Objective Obstructive sleep apnoea (OSA) and obesity are both associated with endothelial dysfunction, which precedes the development of atherosclerosis. As obesity is highly prevalent in OSA, we wanted to test the hypothesis that OSA is associated with endothelial dysfunction independently of obesity.
Design Cross-sectional, population-based study.
Setting Norwegian university hospital.
Patients Seventy-one subjects (median age 44 years, 35% female) were recruited from a population-based study in Norway. Participants were categorised as obese (body mass index (BMI) ≥30 kg/m2), non-obese (BMI<30 kg/m2) with OSA (apnoea–hypopnoea index (AHI)≥10), or non-obese without OSA (AHI<5).
Main outcome measures Endothelial function measured by brachial artery ultrasound and expressed as percentage of flow-mediated dilation (FMD%).
Results When non-obese subjects without OSA were used as the reference (FMD% (mean±SD) 10.1±6.3), endothelial function was found to be impaired in subjects with OSA (FMD% 6.4±3.2) (p=0.003). FMD% did not differ between obese (6.0±3.4) and non-obese (6.7±3.1) OSA subjects (p=0.3). By univariate linear regression analysis, AHI, BMI, gender and baseline brachial artery diameter were significantly associated with FMD%. When these variables were entered into a multivariate model, only AHI was significantly associated with FMD%.
Conclusions OSA is associated with endothelial dysfunction independently of obesity and conventional risk factors.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Obstructive sleep apnoea (OSA) is associated with cardiovascular disease.1 ,2 The mechanisms underlying this association are not fully elucidated, but endothelial dysfunction, an initiating event for atherosclerosis,3 may represent a link between OSA and cardiovascular disease.4 However, because a high proportion of obese subjects have OSA, it is unclear whether the association between OSA and endothelial dysfunction is merely a reflection of obesity, or represents additional pathophysiology. Mechanisms in OSA, besides obesity, that could contribute to endothelial dysfunction are night-time hypoxaemia, enhanced adrenergic tone, and the production of reactive oxygen species.4
There are diverging reports on the effect of OSA on endothelial function, with some studies showing impaired function,5–8 while a large study with 682 participants found no association between OSA and endothelial dysfunction.9 Furthermore, obesity may be a confounding factor for the association between OSA and endothelial function, although a recent study found decreased endothelial function in normal-weight subjects with OSA compared with weight-matched subjects without OSA.10 However, in that highly selected cohort, all subjects with comorbidities were excluded, and no data on relevant laboratory biomarkers of endothelial dysfunction, such as high-sensitivity (hs) C-reactive protein (CRP), were available.
Accordingly, in a sub-study, from a cohort recruited from the general population, we tested the hypothesis that the severity of OSA is independently associated with endothelial dysfunction in a multivariate model that adjusts for obesity, circulating biomarkers of endothelial dysfunction, and other established risk factors.
The recruitment protocol and the inclusion and exclusion criteria of the population-based study, Akershus Sleep Apnoea Project, have previously been reported.11 In short, the Berlin Questionnaire was mailed to 30 000 randomly selected individuals aged 30–65 years (figure 1). The Berlin Questionnaire classifies individuals as having a high or low probability of OSA by screening for OSA risk factors and symptoms (daytime sleepiness, snoring and obesity/hypertension). In total, 16 302 individuals responded; 1772 subjects were further divided into predefined strata based on the Berlin Questionnaire, and 535 were randomly selected to undergo a clinical examination at Akershus University Hospital, Department Stensby between 2006 and 2008.12 All participants who underwent the clinical examination were subjected to a full polysomnography.
For this sub-study, we excluded individuals with cardiovascular disease, diagnosed at a previous examination according to the International Classification of Diseases–10th edition (arrhythmias, peripheral arterial disease, stroke, heart failure, coronary artery disease and valvular heart disease) (figure 1). We also excluded participants with a history of chronic kidney disease or an estimated glomerular filtration rate <60 ml/min. In addition, subjects treated with continuous positive airway pressure and older subjects (≥55 years) were excluded to allow a more direct assessment of OSA per se to endothelial function. Subjects with missing data from the initial clinical examination in 2006–2008 were not eligible for inclusion in this study. For this study, we included OSA subjects with apnoea–hypopnoea index (AHI)≥10, which left 61 subjects with OSA for assessment of endothelial function. Of these 61, 40 (66%) agreed to undergo clinical assessment of endothelial function. We stratified the OSA subjects according to body mass index (BMI), with non-obese defined as BMI<30 kg/m2 and obese as BMI≥30 kg/m2.
Applying the same inclusion and exclusion criteria to subjects with normal polysomnograms (AHI<5) as in the OSA group, 165 participants were eligible for recruitment to the control group. From this group, 41 individuals were selected by a stratified protocol according to non-obesity and a low Berlin Questionnaire score. In total, 20 obese OSA patients, 20 non-obese OSA patients, and 41 non-obese subjects free from OSA were included in this sub-study and examined for endothelial function at Oslo University Hospital, Aker, Oslo, Norway in 2008 and 2009. Non-fasting was predefined as an exclusion criterion, and three subjects (two obese OSA subjects, one control subject) were excluded from the final analysis because of non-fasting. In four subjects (one non-obese OSA subject, three control subjects), the quality of the ultrasound images was too poor to permit flow-mediated dilation (FMD) analysis. Three subjects in the control group had BMI≥30 kg/m2 and were therefore also excluded, resulting in a final study of 18 obese OSA subjects, 19 non-obese OSA subjects, and 34 non-obese subjects free from OSA.
The Regional Ethics Committee of South-Eastern Norway approved the main protocol and all sub-studies of the Akershus Sleep Apnoea Project. All participants gave written informed consent before study commencement, and the study was conducted in accordance with the Declaration of Helsinki.
Clinical examination and polysomnography recordings
Participants in this sub-study were interviewed and examined at the time of polysomnography recording.12 In short, apnoeas and hypopnoeas were recorded during a full night's sleep recording with the Somnologica 3.2 software package (Flaga-Medcare, Buffalo, New York, USA). The AHI was calculated by two US board-certified polysomnography technicians, who were blinded to the clinical characteristics of the participants.
Medication use was recorded at the time of endothelial function examination, and diabetes mellitus was defined as the use of antidiabetic medication. Blood pressure was measured, and BMI was calculated as weight (kg)/height (m)2 based on self-reported values at the time of the brachial artery ultrasound examination.
Endothelial function was assessed by the FMD method.3 The measurements were performed after an overnight fast, with no smoking and no intake of medication; the participants were in the supine position during the examination. A commercially available Vivid E9 scanner (GE Vingmed Ultrasound, Horten, Norway) with a 10 MHz transducer was used for all recordings, and the results were read offline from jpg images extracted from video files using Artery Measurement System (AMS) analysis software (Tomas Gustavsson, Gothenburg, Sweden). The transducer was placed in a fixed position showing the brachial artery during the recordings. FMD% was calculated as the percentage change between the baseline diameter and the largest diameter of the brachial artery within the first 2 min after 5 min occlusion of the right forearm. The examiner had no knowledge of results from the polysomnography recordings, and all recordings and analyses were performed by the same experienced ultrasonographer (JH) in accordance with international guidelines.13
Whole blood was obtained by venepuncture of the intermediate basilic or cephalic veins, after the ultrasound examination, and immediately put on ice. Samples were left at room temperature for 30 min to coagulate and then instantly processed. Serum samples were stored as frozen aliquots of 500 µl, and no sample had previously been thawed. Blood sampling and processing were conducted by dedicated study personnel and performed in a uniform manner throughout the study. Total cholesterol, high-density lipoprotein (HDL)-cholesterol, low-density lipoprotein (LDL)-cholesterol, triglyceride and creatinine levels were measured by standard laboratory methods. CRP was measured by a highly sensitive analysis (CRPL3; Roche Diagnostics, Penzberg, Germany). The limit of detection for the CRP analysis is 0.3 mg/l, and the coefficient of variation is 7.5% in the lower (1.8 mg/l) and 4.6% in the higher (65 mg/l) range.
Continuous data with normal distribution are presented as mean±SD, and non-parametric data as median and quartiles (Q) 1–3. Normal distribution was assessed by normal Q–Q plots and histograms, and continuous data were compared by Student's t test (normal distribution) or the Mann–Whitney U test. Categorical data are presented as absolute numbers and percentages and compared by the χ2 test or Fisher's exact test as appropriate. Correlation analyses were performed with Spearman rank order correlation. Univariate and multivariate linear regression analyses were used to assess the contribution by AHI to FMD%. Covariates, with p<0.2 by univariate analysis, were entered into a stepwise forward multivariate linear regression with FMD% as the dependent variable. In the univariate analyses, we included BMI, age, systolic blood pressure, diastolic blood pressure, total cholesterol, baseline brachial artery diameter and hs-CRP as continuous variables. Gender, smoking and diabetes mellitus were entered as categorical variables. In the multivariate analyses, p<0.05 was considered significant. PASW for Windows V.18.0 (IBM, Chicago, Illinois, USA) was used for all analyses.
Baseline characteristics stratified according to OSA status are presented in table 1. Subjects with OSA had higher diastolic blood pressure, BMI, and triglyceride and hs-CRP levels than subjects without OSA. HDL-cholesterol was lower in subjects with OSA than in subjects without OSA. When obese and non-obese OSA subjects were compared (table 2), obese subjects had higher systolic and diastolic blood pressure and higher triglyceride levels and were more likely to have diabetes mellitus than the non-obese subjects with OSA. In contrast, AHI and lipid and hs-CRP levels did not differ between obese and non-obese OSA patients.
Participants with OSA had reduced FMD% compared with participants without OSA: 6.4±3.2 vs 10.1±6.3, respectively, p=0.003 (table 1). In contrast, we found no difference in FMD% between obese and non-obese subjects with OSA: 6.0±3.4 vs 6.7±3.1, p=0.32 (table 2). FMD% correlated inversely with AHI (r=−0.33, p=0.004) and BMI (r=−0.37, p=0.002), and positively with mean SaO2 (r=0.36, p=0.002) and nadir SaO2 (r=0.29, p=0.02), but not with hs-CRP (r=−0.17, p=0.2). AHI also correlated with baseline brachial artery diameter (r=0.43, p<0.001) and hs-CRP (r=0.47, p<0.001).
By univariate linear regression analysis of all subjects, increasing AHI was significantly associated with a reduction in FMD%: B=−0.09 ((±SE) 0.04), p=0.01 (table 3). Other variables that were significantly associated with FMD% by univariate analyses were male gender (B=−2.87 (1.27), p=0.03), BMI (B=−0.31 (0.13), p=0.03), diastolic blood pressure (B=−0.13 (0.06), p=0.03), diabetes mellitus (B=−4.31 (2.20), p=0.05) and basal brachial artery diameter (B=−2.13 (0.87), p=0.02). After adjustment for the other variables significantly associated with FMD% in univariate analysis, there was an independent association between increasing AHI and a reduction in endothelial function as measured by FMD%: B=−0.09 (0.04), p=0.01. In contrast, the associations between FMD% and gender, diabetes mellitus, diastolic blood pressure, basal brachial artery diameter and BMI were attenuated and no longer significant (table 3).
The main finding in the present study is that the association between severity of OSA and impaired endothelial function is independent of obesity and other cardiovascular risk factors. Thus we demonstrated a lower FMD% in patients with OSA than in those subjects without OSA, while there was no difference in endothelial function in obese versus non-obese subjects with OSA. In a second set of analyses, severity of OSA, as measured by the AHI, was independently associated with endothelial dysfunction in a multivariate model that adjusted for confounders.
Our results extend existing information concerning the association between OSA and endothelial function measured by ultrasonographic or plethysmigraphic techniques. Previous studies on this association have yielded diverging results.5–7 14–17 There was no association between OSA and endothelial function in the Framingham part of the Sleep Heart Health Study.9 This study included older subjects with mainly mild OSA (63% of subjects with sleep-disordered breathing had AHI 5–14.9). In contrast, studies of patients with moderate to severe OSA have found a consistent reduction in endothelial function.5 ,7 ,14 ,16 A predominantly community-based study demonstrated an association between AHI and FMD%, but the association was confined to the subgroup of subjects 50 years and younger.8
Jelic et al10 recently reported lower FMD% in subjects with OSA and normal BMI compared with weight-matched subjects without OSA. The authors also explored the association between OSA and endothelial function in overweight subjects and found a non-significant trend towards reduced FMD% for OSA patients in that group. However, the study cohort was highly selected, and subjects with common comorbidities in OSA, such as hypertension and diabetes mellitus, were excluded. Thus generalisation to the broader OSA population is limited. Finally, the Sleep Heart Health/Cardiovascular Health Study18 found a significant association between increasing AHI and reduction in FMD% after adjustment for three categories of BMI in three-factor analysis by the Wald test, but not by analysis of variance. Moreover, after BMI had been combined with the demographic factors age, gender and race in multivariate linear regression analysis, the association between AHI and FMD% was attenuated and no longer significant. Hence, to our knowledge, this is the first study in adults that has demonstrated an incremental contribution by OSA to endothelial dysfunction in a comprehensive multivariate model. There are also important differences between our study sample and those of previous studies, as we now report data on subjects recruited from a population-based study in which the majority of OSA patients were undiagnosed before their inclusion in our study. Although our study also selected patients from a larger sample and excluded participants with cardiovascular diseases, we did not exclude participants with other important comorbidities, such as diabetes mellitus and arterial hypertension. Thus, compared with other studies exploring the effect of OSA on endothelial function, our results may be more generalisable to a greater part of patients with OSA.
Several mechanisms may contribute to the association between OSA and endothelial dysfunction, including hypoxia-induced sympathetic activation, augmented oxidative stress, and inflammation.1 ,4 Increased adrenergic tone triggered by repetitive attacks of nocturnal hypoxaemia in OSA is well documented.19 ,20 Recent data also suggest that OSA is associated with increased oxidative stress, which may negatively affect endothelial function.10 Enhanced inflammatory activity, reflected in elevated circulating levels of CRP, has also been reported in patients with OSA, but obesity will influence CRP levels, and it remains controversial whether the association between OSA and inflammation is independent of obesity.21 ,22 The present findings of a moderately strong correlation between AHI and CRP levels, as well as an independent association between AHI and FMD%, is compatible with the theory that increased inflammatory activity may contribute to, but not completely explain, the association between OSA and endothelial dysfunction. This is also supported by the lack of correlation between FMD% and hs-CRP in our sample. Although this finding was somewhat unexpected, other epidemiological studies, including the Framingham Heart Study, have also reported a lack of association between hs-CRP and FMD.23 ,24
This is one of the largest studies in adults specifically designed to study the relationship between OSA, endothelial function and obesity. The inclusion of younger subjects is a strength, as older age is associated with endothelial dysfunction per se, possibly obscuring the impact of other risk factors. Because of time and capacity limitations, we did not include a group of obese study subjects without OSA. This is a major limitation of our study. Examining endothelial function in obese non-OSA persons would have strengthened this study. Another limitation is the lack of follow-up data on endothelial function after continuous positive airway pressure, which would be required to determine whether the impaired endothelial function in our patients is modifiable by intervention.
In conclusion, we have found the severity of OSA to be independently associated with endothelial dysfunction, rather than simply a consequence of obesity. Future studies should explore the mechanism by which OSA induces endothelial dysfunction and the potential of intervention to reverse the impaired vascular function.
We thank the staff at the Section for Vascular Investigations, Oslo Vascular Centre, Oslo University Hospital, Aker who contributed to this study. We are also grateful for technical and administrative support from Akershus University Hospital.
Contributors TO conceived and obtained funding for the study. SA and VKS contributed to the design. SKN, JH and AR contributed to data collection. SKN, JH, AR, SA, ES, HR, VKS and TO contributed to analysis and interpretation of data. SKN drafted the first version of the manuscript. All authors critically revised the manuscript and approved the final version. TO is responsible for the overall content as guarantor.
Funding This study was supported by the South-Eastern Norway Regional Health Authority (grant number 2004219 and 2789062) and the University of Oslo, Oslo, Norway. Sponsors had no role in the study design, collection, analysis and interpretation of the data, writing of the report, or the decision to submit the manuscript for publication.
Competing interests TO has received research support from Abbott Diagnostics to Akershus University Hospital, Norway, and speaker's honoraria from Abott Diagnostics, Roche Diagnostics and Siemens Healthcare. VKS is a consultant for Neu Pro, Johnson & Johnson, Apnex Medical, Respicardia, ResMed, Medtronic, Sova Pharmaceuticals, Deshum, has received research funds from Philips Respironics Foundation and NIH, and is working with Mayo Health Solutions and their industry partners on intellectual property related to sleep and cardiovascular diseases. The remaining authors have no conflict of interests to declare.
Ethics approval Regional Ethics Committee of South-Eastern Norway.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No additional unpublished data are currently available to the public.