Cummings SR, Duong T, Kenyon E, Cauley JA, Whitehead M, Krueger KA, for the Multiple Outcomes of Raloxifene Evaluation (MORE) Trial . Serum Estradiol Level and Risk of Breast Cancer During Treatment With
Raloxifene. JAMA. 2002;287(2):216–220. doi:10.1001/jama.287.2.216
Author Affiliations: Coordinating Center, University of California, San Francisco (Drs Cummings and Kenyon and Ms Duong); Department of Epidemiology, University of Pittsburgh, Pittsburgh, Pa (Dr Cauley); Kings College Hospital, London, England (Dr Whitehead); and Eli Lilly and Co, Indianapolis, Ind (Dr Krueger).
Context As endogenous estradiol increases, risk of breast cancer increases.
Raloxifene competes with endogenous estrogen for binding to estrogen receptors
in breast tissue. A woman's estradiol level may alter the effects of raloxifene
on breast cancer and other outcomes.
Objective To test the hypothesis that raloxifene reduces breast cancer risk more
in women with relatively high estradiol levels than in women with very low
Design Analysis of the Multiple Outcomes of Raloxifene Evaluation, a randomized,
double-blind, placebo-controlled trial conducted from 1994 to 1999.
Setting One hundred eighty community settings and medical practices in 25 countries
including the United States.
Participants A total of 7290 postmenopausal women aged 80 years or younger with osteoporosis
who had baseline serum estradiol concentrations measured by a central laboratory
using a sensitive assay. Women with a history of breast cancer or estrogen
use were excluded.
Intervention Participants were randomly assigned to receive 60 mg/d or 120 mg/d of
raloxifene (n = 4843) or matching placebo (n = 2447) for 4 years.
Main Outcome Measure New cases of histopathologically confirmed breast cancer in the treatment
and placebo groups, stratified by estradiol levels.
Results In the placebo group, women with estradiol levels greater than 10 pmol/L
(2.7 pg/mL) had a 6.8-fold higher rate of breast cancer (3.0% per 4 years;
95% confidence interval [CI], 1.8%-4.1%) than that of women with undetectable
estradiol levels (0.6% per 4 years; 95% CI, 0%-1.1%; P
= .005 for trend). Women with estradiol levels greater than 10 pmol/L in the
raloxifene group had a rate of breast cancer that was 76% (95% CI, 53%-88%)
lower than that of women with estradiol levels greater than 10 pmol/L in the
placebo group (absolute rate reduction, 2.2% [95% CI, 1.0%-3.5%; number needed
to treat = 45]). In contrast, women with undetectable estradiol levels had
similar breast cancer risk whether or not they were treated with raloxifene
(risk difference, −0.1%; 95% CI, −0.8% to 0.6%; P = .02 for the interaction). In this cohort, treating women with estradiol
levels greater than 10 pmol/L with raloxifene for 4 years would have avoided
47% of breast cancer cases.
Conclusions Measurement of estradiol level by sensitive assay in postmenopausal
women identifies those at high risk of breast cancer who may benefit most
from raloxifene. If confirmed, this suggests that measuring estradiol and
treating women with high estradiol levels could substantially reduce the rate
of breast cancer among postmenopausal women.
Prospective studies have found that the risk of breast cancer rises
with increases in endogenous estradiol levels.1- 4
The selective estrogen receptor modulators tamoxifen and raloxifene block
the effects of endogenous estrogen in the breast5,6
and reduce the risk of breast cancer.7- 9
Therefore, the effects of selective estrogen receptor modulators might depend
on endogenous levels of estrogen. In a previous analysis of raloxifene effects
in the Multiple Outcomes of Raloxifene Evaluation (MORE) trial, Lippman and
colleagues10 observed that characteristics
that might reflect high lifetime exposure to estrogen, such as high bone mineral
density, were associated with an increased risk of breast cancer and somewhat
greater reduction in cancer risk with raloxifene. They also reported that
raloxifene reduced breast cancer significantly in women whether their estradiol
levels were above or below the upper tertile, but the study did not analyze
the full range of estradiol levels.
We postulated that the risk of breast cancer and the effect of raloxifene
on risk of breast cancer would depend on serum concentrations of estradiol.
Specifically, we hypothesized that women with undetectable serum concentrations
of estradiol would have a very low breast cancer risk that would not be further
reduced by raloxifene and that women with high estradiol levels would have
a high risk of breast cancer that would be substantially reduced by raloxifene.
Raloxifene also reduces the risk of vertebral fractures and increases the
risk of venous thromboembolism (VTE),9,11,12
and we postulated that these effects might also depend on estradiol level.
We tested these hypotheses using data from the MORE trial.9- 11
The MORE trial, a multicenter, randomized, double-blind, placebo-controlled
trial designed to test the primary hypothesis that raloxifene would reduce
the risk of vertebral fracture in postmenopausal women with osteoporosis,9,12 monitored participants for the occurrence
of breast cancer. We enrolled 7705 postmenopausal women with osteoporosis
who were aged 80 years or younger.9 Women with
a history of breast cancer were excluded. Eligible women were randomly assigned
to receive 60 mg/d or 120 mg/d of raloxifene or placebo.
All participants had mammography at baseline and annually after the
first year.9 Participants who declined a mammogram
could undergo breast ultrasonography instead. Diagnosis of breast cancer was
confirmed by reports of histopathology and estrogen receptor status was determined
from medical records. Investigators performing these assessments were blinded
to treatment assignment.
To confirm that participants were postmenopausal, serum was obtained
at baseline after at least a 6-hour fast and shipped the same day to the laboratory
where estradiol was measured. We limited our analysis to the 7290 participants
who had estradiol levels assayed by SciCor (Covance) Central Laboratory Services
(Indianapolis, Ind). Estradiol concentration was determined using a double-antibody
procedure. The serum sample was preincubated with antiestradiol antiserum.
Sodium iodide I 125–labeled estradiol, which competes with estradiol
for binding sites, was then incubated with the sample for a fixed time. Bound
and free estradiol was separated by the polyethylene glycol–accelerated
double-antibody method, the antibody-bound fraction was precipitated and counted,
and the concentration of estradiol in the sample was read from a calibration
curve. The intra-assay coefficient of variation is 6.5% (SD, 2.1 pmol/L [0.6
pg/mL]) at an estradiol concentration of 33 pmol/L (9.0 pg/mL).
Serum estradiol concentrations were initially reported by the laboratory
as less than 5 pmol/L (1.4 pg/mL) or exact value at least 5 pmol/L. To analyze
levels less than 5 pmol/L, we obtained the exact values from Covance Laboratories
for the 24 cases with breast cancer whose values had been reported as less
than 5 pmol/L.
From women who did not develop breast cancer, we randomly selected 222
women as a comparison group (3 times the number of cases). Of these, 112 women
had estradiol levels of less than 5 pmol/L. We obtained exact estradiol levels
for these women also.
The primary outcome of the analyses was rate of breast cancer. Both
dosages of raloxifene produced similar reductions in risk of breast cancer,
so the 2 treatment groups were pooled.9,11
All comparisons were by intention-to-treat analyses.
The prospectively defined analysis grouped estradiol levels as 0 pmol/L
(undetectable in comparison to the curve limit for the assay), more than 0
to less than 5 pmol/L, 5 to 10 pmol/L (2.7 pg/mL), and more than 10 pmol/L.
The risk difference was defined as the difference between breast cancer rates
in the placebo and raloxifene groups within each of the 4 groups stratified
by estradiol level. Its reciprocal is the estimated number needed to treat
for 4 years to prevent 1 case of breast cancer.
To estimate the absolute rates of breast cancer in all women in the
trial who had very low estradiol levels, we assumed that the distributions
of undetectable and low estradiol (>0 to <5 pmol/L) in the breast cancer
cases and 222 participants without breast cancer were similar to the entire
cohort. The 95% confidence intervals (CIs) for the estimated rates were calculated
using the delta method, which accounts for uncertainty in the number of women
in each group, as estimated from the sample of women with and without breast
cancer.13 We used logistic regression models
to test for an interaction between estradiol category and the effect of raloxifene
on risk of breast cancer.
Vertebral fractures, VTEs, and hot flashes were assessed as previously
described.9,11 We compared the
rates in women assigned to placebo and raloxifene stratified by serum estradiol
concentrations that were less than 5 pmol/L, 5 to 10 pmol/L, and more than
10 pmol/L (exact values <5 pmol/L were not available for these women).
A total of 7290 (95%) of the 7705 women enrolled in MORE had baseline
measurement of estradiol performed by the central laboratory, of whom 2447
were assigned to placebo and 4843 to raloxifene, 60 mg/d or 120 mg/d. Mean
age was 66.5 years, and 97% were white (Table 1). All 4 years of follow-up were completed by 1760 (72%)
who were assigned to placebo and 3532 (73%) who were assigned to raloxifene.
Of women who completed the trial, 5025 (95%) in both groups took at least
80% of the study medication.
During 4 years, 74 cases of breast cancer were confirmed, of which 59
were invasive; 44 were estrogen receptor–positive, and 13 were estrogen
receptor–negative; receptor status was not available for 17. Among the
women in this analysis, women randomized to raloxifene had a 70% (odds ratio,
0.3; 95% CI, 0.2-0.5) lower risk of invasive breast cancer and an 80% (odds
ratio, 0.2; 95% CI, 0.1-0.4) lower risk of estrogen receptor–positive
cancer compared with women randomized to placebo. These reductions are similar
to the 72% reduction in invasive cancer and 84% reduction in estrogen receptor–positive
cancer after 4 years in the entire trial.11
A total of 24 women with breast cancer had an estradiol level of less
than 5 pmol/L. Of these, 15 had undetectable estradiol levels and 9 had estradiol
levels between 0 and 5 pmol/L. Of the 222 women without breast cancer selected
for comparison, 112 had levels less than 5 pmol/L. Of these, 75 had levels
that were undetectable and 37 had levels that were more than 0 but less than
5 pmol/L. Age, weight, and family history of breast cancer were similar in
women in the sample of the cohort for whom exact estradiol values were available
and in the remainder of participants (P>.30).
The risk of breast cancer in the placebo group increased from 0.6% per
4 years (95% CI, 0%-1.1%) in women with undetectable estradiol levels to 3.0%
(95% CI, 1.8%-4.1%) in those with levels of more than 10 pmol/L (P = .005 for trend; Table 2
and Figure 1). The risk of breast
cancer in women with estradiol levels of more than 10 pmol/L was 6.8 times
higher (95% CI, 2.2-21.0) than in women with undetectable estradiol. A similar
association was observed between estradiol level and both invasive breast
cancer and estrogen receptor–positive cancer (P
= .003 for both).
The raloxifene group had low rates of breast cancer at all levels of
estradiol (Table 2 and Figure 1). Women with the highest serum concentrations
of estradiol had the greatest reductions in absolute risk of breast cancer
compared with women with similar levels in the placebo group. Among those
with estradiol levels of more than 10 pmol/L who were randomized to raloxifene,
breast cancer risk was reduced by 76% (95% CI, 53%-88%) but had no significant
effect in those with undetectable levels (P = .02
for interaction between estradiol level and breast cancer reduction with raloxifene).
This relationship remained in analyses of invasive cancer (79%; 95% CI, 55%-90%)
and estrogen receptor–positive cancer (81%; 95% CI, 59%-91%) and for
the 60-mg/d and 120-mg/d dosages of raloxifene. The associations between estradiol
level and risk of breast cancer as well as the interactions between estradiol
level and reduction in breast cancer risk with treatment remained significant
(P≤.03) after adjustment for weight or body mass
index, femoral neck bone mineral density, and presence or absence of vertebral
The number needed to treat for 4 years to prevent 1 case of breast cancer
varied from 45 among women with estradiol concentrations of more than 10 pmol/L
to 100 and 125, respectively, for women with estradiol levels of 5 to 10 pmol/L
and more than 0 to less than 5 pmol/L. The numbers needed to treat for invasive
cancer were 48, 91, and 111, respectively.
Among the women in this analysis, raloxifene decreased risk of vertebral
fracture by 40%. The reduction in risk was similar for all levels of estradiol
(Table 3). The increased risk
of VTEs with raloxifene (2.1; 95% CI, 1.4-3.3) and increased occurrence of
hot flashes were also similar in estradiol groups (Table 3).
As in most previous studies,1- 4
estradiol levels were strongly associated with risk of breast cancer in postmenopausal
women in the placebo group; the 36% of women with estradiol levels of more
than 10 pmol/L had a 3% 4-year risk of breast cancer. Among women with estradiol
levels of 10 pmol/L or less, raloxifene reduced the risk of breast cancer
by 76%, and 45 women needed to be treated for 4 years to avoid 1 case of breast
cancer. Additionally, 62% of breast cancer cases occurred in this group, so
treating only those with the highest estradiol would have reduced the overall
rate of breast cancer by about 47%. In contrast, raloxifene did not decrease
the already low risk of breast cancer in women with undetectable estradiol
levels. This is consistent with the belief that raloxifene reduces breast
cancer risk by blocking the effect of estradiol on the breast.5,6
Since tamoxifen may act by a similar mechanism, the interaction between endogenous
estradiol and the effect of tamoxifen on breast cancer risk deserves study.
Treatment with raloxifene reduced the risk of vertebral fractures and
increased the risk of VTE and hot flashes to a similar degree, regardless
of estradiol level. Although women with undetectable levels of estradiol did
not appear to benefit from a reduction in risk of breast cancer with raloxifene,
women with undetectable estradiol levels have an increased risk of vertebral
fractures,14 so they might derive greater benefit
from raloxifene's reduction in vertebral fracture risk.
This study has limitations. Women with osteoporosis tend to have a lower
risk of breast cancer and lower levels of estradiol than other women.15,16 Therefore, the proportion of women
with undetectable estradiol levels may be higher and the proportion with levels
of more than 10 pmol/L is likely to be lower in this cohort than in the general
population. The effect of raloxifene on risk of breast cancer was a secondary
aim of MORE, and raloxifene is not approved by the US Food and Drug Administration
for labeling for prevention of breast cancer.17
Our results are based on 4 years of treatment with raloxifene. The effects
of long-term treatment with raloxifene are being studied in a continuation
of the MORE trial. Finally, we sampled only a portion of women without breast
cancer to obtain exact estradiol levels. However, this random sample's characteristics
did not differ significantly from the entire cohort from which it was drawn.
Lippman et al10 previously reported that
women enrolled in the MORE trial who had estradiol levels of at least 12 pmol/L
and those with levels less than 12 pmol/L had similar and significant reductions
in risk of breast cancer (79% and 64%, respectively) during treatment with
raloxifene. However, that study did not analyze the full range of estradiol
values, did not have exact values of estradiol below 5 pmol/L, and included
women with estradiol measurements by local laboratories.
Risk factors for breast cancer other than estradiol might be useful
predictors of response to raloxifene. Only some of the risk factors used in
the model of Gail et al18 were assessed in
the MORE trial. Lippman et al10 did not find
a significant association between body mass index and risk of invasive breast
cancer but reported greater reductions in risk of breast cancer during raloxifene
in women with the highest bone mineral density and without vertebral fractures.
We found that estradiol level predicted both risk of breast cancer and degree
and reduction in risk of breast cancer even after adjusting for body mass
index, bone mineral density, and baseline vertebral fracture. The value of
estradiol level in addition to potential risk factors for breast cancer warrants
further study in larger trials.
Estradiol levels were measured only once, at baseline, and estradiol
levels may vary. However, Hankinson and colleagues19
found that serum concentrations of estradiol in postmenopausal women remain
relatively stable (r = 0.7) over 2 or 3 years. Estradiol
levels below 5 pmol/L have been considered unreliable because the coefficient
of variation (SD divided by mean value) is large when the mean value is low
(<5 pmol/L). Our results suggest that differentiating undetectable values
from extremely low values may have clinical utility despite this variability.
Variability tends to dilute associations, so the true relationship between
endogenous estradiol level and the effects of raloxifene may be stronger than
we observed in this study.
Sensitive estradiol assays of the type used in this study are not yet
widely available for clinical use. If sensitive measurements of estradiol
are used to make treatment decisions regarding breast cancer risk, it will
be important to standardize the methods and reporting of results.
We conclude that in the MORE trial, measurement of estradiol concentration
in older postmenopausal women identified those at high risk of breast cancer
who benefited most from reduction in risk of breast cancer with raloxifene
treatment. If confirmed by other studies, measuring estradiol to determine
breast cancer risk may help identify women likely to experience the greatest
reduction in breast cancer risk from treatment with raloxifene.