ARTICLE
Risk Factors That Increase Risk of Estrogen
Receptor–Positive and –Negative Breast Cancer
Karla Kerlikowske, Charlotte C. Gard, Jeffrey A. Tice, Elad Ziv,
Steven R. Cummings, Diana L. Miglioretti; on behalf of the Breast Cancer
Surveillance Consortium
Affiliations of authors: Departments of Medicine and Epidemiology and Biostatistics (KK, JAT, EZ) and General Internal Medicine Section, Department of Veterans
Affairs (KK), University of California, San Francisco, San Francisco, CA; Department of Economics, Applied Statistics, and International Business, New Mexico State
University, Las Cruces, NM (CCG); San Francisco Coordinating Center, California Pacific Medical Center Research Institute, San Francisco, CA (SRC); Department of
Public Health Sciences, University of California, Davis, Davis, CA (DLM); Group Health Research Institute, Group Health Cooperative, Seattle, WA (DLM).
Correspondence to: Karla Kerlikowske, MD, San Francisco Veterans Affairs Medical Center, General Internal Medicine Section, 111A1, 4150 Clement Street, San
Francisco CA 94121 (e-mail: karla.kerlikowske@ucsf.edu).
Abstract
Background: Risk factors may differentially influence development of estrogen receptor (ER)–positive vs –negative breast
cancer. We examined associations with strong, prevalent risk factors by ER subtype.
Methods: Of 1 279 443 women age 35 to 74 years participating in the Breast Cancer Surveillance Consortium, 14 969 developed
ER-positive and 3617 developed ER-negative invasive breast cancer. We calculated hazard ratios (HRs) using Cox regression
and compared ER subtype hazard ratios at representative ages or by menopausal status using Wald tests. All statistical tests
were two-sided.
Results: For women age 40 years, compared with no prior biopsy, ER-positive vs ER-negative HRs were 1.53 (95% CI¼ 1.30 to
1.81) vs 1.26 (95% CI ¼ 0.90 to 1.76) for nonproliferative disease, 1.63 (95% CI ¼ 1.23 to 2.17) vs 1.41 (95% CI ¼ 0.78 to 2.57) for prolif-
erative disease without atypia, and 4.47 (95% CI¼ 2.88 to 6.96) vs 0.20 (95% CI ¼ 0.02 to 2.51) for proliferative disease with atypia.
Benign disease proliferation risk was stronger for ER-positive than ER-negative cancer for women age 35 years (Wald P ¼ .04),
age 40 years (Wald P ¼ .04), and age 50 years (Wald P ¼ .06). Among pre/perimenopausal women, body mass index (BMI) had a
stronger association with ER-negative than ER-positive cancer (obese II/III vs. normal weight: HR ¼ 1.52, 95% CI ¼ 1.19 to 1.94; vs
1.21, 95% CI ¼ 1.08 to 1.36). Increasing BMI similarly increased ER-positive and ER-negative cancer risk among postmenopausal
hormone users (Wald P ¼ .15) and nonusers (Wald P ¼ .08). Associations with ER subtype varied by race/ethnicity across all ages
(P < .001) and by family history of breast cancer and breast density for specific ages.
Conclusions: Strength of risk factor associations differed by ER subtype. Separate risk models for ER subtypes may improve
identification of women for targeted prevention strategies.
Primary prevention with selective estrogen receptor modulators
(SERM) decreases the risk of estrogen-receptor (ER)–positive, but
not ER-negative, breast cancer, and detection with digital mam-
mography varies by ER status (1–3). Available risk prediction
models, whose implementation is a key aspect of guiding pre-
vention strategies, predict invasive cancer risk overall rather
than by ER subtype (4–6).
Few studies of risk factors used in risk models have exam-
ined associations by ER subtype and, of those that have, results
are inconsistent. Studies have reported history of benign breast
disease (BBD) increases risk of ER-positive and -negative cancer
among pre- (7) and postmenopausal women (8). While one
study of postmenopausal women reported a stronger associa-
tion of BBD with ER-negative than ER-positive cancer (9) and
ARTICLE
Received: April 23, 2016; Revised: September 17, 2016; Accepted: October 19, 2016
© The Author 2016. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.
1of9
JNCI J Natl Cancer Inst (2017) 109(5): djw276
doi: 10.1093/jnci/djw276
First published online December 31, 2016
Article
Downloaded from https://academic.oup.com/jnci/article/109/5/djw276/2898140 by guest on 21 August 2022
another of premenopausal and postmenopausal women re-
ported similar associations (7), neither study stratified by BBD
diagnosis. A BBD cohort study reported women with atypical
hyperplasia had a higher frequency of ER-positive cancer com-
pared with women with proliferative disease without atypia or
nonproliferative disease (10).
Several studies, including a recent meta-analysis (11), report
no difference in associations between breast density and breast
cancer risk by ER subtype (11–17), while others report stronger
associations for ER-positive cancer (18–20). In contrast, a large
study suggests breast density is more strongly associated with
ER-negative (vs ER-positive) breast cancer, but only among
women younger than age 55 years compared with older women
(21).
Elevated body mass index (BMI) is associated with increased
ER-positive cancer risk in postmenopausal women (22–24),
while results for ER-negative cancer are inconsistent. One study
of postmenopausal women not using hormone therapy (HT) re-
ported an association with ER-negative cancer and elevated BMI
(25), while other studies have found no association (22–24).
Studies have reported no association or a decreased risk with el-
evated BMI by ER subtypes in premenopausal women (23,25,26),
but ER-negative studies have limited statistical power (23–25).
We used data from the large, prospective Breast Cancer
Surveillance Consortium (BCSC) cohort to examine the associa-
tions of ER-positive and ER-negative invasive cancer with
strong, prevalent risk factors used in risk prediction models
(4–6) to assess if risk factors are differentially associated with ER
subtypes.
Methods
Study Population
The National Cancer Institute (NCI)-funded BCSC (http://breast
screening.cancer.gov) (27) is a community-based, geographically
diverse cohort study that broadly represents the population of
women undergoing mammography in the United States (28).
Our sample consisted of 1 279 443 women age 35 to 74 years
who had at least one mammogram with American College of
Radiology Breast Imaging Reporting and Data System (BI-RADS)
breast density reported between 1994 and 2014 and complete in-
formation on race/ethnicity, family history of breast cancer, and
history of breast biopsy, as well as, for those who developed in-
vasive breast cancer, ER status. We excluded women with a di-
agnosis of breast cancer prior to their first eligible mammogram
or within the first three months of follow-up and women who
had breast implants or who had undergone mastectomy. Each
registry obtains annual institutional review board approval for
consenting processes or a waiver of consent, enrollment of par-
ticipants, and data linkages for research purposes. All registries
received a Federal Certificate of Confidentiality that protects the
identities of research participants.
Measurement of Risk Factors
Self-reported age, first-degree family history of breast cancer,
race/ethnicity, prior breast biopsy history, height, weight, men-
opause status, and current postmenopausal HT use were ob-
tained at the time of each mammography examination.
Postmenopausal women were those with both ovaries removed,
whose periods had stopped naturally, with current HT use, or
age 55 years or older. Premenopausal women reported a period
within the last 180 days, were younger than age 40 years, or
were birth control hormone users, while perimenopausal were
not sure if their periods had stopped or their last menstrual pe-
riod was 180 to 364 days ago. Women were considered to have
missing menopausal status if they had a hysterectomy without
bilateral oophorectomy or surgical menopause and were not us-
ing HT or if their menopause status could not be determined
based on available information (29–31). Height and weight were
used to calculate BMI by dividing weight in kilograms by height
in meters squared (kg/m
2
). BMI was examined as a continuous
variable and stratified into five standard categories (under-
weight < 18.5 kg/m
2
, normal weight ¼ 18.5–24.95 kg/m
2
, over-
weight ¼ 25.0–29.95 kg/m
2
, grade I: obese ¼ 30.0–34.95 kg/m
2
;
grade II/III: obese > 35.05 kg/m
2
) using nationally defined cut-
points (32). Race and ethnicity were coded using the expanded
definition currently used in Surveillance, Epidemiology, and
End Results (SEER) and US Vital statistics (non-Hispanic white,
non-Hispanic black, Asian/Pacific Islander, Native American/
Alaskan Native, Hispanic, other/mixed race).
Benign Breast Disease
Community pathologists at each registry classified breast bi-
opsy results using their local clinical practice. We grouped each
diagnosis from pathology reports into one of four categories:
nonproliferative, proliferative without atypia, proliferative with
atypia, and lobular carcinoma in situ (LCIS) using the taxonomy
proposed by Dupont and Page (33–35). Details of benign diagno-
ses included in each category have been described previously
(https://tools.bcsc-scc.org/bc5yearrisk/calculator.htm) (4). If
there was more than one diagnosis per biopsy or multiple biop-
sies were performed within three months, we chose the diagno-
sis with the highest grade (LCIS > atypical hyperplasia >
proliferative without atypia > nonproliferative). We classified
the biopsy as diagnosis unknown if a woman reported a prior
biopsy but pathology results were not available.
Mammographic Breast Density
Community radiologists classified breast density as part of rou-
tine clinical practice using the four BI-RADS density categories
(36): a ¼ almost entirely fat; b ¼ scattered fibroglandular densi-
ties c ¼ heterogeneously dense; d ¼ extremely dense.
Ascertainment of Breast Cancer Cases
Breast cancers diagnosed at least three months after index
mammograms were obtained through linkage with the regional
population-based SEER programs, state tumor registries, and
pathology databases. Invasive cancers were categorized by ER
status.
Vital Status
Vital status was obtained through linkage to SEER and state tu-
mor registries and state death records.
Statistical Analysis
We estimated hazard ratios (HRs) for each risk factor using
partly conditional Cox proportional hazards regression to incor-
porate biopsies occurring after study entry and to facilitate risk
ARTICLE
K. Kerlikowske et al. | 2 of 9
Downloaded from https://academic.oup.com/jnci/article/109/5/djw276/2898140 by guest on 21 August 2022
prediction (37). We fit separate models for ER-positive and ER-
negative invasive breast cancer. We used a robust sandwich es-
timator for repeated measures survival data to account for mul-
tiple observations per woman (38). The model time scale started
three months after the index screening or diagnostic mammo-
gram and possibly again three months after each new, more se-
vere biopsy result. Age and risk factors were updated each time
women entered the model. Each observation was censored at
the time of death, diagnosis of invasive cancer of the other ER
subtype or DCIS, mastectomy, at the end of complete cancer
capture for her BCSC registry, or at 10 years of follow-up. All
models were adjusted for risk factors and interactions in the
BCSC model including age at study entry, age-squared and race/
ethnicity, and interactions between age and breast density,
family history, BBD, and race/ethnicity, all which had a P value
of less than .10 in the ER-positive model. Hazard ratios are re-
ported at representative ages because of statistically significant
interactions of age with other risk factors. As BMI was missing
for 43% of women, we fit separate models to examine the asso-
ciation of BMI by ER subtype including interactions between
BMI and menopause status. We report P
interaction
values for lin-
ear and quadratic age and BMI separately, which makes the
type I error rate higher than 5% but provides more information
for descriptive purposes. We assessed the proportional hazards
assumption by calculating interval-specific hazard ratios (ie,
zero to three months, three to six months, six months to one
year, one to two years, etc.) for each predictor variable and com-
paring for clinically meaningful changes over time. The propor-
tional hazards assumption appeared reasonable for all
predictors except BBD, where hazard ratios within six months
of study entry were higher than in subsequent years. Excluding
these data did not change the model results in a clinically sig-
nificant manner.
We used Wald chi-square tests in a competing risks model
to compare ER-positive and ER-negative model parameters (39).
Comparisons were made within levels of age and menopausal
status for risk factors involving interactions.
Analyses were performed using R version 3.0.3 and SAS ver-
sion 9.3. Statistical tests were two-sided, and the cut-point for
statistical significance was .05.
Results
Study Characteristics
During a median follow-up of 8.3 years, 18 586 women devel-
oped invasive breast cancer, 14 969 ER-positive (80.5%) and 3617
ER-negative (19.5%). Black women had the highest proportion of
ER-negative tumors (37.7%) (Table 1).
BBD Associations by ER Subtypes
The strength of BBD associations with ER-positive (P
interaction
¼
.08 for linear age and P ¼ .07 for quadratic age) and ER-negative
cancer (P
interaction
¼ .17 for linear age and P ¼ .18 for quadratic
age) did not vary with age. ER-positive cancer risk increased
with extent of BBD proliferation for all ages of women (Table 2).
For women age 40 years, compared with no prior biopsy, ER-
positive vs ER-negative hazard ratios were 1.53 (95% CI ¼ 1.30 to
1.81) vs 1.26 (95% CI ¼ 0.90 to 1.76) for nonproliferative disease,
1.63 (95% CI ¼ 1.23 to 2.17) vs 1.41 (95% CI ¼ 0.78 to 2.57) for pro-
liferative disease without atypia, 4.47 (95% CI ¼ 2.88 to 6.96) vs
0.20 (95% CI ¼ 0.02 to 2.51) for proliferative disease with atypia,
and 9.93 (95% CI ¼ 4.79 to 20.6) for LCIS. BBD proliferation in-
creased ER-negative cancer risk the most in women age 60 years
(Table 2). BBD proliferation associations were stronger for ER-
positive than ER-negative cancer for women younger than age
50 years (Wald P ¼ .04, age 35 years; P ¼ .04, age 40 years; P ¼ .06,
age 50 years).
BMI Associations by ER Subtypes
Menopausal status was statistically significantly associated
with ER-positive cancer (P < .001) and ER-negative cancer (P ¼
.03), with statistically significant interactions with BMI (Table 3).
Pre/perimenopausal and postmenopausal overweight/obese
women were at increased risk for ER-positive cancer, while un-
derweight women were at decreased risk compared with nor-
mal weight women. Pre/perimenopausal women and
postmenopausal overweight/obese HT nonusers were at in-
creased ER-negative cancer risk, while underweight women
were at decreased risk compared with normal weight women.
For example, pre/perimenopausal obese, grade II/III vs normal
weight women had a stronger association with ER-negative vs
ER-positive cancer (HR ¼ 1.52, 95% CI ¼ 1.19 to 1.94; vs HR ¼ 1.21,
95% CI ¼ 1.08 to 1.36) (Table 3). Also, postmenopausal obese,
grade II/III HT users were at increased ER-negative cancer risk
compared with normal weight women.
Increasing BMI similarly increased ER-positive and ER-
negative cancer risk among postmenopausal hormone users
(Wald P ¼ .15) and nonusers (Wald P ¼ .08), but associations
were stronger with ER-negative than ER-positive cancer for pre/
perimenopausal women (Wald P ¼ .03) (Table 3).
Race/Ethnicity, Family History, and Breast Density
Associations by ER Subtypes
The strength of race/ethnicity associations with ER subtype
(Figure 1) statistically significantly declined with age for ER-
positive (P
interaction
< .001) and ER-negative cancer (P
interaction
¼
.004). Black, non-Hispanic women had an increased ER-negative
cancer risk for all ages compared with white, non-Hispanic
women with the highest hazard ratio for women age 35 years
(3.05, 95% CI ¼ 2.52 to 3.70) and lowest for women age 70 years
(1.56, 95% CI ¼ 1.28 to 1.90). The effect of race/ethnicity on risk
differed by ER subtype for all ages (Wald P < .001) because of an
increased ER-negative cancer risk for black, non-Hispanic
women.
The strength of family history associations with ER-positive
cancer (Figure 2) statistically significantly declined with age
(P
interaction
¼ .02 for linear age and P ¼ .03 for quadratic age), but
not for ER-negative cancer (P
interaction
¼ .33 for linear age and P ¼
.46 for quadratic age). Family history of breast cancer associa-
tions was stronger for ER-positive compared with ER-negative
cancer for women older than age 50 years (Wald P ¼ .09, age 50
years; P ¼ .02, age 60 years; P ¼ .03, age 70 years).
The strength of BI-RADS density associations with ER-
positive cancer (Figure 3) statistically significantly declined with
age (P
interaction
< .001), but not for ER-negative cancer (P
interaction
¼ .34). For example, hazard ratios for extremely dense breasts
compared with scattered fibroglandular densities remained ele-
vated at 1.68 (95% CI ¼ 1.34 to 2.11) at age 35 years and 1.76 (95%
CI ¼ 1.36 to 2.27) at age 70 years for ER-negative cancer, but de-
clined from 2.13 (95% CI ¼ 1.89 to 2.40) at age 35 years to 1.35
(95% CI ¼ 1.20 to 1.53) at age 70 years for ER-positive cancer.
Women with dense breasts had a stronger association with ER-
ARTICLE
3of9 | JNCI J Natl Cancer Inst, 2017, Vol. 109, No. 5
Downloaded from https://academic.oup.com/jnci/article/109/5/djw276/2898140 by guest on 21 August 2022
positive compared with ER-negative cancer for women younger
than age 40 years (Wald P ¼ .05,age35years;P ¼ .08,age40years).
Discussion
We examined strong, prevalent risk factors included in risk
prediction models and their associations with ER subtypes be-
cause women’s risk of ER subtypes can inform targeted pre-
vention strategies. We found that family history of breast
cancer, BBD, breast density, and race/ethnicity increased the
risk for both ER-positive and ER-negative invasive breast can-
cer but the strength of associations varied by age. For
example, associations with high breast density decreased
with age for ER-positive cancer and remained elevated for ER-
negative cancer, whereas associations with family history of
breast cancer decreased with age for ER-negative cancer to a
greater extent than for ER-positive cancer. In comparison, ele-
vated postmenopausal BMI similarly increased ER-positive
and ER-negative cancer risk while elevated peri/premeno-
pausal BMI increased ER-negative cancer risk to a greater ex-
tent than ER-positive risk. Our results suggest that there are
important differences in associations by ER subtype with age
and menopausal status for common risk factors and that
these associations should be taken into account in developing
risk prediction models (8).
Table 1. Baseline characteristics of study cohort
No breast cancer ER-positive cancer ER-negative cancer
Characteristics No. (%) No. (%) No. (%)
Total 1 260 857 (98.6) 14 969 (1.2) 3617 (0.3)
Age groups, y
35–39 111 400 (8.8) 658 (4.4) 238 (6.6)
40–44 287 067 (22.8) 2145 (14.3) 606 (16.8)
45–49 222 884 (17.7) 2213 (14.8) 689 (19.1)
50–54 204 094 (16.2) 2462 (16.5) 581 (16.1)
55–59 152 357 (12.1) 2327 (15.6) 524 (14.5)
60–64 116 788 (9.3) 2012 (13.4) 394 (10.9)
65–69 95 224 (7.6) 1792 (12.0) 332 (9.2)
70–74 71 043 (5.6) 1360 (9.1) 253 (7.0)
Race/ethnicity
White, non-Hispanic 936 413 (74.3) 12 220 (81.6) 2622 (72.5)
Black, non-Hispanic 94 725 (7.5) 884 (5.9) 536 (14.8)
Asian, Native Hawaiian, or Pacific Islander 84 122 (6.7) 644 (4.3) 167 (4.6)
American Indian or Alaska Native 11 801 (0.9) 75 (0.5) 27 (0.8)
Hispanic 115 082 (9.1) 962 (6.4) 220 (6.1)
Other, mixed (2þ races) 18 714 (1.5) 184 (1.2) 45 (1.2)
1st-degree relatives with breast cancer
No 1 109 616 (88.0) 12 187 (81.4) 3036 (83.9)
Yes 151 241 (12.0) 2782 (18.6) 581 (16.1)
Benign breast disease
No prior biopsy 1 060 408 (84.1) 11 173 (74.6) 2783 (76.9)
Prior biopsy, diagnosis unknown 158 967 (12.6) 3073 (20.5) 704 (19.5)
Nonproliferative 29 038 (2.3) 446 (3.0) 89 (2.5)
Proliferative without atypia 10 349 (0.8) 193 (1.3) 30 (0.8)
Proliferative with atypia 1662 (0.1) 52 (0.4) 7 (0.2)
Lobular carcinoma in situ 433 (0.03) 32 (0.2) 4 (0.1)
BI-RADS breast density*
Almost entirely fat 106 244 (8.4) 713 (4.8) 157 (4.3)
Scattered densities 528 925 (42.0) 5675 (37.9) 1350 (37.3)
Heterogeneously dense 499 894 (39.7) 6821 (45.6) 1679 (46.4)
Extremely dense 125 794 (10.0) 1760 (11.8) 431 (11.9)
Body mass index, kg/m
2
†
Underweight (<18.5) 13 137 (1.8) 140 (1.7) 22 (1.3)
Normal (18.5–24.9) 328 162 (46.0) 3581 (44.6) 722 (43.2)
Overweight (25.0–29.9) 201 841 (28.3) 2402 (29.9) 521 (31.2)
Obese, grade I (30.0–34.9) 100 197 (14.0) 1152 (14.3) 226 (13.5)
Obese, grade II/III (35.0) 70 406 (9.9) 761 (9.5) 181 (10.8)
Menopausal status‡
Pre- or perimenopausal 501 293 (46.2) 4335 (32.5) 1243 (39.4)
Postmenopausal with HT 229 557 (21.2) 3895 (29.2) 824 (26.1)
Postmenopausal no HT 353 610 (32.6) 5115 (38.3) 1089 (34.5)
*BI-RADS density categories: a ¼ almost entirely fat; b ¼ scattered fibroglandular densities; c ¼ heterogeneously dense; d ¼ extremely dense. BI-RADS ¼ Breast Imaging
Reporting and Data System; ER ¼ estrogen receptor; HT ¼ hormone therapy.
†Missing body mass index: 547 114 no breast cancer, 6933 ER-positive, 1945 ER-negative.
‡Missing menopausal status: 176 397 no breast cancer, 1624 ER-positive, 461 ER-negative; 95% premenopausal and 5% perimenopausal.
ARTICLE
K. Kerlikowske et al. | 4 of 9
Downloaded from https://academic.oup.com/jnci/article/109/5/djw276/2898140 by guest on 21 August 2022
Observational studies consistently report that obesity is as-
sociated with increased breast cancer risk in postmenopausal
women (40) and weight gain is associated with increased risk in
women of all ages (40–42). Consistent with our results, Ritte et
al. reported that elevated BMI increases ER-negative cancer risk
among overweight/obese postmenopausal women who never
used HT, but Ritte et al. did not find an association of BMI and
ER-negative cancer diagnosed among premenopausal women
(25). In the Shanghai Breast Cancer Study, premenopausal
breast cancer was not associated with increasing BMI, but both
Table 2. Cox proportional hazards model results for benign breast disease by estrogen receptor–positive or estrogen receptor–negative events
HR (95% CI)*
Characteristic Age 35 y Age 40 y Age 50 y Age 60 y Age 70 y
ER-positive
Benign breast disease
No prior biopsy (ref) (ref) (ref) (ref) (ref)
Prior biopsy, diagnosis unknown 1.39 (1.17 to 1.65) 1.41 (1.27 to 1.57) 1.46 (1.40 to 1.53) 1.52 (1.45 to 1.59) 1.58 (1.48 to 1.68)
Nonproliferative 1.61 (1.22 to 2.13) 1.53 (1.30 to 1.81) 1.49 (1.36 to 1.63) 1.58 (1.44 to 1.74) 1.84 (1.60 to 2.11)
Proliferative without atypia 1.53 (0.95 to 2.46) 1.63 (1.23 to 2.17) 1.83 (1.62 to 2.08) 2.03 (1.78 to 2.33) 2.23 (1.85 to 2.69)
Proliferative with atypia 5.05 (2.44 to 10.43) 4.47 (2.88 to 6.96) 3.68 (2.98 to 4.54) 3.21 (2.59 to 3.97) 2.97 (2.10 to 4.19)
LCIS 19.79 (6.09 to 64.35) 9.93 (4.79 to 20.57) 4.32 (3.04 to 6.12) 3.90 (2.75 to 5.52) 7.30 (4.75 to 11.21)
ER-negative
No prior biopsy (ref) (ref) (ref) (ref) (ref)
Prior biopsy, diagnosis unknown 1.56 (1.13 to 2.15) 1.52 (1.26 to 1.84) 1.47 (1.34 to 1.61) 1.44 (1.30 to 1.59) 1.43 (1.23 to 1.67)
Nonproliferative 1.08 (0.60 to 1.94) 1.26 (0.90 to 1.76) 1.53 (1.27 to 1.84) 1.60 (1.31 to 1.96) 1.43 (1.00 to 2.06)
Proliferative without atypia 1.33 (0.47 to 3.75) 1.41 (0.78 to 2.57) 1.53 (1.15 to 2.02) 1.57 (1.14 to 2.17) 1.54 (0.92 to 2.58)
Proliferative with atypia NE† 0.20 (0.02 to 2.51) 1.23 (0.55 to 2.75) 2.33 (1.22 to 4.48) 1.34 (0.43 to 4.22)
LCIS NE† NE† 1.05 (0.17 to 6.30) 3.73 (1.39 to 10.05) 1.52 (0.23 to 9.94)
P, Wald test§ .04 .04 .06 .61 .16
*Adjusted for age at entry (linear and quadratic terms) and race/ethnicity, with interactions between Breast Imaging Reporting and Data System density and age at en-
try (linear), first-degree relatives and age at entry (linear and quadratic), benign breast disease (linear and quadratic terms), and race/ethnicity and age at entry (linear).
Analyses include 1 361 990 mammograms associated with 16 749 estrogen receptor (ER)–positive and 3990 ER-negative cancers. CI ¼ confidence interval; ER ¼ estrogen
receptor; HR ¼ hazard ratio; LCIS ¼ lobular carcinoma in situ.
†Nonestimable because of small sample sizes.
§Compares estrogen receptor–positive vs –negative cancer parameters within age group. This test was two-sided.
Table 3. Cox proportional hazards model results for 687 601 examinations and 8386 estrogen receptor–positive or 1715 estrogen receptor–
negative events by body mass index
HR (95% CI)*
ER subtype by BMI Pre- or perimenopausal Postmenopausal current HT use Postmenopausal no current HT use
ER-positive†
BMI, kg/m
2
‡
18 (underweight) 0.93 (0.87 to 0.98) 0.90 (0.83 to 0.97) 0.79 (0.74 to 0.84)
22 (normal) (ref) (ref) (ref)
27 (overweight) 1.08 (1.03 to 1.14) 1.11 (1.05 to 1.18) 1.28 (1.21 to 1.35)
32 (obese, grade I) 1.15 (1.06 to 1.25) 1.20 (1.09 to 1.32) 1.53 (1.41 to 1.66)
39 (obese, grade II/III) 1.21 (1.08 to 1.36) 1.28 (1.11 to 1.47) 1.78 (1.61 to 1.98)
ER-negative§
BMI, kg/m
2
‡
18 (underweight) 0.76 (0.66 to 0.87) 0.96 (0.85 to 1.09) 0.88 (0.79 to 0.98)
22 (normal) (ref) (ref) (ref)
27 (overweight) 1.28 (1.14 to 1.44) 1.08 (0.95 to 1.22) 1.17 (1.06 to 1.30)
32 (obese, grade I) 1.48 (1.25 to 1.77) 1.19 (0.96 to 1.47) 1.38 (1.16 to 1.63)
39 (obese, grade II/III) 1.52 (1.19 to 1.94) 1.44 (1.06 to 1.94) 1.72 (1.36 to 2.17)
Pk .03 .15 .08
*Adjusted for age at entry (linear and quadratic terms) and race/ethnicity, with interactions between Breast Imaging Reporting and Data System density and age at en-
try (linear), first-degree relatives and age at entry (linear and quadratic), benign breast disease (linear and quadratic terms), and race/ethnicity and age at entry (linear).
BMI ¼ body mass index; CI ¼ confidence interval; ER ¼ estrogen receptor; HR ¼ hazard ratio.
†P values for continuous BMI linear and quadratic terms ¼ .06 and .19, respectively; P value for menopausal status 0.001; P
interaction
values between continuous BMI
linear and quadratic terms and menopausal status ¼ .02 and .12, respectively. All statistical tests were two-sided.
‡Median of values within range of values for category.
§P values for continuous BMI linear and quadratic terms ¼ .002 and .009, respectively; P value for menopausal status ¼ .03; P
interaction
values between continuous BMI
linear and quadratic terms and menopausal status ¼ .02 and .02, respectively. All statistical tests were two-sided.
kCompares estrogen receptor–positive vs –negative cancer parameters within menopausal group. P values were calculated using a two-sided Wald test.
ARTICLE
5of9 | JNCI J Natl Cancer Inst, 2017, Vol. 109, No. 5
Downloaded from https://academic.oup.com/jnci/article/109/5/djw276/2898140 by guest on 21 August 2022