INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

SPECTRAL ANALYSIS of heart rate (HR) variability (HRV) is a widely accepted method for evaluating the cardiac autonomic modulation in physiological and pathological conditions (27). In particular, it is well established that in the supine position and at a breathing frequency >0.15 Hz the high-frequency (HF) component of the HRV power spectrum (i.e., respiratory sinus arrhythmia) provides a measure of "cardiac parasympathetic responsiveness," since it reflects the modulation of vagal cardiac efferent activity by breathing and the sinoatrial response to that modulation (23, 24). However, a small HF component has been seen in the transplanted denervated human heart (2, 3, 18, 25) and in healthy subjects after autonomic blockade (7, 23), suggesting that breathing-related changes in atrial transmural pressure (6, 21) or in the direction of the cardiac axis (17) may be responsible for at least part of the HRV in the HF range. In young healthy subjects, the contribution of nonneural mechanisms to the HF power of the HRV increases from 1% at rest to 30% during exercise (7). This implies that, in the presence of vagal withdrawal, the HF component may significantly overestimate cardiac parasympathetic responsiveness, especially when its power is normalized for the R-R interval (RR) variance (16). These findings may be applicable to patients with chronic heart failure (CHF) who are known to have impaired cardiac parasympathetic responsiveness and a reduced HRV (5, 9, 22). Experimental data, however, indicate that right atrial stretch may suppress neural and nonneural components of respiratory sinus arrhythmia (14), suggesting that the relative contribution to the HF power in patients with CHF may remain small, despite a reduction in HRV. To test this hypothesis, we calculated the relative contribution of autonomic and nonneural inputs to the HF power of the HRV in patients with mild CHF and in healthy age-matched controls.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Twenty patients with stable heart failure [New York Heart Association (NYHA) Class II, mean age 61 ± 2 yr] and 11 age-matched healthy subjects (mean age 58 ± 2 yr) volunteered for the study. The latter were in good health as indicated by medical history, physical examination, electrocardiogram (ECG), and arterial blood pressure (BP) measurements. None of healthy subjects was taking any medication. All patients had been in stable CHF for 3 mo. Fifteen patients had had a documented myocardial infarction, one had chronic ischemic heart disease, and four had dilated cardiomyopathy. All patients were on angiotensin-converting enzyme inhibitors and furosemide, and one was on a -blocker. None was on digitalis or antiarrhythmic drugs. The average radionuclide left ventricular ejection fraction was 25 ± 6% (mean ± SD).

Subjects with a history of atopic allergy, glaucoma, diabetes mellitus, atrial fibrillation, angina pectoris, or ECG evidence of ischemia-limiting exercise, chronic obstructive pulmonary disease, and liver or renal failure were excluded.

Written consent was obtained after the subjects were informed of the procedures and risks involved in the study. The protocol was approved by the Central Oxford Research Ethics Committee.

Experimental design. Subjects were familiarized with the procedures and asked to avoid exercise and alcoholic or caffeinated drinks on the day of the experiment. Patients with CHF were asked to stop their medications 12 h before the experiments. Studies were conducted in a quiet room at a controlled temperature of 22°C 2 h after a light meal. After 20 min of rest in the supine position, lead II of the ECG, a breathing signal (obtained by a thermistor and from changes in chest impedance), and ventilation (I, in l/min; Harvard Apparatus) were recorded in each subject for 10 min during spontaneous breathing and for 10 min during controlled breathing at 0.16 Hz. Minute averages of tidal volume (VT, ml) were obtained by dividing I by the breathing frequency.

Power spectral analysis of HRV. The ECG and breathing traces were digitized on-line by a 12-bit analog-to-digital converter (AT-Codas, Dataq Instruments) at a sampling rate of 600 Hz. The QRS complex was detected by identifying the points of the low-pass-filtered first derivative of the ECG signal, which exceeded an adaptive threshold. Each QRS complex and its triggering were verified by visual inspection. The breathing signal was sampled once every cardiac cycle. Time series of 256 consecutive RRs were used for spectral analysis. The autoregressive method, which was used to evaluate the power spectral density of the RR series, has been described in detail previously (7). Briefly, a computer program first calculated the autoregressive coefficients using the Levinson-Durbin algorithm, and the model order was chosen by the Akaike information criterion starting from a minimum order of 12. A spectral decomposition method was then applied to evaluate the power and the central frequency of the low-frequency (LF, 0.04-0.15 Hz) and high-frequency (HF, 0.15-0.50 Hz) components. The relative contribution of nonneural mechanisms to the HF power was calculated as follows
The very-low-frequency component (<0.04 Hz) was examined to assess the presence of narrow peaks indicating periodic breathing (19).

Cross-spectral analysis. To assess the relationship between respiration and HF oscillations, the cross spectrum and squared coherence function of the HRV and the breathing signal were calculated as described by Pagani et al. (20). Briefly, cross spectra were computed by fast Fourier transformation, using a triangular window, on successive 50% overlapping series of 64 RRs each. The squared coherence function was used to evaluate the correlation between breathing and RR oscillations at the same frequency. The value of this function is regarded as an analog of the squared correlation coefficient (r²) and ranges from 0 to 1.

Statistical analysis. Values are means ± SE. ANOVA (SuperANOVA, Abacus Concepts) was used to compare the data within the protocol stages and between patients with CHF and healthy subjects. To achieve a normal distribution and to normalize the variance of the data between stages, comparisons were made using the square root of the power of the spectral components (i.e., their amplitude, in ms). Nonparametric statistic was employed to compare the contribution of nonneural mechanisms to the HF power in CHF patients and controls. For this variable, data are given as geometric means. Statistical significance was accepted at P < 0.05.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Two CHF patients and one control subject could not complete the study, since they developed a vasovagal reaction (bradycardia, nausea, sweating, and hypotension) shortly after the start of the infusion of trimethaphan camsylate. In two CHF patients and one control subject, the infusion of trimethaphan camsylate was not attempted, since the presence of frequent ectopic beats did not permit a satisfactory analysis of the baseline ECG recordings.

GB was usually obtained with an infusion rate of trimethaphan camsylate between 5 and 6 mg/min. At these doses, the increase in BP induced by the intravenous injection of phenylephrine was associated with a modest shortening of the RR in four subjects (Fig. 1) and no change in the others, with the exception of one CHF patient, in whom a trimethaphan camsylate infusion of 6 mg/min did not completely abolish the RR lengthening associated with phenylephrine. Since high doses of trimethaphan camsylate can cause respiratory depression, it was considered unsafe to increase the rate of infusion further, and the data from this patient were not included in the final analysis. During GB, all subjects experienced dry mouth and cycloplegia.

View larger version (21K): [in this window] [in a new window]   Fig. 1.   Beat-by-beat plot of systolic blood pressure (SBP) and R-R interval (RR) after intravenous phenylephrine before (A) and during (B) ganglionic blockade (GB) in a patient with heart failure. During GB, the increase in SBP secondary to intravenous phenylephrine is associated with shortening (rather than lengthening) of the RR.

Baseline measurements. The mean RR did not differ between CHF patients (n = 15) and age-matched control subjects (n = 9); however, the standard deviation of the RR was lower in CHF patients during spontaneous (P < 0.05) and controlled (P < 0.001) breathing (Table 1). Likewise, patients had lower baroreflex sensitivity (4.34 ± 0.81 vs. 10.85 ± 2.31 ms/mmHg, P < 0.05) and a reduced HF component (Table 1). There were no significant differences in I, VT (Table 2), or spontaneous breathing frequency between control subjects and CHF patients (0.24 ± 0.02 and 0.28 ± 0.03 Hz, respectively). Controlled breathing at 0.16 Hz increased VT (Table 2) and the amplitude of the HF component (Table 1) in both groups.

View this table: [in this window] [in a new window]   Table 2.   I and VT before and during GB

GB. GB caused a significant reduction in the mean RR and its standard deviation (Table 1) in both groups. The RR fluctuations in the LF range were completely abolished by GB, but some degree of respiratory sinus arrhythmia persisted in all subjects and did not differ in magnitude between CHF patients and healthy controls (Table 1, Figs. 2 and 3). The correlation between RR fluctuations in the HF range and the breathing signal was not affected by GB (coherence values of 0.82 ± 0.03 and 0.89 ± 0.02, respectively), indicating that a close relationship between these signals persisted after neural control of sinoatrial node activity was abolished.

View larger version (30K): [in this window] [in a new window]   Fig. 2.   Tachogram (top) and power spectra of the HRV (middle) and breathing signal (bottom) before and during GB in a healthy subject. Whereas the low-frequency (LF) component (centered at ~0.1 Hz) is completely abolished by GB, the high-frequency (HF) component (0.15-0.50 Hz) is still present, albeit greatly reduced in amplitude (note the different scale in the y-axis).

View larger version (30K): [in this window] [in a new window]   Fig. 3.   Tachogram (top) and power spectra of the heart rate variability (HRV, middle) and breathing signal (bottom) before and during GB in a chronic heart failure (CHF) patient. Note the reduction in amplitude of HF oscillations before GB compared with Fig. 2. As in Fig. 2, nonneural HF oscillations are responsible for most of the residual HRV during GB, whereas LF oscillations are completely abolished.

Nonneural contribution to the HF power in CHF patients and in healthy controls. During spontaneous breathing, the relative contribution of nonneural mechanisms to the HF power was 3% in healthy controls (range 0.7-30%) and 15% in patients with CHF (range 1-76%, P < 0.005 vs. controls). During controlled breathing, nonneural mechanisms accounted for 1% of the HF power in healthy controls (range 0.2-13%) and 5% in CHF patients (range 1-44%, P < 0.005 vs. controls and P < 0.01 for the effect of controlled breathing at 0.16 Hz in both groups; Fig. 4).

View larger version (29K): [in this window] [in a new window]   Fig. 4.   Controlled breathing at 0.16 Hz significantly decreased the contribution of nonneural mechanisms to the HF power of the HRV in healthy controls () and patients with CHF (). Average data are geometric means ± SD. *P < 0.005, CHF patients vs. controls. ¶P < 0.01 vs. spontaneous breathing for both groups.

View larger version (14K): [in this window] [in a new window]   Fig. 5.   Relationship between the RR standard deviation and the nonneural contribution to the HF component in healthy controls () and patients with CHF () during spontaneous (A, r2 = 0.53, P < 0.0001) and controlled breathing (B, r2 = 0.50, P < 0.0001).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Heart failure is associated with neurohumoral activation and a reduction in cardiac vagal responsiveness (10). Spectral analysis of the HRV in these patients has shown that the absolute power in the LF range decreases with the worsening of the disease and is often absent in NYHA Class IV patients (18). Likewise, the absolute power of the HF component has been found to be significantly decreased from the early stages of the disease (8, 11). However, HR fluctuations in the HF range are present even in the most advanced stages of CHF, where they can account for most of the short-term HRV (11, 18). To explain this finding, it has been suggested that respiratory sinus arrhythmia in severe CHF might be entirely mediated by nonneural mechanisms (11, 18). Experiments in anesthetized pigs, however, have shown a reduction in neural and nonneural HR oscillations in the HF range with atrial stretch (14), suggesting that nonneural modulation of HR may also be suppressed in patients with CHF.

The present study shows that small HR fluctuations synchronous with breathing persist after GB, and their magnitude is similar in healthy subjects and in patients with mild CHF. These residual HF fluctuations are unlikely to be vagally mediated, since, in the dose range employed in this study, trimethaphan camsylate completely abolished the reflex bradycardia in response to the phenylephrine-induced increase in BP (Fig. 1). Moreover, the LF component was abolished by GB in all subjects (Table 1). Since sympathetic and vagal activities have been shown to contribute to the HRV in this frequency range (23), its disappearance with high doses of trimethaphan camsylate provides further evidence for the adequacy of GB. We also show that, during spontaneous breathing, nonneural mechanisms contribute significantly (i.e., 15%) to the HF power in patients with mild CHF but not in healthy age-matched controls (3%). Since nonneural modulation of HR was preserved in CHF patients, its contribution to the HF component was inversely related to HRV (as assessed by the RR standard deviation, Fig. 5). Taken together, our results indicate that the HF power of the HRV can overestimate cardiac parasympathetic responsiveness in patients with mild CHF and, indeed, whenever the HRV is significantly reduced. This is especially the case when HF is expressed in "normalized" units (i.e., as a percentage of the RR variance) or as LF-to-HF ratio (1). Our findings would have been more dramatic if we had included patients with severe CHF. Indeed, we found that the contribution of nonneural oscillation to the HF component of the HRV was ~50% in four NYHA Class III patients (data not shown). However, high doses of trimethaphan camsylate are poorly tolerated in this group; thus we decided to limit our study to less symptomatic patients. Nevertheless, the strong inverse relationship between contribution of nonneural mechanisms to the HF power and HRV provides proof of a basic principle that can be applied to virtually all pathophysiological conditions associated with a reduction in vagal activity or responsiveness (7). Predictably, maneuvers that are known to increase the vagally mediated respiratory sinus arrhythmia, such as slow deep breathing (13), significantly decreased the relative importance of nonneural mechanisms (Fig. 4). Since controlling the breathing rhythm also reduces the incidence of periodic or Cheyne-Stokes respiration in patients with CHF (19), this procedure, as well as the use of absolute units for expressing HF power, appears to be strongly advisable when one attempts to assess cardiac autonomic control or prognosis by using spectral analysis of HRV in patients with CHF.

Mechanisms responsible for the nonneural HF oscillations. HR fluctuations in the HF range result from periodic stretching of the sinoatrial node secondary to changes in atrial transmural pressure with breathing (6, 21). Pacemaking activity is known to be enhanced by mechanical stretch (15), and stimulation of mechanosensitive Cl channels has been shown to be at least in part responsible for this phenomenon (1, 12). The degree of atrial stretch is dependent on the breathing pattern, and large increases in I and VT have been shown to enhance nonneural respiratory sinus arrhythmia in heart transplant patients (2, 3, 25). Similarly, Perlini et al. (21) found that the HF power of the HRV in anesthetized rabbits after vagotomy and -adrenergic blockade was positively correlated with VT and, to a smaller extent, breathing frequency. The main determinant of nonneural HR oscillations, however, was respiratory flow, and combinations of VT and breathing frequencies that gave the same I also resulted in similar nonneural HF power (21). Consistent with these findings, controlled breathing at 0.16 Hz did not affect the amplitude of the HF component after GB (Table 1), since the increase in VT was counteracted by the reduction in breathing frequency and I did not change compared with spontaneous breathing (Table 2).

Summary. This study shows that the relative contribution of nonneural mechanisms to the respiratory sinus arrhythmia is significantly increased in patients with mild CHF, accounting on average for 15% (range 1-77%) of the HF component of the HRV (vs. 3% in controls). These findings suggest that spectral analysis of HRV should be interpreted with caution in all patients with CHF (not just in those with terminal disease) or, indeed, whenever the HRV is significantly reduced [e.g., during exercise (7)]. Since RR oscillations in the LF range appear to be entirely neurally mediated whereas those in the HF range are clearly not, the use of normalized units or of the LF-to-HF ratio in CHF patients would lead to a significant overestimation of the level of cardiac vagal responsiveness. However, the accuracy of respiratory sinus arrhythmia as a vagal index can be improved by controlled breathing, since this maneuver appears to increase the vagal contribution to respiratory sinus arrhythmia without affecting its nonneural component.