Professional Documents
Culture Documents
Background: We determined the reliability and re- nificant association between either of the AC parameters
peatability of measurements of arterial compliance (AC) and visit (first or second), time (first or second measure at
and gender- and age-specific normal ranges for a healthy the same visit), and visit-by-time (the interaction of the
European population. two preceding factors), suggesting that order of measure
Methods: Three hundred eight healthy volunteers from had no effect on the final value. Analysis of reliability was
seven sites were evaluated. Two measurements were taken used to develop a strictly parallel model estimate of unbi-
during the first visit, repeated on a second visit 1 to 4 ased reliability. Both intravisit and intervisit estimates of
weeks later. We used the HDI/PulseWave CR-2000 for reliability indicate good repeatability of measure and were
measurements of AC. significant (P ⬍ .0001). The AC values were found to
differ significantly by age group, with an inverse associa-
Results: Intravisit measurements, taken 5 min apart, tion between each of the AC parameters and age group.
differed by less than 3% (range, 0.36% to 2.97%). All
intervisit measures differed by less than 4% (range, 0.24% Conclusions: Measurement of the arterial waveform
to 3.67%); none of these differences was statistically sig- with the CR-2000 system is highly reproducible in healthy
nificant. All correlation coefficients for pairs of AC pa- subjects. Am J Hypertens 2005;18:65–71 © 2005 Amer-
rameters measured 5 min apart at the same visit were ican Journal of Hypertension, Ltd.
significant at P ⬍ .0001. Paired AC parameters at visit 1
and 2 were highly correlated (P ⬍ .0001). Repeated mea- Key Words: Arterial compliance, arterial pulsewave
sures GLM (general linear model) failed to detect a sig- contour, repeatability, reliability, normal range.
E
xtensive clinical evidence has shown that the pres- overt CVD. Therefore, it has been proposed as an indicator
ence of risk factors for cardiovascular disease of cardiovascular status and a means of estimating success
(CVD) is strongly associated with impairment in or failure of treatment.2
arterial compliance (AC).1,2 The AC is impaired when One of the main problems with the use of noninvasive
CVD is present.3,4 Decreased AC develops gradually, re- methods developed for estimation of AC has been repeat-
flects vascular damage, and predicts the appearance of ability and accuracy of the measured parameters. The
Received February 19, 2004. First decision August 7, 2004. Accepted Medical and Surgical Sciences (DR), Breccia, Italy; Servizio de Medi-
August 13, 2004. cina Interna General, Hospital General (ACP), Barcelona, Spain; Kerck-
From the The Brunner Institute for Cardiovascular Research, hoff Clinic (CH), Bad Nauheim, Germany; and The Queen’s University
Department of Medicine (RZ, MS, MB) and Epidemiology Unit (MB), of Belfast, Department of Therapeutics and Pharmacology (GM), Belfast,
Wolfson Medical Center and Tel Aviv University, Tel Aviv, Israel; Northern Ireland, UK.
UZ-Gent, The University Hospital, Department of Cardiovascular Dis- Address correspondence and reprint requests to Prof. Reuven
eases (DD), Gent, Belgium; University of Muenster Medical School Zimlichman, Chief of Medicine and Hypertension, Wolfson Medical
(K-HR), Muenster, Germany; University of Breccia, Department of Center, P.O. Box 5, Holon 58100, Israel; e-mail: zimlich@post.tau.ac.il
Characteristics Data
Age (y) 33.8 ⫾ 10.9
Gender (% females) 68.2
Race (% white) 98.7
Self-reported family history of cardiovascular disease/risk factors
Myocardial infarction, angina pectoris (%) 64 (20.8)
Cerebrovascular accident (stroke) (%) 76 (24.7)
Diabetes mellitus (%) 64 (20.8)
Hypertension (%) 125 (40.6)
Alcohol consumption
Any alcohol consumption (%) 97 (31.5)
Type of alcohol consumed
None (%) 211 (68.5)
Beer (%) 61 (19.8)
purpose of the present study was to estimate the AC sures. This calculation was based on assumptions drawn from
parameters using the HDI CR-2000 Cardiovascular Pro- prior experience with AC parameters measured using the
filing System (Hypertension Diagnostics, Inc., Eagen, HDI/PulseWave CR-2000. Study centers recruited a total of
MN) in healthy subjects. We also aimed to determine the 308 participants who were present for both visits and had two
extent to which these measures are repeatable when taken measurements performed in each visit.
5 min apart at a given visit, and also when taken at two
different visits 1 to 4 weeks apart. In addition, gender- and Subjects
age-specific population normal ranges for a healthy Euro- All participants were healthy volunteers recruited in the
pean population were developed based on these measure- following sites: Belfast, Northern Ireland (75 subjects;
ments. 12.4%), Gent, Belgium (22 subjects; 3.6%), Holon, Israel
(131 subjects; 21.7%), Muenster, Germany (119 subjects;
Methods 19.7%), Brescia, Italy (61 subjects; 10.1%), Barcelona,
Spain (102 subjects; 16.9%), Bad Nauheim, Germany (95
Study Design
subjects; 15.7%).
The reproducibility of arterial compliance parameters us- All 308 healthy subjects were between the ages of 15
ing the HDI/PulseWave CR-2000 was measured twice, and 80 years with a body mass index (BMI) of less than 30
5 min apart during the same visit, and at a second visit. kg/m2. Most subjects were recruited mainly at their work
Comparisons of measures were made between the first and sites as a result of local advertisement. Subjects were
second measures taken on the same visit at each of the two required to complete an Assessment Profile Questionnaire
visits. In addition, each measure taken on the first visit was that queried health history, especially history of cerebro-
compared to each of the measures taken on the second visit. vascular, cardiovascular and peripheral vascular disease,
A sample size of 295 individuals was calculated to be ade- hypertension, diabetes mellitus, and medications. Individ-
quate to provide 80% power to detect a difference of 4%, uals self-reporting any of the above-mentioned diseases or
adjusting the significance level to account for repeated mea- medications, especially those that might affect hemody-
AJH–January 2005–VOL. 18, NO. 1 RELIABILITY AND REPEATABILITY OF ARTERIAL COMPLIANCE MEASUREMENTS 67
Table 2. Mean absolute and relative differences in placed on the contralateral arm. The obtained waveforms
measurements of arterial compliance parameters (mean of 30-sec recording) were calibrated to the systolic
taken 5 min apart at the same visit and diastolic cuff pressures. A computer-based third order
four-element Windkessel model of the circulation was
Variable Mean Absolute Mean Relative
Pair Difference Difference used to match a diastolic pressure decay of the waveforms
and quantify changes in terms of SVR, C1, C2, and L.7
Visit 1
The SVR was calculated as MAP divided by cardiac
LAE 0.388 2.46%
SAE 0.212 2.97% output. The MAP was derived from waveform analysis,
Visit 2 integrating the area under the curve and calculating the
LAE 0.070 0.44% mean area of recordings during 30 sec. Cardiac output was
SAE 0.026 0.36% calculated from a multivariate algorithm using cardiac
None of the comparisons was statistically different. ejection time derived from the radial artery waveform. The
method and its results have been validated and described
in depth previously.5–7
Table 4. Correlation coefficients for measures of 0.81 for LAEI and 0.86 for SAEI (P ⬍ .001). Both
arterial compliance parameters: measurements intravisit and intervisit agreement of measure were signif-
taken 5 min apart of the same visit icant (P ⬍ .0001).
The AC values were found to differ significantly by age
Correlation
Visit Variable Pair Coeficient
group as displayed in Fig. 1. An inverse association is
identified between each of the AC parameters and age
1 LAE1 0.69 group. Values differ significantly by gender. Age and
SAE1 0.72
2 LAE1 0.67 gender normal values are displayed in Table 6. “Normal”
SAE1 0.75 was defined as those above the fifth percentile for a given
gender and age group.
All correlations significant at P ⬍ .0001, two tailed.
Bland-Altman plots were used to describe agreement
between measures of LAE and SAE. When the two meth-
ods are identical, the expectation is that ⬃68% of differ-
considered significant at P ⬍ .05. All analyses were per- ences will be distributed symmetrically around 0 within
aging, there is a gradual deposition of collagen and extra- The Windkessel model was used to compare upper and
cellular matrix.14,15 lower body sites, arm and leg, and showed that compliance
A critical question that has to be answered using vari- values were higher for the leg than for the arm.19 These
ous devices to record noninvasively pulse wave contour is: results may question the validity and utility of Windkes-
To what extent are the results repeatable? Inconsistent sel-derived variables by the absence of between-site cor-
modes of application of the sensor may be a potential relations when measured in the same subject. We believe
problem that may jeopardize potential use of a specific that leg arteries reflect different pattern of disease. Patients
method. Several publications reviewed the reproducibility with severe CVD may have minimal arterial disease in
data in the literature for various devices.14 –17 In a study their legs and vice versa. A study that we completed
that used the HDI equipment, the augmentation index, recently confirms that there is a very weak correlation
which was determined noninvasively, compared favorably between arterial disease in the legs when compared to
with the Windkessel model.18 other sites and that peripheral vascular disease, measured
It should be mentioned that neither C1 nor C2 have as ankle brachial index, does not correlate with the clinical
been shown yet to be predictive of events or outcome; manifestations of disease nor with AC measured at the
however, they have been found to correlate with risk radial artery. It seems that factors that affect arterial dis-
factors and with extent of disease. This implies that mea- ease in the legs are not clear enough.
surement of AC, using the Windkessel model, may corre- Another problem has been interference with measure-
late with the degree of disease of the blood vessel. ment accuracy due to movements of the hand or its mus-
70 RELIABILITY AND REPEATABILITY OF ARTERIAL COMPLIANCE MEASUREMENTS AJH–January 2005–VOL. 18, NO. 1
Table 6. Normal values for large and small arterial ent occasions in the same individual was found. Unfortu-
compliance indices by age group and gender nately, there are very few studies that compare the validity
and repeatability of different methods and devices for
Women Men
evaluation of AC.14 –16
Age Age Sources of potential bias must be thoroughly examined
Range LAEI SAEI Range LAEI SAEI in the evaluation of study results in all research. The
15–19 ⬎16.1 ⬎7.7 15–19 ⬎16.1 ⬎7.7 present study, like any study in which the participants are
20–29 ⬎15.5 ⬎7.1 20–29 ⬎15.5 ⬎7.1 recruited through local ads, is vulnerable to volunteer bias.
30–39 ⬎14.9 ⬎6.6 30–39 ⬎14.9 ⬎6.6
40–49 ⬎14.3 ⬎6.0 40–49 ⬎14.3 ⬎6.0 However, for volunteer bias to influence results, we must
50–59 ⬎13.7 ⬎5.5 50–59 ⬎13.6 ⬎5.4 assume that association exists between AC and willing-
60–69 ⬎13.0 ⬎4.8 60–69 ⬎13.1 ⬎4.9 ness to volunteer for participation in the study. Although it
⬎70 ⬎12.3 ⬎4.1 ⬎70 ⬎12.5 ⬎4.3 is true that volunteers tend to be healthier than nonvolun-
teers, this is of low influence in a study designed to recruit
FIG. 2 (A) Bland Altman plot of between-visit measures of large artery elasticity. (B) Distribution of differences, large artery elasticity. (C)
Bland Altman plot of between visit measures of small artery elasticity. (D) Distribution of differences, small artery elasticity.
AJH–January 2005–VOL. 18, NO. 1 RELIABILITY AND REPEATABILITY OF ARTERIAL COMPLIANCE MEASUREMENTS 71
reported data is well documented, there is no reason to Coppack SW, Gosling RG: Relation between number of
assume that this potential inaccuracy in reporting was in cardiovascular risk factors/events and non-invasive Doppler ultra-
sound assessment of aortic compliance. Hypertension 1998;32:565–
some way associated with AC measurements. Therefore, 569.
there is no reason to believe that people with poor AC 2. Cohn JN, Finkelstein S, McVeigh G, Morgan D, LeMay L, Robin-
would be more likely to inaccurately report medical his- son J, Mock J: Noninvasive pulse wave analysis for the early
tory than individuals with good AC. The impact of infor- detection of vascular disease. Hypertension 1995;26:503–508.
mation bias on the results of the present study can be 3. Hirai T, Sasayama S, Kavasaki T, Yagi S: Stiffness of systemic
arteries in patients with myocardial infarction. Circulation 1989;80:
considered inconsequential. 78 – 86.
We evaluated the change in AC that occurs with age. 4. London GM, Marchais SJ, Safar ME, Genest AF, Guerin AP,
Our data confirm the evidence that AC decreases with Metivier F, Chedid K, London AM: Aortic and large artery com-
age.12,13,15,16 It can be seen that the decrease in AC with pliance in end-stage renal failure. Kidney Int 1990;37:137–142.
increasing age is steeper in small (C2) than in large AC 5. Shargorodsky M, Wainstein G, Gavish D, Leibovitz E, Matas Z,
Zimlichman R: Treatment with rosiglitazone reduces hyperinsulin-
(C1). However, it is well known that in the elderly sys- emia and improves arterial compliance in patients with type 2
tolic, and not diastolic, BP is the major problem. This diabetes mellitus. Am J Hypertens 2003;16:617– 622.