Bone Mineral Density and Stroke
Michael E. Mussolino, MA; Jennifer H. Madans, PhD; R.F. Gillum, MD
Background and Purpose—We sought to assess the long-term predictive usefulness of bone mineral density (BMD) for
stroke incidence and stroke mortality.
Methods—The First National Health and Nutrition Examination Survey data were obtained from a nationally represen-
tative sample of noninstitutionalized civilians. A cohort of 3402 white and black subjects 45 through 74 years of age
at baseline (1971 to 1975) was observed through 1992. Hospital records and death certificates were used to identify a
total of 416 new stroke cases.
Results—Results were evaluated to determine the relative risk (RR) for stroke per 1-SD decrease in BMD, after controlling
for age at baseline, smoking status, alcohol consumption, history of diabetes, history of heart disease, education, body
mass index, recreational physical activity, and blood pressure medication. In Cox proportional-hazards analyses,
incidence of stroke was not associated with a decrease in BMD in any of the 3 race-sex groups: white men (RR, 1.01;
95% CI, 0.86 to 1.19; P?0.88), white women (RR, 1.13; 95% CI, 0.93 to 1.38; P?0.21), or blacks (RR, 0.93; 95% CI,
0.72 to 1.21; P?0.60). No association between BMD and stroke mortality was found (RR, 1.03; 95% CI, 0.86 to 1.23;
Conclusions—In a large national study, no significant associations of BMD and stroke incidence or mortality were found
for whites or blacks. (Stroke. 2003;34:e20-e22.)
Key Words: bone density ? cohort studies ? stroke
stroke death and incidence over a 2-year follow-up.1,2The
early posthoc mortality findings raised the possibility that, if
confirmed, further studies of the association might lead to the
discovery of a mechanism linking calcium metabolism with
cerebrovascular disease or at least a useful risk marker. A
relationship between BMD and stroke was observed in a
case-control study in which female but not male stroke
patients had lower BMD than population control subjects.3A
large, cross-sectional analysis of stroke prevalence and BMD
using data from the Third National Health and Nutrition
Examination Survey (NHANES) found no significant asso-
ciation of BMD and stroke in men.4There was some
indication of elevated prevalence in women with BMD
?0.936 g/cm2, but no dose-response relationship was found.
To further investigate the relationship between BMD and
stroke incidence, we used data from the NHANES I Epide-
miological Follow-up Study (NHEFS) to assess the associa-
tion of phalangeal BMD with 20-year stroke incidence and
cohort study of elderly women reported significant
associations of low bone mineral density (BMD) and
Subjects and Methods
The NHEFS is a longitudinal study of participants in NHANES I
who were 25 to 74 years of age at the time of the survey in 1971 to
1975.5The personal interviews and physical and laboratory exami-
nations of NHANES I provided the BMD and other baseline data for
the NHEFS. The ascertainment of subsequent stroke occurrence in
this analysis was based on 4 waves of follow-up data collection
during 1982 to 1984, 1986, 1987, and 1992. Data collected consisted
of 4 interview surveys, healthcare facility medical records for the
period between baseline and last follow-up, and death certificates for
This analysis was restricted to white and black persons 45 to 74
years of age at baseline in NHANES I. Of the 4107 subjects in this
group who had been randomly selected from the general sample to
receive detailed examination, including a single bone density mea-
surement, 112 (total, 2.7%) were lost to follow-up (no subject or
proxy interview at any follow-up wave) or had no death certificate.
Excluded from all analyses were 99 subjects (2.4%) who had a
history of stroke at baseline and 494 subjects (12.0%) with unknown
baseline history of heart disease, diabetes history, weight, height,
smoking status, alcohol consumption, blood pressure medication,
recreational physical activity, or educational attainment, leaving
3402 for analysis. The length of follow-up for stroke-free survivors
ranged from 8.5 to 21.8 years (median, 18.5 years), with ?50 000
total person-years of follow-up.
Incident stroke cases met at least 1 of the following criteria: (1) a
death certificate with underlying or nonunderlying cause of death
coded 431 to 434.9. 436, or 437.0 to 437.1 from the International
Classification of Diseases, ninth revision (ICD-9), or (2) ?1 hospital
and/or nursing home stays during the follow-up period with any
discharge diagnosis coded 431 to 434.9, 436, or 437.0 to 437.1 from
the clinical modification of ICD-9 (see the Appendix, which is
available online at http://stroke.ahajournals.org). The date of inci-
dence was estimated as 1 of the following: date of first hospital
admission with a stroke diagnosis or date of death for persons dying
Received August 2, 2002; final revision received November 18, 2002; accepted November 19, 2002.
From the Centers for Disease Control and Prevention, National Center for Health Statistics, Hyattsville, Md.
Correspondence to R.F. Gillum, MD, Division of Epidemiology, National Center for Health Statistics, 3311 Toledo Rd, Rm 6208, Hyattsville, MD
© 2003 American Heart Association, Inc.
Stroke is available at http://www.strokeaha.org DOI: 10.1161/01.STR.0000065826.23815.A5
of stroke without any stroke hospital records. A total of 416 new
stroke cases were identified.
Phalangeal BMD was determined from a baseline x-ray of the left
hand by use of radiographic absorptiometry and reported in arbitrary
units (AU). The radiographic absorptiometry process uses direct
exposure (nonscreen) radiographs of the participant’s left hand
alongside an aluminum alloy reference wedge, which are digitized
by a high-resolution camera for computer analysis.6,7Radiographic
absorptiometry has been found to have excellent precision and
accuracy (precision error, 1%; accuracy error, 4.8%)5and to corre-
late well with other accepted methods of bone densitometry.8
Baseline age, smoking status, alcohol consumption, blood pres-
sure medication use, educational attainment, and physical activity
level were obtained by interview, and body mass index was calcu-
lated from measured height and weight. History of heart attack and
diabetes was based on self-reported doctor’s diagnoses. Estimates of
the risk of stroke for persons according to BMD derive from Cox
proportional-hazards regression models computed with the PHREG
procedure.9Proportional-hazards assumptions were tested by meth-
ods using plots and time-dependent covariates.9Separate analyses
were carried out for white men, white women, and blacks. Small
sample sizes did not permit sex-specific models for blacks. BMD
was treated as a continuous variable, and estimates of risk of stroke
were calculated for people with values of 1 SD below the mean
compared with the mean. To assess the effect of complex survey
design on the results, Cox proportional-hazards regression analyses
were confirmed by use of the survival procedure in SUDAAN to
incorporate the stratification, clustering, and sample weights.10
No statistically significant differences in age-adjusted mean
BMD at baseline with versus without stroke were found for
white men (101.7 versus 101.1 AU, P?0.57), white women
(95.8 versus 97.0 AU, P?0.29), and blacks (102.6 versus
100.1 AU, P?0.15). The number of incident cases of stroke
by race and sex is shown in the Table. In stroke-free
survivors, the length of follow-up averaged 18.7 years. In
Cox proportional-hazards analyses, the incidence of stroke
was not associated with a 1-SD decrease in BMD in any of
the 3 race-sex groups (Table). No significant associations
were found among white men (P?0.88), white women
(P?0.21), or blacks (P?0.60). Results remained the same
when the SAS-callable SUDAAN survival procedure was
used to check for changes in stroke relative risk (RR) after
using sampling weights. To exclude nonlinear associations,
RR was estimated in a model with dummy variables for BMD
tertiles for white women. With the first tertile serving as the
reference group, the RR of stroke in the highest tertile was
0.75 (95% CI, 0.47 to 1.22, P?0.24) in multivariate analyses.
In an analysis of BMD and stroke mortality (any mention on
death certificate, n?169), the 3 race-sex groups were com-
bined (controlling for sex and race in addition to other
variables) because of the small number of deaths from stroke.
No association between BMD and stroke mortality was found
(RR, 1.03; 95% CI, 0.86 to 1.23; P?0.77).
This article reports one of the few longitudinal studies of BMD
and incidence of stroke and is the first to use a biracial cohort of
women and men. No association of phalangeal BMD was found
with the subsequent development of stroke. Controlling for
baseline age and other potential confounding factors failed to
measured.) This is consistent with the fact that no satisfactory
causal mechanism for the previously reported association of
BMD may serve as a marker of other causes of stroke in some
samples. For example, the possibility that both osteopenia and
stroke may result from low relative lifetime exposure to estrogen
in women or low vitamin D and its metabolites, which might
affect vascular reactivity, has been raised.3,4Osteoprotegerin, a
cytokine that regulates osteoclastogenesis, was not found to be
related to BMD or incidence of stroke in white women despite
associations with hypertension and diabetes.11A diet high in
protein (acids) and salt but low in potassium might aggravate
osteoporosis and increase risk of stroke.3,12Essential hyperten-
sion may be associated with increased mobilization of calcium
from bone.12Stroke-prone spontaneously hypertensive rats have
lower femoral BMD than Wistar-Kyoto control rats.13,14How-
ever, treatment with antihypertensive agents prevented stroke
but not decreased BMD.13Thus, further in vitro and in vivo
research is needed to determine whether a biological mechanism
exists for an association of BMD with stroke incidence reported
by others. It is important to exclude a causal association of BMD
with stroke because, in clinical practice, finding low BMD in a
woman may result in prescribing estrogen-replacement therapy,
which itself may increase the risk of stroke.15
In summary, in a large national study, no significant
associations of BMD and stroke incidence or mortality were
found for whites or blacks.
1. Browner WS, Seeley DG, Vogt TM, Cummings SR. Non-trauma mor-
tality in elderly women with low bone mineral density. Lancet. 1991;
2. Browner WS, Pressman AR, Nevitt MC, Cauley JA, Cummings SR.
Association between low bone mineral density and stroke in elderly
women. The Study of Osteoporotic Fractures. Stroke. 1993;24:940–946.
3. Jorgensen L, Engstad T, Jacobsen B. Bone mineral density in acute stroke
patients: low bone mineral density may predict first stroke in women.
4. Mussolino ME, Gillum RF, Madans J. Bone mineral density and stroke
risk. Stroke. 2001;32:2956–2957.
5. Cox CS, Mussolino ME, Rothwell ST, et al. Plan and operation of the
NHANES I Epidemiologic Follow-up Study, 1992: National Center for
Health Statistics. Vital Health Stat 1. 1997;35:1–231.
6. CompuMed Inc. OsteoGram Radiographic Absorptiometry of NHANES I
Radiographs. Manhattan Beach, Calif: CompuMed Inc; 1994.
7. Yang S, Hagiwara S, Engelke K, et al. Radiographic absorptiometry for
bone mineral measurement of the phalanges: precision and accuracy
study. Radiology. 1994;192:857–859.
RRs* for Stroke Incidence According to Bone Mineral Density
Measurement in Persons Aged 45–74 Years at Baseline:
NHANES I Epidemiologic Follow-up Study
*RRs and 95% CIs are based on a decrease of 1 SD in bone mineral density.
†Adjusted for baseline age, smoking status, alcohol consumption, history of
diabetes, history of heart disease, education, body mass index, recreational
physical activity, and blood pressure medication. Sex is included in age- and
risk-adjusted models for blacks.
8. Cosman F, Herrington B, Himmelstein S, Lindsay R. Radiographic
absorptiometry: a simple method for determination of bone mass.
Osteoporos Int. 1991;2:34–38.
9. The PHREG Procedure. Cary, North Carolina: SAS Institute, Inc;
1991:1–59. SAS technical report P-217.
10. Shah BV, Barnwell BG, Bieler GS. SUDAAN User’s Manual: Software
for Analysis of Correlated Data. Release 6.40. Research Triangle Park,
NC: Research Triangle Institute; 1995.
11. Browner WS, Lui LY, Cummings SR. Associations of serum osteopro-
tegerin levels with diabetes, stroke, bone density, fractures and mortality
in elderly women. J Clin Endocrinol Metab. 2001;86:631–637.
12. MacGregor GA. Salt’s more adverse effects. Am J Hypertens. 1997;10:
13. Naito S, Ito M, Sekine I, Hirano T, Iwaski K, Niwa M. Femoral head
necrosis and osteopenia in stroke-prone spontaneously hypertensive rats
(SHRSPs). Bone. 1993;14:745–753.
14. Yamori Y, Fukuda S, Tsuchikura S, Ikeda K, Nara Y, Horie R.
Stroke-prone SHR (SHRSP) as a model for osteoporosis. Clin Exp
Hypertens A. 1991;13:755–762.
15. Risks and benefits of estrogen plus progestin in healthy postmenopausal
women: principal results from the Women’s Health Initiative randomized
controlled trial. JAMA. 2002;288:321–333.
Mussolino et alBone Mineral Density and Stroke