Nutritional Epidemiology: The Problem

The health headline shouts: “Fish oil increases your risk of heart disease.” The next week, “Fish oil beneficial for reducing risk of Alzheimer’s disease.” It makes one wonder what is going on in research. I’ve felt that way for a long time, and I know you have as well. I think I’ve found part of the problem, and I’m going to talk about in the next three Memos. As I said, this involves statistics, but I’ll keep it as simple as I can; we’ll be talking about some heavy duty statistical analytics, but there’s no math in my approach.

The frustration with conflicting research outcomes is especially prevalent in nutrition studies. It also happens in other fields as well including treatments for cardiovascular disease. Dr. John Mandrola, a cardiologist, wrote about a paper that examined why nutritional epidemiology is subject to conflicting results. After reading his post, I listened to the interview he had with the principal investigator of that study; then I read the paper itself. Statistics are not my forte but putting aside the high-level math, what they did was pretty amazing. By reviewing the paper here, I think you’ll get a better understanding of why research papers yield differing results even when using the same database of subjects.

Let’s begin with what has become accepted in the nutritional literature: eating red meat increases your risk of all-cause mortality. I know that many readers have given up or severely restricted red meat because so many physician, dietetic, and other health organizations have said it’s bad for your heart, among other organs. But is it really?

I used the phrase “nutritional epidemiology.” Those are studies that use some form of diet record, usually a Food Frequency Questionnaire, to track the food intake of a large group of people. Then they are tracked for 5, 10, even 20 or more years to see the differences in mortality between groups who ate red meat and those who did not. This contrasts with randomized controlled trials; they generally have as few as six subjects up to several hundred. They’re difficult to do because of the labor intensity of collecting data and ensuring subjects adhere to the protocol, whether a diet, specific foods, or dietary supplement.

With that in mind, the researchers examined all the nutritional epidemiological studies on the relationship of eating red meat and all-cause mortality. They then calculated all the ways that the data could be analyzed given the number of outcome variables and covariates used. We’ll pick it up on Tuesday, but as a tease, look up how much a quadrillion is.

What are you prepared to do today?