Harvard Public Health Review Text Only Home
Last summer, Dr. Joann Manson,
MPH'84, DPH'87, was put in an unusual position. As a lead investigator for the
women's health initiative and the nurses' health study, she found herself involved
with two important studies that had produced apparently conflicting results
about postmenopausal hormone therapy (HT) and heart health.
In July 2002, the national Women’s Health Initiative (WHI) terminated its randomized controlled trial of estrogen and progestin because the health risks of the regimen clearly outweighed the benefits. Its most startling finding--that HT increases the risk of heart disease--seemed to contradict the conclusions of several Nurses' Health Study (NHS) reports. During the 1980s and '90s, that observational cohort and several others showed a significant overall reduction in heart disease risk among women taking HT. The end result is an epidemiologic puzzle that Manson and other researchers have only begun to untangle.
"It doesn’t
mean that the observational studies were wrong and the clinical trials are right,"
says Manson, who is a professor at both the Harvard School of Public Health
and Harvard Medical School. "Observational studies have been tremendously
helpful in generating hypotheses. They are often on target, but in this case
there’s a divergence, and we need to get a better understanding of why
that is."
A number of
School-affiliated scientists are hot on the trail of this understanding. Manson
joined Assistant Professor of Epidemiology Francine Grodstein and a non-NHS
researcher in penning a February "Sounding Board" article in the New
England Journal of Medicine outlining some of the possible reasons behind the
disparity. Meir Stampfer, chair of the School's Department of Epidemiology and
a NHS investigator, together with Grodstein, organized a major symposium on
campus earlier this year to explore the scientific issues raised by the hormone
debate. And Dr. Robert Hoover, SD'76, who made the first link between hormone
therapy and breast cancer, has been responding to many of the same questions
as director of the Epidemiology and Biostatistics Program at the National Cancer
Institute (NCI). Their combined efforts are not only critical for women's health
but have enormous implications for the design and interpretation of epidemiologic
studies in general.
Manson has
described the theory that estrogen replacement therapy might protect women from
heart disease as a "beautiful hypothesis" generated by the existence
of several viable biological mechanisms related to estrogen and the observation
that women, prior to menopause, seem to be relatively protected from heart disease.
The theory held up when examined through the lens of the NHS, and other studies
indicated possible means for the protective effect, including improved cholesterol
profiles. But ultimately, those findings didn’t translate into protection
from coronary heart disease (CHD) in the WHI.
Often the results
of a randomized controlled trial like the WHI would transcend the results of
an observational study like the NHS outright--and indeed many doctors and patients
have taken its conclusions to heart. But the NHS is not your typical exercise
in epidemiology. Researchers believe that nurses, as health workers, are more
diligent and precise than the average person in responding to questionnaires
and are also more motivated to participate in a study long term. So NHS findings
often carry more heft with clinicians and policymakers than other studies of
its kind. And more often than not, the results of observational studies correspond
with subsequent clinical trials. For many other endpoints--including fractures,
stroke, pulmonary embolism, and breast and colon cancers--the NHS predicted
the results of the WHI (see chart on page 11). So why the conflict when it came
to CHD?
In looking
back, Manson points out several possible biological explanations. Many of the
NHS subjects used estrogen alone, while the WHI looked at women taking estrogen/progestin
combination therapy. In addition, the two populations may have been different.
Compared with WHI subjects, women in the NHS had a lower body-mass index and
tended to begin hormones at the time of menopause. Methodologic limitations
may have played a role as well, she says. The observational studies may have
suffered from a confounding bias--a failure to control adequately for lifestyle
or socioeconomic factors, as well as an inability to capture short-term effects
of HT. "Every effort should be made to collect information about possible
confounders and to control for them carefully in mathematical models,"
she observes. "But sometimes it is very difficult to control for everything."
Stampfer and
Grodstein agree that biology, along with timing, likely contributed to the difference
in the two studies. If the women using hormones in the NHS were healthier in
ways that could not be controlled for, the data would also have indicated a
lower stroke risk. Instead, Grodstein documented a higher stroke risk among
HT users in the NHS, a finding that suggests that confounding is unlikely. On
the other hand, all three researchers note that the NHS subjects began HT near
the onset of menopause, which usually occurs around the age of 50. In contrast,
about two- thirds of WHI subjects were over the age of 60 when they entered
the study. With that in mind, Stampfer points to animal studies that have shown
hormones to protect newly menopausal monkeys from CHD--a protection that failed
to appear when therapy was delayed.
So the result
may be due to the fact that younger women are less likely to have existing coronary
artery disease, Stampfer says. "Hormones increase the production of enzymes
that make plaque in the coronary arteries less stable and can cause an acute
heart attack," he observes. "If you don’t have plaque, it may
not have a bad effect." Studies similar to the WHI in women with preexisting
heart disease lend support to this notion, finding that hormone therapy led
to more heart attacks, not fewer. Likewise, in the NHS, Grodstein found that
women with heart disease who used HT were initially at higher risk for subsequent
heart attacks.
Stampfer remarks
that each research finding must be weighed both on its own merits and in light
of other relevant data. It may mean that hormones protect some women and are
too risky for others, but it is very difficult to extrapolate findings from
one specific group to other populations. Manson agrees: "One of the lessons
here is that researchers need to be very cautious in how they present the results
of their findings, noting the strengths and limitations of the research."
Still, there
are times when those acting on scientific information have marketing rather
than methods in mind. Scientists repeatedly cite the over-promotion of HT by
the pharmaceutical industry as one reason women and their doctors believed it
could prevent a range of illnesses, including CHD. A 1992 review of 109 HT ads
in medical journals found that 40 percent included "unbalanced" information
on the drugs' efficacy and side effects, according to a study in the Annals
of Internal Medicine.
Since his time
at the School, alumnus Robert Hoover has been doing his part to bring a bit
more balance to the topic. For his doctoral thesis, he tracked 1,500 women who
had started taking estrogen in their mid-40s. The result was the first study
to make the link between breast cancer and estrogen, which was published in
1976 in the New England Journal of Medicine and "sparked a whole lot of
enthusiasm to go out and do similar studies in larger populations," he
says.
Those studies,
which came on the heels of research linking estrogen to uterine cancer, convinced
a lot of women to stop taking the drugs. But his role did not end there. While
a combined estrogen/ progestin therapy seemed to eliminate the uterine cancer
risk, Hoover helped determine that the breast cancer risk of that regimen was
even higher than with estrogen alone. During a FDA hearing on HT cancer risks
in the late 1980s, a participant would make note of his ongoing role in the
hormone controversy. "One of the drug company representatives got up to
criticize the study and said, 'Whenever we see a study that implies HT is bad,
somewhere in the authorship there is this name: Hoover,'" the NCI researcher
recalls. At the same time, biological and epidemiologic studies indicated that
the drugs might protect against heart disease. "The real question was:
do the benefits outweigh the risk?" Hoover says. “That was the driving
factor for the WHI."
In light of
the study's recent results, it would appear that they don't--at least for women
who start hormone use later in life. But that hasn't stopped Hoover, who still
works on questions regarding hormones and cancer. (He helped launch the NCI's
Cohort Consortium, which will use data and blood samples from a number of large
studies--including the NHS. The consortium's first project will examine the
link between gene/hormone interactions and breast and prostate cancers.) Today,
however, women are advised to take HT only to relieve significant menopausal
symptoms and for as short a duration as possible. "It is one of the most
dramatic sea changes in clinical medicine that I've ever seen," says Manson.
The good news is when it comes to the heart, we have alternatives, according to Stampfer. "The main take-home message is that we have plenty of ways to prevent heart disease without having to resort to hormones," he says, estimating that about 80 percent of CHD could be prevented by simple changes in lifestyle. "In contrast, we have few ways to lower the risk of breast cancer. So, based on all our past research, avoiding long-term use of estrogen plus progestin makes a lot of sense."
Tinker Ready