Aims We tested the hypothesis that: (i) obstructive sleep apnoea (OSA) by itself originates pulmonary hypertension (PH); and (ii) the application of continuous positive airway pressure (CPAP) can reduce pulmonary pressure.
Methods and results In this randomized and cross-over trial, 23 middle-aged OSA (apnoea–hypopnoea index, 44.1±29.3 h−1) and otherwise healthy patients and 10 control subjects were included. OSA patients randomly received either sham or effective CPAP for 12 weeks. Echocardiographic parameters, blood pressure recordings, and urinary catecholamine levels were obtained at baseline and after both treatment modalities. At baseline, OSA patients had higher pulmonary artery systolic pressure than control subjects (29.8±8.8 vs. 23.4±4.1 mmHg, respectively, P=0.036). Ten out of 23 patients [43%, (95% CI: 23–64%)] and none of the control subjects had PH at baseline (P=0.012). Two patients were removed from the study because of inadequate CPAP compliance. Effective CPAP induced a significant reduction in the values for pulmonary systolic pressure (from 28.9±8.6 to 24.0±5.8 mmHg, P<0.0001). The reduction was greatest in patients with either PH or left ventricular diastolic dysfunction at baseline.
Conclusion Severe OSA is independently associated with PH in direct relationship with disease severity and presence of diastolic dysfunction. Application of CPAP reduces pulmonary systolic pressure levels.
Obstructive sleep apnoea syndrome (OSA) is the most common form of sleep-disordered breathing.1 Cardiovascular disturbances are the most important complications producing severe morbidity and mortality.2 Many risk factors for OSA, such as male gender, increasing age, and obesity are the same as for cardiovascular disease. This fact makes it difficult to establish a causal relationship between OSA and cardiovascular disorders.
Acute pulmonary artery pressure changes during sleep have been extensively reported in association with obstructive apnoeas.3 However, the role of OSA as an independent risk factor for the development of day-time pulmonary hypertension (PH) is not well known. Indeed, the effects of nasal continuous positive airway pressure (CPAP) therapy, the treatment of choice for OSA, on pulmonary artery pressure has not been addressed in prospective placebo-controlled trials.
The goals of this study were: (i) to compare the levels of pulmonary artery systolic pressure (PASP) between middle-aged adults with OSA and no other concomitant heart and lung diseases that may affect pulmonary haemodynamics, and healthy subjects with a similar age, gender distribution, and body mass index; and (ii) to evaluate the effects of CPAP therapy on pulmonary artery pressure in OSA patients in a double-blind placebo-controlled manner.
Figure 1 shows both the study protocol and the fate of participating patients. Twenty-eight newly diagnosed OSA patients and 15 healthy subjects were recruited into the study. OSA patients fulfilled the following inclusion criteria: (i) apnoea–hypopnoea index (AHI) ≥10 h−1 and excessive day-time sleepiness (Epworth sleepiness scale ≥10 points);4 and (ii) no previous treatment for OSA. Inclusion criteria for control subjects were AHI <5 h−1 and Epworth sleepiness scale <10. Exclusion criteria for both study groups were: (i) obstructive or restrictive lung disease demonstrated on pulmonary function testing; (ii) connective-tissue or chronic thrombo-embolic diseases; (iii) current cardioactive drugs; (iv) cardiac rhythm disturbances, including sinus bradycardia and sinus tachycardia; (v) known hypertension, or 24-h mean blood pressure of 135 and/or 85 mmHg or more; (vi) left ventricular (LV) ejection fraction (EF) <50%, ischaemic or valvular heart disease, cardiomyopathy, pericardial disease or stroke, by history, physical examination, electrocardiogram, chest radiography, conventional stress testing, and echocardiography; (vii) diabetes mellitus, on history or two random blood glucose levels ≥126 mg/dL; (viii) morbid obesity (body mass index >40 kg/m2); (ix) day-time hypoxaemia or hypercapnia; (x) history of cocaine or appetite-suppressant drug use. Withdrawal criteria were: (i) clinical exacerbation leading to a change in medication; (ii) hospital admission for 10 or more days; and (iii) average night nCPAP usage less than 3.5 h.
Finally, only subjects (23 OSA patients and 10 healthy subjects) in whom PASP could be estimated by echocardiography, because of the presence of less than moderate tricuspid regurgitation,5 were evaluated.
Control subjects were recruited from a list of healthy subjects from our sanitary area that had had a health routine test in the previous 3 months. We randomly selected a control subject similar in gender, age (±2 years), weight (±2 kg), and height (±5 cm) with regard to the two preceding patients included in the study.
The study was approved by the Institutional Ethics Committee at the hospital and all subjects gave their written informed consent. We performed a single-centre, prospective, randomized, double-blind, placebo-controlled, and cross-over clinical trial. Patients were randomized by one of the investigators, by means of a computer-generated randomization list using random numbers, to receive either effective CPAP or sham CPAP for two 12-week periods.6 Immediately after, the other treatment was applied with no washout period. The sample size was estimated from a preliminary study of our group, which showed a mean PASP of 27±5 mmHg in OSA patients. By two-sided contrast of paired samples, the required sample size to show a significant reduction after treatment of at least 5 mmHg with an estimated correlation coefficient of 0.22 and assuming a loss of 20% was 23 patients (α-risk of 0.05 and β-risk of 0.1). The sham CPAP device consists of a conventional CPAP device in which the area of the exhalation port is amplified thereby nearly cancelling nasal pressure, and an orifice resistor is connected between the tubing and the CPAP unit that load the blower with the same airflow resistance as in effective CPAP. All patients underwent a full-night CPAP titration study using an automated pressure setting device (Auto Set; ResMed, Sydney, Australia).7 Patients were not informed of the type of therapy they were receiving. Compliance to CPAP or sham CPAP was measured using the run-time counter built into each CPAP device.
All subjects underwent baseline sleep study, 24-h ambulatory blood pressure monitoring, and transthoracic echocardiogram. Urine specimens and demographic data were also collected. Patients were randomized to receive either effective or sham CPAP therapy for 12 weeks. They were then readmitted to the hospital, and the CPAP device was switched to the alternate mode of therapy for another 12 weeks. Echocardiogram and ambulatory blood pressure monitoring (ABPM) were repeated just after each period with either effective or sham CPAP treatment. Urine specimens were obtained at the same time at each visit.
Lung function study
Baseline spirometry was performed in accordance with American Thoracic Society recommendations8 using a pneumotachograph spirometer (MasterLab 6.0, Jaeger, Wurzburg, Germany). The values were expressed as a percentage of predicted normal.9 Day-time blood gases were non-invasively analysed by a finger pulsioximeter (Oscar II, Datex, Helsinki, Finland). Expiratory pressure of carbon dioxide (PECO2) and estimated arterial pressure of carbon dioxide (PaCO2) were obtained by the simultaneous monitoring of capnograph and ventilation at rest (Oxycon Alpha, Jaeger).
The same night when the urine specimens were collected, a sleep study was performed in OSA patients and control subjects. We used a previously validated portable respiratory recording device (Sibel Home-300; Sibel S.A., Barcelona, Spain).10
Respiratory events were classified as either obstructive or central on the basis of the presence or absence of respiratory effort. Respiratory events were scored as apnoea when there was a cessation of oronasal airflow lasting ≥10 s. Hypopnoea was defined as a decrease of 50% in oronasal airflow lasting >10 s, associated with a fall in arterial oxygen saturation (SaO2) >4% of the preceding baseline level. Mean and minimum night-time SaO2, desaturation index, and sleep time with SaO2<90% on nocturnal oximetry were computed as indexes of nocturnal oxygen saturation.
Twenty-four-h ABPM was performed on each patient with a Spacelabs device (model 90207, Redmond, WA, USA), using an oscillometric method as previously described.11
Catecholamines in urine
Urinary excretion of norepinephrine and epinephrine were determined as previously described.12
Examinations were performed using high-quality echocardiograph with 2.0–4.0 MHz probes (Hewlett Packard Sonos 5500, Andover, MA). The parameters were measured from at least three cardiac cycles. All echocardiograms were performed by the same experienced echocardiographer, who was unaware of both the subject’s group and the patient’s treatment assignment at each visit.
Non-invasive estimation of PASP by the Doppler transthoracic echocardiography method was chosen because of the close correlation with invasive measurements.13 Peak tricuspid regurgitant jet velocity (V) was determined by continuous-wave Doppler to calculate the right ventricular systolic pressure using the simplified Bernoulli equation (right ventricular systolic pressure =4V2+right atrial pressure). Right atrial pressure was assumed to be 10 mmHg. For this study, PH was defined as an estimated PASP of greater than 30 mmHg.14
Data on both LV systolic function and structure and left atrial end-systolic dimension were also obtained. LV diastolic function was assessed with both two-dimensional and Doppler echocardiography. The following variables were measured: peak flow velocity in early diastole or E-wave, peak velocity at atrial contraction or A-wave, isovolumic relaxation time, mitral deceleration time, and mitral A wave duration. To obtain pulmonary venous flow, an apical four-chamber view was used and pulsed wave Doppler sample volume was placed 1–2 cm into the right upper pulmonary vein. Peak systolic velocity or S-wave, peak diastolic velocity or D-wave, and the duration of reverse flow at atrial contraction or AR wave were measured. LV filling patterns were classified as either normal, impaired relaxation, pseudonormal, or restrictive by a modification of the Appleton et al.15 approach. Normal pattern was defined by E/A ratio >1, normal deceleration time (160–240 ms), isovolumic relaxation time (70–110 ms), and S/D and A/AR duration ratios >1. Impaired relaxation was determined by E/A ratio <1, prolonged deceleration time (>240 ms), isovolumic relaxation time (>110), and S/D and A/AR duration ratios >1. Pseudonormal pattern was identified by E/A ratio in the range of 1–1.5, normal deceleration time and isovolumic relaxation time, and S/D and A/AR duration ratios <1. Finally, restrictive pattern was defined by E/A ratio >1, short deceleration time (<160), isovolumic relaxation time (<70), and S/D and A/AR duration ratios <1.
Values are expressed as mean±SD or percentage. All statistical tests were two-sided. The comparisons between patient groups were performed using linear mixed models that included a random intercept for each subject trio (two OSA patients and one control). In the model, the study group was selected as factor, whereas gender, age, and BMI were selected as covariates. The group was included as the fixed effect, and the random effects were the factor, covariates, and the intercept. The subject trio was selected as a subject identification variable in the subjects' group. The chi-square was used to compare proportions.
A multiple logistic regression analysis was performed to identify the factors determining PH. The independent variables included in the model were gender, age, AHI, mean nocturnal SaO2, and the variables that reached statistical significance in univariate analysis. In this analysis, a forward Wald stepwise method was used with a F<0.05 and a F>0.10 as criteria to enter or remove, respectively.
Comparisons of effects of the treatments over time were made by linear mixed models with correlated residuals within the random effects (MIXED procedure). For each dependent variable, treatment was selected as factor and the period, the sequence and the baseline value of the variable were selected as covariates. For the specification of the model, the treatment and the covariates were selected as fixed factors. Because of the reduced number of patients, the mixed model used only main effects (i.e. no interactions). The slope and the intercept were modelled as random effects. These comparisons were only performed on 22 patients who completed the trial. The comparison of the number of patients with PH at baseline and after treatment was done by means of the McNemar test.
These analyses were performed using the Statistical Package for the Social Sciences for Windows Release 11.0 software (SPSS Inc., Chicago, IL). A value of P<0.05 was considered statistically significant.
General characteristics and baseline LV size and function
There were no significant differences between OSA patients randomized (n=23) and those finally not included owing to lack of tricuspid regurgitation (n=5), with respect to severity of OSA (AHI and minimum SaO2), lung function, blood pressure, urinary catecholamines, and demographic data. Control subjects included (n=10) and those not evaluated because of the absence of measurable tricuspid regurgitation (n=5) were also comparable for the aforementioned data characteristics.
Baseline characteristics as well as LV size and function in both study groups are shown in Tables 1 and 2, respectively.
Echocardiographic parameters in OSA patients and control subjects
OSA patients (n=23)
Control subjects (n=10)
Deceleration time (ms)
Isovolumic relaxation time (ms)
Left atrial diameter (mm)
LV diastolic diameter (mm)
LV systolic diameter (mm)
LV shortening fraction (%)
Interventricular septum (mm)
Posterior wall (mm)
LV mass (g)
LV mass index (g/m2)
At baseline, an abnormal LV filling pattern was present in 13 of the 23 OSA patients (56%) and only in two of the 10 healthy control subjects (20%) (P=0.045). We observed impaired relaxation pattern in nine OSA patients (39%), and pseudonormal pattern was present in the remaining four patients (17%). In the two control subjects with diastolic dysfunction, impaired relaxation was the LV filling pattern.
Pulmonary artery systolic pressure
The mean values for PASP at baseline were 29.8±8.8 mmHg for the OSA group and 23.4±4.1 mmHg for the control group (P=0.036) (Figure 2). Doppler-defined PH was present in 10 of the 23 OSA patients [43% (95% CI: 23–64%)] and in none of the 10 control subjects (0%) (P=0.012). A detailed comparison of OSA patients with and without PH is shown in Table 3.
There was a direct association between the presence of diastolic dysfunction and PH (P=0.006). Nine (69%) and four (31%) OSA patients without PH presented normal filling and impaired relaxation pattern, respectively. Conversely, only one OSA patient with PH (10%) had a normal pattern, whereas five (55%) and four (40%) presented an impaired relaxation and pseudonormal patterns, respectively. Multiple logistic regression analysis selected both AHI and forced vital capacity (% of predicted) as independent factors related to the development of PH (r2=0.75, P=0.000).
Effects of CPAP
Two patients failed to complete the trial because of an average night usage of CPAP of less than 3.5 h. Both patients had PH at baseline. Twenty-one OSA patients completed both treatment arms. Eleven patients received sham CPAP and 10 patients effective CPAP throughout the first 12-week period of the study. Mean CPAP pressure was 10±2 cm H2O, and average night usage was similar on effective CPAP (6.2±1.1 h) and sham CPAP (5.8±1.4 h).
Main demographic, blood pressure, urinary catecholamines, and echocardiographic data after the two treatment modalities in OSA patients are shown in Table 4. Twelve weeks on effective CPAP therapy induced a reduction of 4.9±3.9 mmHg (95% CI: 3.04–6.67 mmHg) [15.1±11.2% (95% CI: 10.1–20.3%)] in the level of PASP, changing from 28.9±8.6 to 24.0±5.8 mmHg (P<0.0001). Indeed, the number of OSA patients with PH was also reduced from eight (38%) to three (14%) out of 21 patients (P=0.001). Individual values for the PASP at baseline and after both sham and effective CPAP in OSA patients are shown in Figure 3.
Heart rate, weight, blood pressure data, urinary catecholamines and levels of pulmonary artery pressure in OSA patients after sham and effective CPAP
Sham CPAP (n=11)
Heart rate (bpm)
Body mass index (kg/m2)
Day-time SBP (mmHg)
Day-time DBP (mmHg)
Night-time SBP (mmHg)
Night-time DBP (mmHg)
Diurnal norepinephrine (µg/g)
Diurnal epinephrine (µg/g)
Nocturnal norepinephrine (µg/g)
Nocturnal epinephrine (µg/g)
Deceleration time (ms)
Isovolumic relaxation time (ms)
Left atrial diameter (mm)
LV diastolic diameter (mm)
LV systolic diameter (mm)
LV shortening fraction (%)
Interventricular septum (mm)
Posterior wall (mm)
LV mass (g)
LV mass index (g/m2)
Higher reduction in PASP after effective CPAP therapy was observed in OSA patients with either LV diastolic dysfunction (7.3±3.3 vs. 1.6±1.8 mmHg, P<0.001) or presence of PH at baseline (8.5±2.8 vs. 2.6±2.8 mmHg, P<0.001).
The main findings in this study were that severe OSA frequently caused day-time PH in the absence of either significant heart and lung diseases, and CPAP therapy significantly reduced the levels of day-time pulmonary artery pressure.
The used definition of PH as a PASP >30 mmHg may overestimate the prevalence of PH in the OSA patients of our study. For healthy subjects with demographic data similar to our OSA patients, specifically male subjects from 50 to 59 years of age, and those with body mass index higher than 30 and lower than 35 kg/m2, McQuillan et al.16 reported normal ranges of 21.0–40.6 and 20.5–40.9 mmHg, respectively. Indeed, in 6% of subjects older than 50 years of age and 5% of those with body mass index higher than 30 kg/m2, PASP >40 mmHg was reported. However, the presence of sleep-disordered breathing in their patients was not ruled out and this could have contributed to the results because of the high prevalence of OSA in middle-aged male subjects.1
Previous studies dealing with the presence of PH in OSA subjects have estimated its prevalence to be from 17 to 53%.17 Only three studies,18–20 however, controlled the influence of concomitant heart and lung diseases, especially the presence of chronic obstructive pulmonary disease. In the study by Bady et al.20 only pre-capillary PH was reported in 12 out of 44 (27%) patients. Sajkov et al.19 found a PH prevalence of 41% in a group of 27 patients, and Sanner et al.18 found post-capillary and pre-capillary PH in 8 and 12 out of 92 patients, respectively. Similar to our research, day-time PH was mild in these studies. These differences between the studies may be mainly because of inter-individual differences in the magnitude of cardiovascular response to factors such as intermittent hypoxia or increased ventricular afterload, length of illness, different definitions of PH, and the limited number of patients studied.
Contrary to most studies, we found AHI to be the parameter that had the closest correlation with PASP level. Also, we have found a relationship between LV diastolic dysfunction and PH. They probably establish a vicious pathophysiologic cycle in which the contribution of diastolic dysfunction to the development of PH seems to be more likely than the opposite. In a group of 120 patients with PH of different aetiologies, Moustapha et al.21 observed that only when PASP was higher than 60 mmHg impaired relaxation was developed. Although data on LV end-diastolic and pulmonary artery wedge pressures were not assessed in our study, we believe that the presence of diastolic dysfunction supports the hypothesis that high filling pressure may be an important determinant of PH in our OSA patients.
We speculate that day-time PH could have a pre-capillary component related to repetitive hypoxia-reoxygenation22 leading to both pulmonary vasoconstriction and vascular endothelial remodelling, but also a post-capillary component in relation with permanent or episodic elevations in LV filling pressure.14 In patients with PH secondary to chronic heart failure because of elevated LV filling pressure, the main component of PH is reversed in minutes to days by vasodilator agents and is related to deregulation of pulmonary vascular endothelial function. This could be the component of PH reversed after CPAP therapy in our study. However, the possible presence of structural pulmonary vascular abnormalities could play an important role in some patients preventing the reversion of PH after only 12 weeks on CPAP.
It has been reported that OSA patients have higher levels of circulating endothelin-1 and lower levels of nitric oxide than healthy subjects. Also, CPAP therapy readily restores the normal levels of these vasoactive mediators,23,24 that share an opposite regulatory pathway on both pulmonary vascular tone and vascular smooth-muscle cell proliferation. Indeed, the increased production of prostanoids might try to compensate the trend towards endothelial cell proliferation and vasoconstrictor tone.25 The effects of inter-related factors such as elimination of both nocturnal hypoxemia and nocturnal sympathetic surges, improvement in LV diastolic relaxation properties, and decreased LV afterload may restore the balance between these endothelial vasoactive mediators with the application of CPAP. The degree of reduction in PASP may depend on both the balance between the components of PH because of structural remodelling and the functional deregulation in vascular endothelial tone and the duration of therapy. In a short treatment time, only functional effects on endothelial tone might be apparent.
In two prospective non-placebo controlled studies, application of CPAP for 4 and 6 months, respectively, was associated with significant reduction in the level of pulmonary artery pressure (mean pulmonary artery pressure was reduced from 16.8±1.2 to 13.9±0.6 in the total study population in one work26 and from 25.6±4.0 to 19.5±1.6 mmHg in patients with PH at baseline in the other,27 respectively). As occurred in our study, the greatest reduction was observed in patients with day-time PH at baseline. Indeed, Sajkov et al.26 concluded that the effects of CPAP on pulmonary haemodynamics were not because of changes in LV diastolic function. However, they only used the E/A ratio as a parameter to assess the diastolic function, and they could not differentiate the possible different grades of severity of diastolic dysfunction in their patients.
Limitations include, first, the limited number of patients included does not let us to know the real prevalence of PH in OSA patients. Indeed, our OSA and otherwise healthy patients do not represent the general OSA population because of the high morbidity usually present in this patient population. However, the thorough selection process was extremely arduous in order to achieve the objectives of the study. Possible gender differences cannot be established because of the fact that patients were predominantly males. Another limitation is that right ventricular function and structure were not assessed, and pulmonary artery pressure depends on both. In this regard, Guidry et al.28 found only minimal right ventricular hypertrophy in patients with sleep-disordered breathing but a high percentage of patients with systemic hypertension, obstructive pulmonary disease, and diabetes had been included. However, asymptomatic mild PH observed in our patients without these disorders, along with the fact that both LV systolic function and structure were normal in all patients, render unlikely the presence of significant right ventricular hypertrophy or failure that could significantly affect the level of PASP.
Our results show that, even in the absence of other lung and heart diseases, day-time PH can develop in severe OSA patients. The level of pulmonary artery pressure maintained a direct relationship with both the severity of OSA and the presence of LV diastolic dysfunction. Given that the application of CPAP reduced the level of PASP, we speculate that long-term CPAP therapy might avoid the development of irreversible structural pulmonary vascular and right ventricular changes that could impair the prognosis of these patients.
This research was partly supported by a grant from the Fondo de Investigación Sanitaria (F.I.S.; exp. 01/0278) and Neumomadrid (2000).
MarinJM, CarrizoSJ, VicenteE, AgustiAG. Long-term cardiovascular outcomes in men with obstructive sleep apnoea–hypopnoea with or without treatment with continuous positive airway pressure: an observational study. Lancet 2005;365:1046–1053.
ChoongCY, AbascalVM, WeymanJ, LevineRA, GentileF, ThomasJD, WeymanAE. Prevalence of valvular regurgitation by Doppler echocardiography in patients with structurally normal hearts by two-dimensional echocardiography. Am Heart J 1989;117:636–642.
MassieCA, McArdleN, HartRW, Schmidt-NowaraWW, LankfordA, HudgelDW, GordonN, DouglasNJ. Comparison between automatic and fixed positive airway pressure therapy in the home. Am J Respir Crit Care Med 2003;167:20–23.
QuanjerPH, TammelingGJ, CotesJE, PedersenOF, PeslinR, YernaultJC. Lung volumes and forced ventilatory flows. Report Working Party Standardization of Lung Function Tests, European Community for Steel and Coal. Official Statement of the European Respiratory Society. Eur Respir J 1993;16(suppl.):5–40.
BallesterE, SolansM, VilaX, HernandezL, QuintoL, BolivarI, BardagiS, MontserratJM. Evaluation of a portable respiratory recording device for detecting apnoeas and hypopnoeas in subjects from a general population. Eur Respir J 2000;16:123–127.
AppletonCP, HatleLK, PoppRL. Relation of transmitral flow velocity patterns to left ventricular diastolic function: new insights from a combined hemodynamic and Doppler echocardiographic study. J Am Coll Cardiol 1988;12:426–440.
McQuillanBM, PicardMH, LeavittM, WeymanAE. Clinical correlates and reference intervals for pulmonary artery systolic pressure among echocardiographically normal subjects. Circulation 2001;104:2797–2802.
IpMS, LamB, ChanLY, ZhengL, TsangKW, FungPC, LamWK. Circulating nitric oxide is suppressed in obstructive sleep apnea and is reversed by nasal continuous positive airway pressure. Am J Respir Crit Care Med 2000;162:2166–2171.
GuidryUC, MendesLA, EvansJC, LevyD, O'ConnorGT, LarsonMG, GottliebDJ, BenjaminEJ. Echocardiographic features of the right heart in sleep-disordered breathing: the Framingham Heart Study. Am J Respir Crit Care Med 2001;164:933–938.