Origin of Oscillatory Kinetics of Respiratory Gas Exchange in Chronic Heart Failure
Background—Respiratory gas exchange measurements in patients with chronic heart failure (CHF) at rest and during exercise commonly reveal prominent slow oscillations in ventilation (V̇E), measured oxygen uptake (V̇o2), and carbon dioxide production (V̇co2), whose origin is not clear. Voluntary simulation of periodic breathing (PB) in normals has been reported to generate a different pattern of oscillations in gas exchange from that seen in spontaneous PB. This necessitates hypothesizing that PB is caused by a primary oscillation in tissue metabolism or in cardiac output.
Methods and Results—We developed an automated method by which normal controls could be guided to breathe according to a PB pattern. The resultant metabolic oscillations closely matched those seen in spontaneous PB and had several interesting properties. At low workloads (including rest), the oscillations in V̇o2 were as prominent as those in V̇E in both spontaneous PB (αVO2/αVE=0.92±0.04) and voluntary PB (0.93±0.07). However, at increased workload, the oscillations in V̇o2 because less prominent than those in V̇E in spontaneous PB (intermediate workload 0.63±0.05, high workload 0.57±0.04; P<0.001) and voluntary PB (intermediate 0.66±0.03, high 0.48±0.03; P<0.001). There was no difference in the relative size of metabolic oscillations between voluntary and spontaneous PB at matched workloads (P>0.05 at low, intermediate, and high workloads). Furthermore, V̇o2 peaked before V̇E in both spontaneous and voluntary PB. This time delay varied from 6.4±0.4 s at low ventilation, to 11.3±0.9 s at high ventilation (P<0.0001).
Conclusions—The magnitude and phase pattern of oscillations in gas exchange of spontaneous PB can be obtained by adequately matched voluntary PB. Therefore, the gas exchange features of PB are explicable by primary ventilatory oscillation.
During exercise testing of some chronic heart failure (CHF) patients, marked periodic oscillations in measured oxygen uptake (V̇o2), measured carbon dioxide production (V̇co2), and ventilation (V̇E) have been recognized, with a period of approximately one minute,1 a manifestation of periodic breathing (PB). The origin of these oscillations is unclear. There are several possibilities, and these may be conveniently grouped into (1) ventilatory and (2) metabolic- hemodynamic.
The ventilatory hypothesis is that there is a primary cyclic fluctuation in the homeostatic systems which regulate ventilation, which results from the time delay between detection of a disturbance and its correction. Fluctuating ventilation causes variations in the body’s gas stores, which manifest as oscillations in V̇o2 and V̇co2 as determined by respiratory gas exchange.
Three lines of evidence oppose this hypothesis. First, it has been reported that the amplitude of V̇o2 oscillation in spontaneous PB during exercise is more prominent than the concomitant oscillation in ventilation.2 This has been given as evidence of an underlying oscillation in metabolism, driving that in ventilation. Second, the metabolic oscillations do not coincide in time with those in ventilation. The third (and most persuasive) finding opposing a ventilatory origin is that control subjects who volitionally simulated PB failed to reproduce the magnitude and phase of the metabolic oscillations seen in spontaneous PB.2 This can be viewed as proof that primary ventilatory oscillations cannot be the cause of the oscillations observed in V̇o2 and V̇co2 in patients with PB.
The alternative metabolic-hemodynamic hypothesis is, therefore, that the fluctuations in V̇o2 and V̇co2 in patients with spontaneous PB result not from fluctuations in ventilation but instead from true fluctuations in gas exchange of the tissues of the body.2 Because tissue V̇o2 is equal to blood flow multiplied by arteriovenous difference in oxygen content, there must be either an underlying oscillation in cardiac output or in tissue metabolic rate. Changes in either of these can modulate ventilation3 through the action of chemoreceptors4 and metaboreceptors,5 thus causing periodic fluctuations in ventilation.
The rejection of the ventilatory hypothesis is based on the failure of voluntary PB to adequately match spontaneous PB. However, one drawback is that voluntary PB is difficult to perform and therefore may not match spontaneous PB well. A second drawback is that without appropriate guidance, controls find it difficult to simulate PB at low mean ventilation rates. It has been assumed that volitional simulation of the qualitative pattern of PB forms an adequate control group, even if the mean ventilation and workload are much higher than those of patients with spontaneous PB. There has been no study to date which assesses the effect of changing workload on the inter-relationships between metabolism and ventilation.
We therefore developed a computer program that could visually guide controls to simulate PB at different ventilation rates to suit rest and different exercise levels. We could thus examine the effect of different workloads on the interrelationship between ventilatory and measured metabolic oscillations.
We studied 13 patients with CHF who exhibited oscillatory breathing at rest and during clinical exercise testing (spontaneous PB) and 8 normal controls who performed voluntary PB at rest and during exercise. Patients had an average left ventricular ejection fraction of 34±10%. Six had ischemic heart disease, 1 had primary valvular disease, and the remaining 6 had idiopathic dilated cardiomyopathy. The patients were aged 52±20 years, and controls were aged 31±4 years. All subjects gave informed consent.
Patients and controls were studied for a short period at rest and then during treadmill exercise, which followed a Bruce protocol modified by the insertion of an initial stage of 0.5 mph at 5% gradient. An Amis 2000 metabolic monitoring system (Innovision), which used a heated pneumotachograph and respiratory mass spectrometer, obtained breath-by-breath on-line calculations of V̇E, oxygen uptake V̇o2 and carbon dioxide production V̇co2.
Voluntary Periodic Breathing
To enable volunteers to simulate PB, the signal from the pneumotachograph was also monitored by a second computer system (Carrera) with custom-designed software which displayed, for the volunteer, a moving bar representing their breathing in association with a target. We could program this system with a fluctuating ventilatory pattern, whose tidal volume varied sinusoidally with a period of oscillation of 1 minute, and with a controllable mean and amplitude. The software compared the volunteer’s respiratory rate and tidal volume with those of the programmed target. It continuously computed and cumulated the difference between intended and actual ventilation. It used this information to modify the target presented to the volunteer. Thus the subject was guided to correct undershoots or overshoots (of rate and/or volume). The result achieved was a close approximation to a sinusoidal ventilation pattern.
Quantification of Oscillation
For each cycle of oscillation, we calculated the mean and amplitude (half the difference between peak and trough). We defined the relative amplitude of the oscillation (α) as the ratio between amplitude and mean. For example, if ventilation varied between 15 and 25 L/min with mean 20 L/min, then the amplitude would be 5 L/min and the relative amplitude, αVE, would be 5/20=0.25. We calculated αVO2 and αVCO2 in a similar manner.
Ventilatory oscillations in spontaneous PB become less prominent (lower αVE) at higher workloads. This must be taken into account when comparing the sizes of metabolic oscillations at different workloads. We therefore calculated the ratio between αVO2 and αVE. For example, if αVE is 0.25 and αVO2 is 0.20, then the size of the metabolic oscillation in comparison to the ventilatory oscillation is αVO2/αVE=0.8. This ratio was also calculated for oscillations in V̇co2.
The distribution of continuous variables is described by their mean and SD. When means of samples are being compared, each mean is qualified by its standard error (SE). Statistical analysis was carried out using Statview 4.5 (Abacus Concepts). Comparisons between patients and controls, and/or across different workloads, were made using ANOVA. P<0.05 was significant.
Examples of the ventilatory and metabolic data obtained are shown in Figure 1A⇓ (spontaneous PB) and Figure 1B⇓ (voluntary PB). The fine line indicates ventilation, whereas the thick line shows measured oxygen uptake. The 13 patients demonstrated a total of 84 cycles of PB, and the normal controls performed a total of 170 cycles of voluntary PB. Overall workload and ventilation rates were comparable between the 2 groups, as shown in Table 1⇓.
The mean workload of each cycle was categorized into low (V̇o2 <6 mL · kg−1 · min−1, which included the resting state), intermediate (6–12 mL · kg−1 · min−1), and high (>12 mL · kg−1 · min−1), so that similar numbers of cycles of spontaneous PB were in each group. The same categorization was applied to the cycles of voluntary PB.
Magnitude of Oscillations in V̇o2
The relative size of the oscillations in oxygen uptake, in comparison to those in ventilation (αVO2/αVE) averaged 0.74 (SE 0.03) in the spontaneous PB of the patients. It was not significantly different in the controls performing voluntary PB (mean αVO2/αVE 0.71, SE 0.03, P>0.5). However, the means of αVO2/αVE for all subjects at different workloads were 0.93 (SE 0.05) at low, 0.67 (SE 0.03) at intermediate, and 0.52 (SE 0.03) at high workload (Figure 2A⇓). The differences in αVO2/αVE were all significant (low versus intermediate, P<0.0001; intermediate versus high, P=0.003; low versus high, P<0.0001).
Considering the spontaneous and voluntary PB cycles separately (Figure 2B⇑) revealed a similar pattern. Among the spontaneous cycles, the means of αVO2/αVE were 0.92 (SE 0.04) at low, 0.69 (SE 0.05) at intermediate, and 0.57 (SE 0.04) at high workload. The differences were significant for low versus intermediate (P=0.0003) and low versus high (P<0.0001) and close to significant for intermediate versus high (P=0.057). Among the voluntary cycles, the means of αVO2/αVE were 0.93 (SE 0.07) at low, 0.66 (SE 0.03) at intermediate, and 0.48 (SE 0.03) at high workload. All the differences were significant: low versus intermediate, P<0.0001; intermediate versus high, P=0.01; low versus high, P<0.0001.
Magnitude of Oscillations in V̇co2
The relative size of the oscillations in CO2 production, in comparison to those in ventilation (αVO2/αVE) was greater overall than the corresponding measure for oxygen uptake: mean αVCO2/αVE 0.87 (SE 0.02) versus mean αVCO2/αVE 0.72 (SE 0.02), P<0.0001. This difference persisted across different workloads: at low workload, 1.04 (SE 0.05) versus 0.93 (SE 0.05), P<0.0001; at intermediate 0.83 (SE 0.02) versus 0.67 (SE 0.03), P<0.0001; at high 0.70 (SE 0.03) versus 0.52 (SE 0.03), P<0.0001. There was no significant difference in αVCO2/αVE between spontaneous (0.81, SE 0.03) and voluntary PB (0.89, SE 0.03), P>0.05. Figure 3A⇓ shows the variation of αVCO2/αVE with workload. Its mean value was 1.04 (SE 0.05) at low, 0.83 (SE 0.02) at intermediate, and 0.70 (SE 0.03) at high workload. These differences were all significant: low versus intermediate, P<0.0001; intermediate versus high, P=0.009; low versus high, P<0.0001. Again, this difference in αVCO2/αVE was seen when spontaneous and voluntary cycles were considered separately (Figure 3B⇓).
Timing of Oscillations in V̇o2
The peak in V̇o2 preceded rather than coincided with the peak in V̇E. This V̇o2→V̇E time delay (TVO2→VE) averaged 8.3 (SD 4.5) seconds over all the cycles and was significantly different from zero (P<0.0001), although there was a wide dispersion. This wide dispersion in TVO2→VE was seen both in spontaneous PB cycles (mean TVO2→VE 6.7, SD 4.4 seconds) and in voluntary PB cycles (mean TVO2→VE 9.1, SD 4.4 seconds) as shown in Figure 4⇓, although both were significantly different from zero (P<0.0001 for both).
No relationship was seen between TVO2→VE and workload. However, there was a clear relationship between TVO2→VE and ventilation level, as shown in Figure 5A⇓. At low ventilation levels (<300 mL · kg−1 · min−1) TVO2→VE averaged 6.4 s (SE 0.4), at intermediate (300–600 mL · kg−1 · min−1) it was 9.1 s (SE 0.4), and at high (>600 mL · kg−1 · min−1) it was 11.3 s (SE 0.9). The comparisons between these are all statistically significant: low versus intermediate, P<0.0001; intermediate versus high, P=0.011; low versus high, P<0.0001.
Separate analysis of spontaneous and voluntary PB cycles also showed a similar result (Figure 5B⇑). For spontaneous cycles, the TVO2→VE at the 3 ventilation levels averaged 4.8 (SE 0.7), 7.3 (SE 0.6), and 10.8 (SE 1.4) s, respectively. These differences were all significant: low versus intermediate, P=0.008; intermediate versus high, P=0.015; low versus high, P<0.0001. For voluntary cycles, the TVO2→VE means were 7.3 (SE 0.5), 9.8 (SE 0.5) and 11.5 (SE 1.1). The differences between low and intermediate (P=0.0005) and low and high (P=0.0002) were significant, although that between intermediate and high ventilation was not (P=0.11).
Periodic breathing, the waxing and waning of ventilation with a period of about one minute, has been a recognized feature of CHF for almost 2 centuries.6 7 Lately the development of on-line respiratory gas exchange analysis has demonstrated associated fluctuations in measured metabolic rates (V̇o2 and V̇co2). Whether these fluctuations in V̇o2 and V̇co2 are cause or effect of the ventilatory oscillations is as yet an unresolved question.1
Under the metabolic-hemodynamic hypothesis, the V̇o2 and V̇co2 oscillate because of a primary fluctuation in tissue metabolism or in cardiac output. In support of this are echocardiographic8 and radionuclide9 studies which demonstrate fluctuations in stroke volume. CHF is associated with weakened baroreflex regulation of heart rate,10 and experimental interference with baroreflex function has been shown to enhance the natural fluctuations in arterial blood pressure.11 Indeed, blood pressure12 and even cerebral blood flow13 have been shown to oscillate during PB. Moreover, there is a potential animal model in the Mayer wave phenomenon seen with experimental hemorrhage: ventilation and blood pressure, as well as directly-measured sympathetic and phrenic nerve activity are all seen to oscillate together.14 In fact, many lines of evidence support the concept of simultaneous ventilatory and metabolic-hemodynamic fluctuations during PB.15 But the most important clinical evidence showing that the ventilatory oscillations could not be the prime mover has been the failure of voluntary PB to reproduce the pattern of metabolic oscillation seen in spontaneous PB.
Our study is, to our knowledge, the first in which volunteers had not simply an oscillatory target but also automated feedback of the cumulative error in their ventilation, so that they could adapt their rate and depth appropriately. It is also unique in that it considered the effect of different workloads upon the observed metabolic oscillations, in both patients and controls. The central finding is that as workload increases, there is a marked decline in size of the measured metabolic oscillations in relation to the ventilatory ones (a falling αVO2/αVE ratio). This occurs both in patients with spontaneous PB and in normal controls with voluntary PB. This key αVO2/αVE ratio does not differ between patients and controls whether all cycles are considered together or grouped by workload level. The other available measure of metabolic rate, V̇co2, also undergoes oscillations during periodic breathing. Again, the relative size of these oscillations (αVCO2/αVE ratio) is comparable between spontaneous and voluntary PB and reduces with increasing workload in both groups. The phase of the metabolic oscillations also changes with level of ventilation; as mean ventilation rises, the V̇o2 peak moves to a progressively earlier time with respect to the ventilatory peak in both spontaneous PB and voluntary PB.
How can we reconcile these findings with the previous clinical studies2 of voluntary PB which showed that the metabolic oscillations measured during voluntary PB differ dramatically from those occurring in spontaneous PB? We propose that in fact there is no contradiction. In previous work, the spontaneous PB cycles occurred at a relatively low workload, with average oxygen uptake of about 8.7 mL · kg−1 · min−1 (obtained by calculating an average of 653 mL/min from the published data and dividing by a notional weight of 75 kg). In contrast, nearly 3-fold higher a workload (approximately 23.9 mL · kg−1 · min−1) was applied to the normal controls carrying out their voluntary PB cycles. Those volunteers manifested a relatively smaller oscillation in V̇o2 (in comparison to the oscillation in VE) than did the patients; this was interpreted as a sign of an underlying metabolic oscillation driving spontaneous PB. Our study now offers an alternative explanation for the lower value of αVO2/αVE seen in volunteers in previous work: their workload was far higher, which we have now shown leads to smaller αVO2/αVE ratios in both patients and controls. When workload is instead matched, αVO2/αVE is higher and equivalent in patients and controls.
The same study also reported a large 39-degree phase discordance (equivalent to ≈6.5 seconds) between spontaneous and voluntary PB with regard to timing of metabolic oscillation in relation to ventilatory oscillation. This has also been interpreted as evidence against a purely ventilatory origin for spontaneous PB. Yet it can now be explained by the marked difference in ventilation between those patients (which can be estimated at 387 mL · kg−1 · min−1) and their controls (740 mL · kg−1 · min−1). The evidence from our study shows that such higher ventilation can itself cause the observed metabolic fluctuations to move earlier in phase in relation to the ventilatory fluctuations. The difference in phase between low and high ventilation in our patients alone was as high as 6 s.
Our study demonstrates that the metabolic oscillations seen in association with spontaneous PB can be closely emulated by voluntary PB. It removes the need to hypothesize an underlying hemodynamic or metabolic oscillation. Application of Occam’s razor (the preference for minimal complexity in scientific explanations) obliges us to consider pure ventilatory oscillation an adequate primary driving force for spontaneous PB. Ventilation is dependent on arterial blood and tissue gas tensions, which in turn (after a circulation-time lag) are determined by lung gas concentrations whose levels themselves fluctuate according to ventilation.16 This sequence of relationships forms a closed circuit which may develop self-sustaining oscillations if the gains of the various linkages are high enough and the time-delays are adequate.17 In CHF, there is enhanced ventilatory sensitivity to both central18 and peripheral19 chemoreceptor stimulation. Chronically increased ventilation results in reduction in arterial concentration of both CO2 and bicarbonate,20 which weakens the blood’s power to buffer against changes in CO2 levels so that the same change in ventilation yields a larger change in arterial pCO2 and pH. Circulation time is also prolonged in CHF. CHF patients with spontaneous PB manifest more severe forms of these abnormalities of chemoreflex control,21 circulation time,22 and blood buffering capacity.20 Modifying the levels of inspired oxygen23 and CO224 have been found to attenuate PB, as have procedures that may shorten circulation time such as valve operations,25 heart transplantation,26 and administration of cardioactive drugs such as milrinone27 or theophylline.28 Thus a wide body of observational, interventional, and theoretical studies support the concept of a primary oscillation in ventilation resulting from harmonic interactions between reflexes of the ventilatory control system and cardiopulmonary physiology.22 The cardiovascular fluctuations that have been documented to occur during PB may well exist, in both patients and controls, as a consequence of this primary oscillation in respiratory physiological control.
Why does the relationship between measured metabolism and ventilation change as exercise progresses? We speculate that at low workloads and associated low ventilations, lung O2 stores are relatively stable. Consequently, periodic fluctuations in ventilation produce proportional changes in respiratory exchange of oxygen (αVO2/αVE close to unity). As ventilation approaches its peak, the difference in O2 concentration between air and the lung begins to fall. Because V̇o2 is the product of alveolar ventilation and O2 concentration difference between air and lung, it peaks slightly before ventilation. This explains the small but significant time delay between peak V̇o2 and peak V̇E, which in our study averaged 6 s at rest and minimal exercise. With gradually increasing exercise, the mean ventilation rises, generating swifter changes in lung gas stores in response to cyclic fluctuations in ventilation. The nadirs in air-lung O2 concentration difference more closely match the peaks in ventilation. This has 2 effects. First, the peak in V̇o2 (which is related to the product of the 2) occurs even sooner (because the air-lung O2 concentration difference falls more rapidly). In our study, at high ventilation rates, the time difference between peak V̇o2 and peak V̇E extended to 11 s. Second, the amplitude of oscillation of V̇o2 is attenuated, because the 2 contributory components are almost in antiphase. Carbon dioxide has a slightly different quantitative physiology; whereas oxygen can be considered to be extracted from the lung into pulmonary blood at a relatively constant rate irrespective of lung O2 fluctuations, carbon dioxide is effectively buffered by the passage of pulmonary blood, whose CO2 concentration can vary in response to lung CO2 concentration. The result is that end-tidal CO2 is relatively more stable in the face of irregular breathing than is end-tidal O2. For the reasons given above, this leads to relatively larger oscillations in V̇co2 than in V̇o2, particularly at the higher workloads where the pulmonary blood flow (and consequently the buffering effect on CO2 levels) is greater.
One important implication of this study is that planning of future investigations of the periodic breathing phenomenon should incorporate the knowledge that workload affects not only the overall size of the ventilatory oscillations1 but also the magnitude and phase of the observed relationships. Studies focusing purely on resting patients or exercising patients may thus have reason to demonstrate divergent results. Furthermore, when control subjects are involved, they should be matched with the patient group for workload and ventilation, if the comparison is to be appropriate.
In conclusion, this study shows for the first time that the oscillations measured in oxygen uptake and CO2 production in patients with spontaneous periodic breathing can be reproduced by appropriate voluntary periodic breathing in normal controls. We make the novel observations that the relative size of the metabolic oscillations becomes smaller with increasing workload, and that rising ventilation leads to a shift of oxygen uptake to a phase further ahead of that of ventilation itself. Finally, these changes occur not only in spontaneous periodic breathing but also in voluntary periodic breathing, which explains discrepancies observed in previous studies of this phenomenon. It is therefore not necessary to hypothesize a primary metabolic oscillation in the genesis of spontaneous periodic breathing of CHF. Harmonic fluctuations in reflex control, resulting from enhancement of chemoreflex sensitivity, slowed hemodynamics, and increased instability of blood gas stores, are an adequate explanation.
Dr Francis is supported by the Estate of the late Mabel Grace Parker and by the British Heart Foundation (FS 98005); Dr Davies is supported by the Robert Luff Fellowship, Prof Coats, by the Viscount Royston Trust, and Dr Rauchhaus, by the Deutsche Herzstiftung e.V.
- Received February 18, 1999.
- Revision received June 1, 1999.
- Accepted June 14, 1999.
- Copyright © 1999 by American Heart Association
Wasserman K, Whipp BJ, Castagna J. Cardiodynamic hyperpnea: hyperpnea secondary to cardiac output increase. J Appl Physiol. 1974;36:457–464.
Chua TP, Ponikowski PP, Harrington D, Chambers J, Coats AJ. Contribution of peripheral chemoreceptors to ventilation and the effects of their suppression on exercise tolerance in chronic heart failure. Heart. 1996;76:483–489.
Piepoli M, Clark AL, Coats AJ. Muscle metaboreceptors in hemodynamic, autonomic, and ventilatory responses to exercise in men. Am J Physiol. 1995;269:H1428–1436.
Cheyne JA. A case of apoplexy in which the fleshy part of the heart was converted in fat. Dublin Hosp Rep. 1818;2:216–219.
Stokes W. The Disease of the Heart and the Aorta. Dublin: Hodges and Smith; 1854:323–326.
Taylor JA, Eckberg DL. Fundamental relations between short-term RR interval and arterial pressure oscillations in humans. Circulation. 1996;93:1527–1532.
Mortara A, Sleight P, Pinna GD, Maestri P, Prpa A, LaRovere MT. Abnormal awake respiratory patterns are common in chronic heart failure and may prevent evaluation of autonomic tone by measures of heart rate variability. Circulation. 1997;96:246–252.
Franklin KA, Sandstrom E, Johansson G, Balfors EM. Hemodynamics, cerebral circulation, and oxygen saturation in Cheyne-Stokes respiration. J Appl Physiol. 1997;83:1184–1191.
Preiss G, Iscoe S, Polosa C. Analysis of a periodic breathing pattern associated with Mayer waves. Am J Physiol. 1975;228:768–774.
Karp HR, Seiker HO, Hyman A. Cerebral circulation and function in Cheyne-Stokes respiration. Am J Med. 1962;41:861–870.
Bradley TD, Phillipson EA. Central sleep apnea. In: Phillipson EA, Bradley TD, eds. Clinics in Chest Medicine. Philadelphia: WB Saunders; 1992:493–506.
Khoo MCK, Gottschalk A, Pack A. Sleep-induced periodic breathing an apnea: a theoretical study. J Appl Physiol. 1991;70:2014–2024.
Chua TP, Ponikowski P, Webb-Peploe K, Harrington D, Anker S, Piepoli M, Coats AJS. Clinical characteristics of chronic heart failure patients with augmented peripheral chemoreflex. Eur Heart J. 1997;18:480–486.
Naughton MT. Pathophysiology and treatment of Cheyne-Stokes respiration. Thorax. 1998;53:514–518.
Ribiero JP, Knutzen A, Rocco MB, Hartley LH, Colucci WS. Periodic breathing during exercise in severe heart failure: reversal with milrinone or cardiac transplantation. Chest. 187;92:555–556.