Time course of continuous positive airway pressure effects on central sleep apnoea in patients with chronic heart failure


Michael Arzt, Department of Internal Medicine II, Pneumology, University of Regensburg, Franz-Josef-Strauß-Allee 11, 93042 Regensburg, Germany. Tel.: 0049-941-944-7281; fax: 0049-941-944-7282; e-mail: michael.arzt@klinik.uni-regensburg.de


Continuous positive airway pressure (CPAP) causes a variable immediate reduction in the frequency of central apnoeas and hypopnoeas in patients with congestive heart failure (CHF) and central sleep apnoea (CSA), but has beneficial mid-term effects on factors known to destabilize the ventilatory control system. We, therefore, tested whether CPAP therapy leads, in addition to its short-term effects on CSA, to a significant further alleviation of CSA after 12 weeks of treatment on the same CPAP level in such patients. CPAP therapy was initiated in 10 CHF patients with CSA. During the first night on CPAP, the pressure was stepwise increased to a target pressure of 8–12 cmH2O or the highest level the patients tolerated (<12 cmH2O). Throughout the second night (baseline CPAP), the achieved CPAP of the first night was applied. After 12 weeks of CPAP treatment, we performed a follow-up polysomnography (12 weeks CPAP) on the same CPAP level (8.6 ± 1.1 cmH20). We found a significant reduction of the apnoea-hypopnoea index (AHI) between the diagnostic polysomnography and baseline CPAP night (41.8 ± 19.2 versus 22.2 ± 12.6 events per hour; P = 0.005). The AHI further significantly decreased between the baseline CPAP night and the 12 weeks CPAP night on the same CPAP level (22.2 ± 12.6 versus 12.8 ± 11.0 events per hour; P = 0.028). We conclude that, in addition to its immediate effects, CPAP therapy leads to a time-dependent alleviation of CSA in some CHF patients, indicating that in such patients neither clinical nor scientific decisions should be based on a short-term trial of CPAP.


Cheyne Stokes respiration with central sleep apnoea (CSA) is characterized by a typical periodic waxing and waning breathing pattern affecting approximately 25–40% of patients with congestive heart failure (CHF) (Javaheri et al., 1998; Sin et al., 1999; Solin et al., 1999).

In contrast to obstructive sleep apnoea, CSA is characterized by a lack of respiratory effort during cessation of airflow. In CHF patients, a number of factors contribute to instability of ventilatory control system that favours the development of periodic breathing with central apnoeas and hypopnoeas during sleep. One such factor is hypocapnia resulting from the combined effects of elevated left ventricular filling pressures and lung oedema that stimulate pulmonary vagal irritant receptors (Lorenzi-Filho et al., 1999, 2002; Solin et al., 1999), increased ventilatory responsiveness to chemical respiratory stimuli (Arzt et al., 2003; Naughton et al., 1993), and decreased lung volume with reduced oxygen stores that cause more rapid oxygen desaturation for a given apnoea length that augments ventilatory overshoot at apnoea termination (Krachman et al., 2003). Another factor contributing to instability is recurrent arousals from sleep that trigger abrupt increases in ventilation that facilitate reductions in PCO2 below the apnoeic threshold (Naughton et al., 1993).

Nocturnal continuous positive airway pressure (CPAP) is the most extensively tested treatment for CSA in CHF. However, its reported effects on CSA in CHF patients are inconsistent. In randomised short-term trials in which nocturnal CPAP was applied for one night to 2 weeks only, CSA was not significantly alleviated (Buckle et al., 1992; Davies et al., 1993) or reduced by less than 50% (Teschler et al., 2001). In contrast, where patients were acclimatized to CPAP during a gradual 2- to 7-day titration protocol and treated for a period of 2–12 weeks, the frequency of central apnoeas and hypopnoeas fell by 50–65% (Arzt et al., 2005; Bradley et al., 2005; Krachman et al., 1999; Naughton et al., 1994). Moreover, treatment of CSA with CPAP for a period of 2 weeks or longer has been demonstrated to alter factors linked to the pathogenesis of CSA in favour of stabilizing the ventilatory control system. For example, CPAP increases total body oxygen stores (Krachman et al., 2003) and nocturnal transcutaneous PCO2 (Naughton et al., 1994), reduces ventilatory drive (Arzt et al., 2005) and improves cardiac function (Arzt et al., 2005; Bradley et al., 2005; Naughton et al., 1995) in patients with CHF and CSA after 2–12 weeks of therapy. Thus, it appears that improvement of CSA in response to CPAP occurs gradually, but the time course of this effect remains to be determined.

We therefore tested the hypothesis that in CHF patients with CSA, CPAP causes progressive alleviation of CSA between the second night of its application, and 12 weeks later on the same pressure level. To this end, we performed a secondary analysis of polysomnographic data on such patients enrolled in the CPAP-treated arm of our previously published trial on the effects of O2 and CPAP on cardiovascular function in CHF patients with CSA (Arzt et al., 2005).



Inclusion criteria were CHF due to ischemic, hypertensive or idiopathic dilated cardiomyopathy with a left ventricular ejection fraction (LVEF) <45% as determined by resting echocardiography according to the guidelines of the American Society of Echocardiography (Cheitlin et al., 2003) by a single cardiologist who was blinded to the clinical data. Furthermore, patients [New York Heart Association (NYHA) functional class II and III] were stable over the previous 3 months as documented by stable cardiac medication and no hospital admissions. CHF patients with CSA as defined by an apnoea-hypopnoea index (AHI) ≥ 15 episodes per hour of sleep and central apnoeas > 75% of total apnoeas, diagnosed by in-laboratory polysomnography (Brainlab, Schwarzer, Germany), and patients who received during the second night on CPAP and 12 weeks later, the same pressure level were included. Exclusion criteria were a history of unstable angina, cardiac surgery or documented myocardial infarction within 90 days of entry into the study.


During polysomnography body position, eye and leg movements, cardiotachography, nasobuccal airflow, chest and abdominal effort, and arterial oxyhaemoglobin saturation assessed by pulse oximetry were recorded (Brainlab, Schwarzer, Germany). Sleep stages were determined using standard criteria (Rechtschaffen and Kales, 1968). Apnoeas and hypopnoeas were scored as follows. Apnoea was defined as a cessation of inspiratory airflow for ≥10 s. Central apnoeas were those that occurred with an absence of rib cage and abdominal motion; while obstructive apnoeas were those that occurred in the presence of out-of-phase rib cage and abdominal motion. Hypopnoea was defined as a more than 50% reduction of airflow or thoraco-abdominal excursions lasting at least 10 s resulting in a ≥4% drop in arterial oxyhaemoglobin saturation (SaO2). The oxygen desaturation-index was defined as the number of ≥4% oxygen desaturations per hour of sleep.

Protocol and intervention

A total of 134 consecutive ambulatory patients with CHF were screened for CSA by a portable computerized sleep diagnosis system (SOMNOcheck effort; Weinmann, Hamburg, Germany). In 39 CHF patients, the results from the screening device indicated significant CSA. Of these, four were excluded because they had a LVEF > 45% at their baseline study echocardiography. In three patients, CSA was not confirmed by polysomnography. An additional six patients refused therapy.

Of the reminder 26 patients, the first 10 (depending on the date of screening) received nocturnal nasal oxygen treatment as previously described (Arzt et al., 2005), and for the purposes of this substudy, were excluded from further consideration. In the remaining 16 patients, CPAP was initiated, of whom the first 10 consecutive patients received the same pressure level during the second night on CPAP and 12 weeks later as described below. As described previously, in two subjects, who did not use CPAP and withdrew, follow up data were not available (Arzt et al., 2005). All patients gave written informed consent prior to participation in this study, which had been previously approved by the ethics committee of our institution.

All patients underwent polysomnography, spirometry and echocardiography at baseline. Following this baseline assessment, patients allocated to CPAP were brought into the sleep laboratory to initiate it at a target pressure of 8–12 cmH2O. To acclimatize patients CPAP (Somnocomfort; Weinmann) was applied to patients at 4 cmH2O for 1 h while awake. During the first night, CPAP was started at 4 cmH2O, and was slowly increased in 1–2.5 cmH2O increments to reach the target pressure or the highest level the patient could tolerate (<12 cmH2O). Polysomnographic data of the titration night are not reported, because most patients were on subtherapeutic CPAP for >50% of time. On the second night, CPAP was applied with the achieved target pressure or the highest tolerated level of the first night (baseline CPAP). Patients were sent home at this CPAP level and were instructed to use CPAP for at least 6 h per night. The mean nightly usage was calculated by built-in CPAP hour meters. Twelve weeks later, polysomnography was performed on the same CPAP pressure as applied during the second night of CPAP initiation. Medications were kept constant throughout the study period.

Statistical analysis

All data were analyzed by a computer program (spss 11.0; SPSS Inc., Chicago, IL, USA). Data from the baseline assessment, second night of CPAP initiation, and 12 weeks later were compared with the Friedmann test. Differences between two time points were assessed using the Wilcoxon signed-rank test. A two-sided P-value of <0.05 was considered to indicate statistical significance.



The baseline characteristics of the 10 patients, who received the same pressure level during the second night on CPAP and 12 weeks later, are shown in Table 1. They were overweight and CHF was a consequence of ischemic heart disease in all cases. Two patients had chronic atrial fibrillation. All patients were on stable cardiac medication including angiotensin-converting enzyme (ACE) inhibitors, diuretics and β-blockers (Table 1). Severe restrictive or obstructive lung disease was ruled out in all cases by spirometry. Baseline polysomnographic data demonstrated that patients had moderate to severe CSA (Table 2 and Fig. 1) with 94 ± 9% of apnoeas central in origin.

Table 1.   Baseline characteristics of heart failure patients with central sleep apnoea
Variable(n = 10)
  1. Continuous variables are expressed as mean ± SD.

Age, year64 ± 8
Body-mass index, kg m−228.3 ± 4.4
Cause of heart failure, n (%)
 Ischemic10 (100)
 Sinus rhythm8 (80)
 Atrial fibrillation2 (20)
Medications, n (%)
 ACE inhibitors9 (90)
 Diuretics8 (80)
 Digoxin4 (40)
 β-Blocker8 (80)
 Forced expiratory volume in one second, % predicted92.0 ± 18.8
 Vital capacity, % predicted87.3 ± 15.7
Exercise capacity and cardiac function
 NYHA functional class, n (%)
  II9 (90)
  III1 (10)
 Peak oxygen uptake, mL kg−1 min−117.5 ± 4.2
 Left ventricular ejection fraction, %34.4 ± 7.8
Table 2.   Effects of short-term and mid-term CPAP
VariableBaselineBaseline CPAP12 weeks of CPAPP-value*P-valueP-value††
  1. CPAP, continuous positive airway pressure; AHI, apnoea-hypopnoea index.

  2. Continuous variables are expressed as mean ± SD.

  3. *Comparison between the baseline, baseline CPAP and 12 weeks of CPAP study by the Friedman-test.

  4. Comparison between the baseline and baseline CPAP study by the Wilcoxon signed-rank test.

  5. ††Comparison between the baseline CPAP and 12 weeks of CPAP study by the Wilcoxon signed-rank test.

Sleep characteristics
 AHI, no. per hour41.8 ± 19.222.2 ± 12.612.8 ± 11.0<0.0010.0050.028
 Apnoea index, no. per hour16.6 ± 18.46.2 ± 5.13.5 ± 3.30.0030.0470.139
 Hypopnoea index, no. per hour25.2 ± 15.615.9 ± 10.29.3 ± 8.40.112
 Oxygen desaturation index, no. per hour35.3 ± 24.36.6 ± 7.34.5 ± 6.4<0.0010.0050.221
 Mean SaO2, %93.8 ± 0.793.8 ± 0.694.1 ± 1.00.690
 Lowest SaO2, %82.4 ± 5.885.4 ± 5.489.2 ± 3.30.275
 Arousal index, no. per hour16.9 ± 11.216.2 ± 8.412.3 ± 6.40.407
 Sleep efficiency, %85 ± 887 ± 1184 ± 150.905
CPAP characteristics
 Hours of use per day4.8 ± 1.6
 CPAP pressure, cmH2O8.6 ± 1.18.6 ± 1.11.0
Figure 1.

 Comparison of the apnoea-hypopnoea index (AHI) (number of episodes per hour) between the baseline diagnostic sleep study (baseline), the second night on continuous positive airway pressure (CPAP) applying stable optimal pressure throughout the night (baseline CPAP) and a follow-up night after a 12-week treatment period on the same CPAP level (12 weeks CPAP). Individual data for all 10 heart failure patients with central sleep apnoea (CSA) are given. The dashed line marks an AHI of 15 events per hour defining the presence of CSA in this clinical evaluation. Data at baseline, at the baseline CPAP and the 12 weeks of CPAP study were compared with the Friedmann test. Differences between two time points were assessed using the Wilcoxon signed-rank test.

Effects of CPAP on CSA

Effects of CPAP after two nights

Throughout the second night on CPAP (baseline CPAP) the optimal CPAP of the first night was applied (Table 2). Compared with the baseline polysomnography, AHI and the oxygen desaturation index were significantly reduced by 47% (P = 0.005, Fig. 1) and 81% (P = 0.005), respectively (Table 2). While the minimal SaO2 was higher during the baseline CPAP study compared with the baseline polysomnography (P = 0.050), the mean SaO2, the arousal index and the sleep efficiency remained unchanged.

Effects of CPAP after 12 weeks

During the 12-week trial period, patients’ average CPAP usage was 4.8 ± 1.6 h per night. Compared with the second CPAP night, the AHI decreased a further 42% without any further adjustments of CPAP pressure (P = 0.028; Table 2, Fig. 1). There were no significant changes in oxygenation or sleep fragmentation between the second night of CPAP initiation and 12 weeks later.

There was no significant difference in baseline characteristics, CPAP use and changes of cardiac function and exercise capacity of subjects whose AHI was or was not suppressed to below 15 events per hour (CPAP-CSA suppressed and CPAP-CSA unsuppressed groups, respectively) of the entire sample (n = 14) of the primary analysis (Arzt et al., 2005) (Table 3).

Table 3.   Characteristics of patients with CHF and CSA with and without suppression of AHI to less than 15 events per hour
CharacteristicsCPAP-CSA suppressed (n = 10)CPAP-CSA unsuppressed (n = 4)Pvalue
  1. CHF, congestive heart failure; CSA, central sleep apnoea; AHI, apnoea-hypopnoea index; LVEF, left ventricular ejection fraction.

  2. Continuous variables are expressed as mean ± SD.

Age, year64 ± 764 ± 110.95
Body-mass index28 ± 631 ± 70.37
AHI, no. per hour34 ± 1440 ± 180.59
Exercise capacity and cardiac function
 Peak VO2, mL kg−1 min−117.3 ± 4.113.5 ± 3.00.13
 VE/VCO2-slope30.4 ± 6.132.6 ± 3.00.52
 LVEF, %33 ± 1029 ± 100.56
Daytime arteriocapillary blood gases
 Resting Pa CO2, mmHg38.1 ± 3.440.5 ± 5.00.38
 Peak exercise Pa CO2, mmHg38.2 ± 3.340.5 ± 5.00.36
CPAP characteristics
 CPAP, cmH2O8.5 ± 1.28.8 ± 1.00.61
 CPAP usage, hours per day4.8 ± 1.64.7 ± 2.40.82
Change after 12 weeks CPAP treatment
 Δ AHI, no. per hour−29 ± 15−11 ± 150.06
 Δ Peak VO2, mL kg−1 min−1+0.2 ± 1.7−0.8 ± 3.80.49
 Δ VE/VCO2-slope−4.7 ± 6.5−5.7 ± 6.30.80
 Δ LVEF, %+3.8 ± 6.2+4.5 ± 70.86
 Δ Resting Pa CO2, mmHg−0.4 ± 4.4−2.9 ± 3.60.33
 Δ Peak exercise Pa CO2, mmHg+1.9 ± 1.5+1.2 ± 1.60.46


This study demonstrates that, in addition to causing an initial attenuation of CSA after two nights, CPAP therapy leads to a further suppression of CSA in some CHF patients without any further adjustment of the CPAP level over the subsequent 12 weeks. Whereas, CPAP causes immediate and complete suppression of obstructive sleep apnoea in patients with CHF (Tkacova et al., 1998), its effects on CSA are gradual and time dependent. Thus, the approach to CPAP initiation for CSA in patients with CHF must take this time-dependent effect into account, and should differ from that used to initiate CPAP in patients with obstructive sleep apnoea.

The mean reduction of apnoeas and hypopnoeas by 47% during the second night on CPAP and by 69% after 12 weeks of treatment are comparable with previous single centre trials using similar treatment periods (Naughton et al., 1995; Philippe et al., 2006). Krachman et al. (2003) demonstrated an AHI reduction of 66% after approximately 2 weeks of CPAP therapy, indicating that CPAP may exert its maximum effect on CSA earlier than 12 weeks. However, to our knowledge our clinical evaluation is the first that examines CPAP effects on CSA at two time points on the same CPAP level throughout the night and during the treatment period in between: During the 12-week CPAP study, the AHI further significantly decreased by 42% compared with the second night on CPAP.

As only 10 of the 14 patients in the CPAP arm analysed in the original study (Arzt et al., 2005) had polysomnography on the optimal pressure during the entire second night on CPAP, and were, therefore, eligible for the present study, this may have conferred a selection bias. However, we consider false-positive results because of this potential bias as unlikely, because the applied CPAP, the daily CPAP use and the reduction of AHI after 12 weeks of CPAP of the entire group were similar compared with the present subsample (8.7 versus 8.6 cmH2O, 4.8 versus 4.8 h daily and 70% versus 69%, respectively).

In the full sample (Arzt et al., 2005), four of 14 subjects had an AHI that remained ≥15 per hour (CPAP-CSA-non-suppressed) after 12 weeks of CPAP treatment. Similar to the data derived from the secondary analysis of the Canadian continuous positive airway pressure for patients with central sleep apnea and heart failure (CANPAP) trial (Arzt et al., 2007), the comparison of the baseline characteristics (demographics, severity of CHF and CSA) between the CPAP-CSA suppressed and the CPAP-CSA unsuppressed groups does not clearly identify a subgroup of CHF patients whose CSA cannot be fully suppressed by CPAP. In contrast to the data from the secondary analysis of the CANPAP trial (Arzt et al., 2007), improvement of LVEF after 12 weeks of CPAP treatment was not significantly greater in the CPAP-CSA suppressed than in the CPAP-CSA unsuppressed group. However, because in the present study echocardiographic assessment of LVEF was only preformed in a small sample of four subjects whose CSA was not suppressed by CPAP, they do not allow firm conclusions with respect to effects of CPAP on cardiac function. Therefore, these results may not be comparable with those of the secondary analysis of the CANPAP trial in which LVEF was assessed by a different technique (radionuclide angiography) in a much larger number of subjects (n = 43) in whom CPAP did not suppress CSA.

Although the present clinical evaluation is limited by its small sample size and the inability to provide a mechanism for the time-dependent effect of CPAP on CSA, the individual data (Fig. 1) clearly show that in patients with CHF, it is not possible to judge the effects of CPAP on CSA after the second night of treatment (baseline CPAP). There were five patients, whose CSA was suppressed below 15 apnoeas and hypopnoeas per hour at 12 weeks, but not at the baseline CPAP study. In one patient, suppression of CSA to an AHI below 15 was not sustained after 12 weeks of treatment.

We believe that our case series provides a clinically important finding: (i) a stratified analysis of the CANPAP trial database (Arzt et al., 2007) suggested that the ability of CPAP to suppress CSA predicts improved cardiovascular outcomes of CHF patients suggesting that lowering the frequency of central apnoeas and hypopnoeas may play a key role to improve cardiovascular outcomes in CHF patients. In this context, it is crucial to know the time point when CPAP exerts its maximum effect on CSA in order to prevent a premature escalation of therapy to other forms of positive airway pressure support. (ii) In addition, our case series illustrates that, in contrast to treatment initiation in obstructive sleep apnoea, it is not possible to titrate CPAP within one night in order to immediately suppress CSA, because CPAP requires a longer period of time to exert its maximum effect on CSA. Keeping this in mind, inadequately high CPAP levels with poor effect on CSA and potentially harmful effects on hemodynamics in CHF patients can be prevented.

In summary our data demonstrate that, in addition to its immediate effects, CPAP therapy can lead after its second day of use to further alleviation of CSA in some CHF patients. We conclude that our finding strongly supports that in CHF patients with CSA neither scientific nor clinical decisions (e.g. whether CPAP should be continued or another form of positive airway pressure should be initiated) should be based on the short-term effects of CPAP on CSA. In view of the current literature, a follow-up sleep study after a trial of 2–4 weeks on CPAP may be the most appropriate means to judge the efficacy of CPAP on CSA (Arzt et al., 2007; Bradley et al., 2005; Krachman et al., 2003; Naughton et al., 1994).