Power spectral analysis of heart rate and arterial ... · RAFFAELLO FURLAN, PAOLO PIZZINELLI,...

Dell'Orto and E PiccalugaM Pagani, F Lombardi, S Guzzetti, O Rimoldi, R Furlan, P Pizzinelli, G Sandrone, G Malfatto, S

sympatho-vagal interaction in man and conscious dog.Power spectral analysis of heart rate and arterial pressure variabilities as a marker of

Print ISSN: 0009-7330. Online ISSN: 1524-4571 Copyright © 1986 American Heart Association, Inc. All rights reserved.is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231Circulation Research

doi: 10.1161/01.RES.59.2.1781986;59:178-193Circ Res.

http://circres.ahajournals.org/content/59/2/178World Wide Web at:

The online version of this article, along with updated information and services, is located on the

http://circres.ahajournals.org//subscriptions/

is online at: Circulation Research Information about subscribing to Subscriptions:

http://www.lww.com/reprints Information about reprints can be found online at: Reprints:

document. Permissions and Rights Question and Answer about this process is available in the

located, click Request Permissions in the middle column of the Web page under Services. Further informationEditorial Office. Once the online version of the published article for which permission is being requested is

can be obtained via RightsLink, a service of the Copyright Clearance Center, not theCirculation Research Requests for permissions to reproduce figures, tables, or portions of articles originally published inPermissions:

at UniversitÃ degli Studi di Milano on October 1, 2012http://circres.ahajournals.org/Downloaded from

http://circres.ahajournals.org/content/59/2/178http://www.ahajournals.org/site/rights/http://www.lww.com/reprintshttp://circres.ahajournals.org//subscriptions/http://circres.ahajournals.org/

178

Power Spectral Analysis of Heart Rate andArterial Pressure Variabilities as a Marker of

Sympatho-Vagal Interaction in Man andConscious Dog

MASSIMO PAGANI, FEDERICO LOMBARDI, STEFANO GUZZETTI, ORNELLA RIMOLDI,

RAFFAELLO FURLAN, PAOLO PIZZINELLI, GIULIA SANDRONE, GABRIELLA MALFATTO,

SlMONETTA DELL'ORTO, EMANUELA PlCCALUGA, MAURIZIO TURIEL, GlUSEPPE BASELLI,

SERGIO CERUTTI, AND ALBERTO MALLIANI

In 57 normal subjects (age 20-60 years), we analyzed the spontaneous beat-to-beat oscillation in R-Rinterval during control recumbent position, 90° upright tilt, controlled respiration (n = 16) and acute(n = 10) and chronic (n = 12) /3-adrenergic receptor blockade. Automatic computer analysis pro-vided the autoregressive power spectral density, as well as the number and relative power of theindividual components. The power spectral density of R-R interval variability contained two majorcomponents in power, a high frequency at -0.25 Hz and a low frequency at ~0.1 Hz, with anormalized low frequency: high frequency ratio of 3.6 ± 0.7. With tilt, the low-frequency componentbecame largely predominant (90 ± 1%) with a low frequency: high frequency ratio of 21 ± 4. Acute/3-adrenergic receptor blockade (0.2 mg/kg IV propranolol) increased variance at rest and markedlyblunted the increase in low frequency and low frequency: high frequency ratio induced by tilt.Chronic /3-adrenergic receptor blockade (0.6 mg/kg p.o. propranolol, t.i.d.), in addition, reduced lowfrequency and increased high frequency at rest, while limiting the low frequency: high frequency ratioincrease produced by tilt. Controlled respiration produced at rest a marked increase in the high-frequency component, with a reduction of the low-frequency component and of the low frequency:high frequency ratio (0.7 ± 0.1); during tilt, the increase in the low frequency: high frequency ratio(8.3 ± 1.6) was significantly smaller. In seven additional subjects in whom direct high-fidelity arterialpressure was recorded, simultaneous R-R interval and arterial pressure variabilities were examinedat rest and during tilt. Also, the power spectral density of arterial pressure variability contained twomajor components, with a relative low frequency: high frequency ratio at rest of 2.8 ± 0.7, whichbecame 17 ± 5 with tilt. These power spectral density components were numerically similar to thoseobserved in R-R variability. Thus, invasive and noninvasive studies provided similar results. Moredirect information on the role of cardiac sympathetic nerves on R-R and arterial pressure variabilitieswas derived from a group of experiments in conscious dogs before and after bilateral stellectomy.Under control conditions, high frequency was predominant and low frequency was very small orabsent, owing to a predominant vagal tone. During a 9% decrease in arterial pressure obtained withIV nitroglycerin, there was a marked increase in low frequency, as a result of reflex sympatheticactivation. Bilateral stellectomy prevented this low-frequency increase in R-R but not in arterialpressure autospectra, indicating that sympathetic nerves to the heart are instrumental in the genesis oflow-frequency oscillations in R-R interval. (Circulation Research 1986;59:178-193)

THE whole history of brain electrophysiologyhas proved that rhythms can often be markersof normal functional states such as wakeful-ness or sleep, or of abnormal states such as epilepsy. Inthe present context, the general hypothesis is thatrhythmical beat-to-beat oscillations of one cardiovas-cular controlled variable might provide some criteria tointerpret the complex interplay among the neural regu-latory outflows. On the other hand, the existence, un-der stable conditions, of rhythmic fluctuations in car-

From the Istituto Ricerche Cardiovascolari, CNR; Patologia Me-dica. Centre "Fidia," Ospedale "L. Sacco," University Milano;CentraTeoria Sistemi. CNR, Dipartimento Elettronica, PolitecnicoMilano, Italy.

Address for reprints: Dr. Alberto Malliani, Istituto Ricerche Car-diovascolari, Via Bonfadini 214, 20138 Milano, Italy.

Received August 8. 1985; accepted April 7, 1986.

diovascular variables such as heart rate or arterialpressure has been recognized for a long time.1"7 Inrecent years, the possibility of quantifying these oscil-lations by using computer techniques, particularly onthe variability of electrocardiographic R-R interval,has aroused a growing interest.8"13

Indeed, as instantaneous heart rate depends on theinteraction between sympathetic and parasympatheticefferent activities and pacemaker properties, it hasbeen suggested that this analysis could lead to a nonin-vasive assessment of the "tonic" autonomic regulationof heart frequency.12 The nonrandom components ofR-R variability usually are assessed with spectral tech-niques based on the fast Fourier transform, in spite ofthe consideration that the heart rate variability signal isnot strictly periodical, as requested by the determinis-tic nature of the algorithm.16


http://circres.ahajournals.org/

Pagani et al Heart Rate and Arterial Pressure Variabilities 179

In conscious dogs and in man, a high-frequencycomponent (~0.25 Hz) has been consistently found inthe power spectrum and has been interpreted as a quan-titative assessment of respiratory arrhythmia." One ortwo low-frequency components have also been de-scribed, respectively, about 0.1 Hz and 0.03 Hz. Theformer of these, the so-called 10-second periodrhythm, has been considered particularly interesting asit has the same frequency of the better known Mayer5

waves.17

As to the neural mechanisms underlying these fluc-tuations, vagal efferent activity has been interpreted asbeing primarily responsible for the high-frequencycomponent of heart rate variability on the basis ofexperiments on vagotomized decerebrated cats,10 con-scious dogs,12 and man with muscarinic receptorblockade."

Both vagal and sympathetic outflows were consid-ered to determine the low-frequency components.12"

The aim of this study on normal human subjects andconscious dogs was to assess the relative role of vagaland sympathetic activities in determining the variabil-ity in heart rate and arterial pressure at rest and duringinduced changes of autonomic regulation. A totallyautomatic computation of the autospectra and of theirindividual components was implemented, using anautoregressive modeling algorithm, in order to takeadvantage of its inherent better definition, compared tothe more current spectral techniques.1819

Subjects and Methods

Fifty-seven subjects without any clinically evidentdisease (20-60 years old) were used for this study.They were healthy volunteers, randomly selected fromhospital staff, medical students, and their relatives.Seven additional subjects consulted our hypertensionclinic and were eventually found to be normotensive inthe hospital environment. These subjects underwent24-hour continuous monitoring of electrocardiogram(ECG) and high-fidelity intraarterial systemic pressurerecording20 as part of their clinical workup for suspect-ed hypertension.

All subjects included in this study had no signs oforgan or systemic disease. Subjects smoking morethan five cigarettes a day were excluded from thestudy. No one was taking any medication, and eachsubject was instructed to avoid beverages containingalcohol or caffeine after 10 pm of the day preceding thestudy and to come to the laboratory after a light break-fast. Studies were performed between 10:30 AM and12:30 PM. All subjects were carefully instructed aboutthe study, and all gave their informed consent.

TiltingSubjects were placed on an electrically driven tilt

table (provided with a mattress and a footrest) andconnected to a conventional electrocardiographic am-plifier (lead II, Cardioline) and FM tape recorder (Ra-cal). After 15-30 minutes, allowed for stabilization,the ECG was continuously recorded for 30 minutes atrest, followed by 20 minutes after the subjects had

been passively moved to an upright 90° position byelectrically rotating the table. During tilting, the sub-jects rested on their feet, although they had been loose-ly strapped to the table to avoid accidental falls.

The arterial blood pressure was recorded at the be-ginning and at the end of "rest" and "tilt" periods usinga conventional sphygmomanometer.

Reproducibility of the results was assessed by re-peating the study in 10 subjects after a period of 1 weekto 12 months. Additionally, in 4 subjects, a third studywas performed after an additional 3-4 months.

The effects of /3-adrenergic receptor blockade wasassessed in two groups of experiments. In 10 subjects,after the baseline studies, the study was repeated on alater day after IV bolus of 0.2 mg/kg propranolol (In-deral, ICI) in order to block acutely /3-adrenergic re-ceptor-mediated effects (Table 1).

In 12 subjects, the study was repeated after 6 days of0.6 mg/kg p.o. propranolol (Inderal, ICI) t.i.d. in or-der to block chronically /3-adrenergic receptor mediat-ed effects (Table 1).

Controlled RespirationThe effects of controlled respiration were tested in

16 subjects both at rest and during tilt. These patientswere studied first while breathing spontaneously, andon a separate day while breathing at 20 acts/min, fol-lowing a metronome. Respiratory movements weremonitored with an impedance respirometer (Cardio-line) and recorded on FM tape. Seven of these patientsbreathed through a mouthpiece connected to a spiro-meter (Jaeger) in order to quantify their tidal volume.All of these patients had been well acquainted with thelaboratory through 1-3 training sessions, during whichthey learned to breathe quietly and to adjust their tidalvolume to the increased respiratory rate. During spon-taneous and controlled breathing, transcutaneous Pco2and Po2 (Sensor Medics) were monitored in 5 patients.

Systemic Arterial Pressure (Invasive Studies)In 7 additional subjects, a microminiature Millar tip

pressure transducer (03 French) was inserted percuta-neously into the radial artery of the nondominant armwith a Seldinger technique.

Calibrations of the transducer, both in absolute val-ues of millimeters of mercury and millivolts, wereperformed just before the insertion of the catheter andat the end of the recording procedure. After suitableamplification, arterial blood pressure and ECG werecontinuously recorded at rest and during tilt on FMtape. These studies were the last part of a 24-hourprotocol for suspected hypertension, based on a con-tinuous ambulatory recording of direct high-fidelityarterial pressure and ECG.20

Animal ExperimentsTwelve mongrel dogs (25-35 kg body weight) were

used for this part of the study. They were lightly anes-thetized with thiopental sodium (10 mg/kg, IV) (Far-motal, Farmitalia) and, after the skin had been infiltrat-ed with 2% xylocaine (Byk Gulden), a cut-down was



180 Circulation Research Vol 59, No 2, August 1986

TABLE 1. Effects of Age

A. Rest

20-30 yr (n = 30)

30-45 yr (n = 10)

45-60 yr (n = 17)

Acute /3-adrenergicblockade, 20-30 yr(n = 10)

Chronic /3-adrenergicblockade, 20-30 yr(n = 12)

B. Tilt

20-30 yr

30-45 yr

45-60 yr

Acute /3-adrenergicblockade. 20-30 yr

Chronic /3-adrenergicblockade, 20-30 yr

and fi-Adrenergic BlockadeArterial

Systolic(mm Hg)

116±2

116±4

118±3103±3t

106±2t

111±1

109±4

113±3

100±4t97±2t*

pressure

Diastol-ic (mm

Hg)

73±2

79±4

78±2

70±3

66±2t

79±2

78±3

77±2

72±4

69±4t

on

R-Rinterval(msec)

834±34

931±39

926±21076±26t

1115±59t

687±14

719±44'

767±24

X

X

X929

±24t*951

±35t*

the Response

R-Rvariance(msec2)

4097±361'

2581±356'

1354±205

8255±1676t

7180±1712t

2642±217

2160±253'

1212±213

*

13009

±361*

4067±664t*

of R-R Interval VariabilityLow-frequency

component

Normalizedpower

58.2±3.3

62.2±6.3

49.9±4.9

50.9±5.0

39.4±5.7t

89.7±1.4

83.7±4.6'

75.7±4.6

t

X

X74.1

±7.0t*

66.7±5.2tt

Frequency(HzEq)

0.11±0.01

0.10±0.01

0.09±0.01

0.12±0.01

0.12±0.01

0.09±0.01

0.09±0.01

0.08±0.01

0.09±0.01

0.11±0.01

to TillHigh-frequency

component

Normalized Frequencypower (Hz Eq)

24.0±2.0

26.3±4.3

32.0±46

32.7±4.3

53.4±5 4t

7.5±0.9

11.4± 2 . 8 '

16.2±3.4*

13.0±2.5*

0.24±0.02

0.27±0.02

0.27±0.01

0.26±0.02

0.26±0.02

0.29X ±0.03

0.29X ±0.06

0.23X ±0.02

0.22±0.02

24.0 0.22±4.4t* ±0.02

LF:HF ratio

3.62±0.70

3.69±1.23

2.58±0.61

1.74±0.31

1.07±O.35t

20.79±3.68*

17.30±8.19*

12.61±3.21*

7.48±I.77t*

3.78±0.93t*

•Significantly different contrast (p < 0.05).tValue during /3-adrenergic blockade significantly different (p < 0.05) from value obtained in unblocked conditions in the young age group.*Value during tilt significantly different from value at rest (p < 0.05).

performed in the region of the femoral artery. A cath-eter was introduced into it, advanced to the aorta, andsecured with a suture. The wound was closed, and thecatheter was exteriorized at the base of the neck.

After baseline studies, 5 dogs underwent bilateralstellectomy. By sterile techniques, they were anesthe-tized with thiopental sodium (30 mg/kg, IV) followedby a continuous infusion of fentanyl citrate (6.2 /xg/kg,IV) and droperidol (0.2 mg/kg, IV) (Leptofen, CarloErba). The animals were paralyzed with small (0.1mg/kg, IV) intermittent doses of succinylcholine(Wellcome) and artificially ventilated with a positive-pressure pump (Harvard). A thoracotomy was per-formed in the fourth left intercostal space, the leftstellate ganglion and its branches were excised, and thethoracotomy was repaired. Seven to 15 days later,under similar anesthesia and through a right thoracot-omy in the fourth intercostal space, the right stellateganglion was excised as well.

Aortic pressure was measured with the implantedcatheter using a pressure transducer (Statham Instru-ments). Aortic mean pressure was obtained with an RCfilter with a 2-second time constant. The electrocardio-gram (lead IT) was obtained with subcutaneous silverelectrodes and an AC amplifier. Heart rate was moni-tored continuously with a cardiotachometer triggeredby the R-wave.

Respiratory movements were monitored with apneumatic belt connected to a pressure transducer(Statham Instruments). Data were recorded on a multi-channel FM tape recorder (Racal Store 7) and playedback on a direct-writing recorder (Brush Gould).

Experiments were performed after a postoperativeperiod long enough to allow complete recovery of thedogs from the operation, as judged by their normalbehavior, body temperature, and hematocrit. Five to 7days usually were sufficient after the implantation ofthe pressure catheter, while 10-15 days were givenafter the thoracotomy. While the trained dogs werelying quietly on the recording table, aortic blood pres-sure, ECG (lead II), and respiratory movements wererecorded continuously for 20-30 minutes of control,followed by an infusion of nitroglycerin (32 /u.g/kg permin, IV) for 15-20 minutes, in order to excite sympa-thetic activity.21

Data AnalysisOff-line analysis was performed on DEC MNC

11/23 and PDP 11/24 minicomputers. ECG, respira-tion, and pressure data were played back from the FMtape and digitized at 300 samples/sec per channel. Theprinciples of the software for data acquisition and anal-ysis have been described previously.l32022

Briefly, stationary sections of data both at rest and




RR INTERVALS TACHOQRAM HI STOGRAM

m


length. For instance, if the average R-R length were1000 msec, this would correspond to 0.25 Hz Eq. Inthe figures, both units are indicated, whereas, in thetext, only hertz equivalents are used.

De Boer et al23 have compared different methods ofspectral analysis of the heart rate variability signal andfound that the analysis of the interval series that repre-sents heart period as a function of the beat number notonly provides comparable results to methods that rep-resent heart period as a function of time, but is evenbetter suited to the study of the relationship betweenheart period and arterial pressure on a beat-by-beatbasis.

After synchronous acquisition and appropriate cali-bration, a similar procedure is used to compute thespectrum of the arterial pressure data for both systolicand diastolic values.20

Spectral analysis of the impedance respiratory sig-nal was also performed to assess the effects of metro-nome breathing on respiratory waveform.

StatisticsData are presented as means ± SE. Student's / test

was used to determine the significance of the differ-ences between rest and tilt.

Analysis of variance with Scheff6 test was used toassess the effects of age and of /3-adrenergic receptorblockade, as well as the effects of controlled breathing.

Differences were considered significant a tp

Pagani et at Heart Rate and Arterial Pressure Variabilities

REST TILT1.2 I

cc

0.5 I 10 512 0

20

183

0.50

beats512

0.52

FIGURE 3. R-R interval series,

i.e., tachogram at rest and duringpassive upright 9(f tilt. On the auto-spectra (bottom panels), two clearlyseparated low- and high-frequencycomponents are present at rest. Dur-ing tilt, the low-frequency compo-nent becomes preponderant.

cant reduction, as expected. However, there was nosignificant correlation between the increases in LFcomponent and in heart rate induced by tilting(r = 0.l9).

Autospectra of R-R variability in individual sub-jects remained remarkably constant when repeatedover time. As shown in Table 2, there was no differ-ence (p>0.05), at rest and during tilt, in the resultsobtained in two or three different recording sessionswith an interval up to I year.

Age Dependency of the Response to TiltDuring tilt in the old age group, mean R-R was

significantly longer, and variance smaller, than both inthe mid and young age groups (Table I).

The effects of tilt on the autospectra of R-R intervalvariability when examined in the three different agegroups demonstrated only minor differences, as LFcomponent was smaller and HF component slightlygreater in the old age group (Table 1). This difference,however, was no longer apparent in the LF:HF ratio,which was not significantly different in the three agegroups both at rest and during tilt (Table 1).

Effects of Controlled RespirationThe effects of controlled respiration were examined

in 16 young subjects that were first studied whilebreathing spontaneously (Figure 4, top panels) and,subsequently, while breathing following a metronome(Figure 4, bottom panels). Respiratory frequency, in

TABLE 2. Reproducibiliry of the Effects of Tilt on R-R Variability

Rest

First study (n = 10)

Second study (n = 10)

Third study (n - 4)Tilt

First study (n = 10)

Second study (n = 10)

Third study (n = 4)

R-Rinterval(msec)

885 ±43

858 ±32

914±34

672 ± 24*

676 ±33*

700 ±25*

R-Rvariance(msec2)

3264 ±464

3508 ±766

4201 ±781

2612+528

1955 ±303

3381 ±250

Low-frequency

Normalizedpower

62.0±5.0

61.4 + 4.6

56.1+2.6

89.5+1.4*

89.2 + 2.2*

87.0 + 2.0*

component

Frequency(HzEq)

0.11 ±0.010.10±0.01

0.12±0.01

0.09±0.01

0.10±0.010.l0±0.01

High-frequency

Normalizedpower

28.5±3.3

29.0 + 3.7

31.3±3.8

6.0±0.5*

6.9±0.07*

7.7± 1.3*

component

Frequency(HzEq)

0.26 ±0.02

0.25 ±0.02

0.31 ±0.03

0.30±0.04

0.26 + 0.03

0.27 ±0.02

Average delay in time between the first and second study: 138 days (range 7-365). Average delay in time between the second and thirdstudy: 127 days (range 90-150).

*Value during tilt significantly different from value at rest (p < 0.05).



184 Circulation Research Vol 59. No 2. August 1986

REST TILT

IQa.

so(0a.

the latter condition, was set at20/min, i.e., 0.33 Hz, inorder to separate better the low-frequency from thehigh-frequency respiratory component.

This maneuver modified the respiratory waveform,as demonstrated by spectral analysis: during spontane-ous respiration, nonstationarities in the signal are re-flected by a diffuse profile of the spectrum, whereasduring metronome breathing, the near sinusoidal shapeof the respiratory waveform is reflected by a singlenarrow component in the spectrum. In either condi-tion, the HF component of the spectrum of R-R vari-ability coincided with the main respiratory frequency.The LF component never coincided with a major com-ponent of the spectrum of the respiratory waveform.

FIGURE 4. Representative example of

autospectra of R-R interval variabilityduring spontaneous breathing (top pan-els) and during controlled respiration,at 20lmin (bottom panels). Note that,with controlled respiration at rest, therelative power of the high-frequencycomponent becomes predominant.

c/b

Hz sq0 8 3

During controlled respiration, at rest, there was asignificant reduction of the LF component with a con-comitant increase in the HF component (Figure 4, bot-tom; Table 3). The LF:HF ratio was significantly re-duced from 2.5 ± 0 . 3 to 0.7 ± 0 . 1 . As the subjectswere allowed to adjust their tidal volume, during spon-taneous respiration the tidal volume (n = 7) was552 ± 98 ml, and slightly less (533 ± 76 ml) duringcontrolled breathing at 20/min. During metronomebreathing, there was only a small but not significantchange in transcutaneous Pco2 and Po2 (respectively,42 ± 2 , 83 ± 4 mm Hg during spontaneous and38 ± 2, 79 ± 6 mm Hg during controlled breathing,p>0.05). During tilt, the HF component was de-

TABLJE 3. Effects of Metronome Breathing on R-R Variability1 (n = 16)

R-Rinterval(msec)

R-Rvariance(msec2)

Low-frequency

Normalizedpower

component

Frequency(HzEq)

High-frequency

Normalizedpower

component

Frequency(HzEq)

Rest

Free resp frequency Hz 0.26 ± 0.02 884±27 4095±436 56.2±3.9 0 . 1 1 ± 0 0 1*

Control resp frequency Hz 0.33I t

934±29 4220±541 28.9±4.3 0.11 ±0.01

Free resp frequency Hz 0.26 ± 0.03 678±21t 2647±308t 86.4±2.9t 0,l0±0.01

Control resp frequency Hz 0.33 705±22t 2119±428t 71,6±5.3t 0.10±0.0l

213+2.1 0.24±0.02* *

51.6 + 4.4 O.33±O.O1

7.8±l.4t 0.25±0.02• *

15.3±3.0t 0.31 ±0.02

•Significant difference (p < 0.05).tValue during tilt significantly different from value at rest (p < 0.05).




creased while the LF component was increased, asexpected; however, this LF increase was less pro-nounced than during spontaneous breathing (Figure 4).Consequently, the increase in LF:HF ratio induced bytilt was smaller during controlled respiration(8.3 ±1 .6) than during spontaneous breathing(16.0 ±3 .0) .

Furthermore, during tilt, tidal volume was slightlysmaller with controlled respiration than with spontane-ous breathing (574 ± 100 and 657 ± 48 ml, respective-ly), and no significant change (p>0.05) was observedin transcutaneous Pco2 or Po2 (respectively, 39 ± 4,87 ± 5 mm Hg with spontaneous and 43 ± 3, 82 ± 4mm Hg with metronome breathing).

j8-Adrenergic Receptor Blockade

The effects of /3-adrenergic receptor blockade wereassessed in two sets of experiments. Acute /3-adrener-gic receptor blockade was obtained in 10 young sub-jects by IV bolus of 0.2 mg/kg of propranolol thatreduced significantly heart rate and arterial blood pres-sure (Table 1). Under these conditions, R-R intervalvariance was significantly increased from control (Fig-ure 5; Table 1), but normalized autospectral compo-nents were not modified significantly (Table 1). Dur-

REST

ing tilt, there was a significantly lesser reduction inR-R interval and a smaller increase in the LF compo-nent and in the LF:HF ratio (Figure 5; Table 1).

Chronic /3-adrenergic receptor blockade was ob-tained in 12 young subjects with 0.6 mg/kg p.o. pro-pranolol t.i.d. for 6 days, which reduced significantlyresting heart rate and arterial blood pressure. The ef-fects on resting R—R interval variability, like acuteblockade, were characterized by augmented variance(Figure 5; Table 1). However, marked changes wereobserved in the resting autospectra: LF was smaller,HF greater, and hence, LF:HF significantly reduced,compared with controls (Table 1; Figure 5). Duringtilt, the LF component was reduced and HF componentincreased, compared with control. Thus, during chron-ic /3-adrenergic receptor blockade, LF:HF ratio in-creased to only 3.78 ± 0.93, which was smaller thanthat observed in the young age group without /3-block-ade during tilt (20.79 ± 3.68) but was similar to theirresting value (3.62 ± 0.70).

Simultaneous R-R and Blood Pressure VariabilitiesIn a group of 7 normotensive subjects (20-60 years

old) in whom systemic arterial pressure was recordedwith direct high-fidelity techniques (see "Materials

chronicB'blockade

0.53 0.41 0.42Hzeq

TILT

050

0.75

0.50

0.45

0.50

I Hzeq0.48

FIGURE 5. Autospectra of R-R interval variability of a young subject at rest (top panels) and during tilt (bottom panels) undercontrol conditions (left panels) and during acute (0.2 mg/kg IVpropranolol. middle panels), and chronic (0.6 mg/kg p.o. propranololt.i.d.) fi-adrenergic receptor blockade. Note the progressive reduction in the relative power of the low-frequency component duringtilt.




and Methods") simultaneously with the ECG, the anal-ysis of the variability of systolic and diastolic beat-by-beat values was also performed.

Systolic arterial pressure variability was character-ized by small oscillations in the time series of beat-by-beat values (Figure 6), resulting in a tracing compara-ble to the interval tachogram.

Moreover, the autospectra of systolic arterial pres-sure variability demonstrated two major components,respectively, a LF of 41 ± 8% and HF of 19 ± 5%,with a relative ratio of 2.8 ± 0.7 at rest. During tilt, theLF component became largely predominant (73 ±8%), with an increase in the LF:HF ratio of 17 ± 5.Spectral analysis of diastolic blood pressure gave simi-lar results (Table 4). It should be appreciated that thesimultaneous autospectra of R-R variability (Figure 6;Table 4) provided results resembling those obtainedfrom the analysis of arterial pressure variability, withtwo components recognized at the same frequencies,with similar relative ratios.

As shown in Figure 7, cross-spectral analysis ofR-R and blood pressure variability confirmed thisevaluation by showing coherence only between thesame LF and HF components both at rest and duringtilt. The phase difference between R-R and arterialpressure variability was approximately 0° at HF. AtLF, the phase presented different patterns in rest and intilt conditions: In the former case, the phase had alinear relation with frequency, whereas, in the lattercase, it oscillated around a fixed value. In both cases,the phase corresponding to the central frequency of theLF peak was about 60°, i.e., Vt of the entire 360° cycle.As LF corresponds to a period of about 10 beats, thecalculated phase difference amounted to XA of that peri-od, i.e., to a delay of 1.7 beats, with pressure leading.

R-R and Arterial Pressure Variability inConscious Dogs

To assess more directly the effects of sympatheticinnervation on heart rate variability, we performedexperiments on a group of conscious dogs before andafter bilateral stellectomy.

As shown in Figure 8 and Table 5, under controlconditions R-R variability was almost solely repre-

sented by a HF respiratory-linked component, whereasa very small (8 ± 2%) LF component was present inonly 50% of the animals. Similarly (Table 5), the auto-spectra of both systolic and diastolic blood pressurevariabilities demonstrated a major HF component.During moderate hypotension (— 9 ± 2% from 85 ± 3mm Hg mean arterial pressure) obtained by a continu-ous infusion of IV nitroglycerin (32 Aig/kg per minute),heart rate increased 47 ± 9% from 82 ± 5 beats/min asa consequence of sympathetic activation.

Under those conditions, not only was there a signifi-cant reduction of total power of R-R variability, butalso the HF component of the autospectrum was re-duced to 42 ± 3%, whereas, in all animals, a LF com-ponent of similar power (36 ± 6%) became evident.Similar changes were observed in arterial pressureautospectra (Table 5). Bilateral stellectomy did notsignificantly modify arterial pressure, heart rate, orR-R or pressure variability, as expressed both by vari-ance or by their autospectra (Figure 8; Table 5). The IVinfusion of nitroglycerin, repeated at the same dose,reduced arterial pressure and increased heart rate to asimilar extent (Table 5). The response of R-R variabil-ity, however, was modified. Although total power wasreduced to a value similar to that observed in the intactanimals, only a HF component could be observed inthe autospectrum. Conversely, the increase in LF com-ponents of pressure autospectra was essentially pre-served (Table 5).

DiscussionThis study in man examines the beat-to-beat oscilla-

tions which characterize heart rate and arterial pressureunder various steady state conditions, in the hypothesisthat the quantitative information provided by the spec-tral analysis of these oscillations reflects the interac-tion between sympathetic and parasympathetic regula-tory activities.

Various approaches have been proposed to evaluatethe contributions of parasympathetic and sympatheticdischarges23 to the heart rate variability. Parasympa-thetic activity has been clinically inferred from peak-to-peak variations of the heart period either in theclinical laboratory environment26 or with Holter moni-

TABLE 4. Effects of Tilt on Simultaneous R-R Interval and Arterial Pressure Variabilities (n = 7)

Rest

R-R

SAP

DAPTilt

R-R

SAP

DAP

Mean value

748 ± 20 msec

125 ±5 mm Hg

70 ±5 mm Hg

633 ± 29* msec

129 ±6 mm Hg

74 ±4 mm Hg

Variance

1403 ±571 msec2

20±3 mm Hg2

15 ±3 mm Hg2

1483 ±643 msec2

47 ±10* mm Hg2

29 ±7* mm Hg2

Low-frequency

Normalizedpower

41.2 + 8.7

46.7±7.2

48 5± 11.7

73.4±8.6*

78.7±4.1*

73.3±2 5*

component

Frequency(HzEq)

0.12±0.01

0.11 ±0.02

0.10 + 001

0.09 + 0.01

0.09±0.01

0.08 + 0.01

High-frequency

Normalizedpower

I9.4±5.3

27.5±9.O

31 1±7.7

7 4± 1.9*

9.2 ± 2.1

9.9±l,7*

component

Frequency(HzEq)

O.33±O.O3

O.32±O.O3

0.35±0.04

O.33±O.O2

O.32 ±O.O3

0.35 ±0.03

SAP = systolic arterial pressure; DAP = diastolic arterial pressure.*Value during tilt significantly different from value at rest (p < 0.05).



Pagani el al Heart Rate and Arterial Pressure Variabilities

REST T I L T

187

515 beats

FIGURE 6. Simultaneous analysis of R—R in-

terval and systolic arterial pressure variabili-ties in a young subject. Note the reduction inaverage R-R interval and the slight increase inaverage systolic arterial pressure on passingfrom rest to tilt. The two components of theautospectrum of systolic arterial pressure vari-ability are similar to those of R-R interval vari-ability at rest and during tilt.

645C c/b

8 Hz/eq

toring.27 A different approach has been used by oth-ers101215 who utilized spectral techniques to assess thefrequency components of the heart rate variability sig-nal. A possible major advantage of spectral analysis isthe observation that changes in the sympathetic activ-ity to the heart can also be recognized, and hence,some index of the instantaneous balance between sym-pathetic and vagal activity can be obtained.13 It shouldbe noted that, since heart rate variability signal is apseudorandom phenomenon, previous studies usingthe fast Fourier transform to compute the power spec-trum have some technical limitations (see the Appen-dix). They include the deterministic nature of the algo-rithms used, which, in principle, are applicable only toperiodical phenomena,16 the need of windowing thedata, and the difficulty in defining with certainty therelative power of the various spectral components.

In the present study, some of these limitations wereavoided because the autoregressive algorithms, whichare applicable to nonperiodical phenomena,28 comput-ed automatically the number and relative power of the

various spectral components. Moreover, there was noneed for windowing or filtering the data.l3 22

Spectral Analysis of Heart Rate Variability andParasympathetic Activity

In our analysis, we observed at rest two consistentmajor spectral components, a LF at ~0.1 Hz and anHF at ~0.25 Hz. The LF component seems to corre-spond to the Mayer waves,517 while the HF componentis synchronous with the respiration and has been con-sidered as a quantitative evaluation of respiratory ar-rhythmia.15 Since the HF component disappears afteratropine, it could represent a clinically useful index ofvagal activity.15

In our study, during spontaneous quiet breathing atrest, a relatively small area was found to be associatedwith the HF component. However, Pomeranz et al'5

found a predominant HF component in 8 subjects whobreathed following a metronome at 15/min. A similar-ly enhanced HF component was found in this study in16 subjects who breathed following a metronome at




REST T I L T

o

3.14

m

CDQ .

3

FIGURE 7. Cross-spectral analysis(coherence, thick lines; phase, thinlines) between R-R interval and sys-tolic arterial pressure variabilitiescomputed from the data of Table 4.Note that, both at rest and duringtilt, only two spectral componentsshow a coherence greater than 0.5 inboth conditions. Phase is about 0° atthe respiratory component and nega-tive, i.e., pressure leads R-R inter-val, at the O.I-Hz component.

-3.14

20/min. The same subjects showed a predominant LFcomponent when breathing spontaneously. Changes inthe amount of respiratory arrhythmia and, hence, ofthe HF component could reflect changes in volume, aswell as in frequency of respiration.29-30 Important mod-ulatory mechanisms could also be initiated by changesin arterial Po2 and Pco2, and thereby changes in neuralefferent activities,31 as well as by variations in arterialpressure and consequent alterations in baroreceptorinput.32

In this study, we could not define the exact mecha-

nisms responsible for the observed increase in HFcomponent with controlled breathing, as tidal volume,transcutaneous Po2 and Pco2, and arterial pressure didnot change significantly.

However, an important role is likely to be played bythe marked change in respiratory waveform,33 actingeither through a more efficient stimulation of lung re-ceptors34 or through fluctuations induced in arterialpressure.3235 In either case, the increased HF compo-nent of R-R variability would indicate that during met-ronome breathing there is enhanced vagal tone.

TABLE 5. Effects of Moderate Hypotension on Simultaneous R-R Interval and Arterial Pressure Variabilities in Conscious Dogs

Part A

Intact (n = 8)

R-R

SAP

DAP

Nitroglycerin(32 pig/kg

R-R

SAP

DAPPart B

Stellectomy (n

R-R

SAP

DAP

Nitroglycerin(32 Mg/kg

R-R

SAP

DAP

Mean value

control

755 ±62 msec

119±18 mm Hg

71 ±3 mm Hg

per min)

521+37* msec

104±5* mmHg

66 ±2 mm Hg

= 5) control

790 ±41 msec

137±10 mm Hg

76 ±3 mm Hg

per min)

617±19* msec

110±7* mm Hg

73 ±4 mm Hg

Variance

58237 ±17754 msec2

217±112 mmHg2

161 ±71 mm Hg2

7513 ±3941* msec2

45 ± 19 mm Hg2

5O ± 21 mmHg2

42786+14370 msec2

67 ±28 mm Hg2

78±22 mmHg2

3491 + 1696* msec2

16 + 4 mm Hg2

24 ± 13 mm Hg

Low-frequency component

Normalizedpower

4 ± 2

19±4

15 ±5

36 ±6*

55 ±4*

59 ±7*

3 ± 2

8 ± 3

8 ± 4

0±0+

38 ±8*

44±6*

Frequency(HzEq)

0.14±0.01

0.12±0.02

0.10±0.03

0.12±0.01

0.09 + 0.01

0.10±0.01

0.10±0.01

0.11 ±0.01

0.09±0.01

Unmeasurable

0.08 ±0.01

0.09±0.01

High-frequency component

Normalizedpower

63 ±660±4

64±4

42 ±4*

27 ±5*

22 + 5*

61+5

72±4

63±7

76±8t46±10*

37 ±7*

1 Frequency(Hz Eq)

0.24 + 0.02

0.23 ±0.03

0.24±0.03

0.26 ±0.03

0.27±0.06

0.25 ±0.03

0.23±0.01

0.25 + 0.01

0.23 + 0.02

0.24 ±0.02

0.26 ±0.02

0.25 ±0.02

SAP = systolic arterial pressure; DAP = diastohc arterial pressure.•Value during nitroglycerin infusion significantly different from control (p < 0.05).tSignificant difference (p < 0.05) between value in intact and denervated animals.




CONTROL NTR

1500

QV)a.

1500

Q(Aa.

0.50 0

STELLECTOMY

500

0 3 6

0.50Hz «q

0.50

FIGURE 8. Autospectra of R-R inter-val variability of a conscious healthydog in control conditions (left panels)and during an IV infusion of nitrogb/cer-in to excite the sympathetic outflow. Thetop panels were obtained with cardiacnerves intact, the bottom panels afterfull recovery from bilateral stellectomy.Note the presence of a large low-fre-quency component during nitroghcerininfusion only in the neuralty intact situa-tion.

0 3 6c/b

J Hz eq0.50

Spectral Analysis of Heart Rate Variability andSympathetic Activity

The possibility that spectral analysis could providean index of sympathetic activity is less well estab-lished. In animal studies, sympathetic activity hasbeen considered either to play no role10 or else to beinstrumental in the genesis of the LF rhythm.12-36 Fur-thermore, electroneurographic recordings in dogs25

suggest that sympathetic activity might also modulaterespiratory arrhythmias. This apparent discrepancy ofresults is probably a consequence of the differences inthe preparations used, i.e., anesthetized and acutelydecerebrated cats10 as opposed to conscious dogs.12

In our study, we tested the hypothesis that spectralanalysis of heart rate variability could provide an as-sessment of sympathetic tone, by planning experi-ments in man in which sympathetic activity was eitherincreased functionally or blocked pharmacologically.Enhanced sympathetic drive to the heart, as obtainedby an orthostatic stimulus, was constantly associatedwith a marked increase in the LF and with a decrease inthe HF component of the autospectrum. We selected apassive change of posture in order to minimize theinfluence on heart rate exerted by the muscular effortof active standing.37

Acute /3-adrenergic receptor blockade with intrave-nous propranolol, besides reducing heart rate, had amarked effect on R-R variability. At rest varianceincreased, while the relative contribution of LF and HFcomponents remained relatively unchanged. During

tilt, there was a significant reduction of the increase ofthe LF component and of the LF:HF ratio. Similarblunting effects exerted by acute /3-adrenergic receptorblockade were observed during active standing byPomeranz et al ."

After chronic /3-adrenergic receptor blockade, therewas, at rest, not only an increase in R-R interval vari-ance similar to that observed with acute blockade, butalso a significant effect on spectral components. Thus,LF and LF: HF were significantly smaller, and HF wassignificantly greater than under control conditions.During tilt, the increase in LF component and in theLF:HF ratio was markedly reduced; indeed, the nu-merical value of the ratio was similar to that observedunder resting control conditions without /3-blockade.

It might be pertinent to recall that Wallin et a Pdemonstrated, with electroneurographic techniques inhypertensive subjects, that chronic /3-adrenergicblockade was effective in reducing sympathetic effer-ent activity, whereas acute blockade was ineffective.

As to the effects of metronome breathing on LF, thiscomponent was reduced at rest, and its increase duringtilt was blunted, possibly as a consequence of the in-hibitory effect on sympathetic rhythms produced bypulmonary afferent activity.34

As to the effects of aging, it was progressively asso-ciated with a reduced R-R variability as expressed byvariance, and during tilt with a smaller activation of LFand a smaller reduction of HF. However, the changesin LF: HF ratio during tilt were not significantly differ-ent among the various age groups. This approach




would suggest that aging is characterized by a newequilibrium between the two sections of the visceralnervous system rather than by alterations limited to thesympathetic outflow.3940

Simultaneous Variability of R-R Interval andArterial Pressure

An additional major observation of this study is thatthe autospectra of R-R interval and of both systolicand diastolic arterial pressure beat-to-beat variabilitieswere similar at rest and showed parallel changes withtilt. The presence in arterial pressure recordings of anHF respiratory component has been traditionally inter-preted as a mechanical consquence of respiration,which could act directly on intrathoracic vessels orindirectly through changes in stroke volume741 andheart period.3* Sympathetic modulation of arterialsmooth muscle is probably too slow to follow the 0.25-Hz respiratory frequency.42 Obviously, these beat-to-beat pressure changes could affect R-R intervalthrough complex reflex adjustments, among whichbaroreflexes could have a paramount importance.

The LF components correspond to the well-knownMayer3 waves, a phenomenon that, although describedin quite artificial experimental conditions, seems topertain to normal human subjects8 as well. Varioustheories, including myogenic oscillations, centralrhythms, feedback mechanisms, and "resonance" dis-turbances, have been advanced for their interpreta-tion.17

As to the effects of tilt, arterial pressure beat-by-beatvariability, as expressed by variance, increased sig-nificantly for both systolic and diastolic values. Fur-thermore, the LF oscillatory components increasedsignficantly and to an extent similar to that observed inthe autospectrum of R-R interval, whereas HF compo-nents were, in both autospectra, markedly reduced.

In this respect, it should be mentioned that an in-crease in arterial pressure variability has been de-scribed also in essential hypertension,45 and a greaterLF component has been documented in daytime asopposed to nighttime recordings in ambulatory pa-tients.46 Both conditions are possibly characterized bya higher sympathetic activity.

Cross-spectral analysis of systolic arterial pressureand R-R interval variabilities indicated that a highdegree of coherence existed between the fluctuationsof these two variables both in recumbency and duringtilt. In correspondence to the HF component, arterialpressure and R-R interval changes occurred in phase,whereas each LF pressure change preceded R-R inter-val oscillation by about two beats.

Although this cross-spectral analysis provides nodirect insight into the mechanisms linking heart periodand arterial pressure oscillations, with their possibleneural and non-neural components,1247 it supports theconclusion that similar information on oscillatoryrhythms can be obtained from both invasive and nonin-vasive studies, not only at rest, but also during aug-mented sympathetic activity.

In a previous study by De Boer et al,47 only a smallHF component was observed in the autospectrum of

diastolic arterial pressure measured in sitting subjectseither with direct or indirect plethysmographic clampmethod.48 Technical differences in arterial pressure re-cording and the intermediate level of gravitationalstimulation induced by sitting should account for thedifference in our data.

Conscious DogsMore direct information on the role of cardiac sym-

pathetic nerves was derived from a group of experi-ments in conscious dogs before and after bilateral stel-lectomy. In the conscious dog, the LF components are,under control conditions, either absent33 or variable'236

and small (Table 5). An excitation of sympathetic ac-tivity as induced by moderate hypotension obtainedwith phentolamine36 or, as in our experiments, withnitroglycerin, was associated with an increase in LFcomponents. This increase was no longer present afterbilateral stellectomy, which, on the other hand, did notmodify the LF and HF components characterizing thearterial pressure variability (Table 5). Hence, in con-scious dogs that had fully recovered from surgery, itwas possible for the first time to dissociate the LFcomponents present in the heart rate and arterial pres-sure autospectra by selectively interrupting the cardiacsympathetic loop.

Changes in LF: HF Ratio as Markers of Changes inSympatho-Vagal Balance

It has been a rather simple but traditional workinghypothesis that conditions which increase sympatheticactivity decrease vagal tone and vice versa.48 How-ever, in electrophysiological experiments, recordingsof the sympathetic and vagal discharges, directed tothe heart, indicated both a reciprocal50 and a nonreci-procal31 organization.

The actual balance between these two outflows islikely to have paramount importance not only undersuch physiological conditions as gravitational stimuli,exercise, and changes in central command,51 but alsoduring disturbances such as arrhythmias,52 myocardialischemia, arterial hypertension,53 or alterations in therenin-angiotensin system.12

It has recently been suggested that spectral analysisof R-R interval variability might reflect this balance.l5

In our study, the LF: HF ratio appeared to be a conven-ient index of such interaction. Indeed, the simulta-neous increase in vagal and decrease in sympatheticmechanisms produced by metronome breathing werereflected by a reduction of LF: HF ratio at rest and by ablunting of the increase induced by tilt. Furthermore,LF: HF ratio appeared to follow the reductions of sym-pathetic activity produced by acute and chronic /3-adrenergic receptor blockade both at rest and duringtilt. Clinical research into those states characterized bya disturbance in the neural regulatory mechanisms willprobably reveal in the near future whether this ap-proach will have true pathophysiological significance.

AppendixThe basic assumption underlying the proposed signal pro-

cessing methods is that heart rate and systolic and diastolic




blood pressure values fluctuate cycle by cycle, even in stableconditions, around a given mean value. Beat-to-beat heart rateand arterial blood pressure variability signals are obtained firstfrom the original biological signals as a discrete time series y(k)where k is the progressive number of detected events.

Variability signals are intrinsically pseudo-random and canbe considered as the realization of a stochastic process y(k)which, is the output of a linear time-invariant system driven bywhite noise w(k), which constitutes the random component ofthe model and is fully characterized by a mean value which iszero for simplicity of notation and a variance X2 [i.e.,w(k) = WN(O,X2)] according to the model

w(k) -> H(z) -> y(k).

where H(z) is the transfer function in the complex variable z,which is determined in a quantitative way and is the model of thesignal-generating mechanism. A simple structure of input/out-put relation is given by the autoregressive (AR) modelization

y(k)= I a(i)y(k - i) + w(k)i - i

where a(i) are the p unknown parameters of indentification.More complex models are described by Box and Jenkins.28

Obviously,

H(z) =p

- I

which relates the transfer function with the identification coeffi-cients. Hence, given N samples of variability signals, the prob-lem is to estimate H(z) or, more precisely, the vector 6 ofparameters

9 = ), a(2), a(3) a(p),

The algorithm applied for such an identification is the Batchleast squares method via Levinson-Durbin recursion.18

Two tests are used to check the validity of the assumedmodel. The first one is Anderson's test, which measures thewhiteness of the prediction error: The identification is not ac-cepted if the test is not satisfied within 5% confidence interval.After fulfilling Anderson's test, the order of the model is chosenas the one which minimizes Akaike's final prediction error(FPE) figure of merit.54 In this way, the model is completelydetermined by order p and vector 9.

Power spectral density (PSD) estimation P(f) is obtained fromthe following relation

values. The spectrum is calculated on 512 consecutive cardiacbeats; stationarity has been tested in two ways: (1) Calculationof mean values and variance of variability signals on consecu-tive records, by verifying that the values are inside a confidencelevel of 5%, (2) construction of the pole diagram of the auto-regressive model, together with the confidence interval in thepole estimation, by verifying that consecutive records are char-acterized by poles which are inside the fitted confidence interval(wide-sense stationarity). The spectra which are presented inthis paper satisfy both of the preceding conditions.

With the AR spectral estimation presented in this paper, it ispossible to improve the statistical consistency, even of very lowfrequency components (less than 0.02 Hz Eq) which correspondto oscillations in the signal having a period comparable or supe-rior to the considered number of cardiac beats (512). Theserhythms are always present in the signals and may have animportant physiological meaning, the analysis of which is out-side the aim of the present paper. The method of variabilitysignal processing presented has the advantage of decreasing theestimation variance of their power.

The multivariate analysis (cross-spectra and phase spectra)between R-R interval and systemic arterial pressure variabili-ties are obtained through the calculation of the complex cross-spectrum P,y(f) where x(k) and y(k) are two synchronous dis-crete time series.

Pxy(f) = X(f)-Y*(f)/N = Gjy(f)exp[j4>ly(f)]

where X(f) and Y(f) are the discrete Fourier transforms of theseries (* denotes the complex conjugate), Gxy is the amplitudecross-spectrum, and 4>xy is the phase spectrum. The squaredcoherence k2 is then obtained as

The cross-spectrum is calculated via FFT algorithm, using atriangular window (Bartlett method)56 on successive over-lapping records of 64 samples each. Results are averaged overthe whole set of 512 samples. The applied segmentation is acompromise between frequency resolution and estimated con-sistency.57

Squared coherence has values between 0 and 1 and should beconsidered as analogue to the squared correlation coefficient r2

in linear regression analysis. Values over 0.5 indicate a signifi-cant phase link between R-R interval and pressure variabilitysignals. Thus, only spectral components with high coherencedemonstrate a stable phase-shift between instantaneous R-Rinterval and arterial blood pressure. Its value can be evaluated asa function of frequency.

I = I a(i)exp{ - jP(0 =

where f is the frequency and the sampling period is unitary. Theresulting PSD satisfies the criterion of maximum entropy16 andpresents many advantages in respect to the methods based onclassical Fourier analysis (FFT algorithms),18 namely, a moreconsistent and smoother spectral estimation, a spectral resolu-tion which is independent of the number N of samples, and thepossibility of avoiding windowing procedures.

Another important advantage of AR modeling with respect tothe more traditional techniques is the possibility of decompos-ing the autocorrelation function and, hence, its transform in thefrequency domain (i.e., the power spectrum density) in singlecomponents by applying the residuals theorem.55 In this way,the AR spectrum also provides the individual spectral compo-nents in terms of center frequency and of the correspondingpower in absolute, fractional (see Figure 1), and normalized

References1. Ludwig C: Beitrage Zur Kenntnis des Einflusses der Respira-

tionsbewegungen auf den Blutulauf im Aortensystem. ArchAnat Physiql (Muller's Arch) 1847; pp 242-257

2. Traube L: Uber periodische Thatigkeits-Ausserungen des va-somotorischen und Hemmungs-Nervenzentrums. CentralblattMed Wiss 1865;56:880-885

3. Hering E: Uber den Einfluss der Atembewegungen auf denKreislauf. 1. Mitteilung: Uber Atembewegungen des Gefas-systems. Sber Akad Wiss Wien, Math-naturwiss Klasse. 2.Abteilung 1869;60:829-856

4. Cyon E: Zur Physiologie des Gefassnervenzentrums. PflugersArch 1874;9:499-513

5. Mayer S: Studien zur Physiologie des Herzens und der Blutge-fasse: 5. Abhandlung: Uber spontane Blutdruckschwankun-gen. Sber Akad Wiss, 3. Abteilung 1876;74:281-307

6. Anrep GV, Pascual W, Rossler R: Respiratory variations of theheart rate. I. The reflex mechanism of the respiratory arrhyth-mia. Proc R Soc Lond [Biol] 1936;! 19:191—217




7. Schweitzer A: Rhythmical fluctuations of the arterial bloodpressure. J Physiol (Lond) 1945;104:25P

8. Hyndman BW, Kitney RI, Sayers BMcA: Spontaneousrhythms in physiological control systems. Nature 1971;233:339-341

9. Sayers B McA: Analysis of heart rate variability. Ergonomics1973;16:17-32

10. Chess GF, Tam RMK, Calaresu FR: Influence of cardiac neu-ral inputs on rhythmic variations of heart period in the cat. Am JPhysiol 1975;228:775-780

11. Kitney RI, Rompelman O (eds): The Study of Heart Rate Vari-ability. Oxford, Clarendon Press, 1980

12. Akselrod S, Gordon D, Ubel FA, Shannon DC, Barger AC,Cohen RJ: Power spectrum analysis of heart rate fluctuation: Aquantitative probe of beat-to-beat cardiovascular control. Sci-ence 1981;213:220-222

13. Bartoli F, Baselli G, Cerutti S: Application of identificationand linear filtering algorithms to the R-R interval measure-ments, in Computers in Cardiology. Silver Spring, Md., IEEEComputer Society Press, 1982, pp 485-488

14. De Boer RW, Karemaker JM, Strackee J: Beat-to-beat vari-ability of heart interval and blood pressure. Automedica1983;4:217-222

15. Pomeranz B, Macaulay RJB, Caudill MA, Kutz I, Adam D,Gordon D, Kilborn KM, Barger AC, Shannon DC, Cohen RJ,Benson H: Assessment of autonomic function in humans byheart rate spectral analysis. Am J Physiol 1985;248:H151-H153

16. Childers DG (ed): Modern Spectrum Analysis. New York,IEEE Press, 1978

17. Penaz J: Mayer waves: History and methodology. Automedica1978:2:135-141

18. Kay SM, Marple SL: Spectrum analysis: A modern perspec-tive. ProclEEE 1981:69:1380-1419

19. Kitney RI, Fulton T, McDonald AH, Linkens DA: Transientinteractions between blood pressure, respiration and heart ratein man. J Biomed Eng 1985;7:217-224

20. Pagani M, Furlan R, DelFOrto S, Pizzinelli P, Lanzi G, BaselliG, Santoli C, Cerutti S, Lombardi F, Malliani A: Continuousrecording of direct high fidelity arterial pressure and electrocar-diogram in unrestricted subjects. Cardiovasc Res I986;2O:384-388

21. Vatner SF, Pagani M, Rutherford JD, Millard RW, Manders T:Effects of nitroglycerin on cardiac function and regional bloodflow distribution in conscious dogs Am J Physiol 1978;234:H244-H252

22. Brovelli M, Baselli G, Cerutti S, Guzzetti S, Liberati D, Lom-bardi F, Malliani A, Pagani M, Pizzinelli P: Computerizedanalysis for an experimental validation of neurophysiologicalmodels of heart rate control, in Computers in Cardiology.Silver Spring, Md., IEEE Computer Society Press, 1983. pp205-208

23. De Boer RW, Karemaker JM, Strackee J: Comparing spectraof a series of point events, particularly for heart rate variabilitydata. IEEE Trans Biomed Eng 1984;31:384-387

24. Armitage P: Statistical Methods in Medical Research. Oxford,Blackwell Scientific Publications, 1971

25. Koizumi K, Terui N, Kollai M: Effect of cardiac vagal andsympathetic nerve activity on heart rate in rhythmic fluctu-ations. J Auton Nerv Syst 1985;!2:251-259

26. Fouad FM, Tarazi RC, Ferrario CM, Fighaly S, Alicandri C:Assessment of parasympathetic control of heart rate by a non-invasive method. Am J Physiol 1984;246:H838-H842

27. Ewing DJ, Neilson JMM, Travis P: New method for assessingcardiac parasympathetic activity using 24-hour electrocardio-grams. Br Heart J 1984;52:396-^*O2

28. Box GEP, Jenkins GM: Time Series Analysis. Forecasting andControl. New York, Holden-Day, 1976

29. Angelone A, Coulter NA: Respiratory sinus arrhythmia: Afrequency-dependent phenomenon. J Appl Physiol 1964; 19:479-482

30. Hirsch JA, Bishop B: Respiratory sinus arrhythmia in humans.How breathing pattern modulates heart rate. Am J Physiol1981;241:H62O-H629

31. Kollai M, Koizumi K: Reciprocal and nonreciprocal action of

the vagal and sympathetic nerves innervating the heart. J AutonNervSyst 1979; 1:33-52

32. Melcher A: Respiratory sinus arrhythmia in man. A study inheart rate regulating mechanisms. Acta Physiol Scand 1976;435:1-31

33. Haddad GG, Jeng HJ, Lee SH, Lai TL: Rhythmic variations inR-R interval during sleep and wakefulness in puppies anddogs. Am J Physiol 1984;247:H67-H73

34. Gootman PM, Cohen MI: Inhibitory effects on fast sympathet-ic rhythms. Brain Res 1983;270:134-136

35. Eckberg DL, Nerhed C, Wallin BG: Respiratory modulation ofmuscle sympathetic and vagal cardiac outflow in man. J Phys-iol (Lond) 1985;365:181-196

36. Akselrod S, Gordon D, Madwed JB, Snidman NC, ShannonDC, Cohen RJ: Hemodynamic regulation: Investigation byspectral analysis. Am J Physiol 1985;249:H867-H875

37. Borst C, Wieling W, Van Brederode JFM, Hond A, De RijkLG, Dunning AJ: Mechanisms of initial heart rate response topostural change. Am J Physiol 1982;243:H676-H681

38. Wallin BG, Sundlof G, Stromgren E, Aberg H: Sympatheticoutflow to muscles during treatment of hypertension with me-toprolol. Hypertension 1984;6:557-562

39. Rodeheffer RJ, Gerstenblith G, Becker LC, Fleg JL, WeisfeldtML, Lakatta E: Exercise cardiac output is maintained withadvancing age in healthy human subjects: Cardiac dilatationand increased stroke volume compensate for a diminished heartrate. Circulation 1984;69:203-213

40. Sever PS, Birch M, Osikowska B, Tunbridge RDG: Plasma-noradrenaline in essential hypertension. Lancet 1977; 1:1078-1081

41. Domhorst AC, Howard P, Leathart GL: Respiratory variationsin blood pressure. Circulation 1952;6:553-558

42. Pagani M, Schwartz PJ, Bishop VS, Malliani A: Reflex sym-pathetic changes in aortic diastolic pressure-diameter relation-ship. Am J Physiol 1975;229:286-290

43. Abboud FM, Thames MD: Interaction of cardiovascular re-flexes in circulatory control, in Shepherd JT, Abboud FM(eds): Handbook of Physiology, sec 2, The CardiovascularSystem, vol III, Peripheral Circulation and Organ Blood Flow.Washington, D.C., American Physiological Society, 1983, pp675-753

44. Eckberg DL, Abboud FM, Mark AL: Modulation of carotidbaroreflex responsiveness in man: Effects of posture and pro-pranolol. J Appl Physiol 1976;41:383-387

45. Mancia G, Ferrari A, Gregorini L, Parati G, Pomidossi G,Beitinieri G, Grassi G, Di Rienzo M, Pedotti A, Zanchetti A:Blood pressure and heart rate variabilities in normotensive andhypertensive human beings. Circ Res 1983;53:96-l04

46. Pagani M, Furlan R, Dell'Orto S, Pizzinelli P, Baselli G,Cerutti S, Lombardi F, Malliani A: Simultaneous analysis ofbeat by beat systemic arterial pressure and heart rate variabili-ties in ambulatory patients. J Hvperten 1985;3(suppl 3):S83-S85

47. De Boer RW, Karemaker JM, Stackee J: Relations betweenshort term blood-pressure fluctuations and heart rate variabilityin resting subjects: I. A spectral analysis approach. Med BiolEng Compul 1985;23:352-358

48. Wesseling KH, Settels JJ, Van Der Hoeven GMA, Nijboer JA,Butijn MWT, Dorlas JC: Effects of peripheral vasoconstrictionon the measurment of blood pressure in a finger. CardiovascRes 1985;19:139-145

49. Levy MN: Sympathetic-parasympafhetic interactions in theheart. Circ Res 1971 ;29:437-445

50. Schwartz PJ, Pagani M, Lombardi F, Malliani A, Brown AM:A cardio-cardiac sympathovagal reflex in the cat. Circ Res1973;32:215-221

51. Mark AL, Victor RG, Nerhed C, Wallin BG: Microneuro-graphic studies of the mechanisms of sympathetic nerve re-sponses to static exercise in humans. Circ Res 1985;57:461^*69

52. Patrick JM, Gibson Z, Hanley SP: Within-breath modulationof the triggering of extrasystoles in man. Clin Sci 1984;67:225-227

53. Bishop VS, Malliani A, Thore"n P: Cardiac mechanoreceptors,in Shepherd JT. Abboud FM, Geiger SR (eds): Handbook of




Physiology, sec 2, The Cardiovascular System, vol III, Periph- 57. Oppenheim AV, Schafer RW: Digital Signal Processing. En-eral Circulation and Organ Blood Flow. Washington, D.C., glewood Cliffs, N.J., Prentice-Hall, 1975American Physiological Society, 1983, pp 497-555

54. Akaike H: Statistical predictor identification. Am Int Stat Math1970;22:203-217

55. Zetterberg LH: Estimation of parameters for a linear difference

^ application to EEC analysis. Math Biosci 1969; K E Y WORDS • spectral analysis • variability sig-56. Jenkins GM, Watts DG: Spectral Analysis and Its Applica- 5:227-275nals • heart rate • high-fidelity arterial pressure

tions. New York, Holden-Day, 1968 • sympatho-vagal interaction


Power spectral analysis of heart rate and arterial ... · RAFFAELLO FURLAN, PAOLO PIZZINELLI,...

Documents

Transcript of Power spectral analysis of heart rate and arterial ... · RAFFAELLO FURLAN, PAOLO PIZZINELLI,...