INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Approximately 40% of patients with cardiac failure have central sleep apnea and/or Cheyne-Stokes respiration (CSA/CSR) during sleep (1, 2). CSA/CSR is associated with severe sleep disruption (1, 3) and decreased survival compared with subjects with a comparable ejection fraction but no CSA/CSR (6, 8). On the assumption that CSA/CSR is not just a marker for heart disease, but is worth treating in its own right, to improve sleep quality and possibly improve cardiovascular function, various long-term respiratory interventions during sleep have been tried, including nasal oxygen (3, 5, 9), nasal continuous positive airway pressure (CPAP) (12), and bilevel ventilation (17).

Adaptive servo-ventilation (ASV) is a new approach to the treatment of CSA/CSR, in which a small but varying amount of ventilatory support is provided. The intention is to provide the hydrostatic benefits of low levels of nasal CPAP (14) while directly suppressing CSA/CSR and attendant sleep disturbance without causing overventilation. The purpose of the present study was to compare the acute effects of ASV on quality of sleep and breathing with nasal oxygen, nasal CPAP, and bilevel spontaneous plus timed (ST) mode nasal ventilation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Principles of Operation of Adaptive Servo-ventilation

The adaptive servo-ventilator (AutoSet CS; ResMed, Sydney, Australia) provides a baseline degree of ventilatory support (pressure swing 4 cm H₂O) superimposed on 5 cm H₂O CPAP. End-inspiratory pressure is 9 cm H₂O, and mean pressure is approximately 7 cm H₂O. A typical pressure-versus-time waveform during therapy can be seen in the second last tracing of Figure 1B. The subject's ventilation is servo-controlled with a high-gain integral controller (0.3 cm H₂O per L/min per second, clipped to 4 to 10 cm H₂O) to equal a moving target ventilation of 90% of the long-term average ventilation (time constant 3 min). If the subject suddenly ceases all central respiratory effort, machine support (i.e., pressure swing amplitude) will increase from the minimum of 4 cm H₂O up to whatever is required to maintain ventilation at 90% of the long-term average (up to a maximum of 10 cm H₂O, reached in approximately 12 s). An example is shown at the start of the second minute of Figure 1B. If the subject then resumes normal spontaneous effort, support will fall back to the minimum of 4 cm H₂O over a similar time period. Smaller or slower changes in patient effort will result in proportionally smaller, slower changes in the degree of support. In the steady state, ventilation exceeds the 90% target, so support stays at the minimum of 4 cm H₂O.

View larger version (27K): [in this window] [in a new window]   View larger version (42K): [in this window] [in a new window]   Figure 1.   (A) Typical 5-min polygraph recording on the diagnostic night. EMG = submental electromyogram; Thorax = rib cage movement strain gauge (uncalibrated). Abdomen = abdominal movement strain gauge (uncalibrated); Thermistor = oronasal airflow (uncalibrated); SaO2 = pulse oximetry. The subject is in stage 1 sleep. Note five central apneas and associated desaturations and arousals. (B) Typical 5-min polygraph recording on the adaptive servo-ventilation night. Same subject as (A). Pressure = mask pressure. Other abbreviations as for (A). Note the transient increase in pressure modulation amplitude within 1 to 2 breaths of the onset of a central hypopnea early in the trace, and the absence of desaturation or arousal. [Thermistor, rib cage, and abdominal movement signal gains were chosen for visual clarity, are uncalibrated, and therefore differ from (A).]

Subjects

Population pool. A total of 16 sequential subjects were enrolled from the pool of all subjects referred to a secondary referral center (Elisabeth Hospital, Kirchen) for heart failure, who, after optimization of medical treatment, showed a 3% desaturation index of > 15/h on routine pulse oximetry performed on all heart failure patients. Chronic heart failure was diagnosed by the referring cardiologist on the basis of history and examination, chest X-ray, echocardiography, and in most cases by left heart catheter.

Inclusion/exclusion criteria. Inclusion criteria were as follows: adult ambulatory subjects with stable cardiac failure already known to have predominantly CSA/CSR at polysomnography (performed at the secondary referral center), giving written informed consent. Exclusion criteria were as follows: subjects already on CPAP or bilevel therapy, subjects with more than 10 obstructive events per hour of sleep, unstable angina, myocardial infarction within the previous 3 mo, change in medication within the previous 2 wk, subjects with tachycardic atrial fibrillation refractory to drug therapy, or subjects who had been resuscitated. (Four subjects with intermittent atrial fibrillation but a normal heart rate were included.)

Subject characteristics on admission to study. Subject characteristics are shown in Table 1. All patients had been stable and on optimal medical therapy for at least 2 wk and typically 1 mo before admission. No patient had a history of stroke. With some exceptions, subjects typically had ischemic heart disease and hypertension, were elderly, moderately sleepy, had mild mixed spirometric defect, were slightly hypocapnic and mildly hypoxic, were moderately obese, New York Heart Association class III (NYHA III), with enlarged heart and reduced fractional shortening. Epworth Sleepiness Scale (ESS) was slightly elevated and equivalent to that seen in moderate obstructive sleep apnea (17). Chest X-ray showed localized alveolar edema in two subjects, Kerley B lines in eight, and enlarged upper lobe vessels in 12, but no subject showed frank pulmonary edema or effusion. Two of the 16 subjects were subsequently excluded from analysis because of severe intractable mouth leak, particularly on the bilevel night, giving a total of 14 subjects for analysis.

Experimental Design

The study used an acute prospective randomized crossover design. Subjects were studied in the laboratory with polysomnography on five sequential nights: on the first night, the untreated polysomnogram was repeated, and then subjects were tested on four treatment nights in random order: nasal oxygen, nasal CPAP, bilevel ventilation, and ASV. The untreated night always preceded the treatment nights.

On the nasal oxygen night, subjects received nasal oxygen at 2 L/ min through nasal cannulae. On the CPAP night, subjects slept with nasal CPAP (Elite; ResMed, Sydney, Australia). CPAP was set initially to 6 cm H₂O, and adjusted until CSA/CSR was eliminated or a reasonable balance was struck between control of CSA/CSR and patient tolerance. Final pressures in the range 7.5 to 10.5 cm H₂O (mean ± SD, 9.25 ± 1.0 cm H₂O) were used. Mean overnight pressure was 8.7 ± 0.1 cm H₂O (95% of final pressure). On the bilevel night, subjects slept with a VPAP device (VPAP-II ST; ResMed) in ST mode. Expiratory positive airway pressure (EPAP) was set at 5 to 6 cm H₂O (mean overnight EPAP 5.3 ± 0.1 cm H₂O), which eliminated most obstructive events. Inspiratory positive airway pressure (IPAP) was initially set to 11 to 15 cm H₂O, and increased until CSA/CSR was eliminated or a reasonable balance between CSA/CSR and patient tolerance was struck. Final inspiratory pressures in the range 11 to 15 cm H₂O (mean ± SD, 13.5 ± 1.1 cm H₂O) were used. Mean overnight IPAP was 12.0 ± 0.3 cm H₂O (89% of final). Assuming a duty cycle of 40%, this gave a mean overnight mask pressure of 8.0 cm H₂O). The backup rate was set at the patient's spontaneous awake respiratory rate less 2 breaths/min. Rates in the range 13 to 18/min were used.

On the ASV night, end-expiratory pressure was set at 4 to 6 cm H₂O with the intention of eliminating any obstructive sleep apnea. Minimum support (difference between end-inspiratory pressure and end-expiratory pressure) was set to 4 cm H₂O, and maximum support set to 10 cm H₂O. (It was not necessary to adjust the respiratory rate setting from the default of 15/min, because the device was designed to automatically track the patient's spontaneous rate above or below this value.) The same mask was used on the CPAP, bilevel, and ASV nights (mostly ResMed Mirage). During ASV therapy, the mean pressure reaches approximately 9 cm H₂O during apneas or hypopneas, but settles to approximately 7 cm H₂O for most of the night during steady breathing.

Between the untreated night and the first of the four treatment nights, the patient underwent CPAP mask fitting, and three separate CPAP practice sessions of between 1 and 2 h, in order to manage problems related to mask fitting, headstrap tightening, and the like, and to acclimatize the patient to pressure and mask use. On the afternoon of the CPAP, bilevel, and ASV nights, subjects were given a further 1- to 2-h practice session with the appropriate treatment modality. The CPAP practice session was at 6 cm H₂O, the bilevel practice session with IPAP/EPAP of 12/4 cm H₂O, and the ASV practice session at the standard settings listed previously. At bedtime, subjects were given 15 to 20 min further habituation time between putting on the mask and lights out. This time was believed to be important because it permitted subjects' PCO₂ to recover after any period of voluntary (behavioral) hyperventilation at the time of donning the mask. Previous experience had shown that if this was omitted, then voluntary or behavioral overventilation would lead to upper airway (and probably vocal cord) closure at sleep onset, indicated by apneas despite cycling of the machine or the addition of high levels of EPAP.

Measurements

Polysomnography was performed using a digital polygraph (Compumedics, Melbourne) including two electroencephalograms (EEG), two electrooculograms (EOG), submental electromyogram (EMG), tibialis EMG, rib cage and abdominal inductance pneumograms, low-profile oronasal thermistor (Nelcor), pulse oximeter (Nonin OEM2) with response time set to 3 s, and body position. In 10 of 14 subjects, transcutaneous PCO₂ (PtcCO₂) was measured (Radiometer TCM3, Copenhagen). The gain of the transcutaneous PCO₂ analyzer was adjusted to agree with simultaneous awake arterialized capillary PCO₂ (Pa_CO2) taken immediately before commencing therapy. In the morning, on awakening, and while still on ASV, the arterialized capillary measurement was repeated. Morning arterialized capillary PCO₂ was also measured in six to nine patients on the other nights. In the remaining four subjects, the transcutaneous PCO₂ meter was not available and neither transcutaneous nor arterialized capillary samples were taken. The PtcCO₂ was recorded every 5 min throughout the night and the average calculated. Sleep staging and arousals were scored using 30-s epochs with Rechtschaffen and Kales (18) and American Sleep Disorders Association (19) criteria respectively. Arousals were scored as respiratory if they followed an apnea, hypopnea, or (very rarely) loud snoring, and nonrespiratory otherwise. Myoclonic arousals were scored as nonrespiratory. On the oxygen night, the oronasal thermistor was worn together with the oxygen cannulae, and on the CPAP, bilevel, and ASV nights, the oronasal thermistor was worn under the nasal mask. The thermistor is mounted in thin adhesive tape, not heavy plastic, and produces no mask leak.

Apneas were scored if there was a 90% reduction in thermistor signal. Central apneas were scored if there was absence of rib cage and abdominal movement during the apnea, and obstructive apneas scored otherwise. Hypopneas were scored in two ways: from the oronasal thermistor, and from the chest and abdominal bands. In either case, a hypopnea was scored if there was a 50 to 90% decrease in signal amplitude for at least 10 s over the stable baseline. No desaturation criterion was used in either case. The apnea + hypopnea index (AHI) was calculated as the number of apneas plus hypopneas per hour of sleep, with separate calculations for thermistor-scored events and band-scored events. The 2% and 4% desaturation indices were calculated as the number of transient arterial hemoglobin oxygen saturation dips of at least 2% and 4% per hour of sleep time. The respiratory arousal index (RARI) and total arousal index (ARI) were calculated as the number of corresponding events per hour of sleep. Slow-wave sleep (SWS) and rapid eye movement (REM) sleep were expressed as percentage of total sleep time.

All studies, including the original untreated night studies, were scored by a single scorer who was unaware of the purpose of the study or of the treatment modalities. To ensure that the scorer was blinded to the treatment modality, the pressure tracing was removed from the replay montage.

We watched for obvious clinical signs of reduced cardiac output: on first exposure to ASV while awake, we watched for dyspnea, angina, sweating, tachycardia (via pulse oximetry), or distress; during the night we watched for new or worsening arrhythmia (fast atrial fibrillation, runs of ventricular ectopics, etc.). Patients were asked to comment on any adverse symptoms that they noticed during the night. The protocol for this study was approved by the institutional review board of the University of Essen, and written informed consent was obtained.

Subjects were told that they would be treated with oxygen, and with three different machines using a mask. They were aware that two of the treatment methods were experimental, but they were unaware of which was which. After the completion of the study, subjects were asked, "Which of the three machine treatment nights did you prefer: the first, second, or third?" They were not asked to comment on the untreated night or the oxygen night.

Statistical Analysis

Data are given as mean ± SEM unless otherwise stated. Statistical comparisons between the untreated night and the four treatments, and between the ASV night and the other four conditions, were made using repeated measures analysis of variance (ANOVA), followed by multiple planned comparisons with a modified Bonferroni's correction based on treatment degrees of freedom, as described by Keppel (20). Preliminary data analysis using box plots showed that within any one treatment night all variables were roughly normally distributed, but the respiratory variables (apnea indices, AHI, respiratory arousal indices) showed greatly reduced variance with treatment compared with untreated (and occasional outliers), so a rank transform was used for these variables. The effect of the passage of time on sleep and breathing indices was assessed using analysis of covariance. The hypothesis that evening Pa_CO2 was constant across the five treatments was tested using one-way ANOVA. The constancy of the overnight change in Pa_CO2 from evening to morning across the five treatments was tested using repeated measures ANOVA on those subjects who had both morning and evening measurements only. Because the null hypothesis was rejected, the test was then repeated separately for each treatment night.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

No adverse clinical events occurred. There was no significant weight loss over the study period (mean loss 0.54 ± 0.4 kg, p = 0.2). Across all five nights, evening Pa_CO2 increased by 0.4 ± 0.3 mm Hg per night (p = 0.22). Across the four active treatment nights, there was neither a clinically important nor a statistically significant improvement in any parameter with the passage of time alone: AHI decreased by 0.004 event/h per night (p = 0.99); arousal index decreased by 0.8 event/h per night (p = 0.6), and central apnea index decreased by 0.15 event/h per night (p = 0.9). Typical 5-min epochs untreated and on ASV are shown in Figures 1A and 1B respectively.

Results for breathing, saturation, and mean overnight PtcCO₂ are shown in Table 2, and results for sleep are shown in Table 3. Data are given as mean ± standard error, followed by p values for the comparisons against ASV on the second line, and comparisons against control on the final line. Overnight changes in arterialized capillary PCO₂ are shown in Table 4. The drift in PtcCO₂ relative to simultaneous arterialized capillary PCO₂ overnight (n = 41) was 0.0 ± 0.5 mm Hg (not significant [NS]). On the ASV night in particular, the drift was 0.2 ± 1.2 mm Hg (NS).

Figure 2 shows that all four forms of treatment reduced the central apnea index, with an approximate halving for both oxygen and CPAP, a further 50% reduction for bilevel, and a yet further halving for ASV. Broadly comparable results were found for both thermistor and effort band AHI, arousal index (Figure 3), and thermistor and effort band respiratory arousal indices. All forms of treatment produced good improvements in the 4% desaturation index, with oxygen and ASV producing better results than CPAP or VPAP.

View larger version (21K): [in this window] [in a new window]   Figure 2.   Box plots of effect of treatment on central apnea index. Horizontal bar: median; thick vertical line: interquartile range; circles: outliers; thin bar: range excluding outliers. Also shown are statistical significance of comparisons between control and each of the four treatments, and between ASV and the other four conditions.

View larger version (24K): [in this window] [in a new window]   Figure 3.   Box plots of effect of treatment on arousal index. Legend as for Figure 2.

All four forms of treatment produced similar improvements in total sleep time and sleep efficiency. Neither oxygen nor CPAP produced significant changes in percentage of SWS or REM sleep, whereas moderate improvements in both SWS and REM sleep were seen with ASV, with bilevel occupying a position intermediate between CPAP and ASV. Across the five nights, there was a positive linear correlation between arousal index and AHI (r = 0.86, p < 0.001), and a negative correlation between percentage of REM sleep and AHI (r = 0.36, p = 0.025). All but one subject preferred sleeping with ASV to sleeping with either nasal CPAP or bilevel ventilation; the fourteenth subject found ASV and bilevel equally comfortable but preferable to CPAP.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study describes a randomized crossover trial of ASV, a new approach to the treatment of Cheyne-Stokes breathing and CSA in cardiac failure, with three other forms of treatment: nasal oxygen, nasal CPAP, and ST bilevel ventilatory support, for one night each. We confirm the results of Naughton and colleagues (7, 15, 16) showing an improvement in breathing and sleep with nasal CPAP, and the work of Andreas (9, 10), Franklin (3), Staniforth (11), and others showing an improvement with oxygen. With the default settings, CSA/ CSR was reduced to 14% (thermistor AHI) or 17% (band AHI) of untreated, superior to that achieved with the other treatment modalities. The key result was a substantial improvement in sleep quality compared with untreated. This improvement was larger than that seen with one night of nasal oxygen or a nasal CPAP titration night. Severe sleep disturbance is a key feature of cardiac failure and of CSA/CSR. Improvement in sleep quality of a similar magnitude in patients with obstructive sleep apnea syndrome leads to a worthwhile improvement in quality of life (21, 22). The tight correlation between arousal index and AHI is consistent with (but does not prove) the notion that the mechanism of improved sleep is abolition of respiratory arousals caused by CSA/CSR.

Methodological Issues

The untreated night was performed before the four treatment nights. Order effects cannot explain differences between the four treatment nights, because these nights were randomized. Order effects might explain some of the improvement over untreated with all four treatment modalities. However, there was no weight loss during the study period, and negligible progressive change in evening PCO₂, arguing against order effects resulting from resolution of pulmonary edema with bed rest or medication compliance. Our subjects had all been studied in the recent past, in the same room, with the same montage, and the effect is small or nonexistent for SWS, REM sleep, and AHI (23), arguing against first night order effects.

The thermistor's ability to detect hypopneas could be differentially affected by nasal cannulae versus mask. However, the thermistor AHI, respiratory band AHI, and the 2% and 4% desaturation indices all yielded consistent results (Table 2).

One night of 2 L/min nasal oxygen and one night of gradually increasing nasal CPAP produced partial improvements in AHI and arousal, but not sleep staging. The effects of oxygen are comparable with those reported by Andreas and colleagues (9) and others (3, 11). One night of CPAP produced a similar partial response to oxygen. Although the mean pressure delivered during the CPAP titration night was 95% of the final pressure, it is possible that a night at the final fixed pressure would have worked better. However, the response was comparable to that seen during long-term nasal CPAP (at pressures roughly comparable to those used here) by Naughton and colleagues (7, 14). We disagree with those who found no acute or long-term benefit (11, 13). Our results with ST mode bilevel ventilation are comparable to those of Willson and colleagues (26) using volume cycled ventilation. Bilevel ventilation appeared superior to both oxygen and CPAP.

Adaptive Servo-ventilation versus Other Treatments

In our acute study ASV produced a better improvement in sleep and breathing than either nasal CPAP or 2 L/min nasal oxygen. Oxygen and ASV are not mutually exclusive, and may work synergistically. ASV provided an 83% further reduction in the central apnea index compared with nasal CPAP at similar mean pressure, yielded better improvement in sleep quality, and was subjectively preferred. Compared with bilevel ventilation, ASV produced a further halving of AHI, central apnea index, and respiratory arousal index (and a modest further improvement in SWS and REM sleep). However, it took some time to reach optimal settings on the bilevel night, and it is possible that on a subsequent night at these optimal settings, bilevel ventilation would have performed as well as ASV.

Transcutaneous and Arterialized Capillary PCO₂

As expected, subjects showed marked arterial hypocapnia in the daytime. Daytime spot blood gas PCO₂ was very low untreated, as was evening arterialized capillary PCO₂. The 6.0 mm Hg increase in Pa_CO2 with oxygen is to be expected because of direct suppression of the carotid body. There was also a 4 mm Hg increase in overnight Pa_CO2 on the CPAP night. Possible speculative mechanisms include increased lung volume, a reduction in respiratory arousals, a reduction in lung water, and mask dead space, all leading to a reduction in overventilation.

There was a 3.6 mm Hg increase in overnight PCO₂ on the ASV night. A decrease in PCO₂ is ruled out with 95% confidence. If, on the bilevel or ASV nights, the degree of support was more than sufficient to do the entire respiratory work of the subject, the morning arterialized capillary PCO₂ and PtcCO₂ would necessarily fall. However, this did not occur, demonstrating that further overventilation did not occur. Indeed, the morning arterialized capillary PCO₂ actually increased by 3.5 mm Hg. We speculate that the reasons for this are as follows. (1) The device only provides significant ventilatory support during hypopneas. (2) The steady-state and usual degree of support (3 cm H₂O, using an intentionally inefficient waveform) is sufficient to do only part of the respiratory work. Under these conditions, the subject's chemoreflexes will reduce the phrenic neural output by a corresponding amount, so that ventilation per se would produce negligible change in PCO₂. Central chemoreflex gain is particularly high in patients with CSA/CSR (27). (3) As with the CPAP night, numerous factors can potentially reduce ventilatory drive. (4) On ASV, the greater reduction in respiratory arousals and increased sleep efficiency will cause the subject to spend more time at the sleeping PCO₂ set-point, and less time at the awake or aroused set-point.

Leg Movements

Hanly and Zuberi-Khokyar (30) found a 52% incidence of myoclonus in cardiac failure or Cheyne-Stokes breathing. In the present study, we found 11.6 myoclonic events/h, which was completely uninfluenced by short-term treatment. Despite this, there was a huge improvement in sleep quality with treatment, and therefore the myoclonus can at most explain a part of the disordered sleep.

Conclusion

One night of therapy with ASV improved nocturnal breathing pattern and sleep quality in patients with CSA/CSR in heart failure. Sleep and breathing was better during one night of ASV therapy than during one night of 2 L/min nasal oxygen or a nasal CPAP titration night. Patients preferred the ASV night to either the CPAP or bilevel titration nights. A long-term study of the effects of ASV on cardiovascular function is needed.