Oh brother, here we go again…
Last week, the journal Atherosclerosis published a study examining the association between the number of egg yolks consumed per week and the amount of plaque in the carotid artery. The study concluded that eating egg yolks is almost as bad as smoking when it comes to speeding up plaque deposits.
OK, everyone, relax and take a deep breath.
The idea that a food which contains 13 essential vitamins and minerals, high-quality protein, antioxidants, eye-health supporting carotenoids (lutein and zeaxanthin) and brain supporting choline should have almost as bad an effect on the heart as a pack of Salems just doesn’t pass the smell test, even if your nose is clogged. And it doesn’t begin to jibe with previous research.
But that’s just the beginning.
Readers of this newsletter have heard me rail time and time again against drawing conclusions from what are known as observational studies, of which this is a prime example. If you’re new to my column, here’s the difference between clinical study (which this is most definitely not) and an observational one. It’s an important difference, and it’s essential to understanding why the present study is not only horrible science, it’s also useless.
What Exactly Is An “Observational Study”?
In an observational study from which many associations are generated, you take a whole bunch of people- thousands of them—and you gather data about a zillion different things.
Maybe it’s blood pressure and cholesterol, maybe it’s heart disease, maybe it’s what they ate for breakfast, how often they brush their teeth, how many of their parents had diabetes, how many of them own television sets, practice the rhumba, love Lady Gaga, take antidepressants, or pop a Centrum now and then.
OK now you’ve got a statistician’s version of heaven—tons and tons of data. A gazillion gigabytes of numbers from thousands of people, and it’s your job to see if there’s any pattern, to determine which things are “associated”, meaning “found together”. If two things are said to be associated (or correlated), that means there is some relationship between these two things.
Of course, discovering that two things happen to be found together is only the first part of science. The second, important part, is finding out why. Observational studies don’t touch that question; they merely catalogue what’s found together and leave it to the big boys to run the actual clinical studies and find out what the found association actually means (if anything).
This study didn’t even generate tons of data because they only looked at two- count ‘em, two—variables. Smoking and egg yolk consumption. We’ll come back to this in a minute.
So this egg yolk study was an observational study, like all the other studies with weird and inexplicable “results” (i.e. taking a multivitamin will kill you). You cannot- repeat cannot– attribute cause and effect to an observational study. All you can say is that it’s been observed that two things are found together more frequently than might be by chance. You have no idea about why.
(If you want to read a great story about the idiocy of drawing cause and effect conclusions from observational studies, read the following paragraphs. If you’d just as soon move on with the story, skip over the part about storks and babies.)
Storks and Babies: A Short Tangent into Observational Studies
In a certain region of Denmark, there is a correlation between the number of storks and the number of babies. Seriously. It’s a positive correlation, meaning the more storks there are, the more babies there are. According to my old Psych 101 professor, Dr. Scott Fraser—this stork-baby correlation thing has held up for years.
So. A classic observational study shows more storks, more babies. Less storks, less babies. What do we conclude?
Well, here’s what we do not conclude, unless of course, we’re three years old: Storks bring babies. We might not know why the correlation occurs—it’s admittedly puzzling and amusing at the same time– but we’re pretty sure that it’s not cause and effect.
Dr. Fraser used this example back when I was in his psych class to introduce the concept of confounding variables.
A confounding variable is one which is not measured directly, but which accounts for the relationship between the two things that were measured.
See if you can find the “confounding variable” in the storks-babies correlation.
Here’s the answer to the puzzle. In Denmark, the cities are largely populated by singles and by childless couples. When people want to have kids, they move to the suburbs. In the suburbs of Denmark, the houses tend to be A frames with tar-based sloping roofs. Storks love to nest on tar and sloping roofs.
The tar-based sloping roofs are the “confounding” variable. Both storks and newlyweds flock to the same area. Puzzle solved.
Back to Egg Yolks
When statisticians do studies like this “egg yolks kill” study, they try to “correct” for the confounding variables by applying statistical techniques that nullify other potential influences, like age, sex, years smoking and all that other good stuff.
But they can only “control” for the variables they think of. And they often don’t think of precisely the variables that are making the difference, particularly when they already have a point of view about what the outcome is going to be (it’s called “confirmation bias” and it’s rampant in research.)
In this study, for example, investigators claim to have statistically controlled for smoking, but they failed to control for overall lifestyle behavior. “Can they truly statistically control for all health factors of smoking?” asks Donald Layman, PhD. No, folks. They can’t. They also didn’t control for sugar intake, nor for omega-6 intake. (The high ratio of omega-6’s to omega-3’s in the western diet is probably responsible for more heart disease than saturated fat ever was!)
It gets worse. They also didn’t control for exercise. Not only that, they completely failed to mention whether the “high egg users” were also the “high smokers”. (Smoking is one of the biggest factors associated with increasing plaque area.) And did I mention that plaque generally increases with age? Subjects in the “high egg consumption” group had an average age of 69.7 years while subjects in the “low egg consumption” group were on average 12 years younger, with an average age of 55.7 years.
But that’s not even the best part.
Buried at the end of the study is the finding that there was no association between cholesterol and plaque growth. Not total cholesterol, not LDL cholesterol not HDL cholesterol. None of them had any association with greater plaque. So if eating 2 extra eggs a week—the difference between the “low egg consumers” and the “high egg consumers”—is causing plaque growth, it’s got to be by some other mechanism than the cholesterol in the eggs, since cholesterol wasn’t associated with the reported plaque growth at all. Yet they claim that the reason they did the study in the first place is because there’s reason to be concerned about the cholesterol in eggs!
In a very limited observational study of 1200 people, researchers measured carotid artery thickness and then went back and examined data on two variables: smoking and egg yolk consumption. That’s all, folks. And they found that people (average age 57) who ate about 2 ¾ eggs a week had less plaque growth than people (average age 69) who ate about 4.1/2 eggs a week.
Plaque growth might also have been positively correlated with portions of asparagus eaten on Wednesdays or number of pitches thrown in a little league game when the subjects were nine. Of course the researchers wouldn’t have looked at any of that, because they wouldn’t have thought it was relevant. And maybe it’s not. But is sugar consumption? Omega-6 fat consumption? Lifestyle choices? Smoking? Age? The investigators didn’t look at other “suspects” because they had already decided they had the guilty party.
David Jenkins, one of the researchers on this paper, is a wonderful man, a very smart, gentle scholar who has made an enormous contribution to nutrition with his work on glycemic index. I questioned Dr. Jenkins after his presentation at the American College of Nutrition conference in New York in 2010, and it was abundantly clear that he was a strong believer in the cholesterol hypothesis and that he was concerned about the growing trend towards looking at cholesterol as just a minor player in heart disease. For those of us who believe that the cholesterol hypothesis is just a myth, I’m sure he feels this study is a wake-up call
Unfortunately this study is nothing of the kind. Sad to say, it’s just really bad science and I’m honestly amazed that it got published at all.