Are Breast Density and Bone Mineral Density
Independent Risk Factors for Breast Cancer?
ARTICLE
Karla Kerlikowske, John Shepherd, Jennifer Creasman, Jeffrey A. Tice, Elad Ziv,
Steve R. Cummings
Background: Mammographic breast density and bone mineral
density (BMD) are markers of cumulative exposure to estrogen. Previous studies have suggested that women with high
mammographic breast density or high BMD are at increased
risk of breast cancer. We determined whether mammographic
breast density and BMD of the hip and spine are correlated and
independently associated with breast cancer risk. Methods: We
conducted a cross-sectional study (N = 15 254) and a nested
case–control study (of 208 women with breast cancer and 436
control subjects) among women aged 28 years or older who
had a screening mammography examination and hip BMD
measurement within 2 years. Breast density for 3105 of the
women was classified using the American College of Radiology
Breast Imaging Reporting and Data System (BI-RADS) categories, and percentage mammographic breast density among
the case patients and control subjects was quantified with a
computer-based threshold method. Spearman rank partial correlation coefficient and Pearson’s correlation coefficient were
used to examine correlations between BI-RADS breast density and BMD and between percentage mammographic breast
density and BMD, respectively, in women without breast cancer. Logistic regression was used to examine the association of
breast cancer with percentage mammographic breast density
and BMD. All statistical tests were two-sided. Results: Neither
BI-RADS breast density nor percentage breast density was
correlated with hip or spine BMD (correlation coefficient =
−.02 and −.01 for BI-RADS, respectively, and −.06 and .01
for percentage breast density, respectively). Neither hip BMD
nor spine BMD had a statistically significant relationship with
breast cancer risk. Women with breast density in the highest
sextile had an approximately threefold increased risk of breast
cancer compared with women in the lowest sextile (odds ratio =
2.7, 95% confidence interval = 1.4 to 5.4); adjusting for hip
or spine BMD did not change the association between breast
density and breast cancer risk. Conclusion: Breast density is
strongly associated with increased risk of breast cancer, even
after taking into account reproductive and hormonal risk factors, whereas BMD, although a possible marker of lifetime exposure to estrogen, is not. Thus, a component of breast density
that is independent of estrogen-mediated effects may contribute
to breast cancer risk. [J Natl Cancer Inst 2005;97:368–74]
Although increased mammographic breast density is one of
the strongest known risk factors for breast cancer (1–3), little is
known about why it is associated with breast cancer risk. Established breast cancer risk factors are associated with both increased
and decreased mammographic breast density. For example, increasing age and menopause are independent contributors to the
observed decrease in breast density that occurs with aging (4).
368 ARTICLES
Older age at first childbirth is associated with increased breast
density (5–7), whereas pregnancy at an early age is associated
with decreased breast density (5,8). Initiation of postmenopausal
hormone therapy that includes progesterone is associated with
increases in breast density, and discontinuation of this therapy is
associated with decreases in breast density (9–11). Together,
these findings suggest that increased breast cancer risk with increasing breast density may reflect, in part, cumulative estrogen
effects on breast tissue (12).
In addition to its effects on breast tissue, estrogen can also affect bone mineral density (BMD). Indeed, BMD may be a marker
of lifetime exposure to estrogen. Consistent with a role as such a
marker, BMD is a risk factor for breast cancer. The Study of Osteoporotic Fractures (13) demonstrated that women in the highest
quartile of distal radius BMD or metacarpal cortical bone mass
had a two- to threefold higher incidence of breast cancer than
women in the lowest BMD quartile. Three subsequent studies
have found weaker associations. The Fracture Intervention Trial
(14) found that women in the highest quartile of hip BMD had a
non–statistically significant 1.5-fold higher incidence of breast
cancer than women in the lowest quartile. The Rotterdam study
(15) found that women in the highest tertile of spine BMD had a
twofold higher incidence of breast cancer than women in the
middle tertile but found no association with hip BMD and incidence of breast cancer. The Dubbo Osteoporosis Epidemiology
Study (16) found a twofold higher incidence of breast cancer in
women with increased spine BMD and a modest increase in
women with increased hip BMD as compared with women with
low BMD.
Although both BMD and mammographic density are risk
factors for breast cancer, and both may be markers of estrogen
exposure, the relationship between mammographic breast density and BMD has not been well studied. Here, we evaluated
the relationship between breast density and hip and spine BMD
and the relationship between breast density in combination
with hip and spine BMD and breast cancer risk among women
participating in the San Francisco Mammography Registry
(SFMR).
Affiliations of authors: Department of Epidemiology and Biostatistics, University of California, San Francisco, CA (KK, SRC); General Internal Medicine Section, Department of Veterans Affairs, University of California, San Francisco, CA
(KK); Department of Medicine, University of California, San Francisco, CA (KK,
JC, JAT, EZ); Department of Radiology, University of California, San Francisco,
CA (JS); San Francisco Coordinating Center, California Pacific Medical Center
Research Institute, San Francisco (SRC).
Correspondence and reprint requests to: Karla Kerlikowske, MD, San Francisco
Veterans Affairs Medical Center, General Internal Medicine Section, 111A1,
4150 Clement Street, San Francisco, CA 94121 (e-mail: [email protected]).
See “Notes” following “References.”
DOI: 10.1093/jnci/dji056
Journal of the National Cancer Institute, Vol. 97, No. 5, © Oxford University
Press 2005, all rights reserved.
Journal of the National Cancer Institute, Vol. 97, No. 5, March 2, 2005
SUBJECTS
AND
METHODS
Subjects
The SFMR (http://mammography.ucsf.edu/SFMR/) is a populationbased mammography registry of women undergoing mammography in any of 16 radiology facilities in San Francisco and Marin
Counties. The present study included women aged 28 years or
older in the registry who underwent a bilateral mammography examination in San Francisco, indicated by the radiologist as being
performed for screening, and a BMD measurement of the hip
within 2 years of each other at the University of California, San
Francisco (UCSF); California Pacific Medical Center (CPMC); or
San Francisco Kaiser Permanente Medical Center (KPMC) between February 1, 1992, and December 31, 2002. We excluded all
women who had a diagnosis of breast cancer before their first
screening examination in the SFMR because treatments for breast
cancer, including tamoxifen and radiation, may alter breast density on subsequent examinations (17) and because mammograms
performed before the establishment of the SFMR are not easily
accessible. We also excluded women who had had breast augmentation, reduction, or reconstruction; history of mastectomy; or bilateral breast cancer. We included women with invasive breast
cancer or ductal carcinoma in situ (as case subjects) in the analysis if the screening mammogram and BMD measurement occurred
before their breast cancer diagnosis. At least two women without
breast cancer (control subjects) were selected from the same
mammography and BMD facility as case subjects. Breast cancer
case subjects were identified by linkage of the SFMR with the
Northern California Surveillance, Epidemiology, and End Results
(SEER1) program and the California Cancer Registry.
Annual approval was obtained from the UCSF Institutional
Review Board to collect registry and BMD information that included a waiver of signed consent.
Measurements and Procedures
Screening mammography examinations were linked to clinical
BMD databases at UCSF, CPMC, and KPMC to identify women
who had had a total hip (N = 15 254) or total spine (lumbar vertebrae
one to four; N = 14 475 of the 15 254) BMD measurement by dual
X-ray absorptiometry within 2 years of a screening mammography
examination. If a woman had more than one screening examination
or more than one BMD measurement, we selected the mammography examination and BMD measurement that were performed closest in time to each other. BMD was measured in g/cm2 using a
Hologic Delphi/A or W scanner at UCSF, a Hologic Delphi/W scanner at CPMC, and a Hologic QDR 1000 scanner at KPMC (Hologic,
Inc., Waltham, MA). Peak bone mass and standard deviation (SD)
were based on reference data supplied by the manufacturer. These
devices were maintained using standard quality control procedures
recommended by the manufacturer to assure that the BMD calibrations remained constant within plus or minus 1%. Although no attempt was made to cross-calibrate the devices, studies on Hologic
dual X-ray absorptiometry devices in clinical practice have shown
them to be within plus or minus 2% of each other (18).
We retrieved screening examinations to digitize and measure
percentage mammographic breast density in 208 women with
breast cancer (case subjects) and a random sample of 436 women
without breast cancer (control subjects). For women with invasive breast cancer or ductal carcinoma in situ, we selected the
Journal of the National Cancer Institute, Vol. 97, No. 5, March 2, 2005
craniocaudal screening examination of the breast that did not
have breast cancer. The craniocaudal view was selected because
it excludes the pectoralis muscle, which has been shown to create
artifacts when measuring breast density (19). For the control subjects, we selected either the right or left craniocaudal view because breast density measures of the right and left breast are
highly correlated (19).
Mammographic breast density was quantified using a validated computer-based threshold method as described previously
(3). The participant’s craniocaudal screening view was digitized
on a Lumisys LumiScan 200 radiographic films digitizer (Kodak,
Inc.) (100-mm pixel size, 12-bit dynamic range) and archived
onto a CD jukebox. This semiautomated, computer-assisted
method involves dividing the mammographic image into a distribution of gray values, with darker regions indicative of fat tissue
and lighter regions representing dense tissue. The method is
based on the interactive selection of two thresholds in the image
of a digitized mammogram. One threshold separates the breast
image from the background (breast area) and the other identifies
the regions that represent radiographically dense tissue (mammographic density). The percentage of dense tissue in the breast
was determined by dividing the number of pixels outlined in the
dense regions by the total area of the breast as calculated with
dedicated computer software (3). A single radiologist trained in
assessing mammographic breast density with the UCSF Mammography Density Workstation read all study films from case and
control subjects (N = 644) (3). A qualitative assessment of breast
density was also assigned in clinical practice for women (N =
3105) undergoing screening mammography at UCSF, but not
the other centers, using breast density categories established by
the American College of Radiology and reported in the Breast
Imaging Reporting and Data System (BI-RADS) (20). The four
breast density categories were as follows: 1) almost entirely fat,
2) scattered fibroglandular densities, 3) heterogeneously dense,
and 4) extremely dense.
At the time of each screening examination, women completed
a survey that included demographic information and breast
health history questions. The survey included the following
race/ethnicity categories: 1) African American/Black, 2) Caucasian/White, 3) Hispanic/Latina, 4) American Indian, 5) Chinese,
6) Japanese, 7) Filipina, 8) Vietnamese, 9) other Asian, and 10)
other non-Asian. In addition, the survey included questions
about family history of breast cancer in a first-degree relative
(mother, sister, or daughter), current postmenopausal hormone
therapy use, menopausal status, age at first live birth, and weight
and height. Women were considered to be current hormone therapy users if they reported using female hormones for menopause
at the time of a screening examination. Women were considered
to be postmenopausal if both ovaries had been removed, if they
reported that their periods had stopped permanently, if they were
using hormone therapy, or if they were aged 55 years or older.
Age at first live birth was stratified into three categories: younger
than 30 years, 30 years or older, and nulliparous. Height and
weight were determined using self-reported information collected at the time of either screening mammography or BMD
measurement.
Data and Statistical Analysis
Frequency distributions of demographic characteristics and
clinical risk factors were computed for all women with a screening
ARTICLES
369
examination and BMD of the hip or spine within 2 years of each
other. For all analyses, race/ethnicity was collapsed into three
categories that include white, Asian or Pacific Islander, and all
other races. For women without breast cancer, the association of
breast density with clinical risk factors was computed by determining the proportion of women with a BI-RADS density of 3 or
4 and the average percentage breast density for those who had a
quantitative assessment of breast density. The association of
BMD with clinical risk factors was computed by determining
mean hip and spine BMD for all women without breast cancer.
Statistically significant differences for categorical variables were
determined using a chi-square test and logistic regression when
adjusting for age. Statistically significant differences for continuous variables were determined using standard t test and t test of
adjusted means when adjusting for age.
Spearman rank partial correlation coefficients were used to
assess the overall association between BMD measured as a continuous variable and BI-RADS breast density. Pearson’s correlation coefficient was used to assess the overall association between
BMD measured as a continuous variable and percentage breast
density as a continuous variable. We also performed correlation
analyses stratified by age and menopausal status and postmenopausal hormone therapy use adjusting for body mass index, race/
ethnicity, and age at first live birth. We stratified women into two
groups (younger than aged 65 years and aged 65 years or older)
because the strongest association between BMD and breast cancer risk has been found among an elderly population (13), suggesting that if breast density and BMD are correlated, then the
strongest correlation may also be observed in the elderly.
For case and control subjects, we calculated mean time between
screening examinations and BMD measurements. In addition, we
calculated the mean time between screening mammography and
breast cancer diagnosis and between BMD measurement and cancer diagnosis. Chi-square tests were used to determine statistical
significance when comparing clinical and demographic characteristics of case and control subjects.
Multivariable logistic regression was used to determine
whether percentage mammographic breast density and BMD
are independent risk factors for breast cancer after adjusting for
age at diagnosis, race/ethnicity, family history of breast cancer,
age at first live birth, and body mass index. BMD was divided
into quartiles, and percentage mammographic density was divided into sextiles based on the distribution of women without
breast cancer.
The attributable risks of breast cancer for varying percentages
of mammographic breast density were calculated using the formula Attributable risk = (RR-1)Pc/RR, in which RR is the relative risk associated with a given percentage mammographic
density and Pc is the prevalence of that category in the breast
cancer case subjects. The odds ratio (see Table 6) was assumed to
approximate the relative risk.
All statistical calculations were performed using SAS (version 8.2; SAS Institute; Cary, NC). Tests resulting in P values
equal to or less than 0.05 were considered statistically significant.
All statistical tests were two-sided.
RESULTS
Between February 1, 1992, and December 31, 2002, 15 254
women had a screening mammography examination and hip
370 ARTICLES
BMD measurement within 2 years of each other. Of these
women, 3105 were assigned a BI-RADS density measure and
644 had a mammographic breast density reading. The average
age of the study population was 60.1 years, and the majority was
menopausal, with 57% currently using postmenopausal hormone
therapy (Table 1).
Of the 265 women who had a screening examination and hip
BMD measurement before being diagnosed with breast cancer, we
were able to locate mammograms for 208 women. Of these 208
women, 153 had a diagnosis of invasive cancer and 55 had a diagnosis of ductal carcinoma in situ. The mean time (± SD) between
the screening mammography examination and cancer diagnosis
was 2.1 ± 1.6 years and between BMD measurement and breast
cancer diagnosis was 2.1 ± 1.6 years. The mean time between the
screening mammography examination and the BMD measurement
was 0.3 ± 0.4 years for women later diagnosed with breast cancer
and 0.4 ± 0.4 years for women without breast cancer.
To investigate the validity of BI-RADS density assessments,
percentage mammographic breast density measurements, and BMD
measurements, we examined the association between the reported
demographic and clinical characteristics (5,13) and breast density
or BMD. We found that the proportion of women without breast
cancer with BI-RADS density assessment categories 3 or 4 and
mean percentage mammographic density decreased with increasing
Table 1. Clinical and demographic characteristics of women with a hip (N =
15 254) or spine (N = 14 475) bone mineral density (BMD) test and screening
mammography within 2 years
Characteristic
Age, y†
Body mass index, kg/m2†
Hip BMD, g/cm2†
Spine BMD, g/cm2†
% of patients* (N)
60.1 ± 10.6 (15 254)
24.1 ± 4.6 (14 502)
0.82 ± 0.13 (15 254)
0.93 ± 0.16 (14 475)
BI-RADS density
Fat
Scattered
Heterogeneous
Extremely dense
8 (232)
46 (1429)
41 (1279)
5 (165)
Family history of breast cancer‡
No
Yes
84 (12 147)
16 (2312)
Age at first live birth
Nulliparous
<30 y
≥30 y
36 (5221)
47 (6784)
17 (2517)
Postmenopausal hormone therapy
No
Yes
43 (4023)
57 (5305)
Menopause
No
Yes
11 (1571)
89 (12 911)
Race/ethnicity
White
Asian/Pacific Islander§
Other║
65 (9265)
23 (3369)
12 (1748)
*Missing values: 5% for body mass index, family history of breast cancer,
spine BMD, and ages at first live birth and menopause; 80% for BI-RADS density; 39% for postmenopausal hormone therapy; and 6% for race/ethnicity.
†Means ± standard deviation.
‡First-degree relative (mother, sister, or daughter) with breast cancer.
§Chinese, Japanese, Vietnamese, Filipina, or other Asian.
║Includes 3% Hispanic, 2% Black, and 7% other races/ethnicities.
Journal of the National Cancer Institute, Vol. 97, No. 5, March 2, 2005
age, menopause, and body mass index and increased with late age
at first live birth and postmenopausal hormone use (Table 2). As
expected, both hip and spine BMD decreased with increasing age
and menopause and increased with increasing body mass index and
postmenopausal hormone use (Table 3). Asian women had a lower
age-adjusted hip BMD than white women (0.792 ± 0.002 versus
0.834 ± 0.001, respectively); the difference was attenuated but not
eliminated after adjusting for body mass index and postmenopausal
hormone use (data not shown). Thus, the BI-RADS density assessments, percentage mammographic breast density measurements,
and BMD measurements appear to be valid measures.
We next examined associations between BMD and BI-RADS
assessments and between BMD and percentage breast density
measurements. Neither hip nor spine BMD was associated with
BI-RADS assessments among women without breast cancer, regardless of age or postmenopausal hormone therapy use (Table 4).
In addition, neither was associated with percentage mammographic
breast density among women without breast cancer, regardless of
age or postmenopausal hormone therapy use (Table 4).
Among women with a percentage breast density measurement
(n = 644), those with breast cancer were more likely to have a
family history of breast cancer, to be white, and to be nulliparous
than women without breast cancer (Table 5). In addition, the mean
percentage mammographic breast density was statistically significantly higher in women with breast cancer than in women without
breast cancer (49.2% versus 44.6%, P = .006; Table 5). Mean hip
and spine BMD did not vary between case and control subjects.
Table 2. Association of clinical factors with BI-RADS and percentage mammographic breast density among women without breast cancer
Variable
Age, y
<50
≥50
% of women with
BI-RADS 3 or 4
(n = 3057)
65.6
43.9
P*
% BD ± SE*
(n = 436)
P*
<.001
58.2 ± 3.9
43.7 ± 1.0
<.001
.80
43.0 ± 2.1
44.9 ± 1.0
.40
<.001
49.4 ± 1.3
40.1 ± 1.3
<.001
Postmenopausal hormone therapy†
Yes
55.2
No
43.1
<.001
48.3 ± 1.5
38.3 ± 1.5
<.001
Menopause
Yes
No
43.6
66.6
<.001
43.6 ± 1.0
57.6 ± 4.3
.002
Race/ethnicity†
White
Asian/Pacific Islander§
Other║
46.1
55.5
46.1
<.001
44.8 ± 1.1
45.6 ± 1.9
43.4 ± 2.9
.48
Body mass index (kg/m2)†
<25
≥25
57.7
30.8
<.001
49.9 ± 1.1
37.6 ± 1.8
<.001
Family history of breast cancer†‡
Yes
47.0
No
47.5
Age at first live birth†
≥30 y or nulliparous
<30 y
54.5
40.0
*Mean percent mammographic breast density (% BD) ± standard errors (SE).
P values were determined using the chi-square test or linear least-squares regression, as appropriate.
†Age-adjusted proportion or mean.
‡First-degree relative with breast cancer.
§Chinese, Japanese, Vietnamese, Filipina, or other Asian.
║Includes Hispanic, Black, and other races/ethnicities.
Journal of the National Cancer Institute, Vol. 97, No. 5, March 2, 2005
Table 3. Association of clinical factors with hip and spine bone mineral density
(BMD) among women without breast cancer*
Variable
Hip BMD
(g/cm2)
(N = 14 989)
Age, y
<50
≥50
.872 ± .003
.816 ± .001
Family history of breast cancer†‡
Yes
.829 ± .002
No
.826 ± .001
Age at first live birth†
≥30 y or nulliparous
.824 ± .001
<30 y
.829 ± .002
Postmenopausal hormone therapy†
Yes
.837 ± .002
No
.810 ± .002
Menopause
Yes
.814 ± .001
No
.887 ± .003
Race/ethnicity†
White
.834 ± .001
Asian/Pacific Islander§ .792 ± .002
Other║
.859 ± .003
Body mass index (kg/m2)†
<25
.798 ± .001
≥25
.879 ± .002
P
Spine BMD
(g/cm2)
(N = 14 227)
<.001
.981 ± .003
.916 ± .001
.35
.934 ± .001
.926 ± .003
P
<.001
.02
.02
.931 ± .002
.923 ± .002
.006
<.001
.950 ± .002
.892 ± .002
<.001
<.001
.912 ± .002
1.002 ± .004
<.001
<.001
.943 ± .002
.881 ± .003
.935 ± .004
<.001
<.001
.909 ± .002
.964 ± .002
<.001
*Means ± standard errors. P values were derived from linear least-squares
regression.
†Age-adjusted means.
‡First-degree relative (mother, sister, or daughter) with breast cancer.
§Chinese, Japanese, Vietnamese, Filipina, or other Asian.
║Includes Hispanic, Black, and other races/ethnicities.
In a multivariable model, age, family history of breast cancer,
and breast density were each statistically significantly associated
with increased risk of breast cancer (Table 6). Controlling for hip
BMD did not change the association between breast density and
breast cancer risk (Table 6), nor did controlling for spine BMD.
In addition, results were similar when the outcome was invasive
breast cancer, i.e., excluding women with ductal carcinoma in situ,
and when the outcome was breast cancer preceded by a positive
mammography result, excluding interval cancers (data not shown).
Separate logistic models for women aged younger than 65 years
and aged 65 or older yielded results similar to those of the overall
model (data not shown). Because BMD was not statistically significantly associated with risk of breast cancer in any model, we
did not test for an interaction between breast density and BMD.
We determined the proportion of breast cancers attributable
to breast density for women in each category of percentage
mammographic density (Table 7). For women with a percentage
mammographic density of greater than 42.6%, the attributable
risk of breast cancer was 38%; for women with a percentage
mammographic density of greater than 54%, the attributable risk
of breast cancer was 29%; and for women with a percentage mammographic density between 23.0% and 42.7%, the attributable
risk of breast cancer was only 4% (Table 7).
DISCUSSION
We found that mammographic breast density, measured either
qualitatively with BI-RADS density assessment categories or
quantitatively with a computer-based threshold method, showed
ARTICLES
371
no correlation with BMD of the hip or spine. In addition, mammographic breast density, but not BMD, was a strong risk factor
for breast cancer. If the association between mammographic density and breast cancer and between BMD and breast cancer are
both primarily mediated by estrogen (12), then mammographic
breast density and BMD should have been correlated. We found
no such correlation. Consistent with our results, no correlation
was found between BI-RADS breast density categories and hip
BMD among women enrolled in randomized controlled trials of
treatment for osteoporosis (21).
Factors that increase the exposure of breast tissue to estrogens, such as older age at first birth, menopause, obesity, or postmenopausal hormone therapy, have been shown to be associated
with an increased risk of breast cancer (22–24). Consistent with
Table 4. Correlation between hip and spine bone mineral density (BMD) and
breast density by age and postmenopausal hormone therapy use among women
without breast cancer
Variables
Hip BMD & BI-RADS density
Overall†
Age, <65 y†
Age, ≥65 y†
Premenopausal‡
Postmenopausal, uses
hormone therapy‡
Postmenopausal, does not
use hormone therapy‡
Spine BMD & BI-RADS density
Overall
Age, <65 y†
Age, ≥65 y†
Premenopausal‡
Postmenopausal, uses
hormone therapy‡
Postmenopausal, does not
use hormone therapy‡
No. of subjects
Coefficient*
−.02
−.01
−.06
−.00
−.00
.39
.60
.12
.96
.92
982
−.06
.06
1629
1121
508
242
635
−.01
−.01
−.04
−.04
−.00
.66
.82
.42
.50
.98
752
−.03
.43
424
239
185
22
158
−.06
−.03
−.11
−.00
−.06
.22
.65
.12
1.0
.47
149
−.11
.17
Spine BMD & percentage mammographic
breast density
Overall§
405
Age, <65 y†
231
Age, ≥65 y†
174
Premenopausal‡
20
Postmenopausal, uses
153
hormone therapy║
Postmenopausal, does not
142
use hormone therapy║
.01
−.01
.01
.09
.00
.87
.89
.95
.71
1.0
.06
.48
*Spearman partial-correlation coefficient for BI-RADS density and Pearson’s
correlation coefficient for percentage mammographic density.
†Adjusted for age, body mass index, race/ethnicity, age at first live birth, and
postmenopausal hormone therapy.
‡Adjusted for age, body mass index, race/ethnicity, and age at first live birth.
§Adjusted for age, body mass index, race/ethnicity, and age at first live birth;
95 missing postmenopausal hormone therapy.
║Adjusted for age, body mass index, race/ethnicity, and age at first live birth;
90 missing postmenopausal hormone therapy.
372 ARTICLES
Variable*
Age, y†
Family history of breast
cancer‡
Age at first live birth:
nulliparous or ≥30 y
Postmenopausal hormone
therapy
Menopausal
Race/ethnicity
White
Asian/Pacific Islander§
Other║
Body mass index (kg/m2)†
Percentage mammographic
breast density†
Hip BMD (g/cm2)†
Spine BMD (g/cm2)†
Case subjects
(n = 208)
Control subjects
(n = 436)
P
63.6 ± 9.4
28.9%
63.0 ± 9.4
19.6%
.42
.009
56.3%
53.9%
48.4%
48.2%
.06
.2
95.5%
94.8%
.7
76.2%
15.3%
8.4%
23.6 ± 4.1
49.2 ± 19
66.7%
23.3%
9.9%
23.5 ± 3.7
44.6 ± 20
.04
0.816 ± .13
0.921 ± .16
0.810 ± .13
0.911 ± .15
.5
.4
.8
.006
P
2016
1345
671
291
743
Hip BMD & percentage mammographic
breast density
Overall§
Age, <65 y†
Age, ≥65 y†
Premenopausal‡
Postmenopausal, uses
hormone therapy§
Postmenopausal, does not
use hormone therapy§
Table 5. Comparison of clinical and demographic characteristics of women with
breast cancer with a bone mineral density (BMD) test and breast mammogram
before being diagnosed with breast cancer and women without breast cancer
*All results were adjusted for age, with the exception of mean age and menopausal status.
†Mean ± SD. P values were derived using logistic regression.
‡First-degree relative with breast cancer.
§Chinese, Japanese, Vietnamese, Filipina, or other Asian.
║Includes Hispanic, Black, and other races/ethnicities.
the hypothesis that prolonged exposure to endogenous estrogens
influences breast cancer risk is the observation that increased
postmenopausal estradiol levels are associated with increased
breast cancer risk (25). High estradiol levels among postmenopausal
Table 6. Multivariable model of factors associated with breast cancer*
Variable
Case-control subjects
OR (95% CI)*
Age per 5 y
Family history of breast cancer†
Age at first live birth:
nulliparous or ≥30 y
1.1 (1.0 to 1.3)
1.6 (1.1 to 2.4)
1.2 (0.8 to 1.7)
Percentage mammographic breast density by sextiles
<23.9%
23.9–34.2%
34.3–42.6%
42.7–54.0%
54.1–66.7%
≥66.8%
1.0 (referent)
1.2 (0.6 to 2.3)
1.2 (0.6 to 2.4)
1.9 (1.0 to 3.7)
2.8 (1.5 to 5.4)
2.7 (1.4 to 5.4)
Hip BMD, g/cm2§
<0.731
0.731–0.809
0.810–0.880
≥0.881
1.0 (referent)
1.1 (0.7 to 1.9)
1.3 (0.8 to 2.1)
1.2 (0.7 to 2.1)
Race/ethnicity
White or other race
Asian/Pacific Islander‡
1.0 (referent)
0.7 (0.4 to 1.1)
Body mass index, kg/m2
<25
≥25
1.0 (referent)
1.2 (0.9 to 1.5)
*Model adjusted for all variables in the table with 200 of 208 case subjects and
431 of 436 control subjects, for whom we had data for all variables. OR = odds
ratio; CI = confidence interval; BMD = bone mineral density.
†First-degree relative with breast cancer.
‡Chinese, Japanese, Vietnamese, Filipina, or other Asian.
§Equal-sized quartiles among women without breast cancer.
Journal of the National Cancer Institute, Vol. 97, No. 5, March 2, 2005
Table 7. Attributable risk of breast cancer due to breast density*
Breast density range†
Case-control
subjects RR
(95% CI)‡
<23.9%
23.9%–34.2%
34.3%–42.6%
42.7%–54.0%
54.1%–66.7%
≥66.8%
1.0 (referent)
1.2 (0.6 to 2.3)
1.2 (0.6 to 2.4)
1.9 (1.0 to 3.7)
2.8 (1.5 to 5.4)
2.7 (1.4 to 5.4)
Percentage of case
subjects in breast
density range
Attributable
risk
10%
14%
11%
19%
25%
21%
Referent
2%
2%
9%
16%
13%
*RR = relative risk; CI = confidence interval.
†Breast density was quantified with a computer-based threshold method (3).
Cutpoints for each range are based on equal-sized sextiles among women without
breast cancer.
‡Odds ratios are from Table 6 and are assumed to approximate relative risks.
RRs were derived from 200 of 208 case subjects and 431 of 436 control subjects,
for whom we had complete data.
women are also positively associated with high BMD (26,27),
and some studies have shown that postmenopausal women with
the highest BMD are at increased risk of breast cancer risk
(13,28). However, one study has shown that BMD may not be
associated with breast cancer risk independent of its relationship
with endogenous hormones and body mass index (29), suggesting that increasing BMD with increasing body mass index among
postmenopausal women is largely the result of the associated increase in serum estrogens (30). By comparison, mammographic
breast density does not appear to be strongly associated with serum estradiol levels among postmenopausal women (31). Moreover, in our study, the strength of the association of increased
breast density and breast cancer risk is not affected by controlling
for body mass index. Thus, mammographic density (1,2) and
post-menopausal hormone levels (25) are both strongly associated with breast cancer risk and may act independently of each
other, whereas BMD may not act independently of its relationship with endogenous estrogens.
What mechanisms other than cumulative estrogen exposure
may account for the association of increased breast density and
risk of breast cancer? First-degree relatives of women with increased mammographic density have an increased risk of developing breast cancer (32). Thus, genes that determine breast
density may also affect breast cancer risk. Growth factors that
affect the breast have also been shown to be associated with
mammographic density (31,33). Studies have demonstrated that,
in premenopausal women, mammographic density is associated
with higher insulin-like growth factor 1 levels, which in turn are
associated with an increased risk of breast cancer (31,33).
We did not find that BMD was associated with risk of breast
cancer. Our study has twice as many women with breast cancer as
other studies (13,14,16,28), so a lack of statistical power is unlikely to explain our findings. In the study with the strongest association between BMD and breast cancer risk (13), the study
population had a mean age of 71 years and lower mean hip and
spine BMDs than those observed in our study (hip = 0.75 g/cm2
versus 0.82 g/cm2, respectively; spine = 0.84 g/cm2 versus 0.92 g/
cm2, respectively). In addition, the studies with the strongest association with BMD and breast cancer risk found the association
with appendicular (arm) BMD (13,28,34). Studies that measured
axial (hip and spine) BMD found weaker and more inconsistent
associations with BMD (14–16). Lastly, BMD is not merely a
marker for estrogen exposure. Other growth factors that we did not
measure may be involved in the association between BMD and
Journal of the National Cancer Institute, Vol. 97, No. 5, March 2, 2005
breast cancer. It is also possible that different methods were used to
assess spine and hip BMD in different facilities, even though these
measurements were made on the same types of machine. Any variation in this measurement across clinical practices could limit our
ability to find an association with BMD and breast cancer. Our inability to corroborate research studies that used measurements
made in a research protocol suggests that the hip and spine BMD
measured in clinical practice may not be a useful marker to predict
breast cancer risk in clinical practice.
Our study included a large sample size with primarily postmenopausal women of diverse racial and ethnic groups, which
increases the generalizability of the results. In addition, our study
is the first, to our knowledge, to report the associations of breast
density, BMD, and breast cancer and the largest study to date that
has examined the BMD–breast cancer relationship. One limitation of the study is that BMD was measured in clinical practice
for reasons not assessed by our study. If primarily healthy women
elect to undergo BMD measures in clinical practice, then a health
selection bias could be introduced, with lower-than-expected
numbers of breast cancers among women with high BMD limiting our ability to find an association between BMD and breast
cancer. A second potential limitation is the possibility of cancer
detection bias. However, the cancer rates reported for the SFMR
(35) are within the range of those reported in the literature, in
which follow-up has been reported to be 99.6% (36). In addition,
cancer reporting to the SFMR from the SEER program has been
estimated to be more than 94.3% complete (37). Thus, cancer
detection bias is unlikely to explain our results. Lastly, masking
bias is a possibility. However, this bias is not likely to explain our
results because, after we excluded cancers associated with a normal mammography result from the analyses, we observed the
same association between breast density and breast cancer risk
and between BMD and breast cancer risk as we did when such
cancers were included.
Our findings suggest that breast density is strongly associated
with increased risk of breast cancer and that BMD, although a
marker of lifetime exposure to estrogen, is not. Given that breast
density is independently associated with an increased risk of
breast cancer after taking into account reproductive and hormonal
risk factors, it is possible that a component of breast density that
is not estrogen mediated may contribute to breast cancer risk.
This theory is supported by the observation that high breast density is as strongly associated with estrogen receptor–positive
breast cancer as with estrogen receptor–negative breast cancer
(38) and with breast cancer in premenopausal women as with
breast cancer in postmenopausal women (1). Further investigation into factors that have been shown to affect breast density and
breast cancer risk may reveal the biologic basis between breast
density and breast cancer.
BMD is measured by dual energy absorptiometry, an accurate, precise, and automated technique. BMD has become a
widely accepted clinical tool to assess fracture risk and to
guide the use of therapies to prevent fractures. By analogy,
breast density is a strong risk factor present in a substantial
proportion of breast cancer cases and therefore may eventually
be useful to assess breast cancer risk and guide the use of preventive therapies for breast cancer. An accurate, precise, and
automated means to assess breast density in clinical practice
is needed, however, before breast density can be used to identify women who may benefit from breast cancer prevention
measures.
ARTICLES
373
REFERENCES
(1) Byrne C, Schairer C, Wolfe J, Parekh N, Salane M, Brinton LA, et al. Mammographic features and breast cancer risk: effects with time, age, and menopause status. J Natl Cancer Inst 1995;87:1622–9.
(2) Boyd NF, Byng JW, Jong RA, Fishell EK, Little LE, Miller AB, et al. Quantitative classification of mammographic densities and breast cancer risk: results from the Canadian National Breast Screening Study. J Natl Cancer Inst
1995;87:670–5.
(3) Prevrhal S, Shepherd JA, Smith-Bindman R, Cummings SR, Kerlikowske K.
Accuracy of mammographic breast density analysis: results of formal operator training. Cancer Epidemiol Biomarkers Prev 2002;11:1389–93.
(4) Boyd NF, Martin L, Stone J, Little L, Minkin S, Yaffe M. A longitudinal
study of the effects of menopause on mammographic features. Cancer Epidemiol Biomarkers Prev 2002;11:1048–53.
(5) Vachon CM, Kuni CC, Anderson K, Anderson VE, Sellers TA. Association of mammographically defined percent breast density with epidemiologic risk factors for breast cancer (United States). Cancer Causes Control
2000;11:653–62.
(6) Salminen TM, Hakama M, Heikkila MM, Saarenmaa IE. Favorable change
in mammographic parenchymal patterns and breast cancer risk factors. Int J
Cancer 1998;78:410–4.
(7) Warwick J, Pinney E, Warren RML, Duffy SW, Howell A, Wilson M, Cuzick J. Breast density and breast cancer risk factors in a high-risk population.
Breast 2003;12:10–6.
(8) El-Bastawissi AY, White E, Mandelson MT, Taplin SH. Reproductive and
hormonal factors associated with mammographic breast density by age
(United States). Cancer Causes Control 2000;11:955–63.
(9) Rutter CM, Mandelson MT, Laya MB, Seger DJ, Taplin S. Changes in
breast density associated with initiation, discontinuation, and continuing use
of hormone replacement therapy. JAMA 2001;285:171–6.
(10) Greendale GA, Reboussin BA, Sie A, Singh HR, Olson LK, Gatewood O,
et al. Effects of estrogen and estrogen-progestin on mammographic parenchymal density. Postmenopausal Estrogen/Progestin Interventions (PEPI)
Investigators. Ann Intern Med 1999;130:262–9.
(11) Greendale GA, Reboussin BA, Slone S, Wasilauskas C, Pike MC, Ursin G.
Postmenopausal hormone therapy and change in mammographic density.
J Natl Cancer Inst 2003;95:30–7.
(12) Clemons M, Goss P. Estrogen and the risk of breast cancer. N Engl J Med
2001;344:276–85.
(13) Cauley JA, Lucas FL, Kuller LH, Vogt MT, Browner WS, Cummings SR.
Bone mineral density and risk of breast cancer in older women: the study
of osteoporotic fractures. Study of Osteoporotic Fractures Research Group.
JAMA 1996;276:1404–8.
(14) Buist D, LaCroixa A, Barlow W, White E, Weiss NS. Bone mineral density and breast cancer risk in postmenopausal women. J Clin Epidemiol
2001;54:417–22.
(15) van der Klift M, de Laet CE, Coebergh J, Hofman A, Pols H. Rotterdam
Study. Bone mineral density and the risk of breast cancer: the Rotterdam
Study. Bone 2003;32:211–6.
(16) Nguyen T, Center JR, Eisman JA. Association between breast cancer and
bone mineral density: the Dubbo Osteoporosis Epidemiology Study. Maturitas 2000;36:27–34.
(17) Cuzick J, Warwick J, Pinney E, Warren RM, Duffy SW. Tamoxifen and
breast density in women at increased risk of breast cancer. J Natl Cancer
Inst 2004;96:621–8.
(18) Blake G, Tong C, Fogelman I. Intersite comparison of the Hologic QDR1000 dual energy X-ray bone densitometer. Br J Radiol 1991;64:440–6.
(19) Byng JW, Boyd NF, Little LE, Lockwood GA, Fishell EK, Jong RA, Yaffe
MJ. Symmetry of projection in the quantitative analysis of mammographic
images. Eur J Cancer Prev 1996;5:319–27.
(20) American College of Radiology. The American College of Radiology Breast
Imaging Reporting and Data System (BI-RADS). 3rd ed. Reston (VA):
American College of Radiology;1998.
(21) Buist DS, Anderson ML, Taplin SH, LaCroix AZ. The relationship between
breast density and bone mineral density in postmenopausal women. Cancer
2004;101;1968–76.
(22) Kerlikowske K, Miglioretti DL, Ballard-Barbash R, Weaver D, Buist
DSM, Barlow W, et al. Prognostic characteristics of breast cancer among
374 ARTICLES
(23)
(24)
(25)
(26)
(27)
(28)
(29)
(30)
(31)
(32)
(33)
(34)
(35)
(36)
(37)
(38)
postmenopausal hormone users in a screened population. J Clin Oncol
2003;21:4314–21.
Rossouw JE, Anderson GL, Prentice RL, LaCroix AZ, Kooperberg C,
Stefanick ML, Jackson RD, Beresford SA, Howard BV, Johnson KC,
Kotchen JM, Ockene J; Writing Group for the Women’s Health Initiative
Investigators. Risk and benefits of estrogen plus progestin in healthy postmenopausal women: principal results From the Women’s Health Initiative
randomized controlled trial. JAMA 2002;288:321–33.
Harris JR, Lippman ME, Veronesi U, Willett W. Breast cancer. N Engl J
Med 1992;327:319–28.
Key T, Appleby P, Barnes I, Reeves G; The Endogenous Hormones and
Breast Cancer Collaborative Group. Endogenous sex hormones and breast
cancer in postmenopausal women: reanalysis of nine prospective studies.
J Natl Cancer Inst 2002;94:606–16.
Cauley J, Gutai JP, Kuller LH, LeDonne D, Sandler RB, Sashin D, Powell JG.
Endogenous estrogen levels and calcium intakes in postmenopausal women.
Relationships with cortical bone measures. JAMA 1988;260:3150–5.
Ettinger B, Pressman A, Sklarin P, Bauer DC, Cauley JA, Cummings SR.
Associations between low levels of serum estradiol, bone density, and factures among elderly women: the study of osteoporotic fractures. J Clin Endocrinol Metab 1998;83:2239–43.
Zhang Y, Kiel DP, Kreger BE, Cupples LA, Ellison RC, Dorgan JF, et al.
Bone mass and the risk of breast cancer among postmenopausal women.
N Engl J Med 1997;336:611–7.
Buist D, LaCroix A, Barlow W, White E, Cauley JA, Bauer DC, Weiss
NS. Bone mineral density and endogenous hormones and risk of breast
cancer in postmenopausal women (United States). Cancer Causes Control
2001;12:213–22.
Key TJ, Appleby PN, Reeves GK, Roddam A, Dorgan JF, Longcope C,
Stanczyk FZ, et al. Endogenous Hormones Breast Cancer Collaborative
Group. Body mass index, serum sex hormones, and breast cancer risk in
postmenopausal women. J Natl Cancer Inst 2003;95:1218–26.
Boyd NF, Stone J, Martin LJ, Jong R, Fishell E, Yaffe M, et al. The association of breast mitogens with mammographic densities. Br J Cancer
2002;87:876–82.
Ziv E, Shepherd J, Smith-Bindman R, Kerlikowske K. Mammographic
breast density and family history of breast cancer. J Natl Cancer Inst
2003;95:556–8.
Byrne C, Colditz GA, Willett WC, Speizer FE, Pollak M, Hankinson SE.
Plasma insulin-like growth factor (IGF) I, IGF-binding protein 3, and mammographic density. Cancer Res 2000;60:3744–8.
Nelson DA, Darga LL, Simon MS, Severson RK. Radial bone density and
breast cancer risk in white and African-American women. Osteoporos Int
2004;15:535–40.
Kerlikowske K, Carney P, Geller B, Mandelson MT, Taplin S, Malvin K,
et al. Performance of screening mammography among women with and
without a first-degree relative with breast cancer. Ann Intern Med 2000;133:
855–63.
Kerlikowske K, Barclay J. Outcomes of modern screening mammography.
Natl Cancer Inst Monogr 1997;22:105–11.
Ernster VL, Ballard-Barbash R, Barlow WE, Zheng Y, Weaver D, Cutter G,
et al. Detection of ductal carcinoma in situ in women undergoing screening
mammography. J Natl Cancer Inst 2002;94:1546–54.
Ziv E, Tice J, Smith-Bindman R, Shepherd JA, Cummings SR, Kerlikowske K.
Mammographic density and estrogen receptor status of breast cancer. Cancer Epidemiol Biomarkers Prev 2004;13:2090–5.
NOTES
1Editor’s note: SEER is a set of geographically defined, population-based, central cancer registries in the United States, operated by local nonprofit organizations under contract to the National Cancer Institute (NCI). Registry data are submitted electronically without personal identifiers to the NCI on a biannual basis,
and the NCI makes the data available to the public for scientific research.
This work was supported by a National Cancer Institute-funded Breast Cancer
Surveillance Consortium cooperative agreement (U01CA63740).
Manuscript received August 26, 2004; revised December 16, 2004; accepted
January 4, 2005.
Journal of the National Cancer Institute, Vol. 97, No. 5, March 2, 2005