CONGE6TiVE
HEART FAILURE
Clinical, Hemodynamic and Sympathetic Neural Correlates Of Heart Rate Variability in Congestive Heart Failure Michael G. Kienzle, MD, David W. Ferguson, MD, Clayton L. Birkett, MSBME, Glenn A. Myers, PhD, W’illiam J. Berg, MD, and D. James Mariano, MD
Heart rate (HR) variabflity has long been recogniied as a sfgn of cardiac health. In the presence of heart disease, HR variability decrea-, an observation that has been associated with poor prognosfs in .a number of recent studies. HR variability is partkufariy altered in congestive heart faikue (CHP), a condition associated with a number of typkai functional, hemodynamic and neurohumoral alterattons. Yhe refation of meaerements of HR variabiilty to these abnormalitfes in patients with heart failure has not been carefully examined. Twenty-three patients (19 men, ‘4 Lvomen, mean age 49 years) wfth New York Heart Association cku II to IV CHF were studied prospectively without card&c medications; radionuciide ventriculography, right-shied heart catheterization, peroneai mf&oneurography, plasma norepinephrine and 24- to 49hour ambulatory electrocardiography were performed. Average RR interval and its standard deviation, and HR power spectrum (0 to 0.5, 0.05 to 0.15 and 0.2 to 0.5 Hz) were derived from the ambubtory electrocardiographic recordings and compared wfth left ventricubr ejection fraction, thermodilution cardiac output, pulmonary arterial wedge pressure, New York Heart Association class, age, muscfe sympathetii nerve activfty (peroneal nerve) and norepinephrine level by linear regressfon. None of the measures of HR variability were dgnRkantfy related to age, left ventricular ejectfon fraction, cardiac output or functional ctassificatfon, whereas the 0.05 to 0.15 and 0.20 to 0.50 Hz components were weakly but significantly retated to cardiac output (r = 0.49 and 0.42, p = 9.02 and 0.045, respectively). in contrast, a generally .&ronger and negative relation was demonstrated between spectral and nonspectral measurements of HR variability, and indicators of sympathoexcRa&on, muscle sympathetic nerve ac-
tivity and plasma norepinephrine. This relation was particularly strong with the 0.05 to 0.15 and 0.2 to 0.5 Hz spectral components, and RR standard deviation (r = -0.55 to -0.76, p = 0.04 to O.otu). it is concluded that the decrease in spectral and nonspectral measurements of HR variabitity that accompanies CHF is not an indicator of the severity of disease, as usually measured clinically, but rather a marker of sympathoexcitation. Therefore, measurements of HR variability may provide noninvasive and unique information regarding neurohumorai maladaptation in heart failure patients. (Am J Cardioi 1992;69:761-767)
here are prominent neurohormonal alterations that characterize heart failure. Evidenceof sympathoexcitation is readily available; circulating !evelsof norepinephrine, vasopressinand renin are substantially elevated.1-3Recently, direct measurementof sympathetic neural activity was used in humans to examine the sympathoexcitation associatedwith clinical heart failure.4,5 Arterial and cardiopulmonary baroreflex control of sympathetic and parasympathetic activity is impaired in both experimental and human heart failure 6-9 Heart rate (HR) variability has long been noted to be abnormal in the presenceof heart diseaseor in alterations in cardiac innervation, as in diabetes mellitus.l”-13 Recently, investigators showedthat diminished HR variability is associatedwith poor prognostic outcome in patients with organic heart diseaseand may provide unique information comparedwith other important prognostic indicators such as left ventricular ejection fraction and ventricular ectopy.14-l6Some investigators maintained that HR variability reflects the severity of ventricular dysfunction generally, but little information is available to substantiate this belief.12 From the Clinical CardiovascularPhysiologyLaboratory, Cardiovascu- Even less data are currently available comparing HR lar and Clinical Research Centers, and Cardiovascular Division, De- variability with other confirmatory measurementsof the partments of Internal Medicine and Biomedical Engineering,Universi- neurohormonal state in heart failure. The purposeof the ty of Iowa, Iowa City, Iowa. Dr. KienzJeis supportedby grants from the General Clinical ResearchCenter (National Institutes of Health RR presentstudy was to prospectivelycomparespectral and 59, Bethesda,Maryland) and the Iowa Affiliate of the American Heart nonspectral HR variability (derived from long-term Association (IA 88-G-18), Des Moines, Iowa. Manuscript received ambulatory electrocardiographic monitoring) with seAugust 26,199l; revisedmanuscript receivedNovember 13,1991,and lected clinical, hemodynamic and neurohumoral meaacceptedNovember 14. Address for reprints: Michael G. Kienzle, MD, Department of surementsin a group of patients with moderate to seInternal Medicine, 4426 John Colloton Pavilion, University of Iowa vere congestive heart failure (CHF) to better underHospitals and Clinics, Iowa City, Iowa 52242. stand.the significance of alterations of HR variability.
T
HEART RATE VARIABILITY IN HEART FAILURE 761
MmHODS Study patIen& Twenty-three patients (19 men, age range 24 to 74 years) were prospectively studied. The largest number of patients had idiopathic cardiomyopathy, whereas a smaller number had documented coronary artery disease.CHF was moderate to severe,as evidencedby a predominanceof New York Heart Association class III to IV and by a substantially reduced radionuclide ejection fraction in most patients (mean 0.21 f 0.07). Patients uniformly needed Ll cardiac medication for control of symptoms. Patients were not studied within 1 month of myocardial infarction. Digitalis glycosideswere discontinued 1 week before study and confirmed with a negative digoxin assay. Most oral medications were discontinued for L4 half-lives, and diuretics for 12 hours before the study. In all patients, normal electrolyte status and sinus rhythm were confu?nedbefore entering the protocol. Informed written consent was obtained in all cases,and the studies were approved by the institutional review board for human studies. Patients were admitted to the clinical researchcenter. During the first 24 to 48 hours, ambulatory electrocardiographic monitoring was performed. During this time patients were encouragedto engagein activities on the unit that simulated thoseduring usual daily life including sleeping and waking routines, and limited physical activity. Hemodynamic, microneurographic and hormonal measurementswere obtained on day 2 or 3. Details of thesestudies are outlined later. vi v Patients underwent resting hemodynamic measurements at the cardiac catheterization laboratory. Hemodynamic measurements at rest included systemic arterial and right heart pressures. Thermodilution cardiac output was determined using 10 ml of iced saline solution injections (mean of 5 injections) with an SAT-l Oximeter/Cardisc Output Computer (American Edwards Laboratories, Santa Ana, California), and simultaneousFick car-
w2o PulmE;G&tery
diac output was determined by measuring oxygen consumption with a Metabolic Rate Meter (MRM, Waters Instruments, Rochester, MN) and by obtaining blood samplesfor arterial-mixed venous oxygen content difference. Left ventricular ejection fraction was determined during normal sinus rhythm by radionuclide ventriculography within 2 days of the other studies. Mieroneurographii
PAP q 44i22 RAP I 7.4
(O-50arW Rbht Atrial
lhxnIre
(g-20 mmHg)
0
I
Electrocardiogram
MSNA (bursts/mln)
762
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 69
reeordii
of muscle sympathed-
ie nerve activity: In 16 of the 23 patients, we obtained multiunit recordings of postganglionic muscle nerve activity recordedfrom a muscle nerve fascicle in the peroneal nerve posterior to the fibular head using previously validated techniques.4,5~17 Briefly described, recordings were obtained by percutaneousinsertion of tungsten microelectrodes into the peroneal nerve. The electrodes were connectedto a preamplifier, and the recorded signal was directed through a bandpass filter and on through an amplitude discriminator, and visually displayed simultaneously with an audible signal. To facilitate recording and analysis, the filtered neurogram was processedin a resistance-capacitanceintegrating network resulting in a mean voltage display of the neural activity. Resting nerve activity was recorded for up to 10 minutes to ensure that a stable and representative nerve recording had beenobtained. Individual burst fre quency was expressedas bursts/min (intraobservervariability 5% and interobservervariability 10%).An example of measurementsobtained in each patient is shown in Figure 1. Plasma hormone determinathw Arterial blood samplesfor plasma norepinephrine were obtained from indwelling arterial catheters at the completion of the consecutive5minute recordings of hemodynamicsand sympathetic nerve activity. Norepinephrine levels were assayed by a single-isotope radioenzymatic method18 having a variability of 40% of complexes classified as ectopic or noise were not used. Mean RR was subtracted from eachsegment, and a Hanning window was applied. The power spectrum of the RR interval signal was calculated using Welch’si variant of the Bartlett procedure (overlapping segmentsof windowed intervals are used to reduce the variance of the power spectral estimate). The power spectrum was calculated as the squared magnitude of the fast Fourier transform for each segment. Segmentswere overlapped by 50% and averaged for each hour. Amplitude measurementsof the power spectrum were multiplied by a factor of 4, and area measurementsby 2.66 to correct for the Hanning window.20 In this study, we determined the spectral power over 3 frequency regions of interest: 0.05 to 0.15, 0.2 to 0.5 and over the entire 0 to 0.5 Hz band. We also determined hourly mean and standard deviations of the RR interval from the sameRR files as a more general nonspectral indicator of HR variability. Statianalysis: Data are expressedas mean f standard deviation. Variables were comparedstatistically using a linear regression analysis. Power spectral measurementsunderwent common log transformation before application of linear regression.
TABLE
I Group Results
Power 0.05-O. 15 (ms2/Hz) Power 0.20-0.50 (ms2/Hz) Power O-0.50 (ms2) RR interval (ms) RR SD (ms) LVEF TDCO (Iltersimin) LVFP (mm Hg) MSNA (bursts/mid Norepi (picogram/ml)
n
Mean + SD
23
44.7 2 62.7
1.2
23
7.8 t a.2
0.45
23 23 23 23 23 14 16 14
la.6 + 736 r 60.0? 0.21 r 4.3 * 242 62.3 + 745 2
15.5 124 23 0.07 1.7 12 23.1 390
Minimum
0.6 523 14 0.08 1.6 6 la.8 128
LVEF = left ventricular ejectlon fraction; LVFP = left ventricular MSNA = muscle sympathetic nerve activity: Norep] = norepinephrine spectral power; TDCO = thermal dilution cardlacoutput.
Maximum 290
28 62.7 944 104 0.34 9.3 46 99.4 1,530 filling pressure; level; Power =
pressedas average area) was 18.6 f 15.5 ms2 (range 0.6 to 62.7). Rehtions among measumd vahabkm The overall relation betweenthe clinical, hemodynamic and neurohormonal variables, and HR variability is illustrated by the 2 patient examplesshown in Figure 2. These 2 patients resemble each other clinically in regard to age, functional class and left ventricular ejection fraction. The patient on the left had a lower filling pressureand higher cardiac output than the one on the right. However, they differ the greatest relative to the measurements of muscle sympathetic nerve activity and plasma norepinephrine. The patient on the left (with the least evidencefor sympathoexcitation) also had greater power apparent over the entire power spectrum. In contrast, the patient on the right had markedly elevatedlevels of both norepinephrine and sympathetic nerve traffic, no visible peak in the high- and midfrequency ranges, and decreasedpower in the lowest frequencies. The relation between the power spectral measurements and the other measuredvariables are tabulated in Table II and shown for selectedcomparisonsin Figure 3. As is apparent, RR interval variability was not significantly related to age, functional class, left ventricular filling pressure or ejection fraction. In contrast, the RESULTS amount of RR standard deviation and both high- and Group data: Group data for the study cohort is tabu- midfrequency spectral power had a positive and signifilated in Table I. Substantial evidenceis apparent sup cant relation to cardiac output. Both RR standard deviporting the presenceand severity of CHF in these pa- ation and the 3 spectral power measurements(especialtients. As mentioned previously, the left ventricular ly the midfrequencies) generally displayed a negative ejection fraction is markedly reduced for the entire and significant relation to 1 or both indicators of symgroup. Left ventricular filling pressure (indicated by pathcexcitation, plasma norepinephrine and muscle pulmonary capillary wedge pressure) was elevated in sympathetic nerve activity. As one may expect, spectral the vast majority of patients, and cardiac output ranged power, RR interval and its standard deviation were sigfrom 1.6 to 9.3 liters/mm (mean 4.3). Muscle sympa- nificantly related, as were muscle sympathetic nerve acthetic nerve activity was clearly elevatedcomparedwith tivity and plasma norepinephrine (r = 0.79, p
HEART FAILURE
Clinical, Hemodynamic and Sympathetic Neural Correlates Of Heart Rate Variability in Congestive Heart Failure Michael G. Kienzle, MD, David W. Ferguson, MD, Clayton L. Birkett, MSBME, Glenn A. Myers, PhD, W’illiam J. Berg, MD, and D. James Mariano, MD
Heart rate (HR) variabflity has long been recogniied as a sfgn of cardiac health. In the presence of heart disease, HR variability decrea-, an observation that has been associated with poor prognosfs in .a number of recent studies. HR variability is partkufariy altered in congestive heart faikue (CHP), a condition associated with a number of typkai functional, hemodynamic and neurohumoral alterattons. Yhe refation of meaerements of HR variabiilty to these abnormalitfes in patients with heart failure has not been carefully examined. Twenty-three patients (19 men, ‘4 Lvomen, mean age 49 years) wfth New York Heart Association cku II to IV CHF were studied prospectively without card&c medications; radionuciide ventriculography, right-shied heart catheterization, peroneai mf&oneurography, plasma norepinephrine and 24- to 49hour ambulatory electrocardiography were performed. Average RR interval and its standard deviation, and HR power spectrum (0 to 0.5, 0.05 to 0.15 and 0.2 to 0.5 Hz) were derived from the ambubtory electrocardiographic recordings and compared wfth left ventricubr ejection fraction, thermodilution cardiac output, pulmonary arterial wedge pressure, New York Heart Association class, age, muscfe sympathetii nerve activfty (peroneal nerve) and norepinephrine level by linear regressfon. None of the measures of HR variability were dgnRkantfy related to age, left ventricular ejectfon fraction, cardiac output or functional ctassificatfon, whereas the 0.05 to 0.15 and 0.20 to 0.50 Hz components were weakly but significantly retated to cardiac output (r = 0.49 and 0.42, p = 9.02 and 0.045, respectively). in contrast, a generally .&ronger and negative relation was demonstrated between spectral and nonspectral measurements of HR variability, and indicators of sympathoexcRa&on, muscle sympathetic nerve ac-
tivity and plasma norepinephrine. This relation was particularly strong with the 0.05 to 0.15 and 0.2 to 0.5 Hz spectral components, and RR standard deviation (r = -0.55 to -0.76, p = 0.04 to O.otu). it is concluded that the decrease in spectral and nonspectral measurements of HR variabitity that accompanies CHF is not an indicator of the severity of disease, as usually measured clinically, but rather a marker of sympathoexcitation. Therefore, measurements of HR variability may provide noninvasive and unique information regarding neurohumorai maladaptation in heart failure patients. (Am J Cardioi 1992;69:761-767)
here are prominent neurohormonal alterations that characterize heart failure. Evidenceof sympathoexcitation is readily available; circulating !evelsof norepinephrine, vasopressinand renin are substantially elevated.1-3Recently, direct measurementof sympathetic neural activity was used in humans to examine the sympathoexcitation associatedwith clinical heart failure.4,5 Arterial and cardiopulmonary baroreflex control of sympathetic and parasympathetic activity is impaired in both experimental and human heart failure 6-9 Heart rate (HR) variability has long been noted to be abnormal in the presenceof heart diseaseor in alterations in cardiac innervation, as in diabetes mellitus.l”-13 Recently, investigators showedthat diminished HR variability is associatedwith poor prognostic outcome in patients with organic heart diseaseand may provide unique information comparedwith other important prognostic indicators such as left ventricular ejection fraction and ventricular ectopy.14-l6Some investigators maintained that HR variability reflects the severity of ventricular dysfunction generally, but little information is available to substantiate this belief.12 From the Clinical CardiovascularPhysiologyLaboratory, Cardiovascu- Even less data are currently available comparing HR lar and Clinical Research Centers, and Cardiovascular Division, De- variability with other confirmatory measurementsof the partments of Internal Medicine and Biomedical Engineering,Universi- neurohormonal state in heart failure. The purposeof the ty of Iowa, Iowa City, Iowa. Dr. KienzJeis supportedby grants from the General Clinical ResearchCenter (National Institutes of Health RR presentstudy was to prospectivelycomparespectral and 59, Bethesda,Maryland) and the Iowa Affiliate of the American Heart nonspectral HR variability (derived from long-term Association (IA 88-G-18), Des Moines, Iowa. Manuscript received ambulatory electrocardiographic monitoring) with seAugust 26,199l; revisedmanuscript receivedNovember 13,1991,and lected clinical, hemodynamic and neurohumoral meaacceptedNovember 14. Address for reprints: Michael G. Kienzle, MD, Department of surementsin a group of patients with moderate to seInternal Medicine, 4426 John Colloton Pavilion, University of Iowa vere congestive heart failure (CHF) to better underHospitals and Clinics, Iowa City, Iowa 52242. stand.the significance of alterations of HR variability.
T
HEART RATE VARIABILITY IN HEART FAILURE 761
MmHODS Study patIen& Twenty-three patients (19 men, age range 24 to 74 years) were prospectively studied. The largest number of patients had idiopathic cardiomyopathy, whereas a smaller number had documented coronary artery disease.CHF was moderate to severe,as evidencedby a predominanceof New York Heart Association class III to IV and by a substantially reduced radionuclide ejection fraction in most patients (mean 0.21 f 0.07). Patients uniformly needed Ll cardiac medication for control of symptoms. Patients were not studied within 1 month of myocardial infarction. Digitalis glycosideswere discontinued 1 week before study and confirmed with a negative digoxin assay. Most oral medications were discontinued for L4 half-lives, and diuretics for 12 hours before the study. In all patients, normal electrolyte status and sinus rhythm were confu?nedbefore entering the protocol. Informed written consent was obtained in all cases,and the studies were approved by the institutional review board for human studies. Patients were admitted to the clinical researchcenter. During the first 24 to 48 hours, ambulatory electrocardiographic monitoring was performed. During this time patients were encouragedto engagein activities on the unit that simulated thoseduring usual daily life including sleeping and waking routines, and limited physical activity. Hemodynamic, microneurographic and hormonal measurementswere obtained on day 2 or 3. Details of thesestudies are outlined later. vi v Patients underwent resting hemodynamic measurements at the cardiac catheterization laboratory. Hemodynamic measurements at rest included systemic arterial and right heart pressures. Thermodilution cardiac output was determined using 10 ml of iced saline solution injections (mean of 5 injections) with an SAT-l Oximeter/Cardisc Output Computer (American Edwards Laboratories, Santa Ana, California), and simultaneousFick car-
w2o PulmE;G&tery
diac output was determined by measuring oxygen consumption with a Metabolic Rate Meter (MRM, Waters Instruments, Rochester, MN) and by obtaining blood samplesfor arterial-mixed venous oxygen content difference. Left ventricular ejection fraction was determined during normal sinus rhythm by radionuclide ventriculography within 2 days of the other studies. Mieroneurographii
PAP q 44i22 RAP I 7.4
(O-50arW Rbht Atrial
lhxnIre
(g-20 mmHg)
0
I
Electrocardiogram
MSNA (bursts/mln)
762
THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 69
reeordii
of muscle sympathed-
ie nerve activity: In 16 of the 23 patients, we obtained multiunit recordings of postganglionic muscle nerve activity recordedfrom a muscle nerve fascicle in the peroneal nerve posterior to the fibular head using previously validated techniques.4,5~17 Briefly described, recordings were obtained by percutaneousinsertion of tungsten microelectrodes into the peroneal nerve. The electrodes were connectedto a preamplifier, and the recorded signal was directed through a bandpass filter and on through an amplitude discriminator, and visually displayed simultaneously with an audible signal. To facilitate recording and analysis, the filtered neurogram was processedin a resistance-capacitanceintegrating network resulting in a mean voltage display of the neural activity. Resting nerve activity was recorded for up to 10 minutes to ensure that a stable and representative nerve recording had beenobtained. Individual burst fre quency was expressedas bursts/min (intraobservervariability 5% and interobservervariability 10%).An example of measurementsobtained in each patient is shown in Figure 1. Plasma hormone determinathw Arterial blood samplesfor plasma norepinephrine were obtained from indwelling arterial catheters at the completion of the consecutive5minute recordings of hemodynamicsand sympathetic nerve activity. Norepinephrine levels were assayed by a single-isotope radioenzymatic method18 having a variability of 40% of complexes classified as ectopic or noise were not used. Mean RR was subtracted from eachsegment, and a Hanning window was applied. The power spectrum of the RR interval signal was calculated using Welch’si variant of the Bartlett procedure (overlapping segmentsof windowed intervals are used to reduce the variance of the power spectral estimate). The power spectrum was calculated as the squared magnitude of the fast Fourier transform for each segment. Segmentswere overlapped by 50% and averaged for each hour. Amplitude measurementsof the power spectrum were multiplied by a factor of 4, and area measurementsby 2.66 to correct for the Hanning window.20 In this study, we determined the spectral power over 3 frequency regions of interest: 0.05 to 0.15, 0.2 to 0.5 and over the entire 0 to 0.5 Hz band. We also determined hourly mean and standard deviations of the RR interval from the sameRR files as a more general nonspectral indicator of HR variability. Statianalysis: Data are expressedas mean f standard deviation. Variables were comparedstatistically using a linear regression analysis. Power spectral measurementsunderwent common log transformation before application of linear regression.
TABLE
I Group Results
Power 0.05-O. 15 (ms2/Hz) Power 0.20-0.50 (ms2/Hz) Power O-0.50 (ms2) RR interval (ms) RR SD (ms) LVEF TDCO (Iltersimin) LVFP (mm Hg) MSNA (bursts/mid Norepi (picogram/ml)
n
Mean + SD
23
44.7 2 62.7
1.2
23
7.8 t a.2
0.45
23 23 23 23 23 14 16 14
la.6 + 736 r 60.0? 0.21 r 4.3 * 242 62.3 + 745 2
15.5 124 23 0.07 1.7 12 23.1 390
Minimum
0.6 523 14 0.08 1.6 6 la.8 128
LVEF = left ventricular ejectlon fraction; LVFP = left ventricular MSNA = muscle sympathetic nerve activity: Norep] = norepinephrine spectral power; TDCO = thermal dilution cardlacoutput.
Maximum 290
28 62.7 944 104 0.34 9.3 46 99.4 1,530 filling pressure; level; Power =
pressedas average area) was 18.6 f 15.5 ms2 (range 0.6 to 62.7). Rehtions among measumd vahabkm The overall relation betweenthe clinical, hemodynamic and neurohormonal variables, and HR variability is illustrated by the 2 patient examplesshown in Figure 2. These 2 patients resemble each other clinically in regard to age, functional class and left ventricular ejection fraction. The patient on the left had a lower filling pressureand higher cardiac output than the one on the right. However, they differ the greatest relative to the measurements of muscle sympathetic nerve activity and plasma norepinephrine. The patient on the left (with the least evidencefor sympathoexcitation) also had greater power apparent over the entire power spectrum. In contrast, the patient on the right had markedly elevatedlevels of both norepinephrine and sympathetic nerve traffic, no visible peak in the high- and midfrequency ranges, and decreasedpower in the lowest frequencies. The relation between the power spectral measurements and the other measuredvariables are tabulated in Table II and shown for selectedcomparisonsin Figure 3. As is apparent, RR interval variability was not significantly related to age, functional class, left ventricular filling pressure or ejection fraction. In contrast, the RESULTS amount of RR standard deviation and both high- and Group data: Group data for the study cohort is tabu- midfrequency spectral power had a positive and signifilated in Table I. Substantial evidenceis apparent sup cant relation to cardiac output. Both RR standard deviporting the presenceand severity of CHF in these pa- ation and the 3 spectral power measurements(especialtients. As mentioned previously, the left ventricular ly the midfrequencies) generally displayed a negative ejection fraction is markedly reduced for the entire and significant relation to 1 or both indicators of symgroup. Left ventricular filling pressure (indicated by pathcexcitation, plasma norepinephrine and muscle pulmonary capillary wedge pressure) was elevated in sympathetic nerve activity. As one may expect, spectral the vast majority of patients, and cardiac output ranged power, RR interval and its standard deviation were sigfrom 1.6 to 9.3 liters/mm (mean 4.3). Muscle sympa- nificantly related, as were muscle sympathetic nerve acthetic nerve activity was clearly elevatedcomparedwith tivity and plasma norepinephrine (r = 0.79, p
