Modern Diet-Health Epidemiology: a Self-Fulfilling Prophecy? Part II

Certain ideas about diet and health, for better or for worse, have worked their way deeply into the American psyche in the last few decades. We're constantly advised by health authorities, the news media, food advertisements, our doctors and our friends to eat less saturated fat, red meat and sugar, and more fruit, vegetables and whole grains. There is some dissent of course, but this has been the mainstream message for roughly four decades. And people are listening. We've replaced animal fats with unsaturated vegetable oils, red meat with poultry, whole milk with low-fat milk, and we're eating more fruit and vegetables than ever before. Here are two graphs of U.S. Department of Agriculture data to illustrate the point:
Whole grains are a very instructive case. Dr. Dennis Burkitt popularized the idea that fiber is good for health. He spent a number of years in eastern Africa, where he observed that natives on their traditional high-grain-fiber diets were free of many modern degenerative conditions, particularly those involving the digestive system. He found that as these cultures began to rely on Western foods such as white flour and sugar, their health declined dramatically. This is the same observation Dr. Weston Price made, however the two men interpreted their findings differently. Price attributed the effect to a loss of micronutrients, while Burkitt attributed it to the loss of a more or less inert substance: fiber.

There are a number of observational studies that have examined the relationship between whole grain intake and health. The massive Iowa Women's Health Study, for example, showed that women with a high intake of grain fiber had a 17% lower risk of death from all causes combined. In the same group, women in the top quintile (top 20%) of whole grain consumption had a 30% lower risk of heart attack than women in the lowest quintile. These two papers were published in 2000 and 1998. Here's where it starts to get interesting. From the second paper:

Higher whole-grain intake was associated with having more education, a lower body mass index and waist-to-hip ratio [and] being a non-smoker, doing more regular physical activity, and using vitamin supplements and hormone replacement therapy.

Do whole grains prevent smoking too? An alternative explanation is that the women who were eating whole grains were all-around more conscientious and concerned about their health than those eating refined grains. And why not? They "knew" from mainstream diet advice that whole grains are healthier than refined grains. When is the last time you saw someone smoking a cigarette while eating whole grain muesli with skim milk and half a grapefruit for breakfast? Is it easier to imagine someone smoking while eating a donut and sweetened coffee? Women who eat whole grains, on average, are those that care about their health and adopt patterns that they perceive as healthy throughout their lives. This includes behaviors large and small, both measurable and unmeasurable. The investigators factored smoking into their model, but you can't factor in things you didn't measure or don't understand.

Maybe it will come as no surprise, then, that the only controlled trial that has ever evaluated the effect of increasing grain fiber on all-cause mortality showed a strong (but not statistically significant) trend toward increased mortality in the group that doubled its grain fiber intake. Here's the graph of survival in the two groups. You can see that the trend is very consistent and the difference widens over time. This was the Diet and Reinfarction Trial. It's important to mention that the fiber group probably increased its grain fiber haphazardly, using bran and unfermented grains, rather than the traditional processing techniques of healthy grain-based cultures Burkitt described.

Here's the theory. When the public decides that a particular behavior is healthy, at that point it becomes difficult to accurately measure its impact on health using observational studies. This is due to the fact that healthy, conscientious people tend to gravitate toward the recommendation. If a theory manages to become implanted early on, it will become a self-fulfilling prophecy as healthy, conscientious people adopt the behavior and are detected by subsequent observational studies. People who don't care about their health or aren't motivated enough to make a change will keep living how they used to, and that will also be detected.

You can adjust for some of these factors if you measure them. Researchers commonly adjust for age, gender, smoking, exercise and sometimes other factors when they're trying to nail down the effect of a particular factor on health. But you can't measure all the little things that accompany a health-conscious lifestyle. Do the participants take the stairs or the elevator? Do they take supplements, and if so, which ones? How much sunlight do they get? Do they have positive relationships with their friends and family? How often do they shave (kidding)? What is the quality of the foods they buy? How often do they visit the doctor, and how often do they follow her advice? I believe there are too many confounds to measure and correct for. In my opinion, this means that observational data gathered from populations that already have opinions about the factor you're trying to study are unreliable and will tend to reinforce prevailing notions.

This brings us to the recent study on meat intake and mortality. It was a massive observational study that followed the diet and health of 617,119 elderly Americans for 10 years. Researchers found that the highest quintile of red meat intake was at an elevated risk of cancer and cardiovascular disease, and had an overall risk of dying about 1/3 greater than those in the lowest quintile. That's a pretty somber finding for those of us who love a juicy steak. But let's look at a few of the things that came along with red meat intake. I'm going to post a few graphs of factors that associated with red meat. They're organized by ascending quintiles of red meat intake; in other words, the people eating the least (left) through the most (right) red meat.

As compared to men eating the least red meat, men eating the most were three times more likely to smoke, half as likely to exercise regularly, and 22% less likely to take vitamin supplements! These are clearly people who are less concerned about their health in general. The investigators adjusted their model for a number of confounds: education, marital status, family history of cancer, race, body mass index, smoking history, exercise, alcohol intake, vitamin supplementation, fruit and vegetable intake, and hormone replacement therapy. This adjustment weakened but did not eliminate the association between red meat intake and mortality.

But again, you can't adjust for variables you don't measure. How about vitamin D status? Sugar intake? Quality and frequency of doctor's visits? Mental health? Dental health? Quality of food? There's no way to measure all the little things a health-conscious person will do to take care of himself. These unmeasured (and sometimes unmeasurable) factors can add up to have a major impact on health. So in the end, what are these studies really measuring? The association between diet and health, or the association between a health-conscious lifestyle and health? There's no way to know without a controlled trial. I rest my case, ladies and gents.

Here are a few other critiques of the study that are worth reading. Chris Masterjohn points out that the investigators' method of measuring meat intake was stunningly inaccurate, and they may have been measuring wishful thinking more than meat itself. Dr. Michael Eades points out that two other studies appeared at the same time, without fanfare, that contradicted the study's findings. And Jenny Ruhl discusses the implications of the bizarre finding that red meat intake also associates with the risk of accidental death.