Effect of Vagal Nerve Electrostimulation on the Power Spectrum of Heart Rate Variability in Man M.V. KAMATH, A.R.M. UPTON, A. TALALLA, and E.L. FALLEN From the Departments of Medicine and Surgery, McMaster University Medical Centre, ChedokeMcMaster Hospitals, Hamilton, Ontario, Canada KAMATH, M.V., ET AI .: Effect of Vagal Nerve Electrostimulation on the Power Spectrum of Heart Rate
Variability in Man. The power spectrum of heart rate variability contains low frequency (LF - 0.08-0.12 Hz) and high frequency {HF = 0.18-0.30 Hz) components said to represent neurocardiac rhythms. To verify whether such a relationship exists ive report a unique study where the heart rate autospectrum was determined in a 28-year-o]d epiJepfic male patienf with an implanted vagal electrical stimulator. The stimulator was activated at 20 Hz, 300 fxsec pulse, and 1.25 V. Continuous ECG and respiratory waveform records were obtained over 45 minutes every 8 hours (7-8 AM; 3-4 PM; 11-12 PM) with the stimulator ON, then 24 hours OFF and then 24 hours ON again. The overall LF.HF peak ratio increased from 0.64 to J.99 (P < 0.001} after the stimulator was turned OFF. There was a dramatic increase in the LF peak power f> 60%) and a corresponding decrease in the HF peak power f> 65%) when the stimulator was turned OFF. These vaJues were reversed when the stimulator was turned ON again, fn the early morning and late evening hours, there was a significant rightward shift in the LF peak power frequency (average 0.057 to 0.075 Hz) whenever the stimulator was ON. Otherwise, there were no significant circadian variations in any of the autospectral components. The results demonstrate an unequivocal relationship between selective vagal nerve electrostimulation and alterations in the heart rate autospectrum. (PACE, Vol. 15, February 1992) autospectrum, circadian, orthostasis, sympatho-vagai balance
Introduction Under steady-state conditions, the variability of the heartbeat is said to reflect cardiac autonomic activity.^*^ By transforming these beat-to-beat fluctuations into their frequency components a power spectrum is obtained that contains two distinct spectral bands now recognized as physiological rhythms of neurocardiac control.^'* The low frequency ([LFl 0.05-0.12 Hz) component is primarily attributed to sympathetic tone especially dur-
Supported in part by a Grant from the Heart and Stroke Foundation of Ontario and the DeGroote Foundation. Address for reprints: Ernest L. Fallen. M.D., Mc:Master University Medical Centre. 1200 Main St. West. Hamilton. Ontario. Canada LBN 3Z5. Fax; (416) 521-5053. Received July 23, 1991: revision October 1, 1991: accepted October 11, 1991.
PACE, Vol. 15
ing orthostatic stress.^ ^ The high frequency ([HF] 0.15-0.35 Hz) band is said to represent cardiac vagal efferent activity.'^"'^ If true, the power spectrum thus obtained provides a unique window through which one can visualize and measure neural regulation of sinus node activity under different steady-state conditions. Establishing a direct linkage between the power spectrum and nerve impulse traffic has often been frustrating. Using selective pharmacological blockade or denervation procedures, several animal studies have shown a correspondence between the power spectrum and cardiac efferent neural activity.^^'^^ Similarly, under less controlled conditions, muscarinic blockade and beta blockers have been used to demonstrate reproducible changes in the peak power of LF and HF components of the power spectrum in man.'*'•''•^''''^ To date, however, there is little or no clinical evi-
February 1992
235
KAMATH. ET AL.
dence that a direct linkage exists between the energies contained in the heart rate autospectrum and selective stimulation of the cardiac nerves. We report here the results of a unique study in which the power spectrum of heart rate variability (HRV) was measured under different steady-state conditions in a patient with a programmable implanted vagal nerve stimulator.
Methods The study was performed on a 27-year-old male with a 12 year history of intractable complex partial seizures occurring at a rate > 10 per month. The seizures were unresponsive to a host of different anticonvulsant drugs, separately or in combination, at therapeutic serum levels. As part of a multicenter study to determine the ability of direct vagal afferent nerve stimulation to suppress seizure activity, the patient had a Medtronic Itrel stimulator (Medtronic, Inc., Minneapolis. MN, USA) implanted on June 10. 1990. The protocol was approved hy the Scientific and Ethics Committee of McMastor University. Helical electrodes were implanted around the left vagus nerve and a pnlse generator, powered by a single lithium thionyl chloride battery was inserted in the left chest, Tbe generator contained integrated circuitry and a microprocessor allowing transmission and reception of programming signals for parameter changes. A description of the technique and stimulation characteristics has been previously reported.^" The stimulator was activated at 20 Hz with 300 |xsec pulses at 1.25 V. The study was performed 9 months after implantation of the stimulator. Over this period the patient experienced a 50% reduction in seizure frequency without any adverse effects of stimulation. Following informed consent the study was carried out over a 3-day period in tbe patient's hospital hed. The patient was already taking valproic acid, carbamazepine, phenytoin, and pbenobarbital. No changes in medications or doses were made during the study period before or after vagal stimulation. The study began on the second day following hospital admission. The patient was familiar with the protocol having undergone a pretrial run several weeks after implantation. With tbe patient resting supine, three ECG electrodes were affixed to the anterior chest, care being taken
236
to ensure adequate R wave signals while avoiding DC interference and minimizing motion artifacts. The ECG bandwidtb was 1-50 Hz. Respiratory waveforms were recorded from a thermistor probe positioned subjacent to the nares. Continuous analog signals of ECG (lead II) and thermistor output were fed through a Tektronix oscilloscope (Tektronix, Inc.. Beavorton. OR. USA), digitized, and stored on an IBM AT Computer (IBM Gorp.. Armonk. NY, USA). Study Protocol During the first 24 hours of recording tbe vagal stimulator was left ON. The patient was asked to rest comfortably in bed for at least 45 minutes. The bead of the bed was elevated 20°. The room was quiet and care was taken to avoid repetitive external auditory or visual stimuli at frequencies heiow 1 Hz. No attempt was made to influence the pattern, depth, or rate of respiration. Continuous EGG records and respiratory waveforms were obtained over 45 minutes in the rest supine state during 1hour epochs at 7-8 AM, 3-4 PM, and 11-12 PM. TO impose an orthostatic load, tbe patient was then asked to stand erect and maintain a free standing posture for another 10 minutes while continuous ECG and respiratory signals were recorded as before. Tbe stimulator was then turned OEF and. the following day, the exact protocol as described above was repeated over the next 24 hours. The stimulator was then turned ON again and the same 24-hour protocol was repeated during the third day of study. This experimental design enabled us to measure and evaluate reproducibility of autospectral responses, circadian rhythms if present, and the effect of supine and ortbostatic stress during controlled steady-state wakeful conditions. Signal Processing and Analysis The EGG signals were digitized using a 12-bit analog-to-digital converter at 0.5 kHz and processed on an IBM AT computer equipped witb 640 KB RAM and a 100 MB bard disc drive. A peak detection algorithm for locating the R wave was implemented. HRV is a point event series. However power spectrnm analysis requires evenly spaced time domain data that is computed as follows. A beat-to-
February 1992
PACE, Vol. 15
VAGAL STIMULATION AND HR VARIABILITY
beat heart rate series was computed from the successive RR intervals and the resulting heart rate was resampled using linear interpolation to obtain an equally sampled time series. A record length of 256 sample points [approximately 2.2 min) was selected for power spectrum analysis. Due to the small variance in heart rates, the mean value of the HRV data was removed (demeaned]. An autoregressive model using linear predictive coding was then applied to the demeaned heart rate data.^' According to this model, for any given signal at instant 'n,' the signal amplitude x(n), is determined as the weighted linear sum of its 'p' previous samples, plus a random component. u(n):
k=l
Results
n - k] + u(n)
where. u(n), n ^ 1 . . . N form the samples of a stationary Gaussian white noise process. The parameters, ai, a-A, . . . Ap are to be determined and are called autoregressive parameters. The power spectral density for the signal x(n) can then be written as: P(f) =
P
|
1 + X ak exp(-j2nfkAt) k=i
J
where \t is the sampling interval andCT^is the variance of the white noise. The power contained in the frequency band ranging from DC to 0.02 Hz (referred to as DC power) often gives spuriously low values of the power in the LF and HF bands and any variations in the latter. In order to minimize such effects we passed the equally sampled HRV signals through a second order high pass Butterworth filter with a cut-off of 0.02 Hz. This was done twice; initially in the forward direction and then with the time series reversed during the second pass, The overall transfer function of such a process has a zero phase shift and the procedure provides a sharp filtering operation. The model order number used to derive the autoregressive coefficients was 16. Data Analysis HRV and its response to vagal nerve stimulation was analyzed using time domain and frequency domain autoregressive statistics. The coef-
PACE, Vol. 15
ficient of variation for heart rate and each of the spectral components' peak power was used to determine steady-state conditions. In addition to peak values, the area under each spectral component was integrated and normalized in relation to the total energy contained in the power spectrum. These normalized values are expressed as LF:HF area ratios. Over 150 autospectra were analyzed. The effect of nerve stimulation on each spectral variable at corresponding time epochs were analyzed with the two tailed Student's (-test. We accepted P < 0.05 as significant. All data are expressed as the means ± SD.
Average supine heart rates remained relatively constant throughout all recording periods ranging from 64 to 80 beats/min. The heart rate variance was unaltered regardless of whether the vagus nerve stimulator was ON or OFF (Table I). Moreover, there were only minor diurnal variations in mean heart rate. There was. however, a dramatic and reproducible change in the amplitude of LF and HF autospectral components whenever the vagal stimulator was switched either ON or OFF (Table I and Fig. 1). With the stimulator first turned ON, the average HF peak power was 87.1 (beats/min)^/Hz compared to an average LF peak power of 35.2 (beats/ min]^/Hz. This yielded a mean LF:HF peak ratio of 0.45 ± 0.2. The dominant vagal tone as expressed by the augmented HF peak power was seen during all recording sessions throughout days 1 and 3; i.e., whenever the stimulator was ON. Conversely, the LF peak power increased to levels significantly higher than the suppressed HF component when the stimulator was turned OFF (Table I and Fig. 1), The overall average LF:HF peak ratio increased from 0.64 to 1.99 (P < 0.001) for all epochs when the stimulator was OFF and decreased again to 0,76 when the stimulator was turned ON again. These average figures were combined for all hourly epochs because no significant circadian variations of the autospectra were observed. The same changes were seen when the total energies contained within each spectral band were measured as integrated areas underneath the peak power. This is illustrated as the LF:HF area ratios during different time periods (Fig. 2).
February 1992
237
KAMATH. ET AL.
12-Ap-91,
11 pm
ON
0.1Z • 0.24 FREQUENCY {Hz)
'0.42
Figure 1. Effect of vagal elecirostimulation on the heart rate autospectrum. Each panel contains a stacked series of power spectra of 2.2 minutes duration each. ON and OFF refer to 24-our status of sfimuJator. Note thai ihe low and higii frequency peak amplitudes are dramatically altered when stimulator turned OFF and then ON again. Bts/min = beats/min.
OFF STATUS OF STIMULATOR 7 AM
3 PM
ON
11 PM
Figure 2. Effect ofvagal stimulation on the ratio of the area subtended by the low frequency (LFJ band to the area beneath ihe high fHF) frequency band. There is littJe or no circadian variafion except perhaps at 11 PM when stimulator OFF.
238
February 1992
PACE. Vol. 15
VAGAL STIMULATION AND HR VARIABILITY
Table 1. Effect of Vagal Nerve Sfimulation on Heart Rate Autospectra LFpk Day of Test
Status of Stimulator
1
ON
2
OFF
3
ON
1
ON
2
OFF
3
ON
1
ON
2
OFF
3
ON
HFpk
HR
beats/min
(beats/min)^ HZ"^
7-8 am 67 (5.2) 67 (7.1) 63 (8.3) 3-4 pm 68 (8.6) 62 (5.8) 78 (6.3) 11-12 pm 64 (7.0) 80'" (6.2) 62 (5.8)
0.55 (0.09) 1.06*' (0.13) 0.85 (0.12)
87.1 (27.1) 44.5**
54
58.4 (14.8)
0.45 (0.2) 1.52** (0.32) 0.97 (0.27)
(8.3) 63.1' (16.1) 42.7 (11.4)
52.4 (7.6) 39.1* (6.6) 77.3 (23.2)
0.91 (0.25) 1.66" (0.49) 0.62 (0.27)
0.77 (0.09) 0.94* (0.14) 0.80 (0.11)
38 (9.5) 941(19.5) 38.9 (10.1)
78.1 (16.1) 39.1" (18.4) 64.1 (17.9)
0.55 (0.31) 2.79" (10) 0.68 (0.3)
0.61 (0.08) 1.52" (0.32) 0.75 (0.12)
46
(6.1)
HR = heart rate; LF = low frequency; HF = high frequency; pk = peak power; area ratio standard deviation. Each value refers to mean of 12-15 individual spectra. Bracketed numbers are standard deviations. * P < 0.05 when compared to corresponding value when stimulator ON during day 1. *' P < 0.01 when compared to corresponding value when stimulator ON during day 1.
PACE, Vol. 15
LF:HF area ratio
35.2 (9.4) 66.3" (9.0) (7.1)
No less interesting was the reproducible shifts in the LF peak frequency. The frequency at which the LF peak power oscillated shifted to the left when the stimulator was turned OFF (Table II). This was particularly noticeable in the early morning and late evening hours. The LF peak shifted rightward again when the vagus nerve was once again electrically stimulated. An example of the effect of orthostatic stress on the HR autospectrum is shown in Figures 3 and 4. During upright posture there was a dominant peak in the LF hand regardless of whether the vagus nerve was stimulated or not (Figs. 3B and 4B). The effect appeared more dramatic when the stimulator was ON although only a maximum of four spectra could be obtained during the upright
LFHF pk ratio
normalized power ratio; SD =
state. The LF peak appears somewhat broad based whether the stimulator was ON or OFF.
Discussion This n of 1 study supports the belief that a direct coupling exists between vagal nerve stimulation and heat-to-beat heart rate fluctuations as expressed by the power density spectrum. Using autoregressive modeling of the interval tachogram we have shown that electrostimulation of the left cervical vagus nerve caused: [1] reduction in the amplitude of the LF component; (2) a dramatic increase in the HF peak power; and (3) a rightward shift in the LF band. No appreciable change in time
February 1992
239
KAMATH, ET AL.
Tabie ii.
Frequency Shifts of LF Peak Power Time of Day
7-8 am 3-4 pm 11-12 pm
Stimulator ON (Day 1)
Stimulator OFF (Day 2)
Stimuiator ON (Day 3)
0,09 (0.012) 0.071 (0.03) 0,082 (0.022)
0,053 (0.008)*' 0.062 (0.009) 0,06 (0.007)'
0.061 (0,021) 0.086 (0.03) 0.091 (0.023)
Values refer to the frequency (Hz) at which the peak low frequency (LF) power oscillated. Bracketed numbers are standard deviations, ' P < 0,05 when compared to LF when stimulator ON during Day 1. " P < 0,01 when compared to LF when stimulator ON during Day 1.
domain parameters such as heart rate variance or respiratory rate were observed. To date, only indirect evidence exists that the heart rate power spectrum, as a noninvasive tool in man, reflects autonomic modulation of sinus node activity. Vagal blockade, for instance, virtually aholishes the HF component.'^•^^ Conversely, beta blockade attenuates the peak LF power'^'^'^'' and may actually augment the HF (vagal) peak power.^"'^^ Using selective pharmacological blockade, Weise^^'^" concluded that sympathetic neural input influences the behavior of
the LF component in the supine state and LF and HF components on standing. By the same token, parasympathetic stimulation appears to mediate all components in supine and standing states. These clinical studies have helped focus attention on the power spectrum as a noninvasive measure of neurocardiac function. They offer indirect and to some extent conflicting evidence that the two major spectral components are autonomically linked. In chronically instrumented dogs, Rimoldi et al.^^ showed that bilateral stellectomy and si-
FREQUENCY (Hi)
D.ia
€.14
0.30
FREQUENCY (Hi)
Figure 3. An example of the effect of ortbostatic slress on (he heart rate autospectrum. Panel A is a representative sample of tbe power spectrum when patient was supine with stimuJafor ON. Panel B shows the increase in low frequency peak power and suppression of bigb frequency component during standing. Stimulator remained ON. BTS/MIN = beats/min.
240
February 1992
PACE, Vol, 15
VAGAL STIMULATION AND HR VARIABILITY
SUM. OFF
STIM. OFF
0.13
O.IB
0.2
Variability in Man. The power spectrum of heart rate variability contains low frequency (LF - 0.08-0.12 Hz) and high frequency {HF = 0.18-0.30 Hz) components said to represent neurocardiac rhythms. To verify whether such a relationship exists ive report a unique study where the heart rate autospectrum was determined in a 28-year-o]d epiJepfic male patienf with an implanted vagal electrical stimulator. The stimulator was activated at 20 Hz, 300 fxsec pulse, and 1.25 V. Continuous ECG and respiratory waveform records were obtained over 45 minutes every 8 hours (7-8 AM; 3-4 PM; 11-12 PM) with the stimulator ON, then 24 hours OFF and then 24 hours ON again. The overall LF.HF peak ratio increased from 0.64 to J.99 (P < 0.001} after the stimulator was turned OFF. There was a dramatic increase in the LF peak power f> 60%) and a corresponding decrease in the HF peak power f> 65%) when the stimulator was turned OFF. These vaJues were reversed when the stimulator was turned ON again, fn the early morning and late evening hours, there was a significant rightward shift in the LF peak power frequency (average 0.057 to 0.075 Hz) whenever the stimulator was ON. Otherwise, there were no significant circadian variations in any of the autospectral components. The results demonstrate an unequivocal relationship between selective vagal nerve electrostimulation and alterations in the heart rate autospectrum. (PACE, Vol. 15, February 1992) autospectrum, circadian, orthostasis, sympatho-vagai balance
Introduction Under steady-state conditions, the variability of the heartbeat is said to reflect cardiac autonomic activity.^*^ By transforming these beat-to-beat fluctuations into their frequency components a power spectrum is obtained that contains two distinct spectral bands now recognized as physiological rhythms of neurocardiac control.^'* The low frequency ([LFl 0.05-0.12 Hz) component is primarily attributed to sympathetic tone especially dur-
Supported in part by a Grant from the Heart and Stroke Foundation of Ontario and the DeGroote Foundation. Address for reprints: Ernest L. Fallen. M.D., Mc:Master University Medical Centre. 1200 Main St. West. Hamilton. Ontario. Canada LBN 3Z5. Fax; (416) 521-5053. Received July 23, 1991: revision October 1, 1991: accepted October 11, 1991.
PACE, Vol. 15
ing orthostatic stress.^ ^ The high frequency ([HF] 0.15-0.35 Hz) band is said to represent cardiac vagal efferent activity.'^"'^ If true, the power spectrum thus obtained provides a unique window through which one can visualize and measure neural regulation of sinus node activity under different steady-state conditions. Establishing a direct linkage between the power spectrum and nerve impulse traffic has often been frustrating. Using selective pharmacological blockade or denervation procedures, several animal studies have shown a correspondence between the power spectrum and cardiac efferent neural activity.^^'^^ Similarly, under less controlled conditions, muscarinic blockade and beta blockers have been used to demonstrate reproducible changes in the peak power of LF and HF components of the power spectrum in man.'*'•''•^''''^ To date, however, there is little or no clinical evi-
February 1992
235
KAMATH. ET AL.
dence that a direct linkage exists between the energies contained in the heart rate autospectrum and selective stimulation of the cardiac nerves. We report here the results of a unique study in which the power spectrum of heart rate variability (HRV) was measured under different steady-state conditions in a patient with a programmable implanted vagal nerve stimulator.
Methods The study was performed on a 27-year-old male with a 12 year history of intractable complex partial seizures occurring at a rate > 10 per month. The seizures were unresponsive to a host of different anticonvulsant drugs, separately or in combination, at therapeutic serum levels. As part of a multicenter study to determine the ability of direct vagal afferent nerve stimulation to suppress seizure activity, the patient had a Medtronic Itrel stimulator (Medtronic, Inc., Minneapolis. MN, USA) implanted on June 10. 1990. The protocol was approved hy the Scientific and Ethics Committee of McMastor University. Helical electrodes were implanted around the left vagus nerve and a pnlse generator, powered by a single lithium thionyl chloride battery was inserted in the left chest, Tbe generator contained integrated circuitry and a microprocessor allowing transmission and reception of programming signals for parameter changes. A description of the technique and stimulation characteristics has been previously reported.^" The stimulator was activated at 20 Hz with 300 |xsec pulses at 1.25 V. The study was performed 9 months after implantation of the stimulator. Over this period the patient experienced a 50% reduction in seizure frequency without any adverse effects of stimulation. Following informed consent the study was carried out over a 3-day period in tbe patient's hospital hed. The patient was already taking valproic acid, carbamazepine, phenytoin, and pbenobarbital. No changes in medications or doses were made during the study period before or after vagal stimulation. The study began on the second day following hospital admission. The patient was familiar with the protocol having undergone a pretrial run several weeks after implantation. With tbe patient resting supine, three ECG electrodes were affixed to the anterior chest, care being taken
236
to ensure adequate R wave signals while avoiding DC interference and minimizing motion artifacts. The ECG bandwidtb was 1-50 Hz. Respiratory waveforms were recorded from a thermistor probe positioned subjacent to the nares. Continuous analog signals of ECG (lead II) and thermistor output were fed through a Tektronix oscilloscope (Tektronix, Inc.. Beavorton. OR. USA), digitized, and stored on an IBM AT Computer (IBM Gorp.. Armonk. NY, USA). Study Protocol During the first 24 hours of recording tbe vagal stimulator was left ON. The patient was asked to rest comfortably in bed for at least 45 minutes. The bead of the bed was elevated 20°. The room was quiet and care was taken to avoid repetitive external auditory or visual stimuli at frequencies heiow 1 Hz. No attempt was made to influence the pattern, depth, or rate of respiration. Continuous EGG records and respiratory waveforms were obtained over 45 minutes in the rest supine state during 1hour epochs at 7-8 AM, 3-4 PM, and 11-12 PM. TO impose an orthostatic load, tbe patient was then asked to stand erect and maintain a free standing posture for another 10 minutes while continuous ECG and respiratory signals were recorded as before. Tbe stimulator was then turned OEF and. the following day, the exact protocol as described above was repeated over the next 24 hours. The stimulator was then turned ON again and the same 24-hour protocol was repeated during the third day of study. This experimental design enabled us to measure and evaluate reproducibility of autospectral responses, circadian rhythms if present, and the effect of supine and ortbostatic stress during controlled steady-state wakeful conditions. Signal Processing and Analysis The EGG signals were digitized using a 12-bit analog-to-digital converter at 0.5 kHz and processed on an IBM AT computer equipped witb 640 KB RAM and a 100 MB bard disc drive. A peak detection algorithm for locating the R wave was implemented. HRV is a point event series. However power spectrnm analysis requires evenly spaced time domain data that is computed as follows. A beat-to-
February 1992
PACE, Vol. 15
VAGAL STIMULATION AND HR VARIABILITY
beat heart rate series was computed from the successive RR intervals and the resulting heart rate was resampled using linear interpolation to obtain an equally sampled time series. A record length of 256 sample points [approximately 2.2 min) was selected for power spectrum analysis. Due to the small variance in heart rates, the mean value of the HRV data was removed (demeaned]. An autoregressive model using linear predictive coding was then applied to the demeaned heart rate data.^' According to this model, for any given signal at instant 'n,' the signal amplitude x(n), is determined as the weighted linear sum of its 'p' previous samples, plus a random component. u(n):
k=l
Results
n - k] + u(n)
where. u(n), n ^ 1 . . . N form the samples of a stationary Gaussian white noise process. The parameters, ai, a-A, . . . Ap are to be determined and are called autoregressive parameters. The power spectral density for the signal x(n) can then be written as: P(f) =
P
|
1 + X ak exp(-j2nfkAt) k=i
J
where \t is the sampling interval andCT^is the variance of the white noise. The power contained in the frequency band ranging from DC to 0.02 Hz (referred to as DC power) often gives spuriously low values of the power in the LF and HF bands and any variations in the latter. In order to minimize such effects we passed the equally sampled HRV signals through a second order high pass Butterworth filter with a cut-off of 0.02 Hz. This was done twice; initially in the forward direction and then with the time series reversed during the second pass, The overall transfer function of such a process has a zero phase shift and the procedure provides a sharp filtering operation. The model order number used to derive the autoregressive coefficients was 16. Data Analysis HRV and its response to vagal nerve stimulation was analyzed using time domain and frequency domain autoregressive statistics. The coef-
PACE, Vol. 15
ficient of variation for heart rate and each of the spectral components' peak power was used to determine steady-state conditions. In addition to peak values, the area under each spectral component was integrated and normalized in relation to the total energy contained in the power spectrum. These normalized values are expressed as LF:HF area ratios. Over 150 autospectra were analyzed. The effect of nerve stimulation on each spectral variable at corresponding time epochs were analyzed with the two tailed Student's (-test. We accepted P < 0.05 as significant. All data are expressed as the means ± SD.
Average supine heart rates remained relatively constant throughout all recording periods ranging from 64 to 80 beats/min. The heart rate variance was unaltered regardless of whether the vagus nerve stimulator was ON or OFF (Table I). Moreover, there were only minor diurnal variations in mean heart rate. There was. however, a dramatic and reproducible change in the amplitude of LF and HF autospectral components whenever the vagal stimulator was switched either ON or OFF (Table I and Fig. 1). With the stimulator first turned ON, the average HF peak power was 87.1 (beats/min)^/Hz compared to an average LF peak power of 35.2 (beats/ min]^/Hz. This yielded a mean LF:HF peak ratio of 0.45 ± 0.2. The dominant vagal tone as expressed by the augmented HF peak power was seen during all recording sessions throughout days 1 and 3; i.e., whenever the stimulator was ON. Conversely, the LF peak power increased to levels significantly higher than the suppressed HF component when the stimulator was turned OFF (Table I and Fig. 1), The overall average LF:HF peak ratio increased from 0.64 to 1.99 (P < 0.001) for all epochs when the stimulator was OFF and decreased again to 0,76 when the stimulator was turned ON again. These average figures were combined for all hourly epochs because no significant circadian variations of the autospectra were observed. The same changes were seen when the total energies contained within each spectral band were measured as integrated areas underneath the peak power. This is illustrated as the LF:HF area ratios during different time periods (Fig. 2).
February 1992
237
KAMATH. ET AL.
12-Ap-91,
11 pm
ON
0.1Z • 0.24 FREQUENCY {Hz)
'0.42
Figure 1. Effect of vagal elecirostimulation on the heart rate autospectrum. Each panel contains a stacked series of power spectra of 2.2 minutes duration each. ON and OFF refer to 24-our status of sfimuJator. Note thai ihe low and higii frequency peak amplitudes are dramatically altered when stimulator turned OFF and then ON again. Bts/min = beats/min.
OFF STATUS OF STIMULATOR 7 AM
3 PM
ON
11 PM
Figure 2. Effect ofvagal stimulation on the ratio of the area subtended by the low frequency (LFJ band to the area beneath ihe high fHF) frequency band. There is littJe or no circadian variafion except perhaps at 11 PM when stimulator OFF.
238
February 1992
PACE. Vol. 15
VAGAL STIMULATION AND HR VARIABILITY
Table 1. Effect of Vagal Nerve Sfimulation on Heart Rate Autospectra LFpk Day of Test
Status of Stimulator
1
ON
2
OFF
3
ON
1
ON
2
OFF
3
ON
1
ON
2
OFF
3
ON
HFpk
HR
beats/min
(beats/min)^ HZ"^
7-8 am 67 (5.2) 67 (7.1) 63 (8.3) 3-4 pm 68 (8.6) 62 (5.8) 78 (6.3) 11-12 pm 64 (7.0) 80'" (6.2) 62 (5.8)
0.55 (0.09) 1.06*' (0.13) 0.85 (0.12)
87.1 (27.1) 44.5**
54
58.4 (14.8)
0.45 (0.2) 1.52** (0.32) 0.97 (0.27)
(8.3) 63.1' (16.1) 42.7 (11.4)
52.4 (7.6) 39.1* (6.6) 77.3 (23.2)
0.91 (0.25) 1.66" (0.49) 0.62 (0.27)
0.77 (0.09) 0.94* (0.14) 0.80 (0.11)
38 (9.5) 941(19.5) 38.9 (10.1)
78.1 (16.1) 39.1" (18.4) 64.1 (17.9)
0.55 (0.31) 2.79" (10) 0.68 (0.3)
0.61 (0.08) 1.52" (0.32) 0.75 (0.12)
46
(6.1)
HR = heart rate; LF = low frequency; HF = high frequency; pk = peak power; area ratio standard deviation. Each value refers to mean of 12-15 individual spectra. Bracketed numbers are standard deviations. * P < 0.05 when compared to corresponding value when stimulator ON during day 1. *' P < 0.01 when compared to corresponding value when stimulator ON during day 1.
PACE, Vol. 15
LF:HF area ratio
35.2 (9.4) 66.3" (9.0) (7.1)
No less interesting was the reproducible shifts in the LF peak frequency. The frequency at which the LF peak power oscillated shifted to the left when the stimulator was turned OFF (Table II). This was particularly noticeable in the early morning and late evening hours. The LF peak shifted rightward again when the vagus nerve was once again electrically stimulated. An example of the effect of orthostatic stress on the HR autospectrum is shown in Figures 3 and 4. During upright posture there was a dominant peak in the LF hand regardless of whether the vagus nerve was stimulated or not (Figs. 3B and 4B). The effect appeared more dramatic when the stimulator was ON although only a maximum of four spectra could be obtained during the upright
LFHF pk ratio
normalized power ratio; SD =
state. The LF peak appears somewhat broad based whether the stimulator was ON or OFF.
Discussion This n of 1 study supports the belief that a direct coupling exists between vagal nerve stimulation and heat-to-beat heart rate fluctuations as expressed by the power density spectrum. Using autoregressive modeling of the interval tachogram we have shown that electrostimulation of the left cervical vagus nerve caused: [1] reduction in the amplitude of the LF component; (2) a dramatic increase in the HF peak power; and (3) a rightward shift in the LF band. No appreciable change in time
February 1992
239
KAMATH, ET AL.
Tabie ii.
Frequency Shifts of LF Peak Power Time of Day
7-8 am 3-4 pm 11-12 pm
Stimulator ON (Day 1)
Stimulator OFF (Day 2)
Stimuiator ON (Day 3)
0,09 (0.012) 0.071 (0.03) 0,082 (0.022)
0,053 (0.008)*' 0.062 (0.009) 0,06 (0.007)'
0.061 (0,021) 0.086 (0.03) 0.091 (0.023)
Values refer to the frequency (Hz) at which the peak low frequency (LF) power oscillated. Bracketed numbers are standard deviations, ' P < 0,05 when compared to LF when stimulator ON during Day 1. " P < 0,01 when compared to LF when stimulator ON during Day 1.
domain parameters such as heart rate variance or respiratory rate were observed. To date, only indirect evidence exists that the heart rate power spectrum, as a noninvasive tool in man, reflects autonomic modulation of sinus node activity. Vagal blockade, for instance, virtually aholishes the HF component.'^•^^ Conversely, beta blockade attenuates the peak LF power'^'^'^'' and may actually augment the HF (vagal) peak power.^"'^^ Using selective pharmacological blockade, Weise^^'^" concluded that sympathetic neural input influences the behavior of
the LF component in the supine state and LF and HF components on standing. By the same token, parasympathetic stimulation appears to mediate all components in supine and standing states. These clinical studies have helped focus attention on the power spectrum as a noninvasive measure of neurocardiac function. They offer indirect and to some extent conflicting evidence that the two major spectral components are autonomically linked. In chronically instrumented dogs, Rimoldi et al.^^ showed that bilateral stellectomy and si-
FREQUENCY (Hi)
D.ia
€.14
0.30
FREQUENCY (Hi)
Figure 3. An example of the effect of ortbostatic slress on (he heart rate autospectrum. Panel A is a representative sample of tbe power spectrum when patient was supine with stimuJafor ON. Panel B shows the increase in low frequency peak power and suppression of bigb frequency component during standing. Stimulator remained ON. BTS/MIN = beats/min.
240
February 1992
PACE, Vol, 15
VAGAL STIMULATION AND HR VARIABILITY
SUM. OFF
STIM. OFF
0.13
O.IB
0.2
