Ventilatory Instability in Patients With Congestive Heart Failure and Nocturnal Cheyne-Stokes Breathing

Transcription

1 Sleep, 17(6): American Sleep Disorders Association and Sleep Research Society Ventilatory Instability in Patients With Congestive Heart Failure and Nocturnal Cheyne-Stokes Breathing Mansoor Ahmed, C. Serrette, M. H. Kryger and N. R. Anthonisen Section of Respiratory Disease, Department of Medicine, University of Manitoba, Winnipeg, Manitoba, Canada Summary: Many of the factors that appear to cause Cheyne-Stokes Breathing (CSB) in sleeping patients with congestive heart failure (CHF) are present during wakefulness. We studied the stability of ventilatory pattern in nine awake CHF patients (left ventricular ejection fraction 9-48%) who demonstrated CSB only while asleep and compared results with 13 age-matched normals. The test involved brief (3D-SO-second) exposure to hypoxia (endtidal P0 2 = 55 Torr) followed by breathing pure oxygen. During hypoxia, ventilation increased about 40% above air breathing control in both groups, whereas end-tidal CO 2 declined to 92% of control in both groups. During hyperoxia, however, breathing pattern differed between groups. In the normals, ventilation gradually declined to air-breathing levels and did not significantly undershoot. In the patients, ventilation dropped more rapidly to baseline and an overshoot was present with ventilation being 72% and air-breathing control at 45 seconds of hyperoxia. Circulatory delay was calculated from the time interval between alveolar hypoxia and in increase in ventilation, and when corrections for circulatory delay were applied to ventilation during hyperoxia the differences between groups increased in that the patients' ventilation was less than baseline immediately after the delay. In the normals, the gradual decline in hyperoxic ventilation probably represents the decay of short-term potentiation (STP) activated by hypoxic hyperventilation. Results in the patients were compatible with absence of such STP decay, but could also have been due to a reduction in ventilatory drive early in hyperoxia related to prolonged circulation times. In either case, awake patients with CHF and nocturnal CSB demonstrated decreased ventilatory stability in response to transient hypoxia, which may relate to their abnormal breathing at night. Key Words: Hypoxia-Afterdischarge-Short-term potentiation. Instability of ventilation with cyclic breathing patterns occurs more often in sleeping than in waking humans. This probably relates to a number of changes occurring with sleep that act in the direction of reducing ventilatory stability. Gas stores are reduced due to the reduction in functional reserve capacity (FRC) imposed by recumbency (1). Cardiac output falls (2) and circulatory delay may increase. The response to mechanical loads is impaired (3), and the resistive load imposed by the upper airway increases and may become more variable. The apneic threshold apparently increases with sleep (4), and there is evidence that respiratory short-term potentiation (STP decay, afterdischarge) is reduced during sleep (5). Nevertheless, sustained cyclic breathing during sleep Accepted for publication May Address correspondence and reprint requests to N. R. Anthonisen, M.D., Ph.D., Dean of Medicine, Faculty of Medicine, University of Manitoba, 753 McDermot Avenue, Winnipeg, Manitoba, Canada R3E OW is uncommon in normal humans under normal conditions (6). Patients with congestive heart failure (CHF), on the other hand, frequently demonstrate sustained nocturnal Cheyne-Stokes breathing (CSB). This has been the subject of recent reviews (7,8) that cite a number of possible mechanisms. FRC may be abnormally reduced in CHF, although this has not been demonstrated. Patients with CHF may be hypoxemic, which has the effect of increasing the gain of ventilatory controller because the response to a given decrease in P0 2 increases as P0 2 falls, which favors cyclic breathing. Also, patients with CHF may be hypocapnic as well, and therefore operating relatively near their apneic threshold. However, the most widely cited and best understood physiological change in CHF that contributes to CSB is the prolonged circulation time. This prolongation tends to throw the ventilatory controllers out of phase with events in the lung, the controllers reacting to events that occurred some time ago in the lung.

2 528 M. AHMED ET AL. Whatever the individual contributions of these various factors, CSB is thought to occur in response to physiological stimuli that are inadequately damped. We hypothesized that patients with CHF and nocturnal CSB would demonstrate instability oftheir awake ventilatory pattern if stimulated appropriately. We therefore examined a group of these patients using a test that we designed to demonstrate STP decay in normals, which imposes rapid application and withdrawal of an hypoxic stimulus to ventilation. METHODS Nine male patients with documented histories of congestive cardiomyopathy and CSB who were less than 65 years old were selected from our out-patient clinic. They had clinical diagnoses of stable CHF (New York Heart Association class 3-4) and left ventricular ejection fractions (LVEF) of less than 50%, as measured by radionuclide scan. Complete histories and physical examinations were obtained in all to rule out the presence of neurologic disease or significant primary lung disease. In each, a chest x-ray, electrocardiogram (ECG) and pulmonary function tests, including awake arterial blood gases, were obtained. Patients with recent (within 3 months) hospitalization for their underlying condition, or who were otherwise unstable, patients with airways obstruction and those who were using sedatives were excluded from the study. As a control group we recruited 13 subjects (nine male) from a local fitness center. Their ages ranged from 57 to 67, the average being 62. None were obese or hypersomnolent, and they all had a negative medical history and physical examination plus normal spirometry and ECG. The study was approved by the Hospital Ethics Committee and all patients gave their written informed consent prior to study entry. Sleep studies Nocturnal polysomnographic studies were conducted in the CHF patients to document the presence CSB. We recorded the electroencephalogram, electroocculogram, and mental electromyogram from surface electrodes. Arterial oxygen saturation (Sa02) was recorded continuously using a pulse ear oximeter (Biox 30, Ohmeda, Boulder, CO, U.S.A.) set on its fastest response. Respiratory movements were monitored by respiratory inductance plethysmograph (Respitrace, Ambulatory Monitoring, Ardsley, NY, U.S.A.). ECG and heart rate were continuously recorded from standard limb leads. Airflow was detected by monitoring expired CO2 at the nose and mouth, using a CO2 gas analyzer (Datex 223, Puritan-Bennett Corp., Overland Park, KS, U.S.A.). All variables were recorded continuously on a multichannel polygraph. Sleep was staged according to the usual criteria. Apnea was defined as the absence of airflow for more than 10 seconds. Hypopnea was defined as a decrease in respiratory movements to 50% or less of the maximum amplitude observed during the preceding minute. In the case of cyclic breathing, this meant that hypopnea occurred when breathing amplitude was 50% or less of the maximum amplitude of the cycle. CSB was defined as regular cyclic changes in the amplitude of breathing movements (9). The number of complete cycles were counted in each patient and divided by the total sleep time to give an average number of cycles per hour. Hypoxic-hyperoxic ventilatory responses Neither the patients nor the controls were aware of the physiological purpose of the study. They were instructed not to smoke and to have no drinks with caffeine for at least 8 hours before the study. The methods used have been described in previous publications (10,11). During the experiments the patients were seated in a comfortable chair, distracted with nonrhythmic music and encouraged to read. They were observed closely and did not sleep during experiments. The ECG was monitored continuously as was Sa02 using the ear oximeter. Wearing nose clips, the subjects breathed through a low resistance unidirectional valve, the inspiratory side of which was connected to a pneumotachograph, which was in turn connected through large bore tubes to a manifold offering the choice of four inspirates: room air, oxygen, nitrogen and 8.5% oxygen in nitrogen. The lumens of the manifold tubes could be occluded by inflating small balloons so the inspirate could be changed without auditory clues alerting the subjects. The dead space of the device from mouth to the tube entrance was approximately 150 ml The partial pressures of end-tidal O2 (PET o2 ) and CO2 (PET C02) were measured by a mass spectrometer (Perkin-Elmer 1100 Don Mills, Ontario, Canada) sampling gas at the mouth piece. The pneumotachograph output was integrated to give tidal volume (V T ). The pneumotachograph was calibrated with both air and O2 and the latter calibration used during hyperoxic breathing. V T, Sa02, PET Oz and PET C02 were recorded continuously on a strip chart recorder. The subjects breathed room air until ventilation and end-tidal gases were stable. The inspirate was then changed to N2 for 2-3 breaths during which PET o2 decreased rapidly to about Torr. The inspirate was then changed to 8.5% O2, which maintained PET o2 at approximately 55 Torr. After seconds, hypoxia was terminated abruptly by switching the inspirate to

3 VENTILATORY INSTABILITY IN CHF 529 PET02, Torr-o- PET O 2, Torr-o- 110 ~ q t b" PT1 VI % Cont -.- VI % Cont /t, Q tj, t tj-o PT TIME (sec) FIG. I. Calculation of circulatory delay. Shown are means of breathby-breath end-tidal P0 2 and minute ventilation during the onset of hypoxia. The lower arrows indicate the time at which PETo = 75 Torr, and the upper arrows when VI = 110% of control. Th~ time between the two was the circulatory delay or the time taken for an hypoxic stimulus to reach the carotid bodies. Data from the patients with the longest and shortest delays are shown. 100% oxygen so that PET 02 of the first hyperoxic breath was at or above nom oxic baseline and that of the second was above 150 mm Hg. Hyperoxia was maintained for 2 minutes, and the inspirate then again was changed to room air. When PET 02 and PET e02 returned to control levels, the same hypoxic-hyperoxic exposure was repeated to total of four in each subject. Runs associated with swallowing or coughing during the critical period of transition from hypoxia to hyperoxia were excluded from the analysis. Minute ventilation (VI)' V T, inspiratory time (Ti), expiratory time (Te) and total breath duration (TTOT), PET 02 and PET C02 were measured breath by breath for 30 seconds during room air breathing (baseline) prior to hypoxia, during hypoxia and for 2 minutes after switching to 100% oxygen. VI and VT during hypoxia and hyperoxia were expressed as a percentage of the mean air-breathing value. In each subject hypoxic and hyperoxic VI and V T from different runs were averaged according to breath number: first hypoxic breath, etc. These were converted to a time base using the mean Ti and T TOT for each breath, VI and V T being assumed to occur at the end of Ti. To compare groups, curves were drawn by eye through each individual's mean data, and values were interpolated at 5-second intervals. These were in turn averaged to obtain mean data for the group. Hyperoxic ventilation was analyzed both before and after discarding data at the beginning of the hyperoxic period to allow for circulatory delay. Circulatory delay was computed by examining measurements during the beginning of hypoxic exposure, based on the assumption that there was little hypoxic stimulus until PET 02 = 75 Torr. The time lag between this PET 02 and a 10% increase in VI was computed for each VI % Control L---~---L--~----L---~--~ o TIME (sec) FIG. 2. Hyperoxic ventilation with and without corrections for circulatory delay in the patients shown in Fig. 1. Mean breath-bybreath minute ventilation expressed as a percentage of air-breathing control is plated against duration of hyperoxia. Arrows indicate the end of the calculated circulatory delay in each patient. subject (Fig. 1). This time lag was added to the Ti of the first hyperoxic breath and the sum taken as zero time for analysis of hyper oxic ventilation after correction for delay (Fig. 2). We also examined nadir values for hyperoxic ventilation in both groups. The breath-by-breath record of each individual was examined and the breath with the smallest ventilation (V TIT TOT) identified. In an effort to take account of random one-breath nadirs, the ventilation of this breath was averaged with the ones that immediately preceded and followed it, giving a mean three-breath nadir. Data were analyzed by nonparametric Mann-Whitney test and two-way analysis of variance (ANOV A). p < 0.05 was considered statistically significant. Values are expressed as mean ± SE. RESULTS Patient characteristics are shown in Table 1. The average age of our patients was 56, not significantly different from that of our controls. The patients were not hypoxemic when awake and arterial PC0 2 was normal. The average arterial ph was slightly elevated, however, indicating the presence of metabolic alka-

4 530 M. AHMED ET AL. 1 Patient Mean SEM Age (years) TABLE 1. Patient characteristics PaO z PacO z (Torr) (Torr) pha losis, probably secondary to diuretic therapy. Spirometry was normal in most subjects. Two (patients 1 and 2) had reductions of forced expiratory volume in 1 second (FEV!) and forced vital capacity (FVC) ofsimilar magnitude that were probably significant, and a third (patient 3) showed a similar pattern that was not clearly abnormal. Patients 1 and 2 had reduced diffusing capacities, but the average for the eight patients with available data was 86% of predicted normal. The LVEF averaged 22% with a range of 9-48%. The circulatory delays calculated from the hypoxic ventilatory response (Fig. 1) are also shown in Table 1. The average value for the patients was 12.4 seconds, significantly greater than the mean control value of 7.4 seconds, but several patients (Nos. 2, 3, 6 and 7) had delays that were not outside the range of normal. Circulatory delay correlated negatively but not significantly with LVEF (r = -0.48). Results of sleep studies are shown in Table 2. An patients slept more than 3 hours while under study, and 7/9 slept more than 5 hours. About half the sleep was spent in Stage 2 with another 19% in REM sleep, fractions which varied little among patients (Table 2). CSB occurred chiefly during stage 2 sleep and not during REM sleep. The apnea-hypopnea index varied from less than one to more than 30/hour and was always less than the average number of CSB cycles per hour, indicating that not all CSB cycles showed a more than two-fold variation in the amplitude of breathing movements. In some subjects (patients 4, 5 and 9) there was a striking discrepancy between the number ofcsb cycles per hour and the apnea-hypopnea index, implying relatively low-amplitude cycles, whereas in others (patients 1, 2 and 6) the two indices corresponded more closely, showing that most CSB cycles had at least a two-fold variation in breathing movements. There was a greater than la-fold variation in number of CSB cycles among patients, which did not relate significantly to either ejection fraction or circulatory delay. Hypoxic responses did not differ between patients and controls. End-tidal P0 2 fell to Torr in both FEV, FVC LVEF Delay (% predicted) (% predicted) (%) (seconds) groups, with end-tidal PC0 2 falling to 92% of the average prehypoxic value in both groups. Minute ventilation at the end of hypoxia averaged 140% of control prehypoxic values in the normals and 142% in the patients. In both groups the increase in ventilation was achieved by increasing tidal volume; breathing frequency did not change significantly in either group. Figure 1 shows the data used to calculate circulatory delay in the patients with the longest and shortest delays. In one patient ventilation increased 10% over 20 seconds after end-tidal P0 2 fell to 75 Torr, whereas in the other ventilation increased 7.5 seconds after the same end-tidal P0 2 was attained. Figure 2 shows hyperoxic ventilation in the same two patients after the circulatory delay was taken into account; both subjects showed a rapid decline in ventilation to levels below those observed during room air breathing. In Figure 3 mean hyperoxic ventilation in the control and CSB groups is shown, including all data from the onset of O 2 breathing. In the controls ventilation fell gradually, reaching baseline seconds after the onset of hype roxi a and never decreased below 89% of baseline. In the patients ventilation dropped more rapidly, reaching baseline in seconds. A distinct TABLE 2. Sleep study results Apneahypopnea index CSB Test % REM (events/ (cycles/ Patient (minutes) % Stage 2 sleep hour) hour) I Mean SEM

5 VENTILATORY INSTABILITY IN CHF ) VI 120 % Control 7, 100 ~ " 1>'-6..,.6....! ~ '1--~< ~.6-~i'~ y TIME (sec) FIG. 3. Mean hyperoxic ventilation in patients (open symbols) and controls (closed symbols). Ventilation is expressed as a percentage of air-breathing control values, and brackets are SEM. No correction for circulatory delay was made, so zero time is the onset ofhyperoxia. nadir averaging 72% of baseline occurred at 45 seconds, with ventilation subsequently increasing to levels near that of the normals. The two curves were significantly different in terms of position (p < 0.05, ANO Y A), but not in terms of shape. Figure 4 shows mean hyperoxic ventilation in the two groups after correction for circulatory delay. The differences between the two curves are larger than in Fig. 3. Again in the normal subjects there was a smooth and continuous decline in mean ventilation that did not reach baseline until 14 seconds after the end ofthe circulatory delay. In the patients, mean ventilation was less than the air-breathing baseline at the end of the circulatory delay and declined further to a nadir of approximately 75% of baseline at seconds, when ventilation in the control subjects was 92% of baseline. After the nadir the patients' mean ventilation increased toward that of the control subjects. The curves of Fig. 4 differed significantly both as to position and shape (p < 0.05, ANOYA). In both patients and normals, changes in hyperoxic ventilation were due to changes in tidal volume; in neither group did breathing frequency change significantly, though in two CHF subjects ventilatory nadirs were in part due to prolongations ofte. Three breath nadirs of ventilation averaged 71.8 ± 3.5% (SEM) of baseline in the patients and 87.6 ± 5.8% in controls, these values being significantly different (p < 0.05, Mann-Whitney U test). Further, when the individual patients were compared, the depth of the nadir observed during hyperoxic ventilation correlated negatively with the fraction of sleep time spent in CSB (Fig. 5); the deeper the nadir the more time the patient spent in CSB sleep. In the patients, nadirs occurred at seconds of hyper oxic breathing (mean VI % Control TIME (sec) FIG. 4. Mean hyperoxic ventilation in patients (open symbols) and controls (closed symbols) after correction for circulatory delay. Time zero therefore represents the end of the circulatory delay. Ventilation is expressed as a percentage of air-breathing control values and brackets indicate SEM. = 45.5 seconds, SD = 4.7 seconds) or seconds after the circulatory delay (mean = 31.1 seconds, SD = 5.7 seconds). In the normals nadirs were reached at seconds of hype roxi a (mean = 48.9 seconds, SD = 16.6 seconds), or seconds after the circulatory delay (mean = 37.0 seconds, SD = 16.0 seconds). Though the average times to nadir values did not differ between groups, variability was much less in the patient group, accounting for the fact that nadirs were evident in the average patient data (Figs. 3 and 4), but not in the average data of the controls. DISCUSSION We have examined ventilatory responses to brief hypoxia followed by hyperoxia in the past (10,11) and designed our breathing circuit to effect these changes as rapidly as possible. The success of this design is attested to by the fact that PET 0, exceeded 150 Torr with the second hyperoxic breath. A further advantage of our experimental setup was that changes in inspirate were made silently without alerting the subjects. Indeed, most of the subjects were unaware of changes in the inspirate and none were able to identify switches between hypoxia and hyperoxia. Thus we are confident that our results do not reflect behavioral changes in breathing. In our patients with CHF and CSB, post-hypoxichyperoxic ventilation decreased rapidly to values that were less than the air-breathing control, and this tendency was more striking when allowances were made for circulatory delay. In age-matched normals, posthypoxic hyperventilation declined gradually to baseline, as has been described previously (la, 11). Breath-

6 532 M. AHMED ET AL. ing patterns in the normals were more stable than in the patients, which might have related to the patients' nocturnal CSB. Several other issues should be considered before discussing this interpretation, however. We did not do sleep studies on our controls, simply assuming that they did not have sleep-related disorders. In fact these disorders are so common that we might have expected one or more of our controls to have had abnormal sleep studies. On the other hand, our control group was probably not at high risk for nocturnal breathing disorders. Only nine of 13 were male, and none were obese or hypersomnolent; they were recruited from a fitness center. It seems very likely that our "average" control subject did not have periodic breathing at night, whereas our average patient certainly did. A more serious problem is that we did not examine patients with CHF but without CSB. Therefore we cannot be certain that the abnormal patterns of post-hypoxic-hyperoxic ventilation we observed in our patients were related to their CSB and not a feature of CHF per se. However, in other patient groups (12, 13) similar instability of breathing patterns after a similar hypoxic stimulus was associated with central apneas during sleep. Further, in the present study the degree of posthypoxic hypo ventilation was related to the amount of CSB (Fig. 5). If these data are valid, one would expect patients with little or no CSB to demonstrate little hypoventilation after withdrawal of an hypoxic stimulus. Thus, we believe it likely that our patients's ventilatory instability while awake as a manifestation of the same factors that caused their nocturnal CSB. Aside from their increased circulatory delay, our patients demonstrated few of the characteristics thought to predispose to CSB (7,8). They were not hypocapnic; indeed, during air breathing PET CO2 averaged 37.6 Torr in our patients, nearly identical with the average PET C02 of 37.4 Torr demonstrated by our controls. Our patients were slightly alkalemic, probably because of diuretic therapy, but this has not been reported as a risk factor for CSB, and it is unlikely that this minor deviation would have influenced the control of ventilation in an important way. Our patients were not hypoxemic while awake, and only one (no. 7, Table 1) appeared to have an increased alveolar arterial O 2 difference. Lung function in our patients was generally normal, and given a normal vital capacity it would be surprising if other lung volumes were abnormally low in our patients. Finally, the average ventilatory response to hypoxia was closely similar between our patients and controls; there was no evidence of increased controller gain in our patients. Our calculation of circulatory delay depended on several assumptions. First, we assumed that PET 02 related to arterial P0 2 in a consistent way, the latter being about 10 Torr below Nadir VI % Control L---~--~~--~---L--~ % CSB FIG. 5. Nadir hyperoxic ventilation as a function of the fraction of sleep time spent in CSB. Nadirs are means of three consecutive breaths and are expressed as percentages of air breathing control values. Each point represents a single patient (r = -0.66, p < 0.05). the former. Second, we assumed that the Pa0 2 associated with PET 02 = 75 Torr elicited a 10% increase in ventilation. Obviously, these assumptions represented approximations, but their general validity is supported by the average circulatory delay of 7.4 seconds calculated in the normal controls, which agrees with the findings of others (14,15). Though our patients demonstrated a greater average circulatory delay than the controls, several individuals had values that overlapped the normals. It must be recalled that we computed delay in awake seated subjects and that delays might have changed with sleep, possibly more in the patients than in the normals. The time from the end of hypoxia, which usually coincided with peak ventilation during the hypoxichyperoxic exposures, to nadir values of ventilation could be conceptualized as a half cycle of a potentially cyclic breathing pattern. In controls no true nadir was observed, but in the patients there was a consistent nadir seconds after hypoxia. Double this value is at the upper end of the range of cycle lengths for CSB noted by others (16,17), and was considerably less variable than theirs. This cycle length, or time to nadir ventilation, did not correlate with circulatory delay as others have noted (16,17). However, we used delay times measured during wakefulness and compared them with an artificially induced peak to nadir ventilation change, whereas others measured both delay and cycle length during spontaneous CSB in sleeping patients. Under the latter circumstances cycle length and circulatory delay should in theory correlate well (14); this might not necessarily be the case in experiments such as we conducted. The most important observations we made were of Sleep. Vol. ]7. No

7 VENTILATORY INSTABILITY IN CHF 533 Nadir VI % Control DELAY FIG. 6. Nadir hyperoxic ventilation, expressed as a percentages of air-breathing control values, as a function of circulatory delay in seconds. Each point represents a single patient (r = -0.71, p < 0.05). breathing immediately after the hypoxic ventilatory stimulus was "turned off" by hyperoxia. We (10,11) and others (18) have interpreted the gradual decline in ventilation observed in normals under these circumstances as evidence of STP, but other events of significance to the control of ventilation probably were also occurring during the time interval considered (19). First, the control subjects were hypocapnic at the end of hypoxia and during hyperoxia there was a gradual increase in CO 2 in the blood and brain, though PET CO2 did not reach control levels for 30 seconds. Second, cerebral blood flow was probably increased during hypoxia and decreased sharply with the onset of hype roxi a, which would tend to decrease the effects of arterial hypocapnia on ventilation. Third, if there were a buildup of inhibitory neuromodulators such as adenosine during hypoxia, this would tend to be reversed during the subsequent hyperoxia, tending to increase ventilation. Finally, in animal experiments that used powerful stimuli to activate STP, there was a "step down" in ventilatory output when the stimulus stopped and STP decay began (19). If this occurred after our much weaker hypoxic stimulus, the step down might have been included in the ventilatory decay we analyzed. Thus, post-hypoxic-hyperoxic ventilation in our subjects almost certainly reflected a number of influences besides STP, and it is important to analyze whether these differed between our controls and the patients with heart failure and CSB. There is no a priori reason to believe that a step decrease in ventilation of the kind observed in animals should have differed between controls and patients because the response to hypoxia was similar in both groups. However, events in and around medullary centers of ventilatory control probably did differ between groups due to the prolonged circulation time of the patients with CSB. Endtidal PCOz was similar in both groups at the end of hypoxia. It is therefore probable that central PCOz was higher in the patients than in the controls, and that in the patients, central PCO z continued to decline early in hyperoxia, while tending to increase in the controls. Similarly, a hyperoxic decrease in cerebral blood flow would have occurred later in the patients than in the controls, as would any decrease in central adenosine concentrations mediated by hyperoxia. These effects would tend to decrease ventilatory drive early in hyperoxia, in patients as opposed to controls, by prolonging central hypocapnia and possibly by allowing persistence of a central depressant agent. Therefore, the differences noted in Fig. 3 between normals and CHF patients might reflect reduced central drive in the patients due to their prolonged circulation time, and this reduction in drive might have cancelled out any STP that was present. If this were the case, then one would expect correction for circulatory delay to decrease the difference between groups. The opposite occurred; after correction for circulatory delay the difference in hyperoxic ventilation between normals and patients increased (Fig. 4). This supports the hypothesis that activation ofstp or its subsequent decay was abnormal in these patients, but the argument assumes that our corrections for delay were accurate. Because we cannot be certain that they were accurate, we cannot be sure that the differences in ventilation of Fig. 4 are due to differences in STP decay as opposed to differences in circulation times. If the patients' patterns of hyperoxic ventilation were due largely to their prolonged circulation time, one would expect the circulatory delay we measured to relate to hypoxic nadirs of ventilation, which was the case (Fig. 6). Because the nadirs correlated with the amount of CSB observed (Fig. 5), the results are consistent with both the CSB and the post-hypoxic-hyperoxic breathing pattern being caused by prolonged circulation times. However, circulatory delay did not correlate significantly with the amount of CSB either in the present experiment or those conducted by others (16,17). Further, it should be noted that the significant correlations we did find were very dependent on data from three subjects (patients 1, 4 and 5; Table 1) who had the longest circulatory delays, the most CSB and the lowest nadirs. Our normal subjects did not hypoventilate during post-hypoxic hyperoxia, although they were also hypocapnic. Their breathing patterns were relatively stable probably due in large part to STP decay or afterdischarge. Because STP is also demonstrable in sleeping normal subjects (20), it is reasonable to argue that it tends to prevent cyclic breathing during sleep in such individuals. By contrast, our patients with heart failure tended to hypo ventilate after withdrawal of a brief hypoxic ventilatory stimulus, and therefore would seem

8 534 M. AHMED ET AL. more likely to develop cyclic breathing after a transient stimulus. Though they did not do so while awake, they demonstrated CSB while asleep, when a variety of other factors such as reduced lung volumes and cardiac output favor periodic breathing. It seems possible that the unstable responses we observed in these patients while awake contributed to their breathing patterns while asleep, and it would be of interest to study patients with CHF but without CSB in the same way. Acknowledgement: from MRC Canada. This work was supported by grants REFERENCES I. Agostoni E, Hyatt RE. Static behaviour of the respiratory system. In: Macklem PT, Mead J, eds. Handbook of physiology. The respiratory system, vol. 3. Bethesda, MD: American Physiological Society, 1986: Phillipson EA, Bowes G. Control of breathing during sleep. In: Cherniack NS, Widdicombe JG, eds. Handbook of physiology. The respiratory system, vol. 2. Bethesda, MD: American Physiological Society, 1986: Dempsey JS, Skatrud JB. Fundamental effects of sleep on breathing. Curr PulmonoI1988;9: Phillipson EA. Control of breathing during sleep. Am Rev Respir Dis 1978;118: Gleeson K, Sweer LW. Ventilatory pattern following hypoxic stimulation during wakefulness and non-rem sleep. J Appl Physiol 1993;75: Younes M. The physiological basis of central apnea and periodic breathing. Curr PulmonoI1989;10: Bradley TD. Right and left ventricular impairment and sleep apnea. Clin Chest Med 1992; 13: Yamashiro Y, Kryger MH. Sleep in heart failure. Sleep 1993; 16: Hanly PJ, Millar TW, Steljes DG, Baert R, Frais MA, Kryger MH. Respiration and abnormal sleep in patients with congestive heart failure. Chest 1989;96: Georgopoulos D, Bshouty Z, Younes M, Anthonisen NR. Hypoxic exposure and activation of the after-discharge mechanism in conscious humans. J Appl Physiol 1990;69: II. Ahmed M, Giesbrecht GG, Serrette C, Georgopoulos D, Anthonisen NR. Respiratory short-term potentiation (after-discharge) in elderly humans. Respir PhysioI1993;93: Georgopoulos D, Giannouli E, Tsara V, Argisopoulou P, Ptakis P, Anthonisen NR. Respiratory short-term post-stimulus potentiation (after-discharge) in patients with obstructive sleep apnea. Am Rev Respir Dis 1992;146: Georgopoulos D, Mitrouska I, Kolestos K, Riggos D, Patakis D, Anthonisen NR. Respiratory short term post-stimulus potentiation (STP) in patients with brain damage. Am J Respir Crit Care Med, 1984 (in press) (abstract). 14. Khoo MCK, Kronauer RE, Strohl KP, Slutsky AS. Factors inducing periodic breathing in humans: a general model. J Appl PhysioI1982;53: Clement JD, Robbins PA. Latency of the ventilatory chemoreflex response to hypoxia in humans. Respir PhysioI1993;92: Millar TW, Hanly PJ, Hunt B, Frais M, Kryger MH. The entrainment of low frequency breathing periodicity. Chest 1990; 98: Naughton M, Bernard D, Tam A, Rutherford R, Bradley TD. Role of hyperventilation in the pathogenesis of central sleep apnea in patients with congestive heart failure. Am Rev Respir Dis 1993; 148: Fregosi RF. Short-term potentiation of breathing in humans. J Appl PhysioI1991;71: Wagner PG, Eldridge FL. Development of short-term potentiation of ventilation. Respir PhysioI1991;83: Badr MS, Skatraud JB, Dempsey JA. Determinants of poststimulus potentiation in humans during NREM sleep. J Appl PhysioI1992;73: