Advanced heart failure and nocturnal hypoxaemia due to central sleep apnoea are associated with increased serum erythropoietin

Abstract

Aims

Central sleep apnoea (CSA) and increased serum erythropoietin (EPO) concentration have each been associated with adverse prognosis in heart failure (HF) patients. The aim of this study was to examine the relationship between nocturnal hypoxaemia due to CSA and the serum EPO concentration in patients with HF.

Methods and results

Heart failure subjects (n = 33) and healthy controls (n = 18) underwent polysomnography (PSG) for diagnosis of CSA and identification and quantification of hypoxaemia. Blood collection for measurement of EPO was performed immediately post-PSG. For the analysis, HF subjects were dichotomized into subgroups defined by the presence or absence of CSA and by HF severity. Multivariate analyses were performed to evaluate the relationships of hypoxaemia and advanced HF to EPO concentration. Mean EPO concentration was 62% higher for HF subjects with CSA than for healthy controls (P = 0.004). The magnitude of nocturnal hypoxaemia was significantly and positively related to EPO concentration (r = 0.45, P = 0.02). Advanced HF was also significantly and positively related to EPO concentration (r = 0.43, P = 0.02). On multivariate analysis, the presence of combined nocturnal hypoxaemia and advanced HF yielded greater correlation to EPO concentration than either factor alone (r = 0.57, P = 0.04 and P = 0.05, respectively). Linear regression demonstrated that the combination of New York Heart Association Class and CSA was strongly associated with EPO concentration (P < 0.0001).

Conclusion

In non-anaemic HF patients, advanced HF and hypoxaemia due to CSA may each be independently associated with increased serum EPO concentration.

See page 309 for the editorial comment on this article (doi:10.1093/eurjhf/hfq023)

Introduction

Sleep-disordered breathing is classified as either obstructive sleep apnoea (OSA) or central sleep apnoea (CSA). The former is characterized by repetitive collapse of the upper airway, whereas the latter is due to periodic loss of ventilatory drive.¹ In contrast to OSA, it is likely that CSA is a consequence rather than a cause of heart failure (HF), more frequent among HF patients and related to HF severity.^1,2 Furthermore, in HF patients, CSA may promote disease progression and mortality.^3,4

Erythropoietin (EPO) is a glycoprotein hormone produced in the renal cortex that stimulates haematopoiesis and may also be important in cardiovascular disease pathogenesis.⁵ In healthy individuals, the production and release of EPO is regulated via a feedback loop mechanism responding to arterial oxygen concentration such that 2 h of continuous hypoxaemia or four hours of intermittent hypoxaemia during wakefulness increases circulating EPO concentration by 50%.⁶ It has also been demonstrated that patients with hypoxaemia due to OSA have increased EPO concentration which is attenuated by treatment with continuous positive airway pressure.⁷ In HF patients, elevation of serum EPO concentration independent of haemoglobin concentration has been associated with adverse prognosis.^5,8,9 However, EPO may also be cardioprotective in the setting of ischaemia–reperfusion,¹⁰ prevent apoptosis, and promote an anti-inflammatory response by cardiac myocytes.¹¹ The reasons for increased EPO in HF patients have not been established, though it has been suggested that bone marrow resistance may play a role.¹²

Episodic nocturnal hypoxaemia due to CSA is frequent in HF patients and has been implicated as a stimulus for increased sympathetic neural activity as well as increased circulating concentrations of catecholamines.¹³ However, no prior study has established a relationship between the severity of hypoxaemia due to apnoea or hypopnoea and EPO in HF patients with sleep-disordered breathing. We hypothesized that HF patients with hypoxaemia due to CSA have increased serum EPO concentration compared with healthy controls and with HF patients without nocturnal hypoxaemia. Accordingly, the aim of this study was to examine the relationship of nocturnal hypoxaemia due to CSA to the serum concentration of EPO in HF patients.

Methods

Subjects

This study was approved by the Mayo Clinic Institutional Review Board and complies with the Declaration of Helsinki. Consecutive, ambulatory patients with stable HF and left ventricular ejection fraction (LVEF) ≤35% by echocardiography receiving optimal medical therapy (n = 33) and healthy controls (n = 18) were prospectively enrolled to undergo investigational, laboratory-based overnight polysomnography (PSG) for detection of OSA or CSA and quantification of the apnoea–hypopnoea index (AHI) and transcutaneous arterial oxygen concentration. Patients with moderate or severe chronic obstructive pulmonary disease, daytime arterial oxygen saturation <90%, or who had been treated with an EPO-analogue or were active smokers were excluded. Patients with anaemia defined as haemoglobin concentration <11.5 g/dL or polycythemia defined as haemoglobin concentration >18 g/dL were also excluded as were subjects found to have OSA by PSG.

Assessment of New York Heart Association (NYHA) class was performed at the most recent outpatient evaluation and subjects with NYHA class III–IV symptoms were considered to have advanced HF. Body mass index was computed as weight in kilograms divided by height in metres squared. Estimated glomerular filtration rate (eGFR) was calculated by the Cockroft–Gault formula.¹⁴ Other clinical characteristics were summarized from the patient record including medications with doses used for the treatment of HF and laboratory investigations performed within 2 days of PSG.

Polysomnography

Diagnostic PSG was performed in the Clinic Research Unit Sleep Core Laboratory and digitally recorded by a multichannel system (Network Concepts Incorporated, Middleton, WI, USA or PSG Online2 E-Series, Compumedics, Abbotsford, Victoria, Australia) and scored using Uniquant or Profusion2 PSG software. Simultaneously recorded parameters included three-channel electroencephalography, two-channel electro-oculography, oronasal airflow by pressure transducer and thermocouple sensors, submental and limb electromyograms, electrocardiography, transcutaneous pulse oximetry (Ohmeda 3740, Madison, WI, USA), thoracic and abdominal respiratory effort by inductance plethysmography, snoring by tracheal microphone or piezo crystal sensor, and body position by closed-circuit video monitoring.

Scoring of sleep stages, disordered breathing events, oxygen desaturation, and periodic limb movement was performed by an experienced polysomnographer and results reviewed by a qualified physician in accordance with current American Academy of Sleep Medicine guidelines.¹⁵ Apnoeas were defined as cessation of airflow or >90% reduction of airflow from baseline for ≥10 s. Hypopnoeas were defined as a reduction in airflow of ≥50% for ≥10 s followed by an oxygen desaturation of ≥4%. Events were classified as central when the airflow criteria were met in the absence of respiratory effort as recorded by thoracic and abdominal inductance plethysmography and as obstructive when airflow criteria were met despite continued or increased respiratory effort. After classification, disordered breathing events were quantified by AHI and reported as the mean number of events per hour. Subjects were considered to have CSA if the total AHI was ≥15 with ≥50% of disordered breathing events classified as of central origin with a Cheyne–Stokes pattern of respiration. Oxygen saturation was measured prior to and continuously during sleep and quantified by both the mean oxygen saturation and the proportion of sleep time spent with arterial oxygen saturation <90% (T90%).

Erythropoietin measurement

Blood for EPO measurement was collected immediately post-PSG for quantification by automated chemiluminescent immunometric assay using a commercially available kit (IMMULITE 2000, Siemens, Germany); the functional sensitivity of this assay for EPO is <1.0 mIU/mL. For all subjects, an index of resistance to EPO was estimated by the ratio of log (EPO) concentration to haemoglobin concentration as previously described.¹⁶

Statistical analysis

Data are summarized as frequencies for categorical variables and means with standard deviation (mean ± SD) for continuous variables unless otherwise noted. Group differences were evaluated by two-sided t-test or Wilcoxon's rank-sum test depending on distribution. Differences in proportions were tested by Fisher's exact test with results displayed as the odds ratio. A linear trend analysis was performed to evaluate the relationship of combined NYHA Class and CSA to EPO concentration.

Linear least squares regression was performed in a stepwise manner to evaluate the relationship between T90% and EPO concentration with tests for confounding by age, gender, advanced HF, LVEF, haemoglobin concentration and eGFR performed a single variable at a time. A similar analysis was performed to evaluate the relationship of advanced HF to EPO concentration with adjustment for these same factors as well as adjustment for mean nocturnal oxygen saturation and T90%, concurrently. The factors selected for adjustment were chosen because of known associations with either CSA or serum EPO concentration. Analyses were performed with JMP version 7 (SAS Institute, Cary, NC, USA). For all comparisons, P < 0.05 was considered significant.

The authors had full access to the data and take responsibility for its integrity. All authors have read and agree to the manuscript as written.

Results

Subject characteristics

Of the 33 HF subjects enrolled for PSG, 14 (42%) had CSA and 15 (45%) had neither CSA nor OSA. Four subjects (12%) were found to have OSA and excluded from further analysis. Heart failure subjects with CSA were compared with those without CSA and observed to have similar age, body mass index, NYHA class, LVEF, and haemoglobin concentration. While subjects with CSA had a lower eGFR compared with those without CSA, this difference was not statistically significant (64.7 vs. 74.2 mL/min/1.73 m²; P = 0.25). Men were significantly more likely to have CSA than women (odds ratio 11.4, P = 0.04) (Table 1). Among HF subjects, 97% were treated with an angiotensin converting enzyme inhibitor or angiotensin II receptor blocker, 93% were on beta-blocker, 73% on loop diuretic, 76% on digoxin, and 28% on spironolactone. No differences in the proportion of patients treated with or doses utilized of these medications were observed on comparison of HF subjects with CSA vs. those without CSA.

Table 1.

Subject characteristics

	Healthy controls (n = 18)	HF without CSA (n = 15)	HF with CSA (n = 14)
AHI (events/h)	2.3 ± 2.7	3.6 ± 4.4	45.0 ± 21.4a
Age (year)	54.7 ± 16.8	59.9 ± 12.0	65.7 ± 10.7
Gender (% male)	72	53	93*
BMI (kg/m2)	25.1 ± 3.1	27.2 ± 3.8	28.3 ± 4.3
NYHA I-II (n)	–	10	8
NYHA III-IV (n)	–	5	6
LVEF (%)	63.8 ± 8.7	27.6 ± 7.5	24.4 ± 8.6
Haemoglobin (g/dL)	14.0 ± 1.1	13.6 ± 1.2	14.2 ± 1.2
eGFR (mL/min/1.73 m2)	81.9 ± 27.1	74.2 ± 19.3	64.7 ± 24.0

Unless otherwise specified, values presented as mean ± SD.

AHI, apnoea–hypopnoea index; BMI, body mass index; eGFR, estimated glomerular filtration rate; CSA, central sleep apnoea; HF, heart failure; LVEF, left ventricular ejection fraction; n, number; NYHA, New York Heart Association.

^aHealthy controls vs. all HF subjects P = 0.001; HF with CSA vs. HF without CSA P < 0.0001.

*P = 0.04.

Hypoxaemia, advanced heart failure, and erythropoietin

On univariate analysis, HF subjects had more nocturnal hypoxaemia than healthy controls as measured by both mean oxygen saturation (93.7 vs. 96.2%, P = 0.003) and T90% (8.8 vs. 0.4%, P = 0.01). Among HF subjects, mean nocturnal oxygen saturation was similar for those with CSA compared with those without CSA (93.6 vs. 93.8%, P = 0.87). Not unexpectedly, HF subjects with CSA had significantly more oxygen desaturation as measured by T90% (16.9 vs. 0.7%, P = 0.0006) (Table 2), though there was considerable variation in the severity of hypoxaemia (Figure 1). On average, subjects with HF and CSA had EPO concentration 62% greater than healthy controls (19.0 vs. 13.4 mIU/mL, P = 0.01), though haemoglobin concentration and eGFR were similar (Table 1). The subgroup of subjects with advanced HF had mean EPO concentration 42% higher than subjects with non-advanced HF (23.3 vs. 16.4 mIU/mL, P = 0.04) despite similar haemoglobin concentration and eGFR (Table 1).

Table 2.

Hypoxaemia and erythropoietin measures by subgroup

	Healthy controls	HF without CSA	HF with CSA
Awake O2 saturation	96.8 ± 2.2	96.5 ± 1.6	95.9 ± 2.3
Mean nocturnal O2 saturation	96.2 ± 1.4a	93.8 ± 4.3	93.6 ± 2.9
T90%	0.4 ± 0.9b	0.7 ± 1.8	16.9 ± 20.5c
EPO (mIU/mL)	13.4 ± 4.1d	16.5 ± 6.0	21.8 ± 9.1e
Log (EPO): haemoglobin ratio	0.18 ± 0.03f	0.20 ± 0.04	0.22 ± 0.04

Values presented as mean ± SD.

CSA, central sleep apnoea; O₂, oxygen; EPO, erythropoietin; HF, heart failure; T90%, proportion of sleep time with arterial oxygen saturation <90%.

^aHealthy controls vs. all HF patients, P = 0.003.

^bHealthy controls vs. all HF patients, P = 0.01.

^cHF with CSA vs. HF without CSA, P = 0.0006.

^dHealthy controls vs. all HF patients, P = 0.01.

^eHealthy controls vs. HF patients with CSA, P = 0.004.

^fHealthy controls vs. all HF patients, P = 0.01.

Figure 1.

Differences in magnitude of arterial oxygen desaturation in two subjects with CSA. (A) Subject with HF and CSA with typical Cheyne–stokes breathing pattern with marked desaturations. The AHI was 64. Baseline oxygen saturation is 94–95% with desaturations to 84%; T90% was 34.6%. Serum EPO concentration was 44 mIU/mL. (B) Subject with HF and CSA with mild desaturations. The AHI was 28. Baseline oxygen saturation is 99% with desaturations to 91–92%; T90% was 0.7%. Serum EPO concentration was 19 mIU/mL. Notably, the mean EPO values for HF patients with CSA did not differ significantly from patients without CSA (Table 2). However, the measure of duration of time spent below 90% desaturation (T90%) did correlate significantly with EPO concentration. AHI, apnoea–hypopnoea index; CSA, central sleep apnoea; EPO, erythropoietin; HF, heart failure.

The magnitude of nocturnal hypoxaemia as measured by T90% was significantly and positively related to EPO concentration (r = 0.45, P = 0.02). Advanced HF was also significantly and positively related to EPO concentration (r = 0.43, P = 0.02) and this persisted after adjusting for both T90% and mean nocturnal oxygen saturation. By multivariate analysis and after adjusting for mean nocturnal oxygen saturation, the combination of T90% and advanced HF yielded greater correlation to serum EPO concentration than either advanced HF or T90% alone (r = 0.57, P = 0.04 and P = 0.05, respectively).

Apnoea–hypopnoea index and erythropoietin

Linear regression demonstrated that the combination of New York Heart Association Class and CSA was strongly associated with EPO concentration (P < 0.0001, Figure 2). However, among HF subjects, EPO concentration was only weakly correlated to the AHI (r = 0.36, P = 0.06), and the AHI was also weakly related to T90% (r = 0.35, P = 0.07). Subjects with HF and CSA showed only a trend toward higher mean EPO concentration compared with HF subjects without CSA (21.8 vs. 16.5, P = 0.08).

Figure 2.

Erythropoietin (EPO) concentration increases due to combination of severity of HF and CSA. Mean EPO concentration by subgroup with 95% confidence intervals (linear trend analysis, P< 0.0001). The combination of increased severity of HF by NYHA Class and presence or absence of CSA are associated with increased serum EPO concentration. AHI, apnoea–hypopnoea index; CSA, central sleep apnoea; EPO, erythropoietin; HF, heart failure; NYHA, New York Heart Association.

Haemoglobin and erythropoietin

No significant correlation was found between log (EPO) and haemoglobin concentration for HF subjects or healthy controls. The ratio of log (EPO) to haemoglobin concentration was significantly higher for HF subjects compared with healthy controls consistent with resistance to EPO among HF subjects (0.21 vs. 0.18 mIU/mL per mg/dL, P = 0.01, Table 2). However, no difference in the ratio of log (EPO) to haemoglobin concentration was found for HF subjects with CSA compared with those without CSA.

Discussion

The novel findings of this study include that, first, among non-anaemic ambulatory HF patients with CSA, serum EPO concentration appears independently related to the severity of nocturnal hypoxaemia as measured by T90%, and second, that advanced HF may also be associated with increased EPO concentration independent of nocturnal hypoxaemia. Furthermore, the combination of New York Heart Association Class and the presence or absence of CSA was strongly associated with serum EPO concentration. Hypoxaemia has been previously shown to be associated with increased serum EPO concentration in healthy individuals⁶ and our data extend this association to HF patients with hypoxaemia due to CSA. We suggest that among HF patients with CSA, episodic nocturnal oxygen desaturation to <90% may be a stimulus for increased EPO production, whereas less significant desaturations may not lead to increased EPO.

Prior studies describing the association between HF and increased serum EPO concentration did not report assessment for nocturnal hypoxaemia. Given the high frequency of sleep-disordered breathing in HF patients, it seems plausible that a potential factor contributing to EPO elevation in these studies may have been unrecognized nocturnal hypoxaemia due to sleep disordered breathing. The magnitude of EPO elevation observed among the HF subjects in our study was substantial and very similar to that associated with increased mortality in a cohort of HF patients reported by van de Meer et al.¹⁶ Overall, in our study, HF subjects and HF subjects with CSA had 42% and 62% higher mean serum EPO concentration than healthy controls, respectively.

The AHI has been considered the primary measurement of the severity of CSA and an endpoint for the assessment of the efficacy of treatment intervention.¹⁷ However, we observed that the AHI was only weakly correlated to EPO concentration, whereas hypoxaemia as measured by T90% was more strongly related to EPO concentration. This suggests that apnoeas or hypopnoeas per se are not sufficient to promote increased serum EPO concentration and that the magnitude of hypoxaemia due to CSA is a more important consideration (Figure 1). Notably, prior reports have not addressed whether the extent of oxygen desaturation associated with CSA is related to morbidity or mortality.^3,4,18,19 This may be relevant as in HF patients with CSA both the magnitude of the AHI elevation and the severity of associated hypoxaemia may vary greatly. Furthermore, by current diagnostic criteria, CSA may be present in individuals for whom arterial oxygen saturation does not fall below 95% despite a markedly elevated AHI. This situation may be similar to that reported for subjects with OSA for whom the magnitude of hypoxaemia is a risk factor for recurrent atrial fibrillation, while the AHI is not.²⁰

We observed evidence of possible EPO resistance among HF subjects as estimated by the ratio of log (EPO) to haemoglobin concentration. Prior studies have suggested that bone marrow resistance to the haematopoietic effect of EPO promotes elevated EPO concentration¹² and may be mediated by pro-inflammatory cytokines.^14,15,21 Moreover, hypoxaemia per se also appears to cause increased circulating concentrations of pro-inflammatory cytokines.²² Hence, while hypoxaemia may promote increased EPO, this may not necessarily lead to increased haemoglobin in HF patients due to opposing effects of cytokines on erythropoiesis. While a relationship between CSA and inflammation has yet to be reported, there is substantial evidence that hypoxaemia among healthy individuals and among patients with OSA is pro-inflammatory,²³ suggesting a potential mechanistic interaction between HF, CSA, systemic inflammation, and EPO resistance.

Pharmacotherapy for HF may have intra-renal or bone marrow actions that also directly or indirectly alter serum EPO concentration. Angiotensin converting enzyme inhibitor therapy reduces the response to EPO among dialysis patients²⁴ and in vitro evidence suggests that the angiotensin II receptor blocker losartan directly suppresses erythroid blast formation.²⁵ It has been suggested that diuretics which act primarily on the proximal tubule may inhibit EPO production mediated via an increase of local oxygen tension.²⁶ However, our study did not suggest a relationship between HF medication and serum EPO concentration.

Consequences of increased serum EPO appear to include increased release of endothelin-1,²⁷ enhanced tissue renin activity,²⁸ increased expression of receptors for endothelin,²⁹ and a direct pressor effect.³⁰ Accordingly, it appears increased serum EPO concentration may be either a marker or a mediator of increased cardiovascular risk and a potential pathophysiological link between hypoxaemia due to CSA and adverse prognosis in HF patients.

This study has several limitations which may account for the modest correlation coefficients, including the small sample size, limited number of EPO samples, and the biologic variability of EPO. Hence, future studies with greater numbers of observations may be required to confirm or refute this relationship. The frequency of CSA was higher in our study than in some prior reports.³ However, the study group was small and not representative of the broader HF population as nearly 38% of subjects in our study had advanced HF. As subjects were not anaemic and did not have severe renal dysfunction, the observations should not be extrapolated to HF patients with either anaemia or advanced renal failure. Patients with HF and OSA were excluded from the analysis to limit subjects with sleep disordered breathing to those with CSA, so these observations should not be generalized to HF patients with OSA. However, our group has previously shown that serum EPO is elevated in patients with OSA.¹⁰ Resistance to EPO was estimated by a simple ratio of log (EPO) to haemoglobin. Ideally, resistance would have been evaluated by measures of plasma volume and red cell mass which were beyond the scope of this study.

In conclusion, nocturnal hypoxaemia due to CSA appears to be associated with increased serum EPO concentration in HF patients. However, nocturnal hypoxaemia alone does not entirely account for the elevated serum EPO observed in advanced HF patients. Advanced HF and CSA-induced nocturnal hypoxaemia may each be independently associated with elevated serum EPO concentration.

Funding

This work was supported by the Mayo Foundation; American Heart Association (grant no. 04-50103Z); National Heart Lung and Blood Institute (grant numbers HL65176, HL70302, and HL73211); and the National Center for Research Resources (NCRR) (grant no. 1ULI RR024150), a component of the National Institutes of Health (NIH) and the NIH Roadmap for Medical Research. The contents of this report are solely the responsibility of the authors and do not necessarily represent the official view of NCRR or NIH.

Conflict of interest: L.J.O. has received research grants from the Medtronic Corporation and the Sorin Corporation. V.K.S. has served as a Consultant for ResMed, Cardiac Concepts, Sepracor, and Medtronic Corporation and has been a principal investigator or co-investigator on research grants funded by the Respironics Foundation, and the Sorin Corporation.