If you already eat Primal, your email inboxes are most likely filling up with links to the story. Concerned mothers clutching the local paper’s “Health” section are calling (or, if they’re hip, texting). Smug vegetarian Facebook friends are posting the story on your wall, sans commentary. Yes, it’s about that time again. It’s another week, it’s another observational study by data-mining researchers hoping to establish a solid link between red meat and some chronic, horrific illness. So, what’s killing us this time? Well, considering that they’ve already done studies linking red meat to colorectal cancer, heart disease, and outright death, type 2 diabetes is next.
Here’s a link to the full study (PDF). Researchers drew on data from three large-scale dietary habit questionnaires of medical professionals to explore how red and processed meat intakes associated with the incidence of type 2 diabetes. The first set was the Health Professionals Follow-Up Study, which comprised 37,083 male physicians followed for 20 years; the second was the Nurses’ Health Study I, which included 79,570 female nurses followed for 28 years; and the third was the Nurses’ Health Study II, which followed 87,504 women for 14 years. These data were pooled with additional data from 442,101 participants in existing studies, so it was a big pile of numbers with which to work.
Sure enough, they found a link between processed meat intake and type 2 diabetes, with a smaller link between unprocessed red meat and the illness. A daily 50 gram serving of processed red meat was associated with a 51% greater risk of developing type 2 diabetes, while a 100 gram serving of fresh red meat represented a 19% increased risk. Unprocessed meat included “beef, lamb, or pork as main dish” (roasts, steaks, chops), “hamburgers” (but I’m sure they got bunless burgers, right?), and “beef, pork, or lamb as sandwich or mixed dish” (served up on coconut flour bread, no doubt). Processed meat meant “hot dogs,” “bacon,” “sausage, salami, bologna, and other processed red meats.”
To give a complete picture of the data, I’ll briefly discuss what the “heavy meat eaters” did when they weren’t eating red meat. You tell me if you notice any alarming trends that might have something to do with type 2 diabetes. Folks in the highest quintiles of meat intake were the least active and the most sedentary. They exercised the least and smoked the most tobacco. They drank more alcohol than any other quintile. They guzzled more soda and other sweetened beverages. In the high meat quintiles, folks ate 800 more calories per day than folks in the low meat quintiles. They were much heavier, too (all muscle, I’m sure). Trans fat intake was higher in the high-meat quintiles, too, as was potato intake (since these data included the years before trans fats were taken out of fast food deep fryers, I’m thinking these guys enjoyed a burger and French fry value meal on occasion). They ate the least amount of fiber from grains, indicating they probably ate the most refined grains, drank the most coffee, and ate the fewest fruits and vegetables. In short, people who ate the most red and processed meat were also the unhealthiest by both Primal and mainstream standards. And if what they were doing was actually healthy or neutral (like drink coffee and avoid fiber from grains), it wasn’t by design. These people (all health professionals, ironically) most likely didn’t particularly care about their health.
There are also variables that weren’t even considered that could have impacted the results. Added sugar, which many people heap into their coffee, wasn’t covered. They did cover the polyunsaturated fat:saturated fat ratio, but those numbers only incorporated the fat content of whole foods like nuts, meat, eggs, and dairy, not whatever cooking oils were used to fry up all that meat. So there are a couple other significant confounders.
You know how we’re always making the point that people who eat the most meat, except for us weirdo Primal types, are also the most likely to be unhealthy in other areas, to eat fast food, to be sedentary, and to smoke? Well, here’s direct evidence that this actually is the case.
And then there’s the issue of relative versus absolute risk (for a good primer on relative risk, check out Dr. Eades’ post on the subject). 51% greater risk sounds pretty awful, right? I mean, that’s over half. And when most people hear a figure like that, they take notice. They don’t really think about what the number means in terms of absolute risk, but they immediately link whatever risk factor is being highlighted to the big scary percentage figure – and the damage is done. In reality, the effect size is tiny. The absolute risk of getting diabetes was rather small for a participant of the studies mined for their data – about 7% over the course of the 14-28 year range. A 51% increase in risk bumps that up to about 10.5%, not 58%, while a 20% increase bumps risk up to around 8.4%, not 27%. I don’t fault the researchers for using statistics, because they’re totally valid, but I worry that the average person will see those big percentages and think that represents absolute risk. And then they give up their meat.
So what’s the problem with observational studies?
It’s not that there’s anything inherently wrong with an observational study. In fact, they’re extremely useful and downright necessary for generating hypotheses, but they cannot establish causality. As both Mat Lalonde and Tom Naughton consistently emphasized during their presentations at the Ancestral Health Symposium, we do science and we do our cause a disservice when we overstate the evidence drawn from an observational study. The same goes for health researchers, who to their credit usually do a good job keeping their conclusions (if they make any at all) conservative. It’s the science reporters that love definitive headlines and concrete conclusions.
“Associated with”? Ha!
“Relative risk”? Never heard of it!
“Causes”? Now we’re talking!
Check out the headlines, which range from the conservative “Red Meat Linked to Increased Risk of Type 2 Diabetes” to the bordering-on-unequivocal “Bacon ‘increases risk of diabetes.” The former hails from the Harvard School of Public Health’s PR department, which actually conducted the study, while the latter comes from the Irish Independent. It’s a perfect example of what’s wrong with science reporting. Those headlines are designed to draw you in and precondition your expectations for the content. Heck, you might not even have time for the full text, in which case you’ll be left with the idea that bacon causes diabetes. Maybe it does, maybe (probably) it doesn’t, but we can’t know from observation.
As I read the study, I was pleasantly surprised by the restraint of the authors. Sure, the science reporters took the limited evidence and ran with it, but you expect that from them. It doesn’t surprise me anymore. The actual researchers, though, seemed to make it abundantly clear that no concrete conclusions about causality could be inferred from the data analysis. They even went on to propose a few potential mechanisms, as if to suggest that, I dunno, more studies were required to establish any sort of causality. In fact, they fully and continually admitted the limitations of the study as they composed the text. Until, that is, the very last sentence of the full text:
“Therefore, from a public health point of view, reduction of red meat consumption, particularly processed red meat, and replacement of it with other healthy dietary components, should be considered to decrease T2D risk.”
In other (my) words, “Despite our repeated insistence on the limitations of observational data analyses, and the fact that both causality and biological mechanism have yet to be established and indeed cannot ever be established through an observational study such as the one described in this paper, we recommend that the public reduce their consumption of red meat. The evidence in favor of such a recommendation is weak, poor, inconclusive, and highly confounded by listed variables like sedentary lifestyles, smoking, alcohol consumption, caloric intake, soda drinking, and bodyweight, plus the variables we didn’t even consider, including the oil the meat was cooked in, the baked goods the meat was served upon, and sugar intake, but do it anyway. We’re from Harvard, by the way.”
So, in the end, should this observational study cause red meat-eaters to worry? No. The confounding variables are vast and the absolute risk is low. Plus, remember: you are not the typical meat eater. You cook your meat in good fats, you eat plenty of vegetables, you lift heavy, you walk, you enjoy life, you savor relaxation and understand the pitfalls of stress, seed oils, sugar, and sunlight deprivation. Should any observational study in the future force you screaming from your butter/red meat/full-fat dairy/deep squats/barefoot running, remember that fact. No observational study should make you fear anything – not if you’re thinking clearly, can make the distinction between relative and absolute risk, and are able to identify potential confounding variables. But it should keep you questioning things, which is exactly what we need if we’re to keep learning, growing, and progressing.