The combined influence of multiple sex and growth hormones on risk of postmenopausal breast cancer: a nested case-control study.

Abstract

Sex and growth hormones are positively associated with postmenopausal breast cancer risk. However, few studies have evaluated the influence of multiple hormones simultaneously.
We considered the roles of estrone, estradiol, estrone sulfate, testosterone, androstenedione, dehydroepiandrosterone (DHEA), DHEA sulfate and prolactin and, secondarily, insulin-like growth factor 1 (IGF-1) and c-peptide in postmenopausal breast cancer risk among 265 cases and 541 controls in the prospective Nurses' Health Study. We created several hormone scores, including ranking women by the number of hormones above the age- and batch-adjusted geometric mean and weighting hormone values by their individual associations with breast cancer risk.
Women in the top versus bottom quintile of individual estrogen or androgen levels had approximately a doubling of postmenopausal breast cancer risk. Having seven or eight compared to zero hormones above the geometric mean level was associated with total (RR = 2.7, 95% CI = 1.3 to 5.7, P trend < 0.001) and estrogen receptor (ER)-positive (RR = 3.4, 95% CI = 1.3 to 9.4, P trend < 0.001) breast cancer risk. When comparing the top versus bottom quintiles of the score weighted by individual hormone associations, the RR for total breast cancer was 3.0 (95% CI = 1.8 to 5.0, P trend < 0.001) and the RR for ER-positive disease was 3.9 (95% CI = 2.0 to 7.5, P trend < 0.001). The risk further increased when IGF-1 and c-peptide were included in the scores. The results did not change with adjustment for body mass index.
Overall, the results of our study suggest that multiple hormones with high circulating levels substantially increase the risk of breast cancer, particularly ER-positive disease. Additional research should consider the potential impact of developing risk prediction scores that incorporate multiple hormones.

Full text

Page 1

RESEARCH ARTICLE Open Access
The combined influence of multiple sex and
growth hormones on risk of postmenopausal
breast cancer: a nested case-control study
Shelley S Tworoger1,2*, Bernard A Rosner1,3, Walter C Willett1,2,4and Susan E Hankinson1,2
Abstract
Introduction: Sex and growth hormones are positively associated with postmenopausal breast cancer risk.
However, few studies have evaluated the influence of multiple hormones simultaneously.
Methods: We considered the roles of estrone, estradiol, estrone sulfate, testosterone, androstenedione,
dehydroepiandrosterone (DHEA), DHEA sulfate and prolactin and, secondarily, insulin-like growth factor 1 (IGF-1)
and c-peptide in postmenopausal breast cancer risk among 265 cases and 541 controls in the prospective Nurses’
Health Study. We created several hormone scores, including ranking women by the number of hormones above
the age- and batch-adjusted geometric mean and weighting hormone values by their individual associations with
breast cancer risk.
Results: Women in the top versus bottom quintile of individual estrogen or androgen levels had approximately a
doubling of postmenopausal breast cancer risk. Having seven or eight compared to zero hormones above the
geometric mean level was associated with total (RR = 2.7, 95% CI = 1.3 to 5.7, P trend < 0.001) and estrogen
receptor (ER)-positive (RR = 3.4, 95% CI = 1.3 to 9.4, P trend < 0.001) breast cancer risk. When comparing the top
versus bottom quintiles of the score weighted by individual hormone associations, the RR for total breast cancer
was 3.0 (95% CI = 1.8 to 5.0, P trend < 0.001) and the RR for ER-positive disease was 3.9 (95% CI = 2.0 to 7.5, P
trend < 0.001). The risk further increased when IGF-1 and c-peptide were included in the scores. The results did
not change with adjustment for body mass index.
Conclusions: Overall, the results of our study suggest that multiple hormones with high circulating levels
substantially increase the risk of breast cancer, particularly ER-positive disease. Additional research should consider
the potential impact of developing risk prediction scores that incorporate multiple hormones.
Introduction
Circulating levels of estrogens and androgens are positively
associated with the risk of breast cancer in postmenopausal
women [1-4]. Furthermore, growth hormones, including
prolactin and insulin-like growth factor 1 (IGF-1), are
modestly associated with risk [5,6]. However, relatively lit-
tle is known about the possible combined effect of multiple
hormones simultaneously with respect to risk [1,2].
Several studies have assessed the joint effects of sex
and growth hormones on postmenopausal breast cancer
risk [2,5,7,8], and, in general, no multiplicative interac-
tions were observed. However, Yu et al. [7] reported
that women with high levels of IGF-1 and at least one
sex hormone had the highest risk of breast cancer. For
example, compared to women with low IGF-1 and low
testosterone levels, those with high levels of both had a
RR = 2.2 (95% CI = 1.0 to 4.5) versus a RR = 1.7 (95%
CI = 0.8 to 3.6) for high testosterone only or a RR = 1.1
(95% CI = 0.5 to 2.4) for IGF-1 only. Only one small
study (29 cases and 58 controls) considered more than
two hormones at the same time [9]. Estrone, estradiol,
androstenedione,dehydroepiandrosterone
(DHEAS), testosterone and IGF-1 were measured on
prospectively collected serum samples. Women whose
hormone levels were below the age-adjusted mean for
sulfate
* Correspondence: nhsst@channing.harvard.edu
1Channing Laboratory, Department of Medicine, Brigham and Women’s
Hospital and Harvard Medical School, 181 Longwood Avenue, Boston, MA
02115, USA
Full list of author information is available at the end of the article
Tworoger et al. Breast Cancer Research 2011, 13:R99
http://breast-cancer-research.com/content/13/5/R99
© 2011 Tworoger et al.; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative
Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and
reproduction in any medium, provided the original work is properly cited.

Page 2

all hormones had a substantially lower risk of breast
cancer compared to women with one or more hormone
levels above the mean (RR = 0.11, 95% CI = 0.01 to
0.90). The authors suggested that mammotropic hor-
mones may act as permissive factors for breast cancer
growth.
To assess this hypothesis and to extend the concept of
developing hormone scores, we evaluated the role of
multiple hormones on postmenopausal breast cancer
risk using data from 265 patients and 541 controls in
the prospective Nurses’ Health Study (NHS) who had
measures of estrone, estradiol, estrone sulfate, testoster-
one, androstenedione, dehydroepiandrosterone (DHEA),
DHEAS and prolactin. We created a hormone score
similar to that described by Trichopoulos et al. [9], as
well as scores based on other approaches, including
weighting hormone values by their individual association
with breast cancer risk or by their proliferative effect on
breast cancer cell lines. On the basis of prior data indi-
cating that sex hormones tend to be more strongly asso-
ciated with estrogen receptor (ER)-positive disease
[2,5,6], we evaluated the scores for this subset of tumors.
Secondarily, we considered a subset of women who also
had IGF-1 and c-peptide measures.
Materials and methods
Study population
The NHS cohort was established in 1976 among 121,700
US female registered nurses ages 30 to 55 years. All
women completed an initial questionnaire and have been
followed biennially by questionnaire to update exposure
status and disease diagnoses. In 1989 and 1990, 32,826
NHS participants (ages 43 to 69 years) provided blood
samples and completed a short questionnaire [10]. Briefly,
these women arranged to have their blood drawn and
shipped with an ice pack via overnight courier to our
laboratory, where it was processed and separated into
plasma, red blood cell and white blood cell components.
These samples have been stored in liquid nitrogen freezers
since collection. This study was approved by the Commit-
tee on the Use of Human Subjects in Research at Brigham
and Women’s Hospital in Boston.
The women were diagnosed with breast cancer after
blood collection but before 1 June 1998. The follow-up
rate was 99% in 1998, and the maximum follow-up time
was 8.83 years. We confirmed 320 breast cancer patients
who were postmenopausal and not using postmenopausal
hormones (PMH) at the time of blood collection. Cases
were matched to two controls on age (± 2 years), month
of blood collection (± 1 month), time of day of blood
draw (± 2 hours) and fasting status. The Institutional
Review Board of Brigham and Women’s Hospital
approved this analysis, and all participants provided their
written informed consent.
Among the potentially eligible 320 cases and 663 con-
trols, 41 cases and 79 controls were missing data for one
hormone, 10 cases and 21 controls were missing data for
two hormones, and 4 cases and 22 controls were missing
data for three or more hormones. In total, 265 cases (147
ER-positive cases) and 541 controls had assay values for all
estrogens, androgens and prolactin. A subset (n = 182
cases and n = 370 controls) also had IGF-1 and c-peptide
data.
Laboratory assays
Hormone assay methods have been described previously
[10,11] and were conducted by four different laboratories.
Estrone, estradiol, androstenedione, testosterone, DHEA
and DHEAS were assayed at the Quest Diagnostics
Nichols Institute (San Juan Capistrano, CA, USA) using an
extraction step (except for DHEAS) followed by RIA. For
estrone sulfate, the first batch was assayed at the Univer-
sity of Massachusetts Medical Center’s Longcope Steroid
Radioimmunoassay Laboratory (Worcester, MA, USA).
The remaining batches were assayed at the Nichols Insti-
tute. To quantify estrone sulfate levels, estrone was first
extracted from the plasma and then the estrone sulfate
bond was enzymatically cleaved to release estrone, which
was extracted from the plasma by an organic solvent and
studied by chromatography and RIA [12]. The first two
batches of sex hormone-binding globulin (SHBG) and pro-
lactin were assayed at the Longcope Laboratory. The
remaining samples were assayed at Massachusetts General
Hospital’s Reproductive Endocrinology Unit Laboratory
(Boston, MA, USA). Prolactin was measured by micropar-
ticle enzyme immunoassay and SHBG was analyzed using
the AxSYM immunoassay system (Abbott Diagnostics,
Chicago, IL, USA). All assays for IGF-1 and c-peptide
were conducted at the laboratory of Dr Michael Pollak at
McGill University (Montreal, QC, Canada) using an
ELISA from Diagnostic Systems Laboratory (Webster, TX,
USA). Free estradiol and testosterone were calculated
according to the method described by Sodergard et al.
[13].
Cases and matched controls identified through 1 June
1998 were assayed for estrogens, androgens and prolac-
tin. Cases and matched controls identified through 1
June 1996 were assayed additionally for IGF-1 and c-
peptide. Case-control sets were assayed together,
ordered randomly and labeled to mask case-control sta-
tus. The coefficient of variation from blinded replicate
samples was less than 15% for all assays. When hor-
mone values were less than the detection limit, we set
the value to half this limit. The detection limits of the
assays and the number of samples below the limit were
2 pg/ml estradiol (n = 2), 10 pg/ml estrone (n = 22),
40 pg/ml estrone sulfate (n = 7), 5 ng/dl androstene-
dione (n = 3), 2 ng/dl testosterone (n = 4), 10 ng/dl
Tworoger et al. Breast Cancer Research 2011, 13:R99
http://breast-cancer-research.com/content/13/5/R99
Page 2 of 10

Page 3

DHEA (n = 1), 5 μg/dl DHEAS (n = 5), 0.6 ng/ml pro-
lactin (n = 0), 0.09 ng/ml IGF-1 (n = 0) and 0.02 ng/ml
c-peptide (n = 0).
Statistical analysis
We identified statistical outliers using the generalized
extreme studentized deviate many-outlier detection
approach [14]. This resulted in the removal of the
following values: four estradiol (≥76 pg/ml), three testos-
terone (≤1 or ≥459 ng/dl), three androstenedione
(≤5 ng/dl), two DHEA (≤14 ng/dl), five DHEAS (≤5 μg/
dl), six prolactin (≥43 ng/ml), two IGF-1 (≤24 ng/ml)
and one c-peptide (≤0.03 ng/ml). Furthermore, some
assays could not be conducted because of low sample
volume or technical difficulties with the assay.
We used unconditional logistic regression models to
estimate RR and 95% CI for total breast cancer, invasive
breast cancer and ER-positive disease, adjusting for
matching factors, including age at blood collection (con-
tinuous), fasting status (yes or no) and time of day of
blood draw (1 to 8 AM, 9 AM to noon or 1 PM to mid-
night). All analyses were adjusted for age at menopause
(continuous), parity (continuous), family history of
breast cancer (yes or no) and personal history of benign
breast disease (yes or no). Secondary analyses further
adjusted for body mass index (BMI) at the time of blood
draw.
We estimated the association of each hormone individu-
ally with postmenopausal breast cancer risk. Hormones
were modeled in quintiles (based on cutpoints in control
women), and RRs compared the top and bottom quintile.
Because of the relatively high correlations between sex
hormones (mean = 0.36, range = 0.13 to 0.71, 40% > 0.40),
the models were unstable when we included multiple
estrogens, androgens or all hormones simultaneously in
the same multivariate model. As such, a stepwise approach
to evaluating the addition of hormones to the statistical
model was not feasible. Therefore, we created several
scores to combine information about all hormones; these
are described below. The primary hormone scores
included the three estrogens, four androgens and prolac-
tin. The secondary analyses also included IGF-1 and c-
peptide. We calculated each score for estrogens only and
androgens only and included the two scores in the same
statistical model to evaluate the independent associations
of these two important hormone classes. Scores were cate-
gorized as indicated in Tables 2 through 4 and were
included as a continuous term to calculate tests for trend
using the Wald test.
The first goal of this study was to replicate the analysis
by Trichopoulos et al. [9], thus we created a score that
was the sum of the number of hormones for each woman
above the age- and batch-adjusted mean level. The results
derived using the median were similar and therefore are
not shown. We were interested in expanding the concept
of developing a hormone score, as the above score does
not take into account that some hormones are more
strongly associated with risk than others. To account for
the varying strengths of the association across hormones,
we created a score that weighted each hormone by its
individual association with overall breast cancer risk on
the basis of the following equation:
ScoreX
mm
m
h
12
1
=
=∑b *log (),
where m stands for each of the measured h hormones,
bmis the coefficient for that hormone modeled linearly
on the log2scale with breast cancer risk and Xmis the
value of the hormone. We used b coefficients for hor-
mone levels on the log2scale modeled continuously that
were reported in the literature with the largest sample
size evaluating sex hormones [1], prolactin [5] and IGF-
1 [8] (Tim Key for the Endogenous Hormones and
Breast Cancer Collaborative Group, personal communi-
cation for continuous b on the log2scale) and our own
data for c-peptide, as a pooled analysis was unavailable.
The coefficients used were 0.372 for estrone, 0.255 for
estradiol, 0.239 for estrone sulfate, 0.344 for testoster-
one, 0.293 for androstenedione, 0.215 for DHEA, 0.174
for DHEAS, 0.182 for prolactin, 0.166 for IGF-1 and
0.108 for c-peptide.
We also were interested in considering the biological
influence of the hormones on breast cancer cells and
thus created a score that weighted each hormone on the
basis of a proliferation index (PIm) with respect to
MCF-7 breast cancer cells, which is a common cell
model of breast cancer. We used this metric because
biological data were available for all hormones, at phy-
siological concentrations, using this cell line and mea-
suring some marker of proliferation. Since assay metrics
that have been published to measure proliferation dif-
fered across hormones, we calculated PImby dividing
the cell proliferation results for each hormone by the
cell proliferation results for estradiol under the same
conditions (for example, same assay, molar concentra-
tion and starting cell number) and calculated the score
by using the following equation:
ScorePIX
mm
m
h
2
1
=
=∑
*ln(),
where m stands for each of the measured h hormones,
PImis the proliferation index based on published data
and Xmis the value of the hormone. The PImvalues
(minimum and maximum) were 1.0 for estradiol (refer-
ence), 0.75 (0.51 to 1.10) for estrone [15-19], 0.45 (0.40
to 0.50) for estrone sulfate [19,20], 0.60 (0.21 to 0.78)
Tworoger et al. Breast Cancer Research 2011, 13:R99
http://breast-cancer-research.com/content/13/5/R99
Page 3 of 10

Page 4

for testosterone [21-26], 0.49 (0.35 to 0.67) for androste-
nedione [26-29], 0.72 (0.40 to 1.08) for DHEA
[20,23,24,26], 0.82 (0.78 to 0.85) for DHEAS [20,23,24]
and 1.04 (0.88 to 1.25) for prolactin [25,30,31]. All P
values were based on two-sided tests and were consid-
ered statistically significant if P ≤ 0.05. The analyses
were conducted using SAS version 9.1 software (Cary,
NC, USA).
Results
The mean age of cases was 62.1 years (SD = 4.5) and that
of controls was 61.8 years (SD = 4.7). A history of benign
breast disease was more common in cases than in con-
trols (37.7% vs 30.1%), as was a family history of breast
cancer (20.4% vs 15.3%). The mean age at menopause
was 50.2 years (SD = 3.6) in cases and 50.8 years (SD =
3.8) in controls. Mean parity was 3.0 (SD = 1.9) in cases
and 3.3 (SD = 2.0) in controls.
In a multivariate models including standard breast can-
cer risk factors, women with estrogen or androgen levels
in the top versus bottom quintile had approximately a
doubling of postmenopausal breast cancer risk (RR range
= 1.5 to 2.5) (Table 1). For most of the sex hormones, the
association was qualitatively stronger for ER-positive dis-
ease than for total breast cancer. Prolactin was suggestively
associated with ER-positive breast cancer risk (RR = 1.7),
but associations for IGF-1 and c-peptide were not statisti-
cally significant. These results are consistent with prior
reports from the NHS [2,5,6].
Increasing numbers of hormones above the age- and
batch-adjusted geometric mean level were linearly asso-
ciated with total (RR per one-unit increase = 1.16, 95%
CI = 1.08 to 1.24, P trend < 0.001) and ER-positive (com-
parable RR = 1.19, 95% CI = 1.09 to 1.29, P trend < 0.001)
breast cancer risk (Table 2). Comparing those data with
seven or eight hormones above the geometric mean level
versus none, the RR was 2.7 (95% CI = 1.3 to 5.7) for
breast cancer overall and 3.4 (95% CI = 1.3 to 9.4) for ER-
positive disease. Because of the small number of ER-posi-
tive cases with no hormones above the geometric mean
level, we also compared those with seven or eight hor-
mones above the geometric mean versus zero or one hor-
mone and observed that the RR was 3.0 (95% CI = 1.5 to
5.9). A similar linear association with risk was observed for
the score weighted by individual hormone associations.
Women in the top versus the bottom quintile had a RR =
3.0 (95% CI = 1.8 to 5.0, P trend < 0.001) for total breast
cancer and 3.9 (95% CI = 2.0 to 7.5, P trend < 0.001) for
ER-positive tumors. In the score weighted by the PIm,
women in the top versus bottom quintile had a 2.5-fold
increased breast cancer risk and a 3.5-fold increased risk
of ER-positive disease. Scores developed using the mini-
mum or maximum PImhad similar associations and had a
correlation of 0.99 with the score using the mean PIm
(data not shown). Adjustment for BMI at the time of
blood draw did not substantially change the results (data
not shown). For example, the RR comparing the top ver-
sus the bottom quintile of the score weighted by individual
hormone associations, after adjusting for BMI, was 2.9
(95% CI = 1.7 to 4.8) for total breast cancer and 3.7 (95%
CI = 1.9 to 7.2) for ER-positive disease. The three hor-
mone scores were very highly correlated (range = 0.90
between the score summing the number of hormones
above the geometric mean and the score weighting by the
PImto 0.98 between the score weighting by the breast can-
cer association and that weighting by PIm).
The results were similar for invasive breast cancer and
when using free estradiol and free testosterone instead of
total levels, although the sample size was decreased
because of missing SHBG levels (data not shown). The
Table 1 Hormone levels among postmenopausal women not taking hormones in the Nurses’ Health Study and the
individual association between each hormone and breast cancer risk overall and for estrogen receptor-positive tumors
Median (10th to 90th percentile) RR, top vs bottom quintile (95% CI)a, b
HormonesCases Controls All casesER-positive cases
Estrone (pg/ml)
Estradiol (pg/ml)
Estrone sulfate (pg/ml)
Testosterone (ng/ml)
Androstenedione (ng/ml)
DHEA (ng/dl)
DHEAS (μg/dl)
Prolactin (ng/ml)
IGF-1 (ng/ml)
c-peptide (ng/ml)
28 (15 to 49)
8 (4 to 17)
276 (102 to 733)
24 (14 to 49)
63 (35 to 108)
234 (104 to 506)
92 (41 to 224)
8.2 (4.9 to 14.8)
152 (98 to 252)
1.71 (0.66 to 4.11)
24 (14 to 43)
6 (4 to 14)
222 (98 to 577)
22 (12 to 40)
57 (30 to 108)
218 (93 to 435)
85 (33 to 175)
8.0 (4.9 to 14.0)
153 (98 to 242)
1.53 (0.61 to 4.03)
2.1 (1.3 to 3.4)
2.4 (1.4 to 4.1)
2.4 (1.5 to 3.9)
1.8 (1.1 to 2.9)
2.1 (1.3 to 3.6)
1.5 (0.9 to 2.4)
2.5 (1.4 to 4.2)
1.1 (0.7 to 1.7)
1.1 (0.6 to 2.0)
1.4 (0.8 to 2.4)
2.8 (1.5 to 5.3)
2.9 (1.4 to 5.9)
2.2 (1.2 to 4.0)
2.0 (1.0 to 3.7)
2.6 (1.3 to 5.0)
1.6 (0.9 to 2.9)
2.0 (1.0 to 3.8)
1.7 (0.9 to 3.1)
1.4 (0.7 to 2.7)
1.4 (0.7 to 3.0)
DHEA = dehydroepiandrosterone; DHEAS = dehydroepiandrosterone sulfate; ER = estrogen receptor; IGF-1 = insulin-like growth factor 1.aRR and 95% CI were
adjusted for age at blood draw and menopause, parity, history of benign breast disease, family history of breast cancer, fasting status and time of day of blood
draw.bFor all hormones (except IGF-1 and c-peptide), n = 265 total cases (invasive and in situ), n = 147 ER-positive, invasive cases, and n = 541 controls; for IGF-1
and c-peptide, n = 182 total cases (invasive and in situ), n = 105 ER-positive, invasive cases, and n = 370 controls.
Tworoger et al. Breast Cancer Research 2011, 13:R99
http://breast-cancer-research.com/content/13/5/R99
Page 4 of 10

Page 5

associations between the hormone scores and breast can-
cer risk also were similar after excluding cases diagnosed
in the first two years of follow-up and when stratifying by
age (younger than 65 years of age vs age 65 years or
older) or BMI (less than 25 kg/m2vs 25 kg/m2or more)
(data not shown). Each score was qualitatively more
strongly associated with breast cancer risk among women
who had never used PMH versus past users, although
none of the interactions were statistically significant (data
not shown), likely because one assessment of hormone
levels better reflects long-term exposure in women who
have never used PMH.
When a separate score for estrogens and androgens
was included in the same statistical model, the associa-
tion for the estrogen score was slightly stronger than for
the corresponding androgen score, although the associa-
tions were not statistically significantly different (P =
0.52) (Table 3). For example, the RR comparing the top
versus bottom quintile for the estrogen score weighted
by individual hormone associations was 2.1 (95% CI = 1.2
to 3.7, P trend = 0.005) and that for the androgen score
was 1.7 (95% CI = 1.0 to 3.0, P trend = 0.06) for all breast
cancers. The comparable associations for ER-positive dis-
ease were 2.4 (95% CI = 1.2 to 4.9, P trend = 0.003) and
2.0 (95% CI = 1.0 to 4.2, P trend = 0.21), respectively.
Adjustment for BMI at the time of blood draw did not
appreciably change the risk estimates for total breast can-
cer or ER-positive disease (data not shown). The correla-
tion between the estrogen and androgen scores was 0.41
for the score summing the number of hormones above
the geometric mean and 0.47 for the score weighted by
the association with breast cancer risk.
No interactions were observed between the estrogen
and androgen scores (P heterogeneity > 0.33). When we
considered women with both estrogen and androgen
scores below the 75th percentile (4.7 for the estrogen
Table 2 Risk of breast cancer among postmenopausal women not using hormones in the Nurses’ Health Study on the
basis of several hormone scoresa
Sample size RR (95% CI)b
Hormone scores
Number of hormones > geometric meand
0
1 to 2
3 to 4
5 to 6
7 to 8
Per 1-unit increase
Per 1 SD increased
P trend
Score weighted by individual hormone associationse
Quintile 1
Quintile 2
Quintile 3
Quintile 4
Quintile 5
Per 1 SD increasee
P trend
Score weighted by MCF-7 proliferationf
Quintile 1
Quintile 2
Quintile 3
Quintile 4
Quintile 5
Per 1 SD increasef
P trend
Total casesER+ casesControlsAll cases ER+ casesc
11
35
63
82
74
265
265
265
5
18
35
43
46
147
147
147
43
113
146
137
102
541
541
541
1.0
1.1 (0.5 to 2.4)
1.6 (0.8 to 3.3)
2.1 (1.0 to 4.5)
2.7 (1.3 to 5.7)
1.16 (1.08 to 1.24)
1.42 (1.21 to 1.66)
< 0.001
1.0
1.2 (0.4 to 3.4
1.8 (0.7 to 5.0)
2.4 (0.9 to 6.5)
3.4 (1.3 to 9.4)
1.19 (1.09 to 1.29)
1.51 (1.24 to 1.85)
< 0.001
30
42
48
58
87
265
265
14
24
25
32
52
147
147
109
107
109
108
108
541
541
1.0
1.4 (0.8 to 2.4)
1.6 (0.9 to 2.7)
2.0 (1.2 to 3.3)
3.0 (1.8 to 5.0)
1.45 (1.23 to 1.70)
< 0.001
1.0
1.7 (0.8 to 3.5)
1.8 (0.9 to 3.7)
2.3 (1.1 to 4.5)
3.9 (2.0 to 7.5)
1.54 (1.26 to 1.88)
< 0.001
35
37
55
53
85
265
265
16
27
23
29
52
147
147
108
109
108
109
107
541
541
1.0
1.0 (0.6 to 1.7)
1.6 (0.9 to 2.6)
1.5 (0.9 to 2.5)
2.5 (1.5 to 4.1)
1.44 (1.22 to 1.69)
< 0.001
1.0
1.6 (0.8 to 3.2)
1.5 (0.7 to 2.9)
1.8 (0.9 to 3.5)
3.5 (1.8 to 6.6)
1.55 (1.26 to 1.89)
< 0.001
ER = estrogen receptor.aHormones included in the score are estrone, estradiol, estrone sulfate, testosterone, androstenedione, dehydroepiandrosterone,
dehydroepiandrosterone sulfate and prolactin.bRR and 95% CI were adjusted for age at blood draw and menopause, parity, history of benign breast disease,
family history of breast cancer, fasting status and time of day of blood draw.cn = 147 ER-positive cases.dScore is the number of hormones for which the
participant had higher than the age- and assay batch-adjusted geometric mean hormone level. Mean = 4.0 (SD = 2.4).eScore is the sum of the log2-transformed
hormone level multiplied by the b coefficient for that hormone with breast cancer risk. Mean = 10.8 (SD = 1.1) among controls.fScore is the sum (across all
hormones) of the natural log-transformed hormone level multiplied by a standardized proliferation rate (compared to estradiol) for MCF-7 breast cancer cells.
Mean = 20.2 (SD = 2.1) among controls.
Tworoger et al. Breast Cancer Research 2011, 13:R99
http://breast-cancer-research.com/content/13/5/R99
Page 5 of 10