Why did you choose public health and preventive medicine as a career?

While going through medical school I was struck that we spent more and more time in our training treating heart attacks and lung cancers, as one example. We were left to treat the consequences of smoking, rather than think about the prevention of smoking. At the same time, I was interested in the role of policy in changing health care, and I actually stepped back and decided that prevention was where I wanted to put my energy. That ultimately took me to training in public health and from there I have pursued a career in epidemiology and the translation of epidemiological findings into public health recommendations.

  When did you join the Nurses’ Health Study, which provides the data for your research?

I joined in 1982 as a doctoral student. Walter Willett was my thesis advisor.

  Your most cited paper is "The use of estrogens and progestins and the risk of breast-cancer in postmenopausal women"(New England Journal of Medicine, vol 332, 1995). How did you come to work in postmenopausal hormones and cancer?

I started writing on oral contraceptive use and total mortality and then on balancing the apparent benefit of hormones on heart disease. I then wanted to look at postmenopausal hormones and how they’re related to the risk of breast cancer, and over the last 15 years I’ve led our analysis of those data.

  Why did the 1995 NEJM paper have such an impact?

It was probably the first clear report that current users of hormones were at increased risk, and that this risk was most clearly apparent with longer durations of use.

  Were the data unambiguous?

We saw clearly that there was increasing risk with increasing duration of use; we saw that after stopping use, the risk dropped back toward that of a never-user. We were able to control for history of screening and the results persisted. So in some sense, we could look to rule out the typical sources of bias that might affect the findings and so come to the presentation of the increasing risk with increasing duration of use that was most clearly evident among women who were currently or continuing to use postmenopausal hormones.

  Was this the expected result when you initiated the analysis?

The study was fundamentally driven by the hypothesis that when you add postmenopausal hormones to women after menopause, you’re elevating their circulating levels of hormones and this then starts promoting growth of breast cancer. Through the 1980s, the role of tamoxifen as an anti-estrogen to protect the breast against recurrence of tumors in women who had been treated for breast cancer became more clearly described through collaborative re-analysis of clinical trials, and there was growing evidence that obesity in postmenopausal women was related to their circulating levels of estrogens. There was also growing evidence that obesity was related to mortality from breast cancer. So the supporting evidence was more and more clear for the role of hormones in the etiology and progression of breast cancer, such that Ron Ross, Brian Henderson and Malcolm Pike, all from the University of Southern California, had a piece in Science that discussed the synthesis of knowledge on hormones and breast cancer. There was more and more of that coming together. We wanted to see if we could in fact separate out the effect from current use that would reflect ongoing exposure to higher levels of circulating hormones.

  What message would you want the public to take away from the research?

At some fundamental level, the message has to be that taking hormones is not a risk-free undertaking and therefore anyone contemplating the use of hormones, particularly long-term, really needs to stop and consider the risks and benefits. And the balance of those risks and benefits is going to vary depending on the profile of the woman. At some level, I think until our paper came out, the obstetrics and gynecology community had been saying, "All this discussion of hormones and increasing risk of breast cancer is just unnecessary worry. We really don’t need to talk to women about the potential risk." This paper, at some level, said "Time out: there is a strong suggestion here that long-term use increases risk and we need to go back and more carefully reconsider the balance of risks and benefits."

  Since postmenopausal hormones reduce the risk of heart attacks, wouldn’t the most important analysis be postmenopausal hormones and all-cause mortality?

Francine Grodstein in our group has published that analysis. In essence, what we saw was that current use was associated with a lower risk of total mortality but the benefit of use actually diminished with longer-term use. It didn’t all go away, but was certainly attenuated by the increasing risk of breast cancer with longer term use, and the increasing mortality from breast cancer that went with it.

  What do you consider the greatest obstacle to pursuing this research?

I think the first obstacle is making sure that the data are right and there isn’t some sort of clinical indication underlying what we’re seeing that we’ve missed, that would give a misleading answer to these questions. The other challenge is separating out the types of hormones women are using: whether it’s estrogen alone or estrogen plus progestin. Some are using testosterone. You really need to separate out the different preparations, since they can have different impacts on hormone levels. By inference if we’re interested in the underlying mechanism, we need to make sure we have a clear categorization of which type of hormone the women have been using.

  Critics of the Nurses’ Health Study often suggest that nurses are different from other women, and so the results may not be valid. How do you deal with this issue?

We looked at that quite extensively. Their weight is like the weight of other women. Their history of using oral contraceptives is like the history of other American women in their age group. Their screening practices are like those of other American women. We’ve gone through a fairly extensive array of analyses to look at their lifestyle habits. At some level, menopause in women is the same whether it’s in a nurse or not a nurse. It has nothing to do with occupation. Mean age of menopause is the same as the rest of American women, etc. So, in fact, we can show that their lifestyle and practices are quite comparable to those of other U.S. women.

  Has your data had the effect on public health policy that you believe it should?

At this stage, I don’t think our data often has any direct impact on policy unless one goes to the level of thinking that clinical decision-making is policy. It’s just at the individual level rather than at a national or regional level. I think that’s a stretch, but a lot of what we’re doing might be feeding into some of the decision-making going on in clinical practice. When it comes to physical activity, reducing the risk of heart disease or colon and breast cancer, per se, the national policies were already set down before we’d done our analysis. So we were adding evidence to support policies already in place.

  Are there areas where your data suggests policy or guidelines should change?

Yes, the guidelines on diet, where for 20 years the guidelines have suggested reducing your fat intake to reduce the risk of breast cancer. Our data and data from many other studies fail to support that association. In that sense, the lack of an effect starts to inform recommendations from organizations, be they the American Cancer Society or others.

  In an ideal world, what would you like to achieve over the next decade?

A better appreciation by the public of how much we could prevent cancer and other chronic disease if we actually implemented what we know already about tobacco and exercise and avoiding weight gain in adult life and so on. We could prevent more than half of cancer, and a sizable portion of diabetes, heart disease and so on if we were actually able to implement what we already know. If we could actually get to a place where that implementation is happening, that would be wonderful.

  Do you have any idea how to go about getting there?

We’re working on it. Part of it is overcoming inertia. Finding opinion leaders and all the other necessary components of the system that can bring about this change. Some of it is individual behavior change. Some is community-level change. And some of it is changed through the provision of health care providers. So it’s not like there’s one simple place to go and change everything.

  Do you find the way the media covers health care makes this difficult with the study-of-the-week phenomenon?

I think that is most disruptive at some level. Rather than getting an integrated answer, the public gets it piecemeal. That certainly doesn’t help.

  What do you tell the public for making decisions on these complicated issues?

I think you need to be telling them the integrative answer, not one research finding at a time. Science works by integrating it all together. It doesn’t work by rushing off necessarily with one study trumping all previous studies. But that’s sort of the way the media works. It’s the story of the day. That is the way the news is sold, if you will.

  If you had to give five items of crucial advice for improving health, what would they be?

My five would be: don’t smoke, get 30 to 60 minutes of physical activity every day; maintain a healthy weight. Then you have two more. Saying eat a healthy diet may be a cop-out because that could be too generic. That certainly would be the paragraph heading, however, if you will. You would have to break that down to don’t eat too much of this and make sure you take a multi-vitamin with folate and few things like that. And then I would limit alcohol intake as the other overall message.

Dr. Graham A. Colditz
Harvard University Medical School
Channing Laboratory
Boston, MA, USA