The Journal of Applied Research • Vol. 4, No. 2, 2004 245

dard conditions. Further studies areneeded before clinical use is recom-mended.

INTRODUCTIONConventional risk factors are unable tofully predict those individuals atincreased risk from cardiovascular dis-ease. Thus alternative, non-invasive,methods of assessment, which can reli-ably assess cardiovascular risk, aredesirable. It has been demonstrated thatimpaired flow-mediated dilatationassessed by ultrasound correlates withvascular risk factors and future vascularrisk.1 Compliance of the arterial tree isassociated with cardiovascular risk andmay be a reflection of atheroscleroticburden. Aortic compliance is reduced inadults who are at increased risk of pre-mature vascular disease.2 Aortic compli-ance is rarely measured in clinicalpractice because direct measurement isinvasive. However, compliance can bemeasured non-invasively with ultra-sound3 or magnetic resonance4 imaging,although clinical utility of these tech-niques are limited. The most commonlyutilized method of assessing complianceis to measure pulse wave velocity (pulse

Test-Retest Reproducibility of theWideband External Pulse DeviceCara A. Wasywich, FRACPWarwick Bagg, MDGillian Whalley, MScJames Aoina, BScHelen Walsh, BScGreg Gamble, MScAndrew Lowe, PhDNigel Sharrock, MBChBRobert Doughty, MD

Department of Medicine, Faculty of Medical and Health Sciences, The University of Auckland,Auckland, New Zealand

KEY WORDS: Compliance,Echocardiography, Stroke volume

ABSTRACTThe Wideband External Pulse (WEP)device has been developed to non-inva-sively assess arterial compliance, andfrom that datum derive other clinicalparameters such as stroke volume (SV)and cardiac output. The aim of thisstudy was to assess the test-retest repro-ducibility of the WEP device under stan-dard conditions in normal subjects andto compare the results to echocardio-graphically derived SV measurements.Mean pooled SV were not significantlydifferent for WEP (72.1 mL, P=0.98, 73.4mL, P=0.54) and Echo (87.1 mL,P=0.29) on 2 occasions, however WEPSV was on average 15 mL less than echoSV. Limits of agreement were wide forWEP SV (±44 mL) compared to echoSV (±16 mL), as were coefficients ofvariation (CV) (WEP CV 32%, echo CV6.9%). In conclusion, although averageperformance of the WEP device is rea-sonable, test-retest reproducibility ispoor in normal individuals under stan-

wave velocity is higher in non-compliantblood vessels). This method requires theplacement of sensors over the carotidand femoral arteries so that the delaybetween the 2 signals can be measured.

The wideband external pulse device(WEP) has been developed to allowquick, simple and non-invasive measure-ment of aortic compliance, and from thecompliance data, to allow the calculationof other physiological parameters suchas stroke volume and cardiac output.The WEP device consists of a piezoelec-tric sensor that is placed over thebrachial artery underneath a standardautomated blood pressure cuff. The sen-sor is connected to a computer whichrecords signals (with the cuff inflated tosuprasystolic pressures) that correspondto pulse wave reflections from the aortictree (Figure 1). The relative amplitudeof these waves, SS1 and SS2, provide ameasure of aortic compliance.5

AIMThe aim of this study was to establishthe test-retest reproducibility of theWEP device under standard conditions,and to compare WEP derived measure-ments of stroke volume to a validatedmethod (echocardiographic measure-ment of stroke volume).

METHODSTwenty-six healthy volunteers wererecruited from staff and students at theUniversity of Auckland. All studies werecarried out in the CardiovascularResearch Laboratory, Department ofMedicine, University of Auckland, NewZealand. Subjects provided writteninformed consent and the University ofAuckland Ethics Committee approvedthe study protocol.

Subjects fasted for 8 hours prior toeach of the 2 examinations. Both exami-nations were carried out at the same

Vol. 4, No. 2, 2004 • The Journal of Applied Research246

Table 1. Demographic Data for Study Participants

Variable Mean SD

Male (n,%) 11 (44)

Age (years) 24.8 4.0

BMI (kg/m2) 24.1 3.4

Blood pressure 122/74 11/7

Fasting glucose (mmol/L) 4.6 0.3

Total cholesterol (mmol/L) 4.6 0.8

BMI body mass index

Table 2. Stroke Volume Assessed by Echo and WEP on Two Different Visits*

Visit 1 Visit 2 Pooled Difference Paverage (difference)

WEP

C (intraday) 1.59 (0.30) 1.53 (0.36) 1.56 (0.32) 0.06 (0.19) 0.12

C (interday) 1.58 (0.33) 1.56 (0.30) 1.59 (0.28) 0.02 (0.29) 0.77

SV (intraday) 72.0 (15.7) 72.1 (20.4) 72.1 (15.9) 0.1 (17.9) 0.98

SV (interday) 72.0 (15.7) 74.7 (22.6) 73.4 (15.9) 2.7 (22.2) 0.54

EchocardiographySV (interday) 88.0 (16.1) 86.1 (16.9) 87.1 (15.9) 1.8 (8.4) 0.29

*Values are mean (SD). SV indicates stroke volume (mL); SS1, Incident wave; SS2, first reflected wave; and C, com-pliance.

time of day. Subjects rested in a quiet,temperature controlled room in a semi-recumbent position. Blood pressure wasestablished as the mean of 3 readingsusing the automated Dynamap sphyg-momanometer. After ensuring that anadequate signal could be obtained, care-ful note of the position of the WEPdevice was made. The WEP device (IlixirLtd, Auckland, New Zealand) was fittedon the same arm for each study and carewas taken to ensure that the electrodewas placed at the same height above theantecubital fossa on each occasion.Three baseline WEP recordings weremade. The subject was then rested for 5minutes and a further 3 recordings weremade.

Technique for Measuring Compliance Suprasystolic recordings from the sensorwere passed through an analogue signalpre-conditioning filter and digitized. Thedigital data was displayed on a PC usingsoftware written in Java by Ilixir Limited(Auckland, NZ). The waveform was thenanalyzed using Matlab Software

(Mathworks Inc., Natick, Mass, USA).The amplitude of the SS1 and SS2 waveswere measured and compliance (C) wascalculated using the formula C=1.018(SS1/SS2)0.257. This formula has beenderived from preliminary data obtainedfrom a small number of patients.6 Strokevolume (SV) was calculated using theformula SV=C x PP where PP is thepulse pressure measured non-invasivelyusing a Dynamap.

A transthoracic echocardiogram(see below) was performed after whichan additional set of WEP recordingswere made.

All data from the WEP device wasstored on a laptop computer. At the con-clusion of the data collection phase allWEP data was analyzed in randomorder in a blinded fashion by AndrewLowe. For each recording the largestcontiguous interval of the signal collect-ed, when the cuff pressure was above 15mmHg higher than the systolic pressure,was identified. The interval was thensegmented where each waveform seg-ment corresponded to an individual

The Journal of Applied Research • Vol. 4, No. 2, 2004 247

Figure 1. Idealized suprasystolic waveforms.

SS1 indicates the arterial pulse wave; SS2, reflectance wave from the abdominal aorta; andSS3, a reflectance wave from the peripheral circulation (occurs at the beginning of diastole).

heartbeat. These waveform segmentswere time-shifted, such that the maxima(peaks) of the SS1 waves were aligned.The mean signal level at each point intime of the aligned waveform segmentswas then calculated and this generated amean beat.

The amplitudes of the SS1 and SS2waves within the mean beat were takenas the amplitude difference between thepeak of the wave and the followingtrough.

Echocardiographic MethodsStandard 2-D and m-mode images of theleft ventricle were obtained using aPhillips HDi5000 ultrasound machineand 3.5 MHz transducer and measure-ments of left ventricular size and functionmade (Philips, Bothell, Wash, USA).Measurement of stroke volume was per-formed using Doppler and 2D data(SV=cross sectional area x velocity timeintegral). The LVOT diameter was meas-ured from a frozen mid-systolic frameand measured using the leading-edgemethodology. Pulsed wave Doppler wasused to measure the time-velocity inte-gral of the LVOT blood flow, with thesample volume placed immediatelybelow the aortic annulus.7

All echo images were digitallyacquired and analyzed off-line using a

dedicated cardiac measurement package.Each echo variable was measured in trip-licate and the average value used for theanalysis.

At the conclusion of one of thestudy visits subjects had blood speci-mens drawn to measure fasting bloodglucose and lipids. These were assayedby a commercial laboratory (DiagnosticMedlab, Auckland, New Zealand) usingstandard laboratory techniques.

StatisticsThe methods of Bland and Altman8

were used to calculate the limits ofagreement for WEP and echocardio-graphically defined measurements andthe coefficient of repeatability (magni-tude of test-retest repeatability). Paireddata were plotted, and least squaresregression lines fitted, to ensure consis-tency of agreement or repeatability overthe entire range of values.

RESULTSOne patient had a biscuspid aortic valveidentified on transthoracic echocardio-gram and was subsequently excludedfrom the analysis. Demographic data forthe study population are presented inTable 1.

Compliance and SV assessed byWEP are similar when assessed on 2 dif-

Vol. 4, No. 2, 2004 • The Journal of Applied Research248

Figure 2. Limits of agreement for WEP (intraday).

SV1 indicates Pre echo stroke volume; and SV2, Post echo stroke volume.

∆(C

omp 2-

Com

p 1)

∆(S

V2-

SV

1m

l/min

)

The Journal of Applied Research • Vol. 4, No. 2, 2004 249

ferent occasions on the same day (intra-day) and on separate days (interday)(Table 2). Pooled interday echocardio-graphic stroke volume is on average 15mL greater than WEP stroke volume.

Limits of agreement8 were calculat-ed comparing WEP measurements onthe same day (Figure 2), and on differ-ent days (Figure 3). Interday limits ofagreement for compliance measured byWEP are ±0.6 with a coefficient of vari-ation (CV) of 26%, however, interdayWEP stroke volume measurements arebroader (±44 mL), CV 32% meaningthat for any given stroke volume meas-ured by WEP, the true value may be asmuch as 44 mL above or below themeasured value. The least squares

regression line slopes downward, sug-gesting that the error is greater at lowermeasured stroke volumes. Converselythe interday limits of agreement forechocardiographic stroke volume areless (±16 mL), CV 6.9% and the leastsquares regression line is horizontal, sug-gesting similar accuracy at all values ofstroke volume measured.

DISCUSSIONThe WEP device has been developed asa novel technique to non-invasivelyassess the compliance of blood vessels.Compliance data are then translatedinto the clinically useful parameters ofSV and CO. This study has demonstrat-ed that the average performance of

Figure 3. Limits of agreement for WEP and echo (interday).

Comp indicates compliance; and SV, stroke volume.

∆(C

omp 2-

Com

p 1)

∆(S

V2-

SV

1m

l/min

)

∆(S

V2-

SV

1m

l/min

)

WEP derived SV between 2 studies(either 2 studies on the same day or ondifferent days) was similar, but inferiorto echocardiography. Importantly, thelimits of agreement of the WEP-derivedSV are more than twice as great as thoseof SV derived by echocardiography.Reproducible echocardiography isdependent on multiple variables includ-ing operator experience, patient bodyhabitus, and heart rate; despite these fac-tors the test-retest reproducibility andCV for echocardiographically derivedstroke volume was clinically acceptablein this study, and in keeping with otherpublished data.9 It is disappointing thatthe limits of agreement using the WEPtechnique for measurement of SV arewide. One factor that may explain this isthe fact that calculation of SV from theWEP device depends on the repro-ducibility of 2 measurements, thesebeing the reproducibility of the compli-ance measurement and secondly thereproducibility of the blood pressuremeasurements used to calculate pulsepressure. It is known that blood pressureeven under controlled conditions canfluctuate significantly, and non-invasivemeasurement itself involves a significantrandom inaccuracy. Figure 3 illustratesthe fact that the addition of pulse pres-sure measured non-invasively results ina less precise measurement (limits ofagreement for compliance are tighterthan limits of agreement for SV meas-ured by WEP). Additionally the formulaapplied for the calculation of SV from Cand PP is simplistic and does not repre-sent well the actual physics involved.The formula assumes a closed, fully elas-tic, statically linear system, whereas inpractice the system under considerationis somewhat more complex.

WEP derived SV was consistentlylower that echocardiographicallyderived SV. This may reflect relativeinaccuracy of the formula used to derivecompliance with the WEP technique.

This device remains investigational andthe formula used is based on correla-tions with hemodynamic data in a smallnumber of cases.6 Accuracy of compli-ance data is likely to be improved aslarger numbers of subjects are studiedand the formula for calculating compli-ance refined.

Other techniques, such as flowmediated dilatation of an artery, are alsorelatively imprecise in individualpatients, but have proved useful forresearch purposed in large groups.Further research and development ofthis device may validate its role in thisarea, and potentially improve clinicalapplicability.

CONCLUSIONAlthough the average performance ofthe WEP device is reasonable, particu-larly for compliance measurements,WEP derived stroke volume measure-ments based on non-invasive pulse pres-sure measurement have poor test-retestreproducibility in normal individualsunder standard conditions. Limits ofagreement are at least twice as wide asthose for echocardiographically derivedstroke volume. Clinical application ofthis technology in individual patients atthis stage is likely to be limited, howeverit may emerge as a promising techniquefor compliance measurement.

REFERENCES1. Celermajer DS. Endothelial dysfunction: does

it matter? Is it reversible? JAm Coll Cardiol.1997;30:325-333.

2. Lehmann E, Hopkins K, Gosling R. Aorticcompliance measurements using Dopplerultrasound: in vivo biochemical correlates.Ultrasound MedBiol. 1993;19:683-710.

3. Wright JS, Cruickshank JK, Kontis S, Dore C,Gosling R. Aortic compliance measured bynon-invasive Doppler ultrasound: descriptionof a method and it’s reproducibility. ClinScience. 1990;78:463-468.

4. Mohiaddin RH, Underwood SR, Bogren HG,et al. Regional aortic compliance studied bymagnetic resonance imaging: the effects of

Vol. 4, No. 2, 2004 • The Journal of Applied Research250

The Journal of Applied Research • Vol. 4, No. 2, 2004 251

age, training, and coronary artery disease. BrHeart J. 1989;62:90-96.

5. Blank S, West J, Muller F, et al. Widebandexternal pulse recording during cuff defla-tion: a new technique for evaluation of thearterial pressure pulse and measurement ofblood pressure. Circulation. 1988;77:1297-1305.

6. Sharrock N. Information for patent applica-tion. 2001:31-2.

7. Lewis J, Kuo L, Nelson J. Pulsed Dopplerechocardiographic determination of strokevolume and cardiac output: clinical validationof two new methods using the apical window.Circulation. 1984;70:425-431.

8. Bland JM, Altman DG. Statistical methodsfor assessing agreement between two meth-ods of clinical measurement. Lancet.1986;i:307-310.

9. Liang YL, Teede H, Kotsopoulos D, et al.Non-invasive measurements of arterial struc-ture and function: repeatablility, interrelation-ships and trial sample size. Clin Science.1998;95:669-79.