Abnormal Awake Respiratory Patterns Are Common in Chronic Heart Failure and May Prevent Evaluation of Autonomic Tone by Measures of Heart Rate Variability
Background Reduced heart rate variability, particularly in the very-low-frequency (VLF) spectral band, has been found to be a marker for poor prognosis in patients after myocardial infarction, but the origin of the VLF oscillations is unclear. In this study, we demonstrate that the power of cardiovascular oscillations in the VLF band in awake patients with mild to severe chronic heart failure is greatly increased by the common occurrence of unrecognized irregularity of breathing, which may confound the use of heart rate variability measures as indexes of autonomic tone or prognosis.
Methods and Results Among 110 consecutive patients referred for consideration of transplantation, 90 were in sinus rhythm, of whom 10 were excluded as unstable. The remaining 80 patients underwent recordings of ECG, beat-to-beat arterial oxygen saturation (Sao2), and respiration during both spontaneous and controlled breathing. During spontaneous awake breathing, 64% showed periodic breathing or Cheyne-Stokes respiration (CSR), which was associated with dominant power in the VLF band of all signals. This VLF power accounted for 55%, 77%, and 87% of heart rate variability, respectively, in patients with normal breathing, periodic breathing, and CSR. It was reduced by 48% and 62%, respectively, during controlled breathing in patients with periodic breathing or CSR. Controlled ventilation also improved oxygen saturation and markedly reduced its variability.
Conclusions Breathing disorders are surprisingly common in awake patients with poor left ventricular function and produce large VLF oscillations in heart rate variability. If measures of heart rate variability are used for prognostic purposes during both short-term and long-term recordings, the confounding effects of variable respiratory patterns should be excluded. Respiratory rehabilitation might help control potentially hazardous surges in sympathetic tone.
The purpose of the present study was to evaluate the prevalence of respiratory rhythm disorders during awake daytime in patients with ventricular dysfunction but compensated heart failure and to determine their effects on measures of HRV. Preliminary results have been presented.1
Abnormal breathing patterns during sleep are well recognized. In patients with severe CHF,2 both PB (waxing and waning of tidal volume without apnea) and the type of PB known as CSR (waxing and waning of tidal volume with apnea) may disrupt sleep, causing insomnia and dyspnea.3 4 These breathing disorders during sleep have also been observed during awake daytime4 and in patients with poor left ventricular function and compensated heart failure during laboratory recordings or even during exercise.5 6
These alterations of breathing are associated with marked oscillations of arterial oxygen saturation, systolic and diastolic pressures, and heart rate.7 8 How this powerful cardiorespiratory rhythm is generated is still debated (see “Discussion”). Whatever its origin, this rhythm can markedly affect analysis of HRV and blood pressure variability. The slow oscillations of lung volume and arterial saturation are associated with heart rate and blood pressure fluctuations that are predominantly in the VLF (0.01 to 0.04 Hz) band of the spectrum (≥70% of the total variability).9 10 Thus, both the overall variability of heart rate (represented by time-domain parameters such as SDNN or SDANN or by spectral analysis computation of TP) and particularly the power in the VLF band may be altered by the presence of PB. This cardiopulmonary rhythm may be evident not only when HRV analyses are applied to 24-hour ECG recordings but also when short 10- to 30-minute segments are analyzed; CSR and PB have been observed previously even during awake daytime laboratory recordings.5 9 Altered respiratory patterns are not restricted to patients with severe symptomatic ventricular dysfunction but also occur in mild to moderate CHF5 ; we shall show in the present study that these oscillations produce a large increase in HRV and particularly in VLF power. Reduction in VLF power or HRV has been used as a marker for poor prognosis in cardiac patients.11 Hence, the usefulness of such markers (eg, on Holter ECG records) may be limited unless respiration is also measured.
Ninety of 110 consecutive patients with dilated cardiomyopathy and moderate to severe heart failure who were admitted during 1995 to the Heart Failure Unit of Montescano Medical Center for consideration of possible heart transplantation were in sinus rhythm and were considered for HRV analysis. Ten patients were excluded because of refractory decompensation. After optimization of medical therapy, 53 patients were in New York Heart Association class II, 23 in class III, and 4 in class IV. All the patients were studied in a stable condition (no changes in signs or symptoms in the 2 weeks preceding the study). No patient was treated with β-blockers. None had a history of pulmonary or neurological disease or acute myocardial infarction or had undergone cardiac surgery within the previous 6 months.
Informed consent was obtained from all subjects; the study protocol was approved by the local ethics committee.
At the same time in the morning and after 30 minutes of supine rest, which allowed for stabilization of the signals, recordings were performed during 15 minutes of spontaneous respiration and 5 minutes of controlled breathing at 0.25 Hz. Baseline recordings were split in three 5-minute epochs, and the results of at least two epochs were averaged. Epochs with >5% ectopic beats and artifacts were excluded. All patients underwent simultaneous recording of ECG, ILV by inductance plethysmography (Nims, Respitrace Plus), and beat-to-beat Sao2 by a fast-response oximeter with ear probe (Ohmeda, Biox 3740).
Analog signals were acquired in a personal computer with a sampling frequency of 250 Hz. The RR time series were obtained from ECG recordings by a linear interpolation algorithm on the first derivative of the signal, yielding a time resolution of 1 ms. IMV was obtained from lung volume measurements by dividing the tidal volume measured in each cycle by its corresponding duration. The resulting time series was then interpolated by a cubic spline and resampled at 2 Hz. The same resampling procedure was applied to the other time series so as to have all signals sampled synchronously. All time series were finally corrected for linear trends with the use of a least-squares fitting algorithm.
Spectral analysis of the recorded time series of RR, IMV, and Sao2 as well as the cross spectra of different combinations of the signals was computed by both the Blackman-Tukey and autoregressive methods. Windowing of the sample autocovariance function for Blackman-Tukey estimation was performed by the Parzen window with a bandwidth of 0.015 Hz. The model order for the autoregressive estimation was interactively selected starting from a minimum value of 12 and searching for the best overlap with the Blackman-Tukey spectral estimate.12 The coherence function of the different bivariate combinations was then estimated. Coherence expresses the fraction of power at a given frequency in either time series that can be explained as a linear transformation of the other and is thus an index of linear association between the two signals.13 Autoregressive spectral decomposition was used to identify and estimate the central frequency and power of main spectral components in the three frequency bands of the spectrum: VLF (0.01 to 0.04 Hz), LF (0.04 to 0.15 Hz), and HF (0.15 to 0.45 Hz). TP in the overall signal (0.00 to 0.45 Hz) and time-domain parameters of HRV (SD of RR intervals and the rMSSD) were also computed.
Definition of Breathing Disorders
PB was defined as a waxing and waning of tidal volume without periodic phases of apnea, whereas CSR was defined as a type of PB in which the phases of hypoventilation and hyperventilation were separated by apnea.2 4 The breathing pattern was determined by two independent observers (A.M., R.M.) on the basis of both a visual inspection of ILV and IMV recordings and the presence in the power spectrum plots of a well-defined peak in the VLF band of the IMV and Sao2 signals10 (see example of normal respiration, PB, and CSR in Fig 1⇓).
Results are quoted as mean±SD. Normality of the distribution of the data was assessed by χ2 analysis with a goodness-of-fit test. Nonparametric statistical methods were used when the variables did not show a normal distribution (VLF, LF, and HF power of HRV). One-way ANOVA and Kruskal-Wallis ANOVA for continuous measures and χ2 test for categorical variables were used to assess differences according to the respiratory pattern. Post hoc simultaneous multiple comparisons were done by Scheffé’s procedure. Differences between pairs of means (baseline versus controlled ventilation) were subsequently analyzed with a t test for paired samples or with the Wilcoxon signed rank test. Statistical significance was defined at the P<.05 level.
The clinical characteristics of the population are reported in Table 1⇓.
Breathing disorders were observed in 51 (64%) of 80 patients (PB in 30 and CSR in 21); the remaining 29 had a normal respiratory pattern. These abnormalities were seen in awake patients and were not due to intermittent bursts of sleep. Comparison of the clinical and hemodynamic variables in the three groups of patients are shown in Table 2⇓. Patients with PB and CSR had a trend toward a worse ventricular function than patients with normal respiration.
In both PB and CSR, the minute ventilation signal, which includes changes in both tidal volume and respiratory frequency, showed quite regular fluctuations in the VLF band (0.018±0.009 Hz). These oscillations were also observed (but with 180° phase shift) in the Sao2 signal and RR (Fig 2A⇓). A good coherence in the VLF band was found between IMV and Sao2 (0.91±0.20) and between Sao2 and RR (0.80±0.15) (Fig 3⇓). All these data clearly support the concept that in patients with PB and particularly in those with CSR, a slow cardiorespiratory rhythm is present that markedly affects the variability of all signals.
All time- and frequency-domain parameters of HRV are listed in Table 3⇓. Note that central frequency of the VLF band in patients with normal breathing refers to a broadband VLF component that is detectable in CHF patients at ≈0.01 Hz. In the CSR group, SD, TP, and VLF were higher and rMSSD and HF reduced compared with patients with normal respiration, with a clear trend from normal levels for PB and CSR in all HRV indexes. In CSR patients, in whom the slow cardiorespiratory rhythm was stronger, power in the VLF band accounted for 87% of total variability, in PB for 77%, and in patients with normal respiration for 55%.
Controlled breathing eliminated the periods of apnea in all CSR patients and markedly reduced or abolished variations of tidal volume in both PB and CSR (see example in Fig 2B⇑). Mean Sao2 increased in all subjects; this increase was significantly higher in those with previous PB and CSR than in patients with normal respiration (0.7% versus 1.4% in PB [P<.01] and versus 2.5% in CSR [P<.01]) (Fig 4⇓).
Changes in the HRV parameters from baseline to controlled breathing are listed in Table 3⇑. No differences were observed in mean RR, rMSSD, LF, or HF power, whereas controlled breathing, by limiting IMV and Sao2 fluctuations, induced a dramatic reduction of SD, TP, and VLF power in both PB and CSR. In particular, VLF power was reduced by 48% and 63%, respectively, in patients with PB and CSR.
This study has shown that in awake subjects with severe left ventricular dysfunction, abnormal respiratory patterns are more common than had been previously recognized. Similar patterns have also been seen in less severe heart failure.5 At least three tenable hypotheses have been formulated to explain the origin of cardiovascular oscillations seen in CSR or PB7 : (1) reductions in ventilatory amplitude and periodic arterial oxygen desaturation may stimulate chemoreceptor reflexes and thus modulate heart rate and blood pressure via autonomic efferents; (2) oscillations of venous return caused by periodic ventilation lead to fluctuations in stroke volume and blood pressure, which then modulate heart rate via both cardiopulmonary and arterial baroreceptor reflexes; and (3) a primary rhythm in the central nervous system may entrain both heart rate, blood pressure, and ventilation.
Even though the causes and exact mechanisms of these complex oscillations are uncertain, we have clearly demonstrated that they are associated with large and significant increases in power in the VLF band on spectral analysis and with a strong trend to increases in time-domain measures of HRV. In patients with abnormal breathing, both of these measures of increase in HRV are significant, and both are greatly reduced by controlled ventilation. During controlled breathing, HRV is progressively reduced with worsening of left ventricular dysfunction, whereas during spontaneous breathing, HRV increases with worsening of left ventricular dysfunction (Table 3⇑). This latter finding is the reverse of what might be expected from the previously described association between reduced HRV (VLF) and poor prognosis.11 14
Confounding Factors in the Use of VLF as a Prognostic Index
Bigger and colleagues11 reported that in cardiac patients, reduced variability in the VLF band (rather than LF or HF) is the best prognostic index, which is somewhat surprising because it might be expected that measures of HF and LF, which are more closely associated with autonomic function15 or baroreflex gain,16 17 might be better indexes of prognosis. Indeed, the ATRAMI study18 showed that baroreflex sensitivity and HRV are good markers of prognosis after myocardial infarction.
Different hypotheses concerning the origin of VLF fluctuations in cardiovascular parameters have been suggested, eg, temperature control, slow hormonal changes such as in the renin-angiotensin system,15 19 20 and enhancement of peripheral chemosensitivity,21 but we do not know of any clear evidence to confirm or refute these hypotheses. In human subjects, Bernardi et al22 showed that physical activity, either random or in regular cycles, can markedly increase power in the VLF band in 24-hour ECG recordings. If VLF power is to prove a robust and independent index of prognosis, it is clear that activity should be controlled for, because subjects unable to exercise would very likely have a poor prognosis. It is this inactivity that might be associated with poor prognosis, rather than low VLF power, which may be just a consequence of inactivity. Of course, it is quite possible that if activity is controlled for, then VLF power might be an even better index of prognosis.
In a small number of CHF patients, we10 previously found that during PB or CSR, heart rate fluctuations in the VLF band are closely linked to a synchronous oscillation of tidal volumes. The present systematic study of consecutive patients emphasizes the importance of such a common and powerful confounder, which may mask the presence of inherently low VLF power by the addition of VLF variability caused by respiratory fluctuations. Recently, Ponikowski et al21 reported the presence of a discrete peak in the VLF band in 64% of the CHF patients studied. Among these, only 50% had a respiratory pattern associated with a VLF peak in the spectral decomposition of the respiratory signal. Although we also observed few patients with a VLF oscillation in heart rate and a nonperiodic respiratory pattern, the data by Ponikowski et al are not concordant with our results. The discrepancy may be easily explained by a different methodological approach. Ponikowski et al did not measure minute ventilation but simply a respiratory signal obtained by an impedance plethysmography technique (and this limitation is reported in their study). Moreover, autoregressive spectral decomposition was performed by using a fixed model order of 15. By overlapping the power spectra obtained by the classic Blackman-Tukey technique (fast-Fourier transformation) and those obtained by the autoregressive method, Pinna et al12 recently demonstrated that real signals, particularly from heart failure patients, are adequately fitted only with a model order much greater than 15. Indeed, in the present study, we also observed that the best order of the model obtained by overlapping the power spectra by both methods was always ≈18 to 20 or more. It is likely that an approach based on the use of low orders during autoregressive decomposition and without the analysis of tidal volume oscillations may have limited the observation of discrete peaks in the VLF band, in which it is sometimes difficult to resolve spectral components near the zero frequency.
Clinical Importance of PB and CSR in Heart Failure
Patients with heart failure who develop PB and CSR at night experience significant sleep disruption with recurrent arousal during the hyperpneic phase.2 3 4 This loss of refreshing sleep causes excessive daytime sleepiness.23 It is possible that it also depresses cerebral function so that these abnormal breathing patterns carry over into the waking state. The marked fluctuations in oxygen saturation, both in sleep and wakefulness, expose patients to prolonged periods of hypoxia.
Somers’ group24 in Iowa studied muscle sympathetic nerve activity in patients who showed similar patterns of disordered breathing as a result not of heart failure but of obstructive sleep apnea. They not only found that patients with obstructive sleep apnea showed increased sympathetic drive at night, but also that this carried over to the day, even when breathing patterns were normal.24 Our group25 showed increased daytime sympathetic activity in patients with left ventricular dysfunction. Recently, Somers et al reported that this increased daytime sympathetic discharge (during awake normal breathing in patients who at night suffer from obstructive sleep apnea) can be very much reduced by the long-term use (at home) of continuous positive-airway-pressure–assisted ventilation (unpublished data, presented at American Society of Hypertension, New York, NY, May 1996). Similar data have been shown in CHF patients with nocturnal CSR and central apneas.26 It was found that these patients had greater overnight urinary norepinephrine and greater daytime plasma norepinephrine concentrations than those without breathing abnormalities despite comparable degrees of left ventricular dysfunction. Moreover, as also observed by Somers et al, 1-month therapy with nocturnal continuous positive-airway-pressure–assisted ventilation caused a significant reduction in both concentrations of norepinephrine.26
Whatever the underlying pathogenesis, these alterations of breathing during awake daytime, which we observed in the present study, may lead to further impairment of ventricular function by causing hypoxia and increased sympathetic drive. Appropriate therapy to reduce these oscillations in patients with heart failure should be considered, not only for the management of symptoms related to CSR but also to possibly limit the excessive sympathetic drive. Treatments of CSR in heart failure have been proposed, such as continuous positive airway pressure,26 benzodiazepines,27 and oxygen therapy.28
Our results show that the voluntary control of respiration abolishes apnea and markedly reduces oscillations of tidal volume, with a significant increase in the level and stability of oxygenation. Even though the patients were asked not to breathe more deeply during controlled ventilation, it is likely that they may have increased minute ventilation with consequent improvement of Sao2. However, as shown in Fig 4⇑, the increase of Sao2 was much greater in patients with PB and CSR, potentially owing to a more efficient respiration with elimination of apneas and hypoventilation. If confirmed, these data may have important clinical implications by supporting an appropriate respiratory training in the treatment and care of CHF patients.
In conclusion, this study in a large population of patients with mild to severe heart failure demonstrates that irregular and periodic respiration during normal awake daytime is a common event. This finding is clinically relevant because it suggests that breathing disorders and apneas are not limited to sleep. By causing frequent and prolonged periods of hypoxia throughout the day, they may significantly contribute to excessive sympathetic discharge and to further deterioration in ventricular function.
These abnormal breathing patterns lead to a marked increase in HRV, particularly by giving rise to a dominant oscillation in the VLF band of power spectral analyses. Unexpected abnormalities of respiration may thus distort time- and spectral-domain analyses of ECG and Holter recordings and mask prognostic information (ie, low HRV) that could be of importance. Controlled breathing completely abolishes periodic hypoxia, thus preventing its effects on the cardiovascular system. Rehabilitation directed toward training in regular breathing may have considerable clinical potential in patients with severe left ventricular dysfunction.
Selected Abbreviations and Acronyms
|CHF||=||chronic heart failure|
|HRV||=||heart rate variability|
|ILV||=||instantaneous lung volume|
|IMV||=||instantaneous minute ventilation|
|rMSSD||=||root mean square of successive differences|
|Sao2||=||arterial oxygen saturation|
|VLF||=||very low frequency|
- Received October 14, 1996.
- Revision received January 6, 1997.
- Accepted January 15, 1997.
- Copyright © 1997 by American Heart Association
Mortara A, Pinna GD, Maestri R, Prpa A, Cobelli F, Tavazzi L. Heart rate variability in chronic heart failure: is it an index of autonomic tone or a marker of respiratory rhythm disorders? J Am Coll Cardiol. 1996;27(suppl):202A. Abstract.
Cherniack NS, Longobardo GS. Cheyne-Stokes breathing: an instability in physiological control. N Engl J Med. 1973;288:952-957.
Pinna GD, Maestri R, Di Cesare A, La Rovere MT, Mortara A. Very low frequency oscillations of heart period: relationship to respiratory activity and blood pressure in heart failure patients. In: Computers in Cardiology. Los Alamitos, Calif: IEEE Computer Society Press; 1994:553-556.
Mortara A, Bernardi L, Pinna GD, Spadacini G, Maestri R, Dambacher M, Muller C, Sleight P, Tavazzi L, Roskamm H, Frey AW. Alterations of breathing in chronic heart failure: clinical relevance of arterial oxygen saturation instability. Clin Sci. 1996;91S:72-74.
Pinna GD, Maestri R, La Rovere MT, Mortara A. New insights on the origin of the VLF cardiorespiratory rhythm in chronic heart failure patients. In: Computers in Cardiology. Los Alamitos, Calif: IEEE Computer Society Press; 1995:205-208.
Bigger JT Jr, Fleiss JL, Steinman RC, Rolnizky LM, Kleiger RE, Rottman JN. Frequency domain measures of heart period variability and mortality after myocardial infarction. Circulation. 1992;85:164-171.
Malliani A, Pagani M, Lombardi F, Cerutti S. Cardiovascular neural regulation explored in the frequency domain. Circulation. 1991;84:482-492.
La Rovere MT, Mortara A, Pinna GD, Bernardi L. Baroreflex sensitivity and heart rate variability in the assessment of cardiac autonomic status. In: Camm AJ, Malik M, eds. Heart Rate Variability. Armonk, NY: Futura Publishing Co; 1995:189-205.
La Rovere MT, Mortara A, Bigger JT Jr, Hohnloser S, Malik M, Marcus FI, Nohara R, Schwartz PJ, for the ATRAMI investigators. Baroreflex sensitivity and heart rate variability in the prognostic evaluation of post-myocardial infarction patients: the ATRAMI study. Eur Heart J. 1996;17:205. Abstract.
Parati G, Saul JP, Di Rienzo M, Mancia G. Spectral analysis of blood pressure and heart rate variability in evaluating cardiovascular regulation: a critical appraisal. Hypertension. 1995;25:1276-1286.
Ponikowski P, Chua TP, Amadi AA, Piepoli M, Harrington D, Volterrani M, Colombo R, Mazzuero G, Giordano A, Coats AJS. Detection and significance of a discrete very low frequency rhythm in RR interval variability in chronic congestive heart failure. Am J Cardiol. 1996;77:1320-1326.
Bernardi L, Valle F, Coco M, Calciati A, Sleight P. Physical activity influences heart rate variability and very-low frequency components in Holter electrocardiograms. Cardiovasc Res. 1996;32:234-237.
Somers VK, Dyken ME, Clary MP, Abboud FM. Sympathetic neural mechanisms in obstructive sleep apnea. J Clin Invest. 1995;96:1897-1904.
Elam M, Casale R, La Rovere MT, Mortara A, Tavazzi L. Is sympathetic neural hyperactivity in chronic heart failure affected by heart transplantation? Eur Heart J. 1993;14:521-525.
Hanly PJ, Millar TW, Rteljes DG, Baert R, Frais MA, Kryger MH. The effect of oxygen on respiration and sleep in patients with congestive heart failure. Ann Intern Med. 1989;111:777-782.