Rd

JU

a

ARRA

KDLRV

1

or[waefcfio

s

1h

Medical Engineering & Physics 36 (2014) 585–591

Contents lists available at ScienceDirect

Medical Engineering & Physics

jo ur nal ho me p ag e: www.elsev ier .com/ locate /medengphy

eproducibility and variability of dynamic cerebral autoregulationuring passive cyclic leg raising

.W. Elting ∗, M.J.H. Aries, J.H. van der Hoeven, P.C.A.J. Vroomen, N.M. Mauritsniversity of Groningen, University Medical Centre Groningen, Department of Neurology, Hanzeplein 1, 9700 RB Groningen, The Netherlands

r t i c l e i n f o

rticle history:eceived 3 January 2013eceived in revised form 26 August 2013ccepted 15 September 2013

eywords:ynamic cerebral autoregulationeg raisingeproducibilityariability

a b s t r a c t

Dynamic cerebral autoregulation (dCA) estimates require mean arterial blood pressure (MABP) fluc-tuations of sufficient amplitude. Current methods to induce fluctuations are not easily implementedor require patient cooperation. In search of an alternative method, we evaluated if MABP fluctuationscould be increased by passive cyclic leg raising (LR) and tested if reproducibility and variability of dCAparameters could be improved.

Middle cerebral artery cerebral blood flow velocity (CBFV), MABP and end tidal CO2 (PetCO2) wereobtained at rest and during LR at 0.1 Hz in 16 healthy subjects. The MABP–CBFV phase difference andgain were determined at 0.1 Hz and in the low frequency (LF) range (0.06–0.14 Hz). In addition theautoregulation index (ARI) was calculated.

The LR maneuver increased the power of MABP fluctuations at 0.1 Hz and across the LF range. Despitea clear correlation between both phase and gain reproducibility and MABP variability in the rest condi-tion, only the reproducibility of gain increased significantly with the maneuver. During the maneuverpatients were breathing faster and more irregularly, accompanied by increased PetCO2 fluctuations andincreased coherence between PetCO2 and CBFV. Multiple regression analysis showed that these concomi-

tant changes were negatively correlated with the MABP–CBFV phase difference at 0.1 Hz Variability wasnot reduced by LR for any of the dCA parameters.

The clinical utility of cyclic passive leg raising is limited because of the concomitant changes in PetCO2.This limits reproducibility of the most important dCA parameters. Future research on reproducibility andvariability of dCA parameters should incorporate PetCO2 variability or find methods to keep PetCO2 levelsconstant.

. Introduction

Dynamic cerebral autoregulation (dCA) describes the processf how cerebral blood flow (CBF) is regulated after mean arte-ial blood pressure (MABP) alterations of relatively short duration1]. This is different from the concept of static autoregulation,here the static response of CBF is assessed after a prolonged

lteration in MABP, usually induced by vasoactive drugs [2]. Sev-ral methods exist to estimate the dynamic response directlyrom the time domain, such as inflation and deflation of thighuffs [1]. However, these methods are used infrequently due toact that sudden decreases in MABP may pose a risk of brainschemia and sympathetic activation [3,4] due to painful inflation

f cuffs.

In search for alternative methods, research has focused on usingpontaneous MABP fluctuations to estimate dCA [5]. Usually this is

∗ Corresponding author. Tel.: +31 503616161.E-mail address: j.w.j.elting@neuro.umcg.nl (J.W. Elting).

350-4533/$ – see front matter © 2013 IPEM. Published by Elsevier Ltd. All rights reservettp://dx.doi.org/10.1016/j.medengphy.2013.09.012

© 2013 IPEM. Published by Elsevier Ltd. All rights reserved.

quantified by transfer function analysis which results in a phase andgain spectrum [6]. Another parameter is the autoregulation index(ARI) [5]. The ARI describes the system response to a step-like dis-turbance, and is graded from 0 (absence of autoregulation) to 9(fastest and most complete autoregulation) [7]. An inherent prob-lem with these strategies is that spontaneous MABP fluctuationshave to be of sufficient amplitude to obtain reliable estimates of thedCA parameters. Although healthy subjects usually demonstratesufficient spontaneous MABP fluctuations, in 15–30% of patientsthis requirement is not fulfilled [8,9]. Several maneuvers have beendeveloped to increase MABP fluctuations [10–18]. Some authorshave shown improvement of reproducibility and/or reduced vari-ability of dCA parameters with these methods [10,14,19], othersdid not find any differences [20,21]. Reproducibility and variabil-ity remain important issues in dCA and it was recognized thatimprovements in this area are required before its diagnostic value

can be fully appreciated [3]. Reproducibility specifically refers tovariability between repeated measurements within a subject, whilevariability refers to variability between subjects within a group ofsubjects.

d.

5 eering

campawttsifiaflt

1

2

2

2

yccd

2

wma(NgNemtawGqLt

tai(swStLaibw

86 J.W. Elting et al. / Medical Engin

If methods to increase MABP fluctuations are to be used in alinical setting, they have to be practical, applicable at the bedsidend should not require patient cooperation. The latter require-ent applies especially to certain neurological disorders, such as in

atients with decreased consciousness or aphasia. In search for anlternative method to induce MABP fluctuations in a clinical setting,e tested the effect of passive cyclic leg raising (LR) on dCA. During

his maneuver venous blood shifts from the legs into the intra-horacic compartment, increasing both ventricular preloads, andubsequently cardiac output and MABP. In its static form, leg raisings used to predict fluid responsiveness in patients with circulatoryailure [22]. By alternating raising and returning the legs to the hor-zontal plane, MABP fluctuations are likely to occur. Although dCAnalysis in patients may benefit more from an increase in MABPuctuations, we first investigated this method in healthy subjectso assess feasibility. The objectives of this study were:

. to evaluate whether LR is able to increase low frequency (LF)MABP fluctuations, and

. to evaluate if reproducibility and variability of dCA parameterscan be improved with LR.

. Methods

.1. Subjects

We tested 16 healthy subjects (8 males, mean age 32.5 ± 9.5ears). Approval for this study was obtained from the local ethicsommittee of the University Medical Centre Groningen. Informedonsent was obtained from all subjects. None had a history of car-iovascular or neurological illness, hypertension or smoking.

.2. Measurement protocol

Subjects were positioned in an adjustable medical chair, whichas brought in a supine position for the measurement. To avoidovement artifact, a semicircular inflatable pillow was positioned

round the neck to prevent contact of the Transcranial DopplerTCD) fixation system with the surface of the chair during LR.on-invasive MABP estimates were obtained from a plethysmo-raphy system (Portapres, Finapres Medical Systems, Amsterdam,etherlands), with a cuff placed around the index finger. Bilat-ral middle cerebral artery blood flow velocity (MCA–CBFV) waseasured with TCD (Nicolet Pioneer TC8080, Carefusion Corpora-

ion, San Diego, USA) using 2 MHz probes that were fixated withn adjustable headframe. End tidal PCO2 (PetCO2) was measuredith a capnograph (Capnomac Ultima, GE Healthcare, Chalfont Stiles UK). All data were measured continuously at a sample fre-uency of 250 Hz An analog–digital converter in combination withabview 9.0 software (National Instruments, Austin, USA) was usedo capture the data for storage.

All participants had four measurement periods of 5 min each:wo periods at rest, and two periods of leg raising. To minimizeny time effects, an alternating measurement protocol was used.e.; period 1: rest (R1), period 2: leg raising (LR1), period 3: restR2), period 4: leg raising (LR2). After each period of leg raising ahort pause (between 2 and 3 min) during which no measurementsere taken was inserted, to allow for hemodynamic stabilization.

ubjects were instructed to lie completely still, and were asked noto oppose the leg raising. No breathing instructions were given. TheR maneuver was executed as follows: both legs were lifted at the

nkles and brought to about a 60◦ angle from the horizontal planen about 1 s. After holding this position for 4 s, the legs were broughtack to the horizontal plane in about 1 s, after which this positionas kept for 4 s. This maneuver was then repeated, which resulted

& Physics 36 (2014) 585–591

in a period of 10 s (0.1 Hz). LR at 0.1 Hz was maintained throughoutthe 5 min leg raising periods.

2.3. Data analysis

The recorded signals were visually inspected for artifacts. Occa-sional artifacts of short duration which were visible as narrowspikes in the TCD or MABP signal were removed by linear inter-polation. Beat to beat data were obtained from the MABP signalby triggering off the ascending slope during systole. The data wereresampled at 10 Hz by spline interpolation to create a uniform time-base. PetCO2 was determined from the capnography trace as themaximum over each cycle, and was subjected to separate inter-polation at 10 Hz Respiratory frequency was calculated from thePetCO2 cycle length. A high pass 8th order zero phase Butterworthfilter was applied, with a cut-off frequency of 0.04 Hz This was fol-lowed by mean normalization and subtraction of 1 to create zeromean signals. A Hanning window was applied to the data, and theWelch method of spectral estimation based on Fast Fourier Trans-formation (FFT) was used with 50% overlap on data segments of51.2 s. This resulted in spectral estimates averaged over 10 seg-ments. The transfer function (TF) was calculated without attemptsto unwrap the phase spectra, but the phase estimates were visuallyinspected for phase wrap around. TF graphs were produced at eachiteration during averaging. If any sudden change in phase in theLF range occurred, the data were excluded. Coherence was consid-ered significant if it exceeded the 95% confidence interval (CI) levelof no linear association: 1 − (0.05)1/L − 1 [23], where L is the numberof data segments used in the averaging. This resulted in a significantcoherence level of 0.28. Gain and phase parameters were calculatedat the 0.1 Hz frequency, and were only used for further analysis ifsignificant coherence existed. All frequencies were rounded off totwo decimal places. Mean values of gain and circular mean val-ues of phase were also calculated in the LF range (0.06–0.14 Hz) byaveraging the values from these bins, i.e. the mean of 5 frequencybins.

Power spectral density (PSD: i.e. power normalized by frequencybin width) of MABP, PetCO2, and respiratory frequency signals weredetermined at 0.1 Hz and total power in the LF range was calculatedby integration of power in this frequency range. ARI was calculatedby transforming the real and imaginary parts of the TFA back tothe time domain with inverse FFT. The impulse response functionthus created was integrated to yield the step response function.The first 10 s of the step response function were compared with theoriginal Tiecks curves and the best fitting curve was determined byimplementing a least squares fitting procedure, resulting in a ARIranging from 0 to 9 [7].

2.4. Statistics

SPSS 16.0 and SAS 9.2 software were used. Repeated measuresanalysis of variance (ANOVA) tests with post hoc correction formultiple comparisons (Bonferroni and Dunnet T3 tests) were used.The following post hoc comparisons were done: R1 vs R2, LR1 vsLR2, R1 vs LR1 and R2 vs LR2. Normality of variable distribution wasevaluated with the Shapiro–Wilk statistic, and if necessary loga-rithmic transformation was used. After logarithmic transformation,all transformed variables were normally distributed in this study.To assess any differences in variability the Levene statistic wasused. Intraclass correlation coefficient (ICC) analysis was used toassess reproducibility. Differences in ICCs were assessed with a dif-ference test based on the CI for two dependent ICCs [24]. No formal

power calculation was done. The relation between reproducibilityand MABP variability was examined with correlation analysis. Therelation between dCA estimates and independent variables (MABP,Pecos and derivatives) was evaluated with multiple regression

J.W. Elting et al. / Medical Engineering & Physics 36 (2014) 585–591 587

Table 1Physiological characteristics: mean data ± SD, spectral data are expressed as median ± IQR because of non-normal distributions.

Rest (1) Leg raising (1) Rest (2) Leg raising (2) p value ANOVA

MABP (mmHg) 78.1 ± 13.1 81.6 ± 14.5* 81.7 ± 13.7# 84.1 ± 14.4* <0.001Mean CBFV (cm/s) Left MCA 57.1 ± 12.1 57.4 ± 12.3 57.0 ± 12.4 55.9 ± 12.0 0.444

Right MCA 64.6 ± 14.1 66.6 ± 15.5 65.3 ± 14.1 64.9 ± 15.2 0.099PetCO2 (kPa) 4.56 ± 0.50 4.41 ± 0.50* 4.46 ± 0.51# 4.28 ± 0.53*,# <0.001Heart rate (beats/min) 63.5 ± 7.1 64.2 ± 8.2 62.2 ± 7.4 64.2 ± 8.7 0.226RF (cycles/min) 13.8 ± 3.4 15.2 ± 3.1* 13.9 ± 3.3 15.4 ± 2.7* 0.003PSD total MABP (mmHg2 ×10−3)/Hz 18.2 ± 12.3 35.5 ± 49.4* 18.5 ± 12.5 29.6 ± 31.3* <0.001PSD 0.1 Hz MABP (mmHg2 ×10−3)/Hz 4.4 ± 2.5 11.2 ± 21.1* 4.3 ± 3.1 11.5 ± 14.0* <0.001PSD MABP peak frequency (Hz) 0.07 ± 0.02 0.10 ± 0.00* 0.08 ± 0.02# 0.10 ± 0.00* <0.001PSD total CBFV (mmHg2 ×10−3)/Hz Left MCA 21.6 ± 8.9 37.6 ± 45.9* 18.2 ± 16.1 36.3 ± 30.5* 0.019

Right MCA 22.5 ± 8.7 27.0 ± 35.6* 20.1 ± 15.0 34.8 ± 37.8* 0.040PSD 0.1 Hz CBFV (mmHg2 ×10−3)/Hz Left MCA 5.5 ± 4.0 13.0 ± 17.3* 4.8 ± 4.8 14.0 ± 11.7* 0.003

Right MCA 5.9 ± 3.2 11.3 ± 12.1* 5.1 ± 5.4 13.1 ± 17.2* 0.005PSD total CO2 (kPa2 ×10−3)/Hz 3.7 ± 5.8 14.7 ± 44.1* 5.6 ± 5.2 14.2 ± 32.7* <0.001PSD 0.1 Hz CO2 (kPa2 ×10−3)/Hz 0.49 ± 1.4 2.6 ± 10.5* 0.81 ± 0.97 3.0 ± 6.0* <0.001PSD total RF (cycles/min2 ×10−2)/Hz 11.8 ± 15.4 44.1 ± 54.6* 12.0 ± 13.5 25.5 ± 26.0*,# <0.001PSD 0.1 Hz RF (cycles/min2 ×10−2)/Hz 1.9 ± 2.0 11.1 ± 16.8* 1.9 ± 3.9 5.6 ± 9.2* <0.001

MABP = mean arterial blood pressure, CBFV = cerebral blood flow velocity, MCA = middle cerebral artery, PSD = power spectral density, RF = respiratory frequency.ns).

T

aii

3

aev

3

tpLAupdlofouc

3

mdlfi0csCbsb

* p < 0.05 vs previous different condition (after correction for multiple compariso# p < 0.05 vs previous similar condition.otal refers to total PSD in the 0.06–0.14 Hz frequency range.

nalysis using a stepwise forward strategy. We accepted a variancenflation factor of <10 as evidence that the input variables werendeed independent (collinearity diagnostics) [25].

. Results

Good quality measurements were obtained in all 16 subjects,nd no exclusions were necessary. The LR maneuver was well tol-rated by all 16 subjects. Mean data and spectral data of all relevantariables during the experiment can be found in Table 1 and in Fig. 1.

.1. Changes induced by LR

Leg raising at 0.1 Hz significantly increased power of MABP inhe LF band. This was accompanied by a shift in the peak of MABPower from around 0.07 to 0.08 Hz during rest to 0.1 Hz duringR. Absolute levels of MABP changed slightly, albeit significantly.t 0.1 Hz, MABP power markedly increased. Mean CBFV remainednchanged during the experiment, but total LF power and 0.1 Hzower increased significantly during LR. Absolute PetCO2 levelsecreased significantly, although only by a small amount. A much

arger increase was observed in the total LF power and 0.1 Hz powerf PetCO2 fluctuations (Table 1). An increase in absolute respiratoryrequency occurred during LR. The total LF power and 0.1 Hz powerf the respiratory frequency curve also increased. The HR remainednchanged during the experiment. Data of a representative subjectan be found in Fig. 2.

.2. Dynamic cerebral autoregulation parameters

Transfer function and ARI results can be found in Table 2. Circularean phase was not different between the four periods. The phase

ifference between MABP and CBFV at 0.1 Hz was different for theeft side only, with a significant difference between the first rest andrst LR period, but not between other periods. Mean gain, gain at.1 Hz and ARI were not different between the four periods. Meanoherence and coherence at 0.1 Hz between MABP and CBFV wereimilar across all periods. However, coherence between PetCO2 and

BFV at 0.1 Hz was different with significant post hoc differencesetween the R1 and LR1 (both hemispheres, p < 0.05, Table 2). Noignificant differences were found for mean coherence in the LFand between PetCO2 and CBFV.

3.3. Reproducibility and variability

Reproducibility results can be found in Table 3. ICCs for therest period comparison were generally low for mean phase and0.1 Hz phase and this did not improve during LR. The ICC valuesfor mean and 0.1 Hz gain were low in the rest period comparison,but improved significantly during LR, especially for 0.1 Hz gain (lefthemisphere: p = 0.02; right hemisphere: p = 0.05). The ICC valuesfor ARI did not improve during LR. To examine the effects of differ-ent levels of MABP variability on reproducibility, we performed anadditional correlation analysis between MABP variability vs DCAparameters. The absolute difference in DCA parameters betweensessions was correlated with the mean PSD MABP values of the ses-sions (Table 4; online supplement). The results indicate that duringrest, phase and gain reproducibility are related to MABP variabil-ity, while during leg raising this relation is maintained for gain, butlost for phase, probably as a consequence of increased influence ofPetCO2 (see Section 3.4).

Supplementary material related to this article can be found,in the online version, at http://dx.doi.org/10.1016/j.medengphy.2013.09.012.

Variability was not different between the four periods for alldCA parameters (mean phase (left: p = 0.14, right: p = 0.34), 0.1 Hzphase (left: p = 0.51, right: p = 0.56), mean gain (left: p = 0.74, right:p = 0.40), 0.1 Hz gain (left: p = 0.09, right: p = 0.06), ARI (left: p = 0.94,right: p = 0.26).

3.4. Relation between variables

Additional analyses were performed to explore the significantlylower 0.1 Hz phase difference during LR1, while no such differenceoccurred for the 0.1 Hz gain. Absolute levels of MABP and PetCO2as well as (the log-transformed) 0.1 Hz MABP PetCO2 power andrespiratory frequency (RF) power, and the 0.1 Hz coherence levelsof MABP and PetCO2 were entered as independent variables in themultiple regression model. Collinearity analysis revealed that allinput variables were statistically independent. Only the coherenceat 0.1 Hz between PetCO2 and CBFV was significantly related

to the 0.1 Hz phase difference (left hemisphere: p = 0.048; righthemisphere: p = 0.007) during LR1. Fig. 3 shows the linear relationbetween the PetCO2–CBFV coherence at 0.1 Hz and the 0.1 Hzphase difference. For 0.1 Hz gain, only the 0.1 Hz MABP power

588 J.W. Elting et al. / Medical Engineering & Physics 36 (2014) 585–591

Fig. 1. Coherence, MABP power spectral density (PSD), gain and phase spectra for all measured frequency bins. Data of the first rest period (gray bars) and the first leg raisingperiod (black bars) are shown (mean ± s.e.m.). Left and right sides were averaged. Note that coherence does not increase despite a clear increase in MABP power in thefrequency bins around 0.1 Hz There is significant leakage into the 0.08 Hz bin, and to a lesser degree into the 0.12 Hz bin. Gain remains similar for all frequency bins, whilefor phase, a focal decrease is found at the 0.1 and 0.08 frequency bins, but not elsewhere in the LF band or VLF band. Statistics are reported in Table 2 for the LF band data.For the VLF band, no significant differences were found for any of the reported parameters.

3

3.5

4

4.5

5

5.5

6

6.5

7

7.5

8

0

10

20

30

40

50

60

70

80

90

6001540148014201360130012401180112016011

Pet

CO

2 (k

Pa)

Time (s)

MABP

MCA-L

MCA-R

PetCO2

RF

Fig. 2. Data of a representative subject. Left panel: rest period 1. Right panel: leg raising period 1. Only PetCO2 is plotted on the secondary Y-axis, all other variables areplotted on the primary Y-axis. Note the increase in oscillation amplitude of MABP, CBFV in both MCA’s, but also in PetCO2 and RF (respiratory frequency) during the legraising period.

J.W. Elting et al. / Medical Engineering & Physics 36 (2014) 585–591 589

Table 2Transfer function and ARI results.

Rest 1 Leg raising 1 Rest 2 Leg raising 2 p value ANOVA

Mean Phase (degree) L 42.4 ± 12.7 36.9 ± 9.6 40.1 ± 14.5 40.6 ± 13.2 0.500R 44.8 ± 11.9 38.4 ± 11.2 40.9 ± 14.4 43.5 ± 13.9 0.190

Phase 0.1 Hz (degree) L 47.6 ± 11.2 33.8 ± 10.7* 44.1 ± 12.7 37.1 ± 14.5 0.017R 46.1 ± 10.9 35.9 ± 14.4 44.5 ± 15.1 46.6 ± 16.1 0.130

Mean Gain (%/mmHg) L 1.02 ± 0.25 1.02 ± 0.37 1.11 ± 0.28 1.14 ± 0.29 0.404R 0.97 ± 0.26 0.93 ± 0.31 1.07 ± 0.19 1.06 ± 0.23 0.080

Gain 0.1 Hz (%/mmHg) L 0.99 ± 0.29 0.93 ± 0.52 0.91 ± 0.26 1.05 ± 0.49 0.613R 0.99 ± 0.25 0.83 ± 0.34 1.01 ± 0.19 0.94 ± 0.40 0.594

ARI L 6.2 ± 2.2 6.1 ± 1.9 6.1 ± 2.1 5.7 ± 2.2 0.935R 6.9 ± 1.9 6.9 ± 1.4 6.4 ± 1.9 6.4 ± 2.1 0.385

Mean Coh. MABP/CBFV L 0.67 ± 0.14 0.61 ± 0.12 0.62 ± 0.15 0.61 ± 0.15 0.060R 0.67 ± 0.13 0.62 ± 0.11 0.62 ± 0.15 0.61 ± 0.15 0.315

Coh. MABP/CBFV 0.1 Hz L 0.72 ± 0.17 0.71 ± 0.21 0.65 ± 0.24 0.74 ± 0.18 0.843R 0.75 ± 0.16 0.74 ± 0.13 0.69 ± 0.25 0.73 ± 0.21 0.450

Mean Coh. PetCO2/CBFV L 0.10 ± 0.04 0.14 ± 0.07 0.11 ± 0.05 0.15 ± 0.08 0.288R 0.09 ± 0.04 0.15 ± 0.09 0.13 ± 0.06 0.14 ± 0.07 0.133

Coh. PetCO2/CBFV 0.1 Hz L 0.10 ± 0.07 0.24 ± 0.17* 0.15 ± 0.11 0.25 ± 0.15 0.013R 0.10 ± 0.08 0.23 ± 0.17* 0.15 ± 0.13 0.23 ± 0.16 0.009

Means ± SD. Coh. = coherence, ARI = autoregulation index, MABP = mean arterial blood pressure, CBFV = cerebral blood flow velocity.* p < 0.05 vs previous different condition (after correction for multiple comparisons).

Table 3Intraclass correlation analysis results.

Rest 1 vs Rest 2 95% CI Leg raising 1 vs Leg raising 2 95% CI

Mean phase L 0.51 0.05 to 0.79 0.59 0.16 to 0.83R 0.76 0.43 to 0.91 0.31 −0.19 to 0.68

0.1 Hz phase L 0.57 0.11 to 0.84 0.47 −0.04 to 0.79R 0.36 −0.17 to 0.74 0.38 −0.13 to 0.74

Mean gain L 0.46 −0.01 to 0.77 0.67 0.29 to 0.87R 0.26 −0.25 to 0.67 0.62 0.21 to 0.85

0.1 Hz gain L 0.43* −0.09 to 0.77 0.84* 0.60 to 0.94R 0.25* −0.29 to 0.67 0.70* 0.49 to 0.94

ARI L 0.57 0.13 to 0.82 0.65 0.24 to 0.87R 0.82 0.56 to 0.93 0.60 0.17 to 0.84

s.C

wr

4

4

4

tseiiMiwceaatstetld

4.1.2. PhaseThe 0.1 Hz MABP–CBFV phase tended to decrease during the

LR maneuver, which would indicate impaired autoregulation. Theincreased 0.1 Hz PetCO2–CBFV coherence was correlated with a

* Significant difference (p ≤ 0.05) in ICC between the legraising and rest conditionI = confidence interval. ARI = autoregulation index.

as significantly related during both LR periods. Fig. 4 shows theelation between 0.1 Hz gain and the 0.1 Hz MABP power.

. Discussion

.1. Changes during the LR maneuver

.1.1. CoherenceThe results of this study show that the LR maneuver was well

olerated and increased the power of MABP and CBFV at 0.1 Hzignificantly in healthy subjects. However, the MABP–CBFV coher-nce did not increase. This may be partly attributed to the fact thatn the resting condition, the LF coherence was already quite highn our study population, because some degree of fluctuations in

ABP and CBFV was present in most subjects. Furthermore, dur-ng 0.1 Hz LR subjects started to breathe faster and more irregularly

ith significantly increased PetCO2 fluctuations and PetCO2–CBFVoherence at 0.1 Hz, which shows that PetCO2 had a greater influ-nce on CBFV. This may explain why coherence between MABPnd CBFV did not increase after leg raising. Although the aver-ge increase in PetCO2–CBFV coherence at 0.1 Hz did not reachhe significance level of 0.28, individual values were significant inome cases. The results of the multiple regression analysis suggesthat PetCO2–CBFV coherence had an impact on the phase differ-

nce between MABP and CBFV. Other potential explanations forhe change in phase difference with LR, such as changes in MABPevels or changes in respiratory frequency are not supported by theata from this regression analysis.

Fig. 3. Correlation between coherence at 0.1 Hz (PetCO2/CBFV) and the 0.1 Hzphase difference (MABP/CBFV) for both MCA’s during the first legraisingperiod. MABP = mean arterial blood pressure. CBFV = cerebral blood flow velocity.MCA = middle cerebral artery.

590 J.W. Elting et al. / Medical Engineering & Physics 36 (2014) 585–591

Fig. 4. Relation between gain at 0.1 Hz (MABP/CBFV) and PSD of MABP fluctuations. Data of the first and second leg raising period are presented. A significant inverse relationexisted, which was best described by non linear inverse functions. After logarithmic transformation of PSD of MABP fluctuations a linear inverse relation was the best fittingf LR1 = fip

ddPaasTibesci

4

ewtbpTctavatciWer

4

stg

unction, and the transformed data were used in the multiple regression analysis.

ressure. CBFV = cerebral blood flow velocity. PSD = power spectral density.

ecrease in the phase estimate (Fig. 3). We hypothesize that theecrease in phase was caused by the disruptive effect of increasedetCO2 fluctuations during LR. An increase in PetCO2 fluctuationsfter MABP oscillation inducing maneuvers has also been describedfter repeated squat-stand maneuvers in the LF band [14]. In thattudy also a decrease in phase (with unaffected gain) was found.his was not interpreted as a CO2 effect, but rather as an alterationn cerebrovascular tone due to an increase in shear stress causedy the large squat-stand MABP fluctuations. However, the coher-nce between PetCO2 and CBFV was not assessed in the squat-standtudy. Our findings indicate that an increase in CO2 fluctuationsan disturb the MABP–CBFV phase estimate in the LF range signif-cantly.

.1.3. Gain and ARIGain and ARI estimates remained unchanged during LR. In our

xperiment the major determinant for gain was MABP power,hile CO2 fluctuations did not have a significant effect. The rela-

ion between MABP power and gain was present during LR,ut not during rest. The decrease in gain with increasing MABPower is best described by a non-linear inverse function (Fig. 4).his may indicate two things. Firstly, a threshold phenomenon;erebral autoregulation becomes active above a certain MABPhreshold. Secondly, at higher MABP fluctuation levels the vari-bility of gain decreases. However, reduction in inter-individualariability could not be demonstrated with formal statisticalnalysis, although intra-individual variability was lower (see Sec-ion 4.2). Sensitivity of ARI to both static and dynamic CO2hanges have been described before, using autoregressive mov-ng average models with time varying estimates of ARI [26,27].

e did not use a time varying estimate of ARI and this differ-nce in technique is a possible explanation for these differentesults.

.2. Reproducibility and variability

Only the reproducibility of the 0.1 Hz gain estimate increasedignificantly with LR. This is remarkable, since the result ofhe correlation analysis clearly showed that both phase andain reproducibility are related to MABP variability in the rest

rst leg raising period. LR2 = second leg raising period. MABP = mean arterial blood

condition. Apparently, gain is less sensitive to PetCO2 fluctua-tions with reproducibility improving with increasing MABP powerinduced by LR. The fact that inter-individual variability remainedunchanged despite increased reproducibility indicates that a largeproportion of total gain variability is due to inter-individual vari-ability. Reductions in dCA estimate variability with increasingMABP fluctuations have been reported by others [19]. The limitednumber of subjects in our study, and the fact that not every subjectshowed a large increase in MABP power with LR in combinationwith changes in CO2 can be an explanation for this discrepancy.We investigated short term reproducibility only in this study. Wecannot be certain if the leg raising method would be of help inincreasing long-term reproducibility, particularly in patients or inpersons with low spontaneous BP fluctuations.

4.3. Limitations

Several limitations deserve mentioning. Firstly, the physiolog-ical mechanism for the increase in MABP power after LR is notentirely clear. Besides an increase in venous return, the alteredbreathing pattern may also have contributed to the increase inMABP fluctuations. Breathing depth was not measured, so we werenot able to analyze this option extensively. Secondly, changes inautonomic nerve activity may also have contributed to the observedchanges after leg raising. The fact that MABP increased slightlybut significantly could be viewed as an indication that autonomicnerve activity was altered. Although the multiple regression anal-ysis failed to show any major effects of MABP on phase, we cannotrule out that some influence of altered autonomic activity on MABPvariability and DCA parameters was present. Thirdly, in contrast tothe LF band, limited coherence is more often problematic in the verylow frequency (VLF) band (<0.05 Hz) [14]. However, in this experi-ment we chose to focus on the LF band, because in this band MABPis considered to be the most important determinant for coherence[6]. In the VLF band spontaneous oscillations in PetCO2 and non-linear system interactions have been shown to exert a significant

influence on the MABP–CBFV coherence calculation [28,29]. Finally,although we hypothesize that concomitant PetCO2 fluctuations sig-nificantly influenced reproducibility of phase and ARI during LR, wecannot exclude the possibility that limitations in methodology have

eering

cdsnpsPibin

4

Mwfimeucdiibmtaita

F

E

U

A

TuIMt

C

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

J.W. Elting et al. / Medical Engin

ontributed. Our methodology only provides a single estimate ofCA parameters for the entire recording period. Other techniquesuch as autoregressive moving average modeling, phase synchro-ization, multimodal pressure-flow analysis or wavelet analysisrovide multiple estimates over time and are less sensitive to nontationarities [3,30]. Multivariate models may be able to correct foretCO2 changes. Alternatively, methods that filter out short last-ng changes in dCA parameters caused by PetCO2 changes maye developed. However, recent experiments have failed to show

mprovements in reproducibility with these methods and tech-iques [20,21].

.4. Implications for clinical use and future directions

We aimed to obtain a practical bedside test that would increaseABP power with minimal patient cooperation. Although the LRas well tolerated and significantly increased MABP power, therst variability and reproducibility results of our selected dCA esti-ates were disappointing. Only the reproducibility for the gain

stimate improved, most probably because it is less affected bynexpected concomitant PetCO2 changes. Therefore, at present thelinical utility of LR to improve diagnostic properties for testingCA status is limited. In cooperative patients the use of breathing

nstructions at a fixed pace could be considered and in mechan-cally ventilated patients there might be less risk of concomitantreathing and PetCO2 changes. Alternatively, one could employethods to ensure that PetCO2 remains constant, such as sequen-

ial gas delivery breathing circuit devices that can accurately targetnd control PetCO2 [31]. The fact that PetCO2 fluctuations have anmportant influence on dCA estimates further illustrates the needo incorporate or at least report PetCO2 variability in future dCAnalysis [28,29,32].

unding

None

thical approval

Ethical approval was given by the local ethics committee of theniversity Medical Centre Groningen.

cknowledgments

The Authors wish to thank Mr. B. Giraudeau (INSERM, CIC 202,ours, France; CHRU de Tours, CIC 202, Tours, France) for supplyings with the SAS macros for the statistical analysis of two dependent

CC’s. The Authors would like to thank Mr. R. Stewart (Universityedical Centre Groningen, University of Groningen) for help with

he statistical analysis.

onflict of interests

None declared.

eferences

[1] Aaslid R, Lindegaard KF, Sorteberg W, Nornes H. Cerebral autoregulationdynamics in humans. Stroke 1989;20(1):45–52.

[2] Lassen NA. Autoregulation of cerebral blood flow. Circ Res1964;15(Suppl.):201–4.

[3] Panerai RB. Transcranial Doppler for evaluation of cerebral autoregulation. ClinAuton Res 2009;19(4):197–211.

[4] van Lieshout JJ, Secher NH. Point: counterpoint: sympathetic activity does/doesnot influence cerebral blood flow. Point: sympathetic activity does influencecerebral blood flow. J Appl Physiol 2008;105(4):1364–6.

[

& Physics 36 (2014) 585–591 591

[5] Panerai RB, White RP, Markus HS, Evans DH. Grading of cerebral dynamicautoregulation from spontaneous fluctuations in arterial blood pressure. Stroke1998;29(11):2341–6.

[6] Zhang R, Zuckerman JH, Giller CA, Levine BD. Transfer function analysis ofdynamic cerebral autoregulation in humans. Am J Physiol 1998;274(1 (Pt2)):H233–41.

[7] Tiecks FP, Lam AM, Aaslid R, Newell DW. Comparison of static and dynamiccerebral autoregulation measurements. Stroke 1995;26(6):1014–9.

[8] Lipsitz LA, Mukai S, Hamner J, Gagnon M, Babikian V. Dynamic regulation ofmiddle cerebral artery blood flow velocity in aging and hypertension. Stroke2000;31(8):1897–903.

[9] Reinhard M, Rutsch S, Lambeck J, Wihler C, Czosnyka M, WeillerC, et al. Dynamic cerebral autoregulation associates with infarct sizeand outcome after ischemic stroke. Acta Neurol Scand 2012;125(3):156–62.

10] Reinhard M, Muller T, Guschlbauer B, Timmer J, Hetzel A. Transfer functionanalysis for clinical evaluation of dynamic cerebral autoregulation – a compar-ison between spontaneous and respiratory-induced oscillations. Physiol Meas2003;24(1):27–43.

11] Diehl RR, Linden D, Lucke D, Berlit P. Phase relationship between cerebralblood flow velocity and blood pressure. A clinical test of autoregulation. Stroke1995;26(10):1801–4.

12] Birch AA, Neil-Dwyer G, Murrills AJ. The repeatability of cerebral autoregula-tion assessment using sinusoidal lower body negative pressure. Physiol Meas2002;23(1):73–83.

13] Birch AA, Dirnhuber MJ, Hartley-Davies R, Iannotti F, Neil-Dwyer G. Assess-ment of autoregulation by means of periodic changes in blood pressure. Stroke1995;26(5):834–7.

14] Claassen JA, Levine BD, Zhang R. Dynamic cerebral autoregulation duringrepeated squat-stand maneuvers. J Appl Physiol 2009;106(1):153–60.

15] Kwan J, Lunt M, Jenkinson D. Assessing dynamic cerebral autoregulation afterstroke using a novel technique of combining transcranial Doppler ultrasono-graphy and rhythmic handgrip. Blood Press Monit 2004;9(1):3–8.

16] Hughson RL, Edwards MR, O’Leary DD, Shoemaker JK. Critical analy-sis of cerebrovascular autoregulation during repeated head-up tilt. Stroke2001;32(10):2403–8.

17] Gisolf J, Stok WJ, Oei SI, Immink RV, vanLieshout JJ, Karemaker JM. Dynamiccerebral autoregulation under sinusoidal gravitational loading. J Gravit Physiol2002;9(1):P85–6.

18] Panerai RB, Dawson SL, Eames PJ, Potter JF. Cerebral blood flow velocityresponse to induced and spontaneous sudden changes in arterial blood pres-sure. Am J Physiol Heart Circ Physiol 2001;280(5):H2162–74.

19] Liu J, Simpson DM, Allen R. High spontaneous fluctuation in arterial bloodpressure improves the assessment of cerebral autoregulation. Physiol Meas2005;26(5):725–41.

20] Gommer ED, Shijaku E, Mess WH, Reulen JP. Dynamic cerebral autoregula-tion: different signal processing methods without influence on results andreproducibility. Med Biol Eng Comput 2010;48(12):1243–50.

21] van Beek AH, Olde Rikkert MG, Pasman JW, Hopman MT, Claassen JA. Dynamiccerebral autoregulation in the old using a repeated sit-stand maneuver. Ultra-sound Med Biol 2010;36(2):192–201.

22] Cavallaro F, Sandroni C, Marano C, La Torre G, Mannocci A, De Waure C, et al.Diagnostic accuracy of passive leg raising for prediction of fluid responsivenessin adults: systematic review and meta-analysis of clinical studies. IntensiveCare Med 2010;36(9):1475–83.

23] Halliday DM, Rosenberg JR, Amjad AM, Breeze P, Conway BA, Farmer SF. Aframework for the analysis of mixed time series/point process data – theory andapplication to the study of physiological tremor, single motor unit dischargesand electromyograms. Prog Biophys Mol Biol 1995;64(2/3):237–78.

24] Ramasundarahettige CF, Donner A, Zou GY. Confidence interval constructionfor a difference between two dependent intraclass correlation coefficients. StatMed 2009;28(7):1041–53.

25] Myers R. Classical and modern regression with applications. 2nd ed. PWS-Kentpublishing company; 1990.

26] Dineen NE, Brodie FG, Robinson TG, Panerai RB. Continuous estimates ofdynamic cerebral autoregulation during transient hypocapnia and hypercap-nia. J Appl Physiol 2010;108(3):604–13.

27] Panerai RB, Dineen NE, Brodie FG, Robinson TG. Spontaneous fluctuationsin cerebral blood flow regulation: contribution of PaCO2. J Appl Physiol2010;109(6):1860–8.

28] Peng T, Rowley AB, Ainslie PN, Poulin MJ, Payne SJ. Multivariate sys-tem identification for cerebral autoregulation. Ann Biomed Eng 2008;36(2):308–20.

29] Mitsis GD, Poulin MJ, Robbins PA, Marmarelis VZ. Nonlinear modeling of thedynamic effects of arterial pressure and CO2 variations on cerebral blood flowin healthy humans. IEEE Trans Biomed Eng 2004;51(11):1932–43.

30] Panerai RB. Cerebral autoregulation: from models to clinical applications. Car-diovasc Eng 2008;8(1):42–59.

31] Slessarev M, Han J, Mardimae A, Prisman E, Preiss D, Volgyesi G, et al. Prospec-

tive targeting and control of end-tidal CO2 and O2 concentrations. J Physiol2007;581(Pt 3):1207–19.

32] Peng T, Rowley AB, Ainslie PN, Payne SJ, Poulin MJ. Wavelet phase synchro-nization analysis of cerebral blood flow autoregulation. IEEE Trans Biomed Eng2010;57(4):960–8.