Eric Rimm, SD ’91, Professor of Epidemiology and Nutrition
“When unexpected findings come out, media people and even scientists often try to draw parallels to true scientific discovery— like finding a new planet in the solar system. But it’s hard to disprove 40 years of nutritional epidemiology with a single study.”
Recent headlines suggest that red meat and butter raise cholesterol—and that red meat and butter are heart-healthy. That vitamin supplements stave off disease—and that vitamin supplements make no difference in disease risk. That even a thimbleful of alcohol is risky—and that moderate consumption prevents cancer and clogged arteries. Fasting regimens, superfoods, keto diets: The media is a smorgasbord of mixed messages.
How to make sense of the current news feed? Harvard Public Health editor Madeline Drexler looked for guidance from Eric Rimm, professor of epidemiology and nutrition and a scientist who has, for more than three decades and in some 750 journal articles, explored key nutrition questions.
Q: How can nonscientists know whether a nutrition study with surprising results is really new or just fodder for the news?
A: You have to be skeptical. Many reporters and headline writers try to create friction between what we confidently know about nutrition and health and what a new study claims. The field of nutritional epidemiology is becoming mature, and has its underpinnings in biology. When unexpected findings come out, media people and even scientists often try to draw parallels to true scientific discovery—like finding a new planet in the solar system. But it’s hard to disprove 40 years of nutritional epidemiology with a single study.
Q: So what should readers look for in a study that seemingly overturns the conventional wisdom?
A: The best studies of nutrition and health are either large randomized studies or large long-running observational cohort studies. In either case, the best studies keep track of people over time, because people’s diets, physical activity, smoking, and body weight all change and these changes influence health risk.
Say I took all of your friends when you were 35 years old and asked them how much they drank, what they ate, and how much they exercised. Then, 30 years later, I decided to track them all down to see who has died and what they died from. That study would give you a very different answer than if I had gone back to your friends every four years and asked them details about their diet, smoking, exercise patterns. You can’t study something like cancer or heart disease, which likely take 20 to 40 years to develop, just by looking at a single snapshot in time.
Another reason to capture as much as possible about someone’s lifestyle is that eating behaviors are often aligned with other behaviors. Someone who eats healthy every day and has been eating healthy every day probably smokes less, exercises more, and has more friends—all of which benefit health. Turning this around, I hesitate to interpret studies that look at the effects of smoking or exercise unless I know something about participants’ diets as well.
I’ve also become more skeptical of meta-analyses, systematic reviews that synthesize the results of many papers. They are typically conducted by selecting a single aspect of diet—for example, nuts, berries, or dairy products—and a single health outcome, such as heart attacks or diabetes. These analyses dig through the scientific literature to pull findings from all studies that have been published on the association. In the last five to 10 years, the academic literature in our field has been flooded with meta-analyses. The problem is that many of those papers are done by people who know a lot about statistics but little about nutrition. Some of these reviews are good, but many produce wrong or skewed results, because the lead scientists don’t fully understand the scientific literature.
Q: Wasn’t it a 2014 meta-analysis that found that saturated fat was healthy?
A: That was a classic example. Inexplicably, the authors first excluded a few key papers, which tilted the benefits toward polyunsaturated fat. Then they interpreted the results of other studies incorrectly. Diet is so complex that you can’t look only at high- or low-saturated fat intake, because when you give up saturated fat, those calories almost always get replaced by foods with polyunsaturated fat or carbohydrates. For example, if you used to eat lots of red meat, cheese, and butter, and then you switched to white bread and white rice, saturated fat would not look comparatively harmful, because both are bad diets. But if you exchange saturated fat from red meat for polyunsaturated fat from vegetable sources, then saturated fat comes out much worse. The authors of the 2014 review largely ignored that important nuance. Right after that study came out, Frank Hu, chair of our Department of Nutrition, did a rigorous meta-analysis that clearly showed that saturated fat is much worse for you than polyunsaturated fat, and probably equally as bad as refined carbohydrates.
Q: Why can’t the nutrition field just settle these questions with large-cohort studies?
A: The current funding environment for nutrition studies is the worst I’ve seen in my 30 years at Harvard. The National Institutes of Health prefers funding research that is high-tech, cutting-edge, and sexy. It can cost tens of millions of dollars to study new technologies with new equipment or new animal models.
I’ve talked to colleagues who are cardiologists. They might do a trial on one drug—say, a new kind of combined statin drug, to see if it lowers cholesterol and reduces heart disease and mortality. These trials take place in 25 to 50 countries and involve 20,000 people. That type of study can cost $300 million, and almost always is paid for by a pharmaceutical sponsor. In our research group at the Harvard Chan School, we have not received $300 million total over the last 40 years for our large-cohort studies, yet we have contributed substantially to the scientific literature on scores of nutrient and dietary patterns and their relationship to dozens of categories of human health.
Think about how the School’s work has changed the food supply in this country. More whole grains. No more trans fats. An emphasis on nutrient-rich berries and superior protein sources such as nuts. Healthier menus in school lunch programs. The return on investment is immense—and it’s because of large, relatively low-tech, observational cohort studies.
– Madeline Drexler is editor of Harvard Public Health.
Photo: Kent Dayton/Harvard Chan School