INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Arousals from sleep are thought to play an important role in provoking recurrent elevations in blood pressure (BP) during obstructive sleep apnea and Cheyne-Stokes respiration (CSR) with central sleep apnea (1). Two hypotheses have been proposed to explain this effect. One is that arousals from sleep elicit direct reflex activation of the cardiovascular system (2, 5, 6) as part of a generalized startle reflex in preparation for adverse environmental conditions (7, 8). The other is that chemical and mechanical respiratory stimuli related to the apnea-hyperpnea cycle can directly provoke elevations in BP, independent of arousals from sleep (5, 9).

The method most frequently used to isolate the effects of arousal from those of respiratory activity on BP has been to elicit arousals from sleep using external stimulation (5, 14, 15). The resulting elevations in sympathetic nervous system activity, BP, and heart rate (HR) have been attributed to the arousal. However, this method does not eliminate respiratory stimuli associated with sleep and, further, it remains uncertain whether the cardiovascular effects of such artificially induced external arousals are indicative of the cardiovascular effects of arousals induced by apneas.

Several lines of evidence suggest that intrinsic respiratory stimuli may increase cardiovascular activity in response to apneas during sleep. First, in patients with obstructive sleep apnea, sympathetic nervous system activity and BP begin to rise toward the end of apneas prior to arousal, presumably due to the effects of apnea and hypoxia (2, 3, 16). Second, in dogs with experimentally induced obstructive sleep apnea, BP and HR increased abruptly following obstructive apneas even when they were not terminated by arousals (11). Finally, in awake humans, voluntary simulation of CSR produces the same fluctuations in BP as CSR during sleep, suggesting that sleep-wake state fluctuations are not necessary for the occurrence of BP fluctuations in CSR (17).

CSR in patients with heart failure may provide a naturally occurring condition in which to distinguish the effects of arousals from sleep on BP and HR from those due to respiratory stimuli. This is because in these patients the apnea-hyperpnea cycle and arousals from sleep can be dissociated as the cycle can occur during wakefulness, precluding arousals from sleep (18, 19). Furthermore, arousals from sleep can occur at different times in the ventilatory period. Thus, their effects may be assessed by comparing changes in BP and HR following arousals with changes during the equivalent time period in apnea-hyperpnea cycles in which arousals do not occur at that time. Accordingly, the objectives of our study were, first, to determine in patients with congestive heart failure whether fluctuations in BP and HR occur in synchrony with the apnea- hyperpnea cycle during CSR and, second, to determine to what extent any such fluctuations are a consequence of arousals from sleep.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Inclusion criteria were patients with congestive heart failure (CHF) with CSR and central apneas diagnosed on an overnight polysomnogram. CHF was documented by at least one episode of pulmonary edema, a left ventricular ejection fraction of < 35% at rest measured by radionuclide angiography, exertional dyspnea (New York Heart Association class 2-4) despite optimal medical therapy, and clinical stability for at least 1 mo. Patients were defined as having CSR if they had a crescendo-decrescendo pattern of tidal volume (VT) during hyperpnea regularly alternating with central apneas and hypopneas at a rate of > 10/h during both wakefulness and sleep. Participants were not permitted any stimulants, including caffeinated beverages, for at least 24 h or sedatives for at least 48 h before experiments. Written informed consent was obtained from all the patients and the experimental protocols were approved by the Human Subjects Review Committee of the University of Toronto.

Polysomnographic and Physiologic Monitoring

On the experimental night, patients were kept awake for at least 30 min prior to lights out in order to collect data during wakefulness as well as during sleep. Each patient had a routine overnight sleep study performed as previously described (20). Wakefulness and sleep stages were identified by electroencephalogram (C3/A2; C4/A1), electrooculogram, and submental electromyogram recordings obtained from surface electrodes and scored according to standard criteria (21). Arousals were defined according to the American Sleep Disorders Association criteria (22), having a minimum of 3 s of continuous alpha activity. The maximum period of alpha activity was < 15 s. All studies were performed with patients supine. The electrocardiogram was monitored from a precordial lead, from which HR was derived. BP was measured by finger photoplethysmography (Finapres 2300; Ohmeda, Englewood, CO). The arm and hand were maintained in the horizontal position at the level of the heart by a splint. The finger cuff was deflated for 5 to 10 min every half hour to prevent discomfort and arousals. It was also deflated if there was a spontaneous awakening lasting more than 2 min, until the patient returned to sleep. Rib cage and abdominal motions were assessed by respiratory inductance plethysmography (Respitrace; Ambulatory Monitoring, White Plains, NY). VT was taken as the electrical sum of the rib cage and abdominal displacements, which was calibrated against a spirometer as previously described (20, 23). Minute ventilation (VI) was calculated as the product of VT and respiratory frequency. Central apneas were defined by the absence of VT excursion for at least 10 s in the absence of thoracoabdominal movement (19, 23). Oxyhemoglobin saturation (Sa_O2) was continuously measured by an ear oximeter (Oxyshuttle, Sensormedics, Anaheim, CA). All variables were recorded onto a computerized data acquisition and analysis program (Sandman, Mallinckrodt Inc., Minneapolis, MN).

Data Analysis

Classification of apnea-hyperpnea cycles and arousals. Periods of CSR during wakefulness and non-rapid eye movement (NREM) sleep during which BP was recorded were identified. Apnea-hyperpnea cycle was defined as the period from 5 heart beats before the onset of central apnea until 30 heart beats after the onset of hyperpnea. Beat by beat HR and systolic and diastolic BP were then analyzed for this interval for each cycle.

Apnea-hyperpnea cycles were classified into three types: (1) awake (AW) cycles that occurred entirely during wakefulness and, by definition, did not include an arousal, (2) early arousal (EA) cycles during NREM sleep in which arousals from sleep occurred within 3 s before or after termination of central apneas, and (3) late arousal (LA) cycles during NREM sleep in which arousals from sleep occurred at least 8 s following the termination of central apneas.

Heart rate, blood pressure, and respiratory analyses. Each apnea- hyperpnea cycle was divided into three phases: (1) the preapnea phase consisting of the last 5 heart beats at the end of the hyperpnea that preceded the onset of the apnea, (2) the apnea phase, and (3) the postapnea phase consisting of the 30 heart beats following the onset of hyperpnea. To determine the amplitude of BP and HR oscillations for each cycle type, the data were smoothed by calculating 3 point running averages. The cycle amplitude was defined as the difference between the lowest and highest values over the cycle that invariably occurred during apnea and hyperpnea, respectively. Amplitudes of the 3 cycle types were then compared by a one-way analysis of variance.

To more precisely assess the effect of arousals on BP and HR, data were averaged over the 10 beats before (prearousal) and 10 beats after (postarousal) the onset of early and late arousals. These changes in BP and HR from pre- to postarousal for early and late arousals were then compared among cycle types by analyzing data from the same time intervals centered around the equivalent time points in cycles where arousals did not occur at these times. For example, if a late arousal occurred 10 s after the termination of apnea, pre- and postarousal BP and HR changes for this type of cycle would be compared with the same time intervals referenced to a point 10 s after the termination of apnea in AW and EA cycles in which an arousal did not occur at that time, as illustrated in Figure 1. The analyses were conducted separately for early and late arousals. Comparisons pre- and postarousal and comparisons among cycle types were performed using a 2 (pre- and postarousal) by 3 (cycle types AW, EA, and LA) repeated measures analysis of variance.

View larger version (27K): [in this window] [in a new window]   View larger version (24K): [in this window] [in a new window]   View larger version (29K): [in this window] [in a new window]   Figure 1.   Effect of late arousal on blood pressure (BP) during Cheyne-Stokes respiration. Recordings are from wakefulness (A), and from stage 2 sleep (indicated by K-complexes on the EEG tracing during apneas) with an early arousal (B) and with a late arousal (C ). Late arousal (indicated by an arrow in C at the time of a K-complex and an increase in EMG activity) occurs five breaths (10 s) following termination of the central apnea and is followed by a sudden increase in VT associated with an abrupt increase in BP. The effect of the late arousal on systolic BP was determined by comparing the average systolic BP of the 10 heart beats before with that of the 10 heart beats after the onset of arousal. This change was then compared with the average change in systolic BP of 10 heart beats before and after the equivalent time points during which arousals did not occur 10 s after the onset of hyperpnea (indicated by the arrows) during wakefulness (A) and stage 2 sleep with an early arousal (B). Note the absence of abrupt changes in BP at this time in A and B. Early arousal in B is indicated by the increase in EMG and alpha activity. These recordings illustrate that whereas late arousal had an effect on BP, early arousal did not. They also illustrate that the gradual rises in BP during hyperpnea paralleled gradual rises in ventilation, and that these rises in BP could occur in the absence of arousal, as shown in A during wakefulness. EEG, electroencephalogram; EMG, submental electromyogram; ECG, electrocardiogram; VT, tidal volume; SaO2, oxyhemoglobin saturation; BP, blood pressure.

We then performed analyses to determine whether any postarousal rises in systolic BP were related to concomitant rises in VI. First, VI was averaged over the first three breaths following apnea. The difference between this value for EA cycles with the values for AW and LA cycles assessed the effect of early arousals on VI. Second, VI was averaged for the first three breaths following the onset of late arousals in LA cycles and the equivalent time periods for AW and EA cycles. This assessed the effects of late arousals on VI in LA cycles. Postarousal changes in systolic BP were averaged over early and late arousals and were regressed against changes in VI.

The magnitude of the oscillation in Sa_O2 was calculated by subtracting the lowest from the highest value within each cycle. The mean oscillation amplitude was compared over the three cycle types using a one-way analysis of variance. To determine whether there was any relationship between changes in systolic BP and changes Sa_O2, the maximum amplitudes of the systolic BP oscillations for all subjects were averaged for all cycle types and were correlated with average oscillations in Sa_O2.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Characteristics of the Subjects

Eight consecutive patients (one female and seven males) with CSR both awake and asleep on a diagnostic sleep study participated in the study. Their average age was 69 ± 19 yr, their mean body mass index was 25.2 ± 6.4 kg/m², and their left ventricular ejection fractions averaged 15.1 ± 5.6%. Seven of the patients were taking digoxin, six diuretics, six angiotensin-converting enzyme inhibitors, one an anticoagulant, and two nitrates. Two of the subjects had pacemakers and thus HR data were not available for them. Baseline polysomnograms showed a mean total sleep time of 3.55 ± 1.29 h, arousal frequency of 33 ± 9 per h of sleep, central apnea frequency of 36 ± 18 per h of wakefulness and sleep, and a mean of the lowest Sa_O2 of 79 ± 8%.

Effects of Cycle Type and Arousal on BP, HR, VI, and Sa₀₂

On the study night all patients had at least two types of cycles, with four patients having all three types. Thus, seven subjects had AW cycles, with the average number of cycles being 11.7 (range 3 to 22), seven subjects had EA cycles (10.3, range 6 to 17), and six subjects had LA cycles (4.8, range 4 to 8).

Polysomnographic recordings illustrating BP oscillations during CSR in wakefulness and in stage 2 sleep with early and late arousals are shown in Figure 1. An example of the average oscillation in systolic BP over the three apnea-hyperpnea cycle types for one representative patient is illustrated in Figure 2. Minimum BP was reached during apneas and maximums during ventilatory periods. These data indicate that there were large oscillations in BP for all three cycle types. The oscillations in BP paralleled oscillations in ventilation, although with a slight lag, such that the rise in ventilation preceded the increase in BP, just as in experimental CSR (17). Furthermore, these oscillations in BP were not dependent on arousals, as substantial BP oscillations were observed during AW cycles. There was no significant difference between mean systolic BP oscillations between AW (11.3 ± 6.0 mm Hg) and EA cycles (13.7 ± 7.0 mm Hg). However, systolic BP oscillations during LA cycles were greater than during AW cycles (14.2 ± 7.1 mm Hg, p < 0.05). Nevertheless, of the 14.2 mm Hg oscillation in systolic BP swings with LA cycles, only 3 mm Hg (i.e., 21%) could be attributed to arousals. Substantial oscillations in diastolic BP were also observed during CSR, but there were no significant differences in the magnitudes of these oscillations among AW, EA, and LA cycles (8.8 ± 3.8, 9.7 ± 4.2, and 9.8 ± 3.7 mm Hg, respectively). Accordingly, oscillations in VI during the apnea-hyperpnea cycle, not arousals, appeared to be the major contributor to the large oscillation in BP during AW, EA, and LA cycles in these individuals. The two subjects with pacemakers did not differ from the other subjects in the amplitude of their systolic BP oscillations.

View larger version (18K): [in this window] [in a new window]   Figure 2.   Average oscillations in systolic blood pressure (BP) over the apnea-hyperpnea cycle for awake (AW), early arousal (EA), and late arousal (LA) cycles in a representative patient. The data are presented from 5 heart beats before the onset of the apnea to 30 heart beats following the onset of the hyperpnea. The timing of arousals for EA and LA cycles is indicated by arrows. Note that systolic BP falls progressively over the course of the apnea, and rises progressively over the first 30 heart beats during hyperpnea. Methods for data presentation are described in the text.

The pattern of changes in HR was similar to those of BP, with the amplitude of the oscillation in LA cycles being slightly, but significantly higher than for AW and EA cycles (6.8 ± 6.3, 6.8 ± 6.4, and 8.0 ± 7.9 bpm for AW, EA, and LA cycles, respectively, p < 0.05). The mean dips in Sa_O2 over the apnea-hyperpnea cycle did not differ significantly among AW, EA, and LA cycles (7.5 ± 3.8, 8.2 ± 3.3, and 7.5 ± 2.4%, respectively).

The specific effects of arousals on BP and HR are presented in Table 1. There was a significant increase in both BP and HR from pre- to postarousal, at the time of both early and late arousals. However, this was not primarily because of arousals, but because both BP and HR increased in all cycle types as the hyperpnea progressed. Thus, at the time of early arousals the magnitude of the increase in systolic BP (Figure 3), diastolic BP, and HR was the same in all cycle types. There was, however, a significant effect of late arousals on systolic BP (p < 0.05) (Figures 1 and 4), but not on diastolic BP or HR. This observation is consistent with previous studies in showing that systolic BP responses to apnea and arousals are more pronounced than are diastolic BP responses (17, 26).

View larger version (13K): [in this window] [in a new window]   Figure 3.   Average systolic blood pressure (BP) for all eight patients during the 10 heart beats following arousals in early arousal (EA) cycles and for the equivalent time periods during awake (AW) and late arousal (LA) cycles where arousals did not occur. Values shown are changes in systolic BP compared with the average of the 5 heart beats during apnea before the early arousal. There were no significant differences among the three cycle types indicating no effect of early arousal on systolic BP.

View larger version (12K): [in this window] [in a new window]   Figure 4.   Average systolic blood pressure (BP) for the 10 heart beats following late arousals in late arousal (LA) cycles and for the equivalent time periods during awake (AW) and early arousal (EA) cycles where arousals did not occur. Values shown are changes in systolic BP compared with the average of the 5 heart beats before the late arousal. The increase in systolic BP following late arousals in LA cycles was significantly greater than in AW and EA cycles. This indicated that late arousals contributed to the increase in systolic BP during LA cycles (see also Figure 3). **p < 0.05 LA compared with AW and EA cycles.

There was no significant relationship between the magnitude of oscillations in systolic BP, VI, or Sa_O2. However, there was a significant correlation (r = 0.70, p < 0.05) between changes in systolic BP and changes in VI in response to arousals (Figure 5). In addition, peak VI during hyperpnea invariably preceded peak rises in systolic BP, which consistently occurred in mid-hyperpnea (Figure 1). Furthermore, the increment in ventilation from the period following early arousals to the period following late arousals was significantly greater for LA than for AW and EA cycles (15.9 ± 10.0 versus 8.1 ± 8.7 and 6.9 ± 8.5 L/min, respectively, p < 0.05). Thus, greater increases in systolic BP following late arousals than for AW and EA cycles was associated with greater increases in VI. These results support the possibility that increases in BP in CSR are respiratory related and are closely linked to increases in VI (17).

View larger version (10K): [in this window] [in a new window]   Figure 5.   Relationship between average changes in systolic blood pressure (BP) and average changes in ventilation following arousals from sleep (both early and late) in each of the eight patients. In general, the more pronounced the ventilatory responses to arousals the more pronounced the increases in systolic BP (r = 0.70, p < 0.05).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the present study, we observed pronounced fluctuations in BP and HR in heart failure patients with CSR, similar to those described in previous reports (4, 18). As illustrated in Figures 1 and 2, these fluctuations were synchronous with the CSR pattern such that BP and HR fell during apnea and rose during hyperpnea. These data indicate that oscillations in BP and HR are entrained by CSR just as they are during voluntary periodic breathing in healthy subjects (17). Heart failure patients without CSR do not have such periodic fluctuations in BP and HR (26, 27). Our study design differed from previous ones, however, in that it allowed us to separate the effects of the CSR cycle itself on BP and HR from the effects of arousals and hypoxia. The most important and novel observation of the study was that most of the change in BP and HR was related to the apnea-hyperpnea cycle itself, rather than to arousals or hypoxia. This was most convincingly demonstrated by the pronounced fluctuations in BP and HR during the CSR cycle when patients were fully awake in the absence of arousals from sleep (Figure 1). The amplitude of the oscillation in systolic BP and HR, but not diastolic BP, was only slightly larger following late, but not early arousals from NREM sleep. Furthermore, the correlation between the increases in systolic BP and VI following arousals (Figure 5) indicated that the small effect attributable to arousals was likely to have been mediated through respiratory activity. Finally, the critical role played by ventilation is emphasized by the observation that peak increases in ventilation during hyperpnea invariably preceded peak increases in BP in the presence or absence of arousals. Thus, in patients with CSR, periodic oscillations in BP are primarily related to oscillations in VI. Although arousals played a minor role in augmenting systolic BP, they are not essential to the development of periodic oscillations in BP and HR during CSR.

The degree of apnea-related O₂ desaturation did not correlate with the degree of increase in BP or HR over the apnea- hyperpnea cycle. This finding is consistent with observations in two recent studies. Franklin and coworkers (4) found, in CHF patients with CSR, that correction of hypoxia by supplemental O₂ did not affect the degree of elevation in BP following termination of central apneas. Ringler and coworkers (2) made similar observations following termination of obstructive apneas. These investigators concluded that hypoxia alone does not explain BP elevations following termination of either central or obstructive apneas. Our data strongly support their conclusions.

The present findings are consistent with those of a recent study in which we demonstrated that oscillations in VI during voluntary periodic breathing in awake subjects caused fluctuations in BP and HR similar to those described herein, in the absence of arousals and hypoxia (17). These respiratory-related fluctuations in BP and HR at the frequency of the CSR cycles might be analogous to breath-to-breath fluctuations in BP and HR that occur during normal breathing (i.e., respiratory sinus arrhythmia). Alterations in cardiac preload and afterload due to the mechanical effects of changing intrathoracic pressure and lung volume are partly responsible for these breath-to-breath changes in BP and HR (28). Therefore, oscillations of venous return and cardiac output caused by periodic rises and falls in ventilation and pleural pressure during periodic breathing might also lead to fluctuations in BP. The recent observation that cerebral blood flow and BP oscillate in concert with VI during CSR is in keeping with this possibility (4).

The question then arises as to what drives cyclic ventilatory oscillations in CSR. Fluctuations in Pa_CO2 below and above the apneic threshold are the most likely possibility (20, 23, 29). Thus whereas correction of hypoxia during CSR causes only minor reductions in the frequency of central apneas, and does not affect fluctuations in BP (4), inhalation of CO₂ abolishes central apneas and damps oscillations in Pa_CO2 and BP (29, 30). Accordingly, it would appear that cyclic fluctuations in chemical stimuli, central respiratory outflow, and/or ventilation are at least partly responsible for cyclic fluctuations in BP and HR during CSR (4).

Late arousals were associated with significant, but small increases in BP and HR, whereas early arousals were not. One possible explanation for this finding, suggested above, is that the greater increment in VI following late than early arousals caused more pronounced increases in BP and HR (17). This might in turn be related to more intense chemostimulation (mainly Pa_CO2) at the point of late rather than early arousals owing to prolonged lung to chemoreceptor circulatory delay in patients with heart failure (31).

Two forms of evidence are cited in support of a direct effect of arousals from sleep on BP and HR. The first is that arousals induced by external stimulation in normal individuals are associated with increases in BP and HR (5). It is assumed that such spontaneous arousals are not associated with heightened respiratory stimuli. However, this is not the case as reduced responsiveness to respiratory stimuli leads to decreased Pa_O2 and increased Pa_CO2 during NREM sleep. Therefore, arousals from sleep, however elicited, will be associated with elevated respiratory drive when the wakefulness drive to breathe is reinstituted. Further, it remains unclear whether external stimuli presented during sleep are neutral with respect to cardiovascular activity. That is, it may be that the external stimulus, rather than the arousal itself, is generating the cardiovascular response. The observation that BP can increase in response to acoustic stimuli that do not provoke a cortical arousal is consistent with this hypothesis (7).

The second form of evidence cited in support of a direct effect of arousal from sleep is that obstructive apneas still provoke increases in BP in concert with arousal, even when the hypoxia associated with the apnea is relieved (2). However, O'Donnell and colleagues (11) demonstrated, in a dog model of obstructive sleep apnea, that postapneic surges in BP occurred even when apnea was terminated prior to arousal. Thus, although arousals at the termination of obstructive apneas may have different functional significance than arousals following central apneas, the data of O'Donnell and coworkers are consistent with ours and suggest that the postapneic rise in BP was, to some extent, linked to the increase in VI, rather than to arousal per se.

In the present study arousals were defined according to ASDA criteria. During NREM sleep we analyzed only those apnea-hyperpnea cycles in which a clear-cut EEG arousal was present. This strategy had the advantage of maximizing the chances of observing cardiovascular responses to arousal. Thus it is not the case that the relatively small BP and HR responses to EEG arousals were because we were examining cardiovascular responses to subtle, non-EEG arousals (7). Moreover, it is an essential point in this study that fluctuations in BP and HR in synchrony with the apnea-hyperpnea cycle were observed while subjects were awake, excluding a role for either EEG or non-EEG arousals in these cardiovascular responses. These findings are consistent with those of a recent paper in which we have demonstrated, in normal subjects, that voluntary periodic breathing with central apneas while awake produces oscillations in BP and HR that duplicate those observed during CSR in the present study (17).

It is possible that the present results are specific to the patient group studied. Thus, one interpretation of the data is that cardiovascular arousal mechanisms are impaired in patients with CHF, either because of their underlying cardiac disease or because of BP-lowering medications. Against these possibilities are two observations. The first is that wide fluctuations in BP and HR occurred in synchrony with the apnea-hyperpnea cycle, with or without arousal, indicating the capacity for BP and HR to vary widely in these individuals. The second was that late arousals did cause significant increases in BP. Thus the absence of significant BP elevations following early arousals has to be interpreted in light of the demonstrated capacity of the cardiovascular system to respond to late arousals. To determine the generalizability of the present results it would be of interest to test cardiovascular responses to spontaneous respiratory-related arousals in a group of patients with central sleep apnea but without heart disease, such as those with idiopathic central sleep apnea (20). However, because patients with idiopathic central sleep apnea do not have apneas while awake, comparisons of apnea-hyperpnea cycles while awake and asleep, as employed to advantage herein, could not be undertaken.

The sample size of the study was relatively small, reflecting the availability of patients who met the selection criteria and who were able to participate in the study night. However, it is unlikely that the basic results of the study were compromised by lack of power. The synchronous oscillations in BP and HR occurred in every subject in each cycle type and were strikingly similar to those provoked by experimentally induced periodic breathing in human subjects while awake (17). In addition, although the arousal effect may have been found to be more robust with a larger sample size, the basic finding of the study was not that the arousals had no effect, indeed we found a significant effect of late arousals on BP, but that the effect was minor compared with the major effect of the respiratory cycle.

The present findings demonstrate that in patients with heart failure, moderately large increases in BP and HR occur synchronously with the ventilatory portion of the CSR cycle. The clinical significance of this finding is not certain, but surges in BP during hyperpnea may be one factor related to the poorer prognosis in patients with heart failure with CSR, compared with those without it (32, 33). Our data lead us to two further conclusions. First, arousals provoked by intrinsic respiratory stimuli during CSR in NREM sleep have relatively little effect on BP and HR, unless they are accompanied by increases in ventilation. This contrasts with the marked increases in BP and HR that occur in response to extrinsically induced arousals (5, 7, 15). These findings imply that arousals from sleep arising from different stimuli have different effects on the cardiovascular system. Therefore, caution should be exercised in extrapolating the effects of extrinsically induced arousals on BP and HR, to those arising from intrinsic respiratory stimuli during CSR or other forms of sleep apnea. Second, our data strongly suggest that BP and HR oscillations in CSR are respiratory related, and that cycle-by-cycle fluctuations in ventilation play an important role in their genesis. The mechanism(s) by which cyclic oscillations in ventilation provoke cyclic oscillations in BP and HR could be explored, for example, by monitoring of beat-by-beat cardiac output during CSR, and use of experimentally induced CSR in patients with denervated lungs and in the presence of sympathetic nervous blockade.