Reported by Nancy Nelson
July 8, 2002
Does smoking cause lung cancer? Do high-fat diets increase the
risk of breast cancer? Is arsenic in water linked to cancer? These
are the kinds of questions that epidemiologists try to answer. In
this month's Benchmarks, we talked to one of NCI's leading epidemiologists,
Dr. Robert N. Hoover, about epidemiology in general, and about the
research in NCI's Division of Cancer Epidemiology and Genetics (DCEG).
DCEG is the primary focus within NCI for population-based research
on environmental and genetic determinants of cancer.
Robert N. Hoover, M.D., Sc.D., is the director of DCEG's Epidemiology
and Biostatistics Program. Dr. Hoover serves on many national and
international committees concerned with various aspects of epidemiology
and preventive medicine. He was awarded the Public Health Service
Commendation Medal in 1976, the Meritorious Service Medal in 1984,
and the Distinguished Service Medal in 1990. In 1996, he received
the Gorgas Medal for distinguished work in preventive medicine from
the Association of Military Surgeons of the United States. His most
recent awards include the Distinguished Service Award, DES Action,
1999; Public Health Service Physician Researcher of the Year, 2001;
The John Snow Award from the American Public Health Association,
2001; The Distinguished Achievement Award from the American Society
of Preventive Oncology, 2002; and The Harvard School of Public Health
Alumni Award of Merit, 2002.
What are the current research priorities of your division?
Dr. Hoover: Historically, we have built a program which is both very
broad and very deep. Because we are the National Cancer Institute,
we don't have the luxury of focusing on only one disease or one
exposure. We are expected to have a broad program that encompasses
all the major areas that are thought to be interesting and fruitful
in the epidemiology of cancer. So we have large programs in genetics,
occupation, radiation, nutrition, hormones, viruses, biostatisticsour
programs cover the waterfront.
At the same time, we try to be at the cutting edge of research
in each of these programsto have something going on that will
move the field forward. We try to take advantage of the opportunities
of the time. The current opportunities are in interdisciplinary
studies, biochemical or molecular epidemiology. Advances have been
made in the laboratory during the last 10 or 15 years and there
are new ones almost every day that are very useful in our kind of
work. These allow us to identify people who are exposed to specific
agents and people who may be susceptible to diseases, or to get
new insights into mechanisms involved in carcinogenesis. There's
a lot of enthusiasm and a lot of opportunity for combining robust
epidemiology research and advanced laboratory tools to take advantage
of both disciplines. These new tools influence each of our programs.
What's an example?
Dr. Hoover: There's a whole lot of interest in environment-gene interactions,
identifying genetically-susceptible states, and using genetic alterations
to discover previously unrecognized or unappreciated environmental
exposures. We're involved in a major attempt to do this right now
with breast cancer and prostate cancer in something called the Cohort
Consortium. This program is an attempt to put together several of
the large follow-up studies in the worlda couple at NCI, some
from Harvard, some from Europe, some from Southern California, and
the American Cancer Society. All of these cohorts have information
on exposures and the development of disease, as well as blood specimens.
In total, there are 7000 cases of breast and 7000 cases of prostate
available for study. We're going to focus on an area which has been
known for a long time to influence these diseaseshormones.
We're going to use the opportunity to look at genes and hormone
exposures together to try to learn something useful in terms of
prevention. We want to find out how these hormonal exposures might
work, what the real levels of the hormones are, what dictates the
levelsincluding exogeneous sources like drugs, as well as
endogenous sources. Also, we want to know how they interact with
each other and with growth factors, and basically try to understand
the genetic and environmental determinants of these two major cancers.
This is indicative of the kinds of things we are doing throughout
the program.
Another example is a bladder cancer study in New England. This
is an interdisciplinary study looking for clues to the high rates
of bladder cancer in New England. We noticed many years ago in doing
cancer maps that bladder cancer concentrates in major urban areas.
This is what we expected given the relationship between tobacco
use and occupational chemical exposures and risk of this malignancy.
However, we felt that this wouldn't explain the very high rates
also seen in rural areas of New England. This has been a source
of concern for some time, and we've been searching for a hypothesis
that we can pursue.
Then some other work we did for bladder cancer and some work from
Taiwan suggested that drinking water may be a candidate. It's something
that both sexes consume and widespread exposures, even if they were
associated with relatively small increases in risk, could influence
general population rates. These studies suggested that arsenic could
cause bladder cancer. Arsenic can leach into the ground water from
granite sources, which are abundant in New England. This has provided
us with a hypothesis. We are working with Dartmouth University,
the state of New Hampshire, the state of Vermont, the state of Maine,
the U.S. Geological Survey and anyone that can help us understand
water availability patterns. We are assaying arsenic in drinking
water from wells and community water supplies. We're looking at
a battery of potential susceptibility genes, along with other risk
factors that may have a role, such as occupational exposures and
tobacco. So, it's a pretty ambitious undertaking. Having the tools
to measure exposure, as well as susceptibility, makes us excited
that we can evaluate at least some of these hypotheses.
On your web site, you state that the research philosophy
of your division is to serve as a "national program for population-based
studies to identify environmental and genetic determinants of cancer."
How does being a national program affect the focus of your work?
Dr. Hoover: Looking historically around the world, epidemiology has
always had a strong presence in governments, and particularly in
national governments. The primary reason for that is that it's concerned
with epidemics, which is a concern of government. But, there is
another reason as well. In terms of research, epidemiology is different
from laboratory researchyou can't just do it anywhere. You
need populations--large populations--that have undergone natural experimentsthat
have had unusual exposures, or have high or low cancer rates. A
national perspective is helpful in responding to these problems
and opportunities. We have access to national data resources, such
as the National Center for Health Statistics, population-based tumor
registries, the social security administration, medicare, and agencies
that collect large amounts of data that can be used for epidemiological
ends. We can move easily to where the action is. We can look, for
example, at lung cancer in Glynn County, Georgia, or colon cancer
in David City, Nebraska. We can and have gone to these and other
places, found collaborators in local health officials and initiated
studies to investigate high rates of diseases in those areas. It's
relatively easy for us to do that.
You can probably afford to take scientific risks?
Dr. Hoover: Yes. We are encouraged to do things that would be relatively
risky for academics to doperhaps a five-year study that has
a high probability of coming up with nothing, but if it did come
up with something, it would be extremely important for public health.
Oftentimes we're put in those roles. We have the stability to do
that and don't have to stake anyone's entire scientific future on
one particular study.
DCEG's research program is international in scope. Where
and why is DCEG carrying out epidemiology research around the globe?
Dr. Hoover: We are an international program for a couple of reasons.
We are in the federal government and we get involved with bilateral
and other nation-to-nation, government-to-government agreements.
We were some of the first U.S. scientists in China after president
Nixon's opening up of that country, and our relationship with the
Chinese National Cancer Institute was one of the first scientific
bridges to that country. We were prominent in détente initiatives
with the Soviet Union and Soviet bloc nations, and we currently
have a role in the Middle East Cancer Consortium, and in forming
the Ireland-Northern Ireland-NCI Cancer Consortium.
Because of our position, we have an opportunity to take advantage
of major disease differences that occur throughout the world. We
do work in an area of China, where 25 percent of the population
develops esophageal cancer. We do work in Latin America, where the
cervical cancer rates are the highest in the world. We also take
advantage of international differences in availability of resources.
For instance, we collaborate often with Scandinavian countries,
which have record linkage resources, and we use these records as
resources for testing epidemiology observations. Many countries
have single-payer health care systemsi.e., the governmentwhich
makes it a lot easier to do health investigations because you're
dealing with only one health care delivery system.
Right now, we're doing a very large study of benzene workers in
China. We're trying to investigate what levels of benzene are related
to a variety of cancer risks. Many factories in China use benzene
in a variety of activities, so there's a range of exposures and
there's a great collaborative spirit to allow us to measure levels.
We've also done a variety of studies of cervical cancer in Latin
America. Our work there allowed us to identify the major role of
human papillomaviruses (HPV), as well as the specific subtypes of
HPV responsible for the disease. Because of an extremely productive
research collaboration with investigators in the government of Costa
Rica, we are launching a major vaccine trial with a candidate vaccine
against HPV developed by NCI researchers.
Are there any fundamental differences in design between epidemiological
studies focused on genetic factors and those concerned with environmental
factors?
Dr. Hoover: For the most part, the methods in human populations are
quite similar. There are some specific methods that are different.
For example, family studies are used to identify high penetrance
genes-the genes that are responsible for familial cancers. But,
these kinds of studies are not very productive for studying environmental
exposure because most families tend to have the same environmental
exposures. Outside of that, trying to evaluate the impact of natural
genetic variation in the population on cancer risk is very similar
to looking at environmental exposures. It's essentially another
exposure--you are exposed to a particular version of a gene.
One difference, however, is a difference in scale. In environmental
studies, we worry about confounding exposures. So, if you're investigating
coffee drinking, you have to control for the effects of tobacco
because they usually correlate. But, there are usually a limited
number of confounding factors that you know about or that you can
assess. With the new genetic technology, there is a new possibility
that if you're interested in a particular gene, there may be thousands
of others whose effects you may be called upon to control for. This
does serve up a kind of challenge of scale that we haven't seen
before. In general, however, the principles are largely similar
and transferable.
What are the priorities for future epidemiological studies?
Dr. Hoover: Epidemiology is an opportunistic science. It goes where
the action is not only in terms of disease and exposure, but also
where the tools are. Many epidemiologists are most anxious to use
the new molecular tools to assess exposures better, as well as measuring
susceptibility. For example, it would be wonderful to have a biological
dosimeter for your exposure to benzene from gasoline fumes, or your
lifetime level of consumption of fat in your diet. It's difficult
to get at these kinds of exposures by asking questions. We are hopeful
that the emerging technology from measurement science will provide
opportunities in this area. That is an area all of us hope will
come to fruition.
You mean that by assaying blood or DNA samples you can estimate
what someone's exposures have been?
Dr. Hoover: Correct. For example, in the area of diet--dietary fat
exposure: People have difficulty giving an accurate assessment of
their diet last week, let alone their lifetime exposure. Getting
an assessment of lifetime dietary fat exposure is extremely difficult
to do with current tools. However, if there were some change in
your metabolism or some other biologic alteration that would reflect
your long-term exposures to specific nutrients, that would be extraordinarily
useful. Right now, it's mostly talk.
It's part of one of the extraordinary opportunities in the bypass
budget. We're trying to encourage people who work in those areas
of technologies. We need close collaborations between people working
in basic science, chemical carcinogenesis, nutrition, and technology
development in order to make progress.
What are some of the landmark epidemiological studies that
have been published during your career?
Dr. Hoover: Most of what we know about human cancer has come first
from studies in human populations--either clinical observations or
epidemiology studies. It wasn't until the late 1960s and early 1970s
that the whole range of radiation-related solid cancers (other than
leukemia, which was known earlier) began to be recognized, primarily
from a whole series of outstanding work done with atomic bomb survivors.
Then there was a series of discoveries linking infectious agents
to cancer--Hepatitis B and liver cancer, HPV and cervical cancer,
HTVL-1 with leukemia and lymphoma. In the occupational setting,
in the late 1960s, there was a famous study showing increased lung
cancer rates in steel workers working near coke ovens. This was
followed by studies showing the leukemia risks of benzene exposure,
and a number of studies indicting a variety of other occupational
carcinogens, even up to the recent studies of dioxin. And in the
area of drugs, one of the most important findings for public health
was understanding the role of HRT and breast cancer risk. Another
important study showed the connection between immunosuppressive
drugs and non-Hodgkin's lymphoma. Through the study of Li-Fraumeni
families, Joe Fraumeni, Fred Li, and Steve Friend used the tools
of the genetic revolution to give us insights into the role in humans
of the tumor suppressor gene p53, which is relevant to a whole variety
of cancers. In the area of genetics, other noteable advances included
the discovery of BRCA1 & 2 and a whole series of cancer susceptibility
genes.
How do the challenges of doing epidemiology research today
compare to when you first arrived at NCI in 1972? Do the research
tools available today make it easier to do research? What are some
of the problems?
Dr. Hoover: There's been a lot of advancement in measuring exposures
and in ways of collecting history of exposure. Also, all of the
current molecular tools and ways of measuring things in blood and
other biologic specimens make it easier to do research today. There's
also been a whole series of statistical advancements both in our
ability to sample populations and to analyze data. The tools of
the trade have gotten a lot better and continue to improve.
On the other hand, it's been more difficult to do some aspects
of epidemiology because of changes in our culture. In the past,
it was relatively easy to get 80 percent of the normal population
to cooperate in studies as controls. It's extremely difficult now.
People are so assaulted with telemarketers and fund-raisers that
researchers become painted with that brush when we try to call on
the phone, send letters, get answers to questionnaires, get people
to give blood and, generally, do things for medical research. People
are less responsive. In addition, the burgeoning concern over privacy
and confidentiality issues make it much more difficult to do these
studies now than in the past.
Also, historically there has been a focus on risk factors that
are associated with major increases in risk. For example, we know
that tobacco gives you a 10-fold risk of lung cancer and HPV a several
hundred-fold increased risk of cervical cancer. More recently, much
work has begun to focus on a concern on the part of scientists that
many risks may be the sum total of several minor risks--on the order
of 10 percent or 20 percent, rather than 1000%. These are tough.
It's tough to measure the relevant exposure, to distinguish high
vs. low exposures. And the result you get could be due to confounders,
whose effects are much more difficult to exclude when dealing with
relatively low levels of risk. There's been a whole lot of interest
and a whole lot of statistical work in small risks, but it still
remains a major challenge for public health.
How do epidemiology results change public health? How cautious
does the public need to be in interpreting preliminary results from
epidemiological studies?
Dr. Hoover: For better or for worse, epidemiology results generally
get out to the public quickly because they're understandable. They
don't use arcane language; they deal with real people and real exposures
that real people get. Therefore, they are instantly relevant and
recognized. As happens in science in general and in medicine in
particular, there is an increasing interest on the part of lay press
to move information to the general public quickly and in great volume.
It is no longer the case that you can publish something that you
would like to see replicated two-three times before signaling any
public health concern. That's difficult to do these days because
once you publish, it's out. It's of interest. The concern that many
people have is that because of this there is lack of discrimination
for what is a solid finding vs. not-so-solid finding, or what are
ones that do and do not deserve action. This applies to all medical
science. It applies to findings in petri dishes for new cancer treatments
but, in fact, it's more prevalent in epidemiology because it's instantaneously
recognized as a real people finding.
We're not going to change the openness of society, or the interest
of people in these issues or the aggressiveness of media. We're
hoping that education will out--the education of the media on reporting
the finding, the education of scientists on how to report, the education
of the public on how to interpret and deal with the avalanche of
information. Certainly there are misrepresentations of the solidness
of findings. The responsible parties are the scientist, the public,
and the media. There are opportunities for misrepresentation at
each of those levels.
I think that this is not all as bad as some people paint it. I
think it's a process of learning how to communicate the information
better. And the public is learning. I have found in my career describing
the relationship between menopausal hormones and breast cancer that
when I have the opportunity to speak to people (either the general
public or reporters on the phone), everyone is smarter than we give
them credit for. They understand the level of evidence and what
the choices are when you take some time to talk with them. They
can handle uncertainty and the scientific method if it's delivered
to them in an understandable manner. I have a lot of faith and reason
to believe that the public can deal with information if it is delivered
in a reasonable way.
Dietary fat and breast cancer. Where are we now with that
issue?
Dr. Hoover: We've been all over the map with that. There is still
overpowering descriptive data that diet probably is involved in
a substantial amount of breast cancer risk. However, very little
is known about what specific aspect of the diet is involved. The
sum total of the data currently doesn't favor a major role for dietary
fat, but certainly over-interpreted data could lead to that conclusion.
The initial data were international rates: U.S. and Western Europe
have high rates of breast cancer; Japan and China have low rates
of breast cancer; therefore dietary fat is the cause. That kind
of loose thinking led to a lack of distinction about levels of evidence.
In the late 1940s, there was concern about the cause of the major
increase in lung cancer. One very prominent hypothesis was that
the cause was asphalt roads. There certainly was a correlation.
There were a lot of roads where the lung cancer rates were high.
But correlations aren't the same as causes. Tobacco was shown to
be the cause.
Epidemiology has often been considered the field of research
that can suggest causation, whereas clinical studies and laboratory
experiments are really needed to prove causation. What do you think?
Dr. Hoover: That's clearly false. It depends on your definition of
proof. The weight of the evidence from human studies can achieve
a level where people are willing to act on the fact that it's likely
to be a cause, and the result of those actions are the ultimate
test of that causality. I have a very simple definition of cause
myself. That is: If you remove the exposure, the disease goes away,
and if you introduce the exposure, the disease appears.
Now, you can do that in a randomized trial, for example, giving
drugs which are thought to be chemoprevention agents. Or you can
say the weight of the observational evidence is sufficient, such
as the evidence that tobacco is related to a large number of cancers:
the weigh of the evidence is so high you don't need a trial--you
need to get rid of the exposure. Remarkably, as we got rid of the
exposure, the rates for these cancers started plummeting. Does this
mean that I understand the mechanism by which tobacco smoke produces
cancer in humans? No, it doesn't because I don't. And actually,
neither do my laboratory colleagues, although they have a myriad
of hypotheses. But probably neither they nor I need to know that
in order to accept causality.
It's also possible that good, solid laboratory findings can suggest
that an exposure might protect against cancer, and human observations
might determine that that's not the case. On the other hand, good
solid human epidemiology can suggest that something is a carcinogenic
exposure, and then interventions determine that that's not the case.
Are there any examples where epidemiology suggested doing
X and you did X and the rates didn't go down?
Dr. Hoover: Yes, beta carotene [a vitamin A precursor] and the lung
cancer prevention trial. The observational studies suggested that
beta carotene would help prevent lung cancer in people at high risk
for the disease, but the intervention study showed that beta carotene
resulted in more lung cancer cases. And we certainly have plenty
of these examples coming from the laboratory--a suggestion that something
is bad or good but we can't demonstrate it in the population.
But you know, I think it's probably not useful to make the distinction
between the various types of research. We actually look for the
weight of the evidence. If we have good human evidence and very
solid supporting experimental or laboratory evidence, we feel a
lot better than if we just have good human evidence, or just good
laboratory evidence. So we really look for the sum total of the
relevent scientific evidence and make a causal, or at least a practical
causal decision based on the sum total. And I think clinical research,
basic science research, and epidemiologic research all contribute
to that--the weight of the evidence.
Do epidemiologists come from a wide variety of backgrounds?
What are some of the common career paths for epidemiologists?
Dr. Hoover: There's been a change in the discipline over my career,
certainly. When I was first in training, epidemiology was almost
exclusively medical epidemiology and epidemiologists were physicians.
And, in fact, many training programs were restricted to people with
medical degrees. There's been a very healthy trend in the discipline
in the past 30 years towards taking people with a wide variety of
backgrounds and teaching them the epidemiologic method and then
allowing them to use their other background in concert with that
method. Probably the first were the statisticians. Now we have anthropologists,
we have sociologists, we have molecular scientists, we have biochemists--you
name it. We even have a biophysicist-epidemiologist in our own program.
It's actually led to what I think is an extraordinarily healthy
mix in the discipline. When you deal with teams, which is usual
in epidemiologic research, you have complimentary training. You
can have a physician-epidemiologist, a statistician-epidemiologist,
and a behavioral scientist-epidemiologist working on the same project.
I actually think you get a much better product that way.
What would you say are the main tasks of the epidemiologists
working in your division? What takes the most time--designing studies,
collecting data, or number crunching?
Dr. Hoover: Epidemiologists do all those elements. In terms of time,
because epidemiological studies are typically measured in years,
not weeks, you spend more time developing and monitoring field work
than you do any other component. But the other components are critical
to everything. You need time to develop the ideas, and find out
what's practical. Then you have to execute the study and finally
make sense of it by the analysis and interpretation. When people
come here as fellows, I tell them to have an eclectic research repertoire
at different stages; have some studies in the field, some in the
development stage, others in the analysis stage, all at the same
time. This is to keep your sanity, as well as to continue the learning
cycle. You do better field work when you understand the problems
in the analysis that were caused by bad field work. If you know
the practical constraints, you do a better job of designing the
study. I encourage them to be involved in different stages and keep
that mix throughout their careers.
Do people go to Iowa, for example, for three months to do
a study?
Dr. Hoover: We pioneered the nationwide conduct of studies. For the
most part, you travel to the site of study very intensely at the
very beginning to get cooperation of the local infrastructure and
the local medical establishment and to get things through committees.
You also want to insure that the setting up of the field study is
consistent with good research design. You set up systems that allow
you to monitor the study from Bethesda. Then, throughout the study,
the amount of time you spend there depends on how well the work
is going. If things aren't going well, then you'll have to go out
there. It is a constant challenge to maintain control and make sure
you know what's being done, and that it's done well.
A lot of our work is contract-based. What we try to do is have
individuals who are well-trained in field studies, contract with
them to do those specific activities, and have them hire local people
to do the work. That's turned out to be the most effective way of
doing the work.
It used to be that epidemiology was a cottage industry. In my training
you were expected to know the literature, get the idea for study,
develop the protocol, develop the questionnaire, hire interviewers,
train the interviewers, supervise the interviewers, code the data,
key in the data, write the programs to analyze data, and produce
the manuscript. It's probably not bad to do this in training. But
it is exceptionally inefficient and probably bad to do in your career.
What happens now is that the epidemiologist is the orchestrator
of all this. He or she is responsible for making sure that the key
elements of the design are implemented by individuals highly-trained,
experienced and good at each specific task.
What does it take to be a good epidemiologist?
Dr. Hoover: Curiosity is certainly one quality, and to be intrigued
by mysteries. It also helps in the daily conduct of the discipline
to pay attention to detail. Sometimes it's a difficult combination
to find; you need some sort of creativity and curiosity, but you
also need an extraordinary amount of discipline and attention to
detail.
What training do you think is important today for young,
aspiring epidemiologists? Does DCEG have an active training program?
Dr. Hoover: First and foremost, an epidemiologist needs to be well-trained
in epidemiology. Epidemiology is a method, a way of thinking. It
has a structure, and you don't just do it because you have had some
training in medicine or chemistry or you feel you can ask people
questions. It requires good solid training in epidemiology--didactic
training and then working closely with epidemiologists. And certainly
that's what we try to provide here. We have pre-doc, post-masters,
post-docs in various stages of their training and can insinuate
them in a variety of studies and have them work closely with other
epidemiologists and people in other disciplines to get a lot of
hands-on experience. That's what's most important. There is an increasing
interest in learning molecular science, so we have a molecular epidemiology
fellowship that allows people to spend time in labs. This gives
them an opportunity to see what is possible and where things may
go wrong. A lot of people like to sub-specialize in a particular
exposure--occupational, radiation, nutrition, etc. We have training
programs that allow someone to focus not only on the epidemilogic
method but on the subject-matter area of interest as well.
Do you have to have mathematical aptitude to do epidemiology?
Dr. Hoover: It is useful if you're quantitative. You don't have to
be a sophisticated statistician, but you have to appreciate the
quantitative side. One of my own mentors repeatedly informed me,
"if you can't count it, you don't know it."
Where do graduates go after they train here?
Dr. Hoover: They go all over--to industry, academia, state governments,
foreign governments, and regulatory agencies. Because you can apply
the method of epidemiology to many situations, you can work in research,
health care, regulation, litigation or many other areas.
### |