Is red meat guilty by association? Some at Harvard think so

Last updated on April 10th, 2017

The Harvard School of Public Health (HSPH) giveth to the low-carb community, and it taketh away.

In late June 2011, the HSPH published a study singling out potatoes as a particularly fattening food, especially in the form of french fries. Having sworn off fries forever as part of my low-carb way of eating, I applauded the brilliance of the findings.  I also liked the general conclusion that all calories are not equal when it comes to packing on pounds.

Now I’m wondering what could have caused a nose-dive in the quality of HSPH’s research in under two months.

Last week, a team from the HSPH published an epidemiological study claiming an association between the consumption of red and processed meats and the development of type-2 diabetes (Pan et al. Red meat consumption and risk of type 2 diabetes. The American Journal of Clinical Nutrition. Note: this link is to the abstract. The article itself is behind a pay wall.). News agencies pounced on the study, pouring out numerous rehashes of the HSPH press release, some suggesting a causal link exists between eating meat and becoming diabetic.

Of course, the researchers themselves stop short of claiming a causal link. In the language of the press release, there is “a strong association between the consumption of red meat—particularly when the meat is processed—and an increased risk of type 2 diabetes.” (Never mind that the researchers know what the popular press will do with the story!)

Not only is an association not a cause, but in this case, the association is between eating meat and having “an increased risk” of developing diabetes.  Other kinds of evidence are needed to show a causal mechanism. Asked to speculate on the mechanism at work, senior researcher Frank Hu hypothesized that it was an overload of iron. Of course, this hypothesizing was done in an interview, not in the study. So it’s PR, not science. (But then the science status of this kind of research is questionable, anyway.)

This was yet another large-scale, observational study in which researchers analyzed the self-reports of thousands volunteers and then supposedly try to factor out all variables — dietary, genetic, behavioral — except the one they were examining. So the quality of the findings depend on the quality of the volunteers’ reports and on the quality of the researchers’ statistical factoring out process. At best, such a study can yield hypotheses worthy of further testing by clinical and experimental means. At worst, they can lead researchers on a wild-goose chase and mislead or confuse the public.

In this case, I hear a goose honking.

Regarding the quality of the self-reports, I have personal experience to draw on. For several months, I’ve been keeping a diet diary of sorts, using computer software designed for the purpose. I have strong self-motivation to log my eating, and to be accurate in doing so; I see it as a tool in weight loss.  I also see it as giving me good blogging material from time to time. But does that mean I believe my food log is 100% accurate?

No, not at all. In terms of the foods I have eaten, the log is close — say 98-99% accurate overall, and 100% on many days.

In terms of the amounts eaten, though, I believe my log is less accurate — say 90-92%.  The reason is, I don’t always stop to measure servings.  Early on, I did, and it gave me an impression of an ounce of cheddar cheese, a cup of broccoli, or a tablespoon of mayo to use in guesstimates of amounts. Lately, I haven’t been as careful to measure, and have relied more on my guesstimates.

Of course, I don’t pretend to be scientific.  Since I continue to lose weight, I’m not too concerned about inaccuracy. I’m apparently in the ballpark on my amounts consumed.  But if I hit a prolonged plateau, or start to gain weight, then one of the first things I will do is to examine my serving sizes more carefully.

However, if we take my minor guesstimation errors made over just a few months, and multiply them by 400,000 people with varying levels of dedication and motivation filling out questionnaires over years, decades even, then the likelihood of error becomes enormous.

In fact, the current study was a so-called “meta-analysis” of questionnaire responses by several hundred thousand volunteers in separate studies conducted over a period of many years. Are we to believe that these volunteers were so accurate in measuring and recording the amounts of food they ate that the researchers can say with confidence that “a daily 100-gram serving of unprocessed red meat (about the size of a deck of cards) was associated with a 19% increased risk of type 2 diabetes” or that “half that quantity of processed meat—50 grams (for example, one hot dog or sausage or two slices of bacon)—was associated with a 51% increased risk”?

Well, you can believe if you want to, but I don’t believe it. Not if we are talking about a real world associations.

Based on my personal experience and my knowledge of human nature, I just cannot buy the gram-by-gram precision of those claims emerging from that kind of data. Certainly, there would be plenty of random error, but there could also be systematic error.

OK, fine, you say, forget about grams.  But what about the overall sweep of the statement? What about the idea that, in general, the more red meat you consume beyond some level, the more likely you are to develop type-2 diabetes?

First of all, the amount does matter. The study is not associating any red meat consumption with increased risk of diabetes, but only red meat consumption in excess of a very specific amount. Second, this is where the quality of the researchers’ factoring out of all confounding variables comes in.

Just how well can anybody’s statistical methods isolate the effects of hundreds or thousands of variables in play in the development of type-2 diabetes? (I wrote about this kind of issue in one of my first posts.) Of course, to factor out a variable, you first must recognize it as such. Were all possibilities and combinations considered, let alone factored out? For instance, was the level of carbohydrate consumption considered?

Were most of the heavy meat-eaters also heavy carb-eaters? (Or just heavy-eaters, period?) It seems likely since the standard American diet is carb-heavy.  In that case, red meat would truly be guilty by association. (Looking world-wide, it seems that red meat consumption cannot be a requirement for developing diabetes.   Consider the type-2 diabetes epidemic going on among Hindus in India, who are not — to say the least — heavy meat-eaters! So other foods, or combinations of foods, must be involved.)

I’m not a mathematician nor an epidemiologist, and I respect both fields. Nevertheless, factoring out intertwined biological, dietary and lifestyle variables is a tall order.  (To my credit, I said as much in my June 23 post, the one in which I applauded the “anti-potato”  HSPH study: “I am skeptical of the procedures used to isolate the effects of individual foods.” To my discredit, I let my anti-potato bias overwhelm my skepticism about the methodology. Of course, there are other reasons to view French Fries as fattening, such as their huge calorie and carb loads per serving.  Even so, I gave too much credit to the observational methodology, and I pledge to do better.)

The epidemiologists  keep trying this approach, looking at one possible target after another. The results are impressively varied. Here are a few of the other epidemiological studies published in 2011 that looked at diet and type-2 diabetes:

• Dietary omega-3 fatty acids and fish consumption and risk of type 2 diabetes (“Marine but not plant-based omega-3 fatty acids were positively associated with incident T2D [type-2 diabetes].”)
• Sugar-sweetened and artificially sweetened beverage consumption and risk of type 2 diabetes in men (“Sugar-sweetened beverage consumption is associated with a significantly elevated risk of type 2 diabetes, whereas the association between artificially sweetened beverages and type 2 diabetes was largely explained by health status, pre-enrollment weight change, dieting, and body mass index.”)
• Changes in alcohol consumption and subsequent risk of type 2 diabetes in men (“Increases in alcohol consumption over time were associated with lower risk of type 2 diabetes among initially rare and light drinkers.”)

Taken together, the studies cited above and the HSPH study seem to be saying that, depending on who you are, consuming certain amounts of oily fish, red meat and soft drinks will increase your risk of diabetes, but drinking more booze will lower the risk.  If the devil is in the details, this is a devilish situation.  God help researchers if they ever come across a person who eats salmon and beef, and drinks cola and rum!

But come on — what is the chance of that?

Let’s get to the bottom-line.

For me, the bottom-line is what to eat and what to reduce or avoid. I won’t be changing my current eating practices because of this HSPH study. Not one bit.

In their press release, the HSPH researchers offer this dietary advice: “consumption of processed red meat—like hot dogs, bacon, sausage, and deli meats, which generally have high levels of sodium and nitrites—should be minimized and unprocessed red meat should be reduced. If possible . . . red meat should be replaced with healthier choices, such as nuts, whole grains, low-fat dairy products, fish, or beans.”

As I wrote a while ago (No hot dogs here), I have restricted my intake of processed meats that contain sodium nitrites. Provisionally, anyway, I believe nitrites cause cancer. (But why the focus on sodium and nitrites in this context? I thought the hypothesis was that too much iron was the culprit in the association between meat and diabetes. Does processed meat have more iron than fresh meat? It could, I guess, but why not mention that in the statement, instead of high levels of sodium and nitrites? I suspect rhetorical trickery is afoot!)

I eat more fresh red meat than I used to, but not a lot more. I’ve been careful about fresh red meat not because of a fear of fat or iron, but a fear of contaminants. (See Beliefs can cause inflammation of the brain.)

As a low-carber, I have significantly increased my consumption fish and nuts. But I have stopped eating whole grains, low-fat dairy, and beans.

Beans I might add back to my diet in limited amounts. Dairy is already there in the form of butter, cheese and cream — high fat dairy, in other words. Grains I want no part of. (I’ll write more on my problems with grains later. I seem to have sensitivities to them.)

The HSPH researchers’ “low-fat” dairy stipulation betrays their bias; they appear to be working from Conventional Diet Wisdom, which aims to reduce intake of saturated fat at all costs.  This isn’t merely a matter of press release rhetoric.  The study actually examined the impact of swapping out a serving of red meat for a serving of low-fat dairy. Why not also examine swapping out meat for full-fat yogurt or cheese? Is it the fat that is being tested again? If so, name it as a suspect, along with iron and nitrites.  It increases the odds of a hit.

Another indication of Conventional Diet Wisdom in action is the omission of eggs from the list of alternative protein sources.  It’s true that eggs have slightly more iron per serving than almonds, but they have less than lentils and far less than beef. Eggs don’t contain nitrites.  They have some fat, but mostly they just have an undeserved bad reputation courtesy of Conventional Diet Wisdom.

Eggs are food I eat significantly more of on my low-carb diet than I ever ate before.  It seems odd that eggs are omitted as a source of protein.

But, then, everyone has biases, even epidemiologists.

(For another critique of the HSPH study, see Mark Sisson’s blog.  For another perspective on diet and the development of type2 diabetes, see my recent post News report: cut carbs to hold off diabetes.  For an amusing primer on detecting “bad science,” see Tom Naughton’s Science for Smart People talk on YouTube.)