Introduction

Sleep apnea often complicates chronic heart failure (CHF) [1–3]. At least 50 % of patients with CHF due to left ventricular systolic dysfunction suffer from central, obstructive, or mixed sleep apnea. This nocturnal respiratory disorder increases morbidity and mortality, mainly by worsening heart failure. In patients suffering from CHF, noninvasive ventilation systems may be used to treat obstructive or central sleep apnea (CSA), often accompanied by Cheyne-Stokes respiration (CSR). While the effects of these ventilation systems on morbidity and mortality are not entirely clear, they have improved the prognosis of patients presenting with CHF and sleep apnea in several studies [2, 3]. Drugs used to treat CHF and cardiac resynchronization therapy are also at least partially effective in the treatment of CSA/CSR [4, 5].

Sleep apnea causes intermittent and repetitive nocturnal hypoxia, which might further worsen the cardiac performance of patients suffering from CHF [6]. By mitigating the severity of sleep apnea-related hypoxia, nocturnal oxygen therapy (NOT) also lowers the apnea-hypopnea index (AHI) in CSA/CSR [7–19]. However, these observations were made in small populations and, usually, on the short term [7–14]. Few published studies lasted >3 months, and the short-term effects of NOT have not been compared with its longer-term benefits [15–18]. Therefore, we studied the effects of a single night and of 6 months of NOT on sleep apnea, quality of life (QOL), and cardiac performance in patients presenting with CHF.

Methods

Among a sample ambulatory population of 380 patients followed-up for CHF in our cardiovascular department over a 4-year period, we studied 51 consecutive patients who presented with New York Heart Association (NYHA) heart failure functional classes II or III and an echocardiographic left ventricular ejection fraction (LVEF) ≤45 %, due to ischemic or non-ischemic cardiomyopathy. The diagnosis of CHF had been made ≥6 months before inclusion in the study, and all patients had remained clinically stable on an optimal drug regimen throughout that period. Patients who had a history of sleep apnea, chronic respiratory or renal disease, transient ischemic attack, stroke, or neurological disorder were excluded from the study. All patients received standard therapy and were followed in our institution. The investigational protocol complied with the principles outlined in the Declaration of Helsinki. This study was reviewed and approved by our institutional ethics committee, and all patients granted a written, informed consent to participate. Since the study began in March 2005 and its enrolment ended in August of year 2009, it was not entered in a clinical trials registry.

Study protocol

In the absence of dedicated sleep questionnaire for patients suffering from CHF, who are often free or nearly free from symptoms of sleep-disordered breathing, we systematically performed nocturnal recordings in our eligible patients, which served as screening and diagnostic tests simultaneously. All patients underwent baseline nocturnal cardiorespiratory polygraphy in our laboratory. When the recordings revealed the presence of AHI ≥15 events/h, sleep apnea was diagnosed, and the patient was randomly assigned to NOT versus no NOT. Randomization was performed by the cast of three dice; an odd sum of the dice assigned the patient to the no NOT group and an even sum to the NOT group. NOT, delivered through a nasal cannula at a flow rate of 3.0 l/min, was initiated in our sleep laboratory within 24 h after the baseline recording, along with a second nocturnal cardiorespiratory polygraphy. The patients assigned to active therapy were instructed to continue using NOT every night, using a concentrator that delivered oxygen at a rate of 3.0 l/min, from the moment they turned off the lights until they woke up in the morning. The concentrator models were chosen by two providers of home care who were unaware of our study protocol (NewLife Elite and QuietLife, AirSep Corp., USA; Companion 590 OCI, Nellcor Puritan Bennett, USA; Nuvo, Nidek Medical Products, Inc, USA). Compliance was monitored by patient interviews and by inspection of the built-in concentrator tachometer at 2 weeks, 3 months, and 6 months after the onset of NOT. Following 6 months of home therapy, the patients underwent a third nocturnal cardiorespiratory polygraphy during delivery of NOT. Patients randomly assigned to no NOT remained on room air, without home oxygen concentrator throughout the study period. They underwent a single baseline nocturnal cardiorespiratory polygraphy. Like the patients assigned to NOT, they underwent nocturnal cardiorespiratory polygraphy after 6 months of follow-up. All patients were seen in our ambulatory department at 2 weeks, 3 months, and 6 months for interim medical histories, physical examinations, and monitoring of adverse clinical events.

Nocturnal cardiorespiratory polygraphy

Overnight cardiorespiratory polygraphy signals were recorded using an Embletta® PDS system (Medcare Flaga, Reykjavik, Iceland). Our recording procedure and interpretation of measurements, by the same sleep medicine specialist who visually and anonymously scored all recordings, using standard criteria, have been described elsewhere [20, 21]. In patients assigned to NOT, the requirement of a ≥4 % dip in arterial oxyhemoglobin saturation (SaO2) to diagnose hypopnea seemed inappropriate since the dips were eliminated or mitigated by oxygen supplementation despite the presence of hypopnea. Therefore, hypopnea was also diagnosed when, on polygraphy recorded during NOT, the recordings of airflow and of breathing efforts were both similar to the baseline recordings of hypopneic events associated with ≥4 % dips in SaO2. On the baseline recordings, hypopnea was defined as a ≥30 % decrease in inspiratory oronasal airflow amplitude from the surrounding baseline, lasting ≥10 s, associated with a ≥4 % decrease in SaO2 from the preceding stable baseline and in the presence of thoracic and abdominal movements. In obstructive hypopnea, the breathing efforts increased in amplitude during the event, along with out-of-phase thoracic and abdominal movements and, perhaps, prolonged inspiration with flattening of the upper airflow waveform, consistent with airflow limitation. In central hypopnea, the breathing efforts were weak and associated with an in-phase decrease in thoracic and abdominal motion combined with a proportional decrease in airflow amplitude.

Measurements of quality of life and left ventricular function

All patients completed a Short Form 36 (SF-36) questionnaire, before undergoing baseline nocturnal cardiorespiratory polygraphy and at 6 months of follow-up, to measure their health-related QOL. They had undergone cardiac catheterization with coronary angiography, left ventriculography, and measurement of LVEF before entry into the study. A transthoracic echocardiogram was also recorded during the week before random study assignment, to verify that LVEF was ≤45 %, using an Acuson-128 sonograph (Siemens Medical Solutions USA Inc., Malvern, PA, USA) and the modified Simpson method. Baseline and 6-month echocardiograms were recorded on the days after the nocturnal polygraphies, and interpreted by the same physician, who was unaware of the patients’ identities. Venous blood samples were collected for standard screening tests on the same day as the baseline echocardiograms.

Statistical analyses

Parametric or nonparametric tests were chosen after examination of their conditions of use. Sidak adjustment was applied in each comparison to penalize the α risk. The aim was to account for the inflated α risk due to the multiple statistical tests performed on the same samples. Sidak adjustment was 1 − (1 − 0.05)1/c, where c was the number of tested variables. A P value was considered statistically significant when it was below the value calculated by the Sidak adjustment. The means ± SE and counts and percentages of baseline characteristics of the study groups were compared using t tests and chi-square tests, respectively, as appropriate. A P value <0.002 was considered statistically significant. In the group assigned to NOT, the baseline measurements were compared with the measurements made after the first night and after 6 months of NOT. In the group assigned to no NOT, the baseline measurements were compared with the measurements made at 6 months of follow-up. We used paired t tests or Wilcoxon tests and chi-square or Fisher’s exact tests, analysis of variance with repeated measures or Friedman analysis of variance and with post hoc multiple comparisons. A P value <0.002 was considered statistically significant.

AHI, oxygen desaturation index (ODI), SaO2 < 90 % time, and changes in LVEF between baseline and 6 months of the study in the two study groups were compared. Due to the small sample size and the variability of the measurements, these between-groups comparisons required standardization of the data. AHIb i and AHI6m i were, respectively, the AHI values for all individuals i at baseline and at 6 months of the study. For each individual i, we standardized the changes (Δ) in AHI between baseline and 6 months of the study as ΔAHI i = [AHI6m i − AHIb i ]/AHIb i . Means ± SE with 95 % confidence intervals of ΔAHI i in the NOT and no NOT groups were expressed as percentages. Thus, the variables ΔAHI, ΔODI, ΔSaO2 < 90 % time, and ΔLVEF were created, and each variable was compared between the study groups using t test. A P value <0.01 was considered statistically significant.

Results

Study population and patient subgroups

The characteristics of all patients and of the patients without versus with sleep apnea are shown in Table 1. After baseline nocturnal cardiorespiratory polygraphy, 33 of the 51 patients were diagnosed with sleep apnea. A greater proportion of men than women were included in the sleep apnea compared with the no sleep apnea group, which is concordant with the gender distribution usually observed with this disorder. Figure 1 shows the flow of patients between baseline and 6 months of follow-up. In the group assigned to NOT, after the first night and repetition of nocturnal polygraphy, two patients declined to use the concentrator at home on the long term. One of these patients agreed to undergo continuous positive airway pressure (CPAP) ventilation throughout the study. The other patient refused all treatments for sleep apnea. A third patient died from a ruptured abdominal aorta aneurysm, 2 weeks before the end of the study. Thirty patients completed the study, of whom 16 were assigned to NOT and 14 to no NOT. The characteristics of the two groups are presented in Table 1.

Table 1 Characteristics of entire study sample, patients without versus with sleep apnea (SA), and patients without versus with nocturnal oxygen therapy (NOT)
Full size table
Fig. 1
figure 1

Flow of patients between inclusion in the study and 6-month nocturnal cardiorespiratory polygraphy in both study groups. LVEF left ventricular ejection fraction, NYHA New York Heart Association heart failure functional class

Full size image

Nocturnal versus no nocturnal oxygen therapy

Table 2 shows the results of the baseline, first night, and 6 months of recordings made in the 16 patients who underwent NOT and the results of the baseline and 6 months of recordings made in the 14 patients assigned to no NOT. In the latter group, no significant difference was observed between the baseline and the 6-month polygraphic, SF-36 questionnaire and LVEF measurements. In contrast, compared with baseline, the mean AHI decreased by 44 % during the first night and by 50 % after 6 months of NOT, with a predominant 64 and 74 % decrease, respectively, in mean central AHI (P < 0.0001), while no significant change was observed in obstructive or mixed AHI (Table 2; Fig. 2). After the first night and after 6 months of NOT, 69 and 94 % of patients, respectively, had a ≥50 % decrease in central AHI as well as <15 events/h. Likewise, both ODI and the duration of SaO2 < 90 % decreased markedly. No significant difference was observed in any of the polygraphic measurements between the first night of NOT and 6 months of NOT. NYHA heart failure functional class, LVEF, and QOL ascertained by the SF-36 questionnaire did not change significantly between baseline and the 6 months of follow-up in either study group. We observed changes in neither the overall nor the subcategories scores of the SF-36 questionnaire. Figure 3 compares the changes between baseline and the end of the study in AHI, ODI, SaO2 < 90 % time, and LVEF between NOT and no NOT patients. Only AHI and ODI changed significantly in the NOT, compared with the no NOT group.

Table 2 Within-groups comparisons of baseline versus follow-up, nonstandardised polygraphic and other selected variables in both study groups
Full size table
Fig. 2
figure 2

Example of nocturnal cardiorespiratory polygraphy before and during nocturnal oxygen therapy (NOT). a Baseline recording. Mixed apneic events of 49 and 39 s duration associated with dips in oxygen saturation from 97 to 74 % and from 97 to 78 %, respectively. The ECG shows acceleration of ventricular rate upon resumption of breathing. b During delivery of NOT in first night of recording in the same patient. Obstructive apnea of 61 s without hypoxia and oxygen desaturation, despite a prolonged apnea. Top to bottom tracings airflow with nasal cannula, airflow with oronasal thermistor, thoracic motion, abdominal motion, oxygen saturation, heart rate with oximeter, and ECG

Full size image
Fig. 3
figure 3

Changes (Δ) between baseline and 6 months of study in 16 recipients of nocturnal oxygen therapy (NOT) versus 14 patients who did not receive NOT. a AHI Apnea-hypopnea index, b ODI oxygen desaturation index, c SaO 2   < 90 % arterial oxyhemoglobin saturation < 90 %, d LVEF left ventricular ejection fraction. ΔAHI, ΔODI, ΔSaO2 < 90 % time and ΔLVEF were new variables expressed as percentages of means ± SE with 95 % confidence interval. A P value <0.01 was considered statistically significant

Full size image

Delivery and tolerance of nocturnal oxygen therapy

NOT was delivered for 194 ± 21 nights, representing 8.3 ± 1.4 h (range, 6.3–10.1) of oxygen delivery per night throughout the 6 months of follow-up. No patient needed interruption or cessation of NOT because of intolerance or adverse clinical event.

Discussion

Published studies of a single night of NOT in patients suffering from CHF and sleep apnea have reported a nearly 50 % average decrease in AHI [7–12]. The AHI decrease was 52 and 53 % in the CHF-HOT Study Group after 3 and 12 months of NOT, respectively [16, 18]. If, as in these studies, we consider only central AHI in our study sample, the decrease in AHI was 64 % during the first night and 74 % after 6 months of NOT. We used the most common dose of oxygen delivered in other studies, with similar indications for NOT, making our study comparable with these prior publications. We observed no significant difference in any variable between the first night and 6 months of NOT, suggesting that (a) the initial response to NOT was sustained over time and (b) a majority of long-term responders to NOT could be detected as early as the first night of oxygen supplementation. The absence of effects conferred by NOT on obstructive events in our patients confirms previously published observations [9, 10, 14, 16, 18]. It is likely that few patients suffered from obstructive respiratory events promoted by a depressed cardiac performance, and that obstruction of the upper airways during sleep was present before the development of CHF. The similar response to NOT of the mixed and obstructive respiratory events is noteworthy, particularly when the mixed events had a prolonged or predominantly central component [9]. This surprising observation remains unexplained.

Patients in stable CHF have a normal SaO2 and blood CO2 concentration during daytime or while awake, and do not require oxygen therapy [22, 23]. Hypocapnia can be present during sleep in patients with CHF and becomes a main promoter of nocturnal CSA/CSR, in which case hypocapnia can be also observed during daytime. In addition, CHF patients with CSA/CSR may have disturbances of the oxygen and carbon dioxide chemoreceptors and an enhanced sensitivity to CO2. During sleep, periods of hypoventilation and increased arterial PaCO2 may promote an excessive response with marked hyperventilation, resulting in hypocapnia and CSA/CSR. Phases of hyperventilation related to a supine position and sleep have also been described in patients with CHF that may promote hypocapnia [24]. Repetitive episodes of hypoxia-rapid re-oxygenation might cause myocardial reperfusion injury from oxidative stress [25, 26]. These repetitive events are, perhaps, one of the mechanisms in sleep apnea that promote the progression of CHF [6]. Oxygen supplementation may stabilize the respiratory control system by lowering the frequency of oxygen desaturation-reoxygenation episodes and ventilatory drive hypoxic stress. NOT might also mitigate the disturbances of oxygen and carbon dioxide chemoreceptors [7]. It might also attenuate the baseline hyperventilation or hyperventilation that occurs in response to hypoxia and/or hypercapnia due to hypopnea or apnea. In addition, oxygen supplementation promotes the formation of oxyhemoglobin instead of carboxyhemoglobin, which increases the arterial PaCO2 [9, 12]. Changes in PaCO2 as small as 2 mmHg by administration of oxygen may markedly alleviate CSA/CSR [18, 27, 28]. In this study, we observed a significant decrease in ODI in the NOT compared to the no NOT group between baseline and 6 months, though not in the duration of SaO2 < 90 %, probably because of the wide variability in this measurement in the no NOT group and because of the small sample size. With nocturnal ventilation systems, decreases in ODI and SaO2 < 90 % are consequences of a decrease in the apnea-hypopnea load; while they are direct effects and signs of compliance with NOT, they are not necessarily associated with a decrease in AHI.

NOT had no significant effect on LVEF over the 6 months of the study. Likewise, Sasayama et al. observed no significant increase in LVEF in recipients of NOT, up to 12 months, compared to a control group [16, 18]. A significant increase in LVEF, from a mean of 27 to 37 %, was reported only by Toyama et al. after 12 weeks of NOT, measured by nuclear imaging in subgroups of the CHF-HOT Study Group [17]. The effects of CPAP, adaptative servoventilation (ASV), and NOT on LVEF remain uncertain [4, 15–19], perhaps because of small sample sizes, variable study durations, or absence of between-groups comparisons of treatment methods. If one assumes that NOT decreases central AHI partly via its effects on the heart, it might be worthwhile to distinguish its impact on nocturnal as opposed to diurnal cardiac function. In sleep apnea, the decrease in cardiac performance might be due to the simultaneity of hypoxia and sympathetic hyperactivity [29]. Repetitive bursts of central sympathetic activity occur upon arousal at the end of obstructive apneas. These sympathetic bursts may also be caused by the hyperventilation phase during CSA/CSR, with an increasing difference between positive and negative intrathoracic pressure from one ventilatory cycle to another [14, 29]. Thus, myocardial oxygen deprivation due to apnea-related oxygen desaturation is concomitant with the excessive myocardial oxygen consumption due to sympathetic bursts that increase the cardiac work load [6]. In patients suffering from CHF, NOT appears to decrease the sympathetic nervous activity related to sleep apnea, probably by eliminating intermittent hypoxia [14, 16, 30]. The absence of statistically significant effects on LVEF might then be explained by the fact that benefits of NOT may not be related to a cardiac function improvement but more to an avoidance of its nighttime deterioration. This would also explain the acute nocturnal beneficial effects it confers on CSA/CSR. Indeed, as soon as NOT is delivered, intermittent hypoxia is eliminated, mitigating the bursts of sympathetic activity and their adverse cardiovascular consequences. Cardiac resynchronization therapy can improve daytime cardiac function; however, several weeks are often required to identify the responders and who ultimately benefits from an alleviation of CSA/CSR [5].

Sasayama et al. reported a significant decrease in mean NYHA heart failure functional class with NOT compared with a control group [16, 18]. The absence of clinical advantage conferred by the positive effects of NOT on the breathing patterns in our study has several potential explanations. First, our patients were not recruited on the basis of manifestations of sleep apnea. Instead, they were consecutive patients presenting with stable CHF who underwent systematic recording of nocturnal cardiorespiratory polygraphy, which might have decreased the likelihood of observing a clinically significant effect of NOT, even when AHI was decreased. Second, patients suffering from CHF and CSA/CSR often have no or only mild sleep apnea-related symptoms as opposed to patients with the usual form of OSA and their QOL is rarely worse than that of patients presenting with CHF alone [31]. This might have further limited the opportunity to observe a clinical impact of NOT in these patients. Third, the SF-36 questionnaire was ill suited for this study, given the duplication and potential overlap of symptoms related to sleep apnea and to CHF. Fourth, there was no significant increase in LVEF in the study. However, CPAP and ASV could show an increase in exercise capacity despite the absence of evident increase in cardiac performance [13, 15]. Finally, our study sample was too small. Even CPAP and ASV in patients with CHF and CSA/CSR did not have a clear impact on the clinical status, despite a decrease in AHI [8, 13]. Despite the absence of improvements conferred by NOT on a clinical level, our observations of a positive effect on respiratory parameters are noteworthy. AHI levels and the prognosis of patients suffering from CHF and sleep apnea seem correlated, and decreasing AHI improved these patients’ outcomes [2, 3]. In particular, NOT might be beneficial in CHF by alleviating the hypoxic load and by decreasing the episodes of intermittent desaturation and sympathetic nerve hyperactivity [6, 32]. BNP was not measured because we focused our study on changes in cardiorespiratory measurements from nocturnal polygraphy and, with respect to myocardial function, we favoured the measurement of LVEF. BNP was not reliable as a means of observing the evolution of heart failure in studies of sleep-disordered breathing [20]. Staniforth et al. noticed no changes in BNP after 1 month of NOT in patients with CSA/CSR and CHF [14]. In the CHF-HOT Study Group, BNP was not reported at 3 months [16, 17] and no changes were observed between baseline and 1 year of NOT [18].

Two recent studies confirmed that whether they suffered from obstructive or central type of sleep-disordered breathing, patients presenting with CHF had a worse prognosis than patients who did not suffer from severe (AHI <20–22 events/h) sleep apnea [2, 3]. Furthermore, these studies found that nocturnal noninvasive ventilation, CPAP, bilevel PAP, and ASV improved the outcome of patients with severe sleep apnea. SERVE-HF is a first ongoing, international, multicentre, randomized trial to examine the effects of ASV, in particular, on the prognosis of patients with CHF and CSA/CSR [33]. Besides the value of ASV for this indication, SERVE-HF might shed light on the overall merit, or not, to treat CSA/CSR in patients suffering from CHF. In studies comparing CPAP with NOT, a similar decrease in central AHI was observed in these patients [11, 12, 15]. When compared with CPAP, NOT, or both, ASV seemed more effective in decreasing central AHI [12, 19]. In our study, compliance with, and tolerance of NOT were very high, as was observed in a recent cross-over comparison of ASV versus NOT, each used for 2 months in seven patients [19]. At the end of the study, six patients preferred to continue using NOT, versus a single patient who chose ASV. The cost/benefit ratio of NOT was estimated from the first CHF-HOT study [16, 34]. Seino et al. studied patients treated with NOT for the 12 weeks of the protocol and continued NOT for 6 months, and controls who were treated with NOT for 6 months after the end of the protocol. Thus, in 53 patients, the estimated saving was ¥1,854,175 per patient per year (¥1 ≈ $0.01) from fewer and shorter hospitalizations, representing 67 days per patient per year free from hospital care, emergency visits, and scheduled ambulatory visits. The reported cost of NOT was ¥854,400 per patient per year.

Study limitations

Polygraphy is less accurate than polysomnography in the measurement of AHI and provides no information regarding the effects of NOT on the architecture and fragmentation of sleep. From a study design standpoint, the use of a second polygraphy 24 h after randomization would have been appropriate in the no NOT group. However, since subsequent studies revealed minimal day-to-day variability of polygraphic measurements in patients suffering from stable CHF [16–18, 35], the absence of that second polygraphy was not a major design flaw. The NOT group benefited from two interventions, including the presence of a concentrator at home and oxygen supplementation, whereas the no NOT group benefited from neither intervention. The absence of a concentrator supplying room air at home in the control group (sham oxygen therapy) created a bias in our study, which was due to our lack of resources to provide the modified concentrators to supply air throughout the study duration. The home compliance with NOT could not be reliably monitored, and we had to rely on the patients’ reports. The measurements made by the devices are shown in the “Results” section, though we could not directly verify that the nasal cannulas were reset every night throughout the protocol.

Conclusions

In our sample of patients suffering from CHF and sleep apnea, NOT significantly decreased central AHI and ODI within 24 h and for up to 6 months without significantly changing obstructive or mixed AHI, QOL, and LVEF. NOT might be an alternate to nocturnal ventilation systems for patients suffering from CHF and CSA/CSR, notably for its sustained effects on AHI and ODI, its high tolerability, and relatively low cost. The impact of NOT on morbidity and mortality in these patients remains to be determined.