Over 100 of you wrote to me to ask my opinion of the latest “Red Meat Will Kill You” scare, a study which was published March 12 in the Archives of Internal Medicine, and gleefully reported by the mainstream media and commented on by just about everyone on both sides of the red meat controversy.
Now this study has been expertly and brilliantly debunked by a number of extremely smart people, so let’s start by giving credit where it’s due. Denise Minger, as always, has written a superb piece on it, as have Gary Taubes and Zoe Harcombe. And look for my good friend Mike Eades, MD to put something up soon- his take on this stuff is always on-the-money.
So rather than repeat what these brilliant folks have already done so well, I’ll just summarize some of the highlights of what they’ve already covered, add my own two cents, and call it a day. Believe me, this won’t be the last you ever hear of the “red meat will kill you” debate, but hopefully it will give you, dear thoughtful reader, a good place to start if you really want to debunk this stuff at your next cocktail party.
The first thing to know about that as epidemiology goes, it was quite good.
But, as science– frankly it sucks.
And why it sucks, unfortunately, is a much bigger issue than just this study from Harvard. Why it sucks has to do with the whole nature of these kinds of epidemiological studies, which are called observational studies, and are as different from real science as baseball is from thermodynamics. (Nothing against baseball, mind you, and nothing against thermodynamics. They’re just very different things.)
Science is about finding out what causes what. You take a hypothesis—i.e. spaghetti eating cause heart disease—and then you test it. You take two identical groups- of rats, people, monkeys—make sure they’re identical in every way that you can think of—put them into identical conditions except that one group gets a drug, or eats spaghetti, or performs an exercise, or takes a spin class, or does whatever the heck you’re trying to investigate—and then you observe the results. If there’s a big difference between the two groups, you can be pretty sure (not positively sure, mind you, but pretty darn sure) that the one thing that was different between the groups was actually responsible for the results. If, for example, all the rats taking the drug died while all the rats who didn’t take the drug did just fine, it’s a good bet that the drug killed them. If two groups of “identical” kids are assigned different 8th grade English teachers, and Group A aces the English finals while Group B fails miserably, you can be fairly sure the teachers had something to do with it.
This was not such a study.
This study—like many, many before it—is an observational study. No one was put into matching groups, no diet was given out and tested, no one came into a clinic. Instead, thousands of people filled out food frequency questionnaires a few times over the course of thirty years or so. (Food questionnaires that are known to be notoriously unreliable. And why wouldn’t they be? How reliably can you remember what you had for breakfast last month?)
Before we go on to the specifics of this particular study on red meat, let me quote Denise Minger on observational studies in general. Hard to say it much better than this:
“In case you’re skeptical that observational studies can run disturbingly contrary to reality, look no further than the hormone replacement therapy (HRT) craze that peaked a few decades ago. By 1991, 30 observational studies—including this one— based on none other than the Nurses’ Health data—collectively showed that women taking estrogen seemed to have a 44% reduction in heart disease risk compared to their non-hormone-replacing counterparts. Naturally, this led literally millions of women to jump on the estrogen bandwagon in pursuit of better health and longer lives. A very unfortunate oopsie-daisy sprouted up later when some randomized, controlled trials finally emerged and revealed that rather than being protective, hormone replacement therapy actually increased heart disease risk by 29%!”
So back to the red meat study.
The participants answered dozens of questions like “how many vitamins do you take?”, “do you take birth control?”, “have you ever been diagnosed with heart disease?”, and “how many times a week did you eat meat last month?”. (You can view the actual questionnaires here.) These people were followed for decades, and hundreds of “data points” collected, such as how many people died, how many people got cancer and how many people got heart disease. Then the researchers “correlate” (or associate) all these data points in an attempt to find some patterns that will eventually generate hypotheses that can be tested in randomized, controlled studies.
Like the hypothesis that “red meat leads to early death”.
This study did not test that hypothesis. Studies that have tested it—and there have been quite a few—have found quite the opposite. In much research done over the past decade, the Atkins diet (or a variation of it) has been tested against a conventional low-fat diet in randomized, controlled clinical studies and in study after study, every measure of “risk” for heart disease has improved, not declined. (For a start, just Google research by the likes of Jeff Volek, RD, PhD of the University of Connecticut, or Eric Westerman, MD of Duke University, the A-Z study out of Stanford, or look at the copious research reported in “The Art and Science of Low-Carb Living”.)
The first thing every eighth grade science class learns is that correlation is not cause. Because two things are found together, or “associated”, does not mean one causes the other. Best example of this was one I learned from my college psych professor, Dr. Scott Fraser, and I’ve never found a better one. Here it is:
In Denmark, the number of storks is positively correlated with the number of babies. That is to say, the more storks in an area, the more babies are found there, and this correlation (association) has held up year after year for decades.
Yet no one seriously thinks that one causes the other. At least I hope not.
(I know you’re dying to know the real reason storks and babies are found together in Denmark so I won’t keep you in suspense. In Denmark, the cities are pretty much inhabited by single people, and when these folks marry and start a family, they move to the suburbs. The suburbs of Denmark are filled with homes that have sloped roofs. Storks love sloped roofs and flock to areas where there are plenty of them. Mystery solved.)
OK, once again, back to Haaarvard and red meat.
What the Harvard researchers found was that people who report eating more meat also have a slightly higher rate of dying. But if we all weren’t so sure red meat is “bad” we might have examined this association a little more carefully.
The researchers broke the participants up into quintiles, from the 20% who reported eating the least meat (virtually vegetarians) to the 20% who reported eating the most, with three quintiles in the middle.
As we go up the “meat-eating” scale, quintile by quintile from lowest-meat-eaters to highest-meat-eaters, we notice that other things go up as well (it’s right there in the data). Meat eaters smoked more, drank more and exercised less. They also had higher BMI and higher blood pressure. And, as a group, they ate a lot more calories. (A lot more.)
Oh, and by the way. They also had lower cholesterol.
Yes, you read that right. The high meat eaters had the lowest cholesterol of any group (and the vegetarians had the highest).
How come the media didn’t proclaim “Red Meat Lowers Cholesterol?”
Because they already believe red meat kills you.
We call this, affectionately, “Doing a Campbell”. This is exactly what Colin Campbell did in his book about a massive study called “Mortality, Biochemistry, Diet and Lifestyle in Rural China”, a book he called “The China Study”. Out of 100,000 statistical associations in the original 894 page tome, he chose to report in his book only the ones that supported his vegan beliefs, while omitting the ones that did not. Did you know, for example, that wheat flour had a higher correlation (.67) with heart disease than meat? You certainly wouldn’t know it from “The China Study”, but it’s right there in the data.
You mean Campbell left that one out? Whoops.
Speaking of correlations left out in the reporting, the people who ate the least amount of meat in the Harvard study were also the most likely to take vitamins. Interesting that the media didn’t call this study “Proof that taking vitamins helps you live longer!”
(If you’d like to skip all the mind-numbing numbers in the original study, they’re presented here in a very-easy-to-read graph form, courtesy of Denise Minger.)
What’s also interesting is that the size of the effect was about 1/100th of the association found in similar studies between smoking a pack a day and lung cancer. Those early studies were also observational, but the size of the effect—20 fold increase of lung cancer that pack-a-day folks have compared to non-smokers—was so enormous, and duplicated so consistently, that it was absolutely impossible to ignore. The association between reporting meat eating and death was a .02 increase as opposed to a 20-fold increase.
And if you’re still in love with association studies and want to base your eating preferences on them, consider that years ago, the great British doctor John Yudkin showed that sugar was far more associated with heart disease than fat was. So, for that matter, was the purchase of a television set.
Meat eating could be a marker for people who are more or less likely to adopt what they think are healthy habits. As Gary Taubes wryly points out, this is like comparing people who choose to be vegetarians and eat at Alice Waters’ restaurant in San Francisco every night after yoga class, with truckdrivers from West Virigina whose idea of a night out is Denny’s, and to then say the difference in their health outcomes is the red meat.
Back in the 60’s when this study started– and certainly in the 70’s and 80’s– the conventional advice about health was to cut back on red meat, don’t smoke, exercise and eat a lot of vegetables. Health conscious people who listened to the “authorities” ate less meat, exercised more, stayed slimmer and didn’t smoke.
It’s only in the past decade or so that health conscious people have realized that grass-fed meat is an entirely different “animal” than the crap we buy in the supermarket or the fast food restaurants. And no study—not even these quasi-scientific epidemiological ones—has looked for associations between grass-fed meat eating and early death.
Look, I’m no shill for animal protein. I wouldn’t touch the stuff they sell at McDonald’s with a ten-foot fork. Feedlot meat- i.e. meat that comes from factory farms—is filled with antibiotics, steroids and bovine growth hormone. The fat is high in inflammatory omega-6’s and virtually absent of omega-3’s. And processed meat is even worse, filled with sodium and nitrates. If that were the only meat available to me, I’d probably become a vegetarian.
But until the day I see a single randomized, clinical study showing that a diet of grass-fed meat and a ton of vegetables produces a single percentage point increase in early death, heart disease, cancer or diabetes, I’m sticking to the same diet that I’ve been on for over a decade: grass-fed meat, wild salmon, tons of vegetables, berries, nuts, full-fat yogurt and as little sugar and grains as I can manage to live without.
That’s kept me healthy, energetic and without a single serious medical condition for 60-something years.
I realize that’s not much of a scientific experiment—but then again, neither was the study from Harvard.
Could you link to the responses you mentioned in the second paragraph? Thanks!