Kidney International, Vol. 47 (1995), pp. 1443—1452

Effect of hemodialysis on cardiovascular rhythms in end-stagerenal failure

FRANK WEISE, GERARD M. LONDON, BRUNO M. PANNIER, ALAINP. GuERIN, and JEAN-LUC ELGHOZI

Department of Phamzacology at the Facuhy of Medicine IVecker, Paris, and Hospital Manhes, Fleu,y Merogis, France

Effect of hemodiaysis on cardiovascular rhythms in end-stage renalfailure. The purpose of this investigation was to determine non-invasivelythe alteration in autonomic cardiovascular control observed in end-stagerenal disease (ESRD) patients submitted to a hemodialysis (HD) treat-ment. The effect of HD on finger blood pressure (BP) and heart rate (HR)variability was studied by means of spectral analysis in 12 ESRD patientsat supine and during 450 head-up tilt. Amplitude spectra of BP and HRrhythinicity were estimated and integrated amplitudes of the low (60 to140 mHz, Mayer waves) and high frequency (area under the curve at meanrespiration rate 50 mHz) components were computed. The overallvariability of a signal was given by the total area under the curve of thespectrum between 20 and 500 mHz. The transfer function was calculatedbetween systolic BP and HR fluctuations using cross-spectral analysis. Theprinciple findings are as follows: (I) Before HD, overall variability insystolic and diastolic BP as well as in the low frequency component in BPspectra were markedly reduced in ESRD patients compared to controlsubjects. Dialysis produced an immediate improvement in overall BPvariability and the LF BP rhythmicity. (2) Before HD, there was analteration of the HR spectral profile in uremic patients. (3) Chronicuremia and HD induced no changes in the transfer gain characteristic ofmodulation of HR by systolic BP. Taken together, these findings suggestthat volume depletion in ESRD patients during a dialysis treatment isresponsible for the improvement of overall BP variability and its spectralcomponents.

Previous studies have shown a high prevalence of impairedcardiovascular autonomic nervous control in patients with end-stage renal disease (ESRD) 1—6]. Procedures based on bloodpressure (BP) and heart rate (HR) responses to standard stimulihave been used to localize the defect in the autonomic nervoussystem. It has been reported that lesions may occur in both theafferent and the efferent limbs of the autonomic nervous system inchronic uremia [1, 3, 41.

Power spectrum analysis of periodic heart rate (HR) and bloodpressure (BP) fluctuations has become a widely accepted indirectand non-invasive tool of assessing the autonomic control ofcardiovascular function during different physiological perturba-tions or disease states [7—12]. The method of spectral analysis isutilized for the detection and quantitative description of period-icities in cardiovascular fluctuations by deconvolution of theoriginal time series into sinusoidal functions of different frequen-cies. Three major components are usually considered: one is at

Received for publication December 21, 1993and in revised form December 21, 1994Accepted for publication December 22, 1994

© 1995 by the International Society of Nephrology

respiration frequency (around 250 mHz), another around 100mHz, and a third in a very low frequency region around 30 mHz[13]. In HR spectra, the high frequency (HF) component is linkedto respiration and mediated predominantly by cardiac vagalactivity, while the low frequency (LF) and very low frequencycomponents are mediated by sympathetic, parasympathetic andrenin-angiotensin system activity [7—10, 14]. In arterial BP spectra,the HF component is mainly related to mechanical effects ofrespiration on cardiovascular control [15]. The periodic fluctua-tions in systolic BP around 100 mHz (so-called 10-s-rhythm orMayer waves) seem to be a marker of sympathetic modulation [9].

Using spectral analysis methods it has been shown that a singlehemodialysis (HD) in ESRD patients restores the LF componentin HR [16] but not in BP rhythmicity [17]. This led to thesuggestion that toxic substances which are removed by HD may beresponsible for a functional cardiac autonomic rieuropathy inthese patients. Other factors besides "uremic toxicity" are anemiaand overhydration which are related to renal insufflency [18].Recently we found that acute volume expansion of the intratho-racic compartment of the circulation by head-down tilt is able toreduce the overall HR and BP variability [19]. Furthermore, in apilot study (Fig. 1) we observed a normalization of the orthostaticHR response in ESRD patients only when the HD was combinedwith significant ultrafiltration. According to these data the pur-pose of our present study was, therefore, to determine whetherHD in ESRD patients provokes acute changes in autonomiccardiovascular control and in the complex systolic BP-HR rela-tionship, which could be due to the volume depletion during HD.

Methods

Subjects

Twelve ESRD patients (7 men/5 women) undergoing mainte-nance hemodialysis therapy were studied. The age of the patientswas 46 (33 to 56) years (median, 25% to 75% percentiles) with arange from 25 to 58 years. The duration of dialysis therapy was 70(22 to 187) months and ranged from 3 to 240 months. The heightof the patients was 162 (155 to 170) cm and the body weightbefore HD was 57.3 (51.1 to 64.6) kg. The causes of the chronicrenal failure in the patient group were chronic glomerulonephritis(5 patients), interstitial nephritis (4 patients), polycystic kidneydisease (2 patients), angioscierotic nephropathy (1 patient). Nopatient had a past history of ischemic heart disease, cerebrovas-cular disease, diabetes or hyperlipidemia or had undergone

1443

1444 Weise et al: HD and cardiovascular rhythms

NS ESRD ESRD ESRDBefore D After D After D+UF

Fig. 1. Heart rate response from supine to active standing in normal subjects(NS, N = 42, qge = 42 10 years) and end-stage renal disease patients(ESRD, N = 31, age = 45 12years) before dialysis (D), after dialysis andultrafiltration (UF, 3.1 0.8 liter), and after dialysis without ultrafiltration(0.1 0.1 liter). Values are mean SD.

bilateral nephrectomy. No one showed signs and symptoms of anautonomic or peripheral neuropathy [20]. No patient was takingany antihypertensive therapy or vasoactive drugs. Four out of 12patients were mildly hypertensive (BP > 140/90 mm Hg) and 4patients were mildly hypotensive (BP < 100/60 mm Hg) in thebefore-HD state in the supine position. The patients underwentdialysis therapy three times per week and the duration of dialysiswas individually adjusted (3 to 5 hr) to control body fluids and tomaintain the predialysis serum urea at less than 30 mmol/liter,serum creatinine at less than 1.1 mmol/liter and serum electrolyteswithin the normal range. Dialysis was performed using dialyzerswhose surface areas (1.2 to 2.1 m2) were matched to the patient'sbody surface area. Dialysate water was treated with reverseosmosis. The dialysate was delivered by a system which includedbicarbonate delivery and ultrafiltration devices. An internal arte-riovenous fistula was used in all patients. Iron supplements wereroutinely prescribed.

Twelve healthy volunteers (7 men, 5women) aged 42 (36 to 49)years with an age range from 23 to 56 years were studied ascontrols. The height of the controls was 175 (165 to 180) cm andthe body weight was 74 (61 to 80) kg. All control subjects werenormotensive, none was receiving any medical treatment, andnone had any neurologic, cardiac or metabolic diseases.

Informed consent for investigation was obtained from eachsubject, after a detailed description of the procedures had beengiven. The protocol of the study was approved by our institutionalreview board.

Apparatus and experimental protocol

Investigations were performed in a controlled environmentkept at 22 1°C. Subjects were lying on a electronic tilt table.Finger arterial BP and pulse intervals [transformed in HR (beatsper minute, bpm) automatically] were measured by a FINAPRESdevice (model 2300, Ohmeda, Trappes, France). In control sub-jects, the cuff was placed at the third finger of the right hand, whilein patients the cuff was wrapped around the middle or ring fingerof the hand contralateral to the arm in which the arteriovenousshunt was placed. The cuffed finger was kept at the heart levelduring the experiments. The analog output from the FINAPRESwas connected to an analog-to-digital (A/D) converter (AS 1,Notocord Systems, Igny, France) to permit data acquisition,storage, and analysis using a Dynamit computer (PC-AT 286).

The subjects rested in a supine position with recording trans-ducers attached for an adaptation period of 15 minutes. At theend of this period, finger arterial BP and HR were recorded for 10minutes at the supine position. Once these measurements werefinished, the examination table was tilted up, leading to 45°head-up tilt (HUT). In no case was discomfort, dizziness, orsyncopes observed. After two minutes of HUT arterial BP andHR were recorded for 10 minutes. In ESRD patients, this procedurewas performed immediately before and after hemodialysis.

Signal processing and data analysis of HR and BP recordings

The BP signal was digitized using a 12-bit A/D converter at arate of 500 Hz, and processed by an algorithm based on featureextraction to detect and measure the characteristics of an arterialpressure cycle with its maximum in a 0.2-s window. Systolic, meanand diastolic BP and HR, calculated as 60,000/heart period, werestored. The evenly spaced (equidistant) sampling permitted directspectral analysis of each distribution using a Fast Fourier Trans-form algorithm on a 1024-point stationary time series (Anapres3.0, Notocord Systems, Igny, France). This corresponded to a 3minute and 25 second period at this 5 Hz sampling rate. Allspectra are reported as amplitude spectra. The amplitude (mod-ulus) of the HR and BP spectrum (ordinates) had units of[bpm/HzU'2] and [mmHg/Hz112}, respectively. The frequency of theoscillation scale (abscissa) was analyzed up to 500 mHz. Theintegrated amplitude was calculated over the frequency band 20 to500 mHz (total area under the curve, T-AUC, without any directcurrent component), as well as in the LF (60 to 140 mHz) and inthe HF band. The HF band corresponds to the respiration, whichis usually observed between 200 and 350 mHz in the HR spectrum[13]. The respiratory frequency was defined as the frequency atthe maximal spectral peak in this frequency region in the HRspectrum and controlled by counting the breathing rate of thesubjects during the experiment. The integrated amplitude for therespiration related frequency band was calculated at the respira-tory peak amplitude 50 mHz in HR and BP spectra to obtain anarrow frequency band.

The mean of each variable, the T-AUC (a measure of overallvariability) and the absolute and relative [100 multiplied withabsolute valueiT-AUC in (%)j spectral density estimates in thetwo frequency bands of interest in systolic and diastolic BP andHR were calculated in controls and ESRD patients for each bodyposture.

The transfer function analysis was utilized to assess the rela-tionship between spontaneous systolic BP and HR fluctuations inthe frequency domain. The calculation of the transfer function

P<0.05

P<0.0001P<0.0001

25

20

E 15

ci)0:: 10a)I

5

0

—5

>()<>Cx><x

Weise et a!: HD and cardiovascular rhythms 1445

from systolic BP to HR was based on the auto- and cross-spectraltechnique with a five-point triangular window for smoothing in thefrequency domain [211. The transfer function provides informa-tion about magnitude (or gain) and phase relation of both signalsin the frequency range considered. The gain function defines theratio between changes in HR (output) and changes in systolic BP(input) in [bpm/mm Hg] at any given frequency. The gain of theLF band is considered to be a non-invasive index of the arterialbaroreflex control of the heart [22, 23]. The phase functionspecifies the phase difference (lead or lag) between the two signalsin the frequency domain. The gain and phase relationship can beestimated reliably only when the coherence is high. The (squared)coherence function quantifies the amount of linear couplingbetween two time series at any given frequency and is comparableto the (squared) correlation coefficient in the time domain. Thecoherence may have values between 0 and 1. The gain and phasevalues for the LF band were estimated at that frequency where thecoherence was maximal (> 0.5) in this frequency region [24].

Statistical analysis

Results are expressed as medians and 25% and 75% percen-tiles. The absolute values of the LF and HF components in BP andHR spectra were log-transformated in order to reduce the largeinterindividual variability. Comparisons of measurements be-tween supine and HUT in ESRD patients before and afterhemodialysis were performed using non-parametric analysis ofvariance for repeated measures followed by the Wilcoxon signedrank test. The Wilcoxon test was also used to assess differences inthe values between supine and tilt conditions in control subjects.Values obtained in controls and ESRD patients were comparedusing the Mann-Whitney U-test. Spearman's rank correlationcoefficients were calculated to evaluate the relation betweenvariables. A probability level < 0.05 was considered statisticallysignificant.

Results

Subjects

ESRD patients had a lower weight (P < 0.05) and were smallerin stature (P < 0.05) than controls. After HD, the weight of thepatients decreased from 57.3 (51.1 to 64.6) kg to 55,6 (49.0to 61.8) kg (P < 0.01), while the hematocrit increased from31.5 (29.0 to 37.5)% to 35.0 (31.0 to 43.5)% (P < 0.05). Thecalculated relative blood volume change (1 hematocritl,efore I ID'hematocritafter HD in [%]) was —11.8 (—16.7 to —2.8)% during HD.

Responses of averages and overall variabilities

As illustrated in Figure 2, arterial BP and HR always showed aspontaneous variability around their average values in healthysubjects. Table I presents the average HR, systolic and diastolicBP, and respiratory frequency in controls and ESRD patients atthe supine and during HUT positions. Patients after HD revealeda significant increase in average HR in the supine and tiltedposition compared to the before-HD state and to the controls. Asshown in Figure 3, the average HR increased significantly withtilting in both control subjects and ESRD patients; however, thisincrease was more marked in controls and in the after-HD state ofthe patients than in the before-HD state. There was no correlationbetween the augmented orthostatic HR response after HD andthe calculated blood volume change (Spearman's rank correlationcoefficient = 0.155; P = 0.61). In ESRD patients before HD, a

significant increase in systolic BP was noted during tilt (Table 1).This average systolic BP value at tilt was higher than the value inthe after-HD state and in the control groups. In both groupsstudied, the average diastolic BP increased with tilting. Patientsbefore HD showed a significantly higher diastolic BP at tilt thanafter HD.

As shown in Table 2, the overall variability in HR, and systolicand diastolic BP was attenuated in ESRD patients before HDcompared to controls. After HD, the overall variability in BP wassignificantly elevated and was no longer different than thatobserved in the control group. The HR variability showed a strongtrend to higher values in patients after HD (Table 2). In controls,the overall variability of systolic and diastolic BP increasedsignificantly with the change from supine to tilt. In patients afterHD, a significant rise in diastolic BP variability was also foundduring HUT.

Response of amplitude spectra

The spectral analysis of BP and HR variability provided aquantitative assessment of their rhythmic components. Figure 2depicts the original HR and BP time series at supine and tilting inone control subject, and shows the principal spectral profilesaccompanying these experimental conditions. With a change inthe position from supine to 45° HUT, marked increments in theamplitude of the 10 second rhythmicity in HR, and systolic anddiastolic BP time series were seen, which is illustrated by aprominent peak in the LF region of their amplitude spectra.Figure 4 presents the spectral pattern of BP and HR signalsduring orthostatic load in one ESRD patient before and after HD.

In the before-HD state, the absolute LF component wasreduced compared to controls in both HR and BP spectra atsupine and tilt (Table 2). The relative contribution of this LFcomponent to the overall variability was especially diminished inthe BP signal independent of the body posture (Table 2). Theabsolute HF component in HR spectra was significantly lower inpatients before HD than in controls at supine and tilt, while therelative value of the HF band showed a reduction only at thesupine position.

In the after-HD state, the absolute LF component in HR andsystolic BP spectra was no longer different than values obtained incontrol subjects at supine. During tilt, systolic and diastolic BPvariabilities at LF were actually reduced compared to controls, butwere significantly increased versus the before-HD state (Table 2).The HF component in HR spectra was reduced in comparison tocontrols in absolute and relative units during HUT.

There were no significant correlations between spectral densityestimates in HR and BP signals and the duration of HD or thecalculated blood volume change during HD.

Response of transfer function between systolic BP and HR

A high correlation (coherence > 0.5) was found betweenperiodic systolic BP and HR fluctuations around 100 mHz in allcontrol subjects and in 9 out of 12 ESRD patients (Table 3). Thecoherence in controls increased from supine to upright and wassignificantly higher than in ESRD patients before HD in bothbody positions (Table 3).

Figure 5 depicts the coherence, gain and phase functionsobtained from one healthy volunteer during the supine andupright positions. The gain diagram indicates that the transfer-magnitude for the LF band was lower during HUT than at supine.

1446 Weise et al: HD and cardiovascular rhythms

I.-

100

100EEa:a

ha-

EE

0

Supine Tilt

This reflects the systematic diminution of the transfer magnitudeafter changing the body position from supine to upright obtainedin all controls as well as in ESRD patients before and after I-ID

(Table 3). There was no significant difference in the transfermagnitude between controls and patients in either posture.

The phase angle between systolic BP and HR fluctuations in the

0 205 0Time, seconds

205Time, seconds

0

Frequency, mHz0 500

0

250

I-t,LaE

0

140

40

250

QNa-

0

0 500Frequency, mHz

100

0

250

0

140

40

250

0

100

0

250

0 205 0 205Time, secondsTime, seconds

Frequency, mHz0 500 0

Frequency, mHz500

0 205Time, seconds

0

250

0

205

Time, seconds

Frequency, mHz

0__ 500 0 500Frequency, mHz

Fig. 2. Tracings are 205 second samples ofrepresentative heart rate (HR), systolic bloodpressure (SBP), and diastolic blood pressure(DBP) time series obtained from one controlsubject at supine and during 450 head-upposition. The corresponding amplitude spectraof HR, SBP and DBP signals are shown. Notethe marked increase in low frequencyrhythmicity in all time series during tilt. Bpm isbeats per minute.

Weise et al: HD and cardiovascular rhythms 1447

Table 1. Mean and overall variability of heart rate and blood pressure and mean respiratory frequency in controls and end-stage renal diseasepatients

ESRD patients

The decrease in blood volume by HD (around 10%) elicited anincrease in HR at rest and during sympathetic activation (tilt) andsignificantly augmented the HR response to tilt in patients afterHD compared to the before-HD state. The orthostatic HRresponse is considered to be a result of a potent cardiovascularstress which demands a change in autonomic balance from a vagal(supine) to a more sympathetic efferent activity (tilt) in order tomaintain the cardiovascular homeostasis [25]. The normal ortho-static HR response we found after HD suggests no importantneuropathic alteration of the efferent cardiac sympathetic nervesin the patient group studied. The enhanced arterial BP reaction totilt was also changed by HD. In a pilot study we found (Fig. 1) thatHD without ultrafiltration produced only a small change in

Fig. 3. Comparison of orthostatic heart rate response (delta HR, HRdifference supine to 45° head-up tilt) between 12 healthy subjects and 12patients with end-stage renal disease (ESRD) before and after hemodialysis(LID). Data are median (25% to 75% percentiles); bpm is beats perminute.

orthostatic HR response compared to the before-HD state incontrast to HD with ultrafiltration. Therefore, it is fair to assumethat the behavior of hemodynamic data before dialysis is second-ary to the overhydration caused by the underlying disease process[26, 27].

Variable

HR

Controls Before HD

Supine Tilt Supine Tilt

After HD

Supine Tilt

Median 64 76b 75 78b 82d,f 89h,d,f

bpm 61—71 69—79 62—80 67—84 76—99 83—113T-AUC 1535 1649 1106 1021° 1406 1553

bpm/Hz"2 1257—2676 1302—2818 879—1815 647—1638 1024—1751 1017—1817SBP

Median 115 110 119 131"" 104 114"mm Hg 107—120 106—118 100—142 124—144 95—126 85—135T-AUC 2134 2624" 1641° 1700d 2170° 2640fmm Hg/Hz"2 1893—2793 2141—3198 1281—2049 1369—2213 1368—2730 1551—2894

DBPMedian 60 63" 63 72" 53 62""

mm Hg 55—63 61—69 56—73 62—90 49—68 53—73T-AUC 1164 1678" 714" 747d 1035° 1186a,fmm Hg/Hz"2 1025—1271 1125—1862 639—871 704—1156 843—1239 865—1638

RFMedian mHz 230

200—292225

200—283262

250—300257

250—300280

252—315285

252—310

Values are medians (25% to 75% percentiles) for 12 control and 12 end-stage renal disease (ESRD) subjects. Abbreviations are T-AUC, total areaunder the curve of blood pressure and heart rate amplitude spectra, respectively, in the frequency band between 20 and 500 mHz; bpm, beats per minute;HD, hemodialysis; HR. heart rate; SBP, systolic blood pressure; DBP, diastolic blood pressure; RF, respiratory frequency.° P < 0.05 supine vs. tilt

"P < 0.01 supine vs. tilt° P < 0.05 control vs. ESRD"P < 0.01 control vs. ESRD° P < 0.05 before HD vs. after HDP < 0.01 before HD vs. after HD

LF band was not significantly modified by HUT in both groups(Fig. 5, Table 3). At supine, the phase angle was significantly lowerin ESRD patients before and after HD compared to controlsubjects.

Discussion

This study shows that: (1) HD with ultrafiltration provides animmediate improvement of the overall systolic and diastolic BPvariability; (2) the increment in overall BP variability is mainly dueto the restoration in the amplitude of the LF component; and (3)chronic uremia and HD induce no changes in the transfer gaincharacteristics of modulation of HR by systolic BP. We suggestthat the improvement of overall cardiovascular variability by HDis related to the correction of overhydration in the patients duringa dialysis treatment.

Autonomic cardiovascular control in ESRD patients before andafter HD

20

Before HO After HD

Controls ESRO patients

Median 9 5 1225%—75%

(7—15) (1—7) (8—16)percenlilos

1448 Weise et al: HD and cardiovascular rhythms

Table 2. Spectral components of heart rate and blood pressure variability in control subjects and in end-stage renal disease patients

The overall HR variability tended strongly to be lower beforethan after HD and was significantly reduced in the before-HDstate at tilt compared to controls. Spectral analysis of HR timeseries showed a mild alteration of the rhythmic components ofHRV in ESRD patients before HD. Reductions in short-termvariability and rhythmicity of HR have already been described inchronic renal failure and were connected with a cardiac auto-nomic defect [4, 16, 28—32]. Previous studies in humans demon-strated that the HF component in HR is mediated by the variationof the efferent vagus nerve activity, while the LF component ismediated by both the variations of vagal and sympathetic nerveactivity to the sinoatrial node [8, 10, 33, 341. In contrast, ourfindings are not related to a decreased parasympathetic nervoussystem input into the heart in ESRD patients before HD, but tothe hypervolemic state and/or "toxic agents." After HD weobtained nearly normal HR spectral values and an intact barore-

flex control of HR. Also, it seems unlikely that a damage of thesympathetic cardiac innervation is behind the reduced LF com-ponent (especially during sympathetic activation at tilt [35]),because dialysis induced a prompt rise in HR and a normal HRresponse to orthostatic stress. In a previous study, Forsström et al[16] reported a significant increase in overall HR variability and inrelative HR power around 100 mHz by HD. They suggested thattoxic agents affecting the autonomic cardiac innervation areremoved by dialysis. Our data in HR variability and rhythmicitytended in the same direction without reaching significance.

The increment of the overall BP variability by HD was a robustfinding in the patient group studied. The reduced overall variabil-ity in systolic as well as diastolic BP in the before-HD state ismainly due to the alteration of the LF component in both bodypostures. While the LF component in systolic BP is essentiallyrestored by HD, the LF component in diastolic BP was partly

Variable

Controls

ESRD patientsBefore HD After HD

Supine Tilt Supine Tilt Supine Tilt

HRLF component

Absolute 2.63 279" 2.39c 225d 2.46 2.49c

Relative(2.48—2.79)

25

(21—32)

(2.55—2.94)31b

(26—40)

(2.12—2.59)22

(17—27)

(2.06—2.51)20d

(17—24)

(2.28—2.66)23

(16—28)

(2.30—2.72)23C

(17—28)HF component

Absolute 2.66(2.57—2.97)

2.60(2.48—2.92)

2.48c(2.29—2.64)

2.34c

(2.16—2.60)

2.62(2.34—2.66)

2.44"(2.25—2.61)

Relative 29(26—35)

27(23—33)

26"(21—30)

23(20—27)

26(19—31)

23C

(18—24)

SBPLF component

Absolute 2.80 2.99" 252" 260a,d 2.70 2.88"""

Relative(2.73—2.88)

28(25—31)

(2.85—3.05)33"

(29—39)

(2.42—2.70)24'

(18—30)

(2.44—2.76)24d

(21—31)

(2.49—2.90)25

(21—29)

(2.63—2.98)30"

(24—35)HF component

Absolute 2.65(2.48—2.78)

2.80a

(2.58—2.89)

2.55(2.33—2.64)

2.51"(2.44—2.66)

2.54(2.34—2.82)

2.68(2.48—2.74)

Relative 21

(19—23)

24(17—28)

22(18—24)

21(16—27)

18(15—27)

18(16—21)

DBPLF component

Absolute 2.64 2.82" 226" 235b.d 2.41"" 256b,c.

Relative(2.57—2.68)

36(30—39)

(2.67—2.90)40b

(35—43)

(2.18—2.41)26"

(21—31)

(2.26—2.55)29a,d

(25—34)

(2.34—2.55)27"

(22—34)

(2.37—2.84)31,c

(26—39)HF component

Absolute 2.14(2.06—2.36)

2.31(2.01—2.73)

2.10(2.01—2.28)

2.11(2.03—2.22)

2.28(2.06—2.45)

2.38(2.07—2.45)

Relative 15(11—20)

14(10—27)

16(14—20)

14

(13—18)

17(13—24)

17(13—21)

Values are medians (25% to 75% percentiles) for 12 control and 12 end-stage renal disease (ESRD) subjects. Abbreviations are: absolute,log-transformated absolute value of the area under the curve in [bpm/Hz"2] for heart rate amplitude spectra and in [mm Hg/Hz"2] for blood pressureamplitude spectra in the frequency bands considered; relative, 100 multiplied with absolute/total area under the curve (20 to 500 mHz) in (%); bpm,beats per minute; HD, hemodialysis; HR, heart rate; SBP, systolic blood pressure; DBP, diastolic blood pressure; LF, low frequency; HF, high frequency.

"P < 0.05 supine vs. tilt"P < 0.01 supine vs. tiltP < 0.05 control vs. ESRD

dP < 0.01 control vs. ESRD"P < 0.05 before HD vs. after HDP < 0.01 before HD vs. after HD

Weise et al: HD and cardiovascular rhythms 1449

Fig. 4. Amplitude spectra of heart rate (HR),systolic blood pressure (SBP), and diastolic bloodpressure (DBP) in one patient with end-stagerenal disease before and after hemodialysis (HD)at supine and during 450 head-up tilt. Opengraphs refer to the after-HD state, shadedgraphs refer to the before-HD state.

higher after than prior dialysis, but remained significantly lower LF component in systolic BP was found to be increased duringcompared to controls. It has been indicated that the LF compo- orthostasis [9], physical exercise [12], daytime compared to night-nent in BP spectra seems to be a marker of sympathetic nervous time [361,and transient coronary occlusion [37], and was observedmodulation of vascular tone. For example, the magnitude of this to be strongly depressed after a1-adrenergic blockade in conscious

250 Supine 250 Tilt

U)

0E(U

(U

0U)c1(I)0a

U)-D

EU)

U)

0C,)

a

00 Frequency, mHz

0500 0 Frequency, mHz 500

1450 Weise et al. HD and cardiovascular rhythms

Table 3. Transfer function response of systolic blood pressure to heart rate

Variable

Controls

ESRD patientsBefore HD After HD

Supine Tilt Supine TiltSupine Tilt

LF componentCoherence 0.83 0.90° 0.70° 0.69° 0.78 0.86

Gain(0.73—0.92)

0.99(0.80—0.95)

070b(0.63—0.75)

1.00(0.65—0.81)

0.49°(0.71—0.84)

0.91(0.71—0.91)

0.47k'

bpm/mm HgPhase

(0.57—1.30)2.32

(0.46—1.00)2.10

(0.49—1.14)1.74c

(0.28—0.53)1.89

(0.58—1.15)1.72"

(0.34—0.74)1.92

radians (1.82—2.88) (1.63—2.48) (1.60—2.09) (1.49—2.00) (1.34—1.74) (1.79—2.10)

Values are medians (25% to 75% percentiles) for 12 control subjects and 9 end-stage renal disease (ESRD) patients. Abbreviations are: LF, lowfrequency; bpm, beats per minute; HD, hemodialysis.

°P < 0.05 supine vs. tiltbP < 0.01 supine vs. tilt°P < 0.05 control vs. ESRDd P < 0.01 control vs. ESRD

dogs [37] or rats [38, 39] and absent in tetraplegic human subjects[401. According to these studies, the decrease in LF BP fluctua-tions (as well as in overall BP variability) may reflect a functionaldeterioration of the efferent sympathetic vasomotor control inESRD patients before HD. However, it has been reported thatthe sympathetic vasoconstrictor discharge to the skeletal musclecirculation was markedly increased in uremic patients with nativekidneys and not influenced by the expanded plasma volumebetween dialysis treatments [41]. The improvement of BP vari-ability by HD, therefore, seems to be unrelated to changes inefferent sympathetic nerve activity. We suggest that the effect ofHD with ultrafiltration on BP variability and rhythmicity is due to(central) blood volume alterations. Changes in the volume loadingof the cardiovascular system by HD produce changes in arterialgeometry which influence the reactivity of the arteries [421. Anenlarged arterial diameter in the overhydrated state before dial-ysis and a decreased arterial compliance [18] may increase thevariation in pressure because of altered function of arterialbaroreceptors. The consistently increased sympathetic nerve dis-charge could be effective in inducing trterial pressure variationsonly when the overhydration is removed and arterial vesseldiameter is reduced. The increase in BP variability by HD can beexpected as long as no peripheral sympathetic neuropathy isdeveloped. It has been shown in experimental animals and healthysubjects, respectively, that a volume depletion causes a significantsurge in LF BP fluctuations [43—45], while a volume loading of theintrathoracic compartment of the circulation produces a decreasein LF systolic BP fluctuations [19]. In pathophysiological stateslike uremia, many confounding variables may disguise the inter-pretation. For example, factors like a-adrenergic receptor sensi-tivity of vascular smooth muscles or toxic agents or the sodiumcontent in the dialysate may interfere with the volume depletioneffect [27, 46]. Recently, Munakata et al [30] also reported BPspectra in uremic patients, but the BP signal was recorded onceonly at the day prior to HD, So far as their results can becompared with the data in our patients before HD, the samedecreased BP fluctuations were found in the frequency bandaround 100 mHz. In our previous study [17] we did not observe asignificant immediate improvement in the amplitude of the LFblood pressure waves by HD. This could be due to the smallnumber of patients (N = 6) studied. Furthermore, these six

patients were treated with HD at the Hôpital Necker wherepatients are summarized who have some kind of difficulties intheir casual history. Also, these patients were not screened forsigns of peripheral and autonomic neuropathy.

Relation of systolic BP to HR

The transfer function analysis of spontaneous systolic BP andHR fluctuations was based on a closed-loop approach to cardio-vascular regulation [12, 14, 24, 451. The interaction betweensystolic BP and HR fluctuations is realized by feedback andfeedforward mechanisms [47]. For example, HR induces changesin BP fluctuations mainly through the mechanical properties ofthe cardiovascular system (feedforward), while BP induceschanges in HR fluctuations mainly by the baroreceptor reflex(feedback). All other sources of variability acting on HR or BPcan be summarized as noise. If we suppose that noise influencingHR fluctuations (especially in the frequency range of 100 mHz) isnegligible, a transfer function from systolic BP to HR can becomputed which fulfills the closed-loop model [12, 47]. In thiscase, the transfer magnitude of the relationship between systolicBP and HR fluctuations represents the baroreceptor-HR reflex(feedback) gain. The interpretation of our data followed thissimplified model. The transfer characteristics between systolic BPand HR were estimated for the LF band only. It was reported thatalteration of arterial baroreflexes by sinoaortic denervation affectsthe coupling between the two signals at the LF but not at the HFcomponent [471. This illustrates the fundamental contribution ofarterial baroreflexes to the systolic BP-HR relationship around100 mHz.

We found a high coherence at LF and a closer link betweenvariations in systolic BP and HR at least in controls during asympathetic activation (tilt). In ESRD patients before HD, theoverhydration seems to be a source of system nonlinearities whichcaused a loss of coherence (<0.5) in 3 out of 12 patients and asignificantly lower coherence in the remaining patient groupcompared to controls. This suggests that volume depletion andsympathetic effectiveness after HD, as estimated according to thecalculated blood volume change and orthostatic HR response,increased the coherence of the patient group and approached it tocontrol values.

The gain of the transfer function between systolic BP and HR

Weise et al: HD and cardiovascular rhythms 1451

Control

Supine—Tilt

Fig. 5. Transfer function analysis of systolic blood pressure to heart rate at asupine position and during 450 head-up tilt in the same subject as in Fig. 2.A high coherence exists for the relation between systolic blood pressureand heart rate in the low frequency (60 to 140 mHz) band. The gainfunction for the transfer from millimeter of mercury in systolic bloodpressure to beats per minute in heart rate is showing a decrease in the lowfrequency band during head-up tilt. The phase spectra depict positivevalues in the frequency band of interest indicating a phase lag of systolicblood pressure behind heart rate. Supine, —; tilt, - - -; bpm, beats perminute.

variations decreased with tilting in the groups studied. Thissupports the notion of a lower arterial baroreflex sensitivity duringupright posture [481. On the other hand, Wright, Levine andBlomqvist [49] have reported the opposite: an increased arterialbaroreflex gain while standing. Using a causal technique to theclosed-loop ielationship between systolic BP and HR variations,that is, autoregiessive moving average analysis, Triedman, Perrottand Saul [50] also observed an increased baroreflex feedback fromBP to HR during tilt. Further studies are required to clarify thesediscrepancies.

An unexpected finding was that uremic patients had a restingtransfer magnitude which was pomparable to controls. During tilt,the gain in patients before and after HD tended to be lower thanin controls. Recently, Munakata et a! [301 also found a normaltransfer gain around 100 mHz in ESRD patients at the day beforeHD. However, there have been a few reports using reflex slowing

of HR to a phenylephrine induced rise in BP as an index onbaroreceptor-HR reflex sensitivity demonstrating that baroreflexgain is reduced in uremic patients [51-53]. This decreased barore-flex gain seems to be related to the presence of hypertension anduremic polyneuropathy [53]. In our study no patient showed signsand symptoms of a peripheral neuropathy, and only four patientswere mildly hypertensive in the before-HD state at the supineposition. Therefore, our findings of a rather unaffected baroreflexgain in addition to a nearly normal HR rhythmicity after HDmight be interpreted that the parasympathetic cardiac innervationwas not affected in our chronic renal failure patients, suggestingan effective dialysis regimen. The immediate improvement inoverall cardiovascular variability after dialysis appears to berelated to the correction of hypervolemia rather than to theremoval of dialyzable toxins.

The transfer phase at LF provided constantly positive values incontrol subjects and ESRD patients, indicating a phase lag ofsystolic BP behind HR fluctuatioi'Is. Why the transfer phase issignificantly lower in resting patients before and after HD com-pared to controls remains unclear.

To summarize, in ESRD patients we studied the effect of HDon cardiovascular variability and rhythmicity and on the complexsystolic BP-HR relationship. Hemodialysis with ultrafiltrationprovides an immediate improvement of the overall BP variability.The LF component in systolic BP spectra is essentially restored bydialysis. Chronic uremia induces no changes in the transfer gainwhich reflects the modulation of HR by systolic BP. We suggestthat the improvement of overall BP variability and its spectralcomponents with dialysis is related to the volume depletion of thepatients during a dialysis treatment.

Acknowledgments

F. Weise was supported by the Association Claude Bernard. Theauthors are grateful to Dr. Phil Saul, Harvard University of Boston, forreviewing the manuscript.

Reprint requests to Frank Weis,. Laboratoire de Pharmacologie, Faculté deMédecine Necker, 156 me de Vaugirard, 75015 Paris, France.

References

1. LILLEY JJ, GOLDEN J, STONE RA: Adrenergic regulation of bloodpressure in chronic renal failure. J Clin Invest 57:1190—1200, 1976

2. RocIcEL A, HENNEMAN H, STERNAGEL-HAASE A, HE1DLAND A: Urae-

mic sympathetic neuropathy after haemodialysis and transplantation.Eur J Clin Invest 9:23—27, 1979

3. ZOCCALI C, CICCARELLI M, MAG0IORE Q:Defective reflex control ofheart rate in dialysis patients: Evidence for an afferent autoflomiclesion. Clin Sci 63:285—292, 1982

4. HEIDBREDER E, SCHAFFERHANS K, HEIDLAND A: Disturbances ofperipheral and autonomic nervous system in chronic renal failure:Effects of hemodialysis and transplantation. Gun Nephrol 23:222—228,1985

5. MALLAMACI F, ZOCCALI C, CICCARELLI M, BRIGGS JD: Autonomicfunction in uremic patients treated by hemodialysis or CAPD and intransplant patients. Clin Nephrol 25:175—180, 1986

6. THOMSON BJ, MCAREAvEY D, NEILSON JMM, WINNEY RJ, EWINGDJ: Heart rate variability and cardiac arrhythmias in patients withchronic renal failure. Clin Auton Res 1:131—133, 1991

7. 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, Science213:220—223, 1981

8. POMERANZ B, MACAULAY RJB, CAUDILL MA, KUTZ I, ADAM D,GORDON D, KILBORN KM, BARGER AC, SHANNON DC, COHEN RJ,

0

0.8

0.6

.c 0.4000.2

0 500

500

—1.6 Frequency, mHz

0 500

3.2

1.6

0

1452 Weise et al: HD and cardiovascular rhythms

BENSON H: Assessment of autonomic function in humans by heart ratespectral analysis. Am J Physiol 248:H151—H153, 1985

9. PAGANI M, LOMBARD! F, GuzzErrI S, RIM0LDI 0, FURLAN R,PIzzINELLI P, SANDRONE U, MALFATrO 0, DELL'ORTO 5, PICCALUGAE, TURIEL M, BASELLI ci, CERUTFI S, MALLIANI A: Power spectralanalysis of heart rate and arterial blood pressure variabilities as amarker of sympatho-vagal interaction in man and conscious dog. CircRes 59:178—193, 1986

10. WEISE F, FIEYDENREICFI F, RUNGE U: Contribution of sympatheticand vagal mechanisms to the genesis of heart rate fluctuations duringorthostatic load: A spectral analysis. J Auton Nerv Syst 21:127-434,1987

11. SAUL JP, RCA RF, ECKBERG DL, BERGER RD, COHEN RJ: Heart rateand muscle sympathetic nerve variability during reflex changes ofautonomic activity. Am J Physiol 258:H713—H721, 1990

12. PAGANI M, SOMERS V, FURLAN R, DELL'ORTO 5, CONWAY J, BASELLIG, CERUTr1 5, SLEIGHT P, ML1AN1 A: Changes in autonomicregulation induced by physical training in mild hypertension. Hyper-tension 12:600—610, 1988

13. SAYERS BMcA: Analysis of heart rate variability. Ergonomics 16:17—32, 1973

14. AKSELROD 5, GORDON D, MADWED JB, SNIDMAN NC, SHANNON DC,

COHEN RJ: Hemodynainic regulation: Investigation by spectral anal-ysis. Am J Physiol 249:H867—H875, 1985

15. DORNHORST AC, HOWARD P, LEATHART GL: Respiratory variationsin blood pressure. Circulation 6:553—558, 1952

16. FORSSTROM 3, FORSSTROM J, HEINONEN E, VALIMAKI I, ANTILA K:Effects of haemodialysis on heart rate variability in chronic renalfailure. Scand J C/in Lab invest 46:665—670, 1986

17. CLOAREC-BLANCHARD L, GIRARD A, HOUHOU S, GRONFELD JP,ELGHOZI JL: Spectral analysis of short-term blood pressure and heartrate variability in uremic patients. Kidney mt 41 (Suppl 37):S14—S18,1992

18. LONDON GM, MARCHAIS SJ, GUERIN AP, METIVIER F, PANNIER B:Cardiac hypertrophy and arterial alterations in end-stage renal dis-ease: Hemodynamic factors. Kidney mt 43 (Suppl 41):S42—S49, 1993

19. WEISE F, LONDON GM, GUERIN AP, PANNIER BM, ELGHOZI JL: Effectof head-down tilt on cardiovascular control in healthy subjects: Aspectral analytic approach. C/in Sci (in press)

20. CLMun BF, EWING DJ, CAMPBELL 1W: Diabetic autonomic neurop-athy. Diabetologia 17:195—212, 1979

21. MICHAELSON ED, GRASSMAN ED, PETERS WR: Pulmonary mechanicsby spectral analysis of forced random noise. J COn invest 56:1210-.1230, 1975

22. ROBBE HWJ, MULDER UM, RUDDEL H, LANGEWITZ WA, VELDMANJBP, MULDER G: Assessment of baroreceptor reflex sensitivity bymeans of spectral analysis. Hypertension 10:538—543, 1987

23. SAUL JP, BERGER RD, ALBRECHT P, STEIN SP, CHEN HM, COHEN RJ:Transfer function analysis of the circulation: Unique insights intocardiovascular regulation. Am J Physiol 261:H1231—H1245, 1991

24. MULDER UM: Asseshient of cardiovascular reactivity by means ofspectral analysis. Thesis, Groningen, 1988

25. BLOMOVIST CG, STONE HL: Cardiovascular adjustments to gravita-tional stress, in Handbook of Physiology (vol III), Peripheral circula-tion and organ blood flow, Part 2, edited by SHEPHERD JT, ABBOUDFM, Bethesda, American Physiol Soc, 1983, pp 1025—1063

26. WEHLE B, ASABA H, CASTENFORS J, FURST P, GUNNARSSON B,SHALDON S, BERGSTROM J: Hemodynamic changes during sequentialultrafiltration and dialysis. Kidney mt 15:411—415, 1979

27. WEHLE B, BEVEGARD 5, CASTENFORS J, DAVIDSSON 5, LINDBLAD LE:

Carotid baroreflexes during hemodialysis. Cl/n Nephrol 19:236—242,1983

28. AKSELROD 5, LISHNER M, Oz 0, BERNHEIM J, R..vio M: Spectralanalysis of fluctuations in heart rate: An objective evaluation ofautonomic nervous control in chronic renal failure. Nephron 45:202—206, 1987

29. STIW4O S, DE CASTRO 5, URANI C, FERRUCCI A, MARCHESELLI A,POLl L, BERLOCO P, CALCAGNINI G, CERUTrI S, CORTESINI R: Power

spectrum analysis of heart rate variability following kidney transplan-tation. Transplant Proc 25:2600—2601, 1993

30. MUNAKATA M, IMAI Y, SEKINO H, ABE K: Rhythmic oscillations ofblood pressure and R-R interval in patients with chronic renal failure.Med Biol Eng Comput 31 (Suppl):S115—S122, 1993

31. BURGESS ED: Cardiac vagal denervation in hemodialysis patients.Nephron 30:228—230, 1982

32. ENDRE ZH, PERL SI, KRAEGEN EW, CHARLESWORTH JA, MAC-DONALD GJ: Reduced cardiac beat-to-beat variation in chronic renalfailure: A ubiquitious marker of autonomic neuropathy. Clin Sci62:561—562, 1982

33. RACZKOWSKA M, ECKBERG DL, EBERT TJ: Muscarinic cholinergicreceptors modulate vagal cardiac responses in man. J Auton Nerv Syst7:271—278, 1983

34. WEISE F, BALTRUSCH K, HEYDENREICH F: Effect of low-dose atropineon heart rate fluctuations during orthostatic load: A spectral analysis.JAuton Nerv Syst 26:223—230, 1989

35. WEISE F, HEYDENREICH F: A non-invasive approach to cardiacautonomic neuropathy iG patients with diabetes mellitus. C/in Physiol10: 137—145, 1990

36. FURLAN R, GuzzErn 5, CRIVELLARO W, DASSI 5, TINELLI M, BASELLIG, CERUTITI 5, LOMBARD! F, PAGAN! M, MALLIAN! A: Continuous24-hour assessment of the neural regulation of systemic arterialpressure and RR variabilities in ambulant subjects. Circulation 81:537—547, 1990

37. RIMOLDI 0, PIERINI 5, FERRARI A, CERUTrI 5, PAGAN! M, MALLIANIA: Analysis of short-term oscillations of R-R and arterial pressure inconscious dogs. Am J Physiol 258:H967—H976, 1990

38. JAPUNDZIC N, GRIcH0IS ML, ZITOUN P, LAUDE D, ELGHOZI JL:Spectral analysis of blood pressure and heart rate in conscious rats:Effects of autonomic blockers. JAuton Nerv Syst 30:91—100, 1990

39. PERSSON PB, STAUSS H, CI-IUNG 0, WITrMANN U, UNGER T: Spectrumanalysis of sympathetic nerve activity and blood pressure in consciousrats. Am J Physiol 263:H1348—H1355, 1992

40. IN0UF K, MIYAKE S, KUMASHIRO M, OGATA H, UETA T, AKATSU T:Power spectral analysis of blood pressure variability in traumaticquadriplegic humans. Am J Physiol 260:H842—H847, 1991

41. CONVERSE RL, JACOBSEN TN, TOTO RD, JOST CMT, COSENTINO F,FOUAD-TARAZI F, VICTOR RG: Sympathetic overactivity in patientswith chronic renal failure. NEnglJMed 327:1912—1918, 1992

42. FOLKOW B: Physiological aspects of primary hypertension. Physiol Rev62:347—504, 1982

43. MADWED JB, COHEN RJ: Heart rate response to hemorrhage-induced0.05-Hz oscillations in arterial pressure in conscious dogs. Am JPhysiol 29:H1248—H1253, 1991

44. GIRARD A, PHAN ML, WEISE F, ELGHGZI JL: Effects of plasmaphere-sis on short-term variability of blood pressure in healthy donors. ClinAuton Res 2:299—302, 1992

45. TRIEDMAN JK, COHEN RJ, SAUL JP: Mild hypovolemic stress altersautonomic modulation of heart rate. Hypertension 21:236—247, 1993

46. DAUL AE, WANG XL, MICFJEL MC, BRODDE OE: Arterial hypoten-sion in chronic hemodialyzed patients. Kidney mt 32:728—735, 1987

47. PARATI G, OMBONI 5, FRATTOLA A, DI RIENz0 M, ZANCHETTI A,MANCIA G: Dynamic evaluation of the baroreflex in ambulant sub-jects, in Blood Pressure and Heart Rate Variability. Computer Analysis,Modelling and Clinical Applications, edited by Di RIENZO M, MANCIAG, PARATI 0, PEDOTFI A, ZANCI-JETFI A, Amsterdam, lOS Press, 1992,pp 123—137

48. HARRISON MN, HILL CC, SPAUL WA, GREENLEAF JE: Effect ofhydration on some orthostatic and haematological responses tohead-up tilt. EurfAppl Physiol 55:187—194, 1986

49. WRIGHT Si, LEVINE BD, BLOMOVIST CG: Effect of posture on thecarotid baroreflex. (abstract) Circulation 82:111—515, 1990

50. TRIEDMAN JK, PERROTT M, SAUL JP: Autoregressive parameteriza-lion of cardiovascular control: comparison with Fourier techniques.(abstract) Proceedings of the 14th Annual International Conference ofthe IEEE Engineering in Medicine and Biology Society, Paris, 1992, pp.467—468

51. PICKERING TO, GRIBBIN B, OLIVER DO: Baroreflex sensitivity inpatients on long-term haemodialysis. Clin Sci 43:645—657, 1972

52. LAZARUS JM, HAMPERS CL, LOWRIE EG, MERRILL JP: Baroreceptoractivity in normotensive and hypertensive uremic patients. Circulation47:1015—1021, 1973

53. TOMIYAMA 0, SHIIGAI T, IDEURA T, TOMITA K, Mrro Y, SHINOHARA5, TAKEUCHI 3: Baroreflex sensitivity in renal failure. Cl/n Sci 58:21—27, 1980