discrepancies, a third reviewer (Y.-H.W.) was consulted.

Quality assessment, data synthesis, and statistical analysis

The risk of bias of enrolled studies was evaluated by a reviewer (W.-L.H.) using the Cochrane risk of bias tool.²¹ Review Manager 5.4 was used to conduct meta-analysis; sensitivity, heterogeneity, and subgroup analyses; and publication bias evaluation. Outcomes were expressed as mean ± standard deviation (SD). Pooled effect was calculated by comparing the pre- and postintervention unstandardized mean differences (USMD) of AHI, ESS, BMI, and VO₂peak between the exercise and control groups. When data were missing, we contacted the corresponding authors of the studies via e-mail to complete the dataset. Nevertheless, the SDs of the mean change from baseline (SD_change) of the outcome variables were not obtained in several studies. Thus, the following equation^21,22 was used to impute SD_change using the SDs of preintervention (SD_baseline) and postintervention (SD_final) outcomes:

$${SD}_{\mathit{\text{change}}}=\sqrt{{SD}_{\mathit{\text{baseline}}}^2+ {SD}_{\mathit{\text{final}}}^2-(2\times \mathit{\text{Corr}}\times {SD}_{\mathit{\text{baseline}}}\times {SD}_{\mathit{\text{final}}})},$$

where the correlation coefficient (Corr) was assigned a value of 0.7 to provide a conservative estimate.²²

The level of significance was 5%. Heterogeneity was measured using the inconsistency index (I²). When significant heterogeneity was observed (P < .10 or I² > 50%), a random-effects model was generated. Subgroup analyses were conducted to examine the factors influencing treatment efficacy. Sensitivity analysis was performed by leaving out 1 study at a time to detect potential heterogeneity. Publication bias was assessed using funnel plot analysis.

RESULTS

Study selection

Overall, 910 and 556 references were retrieved from PubMed and Embase databases, respectively, and another 4 were retrieved through manual search. Of the 1,470 articles identified, 1,061 remained after removing duplicates. The title and abstract screening yielded only 32 articles, which were subsequently selected for full-text evaluation. Finally, 12 studies were included in the meta-analysis. The PRISMA flow diagram is presented in Figure S1 in the supplemental material.

Study characteristics

Table 1 presents the characteristics of the included studies and exercise interventions. The 12 RCTs involved 526 patients. The intervention adopted in 5 of these studies consisted of aerobic exercise alone,^{17,19,23–26} whereas the other 6 studies included resistance training.^16,27–31 Two studies incorporated CPAP in some or all treatment groups.^23,29 One study randomly assigned participants to 1 of 4 groups: control, exercise, CPAP, or exercise + CPAP, which were then divided into 2 comparisons (control vs exercise; CPAP vs exercise + CPAP) in the meta-analysis.²⁹ The enrolled RCTs incorporated various intervention designs, including exercise types (walking, cycling, and treadmill), frequency (2–6 times/wk), session length (30–150 minutes), and intervention duration (4–36 weeks). Table 2 presents the pre- and postintervention outcome parameters of each study.

Table 1.

Study design and details of exercise intervention in the enrolled studies.

Study	Group 1 (n)	Group 2 (n)	Group 3 (n)	Control (n)	N	Aerobic Exercise				Resistance Training		D
						Types	Intensity	F	L	Contents	F
Ackel-D’Elia et al 2012, Brazil23	A + C (13)	—	—	C (19)	32	Treadmill	Up to AT	3	60	—	—	8
Kline et al 2011, USA16	A + R (27)	—	—	Stretch (16)	43	Treadmill	60% of HRR	4	150	10–12 reps; 8 moves, 2 sets	2	12
Servantes et al 2011, Brazil30	A + R (17)	—	—	Nil (11)	28	Walk	Up to AT	3	30–45	12–16 reps; 7 moves, 1 set	3–4	12
Desplan et al 2014, France27	A + R (11)	—	—	Education (11)	22	Cycling	Up to AT	6	45	NA	6	4
Servantes et al 2018, Brazil29	A + R (17)	C (15)	A + C+R (15)	Education (18)	65	Aerobics	Up to AT	3	30–45	12–16 reps; 7 moves, 1 set	3–4	12
Guerra et al 2018, Brazil28	A + R (21)	—	—	Nil (22)	43	Cycling	Up to AT	3	30–40	10–15 reps; 5–6 moves, 1 set	3	24
Berger et al 2018, France31	A + R (43)	—	—	Education (45)	88	Aerobics	Up to AT	3	40	NA	3	36
Sengul et al 2011, Turkey19	A (10)	—	—	Nil (10)	20	Cycling	60–70% of VO2max	3	45–60	—	—	12
Mendelson et al 2016, Canada17	A (17)	—	—	Nil (17)	34	Treadmill	60% of VO2max	5	30	—	—	4
Yang et al 2018, China25	A (32)	—	—	Nil (35)	67	Cycling	Up to AT	3	30	—	—	12
Jurado-García et al 2020, Spain24	A (29)	—	—	Nil (29)	58	Walk	10,000 steps/d	5	20–45	—	—	24
Karlsen et al 2022, Norway26	A (12)			Nil (14)	26	Cycling	90–95% of HRmax	2	30	—	—	12

A = aerobic exercise, AT = anaerobic threshold, C = continuous positive airway pressure, D = intervention duration (weeks), F = frequency (times/wk), HR_max = maximal hear rate, HRR = heart rate reserve, L = session length (minutes), n = number of patients in the group, N = number of patients in the study, NA = not available, R = resistance training, reps = repetitions, VO₂max = maximum oxygen consumption during exercise.

Table 2.

Summary of the outcome parameters pre- and postintervention.

Study	Group	n	AHI Baseline	AHI Post-Tx	AHI USMD	ESS Baseline	ESS Post-Tx	ESS USMD	BMI Baseline	BMI Post-Tx	BMI USMD	VO2peak Baseline	VO2peak Post-Tx	VO2peak USMD
Ackel-D’Elia,201223	CPAP	19	42.3 ± 21.6	34.7 ± 23.5	−7.6 ± 16.4	13.0 ± 4.8	8.8 ± 24.7	−4.2 ± 21.6*	28.5 ± 2.2	NA	NA	NA	NA	NA
	CPAP+ A	13	40.5 ± 22.9	35.0 ± 19.6	−5.5 ± 15.3	14.0 ± 4.1	9.2 ± 12.1	−4.8 ± 9.7*	28.0 ± 3.1	NA	NA	30.0 ± 5.1	32.9 ± 7.0	NA
Kline, 201116	Control	16	24.4 ± 22.4	28.9 ± 25.6	4.5 ± 9.6	NA	NA	NA	33.6 ± 5.6	33.5 ± 6.0	−0.1 ± 0.6	NA	NA	0.2 ± 0.1
	A + R	27	32.2 ± 29.1	24.6 ± 22.9	−7.6 ± 13.0	NA	NA	NA	35.5 ± 6.2	35.2 ± 6.4	−0.3 ± 1.1	NA	NA	2.3 ± 0.1
Servantes, 201130	Control	11	22.8 ± 17.4	25.9 ± 18.8	3.1 ± 11.5	NA	NA	NA	27.7 ± 3.7	NA	NA	15.7 ± 3.0	12.8 ± 3.2	−2.9 ± 0.5
	A + R	17	26.4 ± 17.6	16.4 ± 11.1	−10.0 ± 11	NA	NA	NA	28.0 ± 4.4	NA	NA	15.6 ± 2.7	20.9 ± 4.2	5.3 ± 0.5
Desplan, 201427	Control	11	39.8 ± 19.2	45.4 ± 22.5	5.6 ± 13.8	8.0 ± 5.7	9.4 ± 5.8	1.4 ± 3.6	31.3 ± 2.5	31.3 ± 2.2	0.0 ± 0.7	23.2 ± 6.0	19.8 ± 4.6	−3.4 ± 2.0
	A + R	11	40.6 ± 19.4	28.0 ± 19.3	−12.6 ± 9.6	13.6 ± 4.5	8.0 ± 5.7	−5.6 ± 3.4	29.9 ± 3.4	29.1 ± 3.1	−0.8 ± 0.6	21.3 ± 5.6	22.9 ± 5.6	1.6 ± 2.0
Servantes, 201829	Control	18	29.0 ± 17.0	31.0 ± 14.0	2.0 ± 10.2	9.0 ± 6.0	9.0 ± 5.0	0.0 ± 4.4*	29.0 ± 4.0	NA	NA	15.0 ± 3.0	14.0 ± 3.0	−1.0 ± 2.3*
	A + R	17	28.0 ± 17.0	18.0 ± 12.0	−10.0 ± 10.3	11.0 ± 3.0	7.0 ± 3.0	−4.0 ± 2.3*	30.0 ± 3.0	NA	NA	15.0 ± 2.0	20.0 ± 4.0	5.0 ± 3.0*
	CPAP	15	32.0 ± 25.0	8.0 ± 11.0	−24.0 ± 19.0*	9.0 ± 5.0	5.0 ± 3.0	−4.0 ± 3.6*	31.0 ± 4.0	NA	NA	14.0 ± 2.0	15.0 ± 3.0	1.0 ± 2.1*
	CPAP + A + R	15	25.0 ± 15.0	10.0 ± 16.0	−15 ± 12.0*	9.0 ± 5.0	6.0 ± 5.0	−3.0 ± 3.9*	30.0 ± 3.0	NA	NA	16.0 ± 3.0	19.0 ± 4.0	3.0 ± 2.9*
Guerra, 201828	Control	22	44.0 ± 28.1	50.0 ± 32.8	6.0 ± 14.1	NA	NA	NA	29.5 ± 3.7	29.7 ± 3.8	0.2 ± 1.4	25.8 ± 5.6	24.9 ± 6.1	−0.9 ± 4.6*
	A + R	21	44.0 ± 32.1	38.0 ± 22.9	−6.0 ± 13.7	NA	NA	NA	29.6 ± 4.0	29.0 ± 3.7	−0.6 ± 1.4	24.1 ± 6.0	29.2 ± 6.4	5.1 ± 4.8*
Berger, 201831	Control	45	20.7 ± 6.1	21.9 ± 11.0	1.2 ± 8.0	7.4 ± 4.3	7.3 ± 3.9	−0.1 ± 3.2	28.3 ± 4.3	28.4 ± 4.9	0.1 ± 0.9	22.8 ± 5.7	23.1 ± 5.9	0.3 ± 2.4
	A + R	43	23.1 ± 8.0	19.0 ± 10.0	−4.1 ± 9.7	8.4 ± 4.4	6.6 ± 5.7	−1.8 ± 4.1	28.5 ± 4.1	28.4 ± 4.3	−0.1 ± 0.9	23.6 ± 6.0	25.6 ± 6.1	2.3 ± 3.1
Sengul, 201119	Control	10	17.9 ± 6.5	17.4 ± 11.2	−0.6 ± 8.9	3.4 ± 5.1	5.3 ± 4.2	1.9 ± 3.1	28.4 ± 5.4	28.3 ± 5.5	−0.1 ± 1.3	16.6 ± 4.9	18.4 ± 3.5	1.8 ± 1.4
	A	10	15.2 ± 5.4	11.0 ± 5.3	−4.2 ± 4.4	8.2 ± 6.1	7.0 ± 6.7	−1.2 ± 4.1	29.8 ± 2.7	29.2 ± 3.1	−0.6 ± 1.4	15.4 ± 3.6	17.5 ± 5.6	2.1 ± 1.4
Mendelson, 201617	Control	17	28.1 ± 13.5	27.0 ± 15.1	−1.1 ± 11.2*	NA	NA	NA	26.2 ± 3.9	NA	0.1 ± 0.5	22.5 ± 8.4	23.1 ± 8.3	0.6 ± 2.5
	A	17	31.1 ± 12.9	20.5 ± 9.4	−10.6 ± 9.2*	NA	NA	NA	28.6 ± 4.5	NA	−1.3 ± 0.5	20.4 ± 4.0	21.8 ± 4.1	1.4 ± 2.5
Yang, 201825	Control	35	19.5 ± 6.1	20.1 ± 7.0	0.6 ± 5.1*	NA	NA	NA	27.1 ± 3.5	26.8 ± 3.7	−0.3 ± 2.8*	29.5 ± 6.7	29.4 ± 8.1	0.6 ± 5.1*
	A	32	20.2 ± 7.5	16.4 ± 5.2	−3.8 ± 5.4*	NA	NA	NA	27.6 ± 4.7	24.5 ± 4.2	−3.1 ± 3.5*	28.4 ± 7.1	32.1 ± 8.5	3.6 ± 6.2*
Jurado-García, 202024	Control	29	27.0 ± 9.9	25.0 ± 16.29	−2.0 ± 16.7	9.8 ± 4.56	9.2 ± 3.89	−0.6 ± 3.4	32 ± 4.3	31 ± 4.5	−1.0 ± 1.3	NA	NA	NA
	A	29	29.0 ± 20.8	23.0 ± 13.2	−6.0 ± 11.8	9.9 ± 4.42	7.7 ± 4.47	−2.2 ± 3.5	32 ± 4.1	31 ± 4.1	−1.0 ± 1.8	NA	NA	NA
Karlsen, 202226	Control	14	47.4 ± 26.1	45.5 ± 29.7	−1.9 ± 21.9*	5.3 ± 4.2	5.8 ± 4.6	0.5 ± 3.4*	37.5 ± 5.5	NA	NA	26.6 ± 8.4	27.1 ± 4.9	0.5 ± 6.1*
	A	12	32.3 ± 25.2	25.6 ± 22.8	−6.7 ± 18.7*	9.9 ± 4.2	7.4 ± 4.3	−2.5 ± 3.3*	37.2 ± 5.9	NA	NA	25.6 ± 6.0	27.7 ± 6.8	2.1 ± 5.0*

^*Estimated by formula-based calculation. A = aerobic exercise, AHI = apnea-hypopnea index, BMI = body mass index, CPAP = continuous positive airway pressure, ESS = Epworth Sleepiness Scale, n = number of patients in the group, NA = not available, post-Tx = posttreatment, R = resistance training, USMD = unstandardized mean difference, VO2peak = peak oxygen consumption during exercise.

Risk of bias in studies

The studies were evaluated using the Cochrane risk of bias tool,²¹ and the results are shown in Figure S2 in the supplemental material. Mendelson et al¹⁷ presented a high risk of bias in allocation concealment, whereas Servantes et al³⁰ did not elucidate the allocation process, making the risk of bias unclear. Blinding of participants and personnel is often not possible in exercise interventions because of their nature, except for Jurado-García et al’s study,²⁴ which reported a low risk of bias because the result was measured using a home-based pedometer. In the blinding of outcome assessments, most studies had a low risk of bias, except that of Sengul et al¹⁹, which was classified as unclear.

Impact of exercise intervention on AHI reduction

The impact of exercise on AHI reduction was evaluated using 13 comparisons from 12 studies.^{16,17,19,23–25,27–31} The intergroup differences in mean AHI changes before and after intervention were obtained (Figure 1). A significant AHI reduction of −7.08 events/h (95% confidence interval [CI]: −9.98 to −4.17, P < .00001) was observed (I² = 58%).

Figure 1. Results of the meta-analysis of the pooled effects of exercise on AHI.

AHI = apnea-hypopnea index, CI = confidence interval, MD = pre- and postintervention mean difference, n = number of patients, SD = standard deviation, USMD = unstandardized mean difference, W% = weight percent of study.

Subgroup analyses were conducted on CPAP use, baseline BMI, baseline AHI, and a combination of resistance training, intervention duration, and BMI change (Figure 2). A significant AHI reduction was observed in the group without CPAP use (−8.11 events/h, 95% CI: −10.77 to −5.64, P < .00001), but not in the group with CPAP use (5.47 events/h, 95% CI: −2.48 to 13.42, P = .18), with a significant intergroup difference (P = .001) (Figure S3 in the supplemental material).

Figure 2. Results of subgroup analyses of the effects of exercise on AHI.

AHI = apnea-hypopnea index, BMI = body mass index, CI = confidence interval, CPAP = continuous positive airway pressure, N = number of studies, USMD = unstandardized mean difference.

A significant AHI reduction was observed in the subgroups performing aerobic exercise alone (−4.52 events/h, 95% CI: −6.58 to −2.46, P < .0001) and a combination of aerobic exercise and resistance training (−9.42 events/h, 95% CI: −14.40 to −4.44, P = .0002), with an intergroup P = .08 (Figure S4 in the supplemental material).

The subgroup with intervention duration ≥12 weeks exhibited a significant AHI reduction of −6.58 events/h (95% CI: −9.56 to −3.60, P < .0001). Although a trend of AHI reduction could be observed in the group with intervention duration < 12 weeks, it was not significant (−8.84 events/h, 95% CI: −18.89 to 1.20, P = .08). No significant difference was found between the 2 groups (P = .67) (Figure S5 in the supplemental material). Subgroup analyses of baseline AHI and BMI showed no significant differences between baseline AHI ≥ 30 vs < 30 (intergroup P = .61) and baseline BMI ≥ 30 vs < 30 (intergroup P = .28) (Figure 2). Another subgroup analysis on controlling BMI changes revealed a significant AHI reduction in both subgroups with and without significant postinterventional BMI change. Furthermore, no significant differences were observed between the groups (P = .52) (Figure S6 in the supplemental material).

Impact of exercise intervention on self-reported sleepiness

Seven comparisons from 6 studies^{19,24,26,27,29,31} were included to evaluate the effect of exercise on self-reported sleepiness (Figure S7 in the supplemental material). A significant ESS reduction of −2.37 (95% CI: −3.21 to −1.54, P < .00001) was observed.

In the subgroup analysis, a significant reduction in ESS was observed in the group without CPAP use (−3.15, 95% CI: −4.63 to −1.67, P < .0001) but not in the group with CPAP use (0.91, 95% CI: −1.70 to 3.52, P = .49). The intergroup difference was also significant (P = .008) (Figure 3 and Figure S8 in the supplemental material). Significant ESS reduction was observed in both subgroups receiving aerobic exercise alone (−2.21, 95% CI: −3.53 to −0.89, P < .001) and aerobic exercise with combination of resistance training (−2.85, 95% CI: −5.68 to −0.02, P = .05) (Figure 3). Subgroup analyses of baseline AHI and BMI showed no significant differences between baseline AHI ≥ 30 vs < 30 (intergroup P = .15) and baseline BMI ≥ 30 vs < 30 (intergroup P = .32) (Figure 3).

Figure 3. Results of subgroup analyses of the effects of exercise on ESS and VO₂peak.

AHI = apnea-hypopnea index, BMI = body mass index, CI = confidence interval, CPAP = continuous positive airway pressure, ESS = Epworth Sleepiness Scale, N = number of studies, USMD = unstandardized mean difference, VO₂peak = peak oxygen consumption during exercise.

Impact of exercise intervention on BMI reduction

To measure the influence of exercise on BMI, 8 studies^{16,17,19,24,25,27,28,31} were analyzed. The results indicated that exercise significantly decreased BMI by 0.72 kg/m² (95% CI: −1.22 to −0.22, P = .005) (Figure S9 in the supplemental material). Subgroup analyses showed no significant differences between aerobic exercise alone and aerobic exercise combined with resistance training (intergroup P = .22), baseline AHI ≥ 30 vs < 30 (intergroup P = .98), or baseline BMI ≥ 30 vs < 30 (intergroup P = .17).

Impact of exercise intervention on cardiopulmonary fitness

The pooled effect of exercise on cardiopulmonary fitness was evaluated using 11 comparisons from 10 studies,^{16,17,19,25–31} which found that VO₂peak significantly improved by 3.46 ml·kg⁻¹·min⁻¹ (95% CI: 1.20 to 5.71, P = .003) (Figure 3 and Figure S10 in the supplemental material). In subgroup analyses, VO₂peak significantly increased in subgroups with CPAP use (2.00 ml·kg⁻¹·min⁻¹, 95% CI: 0.19 to 3.81, P = .03) and without CPAP use (3.61 ml·kg⁻¹·min⁻¹, 95% CI: 1.20 to 6.01, P = .003), without significant difference between groups (P = .30) (Figure 3 and Figure S11 in the supplemental material). The combination of resistance training with aerobic exercise significantly increased VO₂peak (4.46 ml·kg⁻¹·min⁻¹, 95% CI: 1.51 to 7.40, P = .003), which was not observed in the subgroup performing aerobic exercise alone (1.14 ml·kg⁻¹·min⁻¹, 95% CI: −0.18 to 2.46, P = .09), with a significant intergroup difference (P = .04) (Figure 3 and Figure S12 in the supplemental material). VO₂peak significantly increased in the subgroup with intervention duration ≥ 12 weeks (3.58 ml·kg⁻¹·min⁻¹, 95% CI: 1.01 to 6.16, P = .006), which was not observed in the subgroup with intervention duration < 12 weeks (2.90 ml·kg⁻¹·min⁻¹, 95% CI: −1.21 to 7.02, P = .17), and there was no significant difference between subgroups (P = .78) (Figure 3 and Figure S13 in the supplemental material). Subgroup analyses showed no significant intergroup difference in terms of initial BMI (≥ 30 vs < 30) and initial AHI (≥ 30 vs < 30) (Figure 3).

Publication bias and funnel plot

Egger’s test for AHI, ESS, BMI, and VO₂peak yielded insignificant results. The funnel plots for AHI, ESS, and BMI exhibited symmetry (Figure S14A, Figure S14B, and Figure S14C in the supplemental material). However, the funnel plot for VO₂peak showed asymmetry, with several studies lying outside the pseudo 95% CIs (Figure S14D in the supplemental material), indicating potential publication bias.

Sensitivity analysis

Sensitivity analysis was conducted by leaving out 1 study at a time, which demonstrated no significant change in the overall results.

DISCUSSION

In this systematic review and meta-analysis, exercise training significantly reduced AHI (USMD = −7.08 events/h, P < .00001), BMI (USMD = −0.72 kg/m², P = .005), and ESS (USMD = −2.37, P < .00001) and enhanced VO₂peak (USMD = 3.46 ml·kg⁻¹·min⁻¹, P = .003).

The pooled effect of exercise on AHI reduction shown by our meta-analysis was −7.08 events/h, consistent with the findings of most previous studies,^15,32–34 whereas a few others yielded superior or inferior results.^35,36 In terms of efficacy in AHI reduction, exercise appeared to be less potent than CPAP, oral appliances, multilevel surgery, and interdisciplinary lifestyle modification^37–40 but comparable to body weight control,^41,42 diet control,³² and myofunctional therapy.⁴³

Previous studies have also demonstrated that exercise improves self-reported outcomes, such as Pittsburgh Sleep Quality Index^44,45 and ESS.^{15,33,35,36,46,47} Our meta-analysis revealed a significant ESS reduction of −2.37 with exercise, which is comparable to the effect of exercise training and even CPAP reported previously.^47,48 We did not conduct meta-analysis based on the Pittsburgh Sleep Quality Index because it was reported in only 2 enrolled studies. Nevertheless, these 2 studies reported significantly improved Pittsburgh Sleep Quality Index following exercise intervention.^16,27

Body weight loss is undoubtedly a key element in OSA treatment^1,49–53 and exhibits a dose–response relationship with AHI reduction.^53,54 However, several studies have demonstrated that AHI and cardiovascular fitness could be improved after exercise despite minimal body weight change.^16,28,31,55 The underlying mechanism may include body fat redistribution,^16,19 strengthening of airway dilator muscles,¹¹ and rostral fluid shift.⁵⁶ Physical activity has demonstrated varied effects on body weight change, which largely depends on exercise duration and intensity and whether these are combined with dietary restriction.^55,57,58 In several studies, aerobic exercise resulted in the reduction of both AHI and BMI.^25,27,59,60 Conversely, multiple previous systematic reviews have reported that exercise improved AHI independently of body weight change.^{15,32,33,36,61} Among the studies enrolled in our analysis, 3 reported significant BMI reduction after exercise^17,25,27, whereas 5 did not.^{16,19,24,28,31} Our subgroups analysis showed that exercise significantly reduced AHI, whether BMI was decreased or not (Figure S6). Although our study revealed a significantly pooled BMI reduction of −0.72 kg/m², it was apparently not the sole factor contributing to AHI reduction after exercise.

VO₂peak represents the overall condition of the heart and lung systems and oxygen utilization capacity. It is also a reliable predictor of cardiovascular events and all-cause mortality in the general population.^62,63 Previous studies reported that patients with moderate-to-severe OSA exhibited VO₂peak reduction of 2.5 to 8.5 ml·kg⁻¹·min⁻¹, which reflects the adverse influence of OSA on cardiovascular risks.^64,65 In a meta-analysis, Fletcher et al⁶⁶ found that CPAP increased VO₂peak by 2.69 ml·kg⁻¹·min⁻¹. However, whether CPAP alone is sufficient to reverse the adverse cardiovascular impact of OSA remains inconclusive. Previous studies reported that CPAP reduced cardiovascular risks^67,68; conversely, a more recent multinational, multicenter double-blinded RCT by McEvoy et al⁶⁹ reported that CPAP did not prevent cardiovascular events in moderate-to-severe OSA. Our meta-analysis revealed that exercise increased VO₂peak by 3.46 ml·kg⁻¹·min⁻¹, which in consistent with the results of previous meta-analyses.^15,33,34 Nevertheless, the cardiovascular-protective effect of VO₂peak enhancement in OSA requires further confirmation.

The therapeutic efficacy of CPAP for OSA of any severity has been proven.¹ Nevertheless, its efficacy in real-world assessments is conspicuously compromised by poor acceptance and long-term adherence.^70,71 Moreover, residual daytime sleepiness remains a frustrating issue for around 20% of CPAP-adherent patients.⁷² Exercise was expected to provide additional benefits to such patients. However, in our subgroup analyses, significant AHI and ESS reductions were only observed in the subgroup without CPAP. The strong therapeutic effects of CPAP may have almost normalized the AHI and ESS,⁷³ downplaying the benefits of exercise. Furthermore, instead of a simple combination of exercise and CPAP, an interdisciplinary lifestyle modification may be required for CPAP-adherent patients with refractory symptoms, as corroborated in a recent RCT.⁴⁰ In our meta-analysis, despite its minor effects on AHI and ESS, exercise significantly enhanced VO₂peak even in CPAP-adherent patients. However, becuase only 1 study was included in the subgroup with CPAP use, this result should be interpreted cautiously and confirmed in a future investigation.

If exercise is not maintained, its effect on AHI declines with time, even when it could still exert long-term effects on daytime sleepiness and cardiovascular function.^26,74 Similar to CPAP and oral appliances,^38,75 persistent adherence to exercise is usually advised to patients with OSA. However, it is worth exploring for the minimum duration of exercise required to achieve a beneficial impact. Previous studies on cardiopulmonary rehabilitation demonstrated that an intervention duration ≥ 12 weeks may have a greater impact on physical activity performance.⁷⁶ A systematic review by Amiri et al⁴⁶ reported that exercise intervention ≥ 12 weeks had greater effects on sleep quality and daytime sleepiness than that < 12 weeks. The exercise intervention duration ranged from 4 to 36 weeks. Subgroup analyses revealed that significant AHI reduction and VO₂peak enhancement were only noted in the subgroup with ≥ 12-week intervention duration. Nevertheless, future studies are warranted to determine the impact of exercise intervention duration on its long-term efficacy.

Patients with concurrent obesity and severe OSA may be incapable of tolerating and sustaining exercise training owing to their impaired baseline cardiopulmonary function.¹ Interestingly, in our meta-analysis, similar effects of AHI, ESS, and BMI reductions and VO₂peak increase were observed in subgroups with baseline AHI ≥ 30 and < 30 and as well as in subgroups with baseline BMI ≥ 30 and < 30. Thus, exercise may be beneficial in OSA of any severity and regardless of the initial BMI.

It has been established that resistance training can improve muscular strength and endurance.⁷⁷ The combination of aerobic exercise and resistance training has been widely implemented for prevention of cardiovascular disease.^77,78 A previous meta-analysis by Peng et al has highlighted the benefit of combining resistance training with aerobic exercise in patients with OSA³⁴; however, the comparisons were made only based on the postintervention measurements without baseline control. Among the 12 studies included in our meta-analysis, 6 incorporated resistance training in the exercise intervention. Subgroup analyses showed a trend of greater AHI reduction (P = .08) and a significant increase in VO₂peak (P = .04) in the subgroup with combined exercise compared to aerobic exercise alone, and there was no intergroup difference in BMI (P = .22). It is deducible that combining resistance training and aerobic exercise may be more effective than aerobic exercise alone.

Strengths and limitations

Our meta-analysis has some strengths. First, it included more studies and patient numbers than the previous meta-analyses.^15,32–36 This enabled us to extract more data and conduct more comprehensive analyses of self-reported and objective outcomes. Second, to the best of our knowledge, this is the first meta-analysis to conduct subgroup analyses based on multiple clinically relevant factors. Nevertheless, this study has some limitations. First, although we managed to collect unreported parameters by contacting the corresponding authors of each enrolled study, some missing data still had to be estimated through calculation. Second, the enrolled studies exhibited high heterogeneity regarding patient characteristics and intervention design. These limitations hindered us from performing meta-regression and further subgroup analyses based on parameters such as frequency, intensity, and length of exercise session. Third, publication bias was demonstrated in the funnel plot for VO₂peak, likely due to unreported negative studies, which could limit the interpretation of our results. Finally, among the studies enrolled, only 1 (Karlsen et al²⁶) was conducted after the Covid-19 pandemic. Thus, the impact of Covid-19 on OSA could not be evaluated and would need to be addressed by future studies.

CONCLUSIONS

In this meta-analysis, exercise significantly reduced AHI, ESS, and BMI and enhanced VO₂peak in patients with OSA. Subgroup analyses revealed that in CPAP-adherent patients, exercise significantly improved VO₂peak, whereas it did not provide additional benefits in AHI and ESS reductions. The combination of resistance training and aerobic exercise resulted in a trend of greater AHI reduction and significant VO₂peak enhancement. The baseline AHI and BMI did not affect the impact of exercise training on OSA. However, further investigation is warranted to determine the optimal frequency, intensity, duration, and combination of exercise interventions for OSA.

DISCLOSURE STATEMENT

This work was supported in part by Taipei Medical University Shuang Ho Hospital (110TMU-SHH-18) and MacKay Memorial Hospital (MMH-112-118). The authors report no conflicts of interest.

ABBREVIATIONS

AHI

apnea-hypopnea index

BMI

body mass index

CI

confidence interval

CPAP

continuous positive airway pressure

ESS

Epworth Sleepiness Scale

OSA

obstructive sleep apnea

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analyses

RCT

randomized controlled trial

SD

standard deviation

USMD

unstandardized mean difference

VO₂peak

peak oxygen consumption during exercise