Updated: Nov 5
The Harvard School of Public Health loves epidemiological studies. I mean just luurves them. Their newest article, published in the American Journal of Clinical Nutrition says that red meat is positively correlated with an increased risk of Type 2 Diabetes.
I'm not sure where to even start with this one. The study is so rife with issues, it should not have been published. But it was highly successful, in that it secured all the necessary headlines to scare people further from eating the most nutrient-dense foods available to us.
Where to start? Let's start with the headliner: diabetes. Diabetes, as you may know, is when the body is unable to manage glucose (i.e. sugar) in the body. Either your insulin production is impaired, or your cells aren't answering the door when insulin comes knocking. Either way, it's related to glucose regulation. Well, not sure the last time you had a sugary rib eye, but meat doesn't contain any glucose.
Hmm, I wonder which foods do contain lots of glucose that could trigger a potential for diabetes? It's so complicated, I can't think what those could be? What in our modern food landscape contains lots of glucose? Nothing comes to mind. Even the study authors didn't think there was anything that really stood out on that front, so they outright didn't report the actual intake from grains or sugars - which are key sources of glucose in the diet. Is that convenient, negligent, or just ridiculous?
There's more... As I mentioned above, this was an epidemiological study. This means that the researchers gathered a whole bunch of data on people, tracked them for a while, gathered more data, and then tried to see if they could identify any patterns between those groups' behaviors and whatever issue was being studied. The results are based on correlations only; there is no intervention. Particularly meaningful correlations are supposed to inspire researchers to go on to conduct more reliable, robust clinical studies. Because, as we all know, correlation does not imply causation. But given funding constraints and the bigger budget required of high-quality clinical studies, no one bothers to do that much any more. The buck ends up stopping here. So instead, the headlines spew out cooked numbers and weak associations as solemn truths.
Diabetes is about an inability to manage glucose... Meat doesn't contain glucose.
And just so you know, in this study the associations were so weak, they would not have warranted going further with a clinical trial, even if they wanted to.
What else? Well, to gather the nutritional data, they used Food Frequency Questionnaires. These tools are notoriously unreliable. Why? Because they literally ask you to remember and document all the types of foods you ate for the past year. Do you remember how often you ate eggs in 2022? Well does that include just eggs for breakfast - like scrambled, easy-over, etc.? Or does it also include how many may have been in baked goods, like in cake, cookies, and other foods? I don't know. And I suspect a lot of people answering these FFQs don't know either. Yet the authors of these studies take the vague, "semiquantitative," and unreliable, selections of the respondents and convert it into exact data, lending the illusion of precision and accuracy. Take a look at an excerpt of the Harvard FFQ below to see what I mean:
You'll also see above that sandwiches, lasagna, and frozen dinners count as "meat." Think of the last piece of lasagna you had or a sandwich - what percentage of that meal was meat versus bread, pasta, sauce, toppings, etc.? Does that really give a clear indication of how much meat a person is consuming? Fine, let's pretend it does. If I select, say, 2-4 sandwiches per week, how do the authors end up with an exact number to run data with? The semi-quantitative selections are converted by the study authors, using frequency weights. So a selection of 2-4 sandwiches per week becomes 0.43 servings per day. Someone eating 5-6 sandwiches per week becomes 0.8 servings per day. Eating 4 sandwiches versus 5 sandwiches per week can practically halve the daily serving size. You can quickly see how this data becomes muddled and not reflective of an individual person's food intake, much less something useful across a population study where everyone is reporting "data" based on their best recollection and understanding of the question. The Food Frequency Questionnaire is designed to be self-administered, no opportunity for questions asked, no answers given.
Sandwiches, burgers and lasagna are counted as "meat."
There are further limitations of these types of questionnaires. Note above that someone who eats burgers and hot dogs, along with fries and a soda, is lumped in the same group as someone eating just homemade roast beef. The servings don't differentiate based on quality. The authors also, for some reason did not adjust for differences in BMI (body mass index) in the model that provided the press release and headline data. The individuals who ate the most meat had the highest BMI. The authors said that because "weight gain mediates at least part of the association between red meat intake and risk of T2D, we did not adjust for" it. But that's circular reasoning; it's not sound science. BMI may have been higher because this group ate more sugar and grains. We'll never know, because they didn't reveal that data in their paper. They also admitted that half the excess risk is explained by excess body fat. The individuals with higher risk interestingly also consumed more calories and were less physically active.
So, let's back up a second. They saw clear differences in the people with excess risk, and it wasn't just their meat consumption. They claimed to control for the grains and sugar, but here they're just talking about completely different people with different lifestyles.
Ok, last point, and it's a good one... The article's super-sexy claim is that eating red meat even twice a week can increase your risk of developing Type 2 Diabetes later in life by 62%. This is what's known as 'relative risk.' Most studies report relative risk because it almost always sounds more dramatic. For example, imagine your odds of winning the lottery are 1 in 1 million and I offer to double your odds. The relative increase will sound very impressive - I've increased your odds by 100% - woohoo! But the absolute increase of your odds of winning the lottery is negligible - it's increased from 0.0001% to 0.0002%.
In the context of this study, we want to know what the 'absolute' risk increase is of anyone developing diabetes. Studies seldom, if ever, include it, because, as mentioned, it's usually nothing to write home about. So, you have to go in and compute it yourself. This study is no different. Let's take the total number of person years that they tracked, 5,483,981, and the number of individuals who went on to develop diabetes, which is 22,761 cases. That's an incidence rate of 0.42%. A 62% relative risk increase equals 0.68% absolute risk increase, a difference of 0.26%. Umm, are you seeing the decimals in these numbers? That's less than 1% increased risk - assuming their data is solid, which is highly debatable for reasons already discussed above. Hardly makes for a riveting headline, does it?
The absolute risk increase of eating red meat twice per week is less than 1%.
As a rule, when you hear obnoxiously large risk numbers like the 62% that was reported in over a hundred headlines around the world, always assume that they're sharing relative risk. Dig around a little and make your old math teacher proud by computing the absolute risk (or ask chatGPT to do it for you), and you'll more than likely find a far more modest, non-headline-making difference.
Red meat has been eaten by humans for over 2.5 million years. It is the most bio-available form of iron and other key nutrients necessary for good health. In the US, red meat consumption has decreased significantly since the 1970s, and yet Type 2 Diabetes incidence has sky-rocketed dramatically over this same time period. Shouldn't there be some overlap in those two trends if they are indeed correlated as much as this study implies? Take a look at the following table.
Diabetes (the blue line) has had an undeniable and rapid increase. And yes something is definitely driving this disturbing trend. But take a look at the trend in red meat consumption (red line) in the US during the same period. You don't have to be a statistician to see that those two trends are negatively correlated.
In the end, I have to conclude, and I hope you do too, that the Harvard study is not very strong, robust or well-done science. But the disheartening part is that it has been accepted as such. It has made all the headlines across all the major media outlets, online and on TV. The study makes zero comment on the rise of ultra-processed foods in the US diet and it's impact on blood glucose and diabetes trends.
Red meat consumption in the US has steadily decreased while diabetes incidence has rapidly increased.
The studies against red meat have been steady and unrelenting. I have some theories as to why that is, but that's a topic for another post. The study authors offer their opinion (rather unconventional in such papers, tbh), that eating less meat is better for the environment as another reason why red meat should be avoided. I've addressed that claim in this previous post to some extent. And Nina Teicholz, an amazing journalist, author of Big Fat Surprise, and head of the nonprofit Nutrition Coalition addresses an interesting perspective on Harvard's anti-meat leanings in this article.
I hope this post gave you a little food for thought and you might more critically evaluate the headlines that we scroll past each day telling us one thing or another about all manner of information, particularly those related to health and food choices. It's unfortunate that loud, powerful voices can drown out common sense and shared history.
Gu, X., Drouin-Chartier, J. P., Sacks, F. M., Hu, F. B., Rosner, B., & Willett, W. C. (2023). Red meat intake and risk of type 2 diabetes in a prospective cohort study of United States females and males. The American Journal of Clinical Nutrition. https://ajcn.nutrition.org/article/S0002-9165(23)66119-2/fulltext
Food Frequency Questionnaire. Harvard T.H. Chan School of Public Health. https://www.hsph.harvard.edu/nutrition-questionnaire-service-center/general-documentation/