Sex-Specific Association of Age
with Carotid Artery Distensibility:
Multi-Ethnic Study of Atherosclerosis
Dhananjay Vaidya, Ph.D.,1Susan R. Heckbert, M.D., Ph.D.,2
Bruce A Wasserman, M.D.,1and Pamela Ouyang, M.B., B.S.1
Background: Older women have a higher prevalence of systolic hypertension than do men; however, whether or
not this relates to arterial properties, such as distensibility coefficient (DC), is not known. We examined whether
the association of carotid artery DC with age differed by sex in the Multi-Ethnic Study of Atherosclerosis
Methods: B-mode ultrasound-measured carotid diameters and brachial pressures were obtained from 6359
participants (53% female, 38% white, 12% Chinese, 27% black, 22% Hispanic, aged 45–85 years) of the MESA
baseline examination. The within-individual slopes of 2log(diameter) vs. blood pressure fit using mixed models
(MM) are interpreted as the DC, and interaction terms are interpreted as differences in DC. The MM calculation
allows for correction of the confounding caused by the association of age, sex, and race with blood pressure, the
denominator in the calculation of DC.
Results: DC was associated with age, sex, and race (all p<0.001). Women had a greater age-related lowering of
DC compared to men (2.52·10-5vs. 2.16·10-5/mm Hg lower DC per year of age, p=0.006). Mean diameter of
carotid arteries was greater with age (p<0.001); this association also was significantly stronger in women
compared to men (0.24% vs. 0.14% larger mean carotid diameter per year of age, p<0.001).
Conclusions: Greater stiffening and enlargement of arteries are seen in older women compared to older men.
This implies that the afterload on the heart of older women is likely to be greater than that among older men.
sion1and heart failure2that are associated with aging. Older
women have a greater prevalence of hypertension than do
older men.3It is not known, however, if this difference of
aging by sex is because of sex differences in the stiffness of
arteries at different ages.
The material properties of arteries (related to stiffness and
compliance) may be measured invasively,4but this is not
possible in large studies where sex by age interaction can be
assessed. Therefore, direct noninvasive imaging of arteries is
often used. The systolic and diastolic diameters of the carotid
artery obtained by ultrasound and the simultaneous measure
of systolic and diastolic brachial artery pressure5allow the
investigator to determine the distensibility coefficient (DC) in
he stiffening of arteries throughout life is thought to
underlie the pathogenesis of such diseases as hyperten-
the linear approximation. If epidemiologic associations of
these material properties with other variables are sought,
however, variables that are associated with only one or the
other of the component variables (i.e., systolic or diastolic
diameters, systolic or diastolic pressures) may erroneously
appear to be associated with the material properties of the
arteries.6Because the diameters and pressures are a part of a
nonlinear calculation of DC, usual linear regression methods
to adjust for confounding is not appropriate. A recent publi-
cation shows association of incident coronary disease and
stroke with carotid artery stiffness measures, and this asso-
ciation is either fully or greatly attenuated by linear adjust-
ment by blood pressure (BP) measures, likely due to
We have analyzed Multi-Ethnic Study of Atherosclerosis
(MESA) data using mixed model (MM) statistical methods
that appropriately correct for such confounding to test for
1Johns Hopkins University, Baltimore, Maryland.
2Department of Epidemiology, University of Washington, Seattle, Washington.
JOURNAL OF WOMEN’S HEALTH
Volume 21, Number 5, 2012
ª Mary Ann Liebert, Inc.
differences in the material properties of arteries between
groups using these methods. Our aim is to determine if there
is a difference in the relationship of age with the distensibility
of arteries by sex, adjusting for BP levels in the population-
Materials and Methods
We used baseline data from the MESA. The design of
MESA, a study of the prevalence and progression of sub-
clinical atherosclerosis, has been described previously.8
Briefly, 6814 individuals, all free of clinical cardiovascular
(Caucasian, Chinese, Hispanic, and African American), were
enrolled at six U.S. sites. Of these, 6359 participants had ca-
rotid ultrasound data adequate for the calculation of carotid
artery distensibility and compliance.
B-mode ultrasound and brachial BP measurements
Detailed methods have been published previously.9A 20-
section ofthedistal right commoncarotidartery wasacquired
using a Logiq 700 machine (General Electric Medical Sys-
tems). Single measures of brachial systolic and diastolic BPs
were obtained simultaneously (DINAMAPP System, GE
Medical systems). This single measurement corresponded
well with a three-measure average of seated BP (Pearson
correlation coefficients: systolic 0.78, diastolic 0.74, pulse
pressure 0.78). Ultrasound images were analyzed and inter-
preted by the MESA ultrasound reading center, located at
software to extract the carotid artery diameters in the images.
Systolic and diastolic diameters were determined as the
largest and smallest diameters during the cardiac cycle. For
the DC (calculated using the linear approximation in Sup-
plementary Table S1 available online at www.liebertonline
.com), the repeat-study class correlation coefficient was 0.71,
and the repeat-reading class correlation was 0.68.9
The traditional method of calculating DC is as follows:
BP within the artery, D represents the difference between
systolic and diastolic measurements, and D represents the
appropriate mean value ofthediameter. In practice, Dmaybe
approximated to the systolic or diastolic diameter.5
We used MM regression where diameters and BP were
modeled; thus, DC and differences in DC were estimated
within the model, appropriately adjusted for confounders.
This allows for disentangled estimation of the associations of
independent variables with diameter alone and with disten-
sibility. This contrasts with calculation of DC outside the
statistical model, when the association and confounding of
diameter and distensibility are entangled.
Differences in DC were estimated in the multivariate
model, which allowed for adjustment for the confounding of
BP by covariates. For a single binary covariate (e.g., F=0 if
male, 1 if female), the model would be specified as:
for the ith condition (diastole or systole) of the jth individual.
Random effects are estimated for the average log-diameters
(log-diameter at the average of the systolic and diastolic
to pressure slope for every individual, which is determined
using diastolic and systolic pressure (c and g, respectively).
The coefficients b3and b4can be interpreted as the fixed-effect
association of the binary covariate with the mean diameter
and the fixed-effect difference in slopes (i.e., the material
properties of the artery) associated with the covariate. This
specification can be generalized to any categorical or linear
A detailed explanation of the MM regression is provided as
Supplementary material (available online at www.liebertonline
.com). Models with age (centered at 65 years), sex, race, and
height were evaluated. First-order (two-way) interactions be-
tween age, sex, and race were also evaluated. Only significant
interactions (p<0.05) were retained.
The MM-estimated DC for every individual from best lin-
earunbiased predicted (BLUP)values wascompared with the
traditionally calculated DC. The mean of the MM DC was
numerically close to the mean of traditionally calculated DC
(0.0025 vs. 0.0025), although the standard deviation (SD) was
smaller (0.0006 vs. 0.0011), showing a shrinkage of variation
due to the MM estimation. The Spearman correlation of the
traditionally calculated DC with model-fitted DC was 0.81. In
sensitivity analysis, the final regression models were further
adjusted for education (as a proxy for socioeconomic status
[SES]) and the use of BP medications to check for possible
confounding. Another model excluding individuals using BP
medications was also evaluated.
The MESA population has been described previously.10
The demographic and clinical profile of the participants in-
cluded in this analysis are shown in Table 1.
Age, sex, and race associations
of the distensibility coefficient
The MM simultaneously assesses the association of cov-
ariables with mean arterial diameter and the D.C. Age, sex,
race, and the age-sex interaction had significant associations
in the regression analysis (Table 2). In addition, the arterial
diameter at the average of systolic and diastolic pressures is
larger in older persons, but this association is significantly
also seen, with Chinese having the smallest diameter and
Hispanics having the largest diameter. Supplementary Table
S2 (available online at www.liebertonline.com) shows that
adjusting for the traditional risk factors, smoking, total cho-
lesterol, and high-density lipoprotein cholesterol (HDL-C)
levels, and current diabetes, does not change the levels of
significance for any of these results. Supplementary Table S3
AGE AND CAROTID DISTENSIBILITY 517
(available online at www.liebertonline.com) shows various
models assessing the association of carotid diameters and
pressures with age, sex, and race. These traditional linear
models with DC as the dependent variable include BP as a
covariate, which results in overadjustment.
The age-sex interaction found in the mixed model is
graphically illustrated in Figure 1, where model-calculated
DCs, adjusted for race distribution and mean height
(166cm) are plotted against age among women and men.
In both men and women, the DCs are lower with greater
age. Although the DCs in men and women have similar
values in the age range 45–60 years (confidence intervals
[CI] overlap), at older ages, the DCs in women are lower
than those in men. Because numerical measures of DC are
not intuitive, Figure 1 illustrates the sex interaction in age
equivalents. A difference of 10 chronologic years results in
a reduced distensibility equivalent of 11.6 years in women
compared to 10 years in men.
In sensitivity analyses, there was no difference in signifi-
cant associations in the MM after addition of education and
use of BP medications to the model. When analyses were re-
stricted to 4279 individuals who were not using BP medica-
tions, all associations remained significant except for one: the
difference between DC of men compared to women at age 65
was no longer significant (beta=3.59·10-5, p=0.16); how-
ever, the magnitude of the point estimate was similar to that
tabulated for the main analysis. The association of tradition-
ally calculated DC with age, sex, race, and the age-sex inter-
action iscontrasted withtheMMresults inthesupplementary
material (available online at www.liebertonline.com).
We have shown significant sex-age interaction in associa-
tion with both DC and mean carotid artery diameter in a large
well-characterized multi-ethnic population-based study. Our
findings are consistent with those of Redfield et al.,11who
found that vascular and systolic and diastolic ventricular
elastance was more steeply associated with age in women
than men in Olmsted County, Minnesota.
Using appropriate MM statistical methods, we were able to
show associations of DC with covariates with two distinct
advantages over using traditional calculations of DC: (1) the
associations in our models were adjusted for BP levels, which
Table 1. Demographic and Cardiovascular Risk Characteristics of Sample, by Race/Ethnicity
Caucasian American Chinese American African AmericanHispanic American
SBP (mm Hg)
DBP (mm Hg)
Type 2 DM
Mean–standard deviation (SD) or number (%).
DBP, diastolic blood pressure; DM, diabetes mellitus, using American Diabetes Association 2003 criteria; HDL-C, high-density lipoprotein
cholesterol; SBP, systolic blood pressure; TC, serum total cholesterol.
Table 2. Association of Age, Sex, and Race with Carotid Artery Distensibility Coefficient and Diameter
Difference in DC (2*slope)
% Larger carotid diameter
Covariate Beta coefficients (95% CI)
Sex (male vs. female) at 65 years of age
Age (difference per year among females)
Difference in age-associated slope between men and women
Height (difference per cm of height)
Chinese vs. White
African American vs. White
Hispanic vs. White
Overall test for race/ethnicity
Intercept: mixed model intercept coefficient corresponding to group difference between geometric means in percent; slope: mixed model
slope coefficient corresponding to distensibility coefficient difference by group; 2*slope=difference in distensibility coefficient; to allow for
judging of magnitude of coefficients, the standard deviation (SD) of distensibility coefficient=1.1·10-3; for example, the beta coefficient for
age in women is -5.04·10-5/mm Hg. This means that, on average, keeping other covariates constant, if 2 women differ 1 year in age, the
older would have DC lower by -5.04·10-5/mm Hg, which corresponds to 0.045 SD units. The SI unit for distensibility, kPa-1=7.5mm Hg-1.
518VAIDYA ET AL.
is not possible for traditionally calculated DC, where BPs are
part of the nonlinear calculation and (2) using our analytic
model, the MM also provides estimates of association with
mean carotid artery diameter, which has been shown to be
associated with carotid atherosclerosis.12Our results show
that MM analysis can detect significant associations that have
been previously published using within-participant calcula-
tion of DC9in the same study dataset.
Although our analysis reproduced the findings of Blaha
et al.9for some covariates, we also contribute several impor-
tant new findings. We have shown interethnic differences in
mean carotid artery diameters that have not been reported
previously. Whereas the sex differences in arterial diameters
are expected, the age-sex interaction is novel. Older women
have larger arterial diameters than young women, and this
Also, the age-sex interaction in DC means that older women
have stiffer arteries than older men. To give a sense of the
magnitude of the differences, we use the unadjusted hazard
ratios for stroke associated with lower distensibility reported
by Yang et al.7: because of differences in DC, women who are
10 years older would have a 30.4% higher hazard of stroke
than their younger counterparts, whereas men with the same
age difference would have a 25.6% higher hazard of stroke. In
addition, the afterload on the heart of older women is likely to
be greater than that among older men. This may help to ex-
plain why older women predominate among patients with
heart failure with preserved systolic function.13It may also
have apart inexplaining whyalthough the absolute mortality
rate is higher in men than women, the rate of increase of total
heart disease mortality in men >45 years rises less steeply
than in women.14
Because this is a cross-sectional, observational study, we
are circumspect about clinical and treatment-related impli-
cations of this finding. It is noteworthy that current antihy-
in women compared to men in the prevention of death in
heart failure.15It is possible that drugs that improve arterial
distensibility might be more efficacious in prevention of heart
failure mortality in women.16
Strengths and limitations
The large multi-ethnic database of both men and women
with a large range of ages collected using rigorous mea-
surement standards is a significant strength of this study. A
limitation of this study is that the arterial diameter imaging
was of the carotid artery, but pressure measurement was
performed in the brachial artery. There was no invasive
gold standard measurement of distensibility. However, the
aim of study was to analyze the kind of noninvasive
measurement practical in large-scale studies, such as
MESA, which have been reported in current published lit-
erature,9rather than comparison with a gold standard
We have shown sex and ethnic differences in the diameter
and distensibility of the carotid artery, with adjustment for
confounding by BP. Furthermore, we have shown a sex-age
interaction that can be hypothesized to have worse conse-
quences of arterial stiffness among older women compared
to older men. The biologic underpinnings of these strong
findings and whether they translate into differences in heart
failure and other cardiovascular outcomes remain to be de-
termined by future research.
This research was supported by contracts N01-HC-95159
through N01-HC-95169 from the National Heart, Lung, and
Blood Institute. We thank the other investigators, the staff,
and the participants of the MESA study for their valuable
contributions. A full list of participating MESA investigators
and institutions can be found at www.mesa-nhlbi.org. D.V.
was supported by grant Number UL1 RR 025005 from the
National Center for Research Resources, a component of the
National Institutes of Health.
The authors have no conflicts of interest to report.
1. Safar ME. Systolic hypertension in the elderly: Arterial wall
mechanical properties and the renin-angiotensin-aldosterone
system. J Hypertens 2005;23:673–681.
2. Kass DA. Age-related changes in venticular-arterial cou-
pling: Pathophysiologic implications. Heart Fail Rev 2002;
3. Martins D, Nelson K, Pan D, Tareen N, Norris K. The effect
of gender on age-related blood pressure changes and the
prevalence of isolated systolic hypertension among older
adults: Data from nhanes III. J Gend Specif Med 2001;4:10–
4. Salvi P, Lio G, Labat C, Ricci E, Pannier B, Benetos A. Vali-
dation of a new non-invasive portable tonometer for deter-
mining arterial pressure wave and pulse wave velocity: The
Pulsepen device. J Hypertens 2004;22:2285–2293.
5. Reneman RS, Meinders JM, Hoeks AP. Non-invasive ultra-
sound in arterial wall dynamics in humans: What have we
learned and what remains to be solved? Eur Heart J 2005;
Distensibility Coeff (per mmHg)
men (dashed lines) and women (solid lines), adjusted for race
and mean height (166cm), and their 95% confidence intervals
calculated using mixed model analysis.
Fitted age-distensibility coefficient relationship for
AGE AND CAROTID DISTENSIBILITY519
6. Laurent S, Cockcroft J, Van Bortel L, et al. Expert consensus Download full-text
document on arterial stiffness: Methodological issues and
clinical applications. Eur Heart J 2006;27:2588–2605.
7. Yang EY, Chambless L, Sharrett AR, et al. Carotid arterial
wall characteristics are associated with incident ischemic
stroke but not coronary heart disease in the Atherosclerosis
Risk in Communities (ARIC) study. Stroke 2012;43:103–108.
8. Bild DE, Bluemke DA, Burke GL, et al. Multi-Ethnic Study of
Atherosclerosis: Objectives and design. Am J Epidemiol 2002;
9. Blaha MJ, Budoff MJ, Rivera JJ, et al. Relationship of carotid
distensibility and thoracic aorta calcification: Multi-Ethnic
Study of Atherosclerosis. Hypertension 2009;54:1408–1415.
10. McClelland RL, Chung H, Detrano R, Post W, Kronmal RA.
Distribution of coronary artery calcium by race, gender, and
age: Results from the Multi-Ethnic Study of Atherosclerosis
(MESA). Circulation 2006;113:30–37.
11. Redfield MM, Jacobsen SJ, Borlaug BA, Rodeheffer RJ, Kass
DA. Age- and gender-related ventricular-vascular stiffening:
A community-based study. Circulation 2005;112:2254–2262.
12. Montalcini T, Gorgone G, Gazzaruso C, Sesti G, Perticone F,
Pujia A. Large brachial and common carotid artery diameter
in postmenopausal women with carotid atherosclerosis.
13. Bursi F, Weston SA, Redfield MM, et al. Systolic and diastolic
heart failure in the community. JAMA 2006;296:2209–2216.
14. Vaidya D, Becker DM, Bittner V, Mathias RA, Ouyang P.
Ageing, menopause, and ischaemic heart disease mortality
in England, Wales, and the United States: Modelling study
of national mortality data. BMJ 2011;343:d5170.
15. Shekelle PG, Rich MW, Morton SC, et al. Efficacy of angio-
tensin-converting enzyme inhibitors and beta-blockers in the
management of left ventricular systolic dysfunction accord-
ing to race, gender, and diabetic status: A meta-analysis of
major clinical trials. J Am Coll Cardiol 2003;41:1529–1538.
16. Zieman SJ, Melenovsky V, Clattenburg L, et al. Advanced
glycation endproduct crosslink breaker (alagebrium) im-
proves endothelial function in patients with isolated systolic
hypertension. J Hypertens 2007;25:577–583.
Address correspondence to:
Dhananjay Vaidya, Ph.D.
Assistant Professor of Medicine
Johns Hopkins University School of Medicine
1830 E. Monument Street, Suite 8028-A
Baltimore, MD 21287
520VAIDYA ET AL.