Sir Michael Marmot Interview: Conversations with History; Institute of International Studies, UC Berkeley

Redefining Public Health: Epidemiology and Social Stratification: Conversation with Sir Michael Marmot, Professor of Epidemiology and Public Health, University College, London; 3/18/02 by Harry Kreisler
Photo by Jane Scherr

Page 3 of 4

The Whitehall Studies

Tell us a little about what the Whitehall study is and what you found, especially in the second study.

The first Whitehall study was set up as a rather conventional study of risk factors for cardiovascular and respiratory disease. It was in British civil servants. Why civil servants? Because there is such a large number of them, and there they were easily accessible, and Professor Donald, who started the study, had close connections with the chief medical advisor in the civil service. So they set up this study, and they looked at blood pressure and cholesterol and smoking and other things. When I arrived in London, Geoffrey Rose said to me, "As a matter of good housekeeping, we've been looking at what grade men were -- " (these were all men) "-- at what grade men were in the civil service, where they were in the hierarchy."

Marmot and bargraph from Whitehall Study; see text below photo

The screenshot shows a bargraph from the Whitehall Study: "CHD MORTALITY BY GRADE: WHITEHALL MEN." From left to right, the bars show the relative mortality rates in "Admin," "Prof/Exec," "Clerical" and "Other" job classifications. Blue bars are adjusted for age; tan bars are adjusted for risk factors such as smoking, cholesterol levels, etc. -- Marmot/Lancet 1984

And he said, "You're interested in social and psychosocial things. Would you like to look at this?" And that's all we had, essentially, that was social and psychosocial, just what grade they were.

So I spent quite a lot of time looking at it in the first Whitehall study, and the remarkable finding, which ran counter both to my expectations at the time, and I think most other people's, was firstly, just looking at heart disease, it was not the case that people in high stress jobs had a higher risk of heart attack, rather it went exactly the other way: people at the bottom of the hierarchy had a higher risk of heart attacks. Secondly, it was a social gradient. The lower you were in the hierarchy, the higher the risk. So it wasn't top versus bottom, but it was graded. And, thirdly, the social gradient applied to all the major causes of death. I had come out of the model of, "If you want to understand heart disease, look for the cause of heart disease." We know something about cholesterol and blood pressure and overweight and diabetes, "Let's look for these things." But it applied to all the major causes of death -- to cardiovascular disease, to gastrointestinal disease, to renal disease, to stroke, to accidental and violent deaths, to cancers that were not related to smoking as well as cancers that were related to smoking -- all the major causes of death.

This finding of a gradient, not only did it run counter to thinking at the time about heart attacks, that high-status people were supposed to get more heart disease, but secondly, it ran counter then and still does to the way most people think about inequalities in health, or in the United States, disparities in health. We tend to think of them, the poor, that have poor health; and us, the non-poor, who have better health. Most social policy is about the underserved or the served, the underclass or the rest of us. It's not about a gradient, it's not about something that runs from the top all the way to the bottom, and that health follows that social gradient. And that's kept me exercised for the last twenty-five years, trying to understand that, and its implications for policy.

Now, the first question is, "Well, maybe civil servants are atypical. Maybe a top-grade civil servant hasn't a clue what it means to have a stressful job because he's protected and so on." Then when we looked at national data; there was, indeed, a shallow gradient in mortality in national data in Britain. It wasn't as steep as in Whitehall, and that bothered me a bit. Why not? One of the answers I gave myself was that the civil service was better at classifying people than the national data, the national social classes. But in practice over the last twenty years, we've seen the social gradient get steeper and steeper in the national data, so that it's certainly not the case that Whitehall is atypical, it's typical. We see similar findings in the United States, in Canada, in Australia, New Zealand, and most European countries that looked for it. So that was first thing: it's not because Whitehall is atypical.

Secondly, people say, "We've got rid of the problem in infectious disease now." In fact, we haven't, although we're not dealing with dirty water and poor sewage and so on. We're now in the age of 'lifestyle diseases,' it's because people behave badly." So we looked at the usual risk factors that one believes that are related to lifestyle -- smoking prime among them, but plasma cholesterol, related in part to fatty diet and an overweight, sedentary lifestyle. We asked how much of the social gradient in coronary disease could be accounted for by smoking, blood pressure, cholesterol, overweight, and being sedentary. The answer was, somewhere between a quarter and a third, no more. People said, "Oh, well, you didn't measure smoking properly." Well, we looked at never smokers, and we found the same gradient in never smokers as we found in smokers. I had to reassure people, I was not saying that smoking was unimportant. Whitehall confirmed how vitally important smoking is. It's just that smoking wasn't the prime cause of the social gradient in disease. It was very important and, for example, contributed to the social gradient in lung cancer. In fact, asking, "Is there a social gradient in lung cancer in non-smokers?" is a non-question, because all the lung cancers occur in smokers. So I was by no means saying smoking is unimportant. But I was saying when we look at coronary disease, and accidental and violent deaths, and gastrointestinal and renal deaths and so on, smoking was only some part of the story, but there was a lot else. Similarly, with cholesterol, there was no level in cholesterol levels by gradient.

So we published data from the first Whitehall study and said two-thirds, at least, of this gradient is unexplained. We said at the time that there were two candidates for the explanation: psychosocial factors and other aspects of nutrition apart from dietary fat related to cholesterol. We set up the Whitehall II study to test those hypotheses. This was a new cohort, we included women as well as men -- Whitehall I was only men -- and so they were not the same people in the course study. We set this up in the mid-1980s, twenty years after the original Whitehall study. The first finding was that the social gradient in health was as steep as it had been twenty years previously, hadn't got any less, and that fitted with national data. Then what we saw was there were clear social gradients in smoking and lack of physical activity, and so on, as we found in Whitehall I, but in addition now, we showed very clear social differences in people's experience of the workplace -- how much control they had at work, how fairly they were treated at work, how interesting their work was. We found clear social gradients in people's participation in social networks. We found social gradients in psychological attributes like hostility.

So in the end, what you're saying is you don't -- and I'm coming at this as a layman now -- that it's not about eliminating hierarchy so much as finding the factors that can be changed even if the hierarchy might remain the same. What you're finding is a gold mine of information about possible changes that nobody has been looking at, and that were very important in the different rates of death.

That's right, because as I've been thinking about it over the years, and more recently, particularly thinking about the Hitchcock Lectures at Berkeley, I'm struck by the fact that when you look at non-human primates you, of course, find hierarchies, but [also] hierarchies in disease. The subordinate monkeys in captivity or baboons in the wild appear to get more disease than the dominant ones. So that gives me pause. If you think about this social hierarchy and this relation to disease, there are at least two possible ways of thinking about it. First is, we could abolish hierarchies. Well, that doesn't sound very promising. If the baboons have hierarchies and the cynomolgus macaque monkeys have hierarchies, and the civil servants have hierarchies, and everywhere we look there are hierarchies, that doesn't sound very promising. The idea of some egalitarian society, where everybody is exactly the same, doesn't seem to be the way to go.

So then the question is, what is it about position in the hierarchy that determines different rates of disease? And given that, the hierarchy in disease does change. All societies may have hierarchies, but we know that the social gradient in disease is not fixed. It's bigger in some places than others, and it can change over time. That could be because the magnitude of the hierarchies change, but there's always hierarchies. But, more importantly, it suggests what is it about where you are in the hierarchy that's related to disease, and can we do something about that? So you ask, is it money? Is it prestige, self-esteem? And, in fact, what I think is, it is has much more to do with how much control you have over life circumstances, and the degree to which you're able to participate fully in society; what Amartya Sen calls capabilities.

It leads you back to overall human values about how we organize the way we work.

Human values, but in the end -- I'm a British empiricist, so human values, I think, are absolutely crucial here. But I'm also interested in empirical demonstration of how they translate into pathology, because in the end people go and get sick, and a value sounds like something rather abstract -- that it's the mind, where, in fact, what happens in the mind has a crucial impact on what happens in the rest of the body. The mind is part of our biological makeup as well. So the empirical study is how the sets of values translate into people's perception of reality, and that, in turn, changes physiology and leads to risk of disease. So we're trying to deal in a crude way with a mind/body question of how the one translates into the other.

This is interesting, because this is the new phase of your work, trying to find this biological link. So the output, in a way, is disease, and the input is this hierarchy, the social gradient. And the intervening variable is changes in the body over time, and the way stress operates on the hormonal system. Is that rights?

Yes, although there are one or two other intervening variables, so that if we've got position in the hierarchy, then we're asking what does that mean in terms of relationships in the workplace, in the wider society, control over events and participation. And how that, the satisfaction of those needs, or the frustration of those needs for control and participation, how that translates into physiological changes. You used the word "stress." I didn't, although I think it is a useful word. The reason I didn't use the word "stress" is people say, "Oh, stress, you can't define it, you can't measure it."

"You can't do away with it."

"You can't do away with it; one person's stress is another person's happiness," and all the usual things around the stress paradigm. But, in fact, it is a useful way to think about it. The way we've approached the measurement issue is not to say, "Is he or she under stress?'' but to ask, "Can we characterize their work circumstances? Can we characterize their social life in measurable ways?" And then ask, "Is that related to physiological changes, to endocrine changes, and, in turn, to other bodily changes that we can see, that we think are linked to risk of disease?"

And this is why your base in empirical data becomes very important, because it avoids entering into a realm of general discussion about these issues without making it very concrete. How does the comparative perspective enter into this? I know that you're involved in studies in Europe, you've looked at populations from the former Soviet Union and Eastern Europe. What does that bring to the table in furthering the insights that you're already making?

It brings at least two rather important things. The first is just a methodological point, but a crucial one. When I show -- and it's happened from the very beginning of the Whitehall studies -- when I show that people rank according to their social position and that's linked to their health, then one of the early questions -- I've already mentioned two, "Is it because of [being] civil servants?" and "Is it because of behaviors?" The next question people ask is, "Yeah, but people aren't randomly allocated to their positions in society. Maybe people who are genetically predisposed to be superior end up superior, and those same genetic predispositions to being superior lead to better health." Or, alternatively, a variant on that is that people's health status determines where they end up socially, it's not where they end up socially that determines their health. And that I've got the cart before the horse, and that my whole argument about social causation collapses because it's simply genetic differences determining health, and health determining social position.

That's not a straw man, by the way. It doesn't sound very politically correct, but there are a lot of people out there who believe that to be the case. Well, there are a number of ways of trying to answer to that. But look at what's happened to Central and Eastern Europe. Whole societies change radically, incredibly quickly. If we look at life expectancy in Central and Easter Europe, let's say, in 1970, so twenty-five years after the end of World War II, and we look at the countries in Central and Eastern Europe compared with the countries of Western Europe, the democratic countries, then life expectancy was fairly similar, east and west. It was always a bit lower in the former Soviet Union, but fairly similar. And then what happened? Over the last two decades of the Communist period, life expectancy for men, particularly, actually declined in Central and Eastern Europe, actually got worse, whereas it got better year-on-year in the West. So the gap between east and west increased. And then post-1989, things changed dramatically. Life expectancy took a tumble, got very much worse, very quickly in the former Soviet Union, particularly in Belarus, in the Ukraine, and in Russia. Life expectancy after the initial drop started to improve in the Czech Republic, in Poland, and then a bit later in Hungary. So you've got this playing out.

Now, whole countries don't get changes like that because of changes in genetic predisposition. It's just not credible that changes in health status determined the changes in the economy and the economic fortunes of the former Soviet Union; it's just ludicrous, that clearly was not the case. That it doesn't, by itself, argue that the determinants of the social gradient in health within a country can't be genetic or can't be determined by health just because you can't see it. But given that we think that some of the same causes are operating, and this is my second point now, the first one being the technical one, that we do think some of the same causes might account for the differences in health across Europe and account for the social gradient in health within a country. Why? Because, in general, those European countries were all at about the threshold of absolute deprivation. Infant mortality was low. There's a prima facie case that during the seventies and eighties they were not suffering from poverty. We know, for example, the Czech school children, we studied this, their growth was at least equal to Western European school children. So it looked like the children were well nourished and there was investment in children, there wasn't absolute poverty. And yet heath suffered, particularly in middle age men.

That was what led me to ask if maybe the same set of factors are going on. The whole idea that lack of control over your life, lack of opportunity to participate socially in a meaningful way, could affect whole societies, not just people, depending on where they were within the gradient within a society, was really rather powerful. It suggested that the way we ought to think about health policy should change.

Next page: Changing Society

© Copyright 2002, Regents of the University of California