Author + information
- Received November 9, 2006
- Revision received January 19, 2007
- Accepted January 29, 2007
- Published online May 22, 2007.
- Shahrokh Javaheri, MD⁎ (, )
- Rakesh Shukla, PhD,
- Haoyue Zeigler, MS and
- Laura Wexler, MD
- ↵⁎Reprint requests and correspondence:
Dr. Shahrokh Javaheri, Pulmonary Section (111F), VA Medical Center, Cincinnati, Ohio 45220.
Objectives The purpose of this study was to determine whether central sleep apnea (CSA) contributes to mortality in patients with heart failure (HF).
Background Cheyne-Stokes breathing with CSA commonly occurs in patients with systolic HF. Consequences of CSA, including altered blood gases and neurohormonal activation, could result in further left ventricular dysfunction. Therefore, we hypothesized that CSA might contribute to mortality of patients with HF.
Methods We followed 88 patients with systolic HF (left ventricular ejection fraction ≤45%) with (n = 56) or without (n = 32) CSA. The median follow-up was 51 months.
Results The mean (±SD) of apnea-hypopnea index was significantly higher in patients with CSA (34 ± 25/h) than those without CSA (2 ± 1/h). Most of these events were central apneas. In Cox multiple regression analysis, 3 of 24 confounding variables independently correlated with survival. The median survival of patients with CSA was 45 months compared with 90 months of those without CSA (hazard ratio = 2.14, p = 0.02). The other 2 variables that correlated with poor survival were severity of right ventricular systolic dysfunction and low diastolic blood pressure.
Conclusions In patients with systolic HF, CSA, severe right ventricular systolic dysfunction, and low diastolic blood pressure might have an adverse effect on survival.
Heart failure (HF) is a prevalent syndrome with poor prognosis. Identification of variables that contribute to increased mortality and are treatable might lead to improvement in survival. Cheyne-Stokes breathing (CSB) with central sleep apnea (CSA) commonly occurs in patients with HF due to systolic dysfunction (1–5). Recurrent episodes of apnea (cessation of breathing) and hypopnea (decrease in breathing) followed by hyperpnea are associated with recurrent episodes of hypoxemia-reoxygenation, arousals, and relatively large negative deflections in intrathoracic pressure. These pathophysiological consequences of sleep apnea-hypopnea might result in an imbalance in myocardial oxygen delivery/consumption ratio, activation of sympathetic and other neurohormonal systems, and increased right and left ventricular afterload (6,7). Therefore, sleep apnea could contribute to mortality of patients with systolic HF.
Findley et al. (8) were the first to report that CSB/CSA was associated with increased mortality in patients with systolic HF. However, several recent studies have reported conflicting results (9–14), calling into question whether CSA contributes directly to the prognosis of HF or is merely an epiphenomenon (15). The present report represents the largest and most systematic study of a cohort of well-defined patients with compensated congestive HF and systolic dysfunction, some with and some without CSA, who were followed at the Veterans Affairs Medical Center in Cincinnati. The primary end point was all cause mortality to determine whether CSB/CSA is an independent predictor.
Within the last several years, our laboratory has been involved in a series of prospective studies of sleep apnea in patients with stable HF due to systolic dysfunction (left ventricular ejection fraction [LVEF] ≤45%) (2,16). Patients were recruited from cardiology and primary care clinics. These were patients with typical chronic congestive HF who had been treated in our clinics for a long time. At the time of recruitment, no information was sought about symptoms or risks factors for sleep apnea. Patients were examined by the cardiologist investigator to assure they were receiving adequate and stable therapy for treatment of HF. Details have been published previously (2,16,17). Exclusion criteria were unstable angina; unstable HF; interstitial lung disease; obstructive lung defect; intrinsic renal and liver disorders; use of morphine derivatives, benzodiazepines, or theophylline; and presence of obstructive sleep apnea (see subsequent text).
For uniformity, only male patients were studied, because female patients are seldom referred to this center. This protocol was approved by the Institutional Review Board of the University of Cincinnati Medical Center. Each patient signed the consent form.
Of 114 eligible patients, 100 (88% recruitment) agreed to participate. The main reasons for refusal were unwillingness to stay in the hospital for sleep studies or unwillingness to travel to the Veterans Affairs Medical Center because of distance. The mean LVEF of 9 of the 14 non-participating patients in whom data were available was 18 ± 6%, which is similar to that of the participants in the study. One hundred patients underwent polysomnography; 12 of them had predominantly obstructive sleep apnea and were excluded. Therefore, a total of 88 patients are involved in this report.
The patients were admitted to the hospital for 2 consecutive nights. On the first night, the patients slept in the sleep laboratory and electrodes were placed for habituation to the environment of the laboratory. Caffeinated products were proscribed. A detailed history including sleep habits was obtained, and physical examination was performed by 1 author (S.J.). The following tests were also obtained: complete blood count, serum electrolytes, blood urea nitrogen, serum creatinine, digoxin level, radionuclide ventriculography, arterial blood gases and pH, and pulmonary function tests. Details have been published previously (2,16,17).
On the second night, polysomnography was performed with standard techniques as detailed previously (2,16,17). We recorded electroencephalogram, electro-oculogram, and chin electromyogram. Thoracoabdominal excursions and naso-oral airflow (with a thermocouple) were measured qualitatively, and arterial blood oxyhemoglobin saturation was recorded with a pulse oximeter. These variables were recorded on a multichannel polygraph (Model 78D, Grass Instrument Company, Quincy, Massachusetts).
Standard definitions were used for sleep-related disordered breathing (18,19). An apnea was defined as cessation of inspiratory airflow for 10 s or more. An obstructive apnea was defined as the absence of airflow in the presence of rib cage and abdominal excursions. A central apnea was defined as the absence of rib cage and abdominal excursions with absence of airflow. Hypopnea was defined as a discernible reduction of airflow (or thoracoabdominal excursions) lasting 10 s or more and associated with at least a 4% drop in arterial oxyhemoglobin saturation and/or an arousal. An arousal was defined as appearance of alpha wave on electroencephalogram for at least 3 s (19). Hypopneas were defined as obstructive if thoracoabdominal excursions were out of phase. The number of apneas and hypopneas/h of sleep is referred to as the apnea-hypopnea index (AHI). Similarly, the number of arousals/h of sleep is the arousal index. Polysomnograms were scored blindly by 1 author (S.J.). For patients with AHI ≥15/h, supplemental nocturnal oxygen was recommended; however, we do not know how many of these patients received oxygen and who complied. Deceased patients were verified with Hospital Inquiry, which is an information database maintained by the Veterans Affairs regional office.
The primary outcome variable was survival from the time of polysomnogram. Our key “explanatory” variable was CSA (yes/no) as defined by categorizing AHI. We used an AHI ≥5/h to define presence of clinically significant sleep apnea. This threshold has been used to define presence of clinically significant obstructive sleep apnea (18). However, we also used other AHI thresholds for analysis of survival (see subsequent text).
To ascertain independent effect of CSA on patient’s survival, we identified potential confounders (i.e., variables associated with both the key explanatory variable [AHI] and the key outcome variable [time till death]) and covariates (i.e., variables associated with the outcome only). These variables included age, body mass index, heart rate, systolic blood pressure, diastolic blood pressure, smoking status, New York Heart Association (NYHA) functional class, right ventricular ejection fraction (RVEF), LVEF, atrial fibrillation, cause of HF (ischemic vs. idiopathic cardiomyopathy), time during sleep with oxygen saturation below 90%, minimal oxygen saturation during sleep, forced expiratory volume in 1 s (FEV1) percentage predicted, forced vital capacity (FVC) percentage predicted, FEV1/FVC percentage predicted, diffusion capacity for carbon monoxide percentage predicted, partial pressure (P) arterial blood (a) O2, PaCO2, serum [K+], serum [Na+], hemoglobin, and digoxin level <1 nmol/l versus ≥1 nmol/l.
A univariate Cox proportional hazard regression was used to obtain significant associations between continuous cofactors and survival time. The log-rank test was used to ascertain significant association between survival time and categorical cofactors. Two-sample ttest (p < 0.10) was used to ascertain whether significant difference existed between the key predictor-CSA (dichotomous) and factors that were continuous in nature. Chi-square test was used to obtain association between CSA and categorical factors. Cox proportional hazard multiple regression analysis was performed to determine whether significant difference existed in survival between the patients with CSA and those without CSA, after controlling for all identified confounders and covariates in the model. Significant confounders (p < 0.10) identified in the univariate analysis were included in the starting model for the multivariate analyses. Backward elimination procedure was used to identify final reduced model. All identified variables were included in the model, and nonsignificant covariates were removed with backward elimination according to a selection stay criterion of 0.05. The hazard ratio and its 95% confidence interval for survival were used to assess the importance of confounders and covariates. This strategy allowed us to assess independent impact of CSA on mortality after accounting for all relevant and significant covariates and confounders of mortality in this cohort of patients. Adjusted hazard ratio of survival and adjusted survival curves between patients with and without CSA was also calculated.
In addition to AHI threshold of 5/h, we also used AHI cutoff points of 10 or more, 15 or more, 20 or more, 25 or more, and 30 or more to define different patient groups with CSA and calculated hazard ratios of survival after adjusting for potential confounders. Mean values ± SD are reported. Data were analyzed with SAS software for Windows version 8 (SAS Institute, Cary, North Carolina).
Descriptive characteristics and sleep-related breathing disorders of the 2 groups of patients with and without CSA are presented in Tables 1, 2, and 3.⇓⇓⇓The average AHI of patients with CSA was 34/h, and most of these events were central apneas. These events were associated with arousals, desaturation, and decreased rapid eye movement sleep.
The median follow-up was 51 months with a 25th percentile of 21 and 75th percentile of 112 months. In univariate analysis, several of the 24 variables correlated with survival (Table 4).Systolic and diastolic blood pressures, LVEF, RVEF, and minimum saturation during sleep positively correlated with survival, whereas AHI, time below saturation of 90%, heart rate, and NYHA functional class III negatively correlated with survival. Serum digoxin ≥1 nmol/l was associated with a short survival.
In multiple Cox proportional hazard regression analysis, only 3 variables were independently associated with survival. After adjusting for all significant cofactors, patients without CSA had significantly higher survival probability than patients with CSA (hazard ratio = 2.14, p = 0.02) (Fig. 1).The median survival time of patients without CSA and with CSA were 90 and 45 months, respectively (p = 0.01) (Fig. 2).The other 2 variables that independently correlated with survival were RVEF (hazard ratio = 0.97, p = 0.003) and diastolic blood pressure (hazard ratio = 0.96, p = 0.02).
When mortality of patients with AHI <5/h was compared with those with AHI categories of 5/h and <10/h, 10/h and <15/h, and ≥15/h, the survival of patients without CSA was the longest and the respective hazard ratios were 1.52 (p = 0.37), 3.07 (p = 0.03), and 2.26 (p = 0.02). We also compared the median survival of groups below and above various AHI cutoff points (Fig. 2). In all categories, the median survival was significantly longer for patients below each threshold when compared with those above the threshold.
The results show that in patients with systolic HF, the presence of CSA defined as AHI of 5 or more /h is an independent predictor of mortality (Fig. 1). We used an AHI of 5/h as the threshold for this analysis because this threshold is used to define presence of significant obstructive sleep apnea (18). After accounting for various confounders, patients with AHI <5/h lived about twice as long as patients with AHI 5/h (90 months vs. 45 months). The excess mortality of HF patients with CSA was also evident when various groups of different thresholds (below and above various AHI levels) were compared (Fig. 2).
It must be emphasized that the effect of CSA on survival was present after adjustments for a large number of confounders, such as LVEF, NYHA functional class, heart rate, serum digoxin and sodium concentration, hemoglobin, and age, variables that have been shown to be associated with poor survival in patients with systolic HF.
The results of this study are particularly important because previous studies of survival in patients with HF, including those considering the role of angiotensin-converting enzyme inhibitors or beta blockers, have not routinely included sleep studies; therefore, the impact of sleep apnea on survival was not known.
However, there are a few studies of patients with HF in which sleep studies were performed to determine the effect of sleep apnea on survival (8–14). The results have been divergent. A major criticism in most previous studies is the small number of patients, either in the CSA or in the control group. Also, some of the studies combined cardiac transplantation with death as the primary end point. Most studies only reported limited variables, making it difficult to assess the effects of confounders on survival. Inclusion of patients with obstructive sleep apnea with CSA in some studies has resulted in homogeneity, making it difficult to interpret the results. In some studies full polysomnography was not performed.
The strength of the present study includes performing full night polysomnography after a night of habituation, scoring of sleep studies blindly by 1 investigator who also examined all patients, availability of a large number of laboratory studies on the day of polysomnography, and analysis of a variety of factors that could affect survival of HF patients. In the univariate analysis, low LVEF (p = 0.004), NYHA functional class III (p = 0.003), heart rate (p = 0.03), and high serum digoxin (p = 0.007) were significantly correlated with low probability of survival. Therefore, these results are in agreement with studies that show low LVEF and high NYHA functional class and digoxin level (20) contribute to mortality. Yet, in Cox proportional hazard multiple regression analysis, CSA, RVEF, and diastolic blood pressure were the only 3 independent predictors of survival. In regard to the mechanism of CSA’s effect on mortality, we suggest that the composite consequences of apnea (6,7) including desaturation, arousals, and increased negative intrathoracic pressure (during hyperpnea that follows central apnea) adversely affect cardiac function and contribute to excess mortality in HF.
The results of the present study showing that RVEF is an independent predictor of mortality are in agreement with the results of several previous studies (21–28). One of the major determinants of right ventricular systolic function is pulmonary artery pressure (28). In the context of the results of the present study, this is important, because it is likely that patients with sleep apnea and nocturnal desaturation had more severe pulmonary arterial hypertension (sustained or intermittent), which could have resulted in further impairment in right ventricular function. We also found that low diastolic blood pressure was a predicator of poor survival. This is consistent with the results of the Framingham study (29) showing that from 60 years on diastolic blood pressure negatively correlated with coronary heart disease and mortality. Because diastolic blood pressure is a major determinant of coronary blood flow, a low diastolic pressure could result in myocardial ischemia particularly during sleep. Normally during non–rapid-eye-movement sleep, diastolic blood pressure decreases. This could be the main factor contributing to non-demand myocardial ischemia during sleep that has a nadir between 1:00 amand 3:00 am(30). We therefore suggest that low diastolic blood pressure in conjunction with apnea-induced desaturation might contribute to myocardial ischemia and perhaps also to nocturnal arrhythmias, because HF patients with CSA have more ventricular arrhythmias during sleep than those without CSA (2).
Previous studies (31–35) show that obstructive sleep apnea contributes to mortality. Therefore, it is not surprising that this study shows that, in patients with established HF, CSA also contributes to mortality.
Our study suffers from a number of limitations. Many other variables that might affect survival of HF patients were not available for analysis, and like most previous studies of sleep apnea in HF (8–13), only a few patients were taking beta blockers (when the study began, beta-blockers were not yet part of standard therapy). Beta-blockers, by improving cardiovascular function, might improve sleep apnea and survival, but studies to ascertain effect of CSA in patients with HF taking beta blockers are lacking. Another major limitation of the present study was lack of knowledge regarding use and compliance with supplemental nocturnal oxygen, which was recommended only for patients with AHI ≥15/h. Therefore we can’t be certain whether use of oxygen affected survival. However, because we found decreased median survival time of patients with higher AHI than lower AHI across various cutoff points (Fig. 2), our data suggest that sleep apnea by itself is a major predictor of survival in HF.
The important question is whether the effective treatment of CSA is associated with improved survival. Therapeutic options (6,7) to treat CSA include nocturnal administration of supplemental nasal oxygen (36), theophylline (37), acetazolamide (38), and use of positive airway pressure devices (39,40). Among these, only continuous positive airway pressure (CPAP) has been tested for survival (39). This multicenter study (39) was prematurely terminated in part because of increased early mortality of patients in the CPAP arm. However, the high mortality might have been due to adverse hemodynamic effects of CPAP, particularly in those patients who were non-responsive to CPAP (41). Because CPAP results in increased intrathoracic pressure, it could decrease venous return, adversely affecting right ventricular stroke volume, causing hypotension, decreased coronary blood flow, and myocardial ischemia. Therefore, in some patients with HF, treatment with CPAP might further adversely affect central hemodynamic status. This is interesting in the context of the present study, because we found that right ventricular dysfunction and low diastolic blood pressure are predictors of poor survival patients. Other therapeutic modalities (6,7) such as supplemental oxygen that are devoid of adverse hemodynamic effects could improve survival, yet no long-term study has been done (42).
In summary, the results of this study show that presence of CSA is associated with poor survival in patients with systolic HF. This association remained strong after adjustments for a number of known pathophysiological factors that have generally been considered to be major predictors of poor survival in patients with systolic HF.
This work was supported by grants from the Merit Review Program.
- Abbreviations and Acronyms
- apnea-hypopnea index
- continuous positive airway pressure
- central sleep apnea
- Cheyne-Stokes breathing
- heart failure
- left ventricular ejection fraction
- right ventricular ejection fraction
- Received November 9, 2006.
- Revision received January 19, 2007.
- Accepted January 29, 2007.
- American College of Cardiology Foundation
- Tremel F.,
- Pépin J.-L.,
- Veale D.,
- et al.
- Lanfranchi P.A.,
- Somers V.K.,
- Braghiroli A.,
- et al.
- Lanfranchi P.A.,
- Braghiroli A.,
- Bosimini E.,
- et al.
- Roebuck T.,
- Solin P.,
- Kaye D.M.,
- et al.
- Corra U.,
- Pistono M.,
- Mezzani A.,
- et al.
- ↵(1992) ASDA report: EEG arousals: snoring rules and examples. Sleep 15:174–184.
- Polak J.F.,
- Holman L.,
- Wynne J.,
- et al.
- Di Salvo T.G.,
- Mathier M.,
- Semigran M.J.,
- et al.
- Mehta S.R.,
- Eikelboom J.W.,
- Natarajan M.K.,
- et al.
- Ghio S.,
- Gavazzi A.,
- Campana C.,
- et al.
- Franklin S.,
- Larson M.G.,
- Khan S.A.,
- et al.
- Verrier R.L.,
- Mittleman M.A.
- He J.,
- Kryger M.H.,
- Zorick F.J.,
- et al.
- Sasayama S.,
- Izumi T.,
- Seino Y.,
- et al.