Content uploaded by Nathan Wong
Author content
All content in this area was uploaded by Nathan Wong on Jan 22, 2015
Content may be subject to copyright.
American Journal of Epidemiology
ªThe Author 2008. Published by the Johns Hopkins Bloomberg School of Public Health.
All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org.
Vol. 169, No. 2
DOI: 10.1093/aje/kwn303
Advance Access publication December 8, 2008
Original Contribution
Association of Coronary Artery and Aortic Calcium With Lumbar Bone Density
The MESA Abdominal Aortic Calcium Study
Joseph A. Hyder, Matthew A. Allison, Nathan Wong, Agnes Papa, Thomas F. Lang, Claude Sirlin,
Susan M. Gapstur, Pamela Ouyang, J. Jeffrey Carr, and Michael H. Criqui
Initially submitted October 31, 2007; accepted for publication September 4, 2008.
Atherosclerosis and osteoporosis share many risk factors, but their independent association is unclear. The
authors investigated the independent associations between volumetric trabecular bone mineral density (vBMD) of
the lumbar spine and coronary artery calcium (CAC) and abdominal aortic calcium (AAC). During 2002–2005, they
used quantitative computed tomography to assess vBMD and the presence and extent of CAC and AAC among
946 women (mean age ¼65.5 years) and 963 men (mean age ¼64.1 years) in a substudy of the Multi-Ethnic
Study of Atherosclerosis. Prevalences of CAC were 47% and 68% in women and men, respectively, and AAC
prevalences were 70% and 73%. Sequential, sex-specific regression models included adjustment for age, ethnic-
ity, body mass index, hypertension, dyslipidemia, diabetes mellitus, smoking, alcohol consumption, physical
activity, interleukin-6, C-reactive protein, homocysteine, and sex hormones. After full adjustment, lower vBMD
was associated with greater CAC score among women (P<0.002) and greater AAC score among women
(P¼0.004) and men (P<0.001). After adjustment, vBMD quartile was inversely associated with CAC prevalence
(P-trend ¼0.05) in women and AAC prevalence (P-trend <0.01) in men. Partially and fully adjusted models
showed similar results. Though modest, these significant, independent associations suggest that atherosclerosis
and bone loss may be related.
aging; arteries; atherosclerosis; bone and bones; bone density; calcium; coronary vessels; ethnic groups
Abbreviations: AAC, abdominal aortic calcium; CAC, coronary artery calcium; CT, computed tomography; MESA, Multi-Ethnic
Study of Atherosclerosis; vBMD, volumetric trabecular bone mineral density.
Calcium in the coronary arteries and the abdominal aorta
is closely related to the volume of atherosclerosis at arterial
sites (1) and shares histologic patterns with bone tissues at
skeletal sites (2). Both arterial calcium and bone density
have demonstrated the incremental ability to predict future
death from cardiovascular disease (3–5), while in large ob-
servational studies, investigators have reported significant
inverse associations between bone density and atheroscle-
rosis (5–9). Both are strongly correlated with age, sex, and
ethnicity, but other determinants are less robust (10–12).
Laboratory investigations have demonstrated that oxi-
dized low density lipoproteins inhibit normal osteoblast de-
velopment from bone marrow stromal cells and also
promote the calcification of vascular smooth muscle cells
(13), suggesting that the link between bone density and
atherosclerosis may be relevant in the context of calcified
atherosclerosis (14). However, investigators discussing pos-
sible reasons for a bone-artery association typically impli-
cate lipids (15), inflammation (2), or sex hormones (7, 16) to
explain the common disease patterns. To our knowledge, no
epidemiologic investigations to date have been able to eval-
uate simultaneously the roles of each of these factors in the
association. With few exceptions (17), previous studies of
the association have not included a multiethnic sample, and
patterns of atherosclerosis and bone density are known to
vary by ethnicity (10, 18).
Correspondence to Dr. Joseph A. Hyder at Department of Medicine, Harvard Medical School and Brigham and Women’s Hospital, 75 Francis
Street, Boston, MA 02115 (e-mail: jhyder@partners.org).
186 Am J Epidemiol 2009;169:186–194
We hypothesized that calcified atherosclerosis, specifi-
cally coronary artery calcium (CAC) and abdominal aortic
calcium (AAC), would be inversely associated with volu-
metric trabecular bone mineral density (vBMD) of the lum-
bar spine in a population free of symptomatic cardiovascular
disease. We further hypothesized that these associations
would be attenuated upon adjustment for lipids, inflamma-
tory markers, sex hormones, and other potential shared
determinants (19).
MATERIALS AND METHODS
Study participants
The methods of the Multi-Ethnic Study of Atherosclerosis
(MESA) have been described previously (20). In brief, the
MESA cohort was recruited between July 2000 and August
2002 from 6 field centers around the United States. The study
population consisted of 6,814 men and women aged 45–84
years who were free of clinically manifest cardiovascular
disease and identified themselves as non-Hispanic white,
Chinese-American, African-American, or Hispanic.
In this analysis, we used a random sample of MESA
subjects who participated in the MESA Abdominal Aortic
Calcium Study. Participants in the MESA Abdominal Aortic
Calcium Study were recruited during follow-up visits be-
tween August 2002 and September 2005 from 5 MESA field
centers: Chicago, Illinois; Forsyth County, North Carolina;
Los Angeles County, California; New York, New York; and
St. Paul, Minnesota. Of 2,202 MESA subjects recruited,
2,172 agreed to participate, and 1,990 satisfied the eligibility
criteria, including postmenopausal status (for women), no
recent prior diagnostic abdominal computed tomography
(CT), and age and ethnicity subsampling from the MESA.
In all, 1,968 participants (974 women and 994 men) com-
pleted CT scanning. Subsequently, 28 women and 31 men
were excluded because of vertebral abnormalities compli-
cating bone density measurement. Concurrent CAC scores
were available for all but 28 women and 25 men. For these
participants, CAC scores were replaced with scores from
a study visit made 2 years prior. There were no missing
AAC data, leaving 946 female and 963 male participants
for investigation. Signed informed consent was obtained for
all participants, and institutional review board approval was
obtained from all participating institutions.
CT scanning
Participants underwent CT scanning of the chest at 1 of 2
clinical visits between August 2002 and September 2005.
Scans were performed either with an electrocardiogram-
triggered (at 80% of the R-R interval) electron-beam CT
scanner (Imatron C-150; GE Medical Systems, Milwaukee,
Wisconsin) or with prospectively electrocardiogram-
triggered scan acquisition at 50% of the R-R interval with
a multidetector CT system (Somatom Sensation 64
(Siemens, Erlanger, Germany), Lightspeed QXi (GE Medi-
cal Systems), Siemens S4þVolume Zoom (Siemens), and
Siemens Sensation 16 (Siemens)) that acquired 4 simulta-
neous 2.5-mm slices for each cardiac cycle in a sequential or
axial scan mode (New York, Forsyth County, and St. Paul
field centers) (21). For accuracy, 2 chest scans were
performed for each individual.
CT of the abdomen was performed concomitantly with
the CT scans of the chest for CAC. For electron-beam CT,
scanners were set as follows: scan collimation of 3 mm,
slice thickness of 6 mm, reconstruction using 25 6-mm
slices with a 35-cm field of view, and normal kernel. For
multidetector CT, images were reconstructed in a 35-cm
field of view with a 5-mm slice thickness. All scans were
brightness-adjusted with a standard phantom.
Calcium scoring
CT scans were read centrally by the MESA CT Reading
Center, and calcium in the coronary arteries and in an 8-cm
segment of the distal abdominal aorta ending at the aortic
bifurcation was scored. Abdominal calcium score and the
average Agatston CAC score were determined with high
reliability, as has been previously described (21, 22). Rescan
agreement for CAC score was found to be high with both
electron-beam tomography and multidetector CT scanners
(22). Interobserver agreement and intraobserver agreement
were found to be very high (j¼0.93 and j¼0.90, respec-
tively) (21–23).
Bone density measurement
Using CT scans of the abdomen, data were collected
using the Image Analysis QCT 3D PLUS software program
(Image Analysis, Columbia, Kentucky), and scans were read
centrally at the MESA CT Reading Center by a reader
blinded to the results of arterial calcium scoring. Measure-
ments of vBMD in a virtual 10-mm-thick slice of trabecular
bone from each vertebra (L2–L4) used software-directed,
automated placement of the region of interest in the anterior
one-half to one-third of the vertebral body, where it encom-
passed a large area exclusively of trabecular or cancellous
bone, excluded cortical bone, and excluded the basivertebral
plexus. A trained reader examined each region of interest
and changed its placement to exclude vertebral abnormali-
ties, including bone islands and diffuse density variations, or
excluded a vertebra entirely if any of the following abnor-
malities were noted: fractures, metastatic lesions, osteo-
phytes, or benign focal lesions within the vertebra. In the
current analyses, we used bone density from the third
lumbar vertebra.
In a random sample of 25 scans reread on 3 occasions,
there was 100% agreement on inclusion or exclusion for
all vertebrae assessed (L2–L4). Multivariate repeated-
measures analysis of variance indicated no time effect in
the data (Wilks’ test with 18 df: F-equivalent ¼0.18,
P¼0.839) (24). Pearson’s rfor pairwise rereads was
greater than 0.98.
Clinical measurements
Covariate data from the first MESA examination were
used in the present analyses (20). Age, sex, ethnicity, height,
weight, current use of prescription medications, physical
activity patterns (metabolic equivalents 3minutes/week),
Arterial Calcium and Bone Density 187
Am J Epidemiol 2009;169:186–194
smoking history, alcohol consumption (never/former/cur-
rent), and previous medical diagnoses were recorded. Hor-
mone therapy among women was defined as recent if
hormone use in the previous 2 years was reported. Dietary
calcium was calculated using a self-administered food fre-
quency questionnaire and dietary supplement form. Body
mass index was calculated as weight in kilograms divided
by height in meters squared from data concurrent with bone
density assessment. Blood pressure was measured 3 times
with a Dinamap model Pro 100 automated oscillometric
sphygmomanometer (Critikon Company, LLC, Tampa,
Florida) while the participant rested in a seated position.
The average of the last 2 measurements was used. Hyper-
tension was defined as systolic blood pressure 140 mm Hg
or diastolic blood pressure 90 mm Hg or current use of
antihypertensive medication.
Laboratory measurements
Levels of C-reactive protein, interleukin-6, and homocys-
teine were measured using standardized methods (20).
Serum sex hormone concentrations were measured from
stored samples at the University of Massachusetts Medical
Center in Worcester, Massachusetts. Total testosterone and
dehydroepiandrosterone were measured directly using radio-
immunoassay kits, and sex hormone-binding globulin was
measured by chemiluminescent enzyme immunometric assay
using Immulite kits obtained from Diagnostic Products
Corporation (Los Angeles, California). Estradiol was mea-
sured by use of an ultrasensitive radioimmunoassay kit from
Diagnostic Systems Laboratories, Inc. (Webster, Texas).
Levels of total cholesterol, high density lipoprotein cho-
lesterol, triglycerides, and glucose were measured from
blood samples obtained after a 12-hour fast. For the analyses
described here, we used measurements ascertained at the
initial visit to minimize missing values. Low density lipo-
protein cholesterol was calculated with the Friedewald
equation (25). Diabetes was defined as either a fasting
plasma glucose concentration greater than 126 mg/dL, a re-
ported previous diagnosis of diabetes, or the use of hypo-
glycemic medication.
Statistical analyses
All analyses were stratified by sex, given the significantly
different distributions of calcified atherosclerosis (26) and
bone density (27) in men and women. Chi-square tests of
association and generalized linear models were used to com-
pare distributions of categorical and continuous variables,
respectively, across ethnic groups and bone density quar-
tiles. Trend tests used integer scores across quartiles.
To determine the association between bone density and
calcified atherosclerosis, we used 2-part analyses. We tested
the overall association between bone density and arterial
calcium for each bed using multiple linear regressions to
estimate the mean of the natural logarithm of Agatston score
plus 1 for each bone density quartile. In addition, the asso-
ciation between bone density and the presence or absence of
calcified atherosclerosis was directly estimated as the prev-
alence ratio (relative prevalence) using a generalized linear
regression model with a log link, Gaussian error, and robust
estimates of variance. This model was selected because the
odds ratio requires the rare disease assumption to estimate
accurately the prevalence ratio or risk ratio (28).
We aimed to test the contributions of different groups of
variables to the associations between bone density and ath-
erosclerosis. The following terms were defined as mutually
exclusive theoretical concepts, as follows: a confounder is
a covariate that is associated with both bone density and ath-
erosclerosis but is a risk factor for only the outcome (here,
calcified atherosclerosis); mediators are covariates that medi-
ate or lie within the causal pathway between atherosclerosis
and bone density; and common risk factors are covariates that
cause both atherosclerosis and low bone density. The empir-
ical effect of adjusting for a confounder, mediator, or common
risk factor in these serial models is attenuation or augmenta-
tion of an association. No change in the association with
adjustment indicates that the added covariates make no con-
tribution to the association (19).
Associations were adjusted for groups of covariates in
serial models. In model 1, we adjusted for age in quartiles
and ethnicity. In model 2, we adjusted additionally for total
cholesterol, high density lipoprotein cholesterol, use of lipid
medication, hypertension, diabetes mellitus, smoking
(never/former/current), body mass index, physical activity,
the natural logarithm of dietary calcium, alcohol consump-
tion (never/former/current), and (women only) hormone
therapy. In the final model (model 3), we additionally
adjusted the prevalence ratio for novel risk markers—
ln(interleukin-6), ln(C-reactive protein), homocysteine, and
sex-specific quartiles of total testosterone, estradiol, and sex
hormone-binding globulin (but not for recent hormone
therapy). For women, quartiles of sex hormones were inves-
tigated when defined separately for those using and not using
supplemental estrogens in an unordered 8-level variable
and when groups were collapsed to form quartiles making
a 4-level variable. Models were inspected for attenuation
and/or augmentation of an association across quartiles.
Regression diagnostics included tests of interaction by
age, ethnicity, and (in women) hormone therapy. We inves-
tigated possible multicollinearity by screening for tolerance
values less than 0.15 for variables in fully adjusted models;
no multicollinearity was found.
RESULTS
The characteristics of women and men in this study are
shown in Table 1. The average age of the women was 65
years, and the average age of the men was 64 years. Non-
white participants comprised 62% of women and 59% of
men. In women, the prevalences of CAC and AAC were
47% and 70%, respectively. In men, CAC and AAC preva-
lences were 68% and 73%. Among women, 37% had taken
estrogen within 2 years of bone density assessment.
Table 2 demonstrates the distribution of participants by
sex-specific quartile of vBMD. For both sexes, participants
in higher quartiles of vBMD were significantly younger
(P<0.05), although there was substantial overlap in age
across vBMD quartiles. The natural logarithms of CAC and
188 Hyder et al.
Am J Epidemiol 2009;169:186–194
AAC scores were greater with decreasing vBMD quartile (in
both sexes, P<0.01). CAC and AAC prevalences were
greater among persons in lower vBMD quartiles before ad-
justment (in both sexes, P<0.001 for both outcomes), and
associations were attenuated after adjustment for age and
ethnicity, with marginally significant associations in women
for CAC (P¼0.052) and in men for AAC (P<0.001).
Table 3 displays the adjusted quartile differences in mean
values of the natural logarithm of CAC score plus 1 and
AAC score plus 1 among all participants. The adjusted dif-
ferences were calculated for persons in the lower 3 quartiles
as compared with the highest quartile. In women, lower
vBMD was significantly associated with greater CAC score
in all models (in all models, P<0.01 for trend). Lower
vBMD was also significantly associated with greater AAC
score among women, regardless of adjustments (in all mod-
els, P<0.01 for trend). In men, lower vBMD quartile was
associated with greater CAC score, and this association was
Table 1. Characteristics of Women and Men in the MESA Abdominal Aortic Calcium Study,
2000–2005
Characteristic Women (n5946) Men (n5963)
No. % Mean (SD) No. % Mean (SD)
Age, years 65.2 (9.2) 64.1 (9.9)
Ethnicity
Non-Hispanic white 360 38 404 42
Chinese-American 119 13 136 14
African-American 218 23 172 18
Hispanic 249 26 251 26
Bone density
a
, mg/cm
3
111 (41) 121 (39)
Coronary artery calcium
b
Prevalence 442 47 655 68
ln(Score þ1) 4.2 (1.6) 4.7 (1.8)
Abdominal aortic calcium
b
Prevalence 662 70 703 73
ln(Score þ1) 6.2 (1.8) 6.4 (1.7)
Physical activity, MET-minutes/week/100 120 (59) 120 (73)
Body mass index
c
28.4 (5.8) 27.8 (4.4)
Total cholesterol, mg/dL 201 (33) 191 (34)
HDL cholesterol, mg/dL 57 (16) 45 (11)
Use of cholesterol medication 165 17 136 14
Diabetes mellitus 110 12 136 14
Hypertension 447 47 424 44
Cigarette smoking status
Never smoker 60 42
Former smoker 28 45
Current smoker 12 13
ln(Dietary calcium), mg/day 6.4 (0.6) 6.5 (0.6)
ln(Interleukin-6), pg/mL 0.85 (0.36) 0.81 (0.37)
ln(C-reactive protein), mg/L 1.35 (0.78) 1.03 (0.64)
Homocysteine, lmol/L 8.7 (4.6) 10.0 (3.4)
Recent hormone therapy
d
346 37
Estradiol, nmol/L 0.14 (0.18) 0.12 (0.04)
Sex hormone-binding globulin, nmol/L 77.3 (56.8) 43.2 (17.9)
Total testosterone, nmol/L 1.06 (0.98) 15.0 (5.6)
Abbreviations: HDL, high density lipoprotein; MESA, Multi-Ethnic Study of Atherosclerosis;
MET, metabolic equivalent; SD, standard deviation.
a
Volumetric trabecular bone mineral density of the lumbar spine.
b
Mean calcium scores include only participants with positive scores.
c
Weight (kg)/height (m)
2
.
d
Estrogen use in the previous 2 years.
Arterial Calcium and Bone Density 189
Am J Epidemiol 2009;169:186–194
significant after full adjustment (for trend across quartiles,
P¼0.03). In men, lower vBMD quartile was significantly
associated with greater AAC score regardless of adjustment
(P<0.001).
Table 4 displays prevalence ratios for any CAC and AAC
derived from sex-specific sequential models. Among women,
quartile decrement of vBMD was significantly associated
with CAC >0. Comparing the highest quartile of bone den-
sity with the lowest quartile, this association differed little
when results were adjusted for age and ethnicity (prevalence
ratio ¼1.17, 95% confidence interval: 0.95, 1.43; for trend
across quartiles, P¼0.052) or when results were fully ad-
justed (prevalence ratio ¼1.15, 95% confidence interval:
0.96, 1.39; for trend across quartiles, P¼0.048). In contrast,
vBMD quartile was not significantly associated with AAC
presence in any models among women.
In men, vBMD was not significantly associated with CAC
presence but was significantly associated with AAC pres-
ence. After adjustment for age and ethnicity, men in the
lowest quartile of vBMD had a 19% greater prevalence
(95% confidence interval: 1.07, 1.32; for trend across quar-
tiles, P¼0.001) of AAC than men in the highest quartile of
vBMD. The magnitude and significance of this association
did not change after full adjustment.
For the results presented in Tables 3 and 4, we conducted
multiple sensitivity analyses, including substitution of L4
vBMD values for L3 vBMD values, exclusion of partici-
pants without concurrent vBMD and CAC assessment,
alternative adjustment and exclusions for sex hormone
measurements, and redefinition of calcium presence as an
Agatston score of 10 or greater; multiple variable forms for
age, including linear, continuous age, were investigated, as
well as time since the menopause began (women only).
Across models, patterns of associations were similar, and
the results from final models did not differ materially from
those shown. In addition, for any outcome, each analysis
specified by models 1 and 2 was repeated with the smaller
samples available for model 3. Results did not differ mate-
rially from those presented. Tests of interaction by ethnicity
were not statistically significant for either sex or outcome. In
a final rigorous adjustment for age, a series of models in
which vBMD quartile was assigned specifically by sex and
quartile of age yielded findings similar to those shown, with
the exception of a loss of statistical significance of the as-
sociation between linear CAC and vBMD among men in
model 3 and statistical significance of the association
between AAC prevalence and vBMD in women.
DISCUSSION
In this study, we demonstrated inverse, independent asso-
ciations between vBMD and calcified atherosclerosis.
Lower vBMD was significantly associated with greater
Table 2. Distribution of Coronary Artery Calcium and Abdominal Aortic Calcium Scores by Sex-Specific Quartile of Volumetric Trabecular Bone
Mineral Density of the Lumbar Spine, MESA Abdominal Aortic Calcium Study, 2000–2005
Quartile of vBMD
Women Men
Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4
No. of subjects 242 235 235 234 244 236 241 242
vBMD cutpoint, mg/cm
3
<83 <108 <137 137 <94 <118 <145 145
Mean vBMD (SD) 63 (15) 96 (8) 122 (8) 164 (26) 75 (17) 107 (7) 131 (8) 172 (25)
Age, years
Range 51–86 46–85 48–86 46–84 47–88 47–85 47–83 46–85
Mean (SD) 73 (7) 66 (8) 62 (8) 59 (8) 72 (9) 66 (9) 60 (9) 59 (9)
Median (IQR
a
) 73 (9) 66 (12) 62 (12) 57 (12) 71 (12) 67 (15) 59 (14) 57 (13)
ln(CAC þ1)
Median (IQR) 3.5 (5.4) 1.1 (4.3) 0 (3.2) 0 (2.1) 4.8 (3.4) 4.0 (5.7) 2.1 (4.9) 1.5 (4.7)
Mean (SD) 3.1 (2.5) 2.1 (2.4) 1.4 (2.1) 1.1 (1.9) 4.2 (2.4) 3.5 (2.7) 2.6 (2.6) 2.4 (2.5)
Prevalence, % (no.) 67 (163) 51 (120) 37 (88) 30 (71) 83 (202) 72 (170) 59 (141) 59 (142)
Adjusted prevalence
b
(95% CI) 50 (40, 61) 46 (37, 57) 41 (33, 52) 42 67 (59, 75) 64 (57, 72) 62 (55, 71) 66
ln(AAC þ1)
Median (IQR) 7.1 (3.1) 5.4 (7.4) 4.0 (6.3) 3.4 (5.9) 6.9 (2.7) 6.3 (4.2) 4.7 (6.8) 3.4 (6.3)
Mean (SD) 6.0 (2.7) 4.6 (3.1) 3.4 (3.1) 3.2 (3.1) 6.3 (2.5) 5.3 (3.0) 4.0 (3.2) 3.3 (3.3)
Prevalence, % (no.) 88 (212) 74 (175) 60 (140) 58 (135) 91 (222) 81 (190) 66 (159) 55 (132)
Adjusted prevalence
b
(95% CI) 71 (64, 79) 70 (63, 79) 62 (55, 71) 69 76 (68, 84) 72 (64, 81) 71 (62, 80) 63
Abbreviations: AAC, abdominal aortic calcium; CAC, coronary artery calcium; CI, confidence interval; IQR, interquartile range; MESA, Multi-
Ethnic Study of Atherosclerosis; SD, standard deviation; Q, quartile; vBMD, volumetric trabecular bone mineral density.
a
25th percentile–75th percentile range.
b
Prevalence values adjusted for age and ethnicity (African-American, Chinese-American, Hispanic, or non-Hispanic white), with the highest
quartile being used as the reference category for adjustment.
190 Hyder et al.
Am J Epidemiol 2009;169:186–194
Table 4. Prevalence Ratio for the Association of Coronary Artery Calcium and Abdominal Aortic Calcium With Sex-Specific Quartile of
Volumetric Trabecular Bone Mineral Density of the Lumbar Spine, MESA Abdominal Aortic Calcium Study, 2000–2005
Measure and Bone
Density Quartile
Women Men
Model 1
a
(n5946) Model 2
b
(n5865) Model 3
c
(n5769) Model 1
a
(n5963) Model 2
b
(n5887) Model 3
c
(n5848)
PR 95% CI PR 95% CI PR 95% CI PR 95% CI PR 95% CI PR 95% CI
CAC score >0
1 1.17 0.95, 1.43 1.16 0.96, 1.41 1.15 0.96, 1.39 1.02 0.91, 1.14 1.06 0.94, 1.19 1.05 0.93, 1.19
2 1.09 0.88, 1.35 1.06 0.87, 1.29 1.05 0.86, 1.29 0.98 0.87, 1.10 1.01 0.89, 1.14 1.01 0.90, 1.15
3 0.97 0.77, 1.22 0.92 0.74, 1.16 0.96 0.77, 1.20 0.95 0.83, 1.08 0.98 0.86, 1.12 0.97 0.85, 1.11
4
d
1111 1 1
Pfor trend 0.052 0.016 0.048 0.478 0.186 0.231
AAC score >0
1 1.03 0.93, 1.14 1.02 0.93, 1.12 1.02 0.93, 1.12 1.19
a
1.07, 1.32 1.19
a
1.07, 1.33 1.19 1.07, 1.32
2 1.02 0.91, 1.14 0.99 0.89, 1.10 0.98 0.88, 1.09 1.13
a
1.01, 1.27 1.14
a
1.02, 1.28 1.15 1.03, 1.29
3 0.91 0.79, 1.03 0.88 0.78, 0.99 0.88 0.78, 0.99 1.11 0.98, 1.25 1.13 1.00, 1.27 1.11 0.99, 1.26
4
d
1111 1 1
Pfor trend 0.168 0.093 0.145 0.001 0.001 <0.001
Abbreviations: AAC, abdominal aortic calcium; CAC, coronary artery calcium; CI, confidence interval; MESA, Multi-Ethnic Study of Atheroscle-
rosis; PR, prevalence ratio.
a
Adjusted for age and ethnicity.
b
Additionally adjusted for total cholesterol, high density lipoprotein cholesterol, use of lipid medication, hypertension, diabetes mellitus, smoking,
body mass index, physical activity, dietary calcium, alcohol consumption, and (women only) hormone therapy.
c
Additionally adjusted for interleukin-6, C-reactive protein, homocysteine, total testosterone, estradiol, and sex hormone-binding globulin.
d
Reference category.
Table 3. Adjusted Difference From Reference Value for Log-Transformed Coronary Artery Calcium and Abdominal
Aortic Calcium, by Sex-Specific Quartile of Volumetric Trabecular Bone Mineral Density of the Lumbar Spine, MESA
Abdominal Aortic Calcium Study, 2000–2005
Measure and Bone
Density Quartile
Women Men
Model 1
a
Model 2
b
Model 3
c
Model 1
a
Model 2
b
Model 3
c
ln(CAC þ1) n¼946 n¼859 n¼769 n¼963 n¼887 n¼848
1 0.61 (0.22)
d,e
0.81 (0.24)
e
0.77 (0.25)
e
0.34 (0.24) 0.47 (0.25) 0.52 (0.25)
e
2 0.30 (0.21) 0.32 (0.21) 0.27 (0.24) 0.06 (0.23) 0.14 (0.23) 0.26 (0.24)
30.05 (0.20) 0.04 (0.20) 0.01 (0.23) 0.08 (0.22) 0.10 (0.23) 0.10 (0.23)
4
f
0 00000
Pfor trend 0.004 <0.001 <0.002 0.137 0.067 0.034
ln(AAC þ1) n¼946 n¼859 n¼769 n¼963 n¼887 n¼848
1 0.59 (0.30)
e
0.75 (0.29)
e
0.74 (0.30)
e
0.97 (0.27)
e
0.98 (0.26)
e
1.00 (0.27)
e
2 0.23 (0.27) 0.23 (0.26) 0.15 (0.28) 0.52 (0.26)
e
0.57 (0.25)
e
0.64 (0.26)
e
30.33 (0.26) 0.39 (0.24) 0.41 (0.27) 0.35 (0.25) 0.43 (0.24) 0.40 (0.24)
4
f
0 00000
Pfor trend 0.020 0.003 0.004 <0.001 <0.001 <0.001
Abbreviations: AAC, abdominal aortic calcium; CAC, coronary artery calcium; MESA, Multi-Ethnic Study of
Atherosclerosis.
a
Adjusted for age and ethnicity.
b
Additionally adjusted for total cholesterol, high density lipoprotein cholesterol, use of lipid medication, hyperten-
sion, diabetes mellitus, smoking, body mass index, physical activity, dietary calcium, alcohol consumption, and
(women only) hormone therapy.
c
Additionally adjusted for interleukin-6, C-reactive protein, homocysteine, total testosterone, estradiol, and sex
hormone-binding globulin.
d
Numbers in parentheses, standard error.
e
Significant difference from reference category.
f
Reference category.
Arterial Calcium and Bone Density 191
Am J Epidemiol 2009;169:186–194
coronary and aortic calcium in both women and men. In
addition, vBMD was significantly associated with CAC
prevalence in women and with AAC prevalence in men. In
sequential models, these modest associations between
bone density and atherosclerosis showed little evidence of
attenuation or augmentation upon adjustment for lipids,
inflammatory markers, and sex hormones, as well as other
covariates, including shared risk factors for atherosclerosis
and bone density such as physical activity and smoking.
Only adjustment for age attenuated the associations (19).
The modest, significant, inverse associations between
bone density and atherosclerosis in the present study are
consistent with previous investigations in large samples of
white women and extend previous findings to the coronary
arteries, to men, and to nonwhite populations (29–31). To
our knowledge, the present study was the first population-
based study of a possible bone-artery association to assess
bone density volumetrically, which uniquely permits the
specific measurement of trabecular bone, the exclusion of
vertebrae with bony islands or osteophytes, and volumetric
accounting of differences in vertebra size by age, sex, and
ethnicity (32, 33). This study was also the first to examine
the effects of adjustment for lipids, multiple inflammatory
markers, and sex hormones (34, 35). In addition, the present
study was the first of its kind (17) to include both white and
nonwhite participants, and patterns across ethnic groups did
not differ significantly.
Evidence indicating possible roles for lipids, inflamma-
tion, and sex hormones in the association between osteopo-
rosis and atherosclerosis comes indirectly from
epidemiologic studies, animal models, and cell culture
studies (36). In some epidemiologic investigations (37),
but not all (38), investigators have reported that lipid level
or saturated fat intake is associated with bone density
(39). Atherosclerosis, as well as osteoporosis-like disease,
can be induced in atherosclerosis-prone mice by assigning
the animals to a high-fat diet (40). Cell-culture studies
have demonstrated that oxidized low density lipoprotein
can inhibit the differentiation of osteoblasts in bone, as
in osteoporosis, and promote the calcification of smooth
muscle vascular cells, as in atherosclerosis (13). In prospec-
tive observational studies, interleukin-6 has been indepen-
dently associated with cardiovascular disease events (41)
and bone loss (42), and homocysteine has been demonstrated
to be a risk factor for cardiovascular disease (43) and oste-
oporotic fractures (44, 45). Finally, the uncertain role of sex
hormones in cardiovascular disease is contrasted by their
clearer roles in bone loss (27, 46).
In this study, multiple adjustments for all measured fac-
tors did not attenuate the associations between bone density
and atherosclerosis, nor did multiple attempts to adjust for
age fully attenuate the associations. Reports from this and
other large, population-based observational studies, includ-
ing the Framingham (6), Rancho Bernardo (7), Tromsø (8),
and Rotterdam (9) studies, suggest that evidence for the
association is less likely to be spurious. Three separate
explanations for our results seem possible. First, lipids, in-
flammation, and sex hormones may account for the associ-
ation, but the cross-sectional study design did not allow for
a comprehensive assessment of their effects. Alternatively,
factors not measured here, such as genotypes, may explain
what is a noncausal association. Finally, atherosclerosis and
bone density loss may be related processes.
A cross-sectional study is limited in its ability to evaluate
the effects of long-acting factors such as lipids, inflamma-
tion, and sex hormones. However, large cross-sectional
studies often detect associations with lipids, inflammatory
markers, or sex hormones that are subsequently validated in
prospective studies. Regarding other factors, many factors
not measured in the MESA are known to affect disease de-
velopment in both the bones and the arteries. In recent stud-
ies, the cannabinoid system has been implicated in
atherosclerosis and osteoporosis (47, 48). Cannabinoid re-
ceptors are found in atherosclerotic lesions but also exert
effects on bone, partly through the osteoprotegerin/RANKL
system, which affects both atherosclerotic (49) and osteo-
porotic (50) change.
Alternatively, trabecular bone loss may result, in part,
from atherosclerotic disease in bone arteries and may
represent bone vascular disease (51). Recent magnetic
resonance imaging studies have demonstrated that dimin-
ished bone perfusion is independently correlated with
greater carotid atherosclerosis, lower bone density, and
greater bone marrow fat content—the latter 2 being highly
characteristic of osteoporosis (52, 53). Lower bone density
is common in type 1 diabetics among whom microvascular
disease is present (54); and in a trial among rats, assignment
to hormone therapy or nitroglycerin, a potent vasodilator,
resulted in equal improvements in bone density after surgi-
cal menopause (55). The recent application of volumetric
roentgenographic methods has demonstrated that trabecular
bone loss may begin as early as the second and third decades
of life in men, coinciding with the early development of
atherosclerotic disease, typically in the aorta (26, 33). The
rapid loss of trabecular bone in early menopause among
women crudely mirrors the ‘‘catch-up’’ phase of increased
cardiovascular risk that occurs among women during that
time (33). Trabecular bone density may prove useful for
refined cardiovascular disease risk stratification.
There are few strong correlates of arterial calcium other
than age, sex, and ethnicity (10–12). In the present study,
associations between the ranges of CAC in women and AAC
in both sexes, though modest, were highly significant and
changed little after multiple adjustments. The relative prev-
alence estimates in this study were small; however, this was
largely a result of the high prevalence of calcified disease in
the lowest vBMD quartiles. Associations between bone den-
sity and arterial calcium were comparable to associations of
calcium with standard cardiovascular risk factors such as
hypertension and high density lipoprotein cholesterol (not
shown).
Limitations of our study include its cross-sectional de-
sign, the use of bone density measurements from a single
site, the onetime measurements of sex hormone levels and
inflammatory factors, and a lack of data describing vitamin
D status, parathyroid hormone levels, use of selective estro-
gen receptor modulators, and antiresorptive agents such as
bisphosphonates. In light of these limitations, conclusions
about causality from these data are speculative. Local
effects of lipids, inflammation, or sex hormones in the bone
192 Hyder et al.
Am J Epidemiol 2009;169:186–194
arteries—effects not quantifiable by assessment of systemic
venous blood here—may yet explain the association. This is
the first investigation of its kind with a multiethnic sample,
but conclusions about any ethnic differences are further
limited by the small sizes of the ethnic subgroups.
In summary, we investigated a large, multiethnic, population-
based sample with accurate measures of atherosclerosis
from multiple sites, volumetric bone density, and exten-
sive data on risk factors for atherosclerosis and low bone
density. We found that vBMD was significantly associated
with CAC presence and amount in women, AAC amount
in women, and AAC prevalence in men. These associa-
tions did not change upon adjustment for multiple hypoth-
esized correlates of an association, suggesting that bone
loss and atherosclerosis may be linked processes. Clini-
cally, the utility of bone density for refined cardiovascular
disease risk stratification and/or assessment of biologic
age merits further investigation. Joint epidemiologic and
laboratory investigations may further elucidate the deter-
minants and implications of a bone-artery association.
ACKNOWLEDGMENTS
Author affiliations: Department of Medicine, Harvard
Medical School and Brigham and Women’s Hospital, Bos-
ton, Massachusetts (Joseph A. Hyder); Department of
Family and Preventive Medicine, School of Medicine,
University of California, San Diego (UCSD), La Jolla, Cal-
ifornia (Matthew A. Allison, Michael H. Criqui); Heart Dis-
ease Prevention Program, University of California, Irvine,
Irvine, California (Nathan D. Wong); Los Angeles Unified
School District, Gardena, California (Agnes Papa); Depart-
ment of Radiology, School of Medicine, University of Cal-
ifornia, San Francisco, San Francisco, California (Thomas
F. Lang); Department of Radiology, School of Medicine,
UCSD, La Jolla, California (Claude Sirlin); Department of
Preventive Medicine, Feinberg School of Medicine, North-
western University, Chicago, Illinois (Susan Gapstur); Di-
vision of Cardiology, School of Medicine, Johns Hopkins
University, Baltimore, Maryland (Pamela Ouyang); Depart-
ments of Radiology and Public Health Sciences, School of
Medicine, Wake Forest University, Winston-Salem, North
Carolina (Jeffrey Carr); and Department of Medicine,
School of Medicine, UCSD, La Jolla, California (Michael
H. Criqui).
This research was supported by contracts N01-HC-95159
through N01-HC-95165, N01-HC-95169, R01 HL074406,
R01 HL074338, and R01 HL72403 from the National Heart,
Lung, and Blood Institute.
The authors thank the other investigators and staff of the
Multi-Ethnic Study of Atherosclerosis (MESA) for their
valuable contributions. The authors acknowledge ‘‘marc
ve
´ron ag informatik information internet’’ (Allschwil, Swit-
zerland) for donating technical expertise and software for
data transfer.
A full list of participating MESA investigators and
institutions can be found at http://www.mesa-nhlbi.org.
Conflict of interest: none declared.
REFERENCES
1. Sangiorgi G, Rumberger JA, Severson A, et al. Arterial cal-
cification and not lumen stenosis is highly correlated with
atherosclerotic plaque burden in humans: a histologic study of
723 coronary artery segments using nondecalcifying method-
ology. J Am Coll Cardiol. 1998;31(1):126–133.
2. Demer LL, Abedin M. Skeleton key to vascular disease. JAm
Coll Cardiol. 2004;44(10):1977–1979.
3. Greenland P, LaBree L, Azen SP, et al. Coronary artery cal-
cium score combined with Framingham score for risk pre-
diction in asymptomatic individuals. JAMA. 2004;291(2):
210–215.
4. Magnus JH, Broussard DL. Relationship between bone min-
eral density and myocardial infarction in US adults. Osteo-
poros Int. 2005;16(12):2053–2062.
5. Samelson EJ, Kiel DP, Broe KE, et al. Metacarpal cortical area
and risk of coronary heart disease: The Framingham Study. Am
J Epidemiol. 2004;159(6):589–595.
6. Kiel DP, Kauppila LI, Cupples LA, et al. Bone loss and the
progression of abdominal aortic calcification over a 25 year
period: The Framingham Heart Study. Calcif Tissue Int.
2001;68(5):271–276.
7. Bakhireva LN, Barrett-Connor EL, Laughlin GA, et al. Dif-
ferences in association of bone mineral density with coronary
artery calcification in men and women: The Rancho Bernardo
Study. Menopause. 2005;12(6):691–698.
8. Jorgensen L, Joakimsen O, Rosvold Berntsen GK, et al. Low
bone mineral density is related to echogenic carotid artery
plaques: a population-based study. Am J Epidemiol. 2004;
160(6):549–556.
9. Van Der Klift M, Pols HA, Hak AE, et al. Bone mineral
density and the risk of peripheral arterial disease: The Rot-
terdam Study. Calcif Tissue Int. 2002;70(6):443–449.
10. Bild DE, Detrano R, Peterson D, et al. Ethnic differences in
coronary calcification: The Multi-Ethnic Study of Athero-
sclerosis (MESA). Circulation. 2005;111(10):1313–1320.
11. Rosen CJ. Clinical practice. Postmenopausal osteoporosis.
N Engl J Med. 2005;353(6):595–603.
12. Raisz LG. Clinical practice. Screening for osteoporosis.
N Engl J Med. 2005;353(2):164–171.
13. Parhami F, Morrow AD, Balucan J, et al. Lipid oxidation
products have opposite effects on calcifying vascular cell and
bone cell differentiation. A possible explanation for the para-
dox of arterial calcification in osteoporotic patients. Arterios-
cler Thromb Vasc Biol. 1997;17(4):680–687.
14. Demer LL, Tintut Y. Mineral exploration: search for the
mechanism of vascular calcification and beyond. The 2003
Jeffrey M. Hoeg Award Lecture. Arterioscler Thromb Vasc
Biol. 2003;23(10):1739–1743.
15. Parhami F, Garfinkel A, Demer LL. Role of lipids in osteo-
porosis. Arterioscler Thromb Vasc Biol. 2000;20(11):
2346–2348.
16. Tanko LB, Bagger YZ, Christiansen C. Low bone mineral
density in the hip as a marker of advanced atherosclerosis in
elderly women. Calcif Tissue Int. 2003;73(1):15–20.
17. Boukhris R, Becker KL. Calcification of the aorta and osteo-
porosis. A roentgenographic study. JAMA. 1972;219(10):
1307–1311.
18. Ettinger B, Sidney S, Cummings SR, et al. Racial differences
in bone density between young adult black and white subjects
persist after adjustment for anthropometric, lifestyle, and
biochemical differences. J Clin Endocrinol Metab. 1997;
82(2):429–34.
19. Greenland S, Morgenstern H. Confounding in health research.
Annu Rev Public Health. 2001;22:189–212.
Arterial Calcium and Bone Density 193
Am J Epidemiol 2009;169:186–194
20. Bild DE, Bluemke DA, Burke GL, et al. Multi-Ethnic Study of
Atherosclerosis: objectives and design. Am J Epidemiol. 2002;
156(9):871–881.
21. Carr JJ, Nelson JC, Wong ND, et al. Calcified coronary artery
plaque measurement with cardiac CT in population-based
studies: standardized protocol of Multi-Ethnic Study of
Atherosclerosis (MESA) and Coronary Artery Risk Develop-
ment in Young Adults (CARDIA) Study. Radiology. 2005;
234(1):35–43.
22. Detrano RC, Anderson M, Nelson J, et al. Coronary calcium
measurements: effect of CT scanner type and calcium measure
on rescan reproducibility—MESA Study. Radiology. 2005;
236(2):477–484.
23. Agatston AS, Janowitz WR, Hildner FJ, et al. Quantification of
coronary artery calcium using ultrafast computed tomography.
J Am Coll Cardiol. 1990;15(4):827–832.
24. Keselman HJ, Algina J, Kowalchuk RK. The analysis of
repeated measures designs: a review. Br J Math Stat Psychol.
2001;54(pt 1):1–20.
25. Friedewald WT, Levy RI, Fredrickson DS. Estimation of the
concentration of low-density lipoprotein cholesterol in plasma,
without use of the preparative ultracentrifuge. Clin Chem.
1972;18(6):499–502.
26. Allison MA, Criqui MH, Wright CM. Patterns and risk factors
for systemic calcified atherosclerosis. Arterioscler Thromb
Vasc Biol. 2004;24(2):331–336.
27. Riggs BL, Khosla S, Melton LJ III. Sex steroids and the con-
struction and conservation of the adult skeleton. Endocr Rev.
2002;23(3):279–302.
28. Greenland S. Model-based estimation of relative risks and
other epidemiologic measures in studies of common outcomes
and in case-control studies. Am J Epidemiol. 2004;160(4):
301–305.
29. Schulz E, Arfai K, Liu X, et al. Aortic calcification and the risk
of osteoporosis and fractures. J Clin Endocrinol Metab. 2004;
89(9):4246–4253.
30. Hak AE, Pols HA, van Hemert AM, et al. Progression of aortic
calcification is associated with metacarpal bone loss during
menopause: a population-based longitudinal study. Arterios-
cler Thromb Vasc Biol. 2000;20(8):1926–1931.
31. Vogt MT, San Valentin R, Forrest KY, et al. Bone mineral
density and aortic calcification: The Study of Osteoporotic
Fractures. J Am Geriatr Soc. 1997;45(2):140–145.
32. Lang TF, Guglielmi G, van Kuijk C, et al. Measurement of
bone mineral density at the spine and proximal femur by
volumetric quantitative computed tomography and dual-
energy X-ray absorptiometry in elderly women with and
without vertebral fractures. Bone. 2002;30(1):247–250.
33. Riggs BL, Melton LJ III, Robb RA, et al. Population-based
study of age and sex differences in bone volumetric density,
size, geometry, and structure at different skeletal sites. J Bone
Miner Res. 2004;19(12):1945–1954.
34. From AM, Hyder JA, Kearns AM, et al. Relationship between
low bone mineral density and exercise-induced myocardial
ischemia. Mayo Clin Proc. 2007;82(6):679–685.
35. Hyder JA, Allison MA, Criqui MH, et al. Association between
systemic calcified atherosclerosis and bone density. Calcif
Tissue Int. 2007;80(5):301–306.
36. McFarlane SI, Muniyappa R, Shin JJ, et al. Osteoporosis and
cardiovascular disease: brittle bones and boned arteries, is
there a link? Endocrine. 2004;23(1):1–10.
37. Adami S, Braga V, Zamboni M, et al. Relationship between
lipids and bone mass in 2 cohorts of healthy women and men.
Calcif Tissue Int. 2003;74(2):136–142.
38. Solomon DH, Avorn J, Canning CF, et al. Lipid levels and
bone mineral density. Am J Med. 2005;118(12):
1414.e1–1414.e5.
39. Corwin RL, Hartman TJ, Maczuga SA, et al. Dietary saturated
fat intake is inversely associated with bone density in humans:
analysis of NHANES III. J Nutr. 2006;136(1):159–165.
40. Parhami F, Tintut Y, Beamer WG, et al. Atherogenic high-fat
diet reduces bone mineralization in mice. J Bone Miner Res.
2001;16(1):182–188.
41. Ridker PM, Rifai N, Stampfer MJ, et al. Plasma concentration
of interleukin-6 and the risk of future myocardial infarction
among apparently healthy men. Circulation. 2000;101(15):
1767–1772.
42. Scheidt-Nave C, Bismar H, Leidig-Bruckner G, et al. Serum
interleukin 6 is a major predictor of bone loss in women
specific to the first decade past menopause. J Clin Endocrinol
Metab. 2001;86(5):2032–2042.
43. Wald DS, Law M, Morris JK. Homocysteine and cardiovas-
cular disease: evidence on causality from a meta-analysis.
BMJ. 2002;325(7374):1202.
44. McLean RR, Jacques PF, Selhub J, et al. Homocysteine as
a predictive factor for hip fracture in older persons. N Engl J
Med. 2004;350(20):2042–2049.
45. van Meurs JB, Dhonukshe-Rutten RA, Pluijm SM, et al.
Homocysteine levels and the risk of osteoporotic fracture.
N Engl J Med. 2004;350(20):2033–2041.
46. Manson JE, Allison MA, Rossouw JE, et al. Estrogen therapy
and coronary-artery calcification. N Engl J Med. 2007;
356(25):2591–2602.
47. Ofek O, Karsak M, Leclerc N, et al. Peripheral cannabinoid
receptor, CB2, regulates bone mass. Proc Natl Acad Sci U S A.
2006;103(3):696–701.
48. Steffens S, Veillard NR, Arnaud C, et al. Low dose oral can-
nabinoid therapy reduces progression of atherosclerosis in
mice. Nature. 2005;434(7034):782–786.
49. Erdogan B, Aslan E, Bagis T, et al. Intima-media thickness of
the carotid arteries is related to serum osteoprotegerin levels in
healthy postmenopausal women. Neurol Res. 2004;26(6):
658–661.
50. Raisz LG. Pathogenesis of osteoporosis: concepts, conflicts,
and prospects. J Clin Invest. 2005;115(12):3318–3325.
51. Jordan J. Good vibrations and strong bones? Am J Physiol
Regul Integr Comp Physiol. 2005;288(3):R555–R556.
52. Griffith JF, Yeung DK, Antonio GE, et al. Vertebral bone
mineral density, marrow perfusion, and fat content in healthy
men and men with osteoporosis: dynamic contrast-enhanced
MR imaging and MR spectroscopy. Radiology. 2005;236(3):
945–951.
53. Chen WT, Ting-Fang Shih T, Hu CJ, et al. Relationship
between vertebral bone marrow blood perfusion and common
carotid intima-media thickness in aging adults. J Magn Reson
Imaging. 2004;20(5):811–816.
54. Tuominen JT, Impivaara O, Puukka P, et al. Bone mineral
density in patients with type 1 and type 2 diabetes. Diabetes
Care. 1999;22(7):1196–1200.
55. Wimalawansa SJ. Restoration of ovariectomy-induced
osteopenia by nitroglycerin. Calcif Tissue Int. 2000;66(1):
56–60.
194 Hyder et al.
Am J Epidemiol 2009;169:186–194
