1988;12;600-610 - lifemedicalcontrol.com · sure (sap) variabilities. as explained in the text,...

ISSN: 1524-4563 Copyright © 1988 American Heart Association. All rights reserved. Print ISSN: 0194-911X. Online

72514Hypertension is published by the American Heart Association. 7272 Greenville Avenue, Dallas, TX

1988;12;600-610 Hypertensionand A Malliani

M Pagani, V Somers, R Furlan, S Dell'Orto, J Conway, G Baselli, S Cerutti, P Sleight hypertension

Changes in autonomic regulation induced by physical training in mild

http://hyper.ahajournals.orglocated on the World Wide Web at:

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

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

[email protected]. E-mail: Kluwer Health, 351 West Camden Street, Baltimore, MD 21202-2436. Phone: 410-528-4050. Fax: Permissions: Permissions & Rights Desk, Lippincott Williams & Wilkins, a division of Wolters

http://hyper.ahajournals.org/subscriptions/Subscriptions: Information about subscribing to Hypertension is online at

at Countway Library of Medicine on July 16, 2008 hyper.ahajournals.orgDownloaded from

http://hyper.ahajournals.org

http://hyper.ahajournals.org/subscriptions/

mailto:[email protected]

http://www.lww.com/reprints


Changes in Autonomic Regulation Induced byPhysical Training in Mild Hypertension

MASSIMO PAGANI, VIREND SOMERS, RAFFAELLO FURLAN,

SIMONETTA DELL'ORTO, JAMES CONWAY, GIUSEPPE BASELLI,

SERGIO CERUTTI, PETER SLEIGHT, AND ALBERTO MALLIANI

SUMMARY The adaptive effects of physical training on cardiovascular control mechanisms werestudied in 11 subjects with mild hypertension. In these subjects we assessed the gain of the heart period-systolic arterial pressure relationship in the unfit and the fit state by using 1) an open loop approach,whereby the gain is expressed by the slope of the regression of heart period as a function of systolic arterialpressure, during a phenylephrine-induced pressure rise and 2) a closed loop approach with propersimplification, whereby the gain is expressed by the index a, obtained through simultaneous spectralanalysis of the spontaneous variabilities of heart period and systolic arterial pressure. Both methodsindicated that training significantly increased the gain of the relationship between heart period and systolkarterial pressure at rest and reduced arterial pressure and increased heart period significantly. This gamwas drastically reduced during bicycle exercise both in the unfit and fit state. In a second group ofnormotensive (u = 7; systolic pressure, 133 ± 3 mm Hg) and hypertensive (n — 1; systolic pressure,180 ± 10 mm Hg) subjects undergoing 24 -hour diagnostic continuous ekctrocardiographic and highfidelity arterial pressure monitoring, the index a was significantly reduced in the hypertensive group atrest. Furthermore, when analyzed continuously over the entire 24-hour period, this index underwentminute-to-minute changes with lower values during the day and higher values during the night. Wepropose the index a as a quantitative indicator of the changes in the gain of baroreceptor mechanismsoccurring with physical training in mild hypertension and during a 24-hour period in ambulatory subjects.(Hypertension 12: 600-610, 1988)

KEY WORDS • baroreceptor reflex mechanisms • exercise • 24-hour direct arterial bloodpressure • neural control of circulation • power spectrum analysis

PHYSICAL training induces bradycardia anda modest reduction in arterial pressure.1' 2

This raises the possibility of recommendingphysical training as part of the nonpharmacologicaltreatment of hypertension.1- 2 The mechanismsunderlying the observed cardiovascular changes arenot yet understood but are likely to be quite com-plex; they may include metabolic, cardiovascular,and neural factors in addition to changes in skeletalmuscle fiber type. Our study explored the hypoth-esis that training induces new operating conditionsin the neural regulatory mechanisms. We compared

From the Istituto Ricerche Cardiovascolari (CNR), PatologiaMedica, Centre Fidia, Ospedale L. Sacco, Universita Milano, Milano(M. Pagani, R. Furlan, S. Dell'Orto, A. Malliani); John RadcliffeHospital, Oxford, UK (V. Somers, J. Conway, P. Sleight); Diparti-mento Elettronica, Politecnico, Milano (S. Cerutti); and Diparti-mento Automazione Industriale, Universita Brescia, Brescia (G.Baselli), Italy.

Address for reprints: Massimo Pagani, M.D., Istituto RicercheCardiovascolari, via Bonfadini 214, 20138 Milano, Italy.

Received September 22, 1987; accepted August 8, 1988.

two different methods of measuring some indices ofthis autonomic regulation.

Several techniques are available to detect suddenchanges in sympathetic outflow in humans (e.g., thelevel of circulating catecholamines3 or direct elec-trical recording of the impulse activity from sympa-thetic fibers4). However, long-term adaptive changesin parasympathetic outflow, such as those producedby physical training, cannot be assessed by similartechniques. It is possible to gain some insight intovagal and sympathetic control of the heart by mea-suring the reflex bradycardia that results from thearterial pressure rise induced by intravenous injec-tion of a pressor drug5; with this approach thebaroreceptor reflex control of heart rate is quanti-fied by calculating the slope of the regression ofheart period (RR interval) on systolic arterial pres-sure (SAP) during the blood pressure rise inducedby phenylephrine. A reduction in this slope occurstemporarily during physical exercise6-7 or chroni-cally in hypertension or in older subjects.8

600



AUTONOMIC CHANGES WITH PHYSICAL TRAINING/Pa^am et al. 601

In the present study, carried out in a selectedpopulation of patients with mild arterial hyperten-sion undergoing an intense physical training pro-gram, we had the opportunity of comparing the gainof the reflex bradycardia induced by i.v. injection ofphenylephrine with other recently described indicesobtained by power spectral analysis.9- I0 It has longbeen known that beat-by-beat, spontaneous fluctu-ations of RR and arterial pressure (Figure 1) giverise to definite rhythms. We found that these rhythmscan be well quantified by spectral analysis, asshown in Figure 2, and that the powers of the twomajor oscillations can be used as markers ofshort-term" and long-term12 changes in cardiovas-cular neural control activities. In particular, thehigh frequency (HF) respiratory component (at 0.25Hz) seems to reflect mostly vagal modulation11-13-14

while the power of the low frequency (LF) oscilla-tion (0.1 Hz), although amplified by vagalactivity13- M seems mainly to provide an index ofsympathetic modulation and its changes.11- 12

In the present study, we used an automatic bivar-iate parametric spectral analysis9 to obtain the gainof the relationship existing between spontaneousRR and SAP oscillations. The two methods (phen-ylephrine injection and spectral analysis) furnishcomplementary information and both suggest thatimportant changes in the neural regulatory mecha-nisms are induced by physical training.

Subjects and MethodsStudy of the Effects of Training

This part of the study was composed of 11subjects (age, 32 ± 2 years; three women, eightmen) with uncomplicated mild hypertension with-out target organ damage. They participated in aresearch project in the Department of Cardiovascu-lar Medicine, John Radcliffe Hospital (Oxford, UK),designed to assess the systemic arterial blood pres-sure-lowering effect of physical training.2 The studywas approved by the ethics committee of the hos-

200ARTERIALPRESSURE(mmHg)

031200)

MEANPRESSURE(mnJty)

HEARTRATE(Ivmln)

• n<

FIGURE 1. Analog recording of ECG and direct intra-arterial pressure at rest (left panel) and during moderatebicycle exercise (right panel) in a mildly hypertensivesubject in the fit state.

pital and all the subjects gave informed consent.These patients were allocated to one of two groups.

In the first group (n = 7), a first hemodynamicstudy was performed with subjects in the unfit state,and then after a 6-month training period they werestudied again. The other group of subjects {n = 4)underwent a 6-month training period before the firsthemodynamic study. They were studied again aftera 4-month period during which they avoided physicalactivity as much as possible (i.e., the "detraining"period). Thus, with this pseudo-crossover design allsubjects underwent two sessions of direct arterialpressure recording.

The training period consisted of a daily routine of22 minutes of calisthenics15 followed by 20 minutesof jogging at least five times a week. The trainingperiod was supervised by one of the authors (V.S.).The hemodynamic study was performed at theOxford Center. After a light breakfast, with nocoffee or alcohol then or in the previous 12 hours,the subjects, who were not taking any medication,arrived in the laboratories at about 1000. A Tefloncannula (internal diameter, 1 mm) was placed in thebrachial artery of the nondominant arm by theSeldinger technique and was connected to a pres-sure transducer (Statham Gould, Oxnard, CA, USA).The catheter-manometer system had a frequencyresponse flat (±5%) to at least 8 Hz, with maximumresonance at 14 Hz and the -3dB point at 26 Hz.

An indwelling catheter was also positioned in thebasilar vein of the same arm for injecting phenyl-ephrine (40-60 /tig) just prior to the end of the restand exercise periods. The electrocardiogram (ECG;lead II) was also recorded. Data were fed to alaboratory frequency-modulated tape recorder(Racal, Southampton, England) for later analysis.

After a period of equilibration, patients werestudied at rest for 20 minutes. They were then askedto pedal for 10 minutes on a bicycle ergometer at100 W work load (see Figure 1).

The subjects included in this study all had pres-sure recordings adequate for computer analysis andfree of artifacts throughout. Only 10 subjects satis-fied these criteria during exercise.

Study of the Effects of HypertensionAn additional group of 18 subjects, without evi-

dence of any disease and with arterial pressurelevels ranging from normotension to hypertensionwas used for this section of the study. Thesesubjects were part of an ongoing study of bloodpressure variability originating at the Ospedale L.Sacco (Milan, Italy). As part of the protocol, pre-pared in accordance with institutional guidelines,these subjects underwent a 24-hour direct highfidelity blood pressure recording16 that employed aminiature (diameter, 3F) catheter tip transducer(Millar, Houston, TX, USA), which was introducedinto the radial artery of the nondominant arm by theSeldinger technique. The transducer has a highdegree of constancy in gain and stability and a wide



602 HYPERTENSION VOL 12, No 6, DECEMBER 1988

•[««]

kin

FIGURE 2. Example of the simultaneous computeranalysis of the RR interval and systolic arterial pres-sure (SAP) variabilities. As explained in the text, afterappropriate analog-to-digital conversion of the ECGand arterial pressure signals, the RR interval and SAPseries are obtained (top and middle left panels). Then,the corresponding spectra (power spectral density[PSD]) are computed (top and middle right panels).The bottom left panel shows the cross spectrum (Cross);the bottom right panel shows the phase relationship(<&) and the squared coherence (K2, curve with twomajor peaks).

band pass (better than 1 kHz); the pressure signal isrecorded, along with the ECG, on a modified four-track Holter tape recorder (Remco, Milan, Italy).The signal is recorded with frequency modulation,and the frequency response of the entire record-playback system has a - 3 dB point at 50 Hz. Tominimize tape speed errors, a crystal-controlled 160Hz signal is recorded on a channel of the portablerecorder. This signal is used during playback by aphase-locked loop circuitry (Remco), which pro-vides an error signal proportional to any possiblechange in recording tape speed. This voltage is usedto correct instantaneously the voltage supply of theDC motor of the playback unit, thus virtually abol-ishing tape speed variations.

At the end of the 24-hour ambulatory recording,the subjects were disconnected from the portablerecorder and connected to a frequency-modulatedrecorder (Racal) while lying flat in the laboratory.Blood pressure and the ECG were recorded duringa 20-minute rest period.

Data AnalysisOff-line analysis was performed in Milan. Data

were played back from Racal tape at real time andanalog-to-digital (A/D) conversion was performedat 300 samples/sec per channel on a Mine 23 com-puter (Digital Equipment, Maynard, MA, USA). Inthe case of the 24-hour recordings, playback wasperformed at 64 times real time, thus allowingacquisition of the entire recording in about 22minutes. At this speed, A/D conversion (PDP 11/24,Digital Equipment) is performed with 19,200 samples/sec per channel in order to achieve 300 samples/secfor channel real time. Spectral and cross-spectralanalysis was performed using a PDP 11/24 and aVAX 750 computer (Digital Equipment). The prin-ciples of the software for data acquisitions andanalysis have been described previously." Schemat-

ically (see Figure 2), the computer first calculatesthe interval tachogram (i.e., the series of consecu-tive RR intervals) as well as the systogram (i.e., theseries of systolic values synchronized to the beat atthe beginning of each RR interval). The length ofthe tachogram and systogram was usually 512 beats.From these series the computer program automat-ically calculates simple statistics (mean and vari-ance) as well as the autoregressive coefficientsnecessary to define the estimate of the power spec-tral density.11

An important feature of the program is that it alsocalculates the model that provides the best statisti-cal estimate of the power spectrum (maximumentropy spectrum) and prints out the power andfrequency of each component. Every spectral com-ponent is presented in absolute units and also in anormalized form obtained by dividing it by the totalpower minus any DC component." Normalizedvalues allow comparisons among subjects whenlarge interindividual variability exists,17 as is thecase for RR variance." Only stationary (steadystate) sections of recordings were analyzed. Station-arity was defined as a difference of less than 5% inthe spectral components calculated in two succes-sive 256-beat series, with the subject in a steadystate." It should be pointed out that with this tech-nique the duration of the periodical phenomena inthe variability signal was measured as a function ofcardiac beats rather than seconds. However, thisfrequency can be converted into its Hertz equivalentwithout producing harmonic effects by dividing it bythe average RR interval." Accordingly, the units forfrequency in this paper will be indicated by theabbreviation Hz. In the case of the 24-hour record-ings, a recursive version of the program is used.

Assessment of the RR Interval-Systolic ArterialPressure Relationship

As previously described,5 the slope of the linearregression of RR interval as a function of SAP



AUTONOMIC CHANGES WITH PHYSICAL TRMNWG/Pagani et al. 603

during the pressor rise produced by the i.v. injec-tion of phenylephrine was computed using a DataGeneral Eclipse 200 computer (Westborough, MA,USA). This value provides an estimate of the gainof the open loop model of the RR-SAP relationshipas schematically depicted below.

H..

According to this model s and t are intended asbeat-to-beat SAP and RR duration values, respec-tively, //u is the transfer function between s and t. Itshould be noted that s (and i) values (small signalrepresentation) are the difference between individ-ual SAP (and RR interval) values and the mean ofthe respective series. In order to get a mathematicalequation that describes the input-output relation-ship of the model in terms of its transfer function, itis necessary to transform the input and outputsignals into the frequency (f) domain, thus obtain-ing T(f) = Hu- S(f) where T and 5 are the Fouriertransforms of / and s, respectively. In the particularcase considered above, we have Hu = k (where k isthe constant slope of the regression straight linebetween s and t values), and therefore the input-output relation between the RR/SAP magnitudesbecomes T(f) = k-S(f) or t = k-s.

However, changes in / could also induce changesin s. For example, an increase in RR, and hence inthe filling period of the heart, will tend to produce agreater stroke volume, and hence a higher SAPvalue, although blunted by the longer diastolicaortic runoff. On the other hand, s could affect /through various neural reflexes operating with bothpositive and negative feedback mechanisms.18 Toaccount for these effects, a closed loop modelappears more appropriate (Figure 3).

The two transfer functions Ha and Hu describethe simultaneous effect of t on s and of s on t,

FIGURE 3. Schematic representation of the closed loopmodel of the relationship between tachogram (t) and systo-gram (s), with the indication of the two noises n, find n,.

respectively. The interactions under study are rel-evant to variability phenomena that are small inrespect to the set point values (mean values) and arewell suited to be described by linear equations, atleast as a first order approximation. H's are gener-ally complex operators (commonly expressed underthe form of gain and phase) that are functions offrequency / . Two noises, n, and nf, are also indi-cated to describe the external inputs to the system,such as mechanical disturbances, respiration, andsmall adjustments by central command, to take intoaccount all the possible sources that can causevariability of the t or s signal independently of s ort, respectively.

Experimental measurement of the gain of theneural feedback from SAP to the RR period ispossible in two essentially equivalent ways: 1) byforcing a change in SAP and consequently measur-ing the relevant variations in RR duration (thismodeling technique is equivalent to the experimen-tal increase in SAP produced by an intravenouslyadministered vasoconstrictor agent, implying a vir-tually open loop at the level of s); 2) by analyzingthe simultaneous variabilities of SAP and RR dura-tion, which are actually interacting in a closed loop.Using the latter approach, it should be noted that acomplete solution of the system shown in Figure 3is impossible without additional a priori hypothe-ses. The problem involves the computation of Hu,/ / „ , n,, and n, spectral and cross-spectralcharacteristics.10 In general, n, and n, are coloredprocesses (i.e., they present power peaks in spe-cific bands) and are correlated (i.e., they affect RRduration and SAP but not independently). Theseinteractive effects on RR and SAP variabilitycould otherwise be erroneously attributed to effectsof the loop.

One possible way to solve this problem, as sug-gested by Akselrod et al. for the LF components,19

starts from the simplified hypothesis that all distur-bances enter the system (or are generated) only at slevel through n, (let us suppose that nt = O).20 Inthis way, when nt tends to zero, it is possible tocompute the transfer function of the closed loopmodel. Hence, it is easy to obtain the gain of //„(i.e., its modulus \HU\) as a function of frequency/(see also Appendix) from the power spectra (P) of tand s following the formula below.

\Hu\ = [Pt(f)/P,(f)]m

In particular, the modulus \HU\ is calculated in thetwo major bands: those centered at 0.1 Hz (LF) andat the respiratory frequency (HF). This parameter isindicated as aLF and aHF for LF and HF bands,respectively. Therefore, under the hypothesis dis-cussed above, in the closed loop model the indexaLF is the gain of the relationship between SAPvariability and RR interval variability at about 0.1Hz; aHF is the same index relative to the respiratoryfrequency. The correct interpretation of a requiresthat there is high coherence in these bands (see also




Appendix); therefore the calculation of a was madeafter verification of a high value (>0.5) of coherencein the LF and HF bands (see Figure 2).

StatisticsResults are presented as means ± SE. The sig-

nificance of the effects of training and exercise, aswell as the difference between the changes pro-duced by exercise in the fit and unfit states, wasassessed by the two-tailed paired / test. The unpairedt test was used to assess the difference betweennormotensive and hypertensive subjects. Multiplecorrelation analysis of aHF, aLF, and the phenyl-ephrine slopes in the fit and unfit states, with thetraining bradycardia and training hypotension, wasused to assess the effects of training. A probabilitylevel below 0.05 was considered significant.21

ResultsEffects of Training on the Gain of the HeartPeriod-Systolic Arterial Pressure Relationship

In the 11 subjects with mild hypertension, arterialpressure and heart rate, both at rest and duringsteady state exercise, always showed spontaneousbeat-by-beat oscillations around their mean values(see Figure 1). Spectral analysis provided a quanti-tative assessment of their nonrandom components(Table 1; see Figure 2). Consistently, two majorcomponents were observed at ~0.1 (LF) and 0.25(HF) Hz, respectively; the LF component hadgreater power in both RR and SAP variability (seeFigure 2 and Table 1) than the HF (respiration-linked) component. Cross-spectral analysis indi-cated that only at these two frequencies was thecoherence value greater than 0.5.

Physical training significantly increased the aver-age resting heart period (RR) from 796 ± 77 to923 ± 39 msec, thus producing a training bradycar-dia as well as modifying the powers of both the LFand HF components of RR variability. LF wasreduced and the respiration-linked component (HF)increased; however, there was some overlap in thedistribution of data. SAP was significantly reducedby 10 ± 2 from 147 ± 4 mm Hg and diastolic arte-rial pressure (DAP) by 7 ± 1 from 82 ± 3 mm Hg asa result of training.

Training significantly increased the gain of theRR-SAP relationship at rest (see Table 1). Thiseffect was assessed in two ways: from the slope ofthe regression of the RR period during a pressurerise produced by i.v. injection of phenylephrine andby the index a (i.e., by the square root of the ratiobetween spectral components of RR and SAP vari-abilities). This index was computed for both LF andHF components (see Subjects and Methods and theAppendix), and there was a significant correlation(p < 0.001) between both these indices and thephenylephrine RR-SAP slopes in the populationstudied (Figure 4). Thus, the phenylephrine slopeand both the aLF and aHF at rest were significantlygreater after physical training (see Table 1). Fur-thermore, for both aLF (r = 0.82) and aHF (r = 0.80)this increase was positively correlated with theattendant increase in resting RR. On the other hand,the magnitude of the fall in arterial pressure withtraining was not significantly correlated with eitherthe changes in a or the phenylephrine slope.

During 100 W bicycle exercise (Table 2), the RRinterval was reduced significantly: by 274 ± 23 msecin the untrained state and by 372 ± 31 msec in the

TABLE 1.

Group

UntrainedTrained

UntrainedTrained

UntrainedTrained

UntrainedTrained

Effects of Training on

Slope(msecAnm Hg)

14.7+2.519.5+2.7*

RR(interval msec)

796 ±27923 ±39*

SAP(mm Hg)

147+4138+3*

DAP(mm Hg)

82±375 + 3*

RR and Arterial Pressure Variabilities at Rest in 11

(msec/mm Hg) (sec)

10.4+1.3 2.7±0.215.4+3.0* 3.3+0.3

RR(variance msec2)

1831+2805465 ±1853

SAP(variance mm Hg2)

19+318+2

DAP(variance mm Hg1)

12+213 + 2

*(degrees)

Subjects with Mild

<*HF

(msec/mm Hg)

- 9 5 ± 3 12.2+0.7-104±7 21.5+3.2*

LFofRR

(nu) (Hz)

65+2 0.11 ±0.0157±3* 0.10+0.01

LF of SAP

(mmHg2)

7.4+1.56.1+0.9

LFof

(mmHg2)

5.7±0.94.6±0.7

(Hz)

0.10+0.010.09±0.01

DAP

(Hz)

0.10+0.010.09±0.01

Hypertension

<J>(sec)

*(degrees)

0.53±0.10 -5O±90.38±0.04 - 3 6 ± 3

HFofRR

(nu) (Hz)

24±2 0.28±0.0131 ±3* 0.27+0.01

HF of SAP

(mmHg2)

1.6+0.31.6+0.3

HFof

(mmHg2)

0.6±0.11.0+0.3

(Hz)

0.28±0.010.27±0.01

DAP

(Hz)

0.28±0.010.27±0.01

Values are means ± SE. Slope = slope of the regression of RR as a function of systolic arterial pressure (SAP) during the pressurerise produced by i.v. injection of phenylephrine; a u , oHF = gain of the RR-SAP relationship, closed-loop model; <t> = phase angle of thecross spectrum between RR and SAP variability. The negative sign indicates that pressure oscillations lead RR interval oscillations. LF= low frequency component; HF = high frequency component; nu = normalized units; DAP = diastolic arterial pressure.

*p < 0.05 compared with untrained value.



AUTONOMIC CHANGES WITH PHYSICAL TRAINING/Pa^am et al. 605

3 4

SLOPEmsec/mm Hg]

0

' y.e.uasx' .•' . r.aei

'•4 0 4 0

• LF[mi«/nmHg] t ̂ [mt«c/mmHg]

FIGURE 4. Regression analysis of the values of the phenylephrine slopes and the corresponding valuesofaLf and aHF. Subjects were assessed both before and after training, and the analysis was performedon the entire group of data.

trained state. The variance of RR was drasticallyreduced by exercise both before and after trainingto 640 ± 142 and 828 ± 252 msec2, respectively. Inthe fit state, the LF component of RR variabilityincreased by 22 ± 4 normalized units from thevalue observed at rest (Figure 5).

During exercise in the unfit state, SAP and DAPincreased significantly (by 27 ± 4 and 6 ± 2 mmHg), together with their total variance (by 37 ± 5and 14 ± 2 mm Hg2), from the values observed atrest. Furthermore, the LF component of both SAPand DAP variability was markedly augmented dur-ing exercise: by 11 ± 3 and 5 ± 2 mm Hg2, respec-tively.

In the trained state, similar changes in SAP andDAP (29 ± 4 and 4 ± 1 mm Hg, respectively) wereobserved during exercise; however the increases inthe LF component of both SAP and DAP (22 ± 6

and 11 ± 3 mm Hg2) were more evident than in theunfit state.

Finally, 100 W bicycle exercise in the unfit statedrastically reduced from the resting values thephenylephrine slope (by 12 ± 3 msec/mm Hg) aswell as aLF and aHF (by 8 ± 1 and 9 ± 1 msec/mmHg, respectively). Similar changes were observedin the fit state (i.e., the slope was reduced by 17 ± 5msec/mm Hg, aLF by 13 ± 3 msec/mm Hg, and aHFby 19 ± 3 msec/mm Hg).

Cross-spectral analysis allowed a quantificationof the phase angle between RR and SAP beat-by-beat variabilities. At rest (see Figure 2), there was alag of about 50 degrees, corresponding to 0.5 sec-onds, at the respiratory frequency, with pressureleading. The lag was greater in the case of the LFcomponent (i.e., about 100 degrees corresponding

TABLE 2. Effects of Training on RR and Arterial Pressure Variabilities During Exercise in 10 Subjects with Mild Hypertension

Group

UntrainedTrained

UntrainedTrained

UntrainedTrained

UntrainedTrained

Slope(msec/mm Hg)

2.8+0.62.6±0.5

RR(interval msec)

526±35555+26

SAP(mm Hg)

177±5167±6

DAP(mm Hg)

86±378±3*

<*LF *

(msec/mm Hg) (sec)1.8+0.6 2.8+0.32.0±0.4 3.4±0.4

RR(variance msec2)

640±142828+252

SAP(variance mm Hg2)

56±666±ll

DAP(variance mm Hg2)

28±429±5

* «HF(degrees) (msec/mm Hg)-97±U 2.2±0.5

-108±8 2.8±0.6

LF of RR

(nu) (Hz)

69±8 0.11 ±0.0177±6 0.11 ±0.01

LF of SAP

(mm Hg2) (Hz)

19+3 0.10±0.0129±5* 0.10±0.01

LF of DAP

(mm Hg2) (Hz)

11±2 0.10+0.0115±3 0.10±0.01

(sec)*

(degrees)

0.6±0.1 -72±130.5±0.1 -77±15

HF of RR

(nu) (Hz)

23±4 0.34±0.0219±6 0.36±0.03

HF of SAP

(mm Hg2) (Hz)

6±1 0.34±0.026±1 0.36±0.02

HF of DAP

(mm Hg2)

3±0.42±0.4

(Hz)

0.34±0.020.36±0.02

Values are means ± SE. Slope = slope of the regression of RR as a function of systolic arterial pressure (SAP) during the pressurerise produced by i.v. injection of phenylephrine; a^, aHF = gain of the RR-SAP relationship, closed-loop model; * = phase angle of thecross spectrum between RR and SAP variability. The negative sign indicates that pressure oscillations lead RR interval oscillations. LF= low frequency component; HF = high frequency component; nu = normalized units; DAP = diastolic arterial pressure.

*p < 0.05 compared with untrained value.




COWTROt

0

A /v

FIGURE 5. Spectral analysis of simultaneous RR (toppanels) and systolic arterial pressure (bottom panels)variabilities in a trained subject, at rest and duringexercise (inset in lower left panel: x 10 magnification ofamplitude). Note the marked predominance of the lowfrequency component in both spectra during exercise.

to 3 seconds). This lag was not significantly modi-fied by exercise or by training.

Day-Night Variations in RR-SAP RelationshipThe short-term changes in the gain of the rela-

tionship between the RR period and SAP, asreflected by both aLF and aHF, were assessed fromcontinuous recordings of ECG and high fidelityarterial pressure for 24 hours in six ambulatoryMilan subjects (Table 3). Figure 6 illustrates thisanalysis. The well known day-night differences inRR and SAP can easily be appreciated, with maxi-mum RR and minimum pressure occurring duringthe night hours. Similarly, both aLF and OHF demon-strate higher values during the night and lowervalues during the waking hours. The slow trends inday-night oscillations can be well represented byhourly averages, as shown in Figure 6 (right panels).However, only a continuous representation of thedata throughout the recording period can depict thefast oscillations that are superimposed on the slowday-night changes (see Figure 6, left panels).

RR-SAP Relationship in Normotensive andHypertensive Subjects at Rest

In the nine hypertensive subjects SAP and vari-ance were greater than in the nine controls, while

the difference in RR interval and variance was not.Furthermore, in hypertensive subjects the gain ofthe RR-SAP relationship (i.e., % and aHF) wassignificantly smaller (Table 4). However, the age ofthe normotensive subjects was lower, although notsignificantly, than the age of hypertensive subjects(see Table 4); thus, age8 may in part account for thedifferences in aLF and aHF that we have observed.

DiscussionIn this study we employed methods based on

control theory to investigate with a simplified closedloop model the relationships between heart period(RR interval) and arterial pressure. To this purpose,we assessed: 1) the gain of the RR-SAP relationshipusing the phenylephrine method5 during rest andexercise, both before and after physical training;2) the gain of the RR-SAP relationship using theindex a obtained, under the same conditions, withsimultaneous spectral analysis of RR and SAP varia-bilities.9 Furthermore we assessed continuous vari-ability of the index a through the analysis of 24-hour,direct high fidelity arterial pressure recordings inambulatory patients as well as the difference betweennormotensive and hypertensive subjects at rest.

RR-SAP RelationshipIn this study the RR-SAP relationship was eval-

uated in two ways. First, we calculated the slope ofthe regression of RR intervals as a function of SAPduring the pressor rise produced by an i.v. bolusinjection of phenylephrine. This approach, devel-oped and tested at the Oxford Center since 19695

has the advantage of requiring only a simple linearrelationship between the two variables (i.e., RRperiod and SAP). According to this technique, theslope of the regression provides an index of the gainof the reflex arc, the afferent limb from the barore-ceptors, projecting to the medulla, and the efferentlimbs in both the cardiac vagal and sympatheticnerves. This approach revealed that the gain of thebaroreceptor reflex control of heart rate was affectedby a variety of pathophysiological situations.6 Myo-cardial infarction,22-a hypertension,8-24-M exercise,26

and sympathetic activation27 all reduce, while sleep5

increases, the gain of the relationship. Additionally,a marked increase in the gain of the reflex has been

TABLE 3. 24-Hour Changes in Arterial Pressure, Pulse Interval, and a Values in Six Ambulatory Subjects

Measuredvalues

AverageVariance*MaximumMinimum

Systolic(mm Hg)

138±3139±21162±5123±3

Arterial pressure

Mean(mm Hg)

108±375 ±10

122±497±3

Diastolic(mm Hg)

86±456±999±576±5

RR interval(msec)

791 ±2116921±44731040±55525±31

(msec/mm Hg)

7.9±1.553.6±20.828.6±6.52.3±0.3

(msec/mm Hg)

8.4±1.345.3 + 13.422.5+5.13.2±0.4

Values are means ± SE. LF = low frequency; HF = high frequency;closed loop model.

•Variance units = SD2.

, OHF = gain of the RR-systolic arterial pressure relationship,



AUTONOMIC CHANGES WITH PHYSICAL TRAINING/Pagani et al. 607

Rfl-MTERVAL

FIGURE 6. Arterial pressure, RR interval, a^, and aHF,as obtained by computer analysis of direct ECG and highfidelity intra-arterial continuous pressure recordings dur-ing a 24-hour period. In the left panels is a continuousrepresentation of the data. The right panels show bihourlyaverages with standard deviation superimposed.

reported with physical training in both consciousdogs23 and patients22 trained after a prior myocar-dial infarction.

Experimental evidence indicates that other fac-tors modify the gain of this reflex mechanism andshould be included in the analysis of the RR-SAPrelationship (e.g., the central command and the roleplayed by other sensory inputs, such as that sub-served by the cardiovascular sympathetic afferentfibers).18- «.2«

An additional problem in the assessment of RR-SAP relationship is provided by the mechanicaleffects of changes in the RR interval on the follow-ing SAP value. Therefore, a closed loop model, asalready suggested by Akselrod et al.,19 seems moreappropriate to describe the complex interactionbetween beat-by-beat changes in RR (i.e., the tacho-gram) and SAP (i.e., the systogram; see Subjectsand Methods and the Appendix). Hence, weemployed an approach to the study of the RR-SAPrelationship that is based on the analysis of beat-by-beat spontaneous changes in RR and SAP anddoes not require a disturbance in the pressure fromoutside the system such as that produced by i.v.administration of a pressor drug.5

This problem has also been addressed by others,although with different methods. De Boer et al.,29

using stationary recordings of several minutes, com-puted the regression of the difference between suc-cessive RR values of SAP. The slope of that regres-sion was considered analogous to the phenylephrineslope. Bertinieri et al.30 selected short episodes of 3to 5 successive beats, in which changes in SAP andRR period were directionally similar, and computeda baroreceptorlike slope. Robbe et al.10 used spec-tral analysis to compute the modulus of the relation-ship between SAP and RR variability in the band0.07 to 0.14 Hz in order to quantify the gain of theRR-SAP relationship.

In the present study a two-way autoregressivebivariate model provides an index of the gain of theclosed loop relationship between tachogram andsystogram as a whole. The gain of the relationship isequal to the square root of the ratio, at any givenfrequency, of the RR and SAP variabilities, pro-vided the coherence value between the two signalsapproaches unity (see the Appendix and Subjectsand Methods). In this way two indices are obtained(aLF and «HF) that indicate the gain of the relation-ship between tachogram and systogram, from whichthe possible frequency dependency is also ana-lyzed. It is interesting to observe that essentially thesame gain is observed both at LF and HF, inkeeping with the simplified hypothesis of consider-ing n, negligible not only at the LF19 but also at theHF components. On the contrary, the phase differ-ence between tachogram and systogram occurringduring fast respiration-linked oscillations (HF) issmaller than that measured during the slow oscilla-tions (LF). This finding could explain the variablelatency reported for baroreceptor heart rate reflexesin humans.31 In the present study, both the slope and

TABLE 4. Relationship ofRR and Systolic Arterial Pressure inNormotensive and Hypertensive Subjects at Rest

Measured variablesAge (yr)Sex (M/F)SAP (mm Hg)Variance (mm Hg2)RR interval (msec)Variance (msec2)a u (msec/mm Hg)<t>(sec)4> (degrees)OHF (msec/mm Hg)<J> (sec)<t (degrees)

Normotensive(n=9)

41±67:2

133±320±2

828±37601 ±223

10±2-2.67±0.35

-90±99±2

-0.37±0.29-28±19

Hypertensive(«=9)49±33:6

177±9*44±8*

799±271123±189

4±1*-2.24±0.27

-75±84±1*

-0.00±0.22-22±12

Means are values ± SE. SAP = systolic arterial pressure. % ,am = gain of the RR-SAP relationship, closed loop model;<t> = phase angle of the cross spectrum between RR and SAPvariabilities. The negative sign indicates that pressure oscilla-tions lead RR interval oscillations.

*p < 0.05, compared with normotensive subjects.




the spectral analysis method indicate that trainingincreases the gain of the tachogram-systogram rela-tionship while exercise drastically reduces it. Fur-thermore the results obtained employing eithermethod were significantly correlated (see Figure 4).10

The mechanisms responsible for the changes thatwe have observed cannot be specifically identifiedbecause of the complexity of the neural circuitry,which is composed of arterial baroreceptive, vagaland sympathetic afferent fibers, together with theintegration within the central nervous system of thereflex mechanisms that they mediate (Figure 7).18

This modeling approach can only indicate a changein the net gain of the complex interaction withoutproviding information on the changes in the tonicactivity of the major constituents of the sympathova-gal balance, which can instead be inferred from thestudy of RR and arterial pressure variabilities withspectral techniques.11 However, the presentapproach is useful in showing that the net gain (a) ofthe neural circuitry undergoes fast minute-to-minute changes, as indicated by the analysis of24-hour recordings of ECG and direct arterial pres-sure in ambulatory subjects.

As to the possible hemodynamic consequences ofchanges in the gain of the RR-SAP relationship, wecould consider the importance of higher values of ain promoting a buffering of the beat-by-beat varia-tions in arterial pressure by way of changes in theRR period when the system is operating towardstability.32 Conversely, lower values of a, as ob-served in hypertension, may be important in pro-moting sudden changes when the system is shift-ing toward instability.18

Effects of Physical Training and Exercise on RRand Arterial Pressure Variabilities

In this study we have used spectral analysis toquantify the beat-by-beat oscillations present in the

FIGURE 7. Schematic representation of the closed loopmodel of the relationship between tachogram (t) andsystogram (s). In this scheme the neural block containsboth negative (H,,~) and positive (H,^) feedback mech-anisms. Thus, the gain of transfer function depends onthe complex interaction between these functional compo-nents and the central command.

heart rate and arterial pressure signals, in accord-ance with the concept that these oscillations reflectthe interaction between sympathetic and vagalcontrol.11 The HF respiratory component of RRvariability has consistently been considered to reflectmodulation of vagal efferent activity on the basis ofexperiments on vagotomized, decerebrated cats,33

conscious dogs,13 and humans with muscarinic recep-tors blockade.14 In prior studies the power of the LFcomponent has been attributed to both vagal andsympathetic mechanisms.19

We consistently correlated the presumed changesin sympathetic activity with changes in the normal-ized power of LF. This correlation was observed inboth the RR and arterial pressure variability signals. •'However, the possibility of a vagal modulation of thepower of the LF oscillations of RR variability wassuggested by experiments with vagotomy34 or mus-carinic receptor blockade13- M in conscious dogs. TheLF oscillations in arterial pressure, with a 10 secondrhythmicity (i.e., Mayer waves), which have longbeen recognized in short-term animal experiments,35

have more recently been observed in normalhumans.36 Using high fidelity blood pressure record-ing techniques in humans37 in various conditions, wefound that Mayer waves follow the gradual increaseof sympathetic activity from rest, through tilt," totreadmill exercise38 or the decrease of activity duringsleep.37 We should note, however, that spectralmarkers of autonomic control cannot distinguishbetween efferent nerve activity and target organresponse; instead, they provide a quantitative indi-cation of the complex interaction between the vari-ous neural and nonneural components.

The major effects of physical training at rest, wererepresented by an increase in average RR (i.e., by atraining bradycardia), together with a significantreduction in SAP and DAP. Bradycardia at rest is awell-known consequence of physical training andhas been attributed to changes in neural control ofthe heart,39 either an increase in vagal activity40 or adecrease in sympathetic activity,41-42 or else to areduced intrinsic heart rate.43 In this group of sub-jects with mild hypertension spectral markers ofautonomic activity controlling heart rate suggested,during resting conditions after training, a simulta-neous reduction of sympathetic modulation andincrease of vagal modulation, as LF was reduced andHF was increased.

Thus, a readjustment of sympathovagal balance,with an increase in the gain of baroreceptor reflexmechanisms and the attendant enhancement of vagalmodulation, appears, at rest, to be the major adap-tive change in neural regulation produced by phys-ical training. As to spectral markers of sympatheticactivation, LF components of RR and SAP andDAP variabilities during exercise were markedlyincreased in the fit state. This finding should beinterpreted within the context of the substantial roleplayed by the sympathetic nervous system in increas-ing cardiovascular performance with training.44-45



AUTONOMIC CHANGES WITH PHYSICAL TRAINING/Pagani et al. 609

AppendixThe original ECG and arteriai blood pressure signals

are first digitized as described in Subjects and Methods,and the discrete series t{ and s, are obtained automatically,where t, is the RR duration and s> is the value of systolicblood pressure relative to the /-th cardiac cycle. Themodel of the signal generation mechanism is supposed tobe a linear, time-invariant system, which "colors" withrepeatable features a white (completely random) noise.For example, the value of tt series at the i-th cardiac cycle is

t(i) = a,t(i-1) + a2t(i-2) + . . . a^t(i-N) + w(i)That is, it is a linear combination of the t values in the Npreceding cycles through TV constant coefficients(a i . . . aN) and the random value w(0, which is the i-thsample of the white noise w(i) and is characterized by anull mean value and A2 as a variance. As alreadydescribed" it is possible to verify whether the startingautoregressive (AR) modeling hypothesis is satisfied ornot by the actual signal. To this end, a test of whiteness ofthe prediction error is performed. The autocorrelationfunction of the prediction error (defined as the differencebetween the values of the identified model and the actualones) is calculated and Anderson's test34 is applied (con-fidence limit of 5%). The calculation of (a, . . . aN) coef-ficients and of A2 is performed by a least squares method(Levinson-Durbin recursive algorithm).46 All the nonran-dom information of the signal is therefore contained inthese parameters and the residual information is whitenoise.

Parametric auto- and cross-spectral analysis of thevariability signals47 may be used to estimate some param-eters in the closed loop model depicted in Figure 3. Themodel reported, which is derived from a control theoryapproach, cannot be solved without additional a prioriinformation. Therefore, modifying further a recent sug-gestion by Akselrod et al.,19 we have introduced thehypothesis that nx = 0 (i.e., all the external disturbancesenter the loop only at s level), as described in Subjectsand Methods. Thus, the relationship between the powerdensity spectra (P) of input/output signals of the blockreported in Figure 3 is

Pt = \Hu\2-P,

Hence, the gain of the transfer function is expressed by

where //„, P, and P, are functions of/. Hu is computedonly at those bands that present a high squared coherencevalue (>0.5), which usually occurs at LF and HF bands.The program of power spectral density estimation of t ands (including cross spectrum and the squared coherenceK2) computes the bivariate identification coefficients andthe matrix of the bivariate values of A2 (variance of theprediction error).

A spectral decomposition of AR processes (describedin Reference 47) provides the LF and HF components ofP, and P%. From these data, the gain | Hu\ in the two bandsis obtained with

In the present, article the gain \HU\ is indicated by theindex a (i.e., a u and amd-

In summary, the calculation of a values requires thefollowing steps: 1) calculation of the AR power spectrum

and coherence of t and s; 2) automatic determination ofLF and HF components through AR spectral decompo-sition; and 3) calculation of the square root ratio of P, andP, in two major bands, provided that coherence is morethan 0.5.

References1. Seals DR, Hagberg JM. The effect of exercise training on

human hypertension: a review. Med Sci Sports Exerc 1984;16:207-215

2. Somers VK. The effects of exercise and physical training oncardiovascular control [Dissertation]. Oxford: Oxford Uni-versity, 1986. 166 p

3. Sever PS, Birch M, Osikowska B, Tunbridge RDG. Plasmanoradrenaline in essential hypertension. Lancet 1977;l:1078-1081

4. Wallin BG, Sundlof G, Strdmgren E, Aberg H. Sympatheticoutflow to muscles during treatment of hypertension withmetoprolol. Hypertension 1984;6:557-562

5. Smyth HS, Sleight P, Pickering GW. A quantitative methodof assessing baroreflex sensitivity. Circ Res 1969^4:109-121

6. Sleight P. Disorders of neural control of the cardiovascularsystem: clinical implications of cardiovascular reflexes. In:Zanchetti A, Tarazi RC, eds. Pathophysiology of hyperten-sion. Amsterdam: Elsevier, 1986:82-95 (Handbook of hyper-tension; vol 8)

7. Stegemann J, Busert A, Brock D. Influence of fitness onblood pressure control system in man. Aerospace Med 1974;45:45-48

8. Gribbin R, Pickering TG, Sleight P, Peto R. The effect of ageand high blood pressure on baroreflex sensitivity in man.Circ Res 1971^9:424-431

9. Cerutti S, Baselli G, Civardi S, et al. Spectral analysis ofheart rate and arterial blood pressure variability signals forphysiological and clinical purposes. Computers in cardiol-ogy. Washington, DC: IEEE Computer Society Press,1986:435-438

10. Robbe HWJ, Mulder LJM, Ruddel H, Langewitz WA,Veldman JBP, Mulder G. Assessment of baroreceptor reflexsensitivity by means of spectral analysis. Hypertension1987;10:538-543

11. Pagani M, Lombardi F, Guzzetti S, et al. Power spectralanalysis of heart rate and arterial pressure variabilities as amarker of sympatho-vagal interaction in man and consciousdog. Circ Res 1986^9:178-193

12. Lombardi F, Sandrone G, Pernpruner S, et al. Heart ratevariability as an index of sympatho-vagal interaction afteracute myocardial infarction. Am J Cardiol 1987;60:1239-1245

13. Akselrod S, Gordon D, Andrew UF, Shannon DC, BargerCA, Cohen RJ. Power spectrum analysis of heart ratefluctuation: a quantitative probe of beat-to-beat cardiovas-cular control. Science 1981;213:220-222

14. Pomeranz B, Macaulay RJB, Caudill MA, et al. Assessmentof autonomic function in humans by heart rate spectralanalysis. Am J Physiol 1985;248:H151-H153

15. Physical Fitness: two series of exercises developed by theRoyal Canadian Air Force. Markham, Ontario, Canada:Penguin Books, 1960. 75 p

16. Pagani M, Furlan R, Dell'Orto S, et al. Continuous recordingof direct high fidelity arterial pressure and electrocardiogramin ambulant patients. Cardiovasc Res 1986;20:384-388

17. de Trafford JC, Lafferty K, Kitney RI, Cotton, LT, RobertsVC. Modelling of the human vasomotor control system andits application to the investigation of arterial disease. IEEProc 1987;129:646-650

18. Malliani A, Pagani M, Lombardi F. Positive feed-backreflexes. In: Zanchetti A, Tarazi RC, eds. Pathophysiologyof hypertension. Amsterdam: Elsevier, 1986:69-81 (Hand-book of hypertension; vol 8)

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




20. Baselli G, Cerutti S, Civardi S, Rizzo G. Toward a modeli-sation of heart rate and blood pressure interaction throughparametric bivariate analysis. Computers in cardiology. Wash-ington, DC: IEEE Computer Society Press, 1986:431-434

21. Armitage P. Statistical methods in medical research. Lon-don: Blackwell, 1971. 504 p

22. Delia Rovere MT, Specchia G, Mazzoleni C, Mortara A,Schwartz PJ. Baroreflex sensitivity in post-myocardial inf-arction patients: correlation with physical training and prog-nosis [Abstract 2051]. Circulation 1986;74{suppl II):514

23. Billman GE, Schwartz PJ, Stone LH. The effects of dailyexercise on susceptibility to sudden cardiac death. Circula-tion 1984;69:1182-1189

24. Mancia G, Ferrari A, Gregorini L, et al. Blood pressure andheart rate variabilities in nonnotensive and hypertensivehuman beings. Circ Res 1983^3:96-104

25. Lindblad LE, Bevegard S, Castenfors J, Transjo J. Circula-tory effects of carotid sinus stimulation and changes in bloodvolume distribution in hypertensive man. Acta Physiol Scand1981;lll:299-306

26. Vatner SF, Pagani M. Cardiovascular adjustments to exer-cise: hemodynamics and mechanisms. Prog Cardiovasc Dis1976; 19:91-108

27. Pagani M, Pizzinelli P, Bergamaschi M, Malliani A. Apositive feedback sympathetic pressor reflex during stretchof the thoracic aorta in conscious dogs. Circ Res 1982;50:125-132

28. Gnecchi Ruscone T, Lombardi F, Mulfatto G, Malliani A.Attenuation of baroreceptive mechanisms by cardiovascularsympathetic afferent fibers. Am J Physiol 1987;253:H787-H791

29. De Boer RW, Karemaker JM, Van Montfrans GA. Determi-nation of baroreflex sensitivity by spectral analysis of spon-taneous blood pressure and heart rate fluctuations in man.In: Lown B, Malliani A, Prosdocimi M, eds. Neural mech-anisms and cardiovascular disease. Padova: Liviana Press(Fidia Research Series), 1986:303-315

30. Bertinieri G, Di Rienzo M., Cavallazzi A, Ferrari AU,Pedotti A, Mancia G. A new approach to analysis of thearterial baroreflex. J Hypertens 1985;3(suppl 3):s79-s81

31. Smith SA, Stallard TJ, Littler WA. Estimation of sinoaorticbaroreceptor heart rate reflex sensitivity and latency in man:a new microcomputer assisted method of analysis. Cardio-vasc Res 1986;20:877-882

32. Conway J, Boon N, Davies C, Vann Jones J, Sleight P.Neural and humoral mechanisms involved in blood pressurevariability. J Hypertens 1984^:203-208

33. Chess GF, Tarn RMK, Calaresu FR. Influence of cardiacneural inputs on rhythmic variations of heart period in thecat. Am J Physiol 1975;228:775-780

34. Rimoldi O, Pagani M, Sandrone G, Pierini S, Songini MG,Malliani A. Moderate hypotension in the assessment of theneural determinants of heart rate variability in consciousdogs [Abstract 985]. Proceedings of the 10th World Congressof Cardiology, Washington DC, 1986:172

35. Penaz J. Mayer waves: history and methodology. Automed-ica 1978;2:135-141

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

37. Pagani M, Furlan R, DeH'Orto S, et al. Simultaneous analy-sis of beat by beat systemic arterial pressure and heart ratevariability in ambulatory patients. J Hypertens 1985;3(suppl3):S83-S85

38. Furlan R, Dell'Orto S, Crivellaro W, et al. Effects of tilt andtreadmill exercise on systolic arterial pressure short-termvariability in hypertensive men. J Hypertens 1987;5(suppl 5):S423-S425

39. Ordway GA, Charles JB, Randall DC, Billman GE, Wek-stein DR. Heart rate adaption to exercise training in cardiacdenervated dogs. J Appl Physiol 1982;52:1586-1590

40. DeSchryver C, Mertens-Strythagen J. Heart tissue acetyl-choline in chronically exercised rats. Experientia 1975;31:316-318

41. Cousineau D, Ferguson RJ, de Champlain J, Gauthier P,Cdte P, Bourassa M. Catecholamines in coronary sinusduring exercise in man before and after training. J ApplPhysiol 1977;43:801-806

42. Wiegman DL, Harris PD, Joshua IG, Miller FN. Decreasedvascular sensitivity to norepinephrine following exercisetraining. J Appl Physiol 1981;51:282-287

43. Katona PG, McLean M, Dighton DH, Guz A. Sympatheticand parasympathetic cardiac control in athletes and nonath-letes at rest. J Appl Physiol 1982^2:1652-1657

44. Sable DL, Brammell HL, Sheehan MW, Nies AS, Gerber J,Horwitz LD. Attenuation of exercise conditioning by beta-adrenergic blockade. Circulation 1982;65:679-684

45. Liang CS, Tuttle RR, Hood WB, Gavras H. Conditioningeffects of chronic infusions of dobutamine. J Clin Invest1979;64:613-619

46. Durbin J. The fitting of time series model. Rev Int Inst Stat1960;28:233-244

47. Baselli G, Cerutti S, Civardi S, et al. Spectral and cross-spectral analysis of heart rate and arterial pressure variabil-ity signals. Comput Biomed Res 1986;19:520-534



1988;12;600-610 - lifemedicalcontrol.com · sure (sap) variabilities. as explained in the text,...

Documents