Endocrine Reviews 20 (3): 253-278
Copyright © 1999 by The Endocrine Society
Tamoxifen, Raloxifene, and the Prevention of Breast Cancer1
V. Craig Jordan and
Monica Morrow
Departments of Molecular Pharmacology, Biological Chemistry
(V.C.J.), and Surgery (M.M.), Robert H. Lurie Comprehensive Cancer
Center, Northwestern University Medical School, Chicago, Illinois
60611
 |
Abstract
|
|---|
- I. Introduction
- II. Lacassagne’s Prevention Principle: A Target and an Estrogen Antagonist
- III. Tamoxifen as an Antitumor Agent
- A. ER status and the duration of tamoxifen
- B. Contralateral breast cancer
- C. Endometrial cancer
- D. Conclusions
- IV. Selective Estrogen Receptor Modulation
- A. Antiestrogen activity at the ER
- B. Coactivators for ER
- C. Alternate response elements on DNA
- D. An alternate ER-ERß
- V. Biological Basis for Tamoxifen as a Breast Cancer Preventive
- A. Animal models
- B. Bones
- C. Lipids
- D. Uterus
- VI. Risk Factors for Breast Cancer
- A. Interactions among risk factors
- B. Identification of candidates for chemoprevention
- VII. Prevention of Breast Cancer with Tamoxifen
- A. Royal Marsden Pilot Study
- B. NSABP/NCI Study
- C. Italian study
- D. Conclusions
- VIII. Biological Basis for Raloxifene as a Breast Cancer Preventive
- A. Antitumor actions
- B. Bones
- C. Lipids
- D. Uterus
- IX. Study of Tamoxifen And Raloxifene (STAR)
- X. The Future of Prevention
|
I. Introduction
|
|---|
IN 1896, George Beatson demonstrated
that the removal of the ovaries from premenopausal women with
metastatic breast cancer could, in some cases, cause regression of the
disease and improve the prognosis of the patient (1). However, by 1900
Stanley Boyd had established that only one in three patients would
obtain improvement for about 1 yr (2). Despite this disappointment, a
link was established between an ovarian factor and the growth of some
breast cancers. This observation was to become the foundation of modern
clinical practice and the rationale for the use of antiestrogens to
treat breast cancer (3, 4). At the turn of the century, studies were
being conducted in the laboratory to complement the clinical effort.
Inbred strains of mice were being established for medical research, and
it was found that certain strains of mice developed a high incidence of
mammary tumors. Lathrop and Loeb (5) reported that an early ovariectomy
could prevent the spontaneous development of tumors but it was not
until Allen and Doisey (6) identified "estrus stimulating
principle" that ovarian hormones could be linked to the development
of breast cancer. By 1936, Professor Antoine Lacassagne, again working
with high-incidence strains of mice, suggested that if breast cancer
was caused by a special hereditary sensitivity to estrogen, then the
disease could be prevented by developing a therapeutic antagonist to
estrogen action in the breast (7). However, there were no therapeutic
antagonists of estrogen at that time, nor was there a target to design
drug molecules. Nevertheless, exciting developments in the discovery of
nonsteroidal estrogens would establish the structural basis of carrier
molecules, which resulted in the design of the two drugs, tamoxifen and
raloxifene (Fig. 1
), both originally
described as antiestrogens and used today in a clinical trial to
prevent breast cancer in high-risk women (see Section IX).
View larger version (24K):
[in this window]
[in a new window]
|
Figure 1. The discovery of MER-25, and the knowledge that a
strategically placed alkylaminoethoxy side chain confers antiestrogenic
properties, was important to develop the antiestrogens tamoxifen and
raloxifene from the known nonsteroidal estrogens, triphenylethylene and
diethylstilbestrol. Tamoxifen is used for the endocrine treatment of
all stages of breast cancer and is available for the reduction of
breast cancer incidence in high-risk women. Raloxifene is used for the
prevention of osteoporosis in postmenopausal women, but because a
preliminary evaluation shows a reduction in the risk of breast cancer,
raloxifene is to be evaluated for the prevention of breast cancer in
high-risk postmenopausal women. The clinical trial STAR started to
recruit the 22,000 volunteers in 1999.
|
|
Estrogen action in the 1930s was assayed using ovariectomized mice as
originally described by Allen and Doisy (6). Using this technique,
parallel research ventures resulted in the discovery of the
triphenylethylene-based estrogens (8, 9, 10) and the stilbene-based
estrogens (Fig. 1
). The triphenylethylenes are long acting and are
stored in body fat (11, 12, 13, 14), whereas the hydroxystilbene derivatives
are short acting (9, 15, 16), primarily because of rapid-phase II
metabolism after absorption. However, Dodds and associates (17, 18)
described an extremely potent compound, diethylstilbestrol (Fig. 1),
that was widely used in gynecology and also subsequently used, at high
doses, as a treatment for advanced breast cancer in postmenopausal
women (19, 20).
In the 1950s and 1960s it became clear that adrenalectomy, with
glucocorticoid support, could also improve the prognosis of some
postmenopausal women with advanced breast cancer (21). In fact, about
one third of the women responded, i.e., about the same
proportion as premenopausal women after oophorectomy. The reason for
the apparently arbitrary responses, however, would not become clear
until the discovery of the estrogen receptor (ER) by Jensen and
Jacobson (22) and the subsequent application of this knowledge to
predict the hormone responsiveness of a patient’s tumor to endocrine
ablation (23). This was an extremely important finding because it
prevented those patients who had an ER-negative tumor from having
additional major surgery with little hope of a response. Only patients
whose tumors had high levels of ER were likely to respond to endocrine
ablative surgery (24). It is important, however, to stress that in the
early 1970s, there was no significant clinical experience available
with the class of drugs called antiestrogens, and no large clinical
studies had linked the efficacy of antiestrogens with the
presence or absence of the ER. Several antiestrogens had been tested in
small clinical studies, but tamoxifen, the first clinically useful
antiestrogen for the treatment of advanced breast cancer in
postmenopausal women, was not approved by the Food and Drug
Administration in the United States until 1977 (25).
Although antiestrogens are important therapeutic agents today, 40 yr
ago there was very little interest in treating breast cancer
with new hormonal drugs, and most of the research in endocrinology was
focused on an understanding of reproduction. The discovery of the
nonsteroidal antiestrogens was serendipitous and resulted from an
interest in contraception in the 1950s. The first nonsteroidal
antiestrogen to be reported in the literature, MER25 (Fig. 1), was
described by Lerner and co-workers in 1958 (26) as an agent that had no
other hormonal or antihormonal properties in any species tested. In
fact, it was a blocking drug for estrogen action with almost no
estrogenic properties. The drug failed in clinical trial because large
doses were required (MER25 has low potency), which caused serious
central nervous system side effects (27). On one hand this was
disappointing but, it must be stressed, that a pure antiestrogen such
as MER25 would ultimately have been catastrophic as an agent to prevent
breast cancer. Drug discovery switched to the triphenylethylene-based
compounds that resulted first in clomiphene and then tamoxifen (25).
Subsequently, drug discovery concentrated on compounds with a high
affinity for ER (25). Only a research focus on cancer in the 1970s
facilitated tamoxifen’s development as a breast cancer therapy for all
stages of the disease (25).
The critical property of the so called "antiestrogens," which
permitted their subsequent development as long-term preventives for
breast cancer, was that they are antiestrogens at some sites, like the
breast, but had estrogen-like properties at other sites to maintain
bone density and lower circulating cholesterol (28, 29, 30, 31, 32, 33, 34, 35). The unusual
target site-specific action as an estrogen or as an antiestrogen was
true for both tamoxifen and raloxifene (29, 30). Lacassagne’s
prediction (7) of developing an antagonist to estrogen action to
prevent breast cancer in healthy women would not have occurred if the
available drugs had increased the risk of osteoporosis and coronary
heart disease.
The goal of this review is to provide an up-to-date analysis of the
current status of efforts to prevent breast cancer in women by the
strategic use of antiestrogens. One aim of our review is to identify
the principles, established in the laboratory, that have, through the
clinical trial process, proven to be valid in patients with breast
cancer or women at risk for breast cancer. The review will also provide
the scientific basis for the ongoing trial called Study of Tamoxifen
And Raloxifene (STAR). This trial is examining the worth of raloxifene,
a drug approved for the prevention of osteoporosis, to prevent breast
cancer in postmenopausal women with elevated risk factors. The
biological basis for consideration of the antiestrogen raloxifene to be
used as a breast cancer preventive is described in Section
VIII. In the interests of space it is not possible to review the
antiestrogen literature exhaustively, but we intend to provide
sufficient background to link laboratory research with clinical
results.
|
II. Lacassagne’s Prevention Principle: A Target and an Estrogen
Antagonist
|
|---|
In 1962, Jensen and Jacobson (22) demonstrated that
[3H]estradiol bound to and was retained by estrogen
target tissue, e.g., uterus, vagina, and pituitary gland, in
the immature female rat. By contrast, tissues that did not respond to
estrogen did not retain [3H]estradiol. Jensen proposed
that an ER must be present in estrogen target tissues to capture
circulating steroids and initiate the cascade of biochemical events
associated with estrogen action in that particular tissue. Gorski and
co-workers (36, 37) first identified the ER as an extractable protein
from rat uterus. Subsequently, the groups of Gorski et al.
(38) and Jensen et al. (39) independently proposed
subcellular models of estrogen action in target tissues. However,
Jensen and associates (23) took the process one step further by
proposing the clinical ER assay to predict hormone-responsive breast
cancer. Thus, a mechanistic link between estrogen action and the growth
of breast cancer was established.
The MCF-7 breast cancer cell line is ER positive (40, 41), and the
cells have found ubiquitous applications in cancer research
laboratories throughout the world (42). Most importantly, access to
these cells has resulted in a fundamental change in the understanding
of hormone action that has resulted in the discovery of the steroid
receptor superfamily of receptors (43, 44). Jensen and co-workers (45, 46) first developed monoclonal antibodies to ER derived from MCF-7
cells. The antibodies were used to establish that the ER was a nuclear
protein (47), and the technology of immunocytochemistry is now standard
for the determination of receptor status in breast biopsies (48, 49).
However, the application of monoclonal antibodies as probes to clone
and sequence the ER gene (50, 51, 52) is of fundamental significance for
the understanding of the ER as a nuclear transcription factor.
The ER is a nuclear protein (47, 53, 54), which, for convenience, is
subdivided into six functional domains (Fig. 2) (55, 56). The E regions make up the
steroid-binding domain that undergoes a conformation change to lock
estradiol into its hydrophobic pocket (see Section IV.A).
Changes in conformation of the ER permits the binding of coactivators
to the activating function 1 and 2 regions (AF-1 and 2) (57, 58, 59) and
facilitates the interaction of the ER with DNA through the DNA-binding
domain (C region) (60, 61). The transcription unit is selectively
located as a homodimer in the promoter region of estrogen-responsive
genes to initiate the events associated with estrogen-stimulated cell
replication. Unfortunately, it is not possible to provide much more
than the basic concepts in hormone action because our article is
focused on progress in breast cancer chemoprevention. The topic of gene
regulation is a rapidly evolving story, so we strongly recommend that
interested readers consult recently published reviews that provide
further information (62, 63, 64).
View larger version (25K):
[in this window]
[in a new window]
|
Figure 2. A subcellular model of estradiol (E2)
action in a target tissue. Estradiol diffuses into all cells but binds
to the ER specifically located in estrogen target tissues
(e.g., uterus, vagina, some breast cancers, etc.). The
steroid receptor complex undergoes a conformational change and
dimerizes before binding to an estrogen response element (ERE) in the
promotor region of an estrogen-responsive gene. A transcription unit is
formed by interaction with coactivator (CoA) molecules to initiate RNA
synthesis and ultimately the estrogen-stimulated cellular response.
Corepressor (CoR) modules are believed to prevent transcription by
exclusively interacting with antiestrogen ER complexes. The ER is
divided into six regions (A–F). The DNA-binding domain (C) is
essential for the interaction of the ER with the ERE. The
ligand-binding domain (E) is the site of E2 binding and the
site of competitive binding by antiestrogens. The AA351 is identified
in the E region because it is the known site of an interaction with the
alkylaminoethoxy side chain of raloxifene. The activating function
(AF-1 and -2) regions are the areas of the ER that interact with
coactivators to form the transcription unit at an estrogen-responsive
gene.
|
|
The second facet of Lacassagne’s hypothesis is the requirement for an
antiestrogen to block estrogen action. Tamoxifen blocks the binding of
[3H]estradiol to the ER derived from rat uterus (65, 66, 67, 68)
or human tumor (69, 70). However, initial clinical studies with
tamoxifen were conducted exclusively on unselected populations of
postmenopausal women with advanced breast cancer (3, 4), and not until
1977 was it noted that tamoxifen was more likely to be effective in
ER-positive breast cancer (71). Tamoxifen is currently used as a
palliative therapy in the treatment of pre- and postmenopausal patients
with ER-positive advanced (Stage IV) breast cancer. By contrast, the
application of the concept of adjuvant therapy has revolutionized the
treatment of breast cancer. Systemic adjuvant therapy is used after
breast surgery to destroy undetected micrometastases around a woman’s
body.
Adjuvant studies with tamoxifen have proved to be successful in
increasing survival (72, 73, 74) but, perhaps most importantly, the
interaction between laboratory and clinical research endeavors has
ultimately elucidated both the principal mechanism of action of
tamoxifen as an antitumor agent in women and identified those women
most likely to benefit from adjuvant tamoxifen treatment. During the
past 12 yr there has been some confusion in the literature about
whether tamoxifen was active in ER-positive breast cancer exclusively
or whether ER-negative breast cancer could respond (75, 76, 77).
Additionally, there was controversy as to whether tamoxifen was
significantly active as an adjuvant in premenopausal women (78). The
reader is referred to recent reviews on the clinical investigation and
development of tamoxifen (79, 80, 81, 82, 83) for further information, but we will
summarize the latest findings of the world-wide randomized clinical
trials (84). It is important to appreciate that the general principles
derived from the use of tamoxifen as a therapy for breast cancer can be
used as a basis for consideration of tamoxifen as a estrogen antagonist
for the prevention of breast cancer.
|
III. Tamoxifen as an Antitumor Agent
|
|---|
The 1998 Oxford Overview Analysis (84) involved any randomized
trial that began before 1990. The analysis included 55 trials of
adjuvant tamoxifen vs. no tamoxifen before recurrence. The
study population comprised 37,000 women with node-positive and
node-negative breast cancer, thus comprising 87% of world evidence of
known randomized clinical trials. Of these women, fewer than 8,000 had
a very low or zero level of ER and 18,000 were classified as ER
positive. The ER status of the remaining nearly 12,000 women was
unknown, but based on the normal distribution of ER in random
populations, the authors estimated that two-thirds would be ER
positive.
This clinical trial data base (84) can now be used to answer the
questions raised over the past two decades by laboratory results and
hypotheses. In the 1970s three laboratory observations emerged that
merited evaluation in clinical trial: 1) tamoxifen blocks estrogen
binding to the ER, making patients with ER-positive disease more likely
to respond than those with ER-negative disease (85), 2) tamoxifen
prevents mammary cancer in rats (86, 87) so the drug could reduce the
incidence of primary breast cancer, and 3) long-term treatment was
better than short-term treatment to prevent rat mammary carcinogenesis;
therefore, longer adjuvant therapy with tamoxifen should be superior to
short-term adjuvant therapy (88, 89, 90), i.e., 5 yr of
tamoxifen should be superior to 1 yr of tamoxifen. By the late 1980s,
tamoxifen had been shown in the laboratory to block estrogen-stimulated
breast tumor growth but to encourage the growth of human endometrial
cancer implanted in the same athymic mouse (91, 92). The clinical
question therefore became "are patients, who are receiving long-term
adjuvant tamoxifen therapy, at risk for an increased incidence of
endometrial cancer?" (92).
The process of evaluating the impact of translational research is
important to establish what works and achieves clinical progress and
what does not. A clinical trial should not begin without a strong
hypothesis and the incorporation of relevant scientific results. For
convenience, the discussion in this section will be subdivided, but the
end points of duration of tamoxifen usage, menopausal status, and ER
status interact, making the size of a pharmacological effect subject to
change.
A. ER status and the duration of tamoxifen
The ER status of the patient is highly predictive of a treatment
response to long-term tamoxifen therapy. The treatment effect, based on
receptor status, is summarized in Table 1. The recurrence reductions produced by
tamoxifen in ER-positive patients are all highly significant
(2P < 0.00001), and the trend between the different
durations of tamoxifen is also highly significant
(
2 = 45.5, 2P < 0.00001). By contrast,
the therapeutic effect of tamoxifen on ER-negative patients is minimal.
Additionally, the questions could be asked, "Does more ER give a
better response to tamoxifen?" and "Does an additional progesterone
receptor (PgR) assay help to improve the results with tamoxifen?" In
the trials of about 5 yr of tamoxifen treatment, the proportional
reductions of recurrence were 43 ± 5% and 60 ± 6% for
patients with below or above 100 fmol/mg cytosol ER protein. This
translated to a reduction in mortality of 23 ± 6% and 36 ±
7%, respectively. Clearly, one can conclude the ER is a powerful
predictor of tamoxifen response, a conclusion consistent with
tamoxifen’s proven mechanism of action as an estrogen antagonist in
breast cancer (82). Although PgR-positive status might be thought to be
of benefit, these data show that there was little additional value if
the tumor was already ER positive. A comparison of interactions is
shown in Table 2. A comparison of the
2,000 women who had ER-positive and PgR-negative tumors and the 7,000
women who had ER-positive and PgR-positive tumors shows there was no
apparent difference in the effect of tamoxifen on either the recurrence
rates or mortality rates. Additionally, the numbers were too few (602
women) in the Overview Analysis (84) to allow a meaningful prediction
of the benefits of tamoxifen in patients who had an ER-negative but
PgR-positive tumor.
View this table:
[in this window]
[in a new window]
|
Table 2. A comparison of the proportional risk reduction of
adjuvant tamoxifen therapy based on PgR status in populations of
ER-positive patients
|
|
The Overview Analysis also provides unequivocal proof of the laboratory
principle (88, 89, 90) that longer adjuvant tamoxifen therapy was predicted
to provide more benefit. The duration of therapy is extremely important
for the ER-positive premenopausal woman with large amounts of
circulating estrogen that can rapidly reverse the effect of short-term
tamoxifen treatment. The effect of the duration of tamoxifen treatment
on the reduction of recurrence rates and the reduction of death rates
is shown in Fig. 3. The duration of
tamoxifen therapy is critical for the premenopausal patient: the effect
of 1 yr of treatment is virtually nonexistent compared with the benefit
of 5 yr of treatment. It is also important to point out that the
reduction of death rates in women under 50 yr of age and over 60 yr of
age treated with 5 yr of tamoxifen is identical, at around 33% (Table 3). By contrast, the effect of tamoxifen
duration on women over the age of 60 is less dramatic because 1 yr of
tamoxifen is much more effective in postmenopausal women. These data
are illustrated in Table 3, which shows a 2- to 3-fold increase in the
effectiveness of tamoxifen when the duration is increased from 1 to 5
yr, whereas there is a 20-fold increase in tamoxifen’s effectiveness
for premenopausal women with an increased duration of 1–5 yr (Fig. 3).
View larger version (18K):
[in this window]
[in a new window]
|
Figure 3. The relationship between the duration of adjuvant
tamoxifen therapy in ER-positive premenopausal patients and the
reduction in recurrence and death rate. A longer duration of treatment
has a dramatic effective on patient survival. [Derived from Ref.
84.]
|
|
View this table:
[in this window]
[in a new window]
|
Table 3. Proportional risk reductions in 60- to 69-yr-old
breast cancer patients when the known ER-poor patients are excluded
|
|
B. Contralateral breast cancer
Tamoxifen consistently reduces the risk of contralateral breast
cancer (i.e., a second primary breast cancer in the other
breast) independent of age (84). Women have a proportional risk
reduction that is 27 ± 11% or 31 ± 7% if they are below
or above the age of 50, respectively. The principle "longer is
better" is also true for the reduction of risk for contralateral
breast cancer with adjuvant tamoxifen therapy. Five years is better
then 2 yr or 1 yr of adjuvant therapy with tamoxifen (Fig. 4). In fact, 1 yr of adjuvant tamoxifen
does not significantly reduce the incidence of contralateral breast
cancer compared with control because the SD is so large
(13 ± 13% reduction compared with control).
View larger version (17K):
[in this window]
[in a new window]
|
Figure 4. The relationship between the duration of adjuvant
tamoxifen and the reduction in contralateral breast cancer. A longer
duration is clearly superior, and the 5-yr result that produces a 47%
reduction in contralateral breast cancer is equivalent to the result
observed in the tamoxifen prevention trial presented in Fig. 9.
[Derived from Ref. 84.]
|
|
It is interesting to note that a quarter of the women allocated to the
known adjuvant trials in the Overview Analysis (84) were Japanese, who
have an annual incidence of contralateral breast cancer in patients not
receiving tamoxifen of 2 per 1,000 compared with 6 per 1,000 elsewhere
in the world. Therefore, if 5 yr of tamoxifen therapy can halve
contralateral breast cancer, then the absolute benefit for Japanese
women would be 1 per 1,000 and 3 per 1,000 elsewhere for both young and
old women. Finally, the proportional reduction in contralateral breast
cancer appears to be similar in women whose initial tumor being treated
with tamoxifen was ER-poor (29 ± 15%) compared with the rest of
the study population (30 ± 6%). This is an important result for
the potential application of tamoxifen for the reduction of
contralateral breast cancer in the woman with a primary breast cancer
that is unequivocally ER negative.
C. Endometrial cancer
The overall increase in the incidence of endometrial cancer in the
Overview Analysis was 2- to 3-fold (84). There was no association with
dose, i.e., 20 mg and 30–40 mg daily produced relative risk
(RR) ratios of 2.7 and 2.4, respectively. However, there was a
suggestion that 1 and 2 yr of tamoxifen doubled the incidence of
endometrial cancer and 5 yr quadrupled the incidence. However, the side
effect is so rare (i.e., the numbers are too small) that the
risk ratios are not significantly different from one another for each
duration of tamoxifen. It is important, however, to state that the
absolute increase in endometrial cancer was only half as big as the
absolute decrease in contralateral breast cancer.
The Overview Analysis was able to identify 3,673 women who took 5 yr of
adjuvant tamoxifen. With 26,400 woman years of follow-up before breast
cancer recurrence in this group, there were seven endometrial cancer
deaths. It is estimated that during the whole first decade, the
cumulative risk was two deaths per 1,000 women. It is important to
state that the current knowledge about the association of tamoxifen
with endometrial cancer will improve these statistics. In general, the
reported trials were conducted without awareness of the endometrial
side effects of tamoxifen. This is no longer the situation, and early
detection will improve mortality figures associated with tamoxifen.
D. Conclusions
Tamoxifen has been extensively tested in clinical trials of
adjuvant therapy for 20 yr. The Overview shows that the proportional
mortality reductions were similar for women with node-positive or
node-negative disease (84). However, the absolute reductions in
mortality were much greater in node-positive than node-negative
disease. Additionally, patients with ER-positive disease have an
increased reduction in death rate with longer duration of tamoxifen
treatment, whereas patients who are ER-negative do not benefit from
tamoxifen, regardless of the duration of therapy. The value of a long
duration of treatment is most important for the premenopausal patient
(Fig. 3). This latter finding is new, as the results for premenopausal
women could not be ascertained with certainty in earlier overviews
(74). The Oxford Overview Analysis has established the veracity of the
laboratory concepts that tamoxifen would be most effective in
ER-positive disease, longer duration would be more beneficial, and
tamoxifen would prevent primary breast cancer, in this case
contralateral disease (85, 86, 87, 88, 89, 90).
Overall, the absolute improvement in recurrence was greater during the
first 5 yr after surgery, but improvement in survival increased
steadily throughout the first 10 yr. This is an important finding
because the patient is clearly benefiting from tamoxifen even after
therapy has been discontinued. There is an accumulation of the
tumoristatic/tumoricidal actions of tamoxifen for at least the first 5
yr of treatment, but the benefit continues after therapy stops. This is
also true for the reduction in contralateral breast cancer; the breast
seems to be protected so the value remains after therapy stops. This
observation is extremely important for the application of tamoxifen as
a preventive because a 5-yr course of tamoxifen would be expected to
protect a woman from breast cancer for many years afterward.
Finally, the risk/benefit ratio of tamoxifen therapy can be stated to
be strongly in the benefit category. The risk of endometrial cancer, a
concept derived from laboratory studies (92), is of concern, but the
benefits clearly outweigh the risks. In contrast, early concerns about
the carcinogenic effects of tamoxifen in the rat liver (see
Section VII) do not translate to the clinic as there is no
evidence from the Overview Analysis of an increase in either liver or
colorectal cancer in patients who take tamoxifen (84)
|
IV. Selective ER Modulation
|
|---|
Nonsteroidal antiestrogens were originally defined as compounds
that would inhibit estradiol-stimulated rat uterine weight. The
compounds tamoxifen (ICI 46,474) (93), nafoxidine (U, 11,100A) (94),
nitromifene (CI628) (95), and clomiphene (MRL 41)(96) are all partial
estrogen agonists in the uterus that also inhibit
dimethylbenzanthracene (DMBA)-induced rat mammary tumor growth (67, 97, 98) and the growth of ER-positive MCF-7 breast cancer cell growth
in vitro (99). Thus, in the 1960s and 1970s,
antiestrogenicity was correlated with antitumor activity. However, the
finding that the compounds expressed increased estrogenic properties,
i.e., vaginal cornification and increased uterine weight in
the mouse (93, 100), raised questions about the reasons for the species
specificity. One obvious possibility was species-specific metabolism,
i.e., the mouse converts antiestrogens to estrogens via
novel metabolic pathways. However, no species-specific metabolic routes
to known estrogens (101, 102) have been identified, but knowledge of
the mouse model created a new dimension for study that ultimately led
to the recognition of the target site-specific actions of
antiestrogens. This concept was subsequently referred to as selective
ER modulation (SERM) to describe the target site-specific effects of
raloxifene, an antiestrogen originally targeted for an application in
breast cancer but now used, paradioxically, as a preventive for
osteoporosis (see Section VII). Now the whole class of drugs
are known as SERMs.
The ER-positive breast cancer cell line MCF-7 (for a review see Ref.
42) can be heterotransplanted into immune-deficient athymic mice but
the cells will only grow into tumors with estrogen support.
Paradoxically, tamoxifen, an estrogen in the mouse, does not support
tumor growth (103) but stimulates mouse uterine growth with the same
spectrum of tamoxifen metabolites present in both the uterus and the
human tumor (28). To explain the selective actions of tamoxifen in
different targets of the same host, it was suggested that the ER
complex could be interpreted as a stimulatory or inhibitory signal at
different sites (28). The concept was consolidated with experimental
evidence from two further models. First, tamoxifen and raloxifene
maintain bone density in the ovariectomized rat, but both compounds
inhibit estradiol-stimulated uterine weight (29) and prevent
carcinogen-induced mammary tumorigenesis (30). Second, the finding that
tamoxifen would partially stimulate the growth of a human endometrial
carcinoma transplanted into athymic mice (91) allowed the investigation
of two human tumors bitransplanted in the same mouse to determine
whether tamoxifen could inhibit estrogen-stimulated growth of two
tumors in the same host equally (92). Tamoxifen demonstrated target
site specificity: breast tumor growth was controlled but endometrial
tumors continued to grow. Again the range of tamoxifen metabolites were
consistent in all target tissues despite the contrasting biological
responses, so it was concluded that the ER complexes must be
interpreted differently in different target tissues.
During the past decade an intense effort has been made to
discover the reason for the target site-specific effects of
antiestrogens. Not only will this knowledge permit a rational
application of tamoxifen and raloxifene in patients, but also the
discovery of new mechanisms for drug selectively will open the door for
new innovations in drug discovery. For the sake of completeness, we
will briefly consider some of the current hypotheses that could explain
the molecular mechanisms of antiestrogen action in different tissue
sites.
A. Antiestrogenic activity at the ER
The crystallization of the ligand-binding domain of the ER with
estradiol and raloxifene has provided an important insight into the
conformational changes that occur in the receptor (104) liganded with
an estrogen or an antiestrogen, respectively. Estradiol causes helix 12
to seal the ligand inside the hydrophobic pocket of the ligand-binding
domain (Fig. 5A). This causes receptor
activation through the binding of coactivators on the surface of helix
12 (see Section IV.B). By contrast, the binding of
raloxifene prevents helix 12 from sealing the hydrophobic pocket (Fig. 5B), and gene transcription cannot occur because coactivators cannot
bind. Unfortunately, the final shape of the ER and anti-ER
complexes do not tell us how the tertiary changes in protein
structure occur. However, the crystal structure provides proof of the
critical importance of AA351 (aspartate) for raloxifene action. The
alkylaminoethoxy side chain is the essential structural feature of
nonsteroidal antiestrogens (for a review see Ref. 105). The distance
between the nitrogen and the oxygen must be optimal (106), the
conformations available to the side chain must not be restricted (107),
and the basicity of the nitrogen must be correct (108). Removal of the
side chain results in loss of all activity or an increase in estrogenic
properties (109). The side chain was originally predicted (110, 111) to
bind to an "antiestrogenic region" in the ligand-binding domain of
the ER to neutralize the estrogenic properties of the receptor. Simply
stated, the antiestrogen was perceived to act like a stick to prevent
the jaws of the ER from closing around the ligand. Looked at another
way, an estrogenic complex would only be created by the protein
enveloping the ligand. For the nonsteroidal antiestrogens, the
"antiestrogenic region" is now known to be AA351 on helix 3. The
discovery of an ER mutant in a tamoxifen-stimulated MCF-7 breast tumor
(112) and the finding that it can increase the estrogenic properties of
4-hydroxytamoxifen (113, 114), the active metabolite of tamoxifen
(115), and convert raloxifene from an antiestrogen to an estrogen (116, 117) is valuable biological proof that AA 351 is important for the
antiestrogenic activity of these specific compounds. This interaction,
at the critical contact point of helix 3 and helix 12, prevents helix
12 from sealing the ligand into the binding pocket (Fig. 5B).
Interestingly, a recent report of the crystal structure of
4-hydroxytamoxifen and the ligand-binding domain of ER (118) shows a
complex interaction of the side chain with several amino acids
including AA 351. The distance between AA 351 and the nitrogen of
4-hydroxytamoxifen is further than the comparable interaction in
raloxifene. This difference between the crystal structure of the
raloxifene ER complex (104) and the 4-hydroxytamoxifen ER complex
(118) may explain the promiscuous nature of the 4-hydroxytamoxifen ER
complex. However, it must be stressed the AA 351 has no role for the
antiestrogenic action of the pure antiestrogen ICI 182,780 (117) and
other amino acids may be found to be involved in the antiestrogenic
mechanisms of novel compounds in the future.
View larger version (37K):
[in this window]
[in a new window]
|
Figure 5. Comparison of the binding of estradiol (diagram A)
and raloxifene (diagram B) in the ligand-binding domain of the human
ER. The key event is the repositioning of helix 12 to seal the steroid
in the hydrophobic pocket. This event allows the ER complex to recruit
coactivators for the transcription complex. The side chain of
raloxifene prevents recruitment of coactivators by first masking AA 351
in helix 3, which is critical for the relocation of helix 12.
Coactivators cannot bind to the AF2 region of helix 12 because it
cannot seal the ligand-binding domain. [Derived from Ref. 104.]
|
|
Nevertheless, this two-dimensional model, which describes antiestrogen
and estrogen action, is too simple to encompass all the complexities of
the target site-specific actions of the antiestrogens. Furthermore, the
crystallization data do not include the conformational information
about the other half (i.e., the A, B, C, or F domains
illustrated in Fig. 2) of the ER that control interaction with
transcription factors and binding to DNA. Subtle changes in the whole
protein shape could be responsible for changes in the intrinsic
efficacy of the receptor complex at different target sites.
B. Coactivators for ER
A host of coactivator and corepressor proteins has been implicated
in the construction of a transcription complex in target cells (63, 64, 119, 120, 121, 122, 123). The finding that an antiestrogen ER complex could become
increasingly estrogenic in different cell contexts (124, 125) raised
the possibility that the differential distribution of coactivators or
corepressors could be responsible for changes in estrogenicity between
the breast and, for example, bones (126, 127). The activating function
(AF-2) region in the ligand-binding domain (Fig. 2) is known to be
repressed by tamoxifen and raloxifene, but the AF-1 region is
unaffected by tamoxifen binding (125). Clearly, the shape of a
particular complex of ligand and ER will be different for different
drugs (125). Thus coactivators could modulate estrogenicity
differentially in different target sites. The candidate proteins could
therefore amplify the anti-ER complex into an estrogenic complex.
Alternatively, the anti-ER complex might recruit completely new
proteins at a specific target site to induce or to suppress gene
transcription.
C. Alternate response elements on DNA
Anti-ER complexes can bind to an estrogen response element but
cannot recruit coactivators to initiate transcription of
estrogen-responsive genes (114). However, it is possible that the
anti-ER complex can bind to alternate sites in the promoter region to
initiate transcription. Alu DNA repeats were originally thought to be
functionally inert, but these elements may be able to activate gene
transcription with antiestrogens through an ER-related mechanism (128, 129). Additionally, a specific region of the transforming growth
factor-ß promoter is believed to be activated by raloxifene (130). A
raloxifene response element has been identified (131), but the authors
are now convinced that a simple protein DNA interaction does not occur
(132).
D. An alternate ER-ERß
The discovery of a second ER, named ERß (133), has introduced a
new dimension into the possible mechanisms of tamoxifen or raloxifene
action. Although ERß has similar functional homology in the
DNA-binding domain, there is only 55% homology between ERß and ER
in the ligand-binding domain. Clearly, one possibility to explain the
target site specificity and altered estrogenicity of antiestrogens is a
differential distribution of ER and -ß to different tissues (134).
The mechanism of action of the differential pharmacology between ER
and -ß may also involve different methods of gene activation. A novel
signal transduction pathway has been identified as a protein-protein
interaction between ERß anti-ER complexes and AP-1 (fos and jun)
(135) that is capable of activating a reporter gene. Estradiol,
however, does not activate the reporter. Therefore, the pathway would
be of pharmacological rather than physiological significance.
Interestingly, an ER tamoxifen complex will activate AP-1 reporter
systems in the context of an endometrial cancer cell (136). This has
led to speculation that the target site specificity of antiestrogens
could be both receptor and context selective.
|
V. Biological Basis for Tamoxifen as a Breast Cancer Preventive
|
|---|
With the molecular mechanisms of antiestrogens as a background, it
is now appropriate to consider the scientific rationale for selecting
tamoxifen to be tested as a breast cancer preventive, based on its
pharmacological properties. Knowledge converged over the past 25 yr to
make the choice of testing tamoxifen in well women a logical extension
of clinical experience. Tamoxifen was selected for testing as a
preventive based on 1) animal studies that demonstrated it could
prevent carcinogenesis, 2) an extensive clinical experience that showed
few serious side effects, 3) a beneficial profile of estrogen-like
action in maintaining bone density, and 4) tamoxifen reduces
circulating cholesterol. The fact that tamoxifen was already known to
reduce the incidence of contralateral breast cancer made the drug the
primary agent to test in high-risk women. The pharmacological
properties of tamoxifen have been recently reviewed extensively (82);
therefore, the purpose of this section is to act as a framework for a
comparison with raloxifene in Section VIII and as a prelude
to considering the current STAR trial (Section IX).
A. Animal models
Tamoxifen prevents rat mammary carcinogenesis induced by
dimethylbenzanthracene, N-nitrosomethylurea, and
ionizing radiation (86, 87, 88, 89, 30, 137), and long-term treatment prevents
spontaneous carcinogenesis in C3H/OUJ mice infected with mouse mammary
tumor virus (138). The latter result is of interest because tamoxifen
is classified as an estrogen in the uterus and vagina of the mouse (93, 100). This, again, illustrates the target site specificity of tamoxifen
in the mouse model, as mammary cancer is prevented almost completely.
Although the athymic mouse model heterotransplanted with breast cancer
cell lines has been extremely instructive for the use of therapeutic
tamoxifen (103), and valuable as a model for understanding drug
resistance (112), there are few parallels to chemoprevention. The MCF-7
cell line is derived from a pleural effusion, and the cells are,
therefore, metastatic breast cancer (40). As such, the cell line does
not replicate carcinogenesis in the breast or mimic primary breast
cancer cells that have not developed the metastatic phenotype.
B. Bones
Tamoxifen maintains bone density in the ovariectomized rat (29, 31), and these observations have been translated to clinical trials.
Sporadic reports (32, 139) and placebo-controlled randomized trials
(33, 140) demonstrate that tamoxifen can increase bone density in the
lumbar spine, forearm, and neck of the femur by 1–2%. Although the
increases are modest compared with the results obtained with estrogen
use or bisphosphonates (
5% increase in bone density), tamoxifen
produced a significant decrease in hip and wrist fractures as a
secondary end point in the breast cancer prevention trial (141). There
is, however, a report that tamoxifen can reduce bone density in
premenopausal women by 1–2% (142), but the decrease appears to be
without clinical significance as there is no increase in the fracture
rate. The reason for this may be the fact that a small decrease in bone
density in a premenopausal woman is still well above the range of bone
densities observed for women in their late 60s and 70s at risk for
fractures.
C. Lipids
Tamoxifen reduces circulating cholesterol (34, 35). Low-density
lipoprotein cholesterol is reduced by about 15%, but high-density
lipoprotein cholesterol is maintained. It is hypothesized that this
magnitude of fall in circulating cholesterol is a good surrogate marker
for protection from coronary heart disease and atherosclerosis. In this
regard there is evidence that woman who have been treated with 5 yr of
adjuvant tamoxifen for breast cancer have a reduced incidence of fatal
myocardial infarction (143, 144). Additionally, longer treatment (5 yr)
appears to be superior to shorter treatment (2 yr) in reducing the
number of hospital admissions for any cardiac condition (145).
Conversely, a large study in the United States of five or more years of
tamoxifen for the adjuvant treatment of breast cancer found no
statistically strong evidence for the protection of women from coronary
heart disease (146). Nevertheless, the incidence of coronary heart
disease doubled once tamoxifen treatment was stopped, and, most
importantly, there was no evidence for a detrimental effect of
tamoxifen, i.e., tamoxifen did not increase the rate of
coronary heart disease in pre- or postmenopausal women. The reasons for
the disparate results probably reflect the populations studied. Breast
cancer clinical trials usually require a good general health status
before enrollment. Obviously, women at high risk for a second disease,
coronary heart disease, would not be enrolled into a trial evaluating
the efficacy of an antiestrogenic therapy in, for example,
node-negative breast cancer where the prognosis, for the majority of
women, would be expected to be good. Only a prospective randomized
trial in a high-risk population would provide accurate data to support
a claim for cardio-protection. At this point in time, there is no
evidence that tamoxifen is detrimental based on current clinical
evaluations, but there is no prospective clinical evidence that
tamoxifen will reduce the risk of coronary heart disease.
D. Uterus
It is well known that tamoxifen produces a partial agonist action
in the rat uterus (93), but the histology is different than the
epithelial hyperplasia noted with estradiol (147). Until the late
1980s, there was very little information about the actions of tamoxifen
in the normal human uterus. However, it is now clear that a variety of
endometrial changes occur in unselected populations of women (148). The
most significant finding is an increase in the stromal component rather
than endometrial hyperplasia (149, 150). Despite the fact that
tamoxifen has been used to treat endometrial cancer, the laboratory
data suggesting that tamoxifen has the potential to encourage the
growth of preexisting disease harbored in the uterus (91, 92) provoked
an intense investigation of the rates of detection of endometrial
cancer in women using adjuvant tamoxifen treatment for breast cancer.
These data have been reviewed (151), and it is clear from the recent
results of the tamoxifen prevention trial (141) that tamoxifen does not
cause an excess of endometrial cancer in premenopausal women but does
increase risk by 3- to 4-fold in postmenopausal women. This is
consistent with the fact that women harbor 4–5 times the level of
endometrial cancer than is detected clinically (152). In other words,
the increase in the detection of endometrial cancer from 1 per 1,000
women per year to 3 per 1,000 women per year is consistent with the
known rate of occult disease. Most importantly, the stage and grade of
endometrial cancer observed in women taking tamoxifen is the same
as those in the general population (141, 153).
|
VI. Risk Factors for Breast Cancer
|
|---|
If tamoxifen is an appropriate agent to test as a chemopreventive
primarily because of its extensive clinical experience for the
treatment of breast cancer, then the issue becomes identification of
women at risk to select as a target population for recruitment to
definitive clinical trials. Family history is probably the most well
recognized risk factor for breast cancer, and it is now known that two
forms of risk are associated with a family history of the disease. An
inherited gene mutation predisposing to breast cancer is believed to
account for only 5–10% of breast cancer cases (154, 155). Although
infrequent, these mutations are significant since they are associated
with a lifetime risk of breast cancer development of 50–80% (156, 157). At present, two predisposition genes, BRCA1, located on
chromosome 17q21 (158), and BRCA2, located on chromosome 13q12–13
(159), have been identified. Both genes are inherited in an autosomal
dominant fashion and are characterized by an extremely high risk of
breast cancer development, which begins at a young age. Both genes also
confer an increased risk of ovarian cancer development, which in BRCA1
carriers is estimated to be 10% by age 60 (160), and is lower in BRCA2
carriers. In addition, germ line mutations of the tumor suppressor gene
p53, as seen in patients with the Li-Fraumeni syndrome, may account for
about 1% of breast cancer cases occurring in women age 40 and younger
(161, 162).
Most women with a family history of breast cancer do not have the
genetically transmitted form of the disease, and therefore their
increase in risk is much less than that seen in women who have
inherited a predisposition gene. The cumulative probability that a
30-yr-old woman with a mother and sister with breast cancer will
develop breast cancer by the age of 70 is reported to be between 7%
and 18% (163, 164). While this risk increases as the number of
relatives with breast cancer increases, the probability of cancer
development if both a mother and sister have bilateral breast cancer
has been reported to be only 25% (162, 164). The cumulative risk of
breast cancer development in women with a family history of breast
cancer rarely exceeds 30%, making it critically important to
distinguish those women with hereditary breast cancer from those with a
family history of the disease. Factors that should increase the
clinician’s index of suspicion that a woman is at risk for genetically
transmitted breast cancer include multiple relatives (maternal or
paternal) with the disease, a family history of ovarian cancer in
association with breast cancer, and a family history of bilateral
and/or early onset of breast cancer. Although not all women with these
factors will have genetically transmitted breast cancer, a referral for
genetic counseling will allow the construction of a detailed pedigree
to estimate both breast cancer risk and the competing causes of death
due to an increased risk of the development of other types of cancer.
Breast cancer is clearly related to endogenous hormones, and numerous
studies have linked breast cancer risk to the age of menarche,
menopause, and first pregnancy. Although the absolute age-specific
incidence of breast cancer is higher in postmenopausal than
premenopausal women (165), the absolute rate of rise of the curve is
greatest up to the time of menopause, and then slows to one-sixth of
that seen in the premenopausal period. Further support for the
promotional role of estrogen in breast cancer development comes from
the observations that early menarche (166), late menopause (167),
nulliparity, and late age at first birth (168) all increase the risk of
breast cancer development. An increased number of ovulatory cycles is
suggested to be the common mechanism of increased risk.
Other hormonal risk factors have been suggested but are not as well
established. Abortion, whether spontaneous or induced, has been
reported by some authors to increase risk (169, 170), while other
studies have found no relationship between abortion and breast cancer
risk (171, 172). Studies of the effect of lactation on breast cancer
risk have also been inconclusive (173, 174), but recent studies have
suggested that a long duration of lactation reduces breast cancer risk
in premenopausal women (175). Physical activity in adolescence is
reported to decrease risk, perhaps due to a higher rate of anovulatory
cycles (176, 177), but an increased level of physical activity later in
life has not been shown to reduce breast cancer risk (178).
Postmenopausal obesity has also been shown to increase risk (179),
perhaps due to increased peripheral estrogen production, but this
relationship between weight and risk is not observed in premenopausal
women. In fact, some studies have reported an inverse relationship
between weight and risk at a younger age (180).
The effects of exogenous hormones in the form of oral contraceptives
and hormone replacement therapy on breast cancer risk have been studied
extensively, but few firm conclusions may be drawn. Overall, there is
no convincing evidence of an increase in breast cancer risk in women
who have ever used oral contraceptives (181). However, some studies
have suggested that the long-term use of oral contraceptives in young
women before first birth may increase breast cancer risk (182, 183).
Two recent meta-analyses of the effect of estrogen replacement therapy
demonstrate small but statistically significant increases in risk for
users (184, 185). However, Steinberg et al. (184) noted no
increase in risk until after at least 5 yr of estrogen use, after which
a proportional increase in risk for each year of estrogen use was
observed, while Sillero-Arenas et al. (185) did not observe
a significant association between duration of hormone replacement
therapy and breast cancer risk. In summary, although hormonal risk
factors are clearly implicated in the pathogenesis of breast cancer,
most of them are associated with a RR of 3 or less of breast cancer
development (Table 4), and the presence
of a single hormonal risk factor is insufficient to classify a woman as
high risk.
The relationship of benign breast disease to breast carcinoma was a
subject of confusion for many years. The use of a standard
classification of benign breast diseases as nonproliferative,
proliferative, or proliferative with atypia has resolved much of the
controversy. The histological diagnoses comprising these categories are
shown in Table 5. Nonproliferative
disease is associated with no increase in breast cancer risk, while
proliferative disease increases risk by a factor of 1.5–2.0, and
atypical hyperplasia by a factor of 4–5. Approximately 70% of
palpable breast masses contain nonproliferative disease (186), and only
3.6% are atypical hyperplasia. The incidence of atypia is somewhat
higher in biopsies performed for mammographic lesions, ranging from
7–10% (187, 188). However, the risk of breast cancer development 15
yr after a diagnosis of atypical hyperplasia is only 8% in the absence
of a family history of breast cancer. Proliferative breast disease is
also noted more frequently in women with a significant family history
of breast cancer than in controls, further supporting its role as a
risk factor (189).
Another benign breast lesion that is clearly associated with an
increased risk of breast cancer development is lobular carcinoma
in situ (LCIS). In the past, LCIS was thought to be a
malignant lesion, albeit one with a favorable prognosis. However, the
finding that LCIS is associated with a risk of breast cancer
development of approximately 1% per yr, the observation that the risk
of carcinoma is equal in both breasts, and the finding that neither the
extent of LCIS in the breast nor its presence at a margin of resection
influence the risk of subsequent cancer have led LCIS to be regarded as
a risk factor for breast cancer development rather than the actual
precursor of carcinoma (190).
A number of environmental factors have also been linked to breast
cancer risk. Exposure to ionizing radiation, whether secondary to
nuclear explosion or medical procedures, has been clearly demonstrated
to increase breast cancer risk (191, 192, 193). The level of risk varies
with the age of exposure, with a minimal increase in risk observed for
exposure in women older than 40 yr. A larger amount of attention has
been directed toward the role of diet in the etiology of breast cancer.
This link has been suggested by the large international variation in
breast cancer incidence rates and the observation that national per
capita fat consumption correlates with breast cancer incidence and
mortality (194). However, prospective studies of diet and breast cancer
risk have failed to identify a relationship between dietary fat intake
and breast cancer incidence for up to 10 yr of follow-up (195). The
lack of a relationship between dietary fat intake and cancer risk
within the context of a Western diet is confirmed by a pooled analysis
of seven cohort studies involving a total of 337,816 women, which
demonstrated no difference in risk for women with the lowest and
highest quintile of fat intake (196). However, all of these studies
have addressed fat intake during adult life, and they do not exclude
the possibility that fat intake during childhood and adolescence may
influence subsequent breast cancer risk.
Stronger evidence exists to support an association between alcohol and
breast cancer. A meta-analysis of 12 case control studies demonstrated
a RR of 1.4 for each 24 g of alcohol consumed daily (197).
Defining a relationship between age of alcohol consumption and breast
cancer risk is more difficult, with conflicting data on the importance
of drinking early in life (198, 199). A summary of the magnitude of
increase in risk associated with the factors discussed is provided in
Table 4.
A. Interactions among risk factors
A major problem in the clinical identification of the "high risk
woman" is the lack of knowledge of the interactions among the various
factors known to alter breast cancer risk, since the majority of
studies have focused on defining individual risk factors. Most women
have a combination of factors that both increase and decrease risk,
complicating the assessment of an individual’s level of risk. In
addition, it is unclear whether the risk conferred by multiple risk
factors is additive, multiplicative, or varies with the risk factor
under study.
The interactions between a family history of breast cancer and other
risk factors have been examined, often with conflicting results. Dupont
and Page (186) observed that the combination of atypical hyperplasia
and a family history of a first-degree relative with breast cancer
increased the RR of breast cancer to 11 times that of an index
population, compared with a RR of 4.4 for atypia alone. However, Rosen
et al. (200) found that the presence of a family history of
breast carcinoma did not alter the level of risk after a diagnosis of
LCIS, a lesion often considered part of a continuum with atypical
hyperplasia. An analysis of data from the Nurses Health Study (201)
found that in women with a mother or sister with breast cancer, known
risk factors of age at menarche or menopause, parity, age at first
birth, alcohol use, and the presence of benign breast disease did not
further alter risk. In contrast, Anderson and Badzioch (202) and
Brinton et al. (203) have reported that hormonal factors
further modulate risk in women with a family history of breast cancer,
although the effect varies with the factor under study.
Studies of the interaction between estrogen replacement therapy and
other known breast cancer risk factors also have variable results,
depending on the risk factor under study. In a meta-analysis of 16
published studies, Steinberg et al. (184) found that the
effect of estrogen replacement did not differ among parous and
nulliparous women and those with or without benign breast disease.
However, an enhanced risk was observed in women with a family history
of breast cancer. The analysis of the interaction among risk factors is
further complicated by the fact that some factors may be important for
the risk of premenopausal, but not postmenopausal, cancer and
vice versa, and these effects may not be constant over time.
A model to predict the risk of breast cancer development in women at a
given age over a defined time interval was developed by Gail et
al. (204) using data from 4,496 matched pairs of cases and
controls in the Breast Cancer Diagnosis and Demonstration Project. The
model incorporates the risk factors of age at menarche, age at first
live birth, number of first-degree relatives with breast cancer, and
number of previous breast biopsies, and has been shown to predict risk
accurately in two validation studies of women undergoing annual
mammographic screening (205, 206). However, the model overpredicts
breast cancer risk by 33% among women age 60 and younger who do not
undergo annual screening. There are several other limitations of the
model. Because only first-degree relatives are considered, it is not an
appropriate model for women with extensive family histories of breast
cancer, where risk may be underestimated. In women with risk due to
LCIS or atypical hyperplasia, the model underestimates risk, since the
highest RR for breast biopsy is 2.0. Similarly, for the woman with
nonproliferative disease, the model may overestimate risk. In spite of
these limitations, the model is a clinically important tool for
identifying a woman’s level of risk over a clinically relevant time
period, after correction for competing causes of mortality. The Risk
Disk, which is available from the National Cancer Institute, uses the
Gail model to provide a numeric estimate of a woman’s 5 yr and
lifetime risk of developing breast cancer compared with an "average
risk" woman of the same age. Examples are given in Figs. 6 and 7.
View larger version (20K):
[in this window]
[in a new window]
|
Figure 6. Risk assessment for breast cancer in a woman with
two risk factors. The woman illustrated here has a mother with breast
cancer and has never had any children. The combination of these factors
means that her 5-yr risk of breast cancer development is 1.6%,
compared with 0.5% in a woman with no risk factors. If she lives to
the age of 70, her risk will be 18% compared with 6.4% for the woman
with no risk factors. This level of risk would not have qualified the
woman to participate in the recently completed Breast Cancer Prevention
Trial with tamoxifen.
|
|
View larger version (22K):
[in this window]
[in a new window]
|
Figure 7. Risk assessment for breast cancer in a woman with
multiple risk factors. The woman illustrated here has early menarche,
late age at first birth, and has a mother and sister with breast
cancer. In addition, she has had a breast biopsy showing atypical
hyperplasia. This combination of risk factors makes her 5-yr risk of
breast cancer development 7.5% and her lifetime risk 53.4%. She is an
ideal candidate to consider tamoxifen for risk reduction.
|
|
B. Identification of candidates for chemoprevention
Women at increased risk for breast cancer would seem to be ideal
candidates for chemoprevention initiatives. However, from the preceding
discussion it is apparent that the problem of identification of the
high-risk woman is far from solved. There is no consensus regarding
what level of increase in risk is clinically relevant. The interactions
among risk factors and their variability over time are poorly
understood, and most of the data on risk come from studies of white
women, so little is known about the impact of ethnic diversity on risk.
Finally, with the exception of women with mutations of breast cancer
predisposition genes, the majority of women with risk factors will not
develop breast carcinoma. In addition, a recent study of the fraction
of breast cancer cases in the United States due to attributable risk
factors (207) found that fewer that 50% of women who develop the
disease have any identifiable risk factors. A family history of breast
cancer accounted for only 9.1% of cases, while relatively minor risk
factors such as later age at first birth and nulliparity contributed
29.5% of cases. In a similar study, Seidman et al. (208)
noted that only 21% of breast cancer cases in women age 30–54 and
29% of cases in women age 55–84 occurred in women with 1 of 10 common
breast cancer risk factors. The majority of women in the studies
described had minor risk factors, which increased the RR of breast
cancer only 2-fold, and most had only a single risk factor. This level
of "increased risk" would not meet the entry criteria for the
trials of breast cancer prevention in high-risk women discussed below.
These data suggest that even if women with a very small increase in
breast cancer risk were targeted for prevention initiatives, a large
number of cases would continue to be missed.
|
VII. Prevention of Breast Cancer with Tamoxifen
|
|---|
This section will explore progress that has been achieved in the
last decade to answer the question, "Does tamoxifen have worth in the
prevention of breast cancer in high risk women?". Two studies have
claimed to address this question—The Royal Marsden Pilot Study (209)
and the National Surgical Adjuvant Breast and Bowel Project
(NSABP) Protocol P-1 (141). However, the Marsden study was not
designed to answer the question about the prevention of breast cancer.
It was a pilot toxicology study (142) that was subsequently part of a
nationwide clinical trial in Britain that planned to recruit a total of
20,000 high-risk women. The main British study is ongoing.
Additionally, an Italian report of the efficacy of tamoxifen in a small
number of low-risk women (
5,000) has been published (210), but again
this is a small component of a 20,000-volunteer trail that has now been
stopped.
A. Royal Marsden Pilot Study
Powles and co-workers (211) recruited high-risk women aged 30–70
to a placebo-controlled trial using 20 mg of tamoxifen daily for up to
8 yr. Women were eligible if their risk of breast cancer was increased
due to family history. Each participant had at least one first-degree
relative with breast cancer under age 50, or a first-degree relative
affected at any age plus an additional affected first- or sec
Top
Abstract
I. Introduction
