Association of obstructive sleep apnoea with exercise capacity and mortality in a population-based study: results of SHIP-TREND

Abstract

Study objectives

Sleep and physical fitness are interrelated, with cardiovascular risk factors potentially influencing this relationship. Sleep disturbances and reduced exercise capacity are linked to higher mortality. This study examined associations between sleep characteristics, cardiopulmonary exercise testing (CPET) measures and all-cause mortality, and assessed whether reduced exercise capacity mediates the link between impaired sleep and mortality.

Methods

We analysed 1001 participants from the population-based SHIP-TREND cohort (2008–2012) who underwent single-night polysomnography and symptom-limited CPET. Associations between sleep and CPET measures, as well as mortality, were assessed using multivariable linear and Cox regression models.

Results

The mean age of participants was 54 (range 44–63) years; 47.1% were women. The apnoea–hypopnoea index (AHI) was 4.9 (95% CI 1.4–13.7) events·h⁻¹ and higher in men (7.9, 95% CI 2.6–18.8) than in women (2.6, 95% CI 0.7–9.2; p<0.001). Peak oxygen uptake (V′_O2peak) was lower in women (22 (95% CI 18–25) mL·min⁻¹·kg⁻¹) than in men (27 (95% 22–32) mL·min⁻¹·kg⁻¹; p<0.001). Higher AHI and oxygen desaturation index were inversely associated with V′_O2peak. Over a median 10.3-year follow-up, 73 deaths occurred. AHI was significantly associated with all-cause mortality across models (hazard ratio 1.31–1.72). Mediation analysis demonstrated a significant direct effect, whereas the indirect effect via V′_O2peak was not statistically significant.

Conclusions

In the SHIP-TREND-0 cohort, elevated AHI is associated with reduced V′_O2peak and increased all-cause mortality. Mediation analysis suggested a possible, but not statistically significant, contribution of impaired cardiopulmonary fitness to the relationship between sleep apnoea and mortality, highlighting a potential role of fitness that warrants further investigation.

Shareable abstract

Individuals with obstructive sleep apnoea have a higher prevalence of cardiovascular disease, which may be associated with reduced exercise capacity as assessed by cardiopulmonary exercise testing https://bit.ly/4scGmkN

Introduction

Sleep-related breathing disorders, particularly obstructive sleep apnoea (OSA), have a high prevalence in the general population. In middle-aged individuals, OSA prevalence (any severity) is 58% in men and 42% in women in Australia [1] and ∼33% in the USA [2]. In Europe, prevalence is 84% in men and 61% in women in Switzerland [3], and 59% in men and 33% in women in Germany [4]. Analysis from 16 countries suggests that, globally, 936 million people aged 30–69 years have OSA (95% CI 903–970 million) [5]. Evidence indicates that OSA prevalence has increased in recent decades [1].

Considering clinically defined OSA syndrome (OSAS) (apnoea–hypopnoea index (AHI) ≥5 events·h⁻¹ and Epworth Sleepiness Scale (ESS) score >10), international prevalence values of 1–5% in women and 3–8% in men have been reported [6], although these values are increasing. In the German Study of Health in Pomerania (SHIP) [4], OSAS prevalence was 9.7% in men and 3.0% in women between 2008 and 2012 using nocturnal polysomnography (PSG) in 1264 participants.

Restful sleep, defined as four to five sleep cycles (nonrapid eye movement (NREM) and rapid eye movement (REM) sleep) within a consistent period, is essential for optimal health. Adequate sleep duration (6–8 h·night⁻¹), sleep quality and regular sleep time are relevant, with meta-analyses showing increased cardiovascular morbidity and mortality with both shorter and longer sleep [7, 8]. Sleep disruptions can disturb the homeostasis of biological systems [9]. Even mild obstructive sleep apnoea (OSA) increases the risk of cardiovascular risk factors, such as obesity, smoking and low fitness, and diseases including hypertension, arrhythmias, coronary artery disease, metabolic syndrome and heart failure [2, 10]. Intermittent nocturnal hypoxaemia plays a central role, affecting inflammatory and immune processes via oxidative stress, increasing sympathetic activity and contributing to metabolic, cardiovascular and neurocognitive disorders [11, 12]. Physical activity has been shown to improve cardiorespiratory fitness in OSA [13], highlighting the complex, bidirectional links between sleep, mental health, socioeconomic status and physical fitness [14, 15].

Cardiopulmonary exercise testing (CPET) is the gold standard for measuring fitness and cardiovascular performance. CPET results are influenced by age, lifestyle and cardiovascular risk factors and comorbidities. Sleep disorders may also influence CPET results, investigated in insomnia [16] and OSA [17]. A meta-analysis shows peak oxygen uptake (V′_O2peak) decline (exercise capacity surrogate) only in severe OSA [18].

In conclusion, individuals with sleep disturbances and reduced exercise capacity have increased cardiovascular and all-cause mortality risk [19, 20]. We hypothesised that decreased exercise capacity partly mediates the relationship between sleep disturbances and mortality, reflecting deconditioning from impaired sleep. We aimed to examine associations of sleep measures with exercise testing and all-cause mortality.

Materials and methods

Study methods and results are reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for cohort studies [21]. The study was approved by the Ethics Committee of the University of Greifswald (BB 39/08). All participants gave written informed consent.

Study and recruitment of participants

SHIP is a population-based cohort study in West Pomerania (northeastern Germany) aiming to assess the prevalence, incidence and interrelations of risk factors, subclinical disorders and diseases. Participants underwent extensive interviews and examinations, including magnetic resonance imaging, cardiorespiratory, sonographic, neurological, dermatological, dental and ophthalmological assessments, laboratory tests, and (for some) genotyping [22, 23].

The SHIP project includes three cohorts, namely SHIP-START (1997), SHIP-TREND (2008) and SHIP-NEXT (2021) (figure S1). This study used data from SHIP-TREND, a representative sample of 10 002 residents, with 4420 participants (2275 women) after excluding deceased or relocated individuals. Age distribution, cardiovascular risk factors [24], CPET data [25] and PSG data from 1208 participants [4] have been published previously. Additional exclusions included pacemaker use and impaired echocardiographic cardiac function.

Sleep analysis

All SHIP-TREND participants were invited to a full-night PSG in a sleep laboratory following baseline medical and interview assessments, including five sleep-related questions.

All participants in the sleep study underwent a single-night, laboratory-based PSG (Alice 5 System, Philips Respironics, Eindhoven, The Netherlands) according to 2007 American Academy of Sleep Medicine (AASM) standards [26]. Recordings included EEG (F4-M1, C4-M1, O2-M1, F3-M2, C3-M2, O1-M2), EOG, EMG (chin, tibialis), ECG, nasal pressure and thermistor, thoracic and abdominal effort (plethysmography), body position, pulse oximetry, and snoring (microphone) [22].

Target bedtime was 8 h (minimum 5 h). Participants completed the following four sleep questionnaires: ESS, Insomnia Severity Index, Pittsburgh Sleep Quality Index (PSQI) and the Restless Legs Syndrome Diagnostic Index.

Study personnel were trained and certified annually by sleep experts from Charité University Hospital, Berlin, to ensure data quality.

Scoring of sleep and respiration

Sleep and respiration were visually scored according to 2007 AASM criteria under strict quality control [26]. Apnoea was defined as a ≥90% reduction in airflow for ≥10 s and hypopnoea as a ≥30% reduction for ≥10 s with ≥4% desaturation or ≥50% reduction with ≥3% desaturation or arousal. OSA severity was classified by AHI as <5 events·h⁻¹ (none), ≥5 events·h⁻¹ (mild–severe), ≥15 events·h⁻¹ (moderate–severe) and ≥30 events·h⁻¹ (severe). At the study's start, OSA with daytime symptoms was termed OSAS; excessive daytime sleepiness was defined as ESS >10.

Cardiopulmonary exercise test

A symptom-limited cardiopulmonary exercise test (CPET) was performed on a calibrated, electromagnetically braked cycle ergometer (Ergoselect 100; Ergoline, Bitz, Germany) following a modified Jones protocol [27] that included 3–5 min rest, 1 min unloaded cycling, then 16 W·min⁻¹ workload increases to exhaustion.

Gas exchange and ventilatory variables were measured breath-by-breath (VIASYS Healthcare System, Höchberg, Germany). Equipment was calibrated before each test using reference gas and volume calibration. A 12-lead ECG was recorded at rest, at every exercise minute and for 5 min recovery; blood pressure was measured with a cuff sphygmomanometer.

Tests were supervised by trained staff; data were averaged over 10-s intervals. V′_O2peak was defined as the highest 10-s average in the final exercise minute. According to clinical guidelines [28], V′_O2peak >84% pred was considered normal; study-specific reference values were applied [29].

Follow-up of vital status

From enrolment until January 2023, vital status was regularly obtained from the central registration office of Mecklenburg–West Pomerania. This office can only provide data for participants registered in the state at the time of inquiry or death. For relocated participants, vital status could only be confirmed if relatives reported the death. Participants with unknown status were considered withdrawn and censored at their last contact.

Statistical analyses

The data were analysed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). The study population was described using medians (25th, 75th percentile) for continuous variables and proportions for categorical variables. Effect sizes were calculated for sex differences and interpreted as small, moderate or large. Multivariable linear regression models were performed to estimate independent associations of sleep parameters (as continuous exposure variables) with CPET parameters. Restricted cubic splines were used to detect possible nonlinear relationships. Three knots were pre-specified, located at the 5th, 50th and 95th percentiles as recommended by Stone and Koo [30], resulting in one component of the spline function. The likelihood ratio test was used to compare the fit of the linear model and the model with restricted cubic splines. Primary analyses adjusted for confounders (age, smoking, alcohol use, medications) but excluded physical inactivity, estimating the total association of sleep parameters with CPET performance. Secondary analyses additionally adjusted for potential mediators (coronary artery disease, hypertension and physical inactivity) to estimate the direct association, with indirect pathways blocked. Given the cross-sectional design, these associations are interpreted as hypothesis-generating rather than causal.

Cox proportional hazard regression models were used to examine the association between AHI (as a categorical and continuous factor) and all-cause mortality, adjusting for age, sex, body mass index (BMI), smoking, alcohol use and medications. Cardiorespiratory fitness was incorporated in separate models using two alternative operationalisations of oxygen uptake (absolute V′_O2peak (mL·min⁻¹·kg⁻¹) and V′_O2peak % pred) to assess the robustness of the observed associations. A mediation analysis was performed to explore whether V′_O2peak could potentially contribute to the association between AHI and mortality. Direct, indirect and total effects were estimated following the approach proposed by Valeri et al. [31].

When p-values are reported, a two-sided p<0.05 is considered statistically significant.

Results

Participants

In the second SHIP cohort (2008–2012), 4420 participants aged 20–84 years were examined; after applying exclusion criteria, 1001 participants (aged 20–81 years) were included (figure 1). Demographic and functional data are shown in tables 1, 2 and S2.

FIGURE 1.

Flowchart showing the selection of the study cohort. Core: core examination programme; CPET: cardiopulmonary exercise testing; PSG: polysomnography; SHIP: Study of Health in Pomerania.

TABLE 1.

Cohort characteristics

Characteristic	All (n=1001)	Men (n=530)	Women (n=471)	p-value#
Age, years	54 (44, 63)	54 (44, 63)	54 (44, 63)	0.417
BMI, kg·m−2	27.8 (25.1, 31.1)	28.4 (25.8, 31.4)	27.1 (24.1, 31.0)	<0.001
Smoking, n (%)				<0.001
Never-smoker	405 (40.5)	160 (30.2)	245 (52.0)
Ex-smoker	411 (41.1)	258 (48.7)	153 (32.5)
Smoker	185 (18.5)	112 (21.1)	73 (15.5)
Alcohol intake during last 30 days, ethanol in g·day−1	4.2 (1.1, 12.1)	8.6 (2.7, 18.4)	2.2 (0.7, 5.7)	<0.001
Abstinence from alcohol (last 12 months), n (%)	81 (8.1)	41 (7.7)	40 (8.5)	0.661
Physical inactivity, n (%)¶	249 (24.9)	142 (26.8)	107 (22.7)	0.137
Medication (ATC code N01–N07, C07), n (%)	349 (34.9)	166 (31.3)	183 (38.5)	0.012
Disorders, n (%)
Coronary heart disease	56 (5.6)	40 (7.5)	16 (3.4)	0.004
Kidney disease	33 (3.3) n=999	17 (3.2) n=529	16 (3.4) n=470	0.931
Diabetes mellitus	130 (13.0)	78 (14.7)	52 (11.0)	0.084
Hypertension	484 (48.5) n=999	285 (54.0) n=528	199 (42.3)	<0.001

Data are presented as median (25th, 75th percentile) for continuous data or as absolute numbers and percentages for categorical data. AHI: apnoea–hypopnoea index; ATC: Anatomical Therapeutic Chemical; BMI: body mass index. ^#: Wilcoxon test (continuous data) or χ² test/Fisher's exact test (categorical data). ^¶: Participants were classed as physically inactive if they reported <1 h·week⁻¹ of regular physical activity during summer and winter.

TABLE 2.

Cardiopulmonary exercise testing (CPET) parameters by severity of obstructive sleep apnoea and sex

Parameter	5 events·h−1≤AHI<15 events·h−1 (n=263)			15 events·h−1≤AHI<30 events·h−1 (n=156)			AHI≥30 events·h−1 (n=80)
	Men (n=155)	Women (n=108)	p-value#	Men (n=109)	Women (n=47)	p-value#	Men (n=62)	Women (n=18)	p-value#
V′O2peak, mL·min−1	2300 (1905, 2734)	1453 (1300, 1730)	<0.001	2181 (1850, 2587)	1471 (1317, 1766)	<0.001	2170 (1900, 2605)	1192 (1100, 1350)	<0.001
V′O2peak, mL·min−1·kg−1	26 (21, 31)	19 (17, 23)	<0.001	24 (20, 29)	17 (15, 20)	<0.001	23 (20, 27)	15 (14, 16)	<0.001
V′O2peak, % pred¶	95 (84, 108)	96 (86, 108)	0.744	94 (82, 105)	89 (83, 101)	0.705	97 (86, 109)	73 (70, 83)	<0.001
V′O2@AT, mL·min−1	1100 (900, 1300)	850 (800, 950) n=107	<0.001	1100 (900, 1300) n=108	900 (800, 950)	<0.001	1100 (900, 1300)	750 (650, 850)	<0.001
V′O2@AT, % pred¶	85 (71, 96)	89 (80, 99) n=107	0.006	84 (73, 96) n=108	88 (83, 97)	0.074	85 (73, 96)	77 (69, 82)	0.023
V′E/V′CO2 slope	27 (25, 30)	28 (25, 30)	0.801	28 (26, 30)	27 (25, 30)	0.118	29 (26, 31)	28 (26, 32)	0.772
Maximum power output, W	180 (148, 212)	116 (100, 132)	<0.001	180 (148, 196)	116 (100, 132)	<0.001	164 (132, 212)	100 (84, 100)	<0.001
Maximum power output, % pred¶	97 (87, 110)	95 (81, 111)	0.297	95 (85, 108)	93 (81, 113)	0.948	98 (85, 106)	80 (68, 88)	<0.001
Oxygen pulse, mL	15 (13, 18)	11 (10, 13)	<0.001	15 (13, 18)	11 (10, 13)	<0.001	17 (14, 18)	10 (9, 11)	<0.001
Oxygen pulse, % pred¶	99 (87, 110)	104 (94, 116)	0.017	100 (90, 112)	100 (89, 113)	0.969	105 (92, 117)	88 (82, 108)	0.007
HRpeak, beats·min−1	160 (142, 171)	142 (126, 155)	<0.001	148 (133, 162)	139 (121, 155)	0.025	147 (125, 157)	127 (108, 137)	0.014
HRpeak, % pred¶	100 (89, 106)	94 (84, 102)	0.002	95 (85, 104)	93 (81, 105)	0.832	94 (81, 102)	88 (75, 91)	0.041
Chronotropic insufficiency, n (%)+	23 (14.8%)	27 (25.0)	0.039	24 (22.0)	13 (27.7)	0.447	20 (32.3)	7 (38.9)	0.601

Data are presented as median (25th, 75th percentile) for continuous data or as absolute numbers and percentages for categorical data. AHI: apnoea–hypopnoea index; HR_peak, peak heart rate; V′_E/V′_CO2 slope: minute ventilation/carbon dioxide production slope; V′_O2peak: peak oxygen uptake; V′_O2@AT: oxygen uptake at anaerobic threshold. ^#: Wilcoxon test (continuous data) or χ² test (categorical data). ^¶: Reference values from [29]. ⁺: Chronotropic insufficiency is defined as HR_peak <80% of target heart rate (calculated as 220−age).

Characteristics of participants who underwent PSG and CPET are shown in table S4. Differences compared to excluded participants are addressed in the discussion.

The mean age was 54 years, with no sex difference. Mean BMI was 27.8 kg·m⁻², 18.5% were active smokers and 24.9% reported low physical activity. Arterial hypertension was present in 48.5%, diabetes mellitus in 13.0%, coronary heart disease in 5.6% and chronic kidney disease in 3.3%.

Functional data

Polysomnographic data (table S1) revealed declining sleep efficiency with increasing OSA severity (89.1% mild, 85.7% moderate, 84.3% severe). Total sleep time slightly decreased, while wake after sleep onset (WASO) rose across severity levels. N1 sleep increased and N3 sleep decreased with higher AHI, indicating reduced sleep depth; REM sleep also declined in severe OSA. Men showed higher AHI and oxygen desaturation index (ODI) than women, particularly in moderate OSA (p=0.042), while sleep efficiency and total sleep time were unaffected by gender. ODI rose markedly from 5.1 to 35.4 events·h⁻¹ from mild to severe OSA. Women with mild and moderate OSA had significantly higher PSQI scores than men (both p=0.004). Daytime sleepiness (ESS >10) was slightly higher in men, reported by 18.8%, 15.1% and 8.2% of participants with mild, moderate and severe OSA, respectively.

CPET results (table 2) showed progressive declines in cardiorespiratory performance with increasing OSA severity. V′_O2peak was lower in women than men in all OSA categories (p<0.001). As % predicted, V′_O2peak was similar between sexes in mild/moderate OSA but lower in women with severe OSA (73% versus 97%, p<0.001). Oxygen uptake at anaerobic threshold (V′_O2@AT), maximum power output and oxygen pulse also decreased with OSA severity and were lower in women (all p≤0.023). Peak heart rate declined with severity and was lower in women, who more often showed chronotropic insufficiency (25.0–38.9% versus 14.8–32.3%).

Effect sizes for sleep architecture differences were mostly small to moderate, while CPET parameters showed large effects for absolute values but smaller effects for percent-predicted measure (table S5).

Across sexes, median AHI was higher in those with impaired V′_O2peak (≤84% pred) than in those with normal V′_O2peak (>84%) (figure 2). Pulmonary function tests showed no obstructive or restrictive disorders and normal diffusion capacity (table S2).

FIGURE 2.

Apnoea–hypopnoea index (AHI) in men and women with impaired or normal peak oxygen uptake (V′_O2peak). Normal V′_O2peak was defined as >84% pred [28].

Associations

Table 3 shows the associations of selected PSG variables with V′_O2peak (mL·min⁻¹·kg⁻¹ or % pred), maximum power (% pred) and maximum heart rate (% pred). For both sexes, AHI and ODI demonstrated significant inverse associations with the V′_O2peak (mL·min⁻¹·kg⁻¹). Additionally, AHI, ODI and percentage of total sleep time spent in NREM sleep stages showed significant associations with V′_O2peak (% pred) in women. The data show that there is an inverse relationship between the AHI, as well as the ODI, and the cardiopulmonary exercise capacity for both men and women (figure 3).

TABLE 3.

Associations between sleep parameters and cardiopulmonary exercise test parameters

Parameter#	Men		Women
	β (se)	p-value	β (se)	p-value
V′O2peak, mL·min−1·kg−1
AHI, events·h−1	−0.061 (0.019)	0.001	−0.123 (0.023)	<0.001
Sleep efficiency, %	0.002 (0.028)	0.946	0.017 (0.021)	0.410
TST, min	0.002 (0.004)	0.573	0.001 (0.004)	0.704
NREM, %	−0.029 (0.026)	0.258	0.062 (0.052)	0.238
NREM′, %	–	–	−9.75×10−5 (5.82×10−4)	0.094
SWS, %	0.007 (0.032)	0.828	0.013 (0.028)	0.648
REM, %	0.067 (0.043)	0.116	0.143 (0.087)	0.102
REM′, %			−3.31×10-4 (2.20×10−4)	0.133
WASO, min	0.001 (0.007)	0.850	−0.005 (0.005)	0.366
ODI, events·h−1	−0.078 (0.020)	<0.001	−0.152 (0.027)	<0.001
V′O2peak, % pred ¶
AHI, events·h−1	−0.030 (0.053)	0.569	−0.319 (0.080)	<0.001
Sleep efficiency, %	0.030 (0.080)	0.710	0.143 (0.076)	0.060
TST, min	0.001 (0.013)	0.962	0.016 (0.013)	0.196
NREM, %	−0.013 (0.075)	0.864	0.277 (0.195)	0.156
NREM′, %	–	–	−4.32×10−4 (2.17×10−4)	0.046
SWS, %	−0.011 (0.094)	0.904	0.019 (0.104)	0.856
REM, %	0.058 (0.128)	0.649	0.739 (0.324)	0.025
REM′, %	–	–	−0.002 (0.001)	0.054
WASO, min	−0.004 (0.019)	0.841	−0.034 (0.018)	0.063
ODI, events·h−1	−0.004 (0.062)	0.945	−0.410 (0.105)	<0.001
Maximum power output, % pred ¶
AHI, events·h−1	−0.046 (0.049)	0.351	−0.319 (0.088)	<0.001
Sleep efficiency, %	−0.057 (0.075)	0.442	0.257 (0.083)	0.002
TST, min	0.035 (0.024)	0.135	0.034 (0.014)	0.015
TST′, min	−1.77×10−6 (6.25×10−7)	0.005	–	–
NREM, %	−0.013 (0.075)	0.864	−0.078 (0.093)	0.396
SWS, %	0.072 (0.087)	0.409	0.456 (0.269)	0.090
SWS′, %	–		−0.001 (0.001)	0.035
REM, %	0.042 (0.119)	0.727	0.298 (0.148)	0.043
WASO, min	0.012 (0.018)	0.501	−0.060 (0.020)	0.003
ODI, events·h−1	−0.028 (0.057)	0.625	−0.410 (0.105)	<0.001
HRpeak, % pred ¶
AHI, events·h−1	−0.091 (0.033)	0.005	-0.207 (0.050)	<0.001
Sleep efficiency, %	0.223 (0.089)	0.012	0.117 (0.047)	0.014
Sleep efficiency′, %	−2.07×10−4 (1.04×10−4)	0.047	–	–
TST, min	0.004 (0.008)	0.619	0.015 (0.008)	0.065
NREM, %	−0.138 (0.046)	0.003	−0.044 (0.053)	0.398
SWS, %	0.154 (0.058)	0.008	−0.062 (0.065)	0.346
REM, %	0.110 (0.079)	0.167	0.216 (0.084)	0.010
WASO, min	0.051 (0.034)	0.133	0,023 (0.034)	0.510
WASO′, min	−7.42×10−6 (3.39×10−6)	0.025	−5.21×10−6 (3.25×10−6)	0.109
ODI, events·h−1	−0.412 (0.141)	0.003	−0.288 (0.059)	<0.001
ODI′, events·h−1	0.001 (4.18×10−4)	0.015	–	–

Peak oxygen uptake (V′_O2peak) (mL·min⁻¹·kg⁻¹) model: adjusted for age, smoking, alcohol intake during last 30 days and medication. V′_O2peak (% pred) model: adjusted for smoking, alcohol intake during last 30 days and medication. Maximum power output (% pred) model: adjusted for smoking, alcohol intake during last 30 days and medication. Peak heart rate (HR_peak) (% pred) model: adjusted for smoking, alcohol intake during last 30 days and medication. AHI: apnoea–hypopnoea index; N1–3: nonrapid eye movement sleep divided into three stages; NREM: nonrapid eye movement; ODI: oxygen desaturation index; REM: rapid eye movement; SWS: slow-wave sleep; TST: total sleep time; WASO: wake after sleep onset. ^#: NREM′, SWS′, REM′, TST′, Sleep efficiency′, WASO′ and ODI′ represent spline components. ^¶: Reference values from [29].

FIGURE 3.

Inverse association of apnoea–hypopnoea index (AHI) and oxygen desaturation index (ODI) with peak oxygen uptake (V′_O2peak) (mL·min⁻¹·kg⁻¹) in men (blue) and women (yellow). Shaded areas indicate 95% confidence intervals.

Evaluation of the association between the selected PSG variables showed a significant inverse relationship of AHI and ODI with V′_O2@AT (% pred) in women (data not shown).

AHI, ODI, sleep efficiency, total sleep time and WASO were significantly associated with maximum power (% pred) in women (table 3 and figure 4). Only total sleep time was significantly associated with maximum power (% pred) in men. Furthermore, AHI, ODI and sleep efficiency in both sexes, as well as the percentage of total sleep time spent in NREM sleep stages in men were significantly associated with the maximum heart rate (% pred) (table 3). The secondary analysis adjusting for coronary heart disease, hypertension and physical inactivity is provided in supplementary table S6. This analysis estimates the direct association and the difference between total and direct effects suggests the portion of the sleep exercise capacity association potentially mediated by cardiovascular conditions and physical inactivity.

FIGURE 4.

Relationship of total sleep time (TST) and wake after sleep onset (WASO) with maximum power output. TST showed a significant association with maximum power output in women (yellow) and men (blue). WASO showed a significant inverse association with maximum power output in women (yellow), but not in men (blue). Shaded areas indicate 95% confidence intervals.

All-cause mortality

During a median follow-up of 10.3 years up to January 2023, a total of 73 deaths were recorded. Several models were calculated for the association of AHI and ODI with all-cause mortality (table 4). The analyses revealed a significant association between AHI and all-cause mortality (the same applies to the ODI, although the data do not differ substantially from those for the AHI and are therefore not presented). Depending on the model, it is clear that the AHI has a significant relationship with mortality especially with a value of ≥15 events·h⁻¹ (model 4) or ≥30 events·h⁻¹ (models 1–4). Analysing the subgroup of participants aged >40 years, the reported results did not change substantially (table 4). In an analysis of V′_O2peak stratified by severity of OSA and survival status, the lowest median V′_O2peak values (mL·min⁻¹·kg⁻¹ and % pred) were found in the deceased subset with severe OSA (table S3).

TABLE 4.

Association between apnoea–hypopnoea index (AHI) and all-cause mortality based on Cox proportional hazard regression models

	Death, N/n	Model 1	Model 2	Model 3	Model 4
		HR (95% CI)	HR (95% CI)	HR (95% CI)	HR (95% CI)
Total study cohort
AHI<5.0 events·h−1	502/17	Reference	Reference	Reference	Reference
5 events·h−1≤AHI<15 events·h−1	263/18	1.12 (0.56, 2.21)	1.09 (0.55, 2.16)	1.06 (0.54, 2.08)	1.84 (0.94, 3.60)
15 events·h−1≤AHI<30 events·h−1	156/19	1.66 (0.82, 3.34)	1.60 (0.80, 3.23)	1.43 (0.70, 2.92)	3.06 (1.57, 5.98)
AHI≥30 events·h−1	80/19	2.78 (1.35, 5.75)	2.45 (1.19, 5.05)	2.28 (1.11, 4.72)	6.17 (3.13, 12.15)
AHI as continuous variable (ln (AHI+0.1))	1001/73	1.37 (1.09, 1.71)	1.34 (1.07, 1.68)	1.31 (1.05, 1.64)	1.72 (1.40, 2.11)
Subgroup aged >40 years
AHI<5.0 events·h−1	348/16	Reference	Reference	Reference	Reference
5 events·h−1≤AHI<15 events·h−1	243/18	1.16 (0.58, 2.32)	1.13 (0.57, 2.26)	1.09 (0.55, 2.17)	1.53 (0.78, 3.02)
15 events·h−1≤AHI<30 events·h−1	148/19	1.74 (0.86, 3.52)	1.69 (0.83, 3.44)	1.54 (0.76, 3.11)	2.61 (1.33, 5.14)
AHI≥30 events·h−1	77/19	2.90 (1.39, 6.03)	2.58 (1.25, 5.36)	2.36 (1.16, 4.81)	4.92 (2.48, 9.77)
AHI as continuous variable (ln (AHI+0.1))	816/72	1.38 (1.10, 1.73)	1.35 (1.08, 1.70)	1.32 (1.06, 1.65)	1.62 (1.31, 2.00)

Model 1: adjusted for age, sex and body mass index (BMI). Model 2: adjusted for age, sex, BMI, smoking, alcohol (ln+0.1) intake during last 30 days and medication (Anatomical Therapeutic Chemical (ATC) code N01–N07, C07). Model 3: adjusted for age, sex, smoking, alcohol (ln+0.1) intake during last 30 days, medication (ATC code N01–N07, C07) and peak oxygen uptake (V′_O2peak) (mL·min⁻¹·kg⁻¹). Model 4: adjusted for smoking, alcohol (ln+0.1) intake during last 30 days, medication (ATC code N01–N07, C07) and V′_O2peak (% pred). HR: hazard ratio.

To explore whether V′_O2peak contributes to the association between AHI and all-cause mortality, we performed a mediation analysis. The natural direct effect of AHI on mortality was statistically significant (estimate 1.38, 95% CI 1.07–1.78; p=0.012), whereas the natural indirect effect through V′_O2peak did not reach significance (estimate 1.03, 95% CI 1.00–1.06; p=0.062). The total effect was significant (estimate 1.42, 95% CI 1.10–1.83; p=0.007), with an estimated proportion mediated of approximately 9%.

These results indicate a significant direct association of AHI with mortality, while the potential mediating role of V′_O2peak should be interpreted as exploratory and hypothesis-generating given the nonsignificant indirect effect.

Discussion

In this population-based study, individuals with OSA demonstrated reduced cardiopulmonary exercise capacity compared with participants without OSA. A trend toward lower percent-predicted V′_O2peak with increasing OSA severity (AHI) was observed, although the absolute differences were modest. The association between OSA severity and V′_O2peak remained significant after adjustment for age, smoking, alcohol consumption and relevant medications, suggesting an independent relationship between OSA severity and reduced exercise capacity.

The primary analysis estimates the total association, while the secondary analysis (supplementary table S5) additionally adjusts for potential mediators, including coronary heart disease, hypertension and physical inactivity, to estimate the direct association, highlighting the portion of the association that may be mediated by these factors. These results underscore the potential contribution of cardiovascular health and physical activity in mediating the relationship between sleep medicine parameters and exercise capacity.

As the study is cross-sectional, causal inference is limited and longitudinal studies are needed to confirm mediation pathways. We also cannot exclude a bidirectional relationship, whereby reduced cardiopulmonary fitness may contribute to OSA development in addition to OSA affecting fitness.

Beyond V′_O2peak, additional CPET parameters were affected in our cohort, specifically, maximum power in women and maximum heart rate in both sexes. Overall, these findings suggest that sex differences in exercise capacity are largely quantitative rather than qualitative, whereas differences in sleep architecture appear to be relatively modest. Previous studies have reported similar alterations in CPET performance among individuals with OSA, including reduced maximal heart rate, lower maximal end-tidal carbon dioxide pressure, lower oxygen saturations during exertion and higher blood pressure during maximal exercise [32]. Impaired heart rate recovery following exercise has also been observed in OSA [33] and is known to predict poor outcomes in patients with chronic heart failure. These findings together suggest that OSA is associated with a broader pattern of cardiopulmonary dysfunction during exertion. However, it should be noted that the magnitude of the observed associations in our healthy, population-based cohort was relatively modest. While a one-unit increase in AHI was associated with small changes in CPET parameters, the clinical implications for fitness or mortality risk in this population remain uncertain. These results are therefore primarily hypothesis-generating and highlight the need for further studies in both healthy and patient populations.

Previous research on the relationship between OSA severity and exercise capacity has yielded mixed results. Some studies have reported lower V′_O2peak [34], functional aerobic capacity [35] or 6-min walk distance [15] in patients with OSA compared with controls, as well as lower daily physical activity prior to continuous positive airway pressure (CPAP) treatment [36]. Physical activity and cardiopulmonary performance have been shown to correlate closely with sleep quality and quantity [37]. More recent studies have also demonstrated significantly reduced V′_O2peak in patients with OSA [32, 33], whereas others found no clear association [38]. A meta-analysis using individual patient data found comparable V′_O2peak in patients with moderate or severe OSA versus those with no or mild disease [18], while two meta-analyses using aggregate data reported lower V′_O2peak in OSA [18, 39]. The results of our population-based cohort therefore provide further support for an independent association between OSA severity and reduced cardiopulmonary performance.

The age distribution of our cohort (median 54 years, range 20–81 years) was comparable to other large-scale investigations, including the Cleveland Clinic cohort (mean±sd 50.7±1.4 years) [40], the Swiss HypnoLaus study (median 57 years) [3] and meta-analyses on exercise performance in OSA (mean±sd 49.7±15.2 years), supporting the generalisability of our findings. Minor differences in BMI, smoking prevalence and comorbidities across cohorts likely reflect population-specific health characteristics rather than methodological inconsistencies.

All-cause mortality was assessed over a median follow-up of 10.3 years. Both AHI and ODI were positively associated with mortality and, when stratified by OSA severity, only severe OSA (AHI ≥30) remained significantly related to mortality risk. These associations were independent of age, BMI, smoking, alcohol consumption and relevant medication use. Our findings align with prior reports demonstrating increased mortality with higher OSA severity. A meta-analysis including 11 932 patients reported a pooled hazard ratio for all-cause mortality of 1.19 (95% CI 1.00–1.41) for moderate and 1.90 (95% CI 1.29–2.81) for severe OSA [41]. Similarly, the Australian Busselton Health Study and the American Sleep Heart Health Study found significantly increased mortality in severe OSA [19, 42]. Additional large-scale cohort studies, including those from Korea and the US, have also confirmed this association [20, 43].

In our data, higher AHI values were consistently associated with increased all-cause mortality, with the relationship remaining significant in participants over 40 years of age. In contrast to the U-shaped association reported by Azarian et al. [44], we found a nearly linear increase in mortality risk with higher AHI. Differences in population structure, follow-up duration and covariate adjustment, particularly the inclusion of cardiorespiratory fitness (V′_O2peak), may account for this discrepancy. These results support the concept that greater OSA severity is linked to both lower exercise capacity and higher mortality risk. Mediation analysis suggested that V′_O2peak may partially contribute to the association, although the indirect effect did not reach statistical significance, leaving the mediating role of fitness uncertain. Accordingly, these results should be interpreted as exploratory and hypothesis-generating, highlighting a potential avenue for future research rather than confirming a causal pathway.

From a clinical perspective, these findings highlight the importance of maintaining physical fitness as a potential modifiable factor to mitigate risk in individuals with OSA. In addition to established treatments such as CPAP, interventions targeting improvements in V′_O2peak through structured exercise programmes or lifestyle modification may offer complementary benefits. CPAP has been shown to improve maximal exercise capacity in patients with OSA [45], while combined lifestyle-based approaches have been associated with improved sleep quality, physical health and overall quality of life [46]. Longitudinal and interventional studies are warranted to clarify whether enhancing exercise capacity can translate into measurable survival benefits in this population.

Strengths of the SHIP-TREND study include its relatively large sample size, population-based design and comprehensive assessment protocol. Potential selection bias cannot be excluded, as a considerable proportion of invited participants declined PSG. Those who underwent PSG and CPET were slightly older, more often male and more physically active, suggesting that included participants may be somewhat healthier and more motivated than the general cohort. PSG was performed on a single night rather than two consecutive nights, which may have increased measurement variability [22] and loss to follow-up as well as the regional origin of the cohort (West Pomerania) may further limit generalisability [47]. Moreover, several associations appeared stronger or were only observed in women; however, the small number of women with severe OSA limits statistical power and the reliability of subgroup analyses and these findings should therefore be interpreted with caution.

Overall, elevated AHI was associated with reduced V′_O2peak and increased all-cause mortality, independent of age, and these associations remained significant after additional adjustment for cardiovascular risk factors. While only a subset of participants underwent PSG and CPET, our findings are consistent with published data, despite ongoing debate regarding the impact of OSA on exercise capacity and should be interpreted in light of the fact that PSG and CPET completers may not fully represent the entire cohort.

Supplementary material

Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.

Supplementary figure

01520-2025.SUPPLEMENT

Supplementary tables

01520-2025.SUPPLEMENT

Data availability

Restrictions are imposed on the availability of data generated or analysed during this study to ensure the preservation of patient confidentiality or due to the utilisation of data under license. The application for data can be made following a standardised procedure: https://www2.medizin.uni-greifswald.de/cm/fv/ship/daten-beantragen/.