If you look at who eats more butter and what happens to them without doing a randomized experiment you don’t learn about the effect of butter, you learn about the effect of butter, but about the effect of eating butter combined with everything else that is correlated with eating butter. Since most people eat butter with very unhealthy
Miles Kimball considers the following as important:
This could be interesting, too:
Roger Farmer writes The Future of Macroeconomics
Donald Marron writes Donald Marron 1970-01-01 00:00:00
Tim Harford writes Tim Harford 1970-01-01 00:00:00
[email protected] (Cyril Morong) writes Are Farmers Markets An Example Of Perfect Competition?
If you look at who eats more butter and what happens to them without doing a randomized experiment you don’t learn about the effect of butter, you learn about the effect of butter, but about the effect of eating butter combined with everything else that is correlated with eating butter. Since most people eat butter with very unhealthy things such as bread or potatoes, looking at what happens to people who eat butter (and probably eat it on top of things such as bread or potatoes) is not at all the same as an experimental protocol of flipping a coin to decide which experimental subject will be asked to eat more butter straight, then following those experimental subjects for a long time to see the difference in outcomes between those who ate extra butter straight and those who didn’t. (I apply this insight in “Hints for Healthy Eating from the Nurse's Health Study.”)
Similarly, if you look at who eats more eggs and what happens to them, you don’t learn about the effect of eggs, you learn about the effect of eggs and everything else that is correlated with egg eating. In particular, I am concerned that those who eat a lot of eggs might eat a lot of animal products in general. Rather than eggs and their cholesterol being especially bad, maybe it is animal products in general (including eggs) being somewhat bad.
I have written about eating too much animal protein as a possible cancer risk before. (See the section of links on “Anti-Cancer Eating” below.) It seems possible that animal protein is a heart disease risk, too. The study flagged at the top of this post has been all over the news. It has been interpreted as saying that cholesterol is the problem. I think we should be cautious about going there so fast. Just as fat that you eat isn’t the same thing as body fat, cholesterol that you eat is not the same thing as cholesterol in your bloodstream. In both cases, the body makes its own. The amount you eat of either body fat or cholesterol may not have a simple relationship to how much your body ends up with once the food is digested and metabolized and then the body decides how much to make in-house of body fat and cholesterol.
Given that caution, let’s look at the newly analyzed evidence. I want to focus on the evidence about how much of a relationship is left between cholesterol (which is abundant in eggs) and heart disease on the one hand and all-cause mortality on the other after controlling for other obvious risk factors, and controlling for the total consumption of animal protein. Below, I have copy-pasted the key parts of the key Table 2. Focus on the lines that say “animal protein,” which means animal protein is controlled for. Looking at the 95% confidence intervals for the absolute risk difference over a median followup of 17.5 years, the lower end of the confidence interval shows only a .08% higher incidence of heart disease and .84% higher all-cause mortality predicted from eating an extra half egg a day. The lower end of the 95% confidence intervals for the hazard ratio show only a 1.01 times higher incidence of heart disease and a 1.04 times higher all-cause mortality after controlling for animal protein consumption.
These numbers overstate the evidence for an effect of cholesterol. It is never OK to think you are “controlling” for the effect of a variable simply by putting it in a regression if the variable being controlled for is measured with error. I promise you, animal protein consumption will be measured with error. If measurement error in the data for animal protein consumption were the only problem, controlling for animal protein with an imperfect measure seems likely to undercontrol for animal protein consumption, and controlling for it using a measurement error correction would affect the coefficients of interest more (in this case, that would be in the direction of knocking the coefficients of interest down more). Things are more complicated here because cholesterol is also measured with error. But in any case, the authors need to go through the exercise of making measurement error corrections.
An omission is that the authors don’t look at a summary measure of either the insulin index or glycemic index (or insulin load or glycemic load) of the foods eaten. One reason this could matter is that egg consumption is likely to be associated with eating breakfast. Along with eggs people might be eating types of food that have quite a high insulin index, such as toast, cold cereal, pancakes or waffles. (The word “breakfast” does not appear in the article.)
Another omission (perhaps an unavoidable one), omission is that when the authors try to look at things within subgroups, they don’t look at people who fast for at least 16 hours a day as a subgroup. Maybe a rest from food would allow the body to defang otherwise troublesome foods; we simply don’t know. The evidence is primarily about variation around the average diet with the average eating pattern. (A search for “fast” turns up only mentions of “fasting glucose,” which is a test that only necessitates fasting once.)
None of this is to say that you should feel totally safe eating a lot of eggs. But at a minimum, in their analysis, the authors are clear that they have a tough time distinguishing whether eggs are the issue, or eggs and meat combined are the issue.
Don’t miss my other posts on diet and health:
I. The Basics
II. Sugar as a Slow Poison
III. Anti-Cancer Eating
IV. Eating Tips
V. Calories In/Calories Out
VIII. Debates about Particular Foods and about Exercise
IX. Gary Taubes
X. Twitter Discussions
XI. On My Interest in Diet and Health
See the last section of "Five Books That Have Changed My Life" and the podcast "Miles Kimball Explains to Tracy Alloway and Joe Weisenthal Why Losing Weight Is Like Defeating Inflation." If you want to know how I got interested in diet and health and fighting obesity and a little more about my own experience with weight gain and weight loss, see “Diana Kimball: Listening Creates Possibilities” and my post "A Barycentric Autobiography.