INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Cheyne-Stokes respiration with central sleep apnea (CSR-CSA) is commonly observed in patients with congestive heart failure (CHF) (1, 2). It is characterized by a crescendo-decrescendo pattern of hyperpnea alternating with central apneas and hypopneas. Although CSR-CSA is a consequence of CHF, once established, it probably plays a role in the progression of cardiac dysfunction. This possibility is supported by the observations that CSR-CSA in CHF patients is associated with recurrent hypoxemia, arousals from sleep, increased sympathetic nervous system activity and frequency of ventricular arrhythmias, and increased risk of mortality (3). However, the mechanisms leading to CSR-CSA have not been clearly established.

Several studies have demonstrated that CHF patients with CSR-CSA are hypocapnic both awake and asleep (4, 6). In addition, central apneas during sleep in these patients are frequently initiated by abrupt increases in ventilation and falls in Pa_CO2 (7, 9). Thus, chronic hyperventilation may maintain Pa_CO2 close to the apnea threshold, whereas acute augmentation of ventilation may drive Pa_CO2 below this threshold and precipitate central apneas (1, 7, 9). However, it is not clear whether central apneas are triggered by reductions in Pa_CO2 sensed by the peripheral or by the central chemoreceptors.

The cause of hyperventilation and hypocapnia in stable CHF patients with CSR-CSA remains uncertain. One possibility is that hypoxia stimulates hyperventilation, which precipitates central apneas by driving Pa_CO2 below the apnea threshold (6, 10). This is the mechanism proposed for the development of central apneas in high-altitude periodic breathing (9, 11, 12). However, CHF patients with CSR-CSA are generally not hypoxemic (1, 6). Nevertheless, administration of supplemental O₂ has been reported to reduce modestly, but not to abolish, central apneas in CHF patients with CSR-CSA (10, 13, 14).

There is some evidence to suggest that inhalation of low concentrations of CO₂ alleviates CSR-CSA. However, this evidence comes from only a few case reports in which neither the effect of CO₂ on minimal end tidal fraction of CO₂ (FET_CO2) nor on Pa_CO2 was assessed (15). Therefore, it remains unclear whether CO₂ inhalation abolishes CSR-CSA and, if so, through what mechanism. If reduction in Pa_CO2 below the threshold for apnea is the mechanism responsible for central apneas during sleep in CHF patients with CSR-CSA, then raising Pa_CO2 above the apneic threshold should eliminate these events. However, increases in ventilation and abolition of apneas in CSR-CSA are also likely to increase Sa_O2 (18). If raising Pa_CO2, rather than increasing Sa_O2, is the primary mechanism through which CO₂ inhalation abolishes central apneas, then raising Sa_O2 to the same level by administration of inhaled O₂ should not reduce the frequency of central events to the same extent as CO₂ inhalation.

In view of the above, we set out to test the following hypotheses. First, because changes in Pa_CO2 occurring secondary to changes in ventilation should be detected by the peripheral before the central chemoreceptors (18), we hypothesized that central apneas in CSR-CSA are initiated by reductions in Pa_CO2 sensed at the carotid bodies. Second, we hypothesized that inhalation of a CO₂-enriched gas eliminates CSR-CSA primarily by raising Pa_CO2 above the apnea threshold rather than by improving oxygenation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Ten consecutive patients with CHF who were referred to our sleep disorders clinic and found to have CSR-CSA were included in the study. The patients had a history of CHF of at least 6 mo duration due to ischemic cardiomyopathy as documented by: (1) at least one episode of pulmonary edema evidenced by cardiomegaly with pulmonary congestion on a chest radiograph; (2) demonstration of coronary occlusion or flow-limiting stenosis (> 75% stenosis) on coronary angiography, or by a history of documented myocardial infarction; and (3) a left ventricular ejection fraction at rest of  35% as determined by equilibrium radionuclide angiography. Patients were free of neurological disease on the basis of a negative history and physical examination. The presence of CSR-CSA was established during a diagnostic polysomnographic recording and was based on the presence of at least 15 apneas and hypopneas per hour of sleep, of which at least 80% had to be central in nature. The protocol was approved by the Human Subjects Review Committee of the University of Toronto and written informed consent was obtained from all patients.

Experimental Setup

Experimental sleep studies were then performed on each patient using standard techniques described for our laboratory (7). Sleep was staged according to standard criteria (19). Thoracoabdominal movements were measured by a respiratory inductance plethysmograph (Respitrace; Ambulatory Monitoring Inc., White Plains, NY). Tidal volume (VT) was taken as the sum of the rib-cage and abdominal displacements, and was calibrated against a spirometer using the two- position simultaneous equations technique (7, 20). Sa_O2 was measured continuously in all patients with a pulse oximeter (Nellcor N200; Nellcor Puritan Bennett, Inc., Pleasanton, CA), placed on the ear. Transcutaneous PCO₂ (Ptc_CO2) was continuously measured with a transcutaneous capnograph (Kontron Medical; Hoffman-La Roche, Basel, Switzerland), with the electrode placed on the anterior chest wall. The instrument was calibrated as previously described in our laboratory (7) and was recalibrated at the end of the study to a PCO₂ of 23 and 55 mm Hg. The PCO₂ during recalibration was always within 2 mm Hg of the test-gas value. Ptc_CO2 measured with this instrument has been shown to correlate closely with Pa_CO2 under a wide variety of clinical conditions, both acutely and over time (7). Central apneas were defined by the absence of VT excursions for at least 10 s in the absence of rib-cage and abdominal movement. Central hypopneas were defined as a 50% or greater reduction in VT from the baseline value, persisting for at least 10 s, in the absence of phase shift or paradoxical motion of the rib cage and abdomen. The number of apneas per hour of sleep was defined as the apnea index (AI), and the number of apneas and hypopneas per hour of sleep as the apnea-hypopnea index (AHI).

In addition to the previously described methods used during the diagnostic study, during the experimental sleep study expired air was sampled from nasal cannulae, from which the FET_CO2 was measured by an infrared CO₂ analyzer (model LB-2; Beckman, Schiller Park, IL). The instrument was calibrated at the beginning of each study and recalibrated at the end of the study by using dry gas samples of 3, 5, and 8.4% CO₂. The offset was within 0.1%.

Patients breathed through a tight-fitting face mask, with built-in low-resistance inspiratory and expiratory valves. Two different Douglas bags with 60 L capacity were connected to a three-way stopcock, which was in turn connected to the inspiratory port of the face mask by vinyl tubing 17 mm in internal diameter. One bag delivered a CO₂-enriched gas (4% CO₂-21% O₂-76% N₂) mixed with compressed air (CO₂). The inspired fraction of CO₂ (FI_CO2) was adjusted by manually controlling the flow rates of the two gas streams into the Douglas bag that was maintained partially full. The second bag delivered compressed air and O₂. Through this bag the subjects breathed either compressed air or an O₂-enriched mixture (O₂). A flow rate of approximately 30 L/min of the different gas mixtures was supplied, and was varied according to each patient's minute ventilation. Patients expired through a separate low resistance valve that minimized dead space and prevented pressure buildup inside the mask. The circuit allowed the subjects to breathe either the air, CO₂, or O₂ mixtures. The concentration of the different gas mixtures and the switching of the inspired gases was controlled by the experimenter in a separate room from the patient to minimize sleep disruption.

Protocol

Prior to sleep studies, arterial blood gas tensions were determined while the patients were awake and breathing room air. Patients then went to sleep wearing a face mask, initially breathing air. All interventions were performed on a single night during stage 2 (S2) non-rapid eye movement (NREM) sleep. Once S2 sleep with CSR-CSA became established for at least 10 min, the gas mixture was switched to CO₂. The FI_CO2 was slowly increased until central apneas and hypopneas were abolished. CO₂ inhalation was continued for at least 10 min, after which the patient was switched to air inhalation. To control for any potential confounding influence of CO₂-induced increases in Sa_O2 on CSR-CSA, O₂ was administered at a concentration sufficient to increase Sa_O2 to the same level as during CO₂ inhalation. After the first trial of CO₂ inhalation, subsequent trials of either CO₂- or O₂-enriched gas mixtures were randomly administered and were interposed with 10 min of air inhalation. Any time the patient woke up or went to a deeper stage of sleep, he was switched to air inhalation.

Data Analysis

A single technician scored respiratory events and sleep stages. Values of all factors measured during all intervals of inhalation of the same gas mixture were averaged for each patient. Because S2 sleep was the sleep stage in which the majority of apneas were observed, and because we wished to avoid the potential confounding influence of sleep stage change on the frequency of respiratory events, analyses were restricted to S2 sleep. The time spent in CSR-CSA, including apneas, hypopneas, and hyperpneas, was expressed as a percentage of S2 sleep time.

To test the hypothesis that low Pa_CO2 generated during hyperpnea triggers central apneas when it is detected by the carotid bodies, we measured the time lag between the beginning of the breath with the lowest FET_CO2 during hyperpnea to the beginning of a central apnea (CO₂ circulation time or [CO₂ CT]). We then compared this CO₂ CT with an independent measure of lung to carotid body circulation delay, the lung to ear circulation time (LECT) (21). The measurement of LECT takes advantage of the fact that the ear lies in close proximity to the carotid body. LECT was measured from the onset of the first breath after a central apnea to the subsequent nadir of Sa_O2 detected by an oximeter on the ear. The minimum FET_CO2, taken from the end of the expiratory plateau (22), was calculated during S2 sleep for three conditions: the hyperpneic phase of CSR-CSA preceding apneas and hypopneas, and stable breathing, defined as a continuous period of at least 10 min during which there were no apneas or hypopneas during room air breathing. Because not all patients had periods of stable breathing on air, we also compared minimum FET_CO2 preceding apneas, hypopneas, and during stable breathing while inhaling CO₂. For each subject, the minimal FET_CO2 values preceding apneas and hypopneas were averaged over five consecutive CSR-CSA cycles during S2 sleep. The minimal values for two consecutive 5-min periods of stable breathing on air and on CO₂ immediately preceding or following periods of CSR-CSA during S2 sleep were also averaged for each subject.

Mean values of sleep study variables during S2 sleep in all 10 patients while inhaling air or CO₂ were compared by paired t tests where data were normally distributed. Wilcoxon signed rank test was used whenever the data were not normally distributed. Mean values of sleep study variables during S2 sleep in the six patients who received air, CO₂, and O₂ were compared by one-way analysis of variance (ANOVA) for repeated measures. Friedman repeated measures ANOVA on ranks was used where data were not normally distributed. Post hoc analyses were performed by Student-Newman-Keuls test where appropriate. Relationships between variables were assessed by least-squares linear regression analyses. A p value of < 0.05 was considered statistically significant. Data are expressed as mean ± SD.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

All 10 patients were men and were generally nonobese. Their baseline demographic and sleep study data are presented in Table 1. During the baseline sleep study all had moderate to severe CSR-CSA during which events were at least 80% central (range 83 to 100%). The few obstructive events observed occurred during rapid eye movement (REM) sleep. The arterial blood gas analyses while awake before the experimental study showed a Pa_O2 of 89.8 ± 18.5 mm Hg, a Pa_CO2 of 34.3 ± 5.1 mm Hg, and a pH of 7.46 ± 0.04. The mean Ptc_CO2 while asleep on air was 36.8 ± 4.7 mm Hg, and correlated with the awake Pa_CO2 (r = 0.84, p < 0.003).

Figure 1 shows how LECT and the CO₂ CT were determined. Figure 2 illustrates the strong correlation between CO₂ CT and LECT for all subjects. Figure 3 shows a spontaneous transition from CSR-CSA to stable breathing in one patient. Initially, arousals from sleep augment ventilation and lower Pa_CO2 below the apnea threshold triggering a central apnea (7, 23). Then, as arousals dissipate, there is a progressive decrease in ventilation during hyperpneas accompanied by a progressive increase in the minimum FET_CO2 from that preceding apnea, to that preceding hypopnea to that during stable breathing. The minimum FET_CO2 during stable breathing, and during hyperpnea preceding hypopnea and apnea for the five patients who had stable breathing during air inhalation are presented in Figure 4. Minimum FET_CO2 became progressively and significantly higher from pre-apnea, to prehypopnea to stable breathing.

View larger version (26K): [in this window] [in a new window]   Figure 1.   Polysomnographic recordings from a representative patient with Cheyne-Stokes respiration and CSR-CSA during S2 sleep illustrating the method of measuring LECT (top bar) and CO2 CT (bottom bar). LECT is the time between the first breath after the apnea to the maximal dip in SaO2 detected by an oximeter on the ear (adjacent to the carotid body). During the hyperpneic phase of CSR-CSA, FETCO2 was highest during the first few breaths, reflecting the CO2 accumulation that occurred during the preceding apnea and the low VT at this point. It then progressively decreased reaching its lowest point during midhyperpnea before increasing in the last part of the hyperpnea as VT progressively decreased. The CO2 CT is the time lag between the lowest FETCO2 during hyperpnea to the onset of apnea. The LECT and the CO2 CT are identical. Note cardiac oscillations in FETCO2 during apneas, indicating that the upper airway is open and, therefore, that apneas are central. EEG = electroencephalogram; EMG = submental electromyogram.

View larger version (16K): [in this window] [in a new window]   Figure 2.   The strong relationship between CO2 CT and SaO2 LECT for the 10 subjects suggests that a low PaCO2 (below the apneic threshold) triggers central apneas when sensed at the peripheral chemoreceptors.

View larger version (29K): [in this window] [in a new window]   Figure 3.   Recording from one subject during S2 sleep illustrating a spontaneous shift from CSR-CSA to stable breathing. During the first two hyperpneas, arousals occurred which were probably responsible for reducing FETCO2. During the last three hyperpneas, however, there were no arousals so that increases in ventilation were damped, allowing FETCO2 to rise. Thus, as minimum FETCO2 during hyperpneas progressively increases from the left to the right of the figure, there is a transition from apneas, to hypopneas, to stable breathing.

View larger version (13K): [in this window] [in a new window]   Figure 4.   Individual data for minimum FETCO2 during stable breathing (SB), and during the hyperpneic phase preceding apneas and hypopneas for the five subjects who had SB on air. Minimum FETCO2 during periods of SB was 4.18 ± 0.20%, preceding hypopneas was 3.83 ± 0.38%, and preceding apneas was 3.40 ± 0.37%. *p < 0.05 compared with SB on air and apnea. **p < 0.05 compared with SB on air and hypopnea.

While breathing room air, the presence and severity of CSR-CSA was relatively constant during S2 sleep throughout the night. Due to poor sleep quality and constraints imposed by our protocol, not all patients had sufficient sleep time to receive both CO₂ and O₂. All patients underwent at least two CO₂ trials, but only six underwent O₂ trials. In these latter six patients, the average number of CO₂ and O₂ trials was 3.5 ± 1.4 and 2.2 ± 1.0, respectively. The FI_CO2 required to eliminate central respiratory events was 1.85 ± 0.64% and ranged from 0.85 to 2.89%. Figure 5 depicts the raw data of one subject demonstrating abolition of CSR-CSA by CO₂ inhalation, in association with an increase in Ptc_CO2, and elimination of dips in Sa_O2. The effects of O₂ inhalation in another subject are shown in Figure 6. Although O₂ desaturations are eliminated, CSR-CSA continues.

View larger version (28K): [in this window] [in a new window]   Figure 5.   Representative polysomnographic recordings from a patient during S2 sleep while breathing air (A) and CO2 (B). Central apneas are abolished by CO2 inhalation in association with an increase in FETCO2 above the level during hyperpneas preceding apneas, and a 1.6 mm Hg increase in PtcCO2. Abolition of Cheyne-Stokes respiration with central sleep apneas was also associated with elimination of dips in SaO2.

View larger version (19K): [in this window] [in a new window]   Figure 6.   Representative polysomnographic recordings during S2 sleep while breathing air (A) and O2 (B) from a different patient than shown in Figure 5. Although O2 inhalation abolishes dips in SaO2, Cheyne-Stokes respiration with central sleep apneas persists. There were no significant effects of O2 inhalation on either FETCO2 or PtcCO2.

Figure 7 shows the minimum FET_CO2 preceding apnea, hypopnea, and during stable breathing on CO₂ for all 10 patients. Minimum FET_CO2 became progressively and significantly higher from preapnea, to prehypopnea to stable breathing on CO₂. The effects of CO₂ on polysomnographic parameters for all 10 patients are presented in Table 2. In association with a 1.7 mm Hg increase in Ptc_CO2, there is virtual elimination of apneas and hypopneas, and CSR-CSA. There is also a significant reduction in the frequency of movement arousals and a 2.3% increase in mean low Sa_O2.

View larger version (16K): [in this window] [in a new window]   Figure 7.   Individual data of the 10 subjects for minimum FETCO2 during the hyperpneic phase preceding apnea and hypopneas, and during periods of SB while inhaling CO2 in S2 sleep. Minimum FETCO2 was 3.29 ± 0.80% preceding apneas, 3.83 ± 0.62% preceding hypopneas, and 4.66 ± 0.71% during periods of SB while on CO2. *p < 0.05 compared with apnea and SB on CO2. **p < 0.05 compared with apnea and hypopnea.

Figures 8 and 9 show data for the six subjects who received both O₂ and CO₂. Figure 8 illustrates that during CO₂ inhalation the increase in Ptc_CO2 required to eliminate CSR-CSA was only 1.6 ± 0.6 mm Hg above that during air breathing (range from 0.9 to 2.6 mm Hg). Note also that the Ptc_CO2 while breathing CO₂ was only 38.0 ± 4.4 mm Hg (ranging from 37.2 to 44.0 mm Hg), which is within the eucapnic range. Thus, CO₂ inhalation did not induce hypercapnia. In addition, CO₂ inhalation caused a significant increase in mean low Sa_O2 of 3.2 ± 1.4%. O₂ inhalation caused a similar increase in mean low Sa_O2 of 4.2 ± 3.0%. However, in contrast to CO₂ inhalation, it had no significant effect on mean Ptc_CO2. Figure 9 demonstrates that CO₂ inhalation practically abolished central apneas and hypopneas. Although O₂ inhalation caused a significant 38% reduction in the AI (by 9.4 ± 7.0 per hour), this reduction was significantly less pronounced than the 100% decrease (by 24.9 ± 7.3 apneas per hour, p < 0.05) caused by CO₂ inhalation. Moreover, O₂ inhalation had no significant effect on the combined AHI. Similarly, compared with air, there was a marked reduction in the percentage of time spent in CSR-CSA while on CO₂ (from 82.9 ± 10.1 to 6.1 ± 9.5%, p < 0.01), but not while on O₂ (to 84.5 ± 17.1%). Taken together, these data demonstrate that abolition of CSR-CSA by CO₂ inhalation was associated with a significant increase in Ptc_CO2 and in Sa_O2. However, when O₂ was administered at a rate sufficient to raise Sa_O2 to the same level as during CO₂ inhalation, it had no significant effect on either mean Ptc_CO2, AHI, or CSR-CSA time. In addition, the frequency of movement arousals during S2 sleep decreased from 32.6 ± 17.7 per hour on air to 18.7 ± 10.3 and 20.3 ± 14.7 per hour during CO₂ and O₂ inhalation, respectively (p < 0.03).

View larger version (16K): [in this window] [in a new window]   Figure 8.   Individual data for mean PtcCO2 and mean low SaO2 during S2 sleep for the six subjects who underwent air, CO2, and O2 inhalation trials. Mean PtcCO2 was significantly higher during CO2 inhalation (38.0 ± 4.4 mm Hg, *p < 0.05) than during air (36.4 ± 4.6 mm Hg) and O2 inhalation (36.7 ± 4.1 mm Hg). Mean low SaO2 was significantly higher (*p < 0.05) during both CO2 (95.6 ± 2.1%) and O2 inhalation (96.6 ± 1.4%) than during air breathing (92.4 ± 3.2%). However, there was no significant difference in mean low SaO2 between O2 and CO2 inhalation. O2 inhalation had no significant impact in PtcCO2. These data illustrate that the abolition of central apneas and hypopneas during CO2 inhalation was caused by an increase in PtcCO2 and not by an increase in SaO2. The increase in PtcCO2 required to eliminate apneas and hypopneas was only 1.6 ± 0.6 mm Hg above that during air breathing (range from 0.9 to 2.6 mm Hg).

View larger version (15K): [in this window] [in a new window]   Figure 9.   Comparisons of AI and AHI for each of the six subjects who underwent air, CO2, and O2 inhalation trials. The AI fell from 24.9 ± 7.3 per hour on air, to 0 ± 0 per hour on CO2 and to an intermediate level of 15.5 ± 13.0 per hour on O2. The AHI fell from 43.0 ± 8.4 per hour on air to 1.6 ± 2.6 per hour on CO2 but did not change significantly from air to O2 (35.1 ± 12.5 per hour). *p < 0.05 compared with air and O2. p < 0.05 compared with air.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present study provides several important insights into the pathophysiology of CSR-CSA in patients with CHF. We have previously shown that central apneas during CSR-CSA are preceded by hyperventilation and a fall in Pa_CO2 (7). However, in that study, we could not determine whether central apneas were triggered by reductions in Pa_CO2 below the apneic threshold detected by the peripheral, or the central chemoreceptors. In the present study, we observed that the time delay from the lowest FET_CO2 during hyperpnea to the onset of apnea was practically identical to LECT, which is a reasonable estimate of lung to carotid body circulatory delay (21). This strongly suggests that central apneas are triggered by reductions in Pa_CO2 sensed by the peripheral chemoreceptors. Another important finding was the demonstration of the critical role of hypocapnia, as opposed to hypoxemia, in the genesis of central apneas during CSR-CSA. Evidence for this was provided by several observations. First, the minimum FET_CO2 preceding apneas was lower than that preceding hypopneas, which was in turn lower than that during stable breathing on air (Figure 3). Second, inhalation of CO₂ at low concentrations abolished central apneas and hypopneas during CSR-CSA in conjunction with a 1.6 mm Hg mean increase in Ptc_CO2. Third, although CO₂ inhalation caused an overall increase in Sa_O2, a similar increase in Sa_O2 induced by O₂ inhalation had no significant impact on either Ptc_CO2 or the frequency of central apneas and hypopneas.

Both peripheral and central chemoreceptors respond to changes in Pa_CO2. Reductions in Pa_CO2 in the lung due to increases in ventilation are transmitted to the carotid chemoreceptors in 6 to 10 s in subjects with normal cardiac function (12, 21). In patients with CHF, however, we have previously shown that the LECT is two to four times longer than in healthy subjects (i.e., 15 to 40 s) (21), just as we observed in the CHF patients in the present study. The time required to transmit changes in Pa_CO2 in the lung across the blood brain barrier, to alter the pH of the cerebrospinal fluid, and to be detected by the central chemoreceptors is approximately two to three times longer than the lung to carotid body delay (12, 24). In order to test the hypothesis that central apneas are triggered by a low Pa_CO2 sensed at the peripheral chemoreceptor, we compared the CO₂ CT with the LECT (Figure 1). This method is subject to some technical limitations. The LECT was determined using an ear oximeter which gives a continuous measurement, and has a relatively short instrument lag time in detecting changes in Sa_O2 (approximately 3 s). In contrast, the FET_CO2 is an instantaneous measurement, but is not continuous and can only be determined at the end of expiration of each respiratory cycle (approximately every 3 to 4 s). In addition, we assumed that the lowest Pa_CO2 during the hyperpneic phase was the stimulus triggering the central apnea. Despite these limitations, the CO₂ circulation time and LECT were in remarkably close agreement (Figure 2). These data therefore support the concept that it is the reduction in Pa_CO2 detected at the peripheral rather than the central chemoreceptor that triggers central apneas. It is still possible that once central apneas are initiated at the peripheral chemoreceptor, they may be prolonged by the delayed detection of the reduced Pa_CO2 at the central chemoreceptors (11, 27). However, the present data do not allow us to confirm or rule out this possibility.

The effects of CO₂ inhalation on CSR-CSA in patients with CHF have not been thoroughly investigated. In two cases, dating back to the 1930s, inhaled 3 to 7% CO₂ was reported to alleviate Cheyne-Stokes respiration (16, 17). However, in these studies sleep state was not objectively monitored and there was no quantitative documentation of the effects of CO₂ on the frequency of apneas and hypopneas, or on oxygenation. More recently, the effects of 3% CO₂ inhalation on Cheyne-Stokes respiration in CHF patients were studied (15). However, Cheyne-Stokes respiration was not well defined, and included respiratory events that did not meet standard criteria for apneas and hypopneas. In any event, the investigators reported that 3% CO₂ alleviated Cheyne-Stokes respiration, although neither FET_CO2 nor PCO₂ was monitored systematically. Because of these experimental limitations, they acknowledged that they could not determine the mechanism of the effect of CO₂ in CSR-CSA. Therefore, although our findings on the effects of CO₂ on CSR-CSA agree, in general, with those of Steens and coworkers (15), they extend them in several important ways. First, we titrated the FI_CO2 to find the minimum level required to abolish CSR-CSA. Second, we monitored FET_CO2, Ptc_CO2, and Sa_O2. Third, we limited our experiments to S2 sleep to control for any potential influence of different sleep states. Finally, we tested the effects of supplemental O₂ to control for any potential confounding influence of increased Sa_O2 on CSR-CSA that might have occurred during CO₂ inhalation.

Several observations indicate that CO₂ inhalation abolished CSR-CSA primarily by increasing Pa_CO2 above the apneic threshold. During hyperpneas preceding apneas and hypopneas, FET_CO2 was in the hypocapnic range, and was significantly lower than during stable breathing on air. Similarly, Ptc_CO2 was in the hypocapnic range during CSR-CSA. CO₂ inhalation increased FET_CO2, to levels just above those during ventilatory periods preceding apneas and hypopneas, and increased Ptc_CO2 by just 1.6 mm Hg to within the eucapnic range. These findings are remarkably similar to those of Xie and coworkers (18) who demonstrated in hypocapnic idiopathic central sleep apnea patients with normal cardiac function, that CO₂ inhalation eliminated central events in association with just a 1.3 mm Hg increase in Ptc_CO2.

During CO₂ inhalation in the current study, FI_O2was maintained constant at 21%. Therefore, the increase in Sa_O2 associated with CO₂ inhalation was caused by elimination of apneas and hypopneas, and increases in ventilation. Hypoxia increases the gain of the peripheral chemoreceptors for Pa_CO2 and renders the respiratory system more susceptible to posthyperventilatory apneas (11, 12). Therefore, an improvement in Sa_O2 during CO₂ inhalation could have contributed to the elimination of CSR-CSA. However, raising Sa_O2 to comparable levels by O₂ inhalation caused no significant change in Ptc_CO2, AHI, or in the percentage of time spent in CSR-CSA. It did cause a marginal reduction in the AI, but this was significantly less pronounced than during CO₂ inhalation. These data indicate that raising Pa_CO2 into the eucapnic range by CO₂ inhalation has a significantly greater effect on the frequency of central apneas and hypopneas in CHF patients with CSR-CSA than does an isolated increase in Sa_O2 into the normoxic range. Accordingly, CO₂ inhalation abolishes CSR-CSA primarily by raising Pa_CO2 rather than by improving oxygenation.

The marginal, if any effect of inhaled O₂ on central apneas and hypopneas during CSR-CSA observed in our study is consistent with the findings of other investigators (10, 13). For example, Hanly and coworkers (10) reported that low-flow oxygen had no significant effect on the duration of CSR-CSA during S2 sleep, although it did reduce its duration during stage 1 sleep (10). In two recent studies by Franklin and coworkers (13, 14), the effects of inhaled O₂ were examined in groups of patients with central sleep apnea of heterogeneous etiology. Most patients had the combination of CHF and stroke which makes it difficult to compare them to our patients, all of whom had CHF, but none of whom had a stroke. In the first of Franklin and coworkers' studies (13), induction of hyperoxia by administration of high concentration (i.e., at least 50%) O₂ by Venturi mask was associated with a reduction in central events of approximately 50% in association with a significant increase in Ptc_CO2. In the second study, O₂ was administered at a lower concentration than in the first study (14). Although it was sufficient to abolish apnea-related dips in Sa_O2, it had no effect on Ptc_CO2 and caused only a marginal decrease in the frequency of central apneas, but did not reduce the amount of time spent in CSR-CSA. These data indicate that if and when O₂ inhalation causes a reduction in the severity of CSR-CSA, it does so when Sa_O2 is raised into the hyperoxic range in association with a rise in Pa_CO2 (11, 14). Although, in our study, O₂ inhalation decreased the frequency of arousals to the same extent as CO₂ inhalation, it did not damp ventilation sufficiently to raise Pa_CO2. It is possible, nevertheless, that if we increased Sa_O2 further into the hyperoxic, unphysiologic range, there may have been a greater increase in Ptc_CO2 and a more pronounced reduction in the AHI. Under these conditions, we would still be left to conclude that reductions in AHI were related to increases in Ptc_CO2. However, this was not the purpose of our study. Rather, we administered O₂ to control for the increased Sa_O2 that occurred during CO₂ inhalation.

Supplemental O₂ can abolish central sleep apnea associated with simulated high altitude. However, unlike CSR-CSA in patients with CHF, hypoxia plays the key role in stimulating hyperventilation and hypocapnia at high altitude. Accordingly, under these conditions, O₂ inhalation eliminates central apneas by reducing hypoxic drive and allowing Pa_CO2 to rise above the apneic threshold (11). Indeed, central apneas at high altitude can also be abolished by raising Pa_CO2 through inhalation of CO₂, even when subjects remain hypoxic. The reason that O₂ inhalation had little or no effect on CSR-CSA in our patients with CHF is probably because hyperventilation in these patients is not caused by hypoxia. It more likely arises from stimulation of pulmonary vagal afferents by interstitial edema (28).

It has recently been shown that the CB is primarily sensitive to Pa_CO2, whereas hypoxia acts mainly to sensitize the CB to Pa_CO2 (24). However, the CB remains sensitive to changes in Pa_CO2 even when Pa_O2 is increased into the hyperoxic range well above 200 mm Hg. In our experiments we increased Sa_O2 only to within the normoxic range (96.6 ± 1.4%) during O₂ inhalation, so that the CB would have remained sensitive to changes in Pa_CO2. This would explain how CSR-CSA could continue even when Sa_O2 was increased and Ptc_CO2 did not change.

Although CO₂ inhalation abolishes CSR-CSA and reduces the frequency of arousals, it also increases ventilation and would therefore augment the energy and blood flow demands of the respiratory muscles in the face of low cardiac output. Therefore, it is unlikely that CO₂ inhalation would be a useful long-term therapy for CSR-CSA in patients with CHF.

Our findings lead us to several conclusions. First, central apneas in CSR-CSA are most likely triggered by reductions in Pa_CO2 below the apneic threshold sensed at the peripheral chemoreceptors. Second, inhaled CO₂ abolishes CSR-CSA in CHF patients primarily by inducing increases in Pa_CO2 above the apneic threshold. Third, although hypoxic dips during central apneas might accentuate the tendency to hyperventilate and develop central apneas, it is unlikely that hypoxia plays a major role in the genesis of CSR-CSA. Indeed, our data indicate that hypoxia is the consequence of central apneas rather than their cause. Finally, our data strongly suggest that the mechanisms responsible for central apneas in CSR-CSA are similar to those in other forms of central sleep apnea associated with hypocapnia, such as idiopathic central sleep apnea (18, 23, 29). However, the mechanisms leading to hyperventilation and low Pa_CO2 in CHF remain uncertain and warrant further investigation.