Rutter CM, Mandelson MT, Laya MB, Taplin S. Changes in Breast Density Associated With Initiation, Discontinuation, and Continuing Use of Hormone Replacement Therapy. JAMA. 2001;285(2):171-176. doi:10.1001/jama.285.2.171
Author Affiliations: Group Health Cooperative of Puget Sound, Seattle, Wash (Drs Rutter, Mandelson, Laya, and Taplin and Ms Seger); Division of General Internal Medicine (Dr Laya), Departments of Biostatistics (Dr Rutter), Epidemiology (Dr Mandelson), Medicine (Dr Laya), and Family Medicine (Dr Taplin), University of Washington, Seattle.
Context Initiation of hormone replacement therapy (HRT) has been shown to increase
breast density. Evidence exists that increased breast density decreases mammographic
sensitivity. The effects on breast density of discontinuing and continuing
HRT have not been studied systematically.
Objective To examine the effects of initiation, discontinuation, and continued
use of HRT on breast density in postmenopausal women.
Design, Setting, and Participants Observational cohort study of 5212 naturally postmenopausal women aged
40 to 96 years and enrolled in a large health maintenance organization in
western Washington State who had 2 screening mammograms between 1996 and 1998.
Main Outcome Measures Breast density, assessed using the clinical radiologists' BI-RADS 4-point
scale, compared among women who did not use HRT before either mammogram (nonusers);
who used HRT before the first but not before the second mammogram (discontinuers);
who used HRT before the second but not before the first mammogram (initiators);
and who used HRT prior to both mammograms (continuing users).
Results Relative to nonusers, women who initiated HRT were more likely to show
increases in breast density (relative risk [RR], 2.57; 95% confidence interval
[CI], 2.12-3.08), while women who discontinued HRT use were more likely to
show decreases in density (RR, 1.81; 95% CI, 1.06-2.98) and women who continued
to use HRT were more likely to show both increases in density (RR, 1.33; 95%
CI, 1.13-1.55) and sustained high density (RR, 1.45; 95% CI, 1.33-1.58).
Conclusions These results indicate that breast density changes associated with HRT
are dynamic, increasing with initiation, and decreasing with discontinuation.
Several studies have shown that initiation of hormone replacement therapy
(HRT) increases parenchymal breast density,1- 5
and there is growing evidence that opposed estrogen has a stronger effect
on breast density than unopposed estrogen.4,5
Increases in density induced by HRT can have important consequences. Increased
density reduces the accuracy of screening mammography.6
Hormone replacement therapy has been directly associated with decreases in
both sensitivity and specificity of mammography,7- 10
which is likely a result of corresponding increases in density. Studies have
also associated increased density with increased risk of breast cancer.11- 14
Although the effect of initiating HRT on breast density has been well
studied, the effects of discontinuing HRT and continuing HRT have not been
systematically examined. In this study, we investigated the relationship between
HRT use and density in a population-based cohort of women undergoing at least
2 mammogram screenings.
Subjects were selected from women enrolled in Group Health Cooperative
(GHC) of Puget Sound, a health maintenance organization with more than 400 000
members in western Washington State. Most mammographic screening at GHC is
delivered through a breast cancer screening program (BCSP), which was established
in 1985.15 The BCSP collects demographic data,
health and screening history, and risk factor information through a self-administered
survey mailed to women aged 40 years or older, and generates letters that
invite women to begin breast cancer screening and periodically remind them
to return for regular screening. During the study period, women were sent
screening reminders every 1 to 2 years, with the reminder interval based on
their breast cancer risk factors. The GHC physicians may also order mammography
screening as part of well care.
Women were eligible for our study if they were postmenopausal and had
at least 2 screening examinations occurring between January 1996 and December
1998, with the second screening examination occurring at least 11 months,
but no more than 25 months, after the first. When women had more than 2 screening
examinations during the study period, we chose the pair whose timing was closest
to 2 years apart. Screening consisted of a 2-view mammogram and clinical breast
examination at dedicated centers within the GHC delivery system.
Women were excluded from our study if they were younger than 40 years,
had a hysterectomy, had a self-reported history of breast cancer, had a diagnosis
of cancer prior to either screening mammogram, or had undergone breast augmentation.
Because we relied on pharmacy data to estimate HRT, we restricted our sample
to women who were continuously enrolled in GHC during the year prior to each
Use of HRT was based on automated pharmacy records that capture all
prescriptions filled at GHC pharmacies. We defined HRT to include estrogens
alone and estrogens in combination with a progestin, delivered orally or by
patch. We excluded women who used vaginal rings from our study because their
use is not associated with higher blood levels of estrogen.16
Because estrogen creams were almost exclusively prescribed for use on an as-needed
basis, we did not consider women using creams to be HRT users. We combined
pharmacy dose and text instructions to estimate the duration of each prescription
and the average dose per day of estrogen and progestin. We estimated the timing
of HRT use by assuming that a woman began taking HRT the day after she filled
her prescription, with refills extending the duration of HRT use. When a woman
filled a prescription for a different HRT drug, or the same drug at a different
dosage within 10 days of an earlier fill, we assumed that her physician had
changed either the dose or formulation, effective the day after the new prescription
We classified HRT use (yes/no) prior to each mammogram using the date
of prescription fills and the estimated duration of the prescription. Women
classified as HRT users at the time of screening filled a prescription for
estrogen that lasted for at least 30 days and was estimated to run out no
more than 6 weeks before the screening mammogram. Women classified as nonusers
at the time of screening had not filled a prescription for estrogen in the
prior year, or had filled a prescription that was estimated to run out more
than 24 weeks before the screening mammogram.
We compared the following 4 patterns of HRT use: nonusers were those
who did not use HRT before either mammogram; discontinuers, those who used
HRT before the first mammogram, but not before their second mammogram; initiators,
those who did not use HRT before their first mammogram, but began using HRT
before their second mammogram; and continuing users, those who used HRT prior
to both mammograms. We confirmed analyses using a subset of women whose self-reported
current HRT use agreed with pharmacy records. However, we did not use self-reported
HRT use for primary analyses because as-needed creams were not distinguishable
from daily preparations; there was no information about duration or recency
of use; and self-reported data are subject to reporting errors and missing
Breast density was coded on a 4-point scale at the time of each mammogram
using American College of Radiology Breast Imaging Reporting and Data System
(BI-RADS) coding.17 A score of 1 indicated
almost entirely fat; 2, scattered fibroglandular tissue; 3, heterogeneously
dense; and 4, extremely dense. Density was coded by clinical radiologists
and was captured using an automated reporting system. Radiologists rated density
separately for each breast, and the breast with the highest density was used
for analysis. To focus on clinically important changes in density, we dichotomized
density ratings into low (almost entirely fat and scattered fibroglandular
tissue) and high (heterogeneously and extremely dense). Change in breast density
was coded into 4 groups: low density (1,2) at both evaluations, decrease in
density (3,4 to 1,2), increase in density (1,2 to 3,4), and high density (3,4)
at both evaluations.
We examined the association between HRT and density adjusting for 2
covariates associated with changes in breast density: age at first study mammogram
and change in body mass index (BMI), which is calculated as weight in kilograms
divided by the square of height in meters. Change in BMI was based on a 5-category
measure that captured clinically important changes in BMI: (1) lean at both
examinations (BMI <20); (2) initial BMI between 20 and 25 and a change
of less than 1 BMI unit; (3) initial BMI between 20 and 25 and a decrease
of at least 1 BMI unit; (4) initial BMI between 20 and 25 and an increase
of at least 1 BMI unit; and (5) heavy at both examinations (BMI ≥25). For
a woman who is 167.6 cm (5'6"), a 1-unit change in BMI roughly corresponds
to a 2.7-kg (6-lb) weight change. We examined the relationships between HRT
and density while controlling for age and change in BMI using 3 separate logistic
regression models to describe the probability of (1) increased density relative
to all other changes, (2) decreased density relative to all other changes,
or (3) high density at both examinations relative to all other changes. We
added interaction effects to logistic regression models to test for differential
effects of HRT change by age and BMI. Age was categorized into 3 groups of
approximately equal size (40-49, 50-69, ≥70 years). Body mass index was
grouped into low (<25 kg/m2) vs high (≥25 kg/m2).
Adjusted relative risks (RRs) were approximated using a transformation of
adjusted odds ratios (ORs): RR = OR divided by 1 + IR (OR−1),
where IR is the rate of the outcome in the reference group.18 The RRs and 95% confidence intervals (CIs) were estimated
by transforming confidence limits for corresponding ORs.
Among the 6313 women who met initial criteria for inclusion in our sample,
497 (7.9%) had HRT use patterns that did not correspond to 1 of our 4 groups.
Among the remaining women, 604 (10.4%) were excluded from analyses because
of missing data on breast density (1.1%) or BMI (9.4%). Our final sample included
5212 women with complete density and covariate information.
At the time of the initial mammogram, the mean age of women in our sample
was 64.5 (SD, 9.5; range, 40-96) years. There were similar age ranges across
the women with the 4 HRT use patterns: nonusers ranged from 43 to 96 years;
discontinuers, 42 to 80 years; initiators, 42 to 91 years; and continuing
users: 40 to 92 years. As shown in Table
1, women who were using HRT at the time of the first mammogram (discontinuers
and continuing users) tended to be younger than women in other groups. Nonusers
tended to be older than other groups. Across all groups, approximately one
third of women were aged 60 to 69 years. However, 44.4% of nonusers were 70
years or older, while 25.7% of initiators, 18.0% of discontinuers, and 16.3%
of continuing users were 70 years or older.
There were few differences in other risk factors across patterns of
HRT use. Our sample was predominately white (92.0%), reflecting the overall
racial composition of women enrolled in GHC and the surrounding population;
96.1% had a prior mammogram available to the radiologist at the time of their
initial screening mammogram; 15.9% were nulliparous; and 13.8% had their first
child after age 30 years. First-degree family history of breast cancer (mother,
sister, or daughter) differed across groups with 24.7% of nonusers, 20.2%
of initiators, 17.1% of discontinuers, and 16.6% of continuing users reporting
a family history.
Differences in average BMI across patterns of HRT use were small, though
discontinuers and continuing users tended to be somewhat leaner than other
groups (Table 1). About half of
the women (51.7%) in this sample had a BMI that was higher than 25 kg/m2 at both examinations. Approximately equal numbers gained or lost 1
or more units on the BMI scale (7.6% and 7.2%, respectively). The BMI for
about one third (29.6%) changed less than 1 BMI unit, and only a few (3.8%)
had a BMI that was less than 20 kg/m2at both examinations.
The HRT dose and drug type were similar across the 3 groups of women
who used HRT (initiators, discontinuers, and continuing users). Most women
who received HRT (92.4%) received a combination of an estrogen and a progestin.
The most common average daily dose of conjugated estrogen was 0.625 mg/d (examination
1, 62.1%; examination 2, 50.0%). Few women received estrogen doses that were
greater than 0.625 mg/d (examination 1, 5.1%; examination 2, 5.3%). Most women
were prescribed Estratab (Solvay, Brussels, Belgium; examination 1, 86.5%;
examination 2, 86.8%). Among women receiving combination therapy, 99.4% were
prescribed medroxyprogesterone acetate, with the remainder prescribed norethindrone
Most women (80.6%) had index screening examinations during 1996, and
most (64.7%) had their second mammogram between 21 and 25 months after their
first. There were no differences in timing of the index mammogram or time
between mammograms across patterns of HRT use. Almost all of the women in
our sample (99.0%) had equal density ratings in both breasts. As shown in Table 1, women using HRT at the time of
the first mammogram (discontinuers and continuing users) tended to have higher
density at the initial screening than women who were not using HRT at the
time of their first mammogram (initiators and nonusers).
Table 2 shows the overall
relationship between HRT use and breast density. Relative to the nonuser group,
initiators were more likely to have an increase in density, discontinuers
were more likely to have a decrease in density, and continuing users were
more likely to have high density at both examinations.
Table 3 shows estimated
associations between HRT use and breast density that were adjusted for age
at initial examination and change in BMI via logistic regression. Because
few women had BMI below 20 kg/m2 at both examinations, these women
were grouped with women who had a less than 1-unit change in BMI between evaluations.
Relative to nonusers, initiators were more likely to have increased density
and less likely to have a decrease in density; discontinuers were more likely
to have a decrease in density; and continuing users were more likely to have
an increase in density and more likely to have high density at both examinations.
There was a significant interaction between age and the effect of HRT
initiation with greater risk of increased density associated with initiation
among older initiators (RR not significantly different for <60 years vs
60-69 years [P = .05] and RR significantly greater
for ≥70 years vs <60 years [P = .001]). These
differences in RR across age groups reflect observed age differences in the
proportion of initiators with increases in density (<60 years, 17.9%; 60-69
years, 29.6%; ≥70 years, 43.0%). There was also a significant interaction
between baseline BMI and the risk of sustained high density among continuing
users relative to nonusers (P = .002). However, this
reflects observed differences in proportion of nonusers with sustained high
density (39.0% with BMI <25 kg/m2; 14.2% with a BMI of ≥25
kg/m2, respectively) since relatively more continuing users with
low BMI had sustained high density than continuing users with high BMI (56.9%
with BMI <25 kg/m2; 30.5% with a BMI of ≥25 kg/m2,
respectively). We found similar differentially increased risk for sustained
high density among initiators relative to nonusers, with greater RR among
women with high BMI (P = .04). These differential
risks were also driven by differences between the 2 nonuser groups.
This study has several strengths that distinguish it from earlier research.
Automated pharmacy data allowed us to measure HRT use across a large group
of mammographically screened women whose breast density was routinely recorded.
We believe that this is the largest study to date of HRT use and breast density
changes. This is the only study to simultaneously examine HRT initiation,
discontinuation, and continuing use relative to women not using HRT. We focused
on clinically significant changes in breast density, and distinguished women
with fatty breasts at both examinations from women with dense breasts at both
examinations. This is also the first published study to explore changes in
density adjusting for co-occurring changes in BMI.
Our analyses provide important new information about women who discontinue
HRT use and women who are continuing HRT users. Discontinuation of HRT was
associated with subsequent decreases in density, and increases in breast density
were sustained by continued HRT use. We also found that initiation of HRT
was associated with increases in parenchymal breast density. These results
provide strong evidence that breast density changes associated with HRT are
dynamic, increasing with initiation, and decreasing with discontinuation.
Our analyses confirmed findings from previous studies demonstrating
an association between initiation of HRT use and increases in parenchymal
breast density.1- 5
Like these earlier studies, we examined changes in breast density among women
who began using HRT. Studies that have failed to find an association compared
the parenchymal patterns in women using HRT with patterns in women not using
HRT, rather than examining within-woman changes as we have.19,20
We measured density using American College of Radiology BI-RADS coding,
with assessments made by a variety of clinical radiologists. Studies finding
increases in density associated with initiation of HRT have used a variety
of measures, though all relied on expert readers. Greendale et al4 used BI-RADS coding. Several earlier studies were
based on the Wolfe classification scheme2,3,5
and 1 study1 used a simple measure of dense,
heterogeneous, and fatty. These studies consistently demonstrated an association
between initiation of HRT and increased breast density across a variety of
density measures. Our study is consistent with previous research, showing
clear associations between initiation of HRT and increased breast density
as measured in clinical practice.
Our findings suggest that the probability of experiencing an increase
in density following HRT initiation, or sustained high density following continued
HRT use, may be affected by age. This finding is important because the risk
of breast cancer increases with age. Because increased density is associated
with decreased mammographic accuracy, HRT use could have particularly deleterious
effects among older, higher-risk women.
We were able to examine changes in HRT use and breast density changes
in a relatively large group of women because of automated data collection.
Previous studies of initiators have also used detailed HRT-dose information,
though the source of these data is sometimes unclear. Two studies appear to
rely on a combination of medical records and self-report for HRT-dose information,1,2,5 while others used information
from randomization within a treatment study.3,4
Automated pharmacy data allow broad capture of information, but these data
have limitations. Women who filled prescriptions but do not subsequently take
HRT may be misclassified as users. Women who filled HRT prescriptions at outside
pharmacy facilities may have been misclassified as nonusers. Timing information
is especially sensitive since we cannot determine when a woman actually begins
to take her prescription. To overcome these limitations, we used relatively
stringent requirements for categorization into use and nonuse groups, resulting
in the exclusion of women who had intermediate use patterns. We also examined
a subset of women whose self-reported current HRT use was consistent with
pharmacy-estimated HRT use and found virtually identical results. Overall,
only 5.1% of women's self-reported HRT use differed from pharmacy records.
Thus, there is little evidence of misclassification of HRT use and such misclassification
would result in attenuation of the effect of HRT use on density change.
Like all previous studies of HRT use, the current observational study
is subject to bias. Women chose whether and when to initiate, continue, or
discontinue HRT use, and these choices may be related to unmeasured factors
that affect study findings. Randomized trials that include a placebo control4 come closest to avoiding this potential bias since
all women, including the placebo group, were willing to initiate HRT. Although
a recent study of HRT users suggests that there are fewer differences between
users and nonusers than previously expected,21
our results must be viewed in light of their observational nature.
Finally, these data did not allow us to address several important factors
that may influence the effect of HRT on breast density. While this is perhaps
the largest study of HRT and density conducted to date, our sample sizes were
moderate, particularly for the groups of women who initiated or discontinued
use during the study period. Although sample sizes were sufficiently large
to show main effects of HRT use on breast density, we may not have had power
to find some interactions between age, BMI, and HRT use. We were unable to
examine the effects of opposed vs unopposed estrogen and the effects of type
of drug prescribed because there was not enough variability in these factors
within our sample, reflecting the selection of naturally postmenopausal (ie,
nonhysterectomized) women. In addition, automated pharmacy data did not allow
us to distinguish between cyclical and combination estrogen and progestin.
We also lacked information about women's overall duration of use.
This study shows strong associations between patterns of HRT use and
changes in breast density. Our findings suggest that in some women, HRT increases
breast density but these increases are potentially reversible with cessation
of HRT. This result has important implications for breast cancer screening.
Increased density adversely affects the accuracy of screening mammography
and is a strong, if not the strongest, risk factor for cancer missed at screening.6 Hormone replacement therapy is associated with decreases
in both the sensitivity and specificity of mammography.7- 10
Observed decreases in mammographic accuracy among women using HRT are a likely
result of corresponding increases in density.