Steroid Receptor Status and the Epidemiology of Breast Cancer Nancy Kreiger, PhD, Will D. King, MSc, Lynn Rosenberg, E. Aileen Clarke, MB, MSc, Julie R. Palmer, ScD, and Samuel Shapiro, MB, FRCP(E)
ScD,
This case-control study examined risk factors for breast cancer according to tumor estrogen receptor (ER) status and progesterone receptor (PR) status. The data included 607 cuse putients and 12 14 control subjects matched by age and residence. Of528 case patients with steroid receptor information, 67% had ER-positive tumors and 55% had PR-posirive tumors. Odds ratios for ERpositive and ER-negative breast cancer were similar with respect to menopausal smtus, uge at menarche, history of cystic breast disease, and Quetelet Index. Family history of breast cancer was a stronger risk factor for ER-negative than for ER-positive breast cancer and the odds ratios for number of births were suggestive of a diff erent effect. While ER and PR status were highly correlated, there were some differences in their associations with risk factors. Odds ratios for PRpositive and PR-negative breast cancer differed f or number of births and were suggest&e of differences with respect to menopausal status, Quetelet Index, and family history of breast cuncer. These findings do not suggest diff erent causal pathways for ER-positive and ER-negative breast cancer. However, they do indicate that PR status may play a role in the etiology of breast cuncer. Ann Epidemiol 1991; J:513-523. KEY WORDS:
case-control
Breast neophms, study.
estrogen receptors, progesterone receptors, risk fucfors,
INTRODUCTION
Both tumor estrogen receptor (ER) status and progesterone receptor (PR) status are established indicators of prognosis in the clinical course of breast cancer. ER-positive (ER+) tumors and PR-positive (PR+) tumors are associated with improved response to hormonal therapy, a longer disease-free interval, and improved survival (l-5). To determine whether steroid receptor status is important in the etiology of breast cancer, investigators have examined the relationship between receptor status and known risk factors. ER status is reportedly associated with age and race, but the literature contains inconsistent reports on the association between ERs and menopausal status, age at menopause, age at menarche, number of births, body mass, history of cystic breast disease, and family history of breast cancer (6). A smaller amount of research is available on the subject of PRs and risk factors for breast cancer. The study reported here examined the relationship between steroid receptor status and known risk factors for breast cancer. Associations with known risk factors are presented in a case-control analysis of receptor-positive and receptor-negative tumors.
From the Division of Epidemiology and Statistics, Ontario Cancer Treatment Toronoto, Ontario, Canada (N.K., W.D.K., E.A.C), and Slone Epidemiology Health, Boston University School of Medicine, Brookline, MA (L.R., J.R.P., Address reprint requests to: WIII D. King, MSC, Diwsion of Epidemiology and Treatment and Research Foundation, 7 Overlea Boulevard, Toronto, Ontario, Received November 16, 1990; revised December 2 1, 1990. 0 1991 Elsevier Science Puhhshing Co., Inc.
and Research Foundation, Unit, School of Public S.S.). Statistics, Ontario Canc.er Canada M4H 1A8.
1047-27971911503.50
514
Kreiger et al. STEROID RECEF’TORS AND BREAST CANCER
AEP Vol. I, No. 6 November 1991: 513-523
METHODS
Eligible case patients were all women who had a histologically confirmed primary carcinoma of the breast; who were newly diagnosed between 1982 and 1986; who had no previous history of cancer (with the exception of nonmelanotic carcinoma of the skin); who were 20 to 69 years old at diagnosis; who were residents of metropolitan Toronto, Ontario, Canada; and who had been referred to the Princess Margaret Hospital, the major cancer treatment hospital in metropolitan Toronto. Populationbased control subjects were matched (2 : 1) to the case patients for age (by decade) and neighborhood of residence. Potential controls were selected from tax assessment rolls for residents maintained by the Ontario Ministry of Revenue. The collection of these records is mandated by provincial law, and the records are completely updated every 3 years. Interview questionnaires were completed by 630 case patients and 1214 control subjects, representing 79% of potential case patients and 65% of potential controls. Nonparticipation among the case patients was due to patient refusals (12%) and physician refusals (9%). A further 23 case patients were excluded because controls could be not found. For the remaining 607 case patients, 1874 eligible neighbors were contacted: 1214 (65%) were interviewed, and 660 (35%) refused. Records of 528 case patients were linked to the steroid receptor files of the Ontario Cancer Registry to obtain both ER and PR concentrations. The receptor assays were carried out using the dextran-coated charcoal method to determine the amount of receptor protein in the tumor (7). An assay result was judged acceptable if the assay was done within 6 weeks of diagnosis. If more than one assay was done within a l-day period, the higher result was taken. If more than one assay was done in the 6-week period after diagnosis, the one closest in time to the date of diagnosis was taken. Data were collected with the use of a standardized structured questionnaire adAll interviews were conducted in the ministered by trained female interviewers. subjects’ homes within 6 months of diagnosis. The questionnaire covered the following areas: sociodemographic factors; medical and reproductive histories; family history of breast cancer; and histories of caffeine, alcohol, and tobacco use. Study results pertaining to alcohol consumption and diazepam use in relation to risk of breast cancer have been reported elsewhere (8, 9). Factors selected in this analysis were those considered to be established risk factors for breast cancer and those potentially hormone-related: age, menopausal status, age at menarche, number of births, Quetelet Index, history of cystic breast disease, and family history of breast cancer. Subjects who reported having menstruated within 1 year prior to diagnosis were classified as premenopausal. From the subjects’ reported height and weight, the Quetelet Index, a measure of body mass, was calculated as weight divided by height squared (kg/m3). Body mass definitions, at age 20 and at the time of diagnosis, were determined on the basis of the distribution of the Quetelet Index values for the control group: The lowest 10% were classified as low; the upper lo%, as high; and the middle 80%, as normal. Family history of breast cancer was positive if the subject reported a history of breast cancer in either her mother or her sister(s). For both ERs and PRs, receptor positive was defined as 10 femtomoles or greater specific binding sites per milligram of protein (fmolimg); receptor negative was less than 10 fmolimg.
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. I, No. 6 November 1991: 513-523
515
Data analysis was divided into two parts. First, an analysis of breast cancer case patients (case-case comparison) was done to determine if tumor receptor status varied by age and menopausal status. Second, a matched case-control analysis was carried out to estimate the relative risk of breast cancer associated with suspected risk factors, according to the receptor status of the tumor. The odds ratio (OR) was used to estimate relative risk. Conditional logistic regression models containing a term representing the interaction between the factor under study and receptor status were used to obtain matched risk estimates according to tumor status and to test for differences .between these estimates (10). Risk estimates were adjusted for all potential confounders (main effects). The P value testing for a difference between the estimates for receptor-positive and -negative disease is provided. For both case-case and case-control comparisons, separate analyses were performed for ER and PR status.
RESULTS Receptor concentrations were available for 528 (87%) of the case patients. Table 1 displays the distributions of ER and PR concentrations; 67% of case patients were ER+, and 55% were PR+. A strong positive association was found between ER and PR status: 76% of tumors were concordant with respect to their ER and PR status.
Associations with Estrogen Receptor Status Figure 1 shows the percentage of ER+ tumors among case subgroups by age and menopausal status. Excluded are postmenopausal case patients aged 20 to 39 (n = 7) and one premenopausal case patient in the age group 60 to 69 because of small frequencies and the chance of misclassification. The proportion of ER+ tumors increased with age in premenopausal case patients, from 38% in those aged 20 to 39 to 80% in those aged 50 to 59. No effect of age was evident in postmenopausal case patients: Approximately 70% in each age group had ER+ tumors. Removing case patients younger than 60 who had a hysterectomy but retained one or both ovaries did not change these results. An effect of menopausal status independent of age could be assessed only among age groups in which there was some variation in menopausal status. In those aged 40 to 49 and 50 to 59, pre- and postmenopausal case patients had a similar percentage of
TABLE Receptor status Negative
1
Distribution
of cases by tumor receptor concentrations
Concentration (fmol/mg) O-4 5-9
Total receptor-negative Positive 10-19 20-49 50-99 100-199 200-399 400-599 600+ Total receptor-positive
Estrogen No. 141 36 177 58 115 72 52 33 15 6 351
receptor Percentage 26.7 6.8 33.5 11.0 21.8 13.6 9.8 6.3 2.8 1.1 66.5
Progesterone No. 160 76 236 57 91 48 54 28 8 6 292
receptor Percentage 30.3 14.4 44.7 10.8 17.2 9.1 10.2 5.3 1.5 1.1 55.3
516
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. 1, No. 6 November 1991: 513-523
8070+ f! s
3 Post-menopausal
6050.
403020lo0
I 20 to 39
I 40 to 49
I 50 to 59
1 60 to 69
Age group
FIGURE 1 Percent of tumors that ate estrogen receptor-positive (ER+) according to subgroups by age and menopausal status-premenopausal (n = 209) and postmenopausal (n = 311).
ER+ tumors. For instance, in case patients aged 50 to 59, 78% of tumors were ER+ in premenopausal case patients, and 72% of tumors were ER+ in postmenopausal case patients. Case patients and matched controls with complete information on all risk factors were included in the conditional regression analysis (501 case patients, 969 control subjects). Case and control distributions according to risk factors and receptor status for those included in the analysis are shown in Table 2. Table 3 presents adjusted ORs from the matched case-control analysis of risk factors according to ER status. Postmenopausal women and women who were older at menarche were at a nonsignificantly decreased risk for both ER+ and ER-negative (ER-) disease. Women who had given birth four or more times were at decreased risk of ER+ breast cancer and at elevated risk for ER- breast cancer, with ORs of .67 and 1.34, respectively. However, neither of these risk estimates was statistically different from 1.0, and the difference between them also was not statistically significant (P = .13). An analysis that included age at first birth (in place of number of live births) in the model, compared different categories of age at first birth and nulliparity to those giving birth before the age of 2 1. The relative risks for those whose first birth came after the age of 30 were protective and not significantly different from 1.0 for both ER+ and ER- breast cancer. Nulliparity slightly increased risk for ER+ breast cancer (OR, 1.20; 95% confidence interval (CI), .72-1.89) and slightly decreased the risk for ER- breast cancer (OR, .80; 95% CI, .34-l .69). This difference was not significant (P = .41). A family history of breast cancer elevated the risk for both ER+ and ER- breast cancer. However, this risk was greater for ER- (OR, 5.69) than for ER+ (OR, 1.75) breast cancer (test of difference, P = .Ol ). A history of cystic breast disease was associated with elevated risk for both ER+ and ER- breast cancer, ORs being 3.01 and 4.2 1, respectively.
Quetelet
Index (current)
Family history of breast cancer History of cystic breast disease Quetelet Index (age 20)
Age (y) at first hirth
No. of births
Age (y) at menarche
Menopausal status
20-39
40-49 50-59 60-69 PrePost511 12-14 E-15 0 1-3 24 220 21-30 231 No births No Yes No Yes Low Normal High LOW Normal High
34.4 29.9 34.1 45.9 20.4 65.6 14.1 24.9 59.3 15.9 14.1 50.0 11.1 24.9 87.1 12.9 85.6 14.4 15.3 71.6 13.2 12.9 68.6 18.6
8.4 27.2
ER+
and controls
Cases (n = 334)
(%) of cases
Age (Y)
Distribution
Level
2
Risk factor
TABLE
8.7 27.2 34.6 29.5 35.9 64.1 16.6 68.5 14.9 18.6 65.7 15.7 11.8 58.7 10.9 18.6 93.0 7.0 94.9 5.1 12.7 73.8 13.5 11.8 75.3 12.9
Controls (n = 644)
by risk factors
14.4 24.0 59.9 16.2 19.8 47.9 8.4 24.0 85.0 15.0 83.8 16.2 18.0 74.9 7.2 18.0 67.1 15.0
25.7 29.9 23.4 21.0 52.1 47.9 19.8 65.9
Cases (n = 167)
ER-
receptor
25.2 30.8 23.4 20.6 52.3 47.7 17.5 66.8 15.7 22.7 62.8 14.5 16.3 50.5 10.5 22.7 95.7 4.3 94.1 5.9 16.9 72.6 10.5 13.5 76.6 9.9
Controls (n = 325)
and by estrogen
12.3 30.7 32.5 24.5 44.0 56.0 22.4 64.3 13.4 25.6 61.4 13.0 15.9 47.3 11.2 25.6 87.0 13.0 85.2 14.8 14.8 71.1 14.1 13.4 68.6 18.0
PR+
12.4 30.7 33.0 24.0 42.7 57.3 18.0 68.2 13.9 18.5 65.9 15.5 13.1 57.9 10.5 18.5 93.3 6.7 94.0 6.0 12.5 73.6 13.9 11.8 74.9 13.3
Controls (n = 534)
and urogesterone
Cases (n = 277)
(ER) status
74.6 7.6 16.1 67.4 16.5
16.5 25.0 28.6 29.9 35.3 64.7 17.4 67.4 15.2 23.2 57.1 19.6 16.1 51.8 8.9 23.2 85.7 14.3 84.8 15.2 17.9
PR-
4.6 16.1 73.1 10.8 13.1 76.8 10.1
16.6 25.5 28.3 29.7 39.8 60.2 15.6 67.6 16.8 21.8 63.2 14.9 13.6 53.6 11.0 21.8 94.7 5.3 95.4
Controls (n = 435)
(PR) status”
Cases (n = 224)
rece!xor
518
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. 1, No. 6 November 1991: 513-523
TABLE 3 Adjusted” odds ratios (OR) and 95% confidence intervals receptor-positive (ER+) and estrogen receptor-negative (ER-) breast
ER+
Risk factor Menopausal
stat&
Age (y) at menarche
No. of births
Age (y) at first birth
Family history of breast cancer History of cystic breast disease Quetelet
Index (age 20)
Quetelet Index (current)d
ER-
Level
OR
Pre-
1.00
1.00
Post-
.88 .55-1.41 1.00 .76 .52-1.10 .77 .47-1.25 1.00 .65 .45- .93 .67 .41-1.10 1.00 .71 .45-1.10 .87 .49-1.55 1.20 .72-1.98 1.00 1.75 1.11-2.76 1.00 3.01 1.84-4.93 1.24 .82-1.87 1.00 1.06 .69-1.61 1.17 .76-1.82 1.00 1.73 1.17-2.56
.91 .53- 1.57 1.00 .91 .53- 1.58 .71 .34- 1.49 1.00 .88 .49- 1.55 1.38 .64- 2.96 1.00 .72 .38- 1.37 .55 .23- 1.33 .80 .34- 1.69 1.00 5.69 2.67-12.16 1.00 4.21 2.00- 8.87 1.04 .60- 1.78 1.00 .65 .31- 1.36 1.67 .93- 3.05 1.00 2.03 1.06- 3.87
511 12-14 z-15 0 1-3 24 520 21-30 231 No births No Yes No Yes Low Normal High Low Normal High
95% CI
(Cl) for estrogen cancers
OR
95% CI
Test
for difference
between groups P .77 .57 .87 .38 .I3 .89 .44 .41 .Ol
.46 .61 .26 .33 .68
’Estimates are adjusted in a conditmnal loglstx regressionmodel for age, menopausal status, age-menopausal status mteraction, age at menarche, number of brths, famlly history, cystic breast disease, and Quetelet Index (age 20). h Postmenopause estimate 1s a point esnmate for the odds ratm at age ot d~agnosn of 50 years. ’Age at first birth replaces no. of births in the model. ’ Quetelet Index (current) replaces Quetelet Index (age 20) m the model.
High Quetelet Index at age 20 was associated with a small elevation in the risk for ER+, but was slightly protective of ER- breast cancer. The difference between these estimates was not statistically significant (P = .26). High Quetelet Index at diagnosis elevated the risk for both ER+ and ER- disease.
Associations
with Progesterone Receptor Status
Figure 2 shows the percentage of PR+ tumors among case subgroups by age and menopausal status. As with ER status, the proportion of PR+ tumors increased with age in premenopausal case patients, from 45% in those aged 20 to 39 years, to 84% in those aged 50 to 59. No effect of age was evident in postmenopausal case patients. In age groups where there was some variation in menopausal status, premenopausal case patients accounted for a greater proportion of PR+ tumors than did postmenopausal case patients. In case patients aged 50 to 59 years, 84% of tumors were PR’ in premenopausal case patients compared to only 50% in postmenopausal case patients. Table 4 presents the ORs from the matched case-control analysis of risk factors according to case PR status. A nonsignificant difference in risk for PR+ and PRnegative (PR-) tumors was found for menopausal status (P = .ll). Postmenopausal women were found to be at decreased risk for PR’ tumors (OR, .68) and elevated risk
AEP Vol. November
519
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
6 1991: 513-523
1, No.
3020100
I
I 20 to 39
I
I
50 to 59
40 to 49
60 to 69
Age group FIGURE 2 Percent of tumors that are progesterone subgroups by age and menopausal status-premenopausal 311).
TABLE
4
receptor-positive (PR’) according to (n = 209) and postmenopausal (n =
Adjusted” odds ratios (OR) and 95% confidence (PR+) and progesterone receptor-negative
receptor-positive
PR+
intervals (CI) for progesterone (PR-) breast cancers PR-
Risk factor
Level
OR
Menopausal statusb
PreI’ost511 12-14 215 0 l-3 24 520 21-30 231 No births No Yes No Yes LOW Normal High LOW Normal High
1.00 1.00 .68 .41-1.13 1.76 .80-3.87 1.00 1.00 .71 .48-1.05 .93 .56-1.54 .77 .45-1.32 .il .37-1.35 1.00 1.00 .61 .41- .92 .93 .57-1.51 .48 .27- .84 1.64 .86-3.09 1.00 1.00 .65 .40-1.05 .44-1.32 .76 .86 .45-1.64 .57 .28-1.18 1.25 .73-2.14 .75 .39-1.44 1.00 1.00 1.86 1.12-3.09 4.05 2.21-7.45 1.00 1.00 2.51 1.48-4.25 4.88 2.54-9.40 1.21 .77-1.90 1.14 .70-1.85 1.00 1.00 1.10 .69-1.74 .65 .35-1.21 1.24 .77-2.01 1.35 .80-2.25 1.00 1.00 1.68 1.09-2.57 2.10 1.22-3.64
Age (y) at menarche
No. of births
Age (y) at first birth‘
Family history of breast cancer History of cystic breast disease Quetelet Index (age 20)
Quetelet Index (current)d
’ Estimatesare adjusted in mteractmn, age h Postmenopause
at menarche. e&mate
a conditional number IS a point
log~st~
of births, esnmate
‘Age at first birth replacesno. of buths in d Quetelet Index (current) replacesQuetelet
regresswn
famdy for the the
95%
CI
model
OR
for age,
menopausal
history. cystic breast disease, udds ran” at age of dqnosis
model.
Index
95%
(age
20)
in the
model.
CI
staws,
and Quetelct of 50 years.
Test for difference between groups P
.I1 .41 .83 .19 .Ol .68 .42 .24 .05 .12 .87 .18 .82 .52 age-menopausal Index
(age
smtus 20).
520
Kreiger et al. STEROID RECEPTORS
for PRcant
tumors
reduced There
AEP Vol. 1, No. 6 November 1991: 513-523
AND BREAST CANCER
(OR,
1.76).
risk for both
A later age at menarche PR+ and PR-
were differences
.Ol).
Women
(OR,
.48) and at higher
breast
in risk according
who gave birth
was associated
to PR status
four or more times
risk for PR-
tumors
for number
women
decreased
were
risk for PR-
significantly
different
significant
at elevated
breast
(OR,
1.64).
risk for PR+
cancer
(OR,
.75).
of births
(P =
were at lower risk for PR+ tumors In an analysis
age at first birth (in place of number of live births) in the model, birth (>30) was nonsignificantly protective for both PR’ and Nulliparous
with a nonsignifi-
cancer.
breast
Neither
from 1.0, and the difference
cancer
(OR,
of these
between
that
included
late age at first live PR- breast cancer.
them
1.25)
and
risk estimates
was
was not statistically
(I’ = .24).
A family
history
of breast
cancer
cancer
(OR,
1.86),
but was associated
cancer
(OR,
4.05).
The difference
cal significance PR+ (OR,
(P = .05).
2.51)
High Quetelet had a slight
(OR,
effect
these
breast
for PR-
the
risk for PR+ breast
elevated
estimates
risk for PR-
was of borderline elevated
breast statisti-
the risk for both
cancer.
elevated
breast
(I’ = .18). High Quetelet
cally significant
elevated
of cystic breast disease
4.88)
Index at age 20 slightly
protective
PR+ and PR-
between
A history
and PR-
moderately
with substantially
the risk for PR+ breast cancer
cancer.
This difference
Index at diagnosis
and
was not statisti-
elevated
the risk for both
tumors.
DISCUSSION Risk factors status.
were examined
An attempt
cation,
which
might
different
subgroups
studies
of risk factors
Both (11-14). tumors status
in women
with breast cancer,
was made to identify
risk groups,
be differentially be identifiable, within
associated then
breast
with
it would
cancer
stratified
within
by steroid
the larger disease
various
risk factors.
be appropriate
recegtor classifiShould
to conduct
future
subtype.
the ER and PR distributions
are highly
skewed,
in our data and in others
Good
has been
reported
on
it should
be noted
agreement,
as positive are highly
however,
or negative correlated,
(14). and
Additionally, that
similar
associated with the two measures. A strong positive association between in this analysis
concurs
with previous
on is the lack of independent after adjustment A number
increasing
findings
association
for age (10, of studies
risk factors
the
might
ER status
of
that ER and PR
be expected
age and ER+ tumor
( 12, 13, 15-19).
between
categorization
to be
status found
Also generally and menopausal
agreed status,
11, 14, 16).
that black women have a lower prevalence of ER+ too few black case tumors than do white women (13, 17, 20). 0 ur study contained patients to analyze this factor; however, removing these case patients did not alter study results. In our case-control higher
risk for ER-
reported
analysis, than
family history
for ER+ breast
of breast cancer McTiernan
cancer.
was associated with a and colleagues (11)
reported a like result, while other case-control studies reported an elevation in risk that is similar for both ER+ and ER- breast cancer ( 17, 18, 2 1). Our results concerning number of births are in agreement with several casecontrol studies that reported breast cancer but increased literature
is somewhat
a protective effect of a greater number of births for ER+ risk for ER- breast cancer (11, 15, 21). However, the
inconsistent
(17,
18).
521
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. I, No. 6 November 1991: 513-523
In general, menopausal cerning
our findings
status
the relationship
strong
confirm
with ER status.
associations
of other
previous
of the relationships
reports
It is difficult risk factors
to make substantial
to ER status
in our data and the inconsistent
of age and
conclusions
because
of both
results reported
con-
the lack of
by other
investiga-
tors. Differences status,
race,
tency
in the various
of results
and
cutoff
reported
points
varied
between
cutoff
chosen
for the studies.
results
analysis,
our case-control tumors. With breast the
tumor
consistent
basis.
status
status,
of number
Quetelet
of menopausal many
our data suggest
for breast
(2) Relative
accuracy
with
may identify
other
differences
have
examined
and precision
distinct
of breast
the
those
cancer
PRs in their
analysis.
tumors
on a
for the PR assay then (11).
(2, 5, 22, 23),
cancer
in
between
for case patients
are poorer
research
PR-
found
relationship
with ER concentration
breast
In our
for PR+ and PR-
than
few have included
correlated
study.
with age and in
in risk for PR+ and
(1) It is not available
epidemiologic
etiologically
differed
and family history
but more pronounced,
cancer,
because:
under
tumors.
status,
studies
the (15).
ER and PR classifica-
However,
of births
Index,
for PR’ and PR-
(3) PR is highly
along
tumor
was found to increase
the effects
have
to date,
(21) to 30 fmol/mg
in the two analyses.
included
the ER assay (7).
between
for study,
status,
conducted
of PR+ tumors
ERs and risk factors
our data,
of concordance
receptor
studies
the percentage
Although
is seldom
negative
from 3 fmolimg
were in the same direction,
PR status
and
inconsis-
chosen
by the risk factors
of differences
ER analysis.
parameters
breast cancers
the exception
cancer
ranged
to age, menopausal
to the general
distinguishes
menopausal
were suggestive
In addition,
in case-control
ER+ tumors
analysis,
Also,
study according
of positive
were anticipated
that PR status better case-case
definition
For example,
of the high degree
similar
under
and time of test may contribute
in the literature.
to define
Because tions,
populations
stage of disease,
However,
suggest
to a greater
that
extent
PR than
does ER status. In our study,
neighborhood
by socioeconomic factors
under
status.
study,
not have introduced status
and,
hence,
cancer.
Although
pronounced since
resulting
spurious
associations,
there
differences are needed
ER and PR status.
that
to suggest were
or differences
differences different
between
pathways
according
This finding
and PRs are biologically
our results
and to clarify the relationship
important
anisms.
It is, of course,
possible
to investigate are indicative
that the relatively
further of distinct
small observed
more
is unexpected
correlated
in our data actually
breast
PR analyses,
to confirm
It is particularly
receptor
to ER status
are highly
of association
it would
to steroid
for ER+ and ER-
our ER and
to PR status were found.
patterns
on some
However,
detected.
in risk factors causal
confounding
in overmatching
according
the associations
similarities
according
were used to minimize
in the loss of associations.
not invalidate
we found enough
ER and PR status
ERs. Studies
subjects
may have resulted
possibly should
In conclusion, were not strong
controls
This approach
dependeiit
whether
or not the
pathogenic deviations
on
between mechin risk are
due to chance, but it may be that different subtypes of breast cancer are determined, in part, by different constellations of risk factors. This research was supported in part by Ontario
Ministry
of Health
research
grant 01617; the Ontario
Research Foundation, Toronto, Ontario; search Foundation, Baltimore, MD.
and the Alcoholic
Cancer Beverage
Treatment Medical
and Re-
The authors are grateful for the collaboration of the attending physicians at Princess Margaret Hospital, Sunnybrook MedIcal Centre, Toronto General Hospital, and Mt. Sinai
522
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. 1, No. 6 November 1991: 513-523
Hospital. We are also grateful to the interviewers who collected the data-Anne Allen, Rosemary Chepa, Suzanne Gale, Kathleen Gillespie, Lindsay Hall, Linda Harrel, Virginia Hunter, Beverly Linden, Muriel Relton, Shirley Rothery, Irene Servos, Ann Skene-Melvin, Sybil Tinker, Kathy Carroll, Binny Heaton, Helene Katz, Maxine Newton, Barbara Walker, and Linda West; Verna Cundari and Phyllis Disenhouse who coordinated the study; and Leonard Gaetano and Marguerite Angeloni who managed the data.
REFERENCES 1. Hawkins RA, Roberts MM, Forrest APM, Oestrogen receptors and breast cancer: Current status, Br J Surg. 1980;67:153-69. 2. Fisher B, Redmond C, Brown A, et al. Influence of tumor estrogen and progesterone receptor levels on the response to tamoxifen and chemotherapy in primary breast cancer, J Clin Oncol. 1983;1:227-41. 3. Wittliff JL. Steroid hormone receptors in breast cancer, Cancer. 1984;53:630-43. 4. Ruder AM, Lubin F, Wax Y, Geier A, Alfundary E, Chetrit A. Estrogen and progesterone receptors in breast cancer patients: Epidemiologic characteristics and survival differences, Cancer. 1989;64:196-202. 5. Clark GM, McGuire WL, Hubay CA, Pearson OH, Marshall JS. Progesterone receptors as a prognostic factor in stage II breast cancer, N Engl J Med. 1983;309:1343-7. 6. Stanford JL, Szklo M, Brinton LA. Estrogen receptors and breast cancer, Epidemiol Rev. 1986;8:42-59. 7. Ryan ED, Clark AF, Mobbs BG, Ooi TC, Sutherland DA, Tustanoff ER. Interlaboratory quality control of estrogen and progesterone assays in breast cancer tissue using lyophilised cytosols, Clin Biochem. 1985;18:20-6. 8. Rosenberg L, Palmer JR, Miller DR, Clarke EA, Shapiro S. A case-control study of alcoholic beverage consumption and breast cancer, Am J Epidemiol. 1990;131:6-14. 9. Kaufman DW, Werler MM, Palmer JR, et al. Diazepam use in relation to breast cancer: Results from two case-control studies, Am J Epidemiol. 1990;131:483-90. 10. Dubin N, Pastemack BS. Risk assessment for case-control analysis by polychotomous logistic regression, Am J Epidemiol. 1986;123:1101-17. 11. McTiernan A, Thomas DB, Johnson LK, Roseman D. Risk factors for estrogen receptor-rich and estrogen receptor-poor breast cancers, J Nat1 Cancer Inst. 1986;77:849-54. 12. Ballard-Barbash R, Griffin MR, Fisher LD, Covalciuc MA, Jiang N. Estrogen receptors in breast cancer: Association with epidemiologic risk factors, Am J Epidemiol. 1986;124:77-84. 13. Hulka BS, Chambless LE, Wilkinson WE, Deubner DC, McCarty KS Sr, McCarty KS Jr. Hormonal and personal effects on estrogen receptors in breast cancer, Am J Epidemiol. 1984;119:692-704. 14. Oxley DK, Haven GT, Wittliff JL, Gilbo D. Precision in estrogen and progesterone receptor assays, Am J Clin Pathol. 1982;78(suppl):587-96. 15. Hildreth NC, Kelsey JL, Eisenfeld AJ, LiVolsi VA, Holford TR, Fischer DB. Differences in breast cancer risk factors according to the estrogen receptor level of the tumor, J Nat1 Cancer Inst. 1983;70:1027-31. 16. Elwood JM, Godolphin W. Oestrogen receptors in breast tumours: Associations with age, menopausal status and epidemiological and clinical features in 735 patients, Br J Cancer. 1980;42:635-44. 17. Stanford JL, Szklo M, Boring C, et al. A case-control study of breast cancer stratified by estrogen receptor status, Am J Epidemiol. 1987;125:184-94. 18. Cooper JA, Rohan TE, Cant EL, Horsfall DJ, Tilley WD. Risk factors for breast cancer by oestrogen receptor status: A population-based case-control study, Br J Cancer. 1989;59:119-25. 19. Mckeown-Eyssen GE, Rodgers-Melamed 1, Clarke EA. Estrogen receptor status of breast cancer in Ontario, Can Med Assoc J. 1985;133:997-1000.
AEP Vol. 1, No. 6 November 1991: 513-523
STEROID RECEPTORS
Kreiger et al. AND BREAST CANCER
523
20. Lesser ML, Rosen PP, Senie RT, Duthie K, Botet C, Schwartz MK. Estrogen and progesterone receptors in breast carcinoma: Correlations with epidemiology and pathology, Cancer. 1981;48:229-309. 21. Hislop TG, Coldman AJ, Elwood JM, Skippen DH, Kan L. Relationship between risk factors for breast cancer and hormonal status, Int J Epidemiol. 1986;15:469-76. 22. Bertuzzi A, Vezzoni P, Ronchi E. Prognostic importance of progesterone receptors alone or in combination with estrogen receptors in node-negative breast carcinoma (abstract), Proc Am Assoc Cancer Res Am Sot Clin Oncol. 1981;22:447. 23. Ellis LM, Wittliff JL, Bryant MS, Hogancamp WE, Sitren HS, Bland KI. Correlation of estrogen, progesterone, and androgen receptors in breast cancer, Am ] Surg. 1989;157:577-80.
ScD,
This case-control study examined risk factors for breast cancer according to tumor estrogen receptor (ER) status and progesterone receptor (PR) status. The data included 607 cuse putients and 12 14 control subjects matched by age and residence. Of528 case patients with steroid receptor information, 67% had ER-positive tumors and 55% had PR-posirive tumors. Odds ratios for ERpositive and ER-negative breast cancer were similar with respect to menopausal smtus, uge at menarche, history of cystic breast disease, and Quetelet Index. Family history of breast cancer was a stronger risk factor for ER-negative than for ER-positive breast cancer and the odds ratios for number of births were suggestive of a diff erent effect. While ER and PR status were highly correlated, there were some differences in their associations with risk factors. Odds ratios for PRpositive and PR-negative breast cancer differed f or number of births and were suggest&e of differences with respect to menopausal status, Quetelet Index, and family history of breast cuncer. These findings do not suggest diff erent causal pathways for ER-positive and ER-negative breast cancer. However, they do indicate that PR status may play a role in the etiology of breast cuncer. Ann Epidemiol 1991; J:513-523. KEY WORDS:
case-control
Breast neophms, study.
estrogen receptors, progesterone receptors, risk fucfors,
INTRODUCTION
Both tumor estrogen receptor (ER) status and progesterone receptor (PR) status are established indicators of prognosis in the clinical course of breast cancer. ER-positive (ER+) tumors and PR-positive (PR+) tumors are associated with improved response to hormonal therapy, a longer disease-free interval, and improved survival (l-5). To determine whether steroid receptor status is important in the etiology of breast cancer, investigators have examined the relationship between receptor status and known risk factors. ER status is reportedly associated with age and race, but the literature contains inconsistent reports on the association between ERs and menopausal status, age at menopause, age at menarche, number of births, body mass, history of cystic breast disease, and family history of breast cancer (6). A smaller amount of research is available on the subject of PRs and risk factors for breast cancer. The study reported here examined the relationship between steroid receptor status and known risk factors for breast cancer. Associations with known risk factors are presented in a case-control analysis of receptor-positive and receptor-negative tumors.
From the Division of Epidemiology and Statistics, Ontario Cancer Treatment Toronoto, Ontario, Canada (N.K., W.D.K., E.A.C), and Slone Epidemiology Health, Boston University School of Medicine, Brookline, MA (L.R., J.R.P., Address reprint requests to: WIII D. King, MSC, Diwsion of Epidemiology and Treatment and Research Foundation, 7 Overlea Boulevard, Toronto, Ontario, Received November 16, 1990; revised December 2 1, 1990. 0 1991 Elsevier Science Puhhshing Co., Inc.
and Research Foundation, Unit, School of Public S.S.). Statistics, Ontario Canc.er Canada M4H 1A8.
1047-27971911503.50
514
Kreiger et al. STEROID RECEF’TORS AND BREAST CANCER
AEP Vol. I, No. 6 November 1991: 513-523
METHODS
Eligible case patients were all women who had a histologically confirmed primary carcinoma of the breast; who were newly diagnosed between 1982 and 1986; who had no previous history of cancer (with the exception of nonmelanotic carcinoma of the skin); who were 20 to 69 years old at diagnosis; who were residents of metropolitan Toronto, Ontario, Canada; and who had been referred to the Princess Margaret Hospital, the major cancer treatment hospital in metropolitan Toronto. Populationbased control subjects were matched (2 : 1) to the case patients for age (by decade) and neighborhood of residence. Potential controls were selected from tax assessment rolls for residents maintained by the Ontario Ministry of Revenue. The collection of these records is mandated by provincial law, and the records are completely updated every 3 years. Interview questionnaires were completed by 630 case patients and 1214 control subjects, representing 79% of potential case patients and 65% of potential controls. Nonparticipation among the case patients was due to patient refusals (12%) and physician refusals (9%). A further 23 case patients were excluded because controls could be not found. For the remaining 607 case patients, 1874 eligible neighbors were contacted: 1214 (65%) were interviewed, and 660 (35%) refused. Records of 528 case patients were linked to the steroid receptor files of the Ontario Cancer Registry to obtain both ER and PR concentrations. The receptor assays were carried out using the dextran-coated charcoal method to determine the amount of receptor protein in the tumor (7). An assay result was judged acceptable if the assay was done within 6 weeks of diagnosis. If more than one assay was done within a l-day period, the higher result was taken. If more than one assay was done in the 6-week period after diagnosis, the one closest in time to the date of diagnosis was taken. Data were collected with the use of a standardized structured questionnaire adAll interviews were conducted in the ministered by trained female interviewers. subjects’ homes within 6 months of diagnosis. The questionnaire covered the following areas: sociodemographic factors; medical and reproductive histories; family history of breast cancer; and histories of caffeine, alcohol, and tobacco use. Study results pertaining to alcohol consumption and diazepam use in relation to risk of breast cancer have been reported elsewhere (8, 9). Factors selected in this analysis were those considered to be established risk factors for breast cancer and those potentially hormone-related: age, menopausal status, age at menarche, number of births, Quetelet Index, history of cystic breast disease, and family history of breast cancer. Subjects who reported having menstruated within 1 year prior to diagnosis were classified as premenopausal. From the subjects’ reported height and weight, the Quetelet Index, a measure of body mass, was calculated as weight divided by height squared (kg/m3). Body mass definitions, at age 20 and at the time of diagnosis, were determined on the basis of the distribution of the Quetelet Index values for the control group: The lowest 10% were classified as low; the upper lo%, as high; and the middle 80%, as normal. Family history of breast cancer was positive if the subject reported a history of breast cancer in either her mother or her sister(s). For both ERs and PRs, receptor positive was defined as 10 femtomoles or greater specific binding sites per milligram of protein (fmolimg); receptor negative was less than 10 fmolimg.
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. I, No. 6 November 1991: 513-523
515
Data analysis was divided into two parts. First, an analysis of breast cancer case patients (case-case comparison) was done to determine if tumor receptor status varied by age and menopausal status. Second, a matched case-control analysis was carried out to estimate the relative risk of breast cancer associated with suspected risk factors, according to the receptor status of the tumor. The odds ratio (OR) was used to estimate relative risk. Conditional logistic regression models containing a term representing the interaction between the factor under study and receptor status were used to obtain matched risk estimates according to tumor status and to test for differences .between these estimates (10). Risk estimates were adjusted for all potential confounders (main effects). The P value testing for a difference between the estimates for receptor-positive and -negative disease is provided. For both case-case and case-control comparisons, separate analyses were performed for ER and PR status.
RESULTS Receptor concentrations were available for 528 (87%) of the case patients. Table 1 displays the distributions of ER and PR concentrations; 67% of case patients were ER+, and 55% were PR+. A strong positive association was found between ER and PR status: 76% of tumors were concordant with respect to their ER and PR status.
Associations with Estrogen Receptor Status Figure 1 shows the percentage of ER+ tumors among case subgroups by age and menopausal status. Excluded are postmenopausal case patients aged 20 to 39 (n = 7) and one premenopausal case patient in the age group 60 to 69 because of small frequencies and the chance of misclassification. The proportion of ER+ tumors increased with age in premenopausal case patients, from 38% in those aged 20 to 39 to 80% in those aged 50 to 59. No effect of age was evident in postmenopausal case patients: Approximately 70% in each age group had ER+ tumors. Removing case patients younger than 60 who had a hysterectomy but retained one or both ovaries did not change these results. An effect of menopausal status independent of age could be assessed only among age groups in which there was some variation in menopausal status. In those aged 40 to 49 and 50 to 59, pre- and postmenopausal case patients had a similar percentage of
TABLE Receptor status Negative
1
Distribution
of cases by tumor receptor concentrations
Concentration (fmol/mg) O-4 5-9
Total receptor-negative Positive 10-19 20-49 50-99 100-199 200-399 400-599 600+ Total receptor-positive
Estrogen No. 141 36 177 58 115 72 52 33 15 6 351
receptor Percentage 26.7 6.8 33.5 11.0 21.8 13.6 9.8 6.3 2.8 1.1 66.5
Progesterone No. 160 76 236 57 91 48 54 28 8 6 292
receptor Percentage 30.3 14.4 44.7 10.8 17.2 9.1 10.2 5.3 1.5 1.1 55.3
516
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. 1, No. 6 November 1991: 513-523
8070+ f! s
3 Post-menopausal
6050.
403020lo0
I 20 to 39
I 40 to 49
I 50 to 59
1 60 to 69
Age group
FIGURE 1 Percent of tumors that ate estrogen receptor-positive (ER+) according to subgroups by age and menopausal status-premenopausal (n = 209) and postmenopausal (n = 311).
ER+ tumors. For instance, in case patients aged 50 to 59, 78% of tumors were ER+ in premenopausal case patients, and 72% of tumors were ER+ in postmenopausal case patients. Case patients and matched controls with complete information on all risk factors were included in the conditional regression analysis (501 case patients, 969 control subjects). Case and control distributions according to risk factors and receptor status for those included in the analysis are shown in Table 2. Table 3 presents adjusted ORs from the matched case-control analysis of risk factors according to ER status. Postmenopausal women and women who were older at menarche were at a nonsignificantly decreased risk for both ER+ and ER-negative (ER-) disease. Women who had given birth four or more times were at decreased risk of ER+ breast cancer and at elevated risk for ER- breast cancer, with ORs of .67 and 1.34, respectively. However, neither of these risk estimates was statistically different from 1.0, and the difference between them also was not statistically significant (P = .13). An analysis that included age at first birth (in place of number of live births) in the model, compared different categories of age at first birth and nulliparity to those giving birth before the age of 2 1. The relative risks for those whose first birth came after the age of 30 were protective and not significantly different from 1.0 for both ER+ and ER- breast cancer. Nulliparity slightly increased risk for ER+ breast cancer (OR, 1.20; 95% confidence interval (CI), .72-1.89) and slightly decreased the risk for ER- breast cancer (OR, .80; 95% CI, .34-l .69). This difference was not significant (P = .41). A family history of breast cancer elevated the risk for both ER+ and ER- breast cancer. However, this risk was greater for ER- (OR, 5.69) than for ER+ (OR, 1.75) breast cancer (test of difference, P = .Ol ). A history of cystic breast disease was associated with elevated risk for both ER+ and ER- breast cancer, ORs being 3.01 and 4.2 1, respectively.
Quetelet
Index (current)
Family history of breast cancer History of cystic breast disease Quetelet Index (age 20)
Age (y) at first hirth
No. of births
Age (y) at menarche
Menopausal status
20-39
40-49 50-59 60-69 PrePost511 12-14 E-15 0 1-3 24 220 21-30 231 No births No Yes No Yes Low Normal High LOW Normal High
34.4 29.9 34.1 45.9 20.4 65.6 14.1 24.9 59.3 15.9 14.1 50.0 11.1 24.9 87.1 12.9 85.6 14.4 15.3 71.6 13.2 12.9 68.6 18.6
8.4 27.2
ER+
and controls
Cases (n = 334)
(%) of cases
Age (Y)
Distribution
Level
2
Risk factor
TABLE
8.7 27.2 34.6 29.5 35.9 64.1 16.6 68.5 14.9 18.6 65.7 15.7 11.8 58.7 10.9 18.6 93.0 7.0 94.9 5.1 12.7 73.8 13.5 11.8 75.3 12.9
Controls (n = 644)
by risk factors
14.4 24.0 59.9 16.2 19.8 47.9 8.4 24.0 85.0 15.0 83.8 16.2 18.0 74.9 7.2 18.0 67.1 15.0
25.7 29.9 23.4 21.0 52.1 47.9 19.8 65.9
Cases (n = 167)
ER-
receptor
25.2 30.8 23.4 20.6 52.3 47.7 17.5 66.8 15.7 22.7 62.8 14.5 16.3 50.5 10.5 22.7 95.7 4.3 94.1 5.9 16.9 72.6 10.5 13.5 76.6 9.9
Controls (n = 325)
and by estrogen
12.3 30.7 32.5 24.5 44.0 56.0 22.4 64.3 13.4 25.6 61.4 13.0 15.9 47.3 11.2 25.6 87.0 13.0 85.2 14.8 14.8 71.1 14.1 13.4 68.6 18.0
PR+
12.4 30.7 33.0 24.0 42.7 57.3 18.0 68.2 13.9 18.5 65.9 15.5 13.1 57.9 10.5 18.5 93.3 6.7 94.0 6.0 12.5 73.6 13.9 11.8 74.9 13.3
Controls (n = 534)
and urogesterone
Cases (n = 277)
(ER) status
74.6 7.6 16.1 67.4 16.5
16.5 25.0 28.6 29.9 35.3 64.7 17.4 67.4 15.2 23.2 57.1 19.6 16.1 51.8 8.9 23.2 85.7 14.3 84.8 15.2 17.9
PR-
4.6 16.1 73.1 10.8 13.1 76.8 10.1
16.6 25.5 28.3 29.7 39.8 60.2 15.6 67.6 16.8 21.8 63.2 14.9 13.6 53.6 11.0 21.8 94.7 5.3 95.4
Controls (n = 435)
(PR) status”
Cases (n = 224)
rece!xor
518
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. 1, No. 6 November 1991: 513-523
TABLE 3 Adjusted” odds ratios (OR) and 95% confidence intervals receptor-positive (ER+) and estrogen receptor-negative (ER-) breast
ER+
Risk factor Menopausal
stat&
Age (y) at menarche
No. of births
Age (y) at first birth
Family history of breast cancer History of cystic breast disease Quetelet
Index (age 20)
Quetelet Index (current)d
ER-
Level
OR
Pre-
1.00
1.00
Post-
.88 .55-1.41 1.00 .76 .52-1.10 .77 .47-1.25 1.00 .65 .45- .93 .67 .41-1.10 1.00 .71 .45-1.10 .87 .49-1.55 1.20 .72-1.98 1.00 1.75 1.11-2.76 1.00 3.01 1.84-4.93 1.24 .82-1.87 1.00 1.06 .69-1.61 1.17 .76-1.82 1.00 1.73 1.17-2.56
.91 .53- 1.57 1.00 .91 .53- 1.58 .71 .34- 1.49 1.00 .88 .49- 1.55 1.38 .64- 2.96 1.00 .72 .38- 1.37 .55 .23- 1.33 .80 .34- 1.69 1.00 5.69 2.67-12.16 1.00 4.21 2.00- 8.87 1.04 .60- 1.78 1.00 .65 .31- 1.36 1.67 .93- 3.05 1.00 2.03 1.06- 3.87
511 12-14 z-15 0 1-3 24 520 21-30 231 No births No Yes No Yes Low Normal High Low Normal High
95% CI
(Cl) for estrogen cancers
OR
95% CI
Test
for difference
between groups P .77 .57 .87 .38 .I3 .89 .44 .41 .Ol
.46 .61 .26 .33 .68
’Estimates are adjusted in a conditmnal loglstx regressionmodel for age, menopausal status, age-menopausal status mteraction, age at menarche, number of brths, famlly history, cystic breast disease, and Quetelet Index (age 20). h Postmenopause estimate 1s a point esnmate for the odds ratm at age ot d~agnosn of 50 years. ’Age at first birth replaces no. of births in the model. ’ Quetelet Index (current) replaces Quetelet Index (age 20) m the model.
High Quetelet Index at age 20 was associated with a small elevation in the risk for ER+, but was slightly protective of ER- breast cancer. The difference between these estimates was not statistically significant (P = .26). High Quetelet Index at diagnosis elevated the risk for both ER+ and ER- disease.
Associations
with Progesterone Receptor Status
Figure 2 shows the percentage of PR+ tumors among case subgroups by age and menopausal status. As with ER status, the proportion of PR+ tumors increased with age in premenopausal case patients, from 45% in those aged 20 to 39 years, to 84% in those aged 50 to 59. No effect of age was evident in postmenopausal case patients. In age groups where there was some variation in menopausal status, premenopausal case patients accounted for a greater proportion of PR+ tumors than did postmenopausal case patients. In case patients aged 50 to 59 years, 84% of tumors were PR’ in premenopausal case patients compared to only 50% in postmenopausal case patients. Table 4 presents the ORs from the matched case-control analysis of risk factors according to case PR status. A nonsignificant difference in risk for PR+ and PRnegative (PR-) tumors was found for menopausal status (P = .ll). Postmenopausal women were found to be at decreased risk for PR’ tumors (OR, .68) and elevated risk
AEP Vol. November
519
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
6 1991: 513-523
1, No.
3020100
I
I 20 to 39
I
I
50 to 59
40 to 49
60 to 69
Age group FIGURE 2 Percent of tumors that are progesterone subgroups by age and menopausal status-premenopausal 311).
TABLE
4
receptor-positive (PR’) according to (n = 209) and postmenopausal (n =
Adjusted” odds ratios (OR) and 95% confidence (PR+) and progesterone receptor-negative
receptor-positive
PR+
intervals (CI) for progesterone (PR-) breast cancers PR-
Risk factor
Level
OR
Menopausal statusb
PreI’ost511 12-14 215 0 l-3 24 520 21-30 231 No births No Yes No Yes LOW Normal High LOW Normal High
1.00 1.00 .68 .41-1.13 1.76 .80-3.87 1.00 1.00 .71 .48-1.05 .93 .56-1.54 .77 .45-1.32 .il .37-1.35 1.00 1.00 .61 .41- .92 .93 .57-1.51 .48 .27- .84 1.64 .86-3.09 1.00 1.00 .65 .40-1.05 .44-1.32 .76 .86 .45-1.64 .57 .28-1.18 1.25 .73-2.14 .75 .39-1.44 1.00 1.00 1.86 1.12-3.09 4.05 2.21-7.45 1.00 1.00 2.51 1.48-4.25 4.88 2.54-9.40 1.21 .77-1.90 1.14 .70-1.85 1.00 1.00 1.10 .69-1.74 .65 .35-1.21 1.24 .77-2.01 1.35 .80-2.25 1.00 1.00 1.68 1.09-2.57 2.10 1.22-3.64
Age (y) at menarche
No. of births
Age (y) at first birth‘
Family history of breast cancer History of cystic breast disease Quetelet Index (age 20)
Quetelet Index (current)d
’ Estimatesare adjusted in mteractmn, age h Postmenopause
at menarche. e&mate
a conditional number IS a point
log~st~
of births, esnmate
‘Age at first birth replacesno. of buths in d Quetelet Index (current) replacesQuetelet
regresswn
famdy for the the
95%
CI
model
OR
for age,
menopausal
history. cystic breast disease, udds ran” at age of dqnosis
model.
Index
95%
(age
20)
in the
model.
CI
staws,
and Quetelct of 50 years.
Test for difference between groups P
.I1 .41 .83 .19 .Ol .68 .42 .24 .05 .12 .87 .18 .82 .52 age-menopausal Index
(age
smtus 20).
520
Kreiger et al. STEROID RECEPTORS
for PRcant
tumors
reduced There
AEP Vol. 1, No. 6 November 1991: 513-523
AND BREAST CANCER
(OR,
1.76).
risk for both
A later age at menarche PR+ and PR-
were differences
.Ol).
Women
(OR,
.48) and at higher
breast
in risk according
who gave birth
was associated
to PR status
four or more times
risk for PR-
tumors
for number
women
decreased
were
risk for PR-
significantly
different
significant
at elevated
breast
(OR,
1.64).
risk for PR+
cancer
(OR,
.75).
of births
(P =
were at lower risk for PR+ tumors In an analysis
age at first birth (in place of number of live births) in the model, birth (>30) was nonsignificantly protective for both PR’ and Nulliparous
with a nonsignifi-
cancer.
breast
Neither
from 1.0, and the difference
cancer
(OR,
of these
between
that
included
late age at first live PR- breast cancer.
them
1.25)
and
risk estimates
was
was not statistically
(I’ = .24).
A family
history
of breast
cancer
cancer
(OR,
1.86),
but was associated
cancer
(OR,
4.05).
The difference
cal significance PR+ (OR,
(P = .05).
2.51)
High Quetelet had a slight
(OR,
effect
these
breast
for PR-
the
risk for PR+ breast
elevated
estimates
risk for PR-
was of borderline elevated
breast statisti-
the risk for both
cancer.
elevated
breast
(I’ = .18). High Quetelet
cally significant
elevated
of cystic breast disease
4.88)
Index at age 20 slightly
protective
PR+ and PR-
between
A history
and PR-
moderately
with substantially
the risk for PR+ breast cancer
cancer.
This difference
Index at diagnosis
and
was not statisti-
elevated
the risk for both
tumors.
DISCUSSION Risk factors status.
were examined
An attempt
cation,
which
might
different
subgroups
studies
of risk factors
Both (11-14). tumors status
in women
with breast cancer,
was made to identify
risk groups,
be differentially be identifiable, within
associated then
breast
with
it would
cancer
stratified
within
by steroid
the larger disease
various
risk factors.
be appropriate
recegtor classifiShould
to conduct
future
subtype.
the ER and PR distributions
are highly
skewed,
in our data and in others
Good
has been
reported
on
it should
be noted
agreement,
as positive are highly
however,
or negative correlated,
(14). and
Additionally, that
similar
associated with the two measures. A strong positive association between in this analysis
concurs
with previous
on is the lack of independent after adjustment A number
increasing
findings
association
for age (10, of studies
risk factors
the
might
ER status
of
that ER and PR
be expected
age and ER+ tumor
( 12, 13, 15-19).
between
categorization
to be
status found
Also generally and menopausal
agreed status,
11, 14, 16).
that black women have a lower prevalence of ER+ too few black case tumors than do white women (13, 17, 20). 0 ur study contained patients to analyze this factor; however, removing these case patients did not alter study results. In our case-control higher
risk for ER-
reported
analysis, than
family history
for ER+ breast
of breast cancer McTiernan
cancer.
was associated with a and colleagues (11)
reported a like result, while other case-control studies reported an elevation in risk that is similar for both ER+ and ER- breast cancer ( 17, 18, 2 1). Our results concerning number of births are in agreement with several casecontrol studies that reported breast cancer but increased literature
is somewhat
a protective effect of a greater number of births for ER+ risk for ER- breast cancer (11, 15, 21). However, the
inconsistent
(17,
18).
521
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. I, No. 6 November 1991: 513-523
In general, menopausal cerning
our findings
status
the relationship
strong
confirm
with ER status.
associations
of other
previous
of the relationships
reports
It is difficult risk factors
to make substantial
to ER status
in our data and the inconsistent
of age and
conclusions
because
of both
results reported
con-
the lack of
by other
investiga-
tors. Differences status,
race,
tency
in the various
of results
and
cutoff
reported
points
varied
between
cutoff
chosen
for the studies.
results
analysis,
our case-control tumors. With breast the
tumor
consistent
basis.
status
status,
of number
Quetelet
of menopausal many
our data suggest
for breast
(2) Relative
accuracy
with
may identify
other
differences
have
examined
and precision
distinct
of breast
the
those
cancer
PRs in their
analysis.
tumors
on a
for the PR assay then (11).
(2, 5, 22, 23),
cancer
in
between
for case patients
are poorer
research
PR-
found
relationship
with ER concentration
breast
In our
for PR+ and PR-
than
few have included
correlated
study.
with age and in
in risk for PR+ and
(1) It is not available
epidemiologic
etiologically
differed
and family history
but more pronounced,
cancer,
because:
under
tumors.
status,
studies
the (15).
ER and PR classifica-
However,
of births
Index,
for PR’ and PR-
(3) PR is highly
along
tumor
was found to increase
the effects
have
to date,
(21) to 30 fmol/mg
in the two analyses.
included
the ER assay (7).
between
for study,
status,
conducted
of PR+ tumors
ERs and risk factors
our data,
of concordance
receptor
studies
the percentage
Although
is seldom
negative
from 3 fmolimg
were in the same direction,
PR status
and
inconsis-
chosen
by the risk factors
of differences
ER analysis.
parameters
breast cancers
the exception
cancer
ranged
to age, menopausal
to the general
distinguishes
menopausal
were suggestive
In addition,
in case-control
ER+ tumors
analysis,
Also,
study according
of positive
were anticipated
that PR status better case-case
definition
For example,
of the high degree
similar
under
and time of test may contribute
in the literature.
to define
Because tions,
populations
stage of disease,
However,
suggest
to a greater
that
extent
PR than
does ER status. In our study,
neighborhood
by socioeconomic factors
under
status.
study,
not have introduced status
and,
hence,
cancer.
Although
pronounced since
resulting
spurious
associations,
there
differences are needed
ER and PR status.
that
to suggest were
or differences
differences different
between
pathways
according
This finding
and PRs are biologically
our results
and to clarify the relationship
important
anisms.
It is, of course,
possible
to investigate are indicative
that the relatively
further of distinct
small observed
more
is unexpected
correlated
in our data actually
breast
PR analyses,
to confirm
It is particularly
receptor
to ER status
are highly
of association
it would
to steroid
for ER+ and ER-
our ER and
to PR status were found.
patterns
on some
However,
detected.
in risk factors causal
confounding
in overmatching
according
the associations
similarities
according
were used to minimize
in the loss of associations.
not invalidate
we found enough
ER and PR status
ERs. Studies
subjects
may have resulted
possibly should
In conclusion, were not strong
controls
This approach
dependeiit
whether
or not the
pathogenic deviations
on
between mechin risk are
due to chance, but it may be that different subtypes of breast cancer are determined, in part, by different constellations of risk factors. This research was supported in part by Ontario
Ministry
of Health
research
grant 01617; the Ontario
Research Foundation, Toronto, Ontario; search Foundation, Baltimore, MD.
and the Alcoholic
Cancer Beverage
Treatment Medical
and Re-
The authors are grateful for the collaboration of the attending physicians at Princess Margaret Hospital, Sunnybrook MedIcal Centre, Toronto General Hospital, and Mt. Sinai
522
Kreiger et al. STEROID RECEPTORS AND BREAST CANCER
AEP Vol. 1, No. 6 November 1991: 513-523
Hospital. We are also grateful to the interviewers who collected the data-Anne Allen, Rosemary Chepa, Suzanne Gale, Kathleen Gillespie, Lindsay Hall, Linda Harrel, Virginia Hunter, Beverly Linden, Muriel Relton, Shirley Rothery, Irene Servos, Ann Skene-Melvin, Sybil Tinker, Kathy Carroll, Binny Heaton, Helene Katz, Maxine Newton, Barbara Walker, and Linda West; Verna Cundari and Phyllis Disenhouse who coordinated the study; and Leonard Gaetano and Marguerite Angeloni who managed the data.
REFERENCES 1. Hawkins RA, Roberts MM, Forrest APM, Oestrogen receptors and breast cancer: Current status, Br J Surg. 1980;67:153-69. 2. Fisher B, Redmond C, Brown A, et al. Influence of tumor estrogen and progesterone receptor levels on the response to tamoxifen and chemotherapy in primary breast cancer, J Clin Oncol. 1983;1:227-41. 3. Wittliff JL. Steroid hormone receptors in breast cancer, Cancer. 1984;53:630-43. 4. Ruder AM, Lubin F, Wax Y, Geier A, Alfundary E, Chetrit A. Estrogen and progesterone receptors in breast cancer patients: Epidemiologic characteristics and survival differences, Cancer. 1989;64:196-202. 5. Clark GM, McGuire WL, Hubay CA, Pearson OH, Marshall JS. Progesterone receptors as a prognostic factor in stage II breast cancer, N Engl J Med. 1983;309:1343-7. 6. Stanford JL, Szklo M, Brinton LA. Estrogen receptors and breast cancer, Epidemiol Rev. 1986;8:42-59. 7. Ryan ED, Clark AF, Mobbs BG, Ooi TC, Sutherland DA, Tustanoff ER. Interlaboratory quality control of estrogen and progesterone assays in breast cancer tissue using lyophilised cytosols, Clin Biochem. 1985;18:20-6. 8. Rosenberg L, Palmer JR, Miller DR, Clarke EA, Shapiro S. A case-control study of alcoholic beverage consumption and breast cancer, Am J Epidemiol. 1990;131:6-14. 9. Kaufman DW, Werler MM, Palmer JR, et al. Diazepam use in relation to breast cancer: Results from two case-control studies, Am J Epidemiol. 1990;131:483-90. 10. Dubin N, Pastemack BS. Risk assessment for case-control analysis by polychotomous logistic regression, Am J Epidemiol. 1986;123:1101-17. 11. McTiernan A, Thomas DB, Johnson LK, Roseman D. Risk factors for estrogen receptor-rich and estrogen receptor-poor breast cancers, J Nat1 Cancer Inst. 1986;77:849-54. 12. Ballard-Barbash R, Griffin MR, Fisher LD, Covalciuc MA, Jiang N. Estrogen receptors in breast cancer: Association with epidemiologic risk factors, Am J Epidemiol. 1986;124:77-84. 13. Hulka BS, Chambless LE, Wilkinson WE, Deubner DC, McCarty KS Sr, McCarty KS Jr. Hormonal and personal effects on estrogen receptors in breast cancer, Am J Epidemiol. 1984;119:692-704. 14. Oxley DK, Haven GT, Wittliff JL, Gilbo D. Precision in estrogen and progesterone receptor assays, Am J Clin Pathol. 1982;78(suppl):587-96. 15. Hildreth NC, Kelsey JL, Eisenfeld AJ, LiVolsi VA, Holford TR, Fischer DB. Differences in breast cancer risk factors according to the estrogen receptor level of the tumor, J Nat1 Cancer Inst. 1983;70:1027-31. 16. Elwood JM, Godolphin W. Oestrogen receptors in breast tumours: Associations with age, menopausal status and epidemiological and clinical features in 735 patients, Br J Cancer. 1980;42:635-44. 17. Stanford JL, Szklo M, Boring C, et al. A case-control study of breast cancer stratified by estrogen receptor status, Am J Epidemiol. 1987;125:184-94. 18. Cooper JA, Rohan TE, Cant EL, Horsfall DJ, Tilley WD. Risk factors for breast cancer by oestrogen receptor status: A population-based case-control study, Br J Cancer. 1989;59:119-25. 19. Mckeown-Eyssen GE, Rodgers-Melamed 1, Clarke EA. Estrogen receptor status of breast cancer in Ontario, Can Med Assoc J. 1985;133:997-1000.
AEP Vol. 1, No. 6 November 1991: 513-523
STEROID RECEPTORS
Kreiger et al. AND BREAST CANCER
523
20. Lesser ML, Rosen PP, Senie RT, Duthie K, Botet C, Schwartz MK. Estrogen and progesterone receptors in breast carcinoma: Correlations with epidemiology and pathology, Cancer. 1981;48:229-309. 21. Hislop TG, Coldman AJ, Elwood JM, Skippen DH, Kan L. Relationship between risk factors for breast cancer and hormonal status, Int J Epidemiol. 1986;15:469-76. 22. Bertuzzi A, Vezzoni P, Ronchi E. Prognostic importance of progesterone receptors alone or in combination with estrogen receptors in node-negative breast carcinoma (abstract), Proc Am Assoc Cancer Res Am Sot Clin Oncol. 1981;22:447. 23. Ellis LM, Wittliff JL, Bryant MS, Hogancamp WE, Sitren HS, Bland KI. Correlation of estrogen, progesterone, and androgen receptors in breast cancer, Am ] Surg. 1989;157:577-80.
