JOURNAL OF THE AMERICAN COLLEGE OF CARDIOLOGY
VOL. 75, NO. 8, 2020
ª 2020 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION PUBLISHED BY ELSEVIER
Trajectories of Age-Related Arterial Stiffness in Chinese Men and Women Yao Lu, MD, PHD,a,b,* Raimund Pechlaner, MD, PHD,c,* Jingjing Cai, MD, PHD,d,* Hong Yuan, MD, PHD,a,d,e Zhijun Huang, MD,a Guoping Yang, MD, PHD,a Jiangang Wang, MD, PHD,e Zhiheng Chen, MS,e Stefan Kiechl, MD,c,f Qingbo Xu, MD, PHDg,h
ABSTRACT BACKGROUND Arterial stiffening is central in the vascular aging process. Traditionally, vascular research has focused on atherosclerotic vascular disease, whereas arterial stiffness has not attracted similar attention. OBJECTIVES The purpose of this study was to assess lifetime trajectories of arterial stiffening in Chinese populations facing a high burden of cardiovascular disease, with a particular focus on age–sex interactions and potential determinants. METHODS This large-scale observational study comprised 2 independent cross-sectional population samples and 1 prospective cohort totaling 80,415 healthy subjects with brachial-ankle pulse wave velocity (baPWV) measurements available. Associations with potential risk conditions were analyzed using linear regression, linear random intercepts mixed models, and L1-regularized linear models. RESULTS The dynamics of age-dependent arterial stiffening differed in sexes, with stiffer vessel observed in men from adolescence to age 58 years and in women thereafter. The steeper increase in baPWV in women after menopause is partly explained by the fact that vascular risk factors are more strongly associated with arterial stiffness in women than in men. Age and systolic blood pressures were the strongest determinants of baPWV, whereas other vascular and metabolic risk factors, except low-density lipoprotein cholesterol, showed consistent associations of moderate strength. CONCLUSIONS The significant age–sex interaction in arterial stiffening provides an important clue of explanation for the heightened cardiovascular disease risk in postmenopausal women. Detailed knowledge on lifetime trajectories of arterial stiffening, and its potential risk factors is a prerequisite for the development of new prevention strategies counteracting vascular aging. (J Am Coll Cardiol 2020;75:870–80) © 2020 by the American College of Cardiology Foundation.
C
ardiovascular disease (CVD) has climbed to
of pathologies, including atherosclerosis, vascular
the number 1 cause of death worldwide (1).
calcification
Arterial stiffening is central in the vascular
end-stage renal disease, matrix degeneration, and
aging process underlying CVD and reflects a variety
genuine vascular aging featured by smooth muscle
and
inflammation,
vasculopathy
of
From the aCenter of Clinical Pharmacology, The Third Xiangya Hospital, Central South University, Changsha, China; bDepartment of Life Science and Medicine, King’s College London, London, United Kingdom; cDepartment of Neurology, Medical University of Innsbruck, Innsbruck, Austria; dDepartment of Cardiology, The Third Xiangya Hospital, Central South University, Changsha, China; eHealth Management Center, The Third Xiangya Hospital, Central South University, Changsha, China; fVASCage, Research Centre for Promoting Vascular Health in the Ageing Community, Innsbruck, Austria; gSchool of Cardiovascular Medicine and Sciences, King’s College London British Heart Foundation Centre, London, United Kingdom; and the hDepartment of Cardiology, Listen to this manuscript’s
the First Affiliated Hospital, School of Medicine, Zhejiang University, Zhejiang, China. *Drs. Lu, Pechlaner, and Cai contributed
audio summary by
equally to this work. This study was supported by the excellence initiative VASCage (Centre for Promoting Vascular Health in the
Editor-in-Chief
Ageing Community, project number 868624) of the Austrian Research Promotion Agency FFG (COMET program–Competence
Dr. Valentin Fuster on
Centers for Excellent Technologies) funded by the Austrian Ministry for Transport, Innovation and Technology; the Austrian
JACC.org.
Ministry for Digital and Economic Affairs; and the federal states Tyrol (via Standortagentur), Salzburg, and Vienna (via Vienna Business Agency); as well as by the National Science Foundation of China (81800393, 81870171, and 81570271) and the Science and Technology Planning Project of Hunan Province (2019RS2014). Dr. Kiechl is CSO of VASCage, a competence center of the Austrian Research Promotion Agency, with 50 university and company partners. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Manuscript received September 23, 2019; revised manuscript received November 27, 2019, accepted December 10, 2019.
ISSN 0735-1097/$36.00
https://doi.org/10.1016/j.jacc.2019.12.039
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
871
Lifetime Age Trajectories of Arterial Stiffness
cell senescence and stiffening, fragmentation and
cerebral microbleeds, cognitive impairment,
ABBREVIATIONS
degeneration of elastic fibers, and collagen accumula-
and dementia as well as lacunar stroke (2,18).
AND ACRONYMS
tion (Central Illustration) (2–6). The heightened CVD
A recent meta-analysis reported a 15% in-
risk related to preeclampsia and prematurity has
crease in all-cause mortality for each 1 m/s
also been tied to sustainable arterial stiffening
increase in aortic PWV (15), and PWV ranks
afforded by hypertensive damage to the vessel wall
among the strongest predictors for mortality
and inadequate elastin synthesis (5,7–9). On top of
independent of conventional vascular risk
structural changes in vessel wall composition, arterial
factors (15,19).
ABI = ankle-brachial index baPWV = brachial-ankle pulse wave velocity
BP = blood pressure CVD = cardiovascular disease LASSO = least absolute
stiffening commonly comprises a functional compo-
Despite the outstanding clinical relevance
nent potentially amenable to pharmacological and
of arterial stiffening, measurement of the
operator
lifestyle interventions (2,5).
PWV has not yet found broad access to clin-
NAFLD = nonalcoholic fatty
ical routine nor attracted adequate research
liver disease
interests. Only few and mainly small-scale
PWV = pulse wave velocity
SEE PAGE 881
Measurement of the pulse wave velocity (PWV) is a
shrinkage and selection
studies have compared PWVs in men and
validated approach to noninvasively quantify arterial
women and obtained inconsistent results (20–24).
stiffness (2,10). It represents the speed of the pressure
Here, we present lifetime trajectories of arterial
wave propagating along the artery and is directly
stiffness from adolescents to the elderly along with a
correlated to the rigidity of the vessel with a higher
comprehensive assessment of its predictors and spe-
velocity indicating stiffer vessels (2,10). Different
cial consideration of age–sex interactions, using 2
measurement techniques
carotid-
large cross-sectional studies and a longitudinal study
femoral pulse wave velocity (cfPWV) (applanation
from China, which total an unprecedented 80,415
tonometry),
individuals.
brachial-ankle
focus
on
pulse
the
wave
velocity
(baPWV) (oscillometry), and segment-specific PWVs (11). cfPWV was the prime technique when distensi-
METHODS
bility measurements accessed published medical data in the mid-80s (12) and was recommended as a
STUDY
marker of asymptomatic organ damage by the 2013
population comprised 78,000 individuals from a
European Society of Hypertension guidelines for
mixed urban and rural area who visited the Health
the management of arterial hypertension. baPWV,
Management Center in the Third Xiangya Hospital of
which reflects the stiffness of both the central
Central South University (Changsha), the largest
conduit and peripheral arteries, was used more
medical institution in central China, between 2004
recently, but already offers the largest resource of
and 2016 (Online Figure 1). The participants came
PWV data. It is an inexpensive technique suitable
from hundreds of institutions in Changsha to check
for large-scale epidemiological studies and clinical
their health status and had a diverse socioeconomic
routine, and is less operator-dependent than cfPWV.
background (public services employees, workers,
Measurement of baPWV bears the advantages of
self-employed persons, farmers, and others). Partici-
high
simple
pation in the health examinations was on a voluntary
simultaneous
basis, but was encouraged by the employers and was
recording of the ankle-brachial index (10), and it
offered free of charge. A total of 9,899 individuals
correlates well with cfPWV in European and Asian
underwent a more detailed examination protocol and
populations (11,13,14).
formed the cross-sectional population I, and the
reproducibility, low
handling
in
obese
time
subjects,
demand, and
POPULATION. The
cross-sectional
study
The clinical sequela of arterial stiffening is sum-
remainder formed the cross-sectional population II.
marized in the Central Illustration. In brief, it is a main
The prospective cohort was enrolled at the same
cause of isolated systolic hypertension in the elderly
center (2012 to 2016) and comprised 2,415 individuals,
and increases left ventricular afterload, giving rise to
most with repeated annual measurements. The study
cardiac remodeling, ventricular hypertrophy, dia-
protocols were approved by the Ethics Committee of
stolic dysfunction, atrial fibrosis, and impaired coro-
the Central South University.
nary perfusion (2,5,6,15). Serious consequences are
CLINICAL CHARACTERISTICS. All participants un-
heart failure and aortic aneurysm development based
derwent
on local stiffness gradients (2,8,16,17). Arterial stiff-
completed detailed questionnaires. Age, sex, height,
ening also augments central blood and pulse pres-
weight, current medication from pill bottles, previous
sure, which damages the microcirculation of organs
medical diagnoses, exercise, smoking history, and
with low vascular resistance, especially the kidney
alcohol consumption were recorded according to
and brain, resulting in chronic kidney disease,
standard protocols. Assessment and definitions of
a
routine
clinical
examination
and
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
Lifetime Age Trajectories of Arterial Stiffness
C E NT R AL IL L U STR AT IO N Lifetime Age Trajectories of Arterial Stiffness Extracellular matrix degeneration Elastic fiber fragmentation Collagen accumulation Smooth muscle cell senescence
End-organ damage: Chronic kidney disease Cerebral microangiopathy Cognitive impairment and dementia Lacunar stroke Cerebral microbleeds Arterial stiffness
Lifetime Trajectories of Arterial Stiffening Advanced atherosclerosis
Systolic hypertension Augmentation of central blood pressure
2,500 Pulse-Wave Velocity, cm/s
872
Vascular calcification and inflammation Vasculopathy of chronic kidney disease Diabetic macroangiopathy Contributors to arterial stiffness
Heart failure with and without preserved ejection fraction Cardiac remodeling Aortic aneurysm formation
2,000
Consequences of arterial stiffness
1,500
1,000 25
50 58
75
Age, Years Sex
Male
Female
Lu, Y. et al. J Am Coll Cardiol. 2020;75(8):870–80.
Contributors to and consequences of arterial stiffness and lifetime trajectories of arterial stiffness as measured by brachial-ankle pulse wave velocity (baPWV) by age and sex in 80,415 healthy Chinese subjects. The vascular aging processes depicted on the left reduce arterial distensibility, causing hemodynamic changes that inflict damage on end organs, in particular the heart and organs with low vascular resistance like the kidney and brain, as summarized on the right. From adolescence to age 58 years, men featured higher baPWV than women, whereas a steeper increase in baPWV in women resulted in higher female baPWV afterwards.
the
clinical
characteristics
are
detailed
in
the
the brachial and posterior tibial artery waveforms.
Online Appendix.
Measurements
MEASUREMENT OF baPWV. The ankle-brachial index
average values of the left- and right-side assessments
(ABI) and baPWV and were measured simultaneously
was calculated. Two trained technicians performed
with an automatic waveform analyzer (BP-203 RPE
all measurements, and the interobserver and intra-
were
performed
twice,
and
the
III, Omron Health Medical, Dalian, China). After a
observer coefficients of variation were 4.9% and
minimum rest of 5 min in the supine position, 4 cuffs
7.8%, respectively, well within the range of 3.6% to
were wrapped around upper arms and ankles and
8.4% and 3.8% to 10.0%, respectively, reported in
connected to a plethysmographic sensor (volume
previous studies (10). baPWV measurements may be
pulse form) and oscillometric pressure sensor. ABI
biased in patients with severe atherosclerosis in the
was measured as the ratio of the ankle systolic blood
lower legs (10,25). Therefore, in case of an ABI <0.9,
pressure (BP) divided by the brachial systolic BP.
we only considered the baPWV of the other side. In
Pressure waveforms were recorded at both the
case of bilateral ABIs <0.9, subjects were excluded
brachial and tibial arteries to assess the transmission
from the analysis. Moreover, substantial side differ-
time between the initial rises in these waves. The
ences in the baPWV of more than 10 m/s indicate
baPWV was calculated using the formula (La Lb)/
problems with measurement. Participants with such
OTba, in which La is the distance from the heart to
pronounced differences were excluded (w0.01%)
the ankle, Lb the distance from the heart to the
as
brachium, and OTba the transmission time between
55 m/s (<0.01%) or age younger than 10 years (10,25).
were
participants
with
a
baPWV
beyond
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
Lifetime Age Trajectories of Arterial Stiffness
T A B L E 1 Associations of Clinical and Laboratory Characteristics With Vessel Stiffness (baPWV) in 2 Cross-Sectional and 1 Longitudinal
Community-Based Study From Changsha, Adjusted for Age, Sex, and Age–Sex Interaction Cross-Sectional Study I Estimate
p Value
Cross-Sectional Study II
Longitudinal Study
n
Estimate
p Value
n
Estimate
p Value
n
5,430
Anthropometry Height, cm
37.2
<0.0001
5,514
37.7
<0.0001
40,718
0.2
0.285
Weight, kg
2.7
0.480
5,519
3.6
0.012
40,721
0.3
0.058
5,432
Body mass index, kg/m2
13.4
<0.0001
5,514
14.5
<0.0001
54,097
1.4
0.002
5,434
Obesity, yes/no
38.8
0.012
5,514
45.2
<0.0001
54,097
16.6
0.022
5,434
Waist circumference, cm
24.9
<0.0001
9,620
22.4
<0.0001
65,766
0.8
<0.0001
5,409
10.2
<0.0001
58,978
3.9
0.280
4,585
8.5
0.226
5,330
2.0
0.460
33,314
5.7
0.061
4,602
2.0
0.496
5,269
Lifestyle and social status Current smoking, yes/no
NA
Regular alcohol drinking, yes/no Physical activity, every day vs. other
30.6
<0.0001
5,405
20.2
<0.0001
34,717
Marriage, married vs. single/divorced
59.3
<0.0001
8,950
91.9
<0.0001
40,150
Occupation, worker vs. other
8.9
0.119
7,285
12.8
<0.0001
49,484
NA
Annual income, <50 KRMB vs. $50 KRMB
48.4
0.102
174
22.9
0.093
1,008
NA
Education, university degree vs. other
NA
NA 8.6
NA
0.003
5,489
Classic vascular risk factors Hypertension, yes/no
175.1
<0.0001
9,812
177.7
<0.0001
67,583
11.8
<0.0001
5,489
Systolic blood pressure, mm Hg
130.6
<0.0001
9,736
130.5
<0.0001
66,736
1.1
<0.0001
5,440
Diastolic blood pressure, mm Hg
92.4
<0.0001
9,737
94.6
<0.0001
66,739
1.5
<0.0001
5,440
Pulse pressure, mm Hg
92.5
<0.0001
9,736
90.1
<0.0001
66,733
0.9
<0.0001
5,440
Heart rate, beats/min
72.8
<0.0001
9,879
67.4
<0.0001
67,916
0.7
<0.0001
3,449
Diabetes, yes/no
86.4
<0.0001
9,812
94.9
<0.0001
67,583
22.4
<0.0001
5,489 5,486
Fasting blood glucose, mmol/l
42.8
<0.0001
9,832
40.3
<0.0001
67,574
2.5
0.005
HDL cholesterol, mmol/l
13.2
<0.0001
9,833
7.6
<0.0001
67,611
5.6
0.123
5,486
LDL cholesterol, mmol/l
2.0
0.392
9,818
2.6
0.004
67,585
2.8
0.063
5,486
Triglycerides, mmol/l
28.9
<0.0001
9,837
26.0
<0.0001
67,618
0.3
0.796
5,486
5.3
0.802
9,879
1.4
0.872
67,231
Lipid-lowering drugs, yes/no
NA
Emerging risk factors and other Creatinine, mmol/l
8.0
0.004
9,836
8.3
<0.0001
67,607
0.0
0.709
Renal disease, yes/no
178.6
<0.0001
9,836
163.7
<0.0001
67,607
7.8
0.228
5,485
Uric acid, mmol/l
24.6
<0.0001
9,836
28.1
<0.0001
67,605
0.0
0.613
5,485
NAFLD, yes/no
50.7
<0.0001
9,812
45.5
<0.0001
67,583
ALT, U/l
14.8
<0.0001
9,833
13.6
<0.0001
67,545
0.2
0.039
AST, U/l
16.3
<0.0001
7,425
13.1
<0.0001
51,039
Albumin, g/l
24.2
<0.0001
9,834
26.8
<0.0001
67,560
5,485
NA 5,486
NA 1.0
0.037
5,486
WBC, 109/l
NA
NA
1.7
0.045
5,486
Hemoglobin, g/l
NA
NA
0.3
<0.001
5,486
Effects represent the difference in PWV associated with a 1-SD higher level of or versus reference category of each row variable. For longitudinal analyses, associations of baseline level of each row variable with change in PWV per year are shown. Analyses are adjusted for age, sex, and their interaction. Cross-sectional results are derived from linear regression, and longitudinal results from linear random intercepts mixed models. Variation in number of subjects used for each variable are due to missing data. 1 SD of PWV was 289 cm/s in cross-sectional study I, 287 cm/s in cross-sectional study II, and 358 cm/s in the longitudinal study. Smoking was not computed in cross-sectional study I because 100% were smokers in this population sample. ALT ¼ alanine transaminase; AST ¼ aspartate transaminase; baPWV ¼ brachial-ankle pulse wave velocity; HDL ¼ high-density lipoprotein; KRMB ¼ 1,000 Renminbi; LDL ¼ low-density lipoprotein; NA ¼ variable not available in this population; NAFLD ¼ nonalcoholic fatty liver disease; WBC ¼ white blood cell count.
In
100
individuals,
both
baPWV
and
cfPWV
were measured. LABORATORY
trajectories of baPWV by age for men and women MEASUREMENTS. Fasting
venous
blood samples were collected and were immediately processed and analyzed at the clinical laboratory of Third
density estimates using Gaussian kernels. Smooth
Xiangya
Hospital,
as
detailed
in
the
Online Appendix.
were calculated using generalized additive models with the smoothness of fit determined by generalized cross-validation. The crossing point of sex-specific age trajectories was defined as the point of minimal predicted difference of the generalized additive models fits along a finely spaced grid.
STATISTICAL ANALYSIS. Distributions of baPWV by
Cross-sectional associations with baPWV were
age and sex groups were illustrated using kernel
analyzed using linear regression, adjusted for age,
873
874
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
Lifetime Age Trajectories of Arterial Stiffness
sex,
Multivariable
(Online Figure 1). In the prospective cohort, 5,489
(independent) associations were investigated using
and
age–sex
data points were available for analysis. Population
L1-regularized linear regression (LASSO) (26). The
characteristics are summarized in Online Table 1.
hyperparameter lambda governing model sparsity
AGE AND SEX TRAJECTORIES OF ARTERIAL STIFFNESS.
was determined by cross-validating mean squared
Figures 1A and 1B display absolute values and distri-
error along the regularization path. The 1-SE rule was
butions of baPWV for each decade of age in men and
used to select the model with the fewest nonzero
women. Whereas baPWV was low in adolescents, it
parameters whose error is within 1 SE of the error of
gradually increased and became more disperse
the
mean
throughout adulthood. Importantly, women had
squared error. The following variables featuring
more dispensable vessels than men until after
model
with
interaction.
minimal
cross-validated
>20% missing values were excluded from the LASSO
menopause, whereas elderly women exhibited stiffer
analysis to facilitate adequate numbers of subjects
vessels than men had. Generalized additive models
available for complete case analysis: height, weight,
suggest a crossing point of age trajectories at about 58
body mass index, obesity, aspartate transaminase,
years of age (Figure 1C). The age–sex interaction was
alcohol consumption, physical activity, marriage,
highly significant (p ¼ 4.46 10151 ).
education, occupation, income, white blood cell
ASSOCIATIONS BETWEEN POPULATION CHARACTERISTICS
count, and hemoglobin.
AND baPWV. Cross-sectional relationships are sum-
Longitudinal associations with baPWV in the pro-
marized in Table 1. Effects represent the difference in
spective cohort were analyzed using linear mixed
baPWV associated with a 1-SD higher level of each
models on change in baPWV. Adjustment variables
variable adjusted for age, sex, and the age–sex inter-
included were the same as used for cross-sectional
action. Findings were highly consistent in the 2 in-
analyses in addition to baseline baPWV and time
dependent
difference in years. The effect of each variable on
associations were observed for hypertension, blood
changes in baPWV per year from baseline was calcu-
and pulse pressure, heart rate, diabetes, renal disease
lated using an interaction effect of the variable of
and various components of the metabolic syndrome,
interest with time difference from baseline.
body mass index and central obesity, smoking, renal
cross-sectional
populations.
Positive
Associations with baPWV in age and sex subgroups
function, uric acid, nonalcoholic fatty liver disease
were analyzed using linear regression. Age groups
(NAFLD), and albumin levels, whereas inverse asso-
were defined by a cut-off equivalent to the crossing
ciations emerged for height, physical activity, and
point of sex-specific age trajectories. Within both age
high-density lipoprotein (HDL) cholesterol as well as
groups, models included age as a continuous variable,
being married and pursuing an occupation with a high
sex, and appropriate terms for interaction by sex.
level of physical activity (workers). No consistent
Main effects for men and women as well as interac-
associations were found for regular alcohol drinking,
tion effects were extracted simultaneously using
low-density lipoprotein cholesterol, and use of lipid-
linear combinations of parameters.
lowering drugs.
Continuous variables were scaled to unit variance
Most associations were successfully replicated in
before analysis to determine effects for a 1-SD dif-
the longitudinal cohort (Table 1, right-hand columns:
ference. Missing values were managed by performing
effects represent 1-year changes in baPWV associated
complete case analyses separately for each variable,
with a 1-SD–higher baseline level of population char-
and the respective numbers of subjects considered is
acteristics). This was true for hypertension, blood and
stated in Table 1.
pulse pressure, heart rate, diabetes, body mass index,
The p values are 2-sided, and an alpha level of 0.05
obesity and central obesity, HDL cholesterol, alanine
was used to define statistical significance. We did not
transaminase as a laboratory surrogate of NAFLD,
formally correct for multiple comparisons, but repli-
albumin, and social variables. Replication failed for
cated key findings in the independent population
smoking, physical activity, renal function and renal
samples. All analyses were conducted using R version
disease, and uric acid. These findings, however,
3.6.0
should be interpreted in view of a lower statistical
(R
Foundation
for
Statistical
Computing,
Vienna, Austria).
power in the longitudinal analyses.
RESULTS
LASSO based on a combined dataset of the 2 cross-
Multivariable associations were tested by means of sectional studies and baseline measurements of the STUDY POPULATION. The overall study population
longitudinal study (n ¼ 66,058 subjects with com-
comprised 80,415 individuals age 14 to 96 years
plete variable information). The final selection of
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
Lifetime Age Trajectories of Arterial Stiffness
F I G U R E 1 Distribution of Mean baPWV by Age and Sex
C Pulse-Wave Velocity, cm/s
Pulse-Wave Velocity, cm/s
A 5,000 4,000 3,000 2,000 1,000
2,500
2,000
1,500
1,000
-2 0, n -3 = 0, 74 30 n = -4 5, 0, 14 n 6 40 = -5 14 0, ,9 19 50 n = 27 -6 ,8 0, 46 n = 60 1 8, -7 67 0, 2 n 70 = 8, -8 74 0, 6 n 80 = -1 4 ,11 00 3 ,n = 69 5
25
Male
Female
20
10
Sex
50 58 75 Age, Years
Age Category Sex
Male
Female
B 10-20, n = 74
20-30, n = 5,146
30-40, n = 14,919
40-50, n = 27,846
50-60, n = 18,672
60-70, n = 8,746
70-80, n = 4,113
80-100, n = 695
0.003
0.002
Density
0.001
0.000
0.003
0.002
0.001
0.000 1,000 3,000 5,000 1,000 3,000 5,000 1,000 3,000 5,000 1,000 3,000 5,000 Pulse-Wave Velocity, cm/s Sex
Male
Female
(A) Box plots showing baPWV by sex and decades of age; (B) kernel density estimates using Gaussian kernels to display an overlay of female and male baPWV distributions by decades of age; (C) smooth trajectories of mean baPWV by age and sex with 95% confidence band, based on generalized additive model fits, with the crossing point of sex-specific baPWV curves marked. A steeper increase of baPWV in young adult men resulted in higher male than female baPWV up to middle age, but at age 58 years the curves crossed, with women featuring higher baPWV afterwards.
875
876
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
Lifetime Age Trajectories of Arterial Stiffness
variables in the multivariable model is summarized in Table 2. Characteristics of the population sample used
T A B L E 2 Multivariable Associations Between Clinical/
Laboratory Characteristics and Vessel Stiffness (baPWV)
in this analysis closely matched those of the entire Coefficient (cm/s)
study population (Online Table 1). Intercept*
1,391.3
ASSOCIATIONS BETWEEN POPULATION CHARACTERISTICS
Systolic blood pressure
108.4
AND
Age
105.9
baPWV
ACCORDING
TO
AGE
AND
SEX.
Figures 2A and 2B plot effect sizes in women (x-axis)
Hypertension
59.5
against effect sizes in men (y-axis) in 2 age groups
Renal disease
46.7
according to the 58-year age cut-off. With few ex-
Heart rate
41.4
Fasting blood glucose
15.6
Age female interaction
15.1
Smoking
8.6
Uric acid
4.8
ceptions, characteristics were more strongly associated with baPWV in women (indicated by blue dots in Figures 2A and 2B). Findings were almost identical if the analysis was confined to the LASSO cohort (Online
Diabetes
Figures 2A and 2B).
Female
3.4
Waist circumference
2.8
DISCUSSION Arterial stiffness represents a promising therapeutic
3.5
Pulse pressure
1.9
Creatinine
1.0
Triglycerides
0.9
target to counteract vascular aging (2,5). For quantification of arterial stiffness, we assessed the baPWV, a valuable alternative to the traditional cfPWV, applicable to large-scale epidemiological studies. Both measures are highly correlated (r ¼ 0.46 to 0.75 [13,27] and r ¼ 0.56 in a subsample of our own population), show consistent patterns of association with risk factors (11,13,27), and qualify as excellent predictors for mortality (15,19). This study is the first to present lifetime trajec-
Effects are for a 1-SD increase or versus reference category. This analysis uses a combined dataset (n ¼ 66,058, see the Methods section) with complete case analysis to estimate associations with PWV by an L1-regularized linear model (LASSO). The following variables were not considered because more than 20% of values were missing: height, weight, body mass index, obesity, alcohol consumption, physical activity, marriage, education, occupation, income, aspartate transaminase, white blood cell count, and hemoglobin. The following variables were not selected by LASSO: diastolic blood pressure, HDL cholesterol, LDL cholesterol, lipid-lowering drugs, NAFLD, and alanine transaminase. *The intercept is different from the other variables listed because it does not reflect any association with PWV, but rather the estimated value of PWV when all the other variables are zero. Abbreviations as in Table 1.
tories of arterial stiffening from adolescents to the elderly in a very large community-derived dataset of
our study demonstrating a greater association with
80,415 individuals (Figure 1) with a particular focus on
baPWV
age–sex interactions (Figure 1C). As a key finding and
(Figures 2A and 2B). This finding accords with the
for
almost
all
risk
factors
in
women
in close agreement with previous studies, vessels
more injurious effects of several vascular risk factors
were stiffer in men from early adulthood onward with
like hypertension and diabetes in women compared
a near linear increase in PWV during life (5,20–24).
with men for heart failure (8) and CVD (32). Pre-
Women, in contrast, showed a substantial augmen-
clinical studies also suggest mechanistic differences
tation of the age-related increase in arterial stiffness
in vascular aging and arterial stiffening in men and
after menopause resulting in a curve-linear increase
women (5,29,33).
in baPWV and crossing of the male age trajectories at
As to the age–sex interaction, similarities exist
an age of about 58 years (Figure 1C). In line, a
with atherosclerosis, which, at the carotid level, starts
number of previous studies have demonstrated
about 10 years earlier in men compared with women
accelerated
menopause
(34). In the decade after menopause, the sex gap in
(5,10,20,24,28–30) and suggested a 6- to 10-year
atherosclerosis prevalence narrows; however, the
transition period to achieve an unfavorable post-
plaque burden remains lower in women throughout
menopausal vessel status (30), which fits well to the
life (34). Atherosclerosis is 1 and maybe not the most
crossing point of sex-specific age trajectories in our
compelling source of arterial stiffening in humans,
study (mean menopausal age in the survey area, 49
given its predilection site in the intima, whereas the
years). Potential reasons for the steeper increase in
tunica media is the vessel’s load-bearing structure
arterial stiffness after menopause include hormonal
mainly responsible for its biomechanical properties
changes, oxidative stress, and body iron accumula-
(2). Arterial stiffness is a comprehensive marker of
tion, as well as a higher susceptibility to conventional
vascular aging, subsuming the injurious effects of
vascular risk factors accumulating after menopause
diverse pathologies including vascular calcification
(5,8,29–31). The latter issue is directly supported by
and inflammation, vasculopathy of end-stage renal
vascular
stiffening
with
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
Lifetime Age Trajectories of Arterial Stiffness
F I G U R E 2 Associations Between Risk Factors and baPWV by Sex and Age
A 150
Hypertension
Stronger in males
Renal disease
Systolic blood pressure
100
Diastolic blood pressure
Effect in Males
Diabetes Heart rate Obesity
50
0
Pulse pressure
NAFLD
Fasting blood glucose Albumin Waist Uric acid Triglycerides Alcohol drinking BMI AST Lipid lowering drugs ALT Smoking Annual income (<50 vs. >50) Creatinine Weight LDL cholesterol HDL cholesterol Occupation (worker vs. other)
Stronger in females
Physical activity (everyday vs. other) Height
–50
Marriage (married vs. single/divorced)
–50
0
50
100 150 Effect in Females
200
250
B Systolic blood pressure
150
Stronger in males Diastolic blood pressure Pulse pressure
Hypertension
Renal disease
Annual income (<50 vs. >50)
Diabetes
100 Effect in Males
Heart rate
NAFLD
50
Triglycerides Fasting blood glucose
Alcohol drinking
0
Marriage (married vs. single/divorced) Albumin Uric acid Creatinine Waist ALT BMI AST Smoking Obesity LDL cholesterol
HDL C Physical activity (everyday vs. other)
Stronger in females
Weight
Occupation (worker vs. other)
–50
Height
–50
Lipid lowering drugs
0
50
100 150 Effect in Females
Significant Interaction by Sex
200
250
No Significant Interaction by Sex
(A) Individuals age <58 years; (B) individuals age $58 years. Dark blue dots indicate significantly stronger effects in women, and red dots in men. Within both age groups, models included age as a continuous variable, sex, and appropriate terms for interaction by sex. Effects represent difference in baPWV per 1-SD increase or versus reference group in cm/s. Effects of risk factors were stronger in female than male participants for most risk factors, both before and after the age of 58 years. ALT ¼ alanine aminotransferase; AST ¼ aspartate aminotransferase; BMI ¼ body-mass index; HDL ¼ high-density lipoprotein; LDL ¼ low-density lipoprotein; NAFLD ¼ nonalcoholic fatty liver disease; (<50 vs. >50) ¼ income greater than vs. less than 50,000 Renminbi.
877
878
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
Lifetime Age Trajectories of Arterial Stiffness
disease, matrix degeneration, and genuine vascular
glomerular filtration rate <60 ml/min/1.73 m 2 in our
aging (2–6).
study. Effect sizes related to diabetes, however, were
Overall, arterial stiffening may be a key driver of
more variable: þ7% (baPWV) in the current Chinese
the enhanced cardiovascular risk burden of women
population compared with þ20% and þ25% (cfPWV)
after menopause. On top of having stiffer vessels,
in studies from the United States and Brazil (40,41).
consequences of arterial stiffening have been re-
Potential explanations include differences in patient
ported to be more serious in women than in men.
characteristics and management as well as effects by
Women develop greater increases in pulse pressure,
ethnicity.
and the predictive capacity of PWV for all-cause
The extent to which each individual predictor of
mortality is up to 2-fold higher in women (5,22,35).
baPWV in this study generalizes to cfPWV measure-
Moreover, relevant consequences of arterial stiff-
ments is unclear, but in light of their strong correla-
ening like aortic aneurysm rupture, heart failure
tion, good agreement may be expected. Validation
with
coronary
studies from Europe and China yielded broadly
microvascular dysfunctions are more common in
consistent patterns of associations with vascular risk
women (5,8).
factors and outcomes for both techniques, even
preserved
ejection
fraction,
and
Absolute values of baPWV in our study are
though the strength of association may be higher for
strikingly similar to published reference values from
cfPWV regarding organ damage and for baPWV
Asia, but are higher than those observed in Euro-
regarding risk factors and coronary calcium, at least
pean populations (11,28,36). Our study is confirma-
in Europeans (11,14,42).
tory (2,6,10,11), in that: 1) age, hypertension, blood
Overall, our study adds significantly to the previ-
and pulse pressure, metabolic diseases (diabetes,
ous published data in that it reports associations from
glucose level, and central obesity), uric acid, and
a large dataset (for some risk predictors exceeding the
renal function all were positively related and regu-
size of the largest available meta-analysis) and repli-
lar physical activity, height, HDL cholesterol, and
cates risk factor associations in a prospective cohort
measures of higher social status all were inversely
with multiple measurements of baPWV. Only few
related to arterial stiffness; 2) low-density lipopro-
studies so far have presented longitudinal data on
tein cholesterol as a key risk factor of atheroscle-
baseline risk factor levels and changes of PWV over
rosis emerged as unrelated; and 3) age and systolic
time (36).
BP exhibited by far the strongest relationships STUDY
(Table 1).
STRENGTHS AND
LIMITATIONS. As
main
Uncertainties still surround the role of smoking in
strengths, high statistical power and inclusion of the
arterial stiffening, with some studies reporting pos-
entire relevant age range enabled thorough assess-
itive, null, or even negative associations (2,6,11). Our
ment of lifetime trajectories and age–sex interactions.
study yielded a positive cross-sectional association
Our study relies on by far the largest PWV dataset
of modest strength, which withstood multivariable
currently available and enables broad internal repli-
adjustment, but was not confirmed in the prospec-
cation by consideration of 3 independent population
tive analysis. Intriguingly, a strong and consistent
samples. Data accuracy for baPWV measurements was
association of NAFLD and its prime laboratory
high given very low numbers of implausible absolute
markers
amino-
baPWV values or side differences, demonstration of
transferase with arterial stiffness emerged. This
good in-house reproducibility, careful consideration
finding accords with a recent meta-analysis of 12
of heart rate and height as physical determinants of
studies (37,38). Insulin resistance, adipokine imbal-
PWV, and availability of ABI enabling exclusion of
ance, and chronic low-grade inflammation have been
individuals with severe peripheral artery disease.
suggested
Limitations include the number of missing values
alanine
as
transaminase/aspartate
potential
pathophysiological
links
(37,38).
(Table 1). It merits attention, however, that the char-
Renal disease and diabetes are key contributors to
acteristics of the subpopulation with complete vari-
vascular stiffness. In Australian patients, severely
able assessment and the entire population were
decreased kidney function (mean estimated glomer-
strikingly similar (Online Table 1, right-hand col-
ular filtration rate ¼ 19.5 ml/min/1.73 m 2) conferred an
umns).
approximately 20% higher cfPWV (39), which is in
community-derived (volunteers) but do not represent
good proportion to the 12% higher baPWV among
random samples, and study data may not necessarily
patients with any renal disease defined by estimated
apply
Moreover,
to
the
populations
study
of
populations
non-Chinese
were
heritage.
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
Lifetime Age Trajectories of Arterial Stiffness
Observational studies cannot infer causality for the
RAAS system (inhibition of angiotensin II) or with
associations obtained (Table 1). Finally, information
systemic inflammation (2,5).
on atrial fibrillation, which may affect PWV measurement and on the exact age of menopause were
ADDRESS
not available.
Kiechl, Department of Neurology, Medical Univer-
FOR
CORRESPONDENCE:
Dr.
Stefan
sity of Innsbruck, Anichstr. 35, A-6020 Innsbruck,
CONCLUSIONS
Austria. E-mail:
[email protected]. Twitter: @imed_tweets. OR Dr. Qingbo Xu, Department of
The current study unraveled a significant age–sex
Cardiology, the First Affiliated Hospital, Zhejiang
interaction in arterial stiffness with higher PWV in
University, 79 Qingchun Road, Hangzhou 310003,
men compared with premenopausal women but
Zhejiang, China. E-mail:
[email protected].
higher PWV in postmenopausal women compared with men, which is partly explained by differential effects of risk factors on the vascular aging process in both sexes. These data provide a valuable clue for a
PERSPECTIVES
better understanding of the enhanced CVD risk in
COMPETENCY IN MEDICAL KNOWLEDGE: Arterial stiff-
women after menopause. Availability of age trajec-
ening is an inherent feature of vascular aging. Although men
tories of arterial stiffness and more detailed infor-
have stiffer vessels than women until about 58 years of age, the
mation on potential underlying risk factors sets the
opposite is true thereafter, due to accelerated stiffening in
groundwork for the future clinical use of PWV mea-
women after menopause.
surements as a surrogate of vascular and biological age (2,6,15,19) and as an appealing target for preventive pharmacological and lifestyle interventions. Potentially effective therapeutic measures include caloric restriction (weight loss and induction of autophagy) (5,43,44), special diets and supplements (e.g., low-sodium high-potassium diet, polyphenol supplements, curcumin, and nicotinamide) (5,45,46), exercise programs (e.g., 12 weeks of moderate aerobic
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Vascular stiffness, as assessed by measurement of PWV, responds favorably to lifestyle and pharmacological interventions. TRANSLATIONAL OUTLOOK: Future studies in large patient cohorts should incorporate PWV measurements as a potential target for therapeutic intervention to reduce atherosclerotic risk.
exercise) (5,47,48), and drugs interfering with the
REFERENCES 1. WHO. World report on aging and health. 2015. Available at: https://apps.who.int/iris/bitstream/ handle/10665/186463/9789240694811_eng.pdf; jsessionid¼E95A9FA38A912FEAAC01A931A27757 14?sequence¼1. Accessed January 8, 2020. 2. Schellinger IN, Mattern K, Raaz U. The hardest part. Arterioscler Thromb Vasc Biol 2019;39: 1301–6. 3. Ungvari Z, Tarantini S, Donato AJ, Galvan V, Csiszar A. Mechanisms of vascular aging. Circ Res 2018;123:849–67. 4. Greenwald SE. Ageing of the conduit arteries. J Pathol 2007;211:157–72. 5. DuPont JJ, Kenney RM, Patel AR, Jaffe IZ. Sex differences in mechanisms of arterial stiffness. Br J Pharmacol 2019;176:4208–25. 6. Lyle AN, Raaz U. Killing me unsoftly: causes and mechanisms of arterial stiffness. Arterioscler Thromb Vasc Biol 2017;37:e1–11.
pathophysiology: why women are overrepresented in heart failure with preserved ejection fraction. Circulation 2018;138:198–205. 9. Chen CW, Jaffe IZ, Karumanchi SA. Preeclampsia and cardiovascular disease. Cardiovasc Res 2014;101:579–86. 10. Munakata M. Brachial-ankle pulse wave velocity: background, method, and clinical evidence. Pulse (Basel) 2016;3:195–204. 11. Baier D, Teren A, Wirkner K, Loeffler M, Scholz M. Parameters of pulse wave velocity: determinants and reference values assessed in the population-based study LIFE-Adult. Clin Res Cardiol 2018;107:1050–61. 12. Bouthier JD, De Luca N, Safar ME, Simon AC. Cardiac hypertrophy and arterial distensibility in essential hypertension. Am Heart J 1985;109: 1345–52.
velocity in association with target organ damage in the community-dwelling elderly Chinese: The Northern Shanghai Study. J Am Heart Assoc 2017; 6:e004168. 15. Vlachopoulos C, Aznaouridis K, Stefanadis C. Prediction of cardiovascular events and all-cause mortality with arterial stiffness: a systematic review and meta-analysis. J Am Coll Cardiol 2010; 55:1318–27. 16. Luft FC. Molecular mechanisms of arterial stiffness: new insights. J Am Soc Hypertens 2012; 6:436–8. 17. Raaz U, Zollner AM, Schellinger IN, et al. Segmental aortic stiffening contributes to experimental abdominal aortic aneurysm development. Circulation 2015;131:1783–95. 18. O’Rourke MF, Safar ME. Relationship between aortic stiffening and microvascular disease in brain and kidney: cause and logic of therapy. Hypertension 2005;46:200–4.
7. Odri Komazec I, Posod A, Schwienbacher M, et al. Aortic elastic properties in preschool children
13. Tanaka H, Munakata M, Kawano Y, et al. Comparison between carotid-femoral and brachial-ankle pulse wave velocity as measures
born preterm. Arterioscler Thromb Vasc Biol 2016; 36:2268–74.
of arterial stiffness. J Hypertens 2009;27: 2022–7.
19. Ohkuma T, Ninomiya T, Tomiyama H, et al. Brachial-ankle pulse wave velocity and the risk prediction of cardiovascular disease: an individual
8. Beale AL, Meyer P, Marwick TH, Lam CSP, Kaye DM. Sex differences in cardiovascular
14. Lu Y, Zhu M, Bai B, et al. Comparison of carotid-femoral and brachial-ankle pulse-wave
participant data meta-analysis. 2017;69:1045–52.
Hypertension
879
880
Lu et al.
JACC VOL. 75, NO. 8, 2020 MARCH 3, 2020:870–80
Lifetime Age Trajectories of Arterial Stiffness
20. Nethononda RM, Lewandowski AJ, Stewart R, et al. Gender specific patterns of age-related decline in aortic stiffness: a cardiovascular magnetic resonance study including normal ranges. J Cardiovasc Magn Reson 2015;17:20. 21. Cunha PG, Cotter J, Oliveira P, et al. Pulse wave velocity distribution in a cohort study: from arterial stiffness to early vascular aging. J Hypertens 2015;33:1438–45. 22. Smulyan H, Asmar RG, Rudnicki A, London GM, Safar ME. Comparative effects of aging in men and women on the properties of the arterial tree. J Am Coll Cardiol 2001;37:1374–80. 23. Vermeersch SJ, Rietzschel ER, De Buyzere ML, et al. Age and gender related patterns in carotidfemoral PWV and carotid and femoral stiffness in a large healthy, middle-aged J Hypertens 2008;26:1411–9.
the age-related increase in arterial stiffness in the early postmenopausal phase. Atherosclerosis 2006;184:137–42. 31. Kiechl S, Willeit J, Egger G, Poewe W, Oberhollenzer F. Body iron stores and the risk of carotid atherosclerosis: prospective results from the Bruneck study. Circulation 1997;96:3300–7. 32. Howard VJ, Madsen TE, Kleindorfer DO, et al. Sex and race differences in the association of incident ischemic stroke with risk factors. JAMA Neurol 2019;76:179–86. 33. Georgiopoulos GA, Lambrinoudaki I, Athanasouli F, et al. Free androgen index as a predictor of blood pressure progression and accelerated vascular aging in menopause. Atherosclerosis 2016;247:177–83.
population.
24. Lin HF, Liu CK, Liao YC, Lin RT, Chen CS, Juo SH. The risk of the metabolic syndrome on carotid thickness and stiffness: sex and age specific effects. Atherosclerosis 2010;210:155–9. 25. Ato D. Pitfalls in the ankle-brachial index and brachial-ankle pulse wave velocity. Vasc Health Risk Manag 2018;14:41–62. 26. Hastie T, Tibshirani R, Friedman JH. The Lasso. In: Hastie T, Tibshirani R, Friedman JH, editors. The elements of statistical learning: data mining, inference, and prediction. 2nd ed. New York: Springer, 2009:86ff. 27. Cheng YB, Li Y, Sheng CS, Huang QF, Wang JG. Quantification of the interrelationship between brachial-ankle and carotid-femoral pulse wave velocity in a workplace population. Pulse (Basel) 2016;3:253–62. 28. Ai ZS, Li J, Liu ZM, et al. Reference value of brachial-ankle pulse wave velocity for the eastern Chinese population and potential influencing factors. Braz J Med Biol Res 2011;44: 1000–5. 29. Hildreth KL, Kohrt WM, Moreau KL. Oxidative stress contributes to large elastic arterial stiffening across the stages of the menopausal transition. Menopause 2014;21:624–32. 30. Zaydun G, Tomiyama H, Hashimoto H, et al. Menopause is an independent factor augmenting
34. Kiechl S, Willeit J. The natural course of atherosclerosis. Part I: incidence and progression. Arterioscler Thromb Vasc Biol 1999;19:1484–90. 35. Regnault V, Thomas F, Safar ME, et al. Sex difference in cardiovascular risk: role of pulse pressure amplification. J Am Coll Cardiol 2012;59: 1771–7. 36. Taniguchi Y, Kitamura A, Shinozaki T, et al. Trajectories of arterial stiffness and all-cause mortality among community-dwelling older Japanese. Geriatr Gerontol Int 2018;18:1108–13. 37. Jaruvongvanich V, Chenbhanich J, Sanguankeo A, et al. Increased arterial stiffness in nonalcoholic fatty liver disease: a systematic review and meta-analysis. Eur J Gastroenterol Hepatol 2017;29:e28–35. 38. Villela-Nogueira CA, Leite NC, Cardoso CR, Salles GF. NAFLD and increased aortic stiffness: parallel or common physiopathological mechanisms? Int J Mol Sci 2016;17:E460. 39. Krishnasamy R, Tan SJ, Hawley CM, et al. Progression of arterial stiffness is associated with changes in bone mineral markers in advanced CKD. BMC Nephrol 2017;18:281. 40. Elias MF, Crichton GE, Dearborn PJ, Robbins MA, Abhayaratna WP. Associations between type 2 diabetes mellitus and arterial stiffness: a prospective analysis based on the MaineSyracuse Study. Pulse (Basel) 2018;5:88–98.
41. de Oliveira Alvim R, Santos P, Musso MM, et al. Impact of diabetes mellitus on arterial stiffness in a representative sample of an urban Brazilian population. Diabetol Metab Syndr 2013;5:45. 42. Venkitachalam L, Mackey RH, SuttonTyrrell K, et al. Elevated pulse wave velocity increases the odds of coronary calcification in overweight postmenopausal women. Am J Hypertens 2007;20:469–75. 43. Barinas-Mitchell E, Kuller LH, Sutton-Tyrrell K, et al. Effect of weight loss and nutritional intervention on arterial stiffness in type 2 diabetes. Diabetes Care 2006;29:2218–22. 44. Donato AJ, Walker AE, Magerko KA, et al. Lifelong caloric restriction reduces oxidative stress and preserves nitric oxide bioavailability and function in arteries of old mice. Aging Cell 2013; 12:772–83. 45. Fleenor BS, Sindler AL, Marvi NK, et al. Curcumin ameliorates arterial dysfunction and oxidative stress with aging. Exp Gerontol 2013;48: 269–76. 46. de Picciotto NE, Gano LB, Johnson LC, et al. Nicotinamide mononucleotide supplementation reverses vascular dysfunction and oxidative stress with aging in mice. Aging Cell 2016;15: 522–30. 47. Endes S, Schaffner E, Caviezel S, et al. Physical activity is associated with lower arterial stiffness in older adults: results of the SAPALDIA 3 Cohort Study. Eur J Epidemiol 2016;31:275–85. 48. Madden KM, Lockhart C, Cuff D, Potter TF, Meneilly GS. Short-term aerobic exercise reduces arterial stiffness in older adults with type 2 diabetes, hypertension, and hypercholesterolemia. Diabetes Care 2009;32:1531–5.
KEY WORDS arterial stiffening, gender, pulse wave velocity, risk factors, vascular aging
A PPE NDI X For an expanded Methods section as well as a supplemental table and figures, please see the online version of this paper.