Is Red Meat Giving People Diabetes?
The most contentious food online.
Red meat is one of the most contentious foods online. We love to hate it for pretty obvious reasons - there’s a certain moral righteousness aspect to it. The fact that red meat production is generally very harmful to our planet in a number of ways means that any story that shows it’s also unhealthy will get everyone very excited. There’s also the whole keto/carnivore movement, full of its own self-righteous steroid-taking hypocrites, and it’s always funny to see people like that taken down a peg or two.
The latest furore about red meat comes from a study conducted at the Harvard school of public health. According to the reporting, red meat in our diets may be “fuelling the type 2 diabetes epidemic” and that you can prevent diabetes simply by leaving red meat out of your diet.
Unfortunately (or fortunately, depending on your point of view), the story about red meat is much more complicated than that. We have known for some time that people who eat more red meat are less healthy in many ways than people who eat less of the stuff, but there’s no strong evidence that red meat causes OR prevents diabetes.
Let’s look at the science.
Nutritional Epidemiology
The new study is a relatively complex example of a field that I’ve written about quite a bit over the years: nutritional epidemiology. Broadly, nutritional epi is the practice of asking people about their dietary habits in the form of what’s called a food frequency questionnaire (FFQ), which is a massive form with 100s of questions about what you’ve eaten in the last year, taking a bunch of biological measurements, and then following them up in a few decades to see how many of them have gotten sick, died, or similar*.
In this case, the authors took three large databases that assessed food intake and health metrics, and looked at the association between various different types of iron intake and whether people got diabetes over the course of the study. They found that people who had more haem iron in their diet were more likely to get diabetes than those who had less, but other types of dietary iron weren’t associated with a greater risk of diabetes.
Now, you may be a bit confused - haem iron isn’t quite the same as red meat. Technically, this is a surrogate marker that gives us an indication of red meat intake, because most dietary haem iron comes from red meat, rather than a specific measurement of meat intake.
However, in this case the researchers actually calculated the haem iron intake directly from the FFQs where people reported their food intake, which means that haem iron in this study is mostly a direct measurement of how much meat people reported eating. While you do get some haem iron from other sources like chicken and salmon, the dietary contribution is much lower than red meat.
The authors of this study also did some more complex analysis on blood biomarkers to assess how the association between haem iron and diabetes might be coming about. They found some evidence that various blood markers were associated in fairly complex ways with diabetes, and then constructed a score which predicted diabetes with some accuracy. They also ran some sensitivity analyses including a variety of additional potential confounding factors which all showed pretty much the same relationship between haem iron intake and diabetes.
As far as nutritional epi studies go, this was quite a solid piece of work. There are some genuinely interesting relationships here that would be very useful to read more about, and I hope the research team does more work on this in the future.
However, I’m not putting my steak knives away just yet. Why?
This sort of research is generally just not that convincing for a number of reasons. There’s the big issue with FFQs and measuring dietary intake. People are extremely bad at reporting how much of various foods they eat. We’ve known for some time that relying on FFQs - even repeated FFQs as this study did - does not give a very reliable estimate of diet. In this paper, the researchers actually compared FFQs with another form of dietary measurement where people report only the foods they ate in the last 7 days and found that using the two methods gave quite different estimates of total iron intake.
It’s also just not possible to run a study like this and interpret the outcomes causally. We’ve known for some time that most observational studies show an association between how much red meat people report eating and negative health outcomes like diabetes and heart disease. I’ve written about the issues with these studies many times before. You can statistically control for the things you measure, but there’s just no way of knowing whether unmeasured or unaccounted factors may be causing the numeric relationships you see in the study itself.
For example, this particular study did not adjust for socioeconomic status in their main models. In a sensitivity analysis, they did adjust for area-level socioeconomics, but this is a very crude measure of people’s relative wealth and wellbeing. It’s entirely possible that the associations seen here are more about what rich people tend to eat (and which groups in America eat more red meat) than they are about the health effects of red meat itself.
It’s also problematic, in these studies, to put too much into your statistical model. It’s important to mitigate the effect of confounders, but if you include things like mediators - factors that impact the relationship between two variables - you can also introduce problems in your analysis. As one example, the authors controlled for total energy intake as a confounder, but given that energy intake was also calculated from the same FFQ as haem iron it’s entirely possible that this is inappropriate to include in the statistical model.
The point isn’t that this is a bad study, it’s that this study design is limited. We can say that most large datasets in the world show some association between red meat intake and diabetes, but that doesn’t tell us whether red meat intake causes diabetes.
Interventional Studies
There is a type of research that can give us a decent answer as to whether red meat causes diabetes, and that’s randomized controlled trials. If you randomize people to eat more or less red meat, and then follow them up months later, you can get a good estimate of whether red meat is making diabetes markers like blood sugar get worse.
Here’s a 2022 systematic review and meta-analysis of such studies. After scouring the internet for reserach, the authors included 21 studies into a statistical model estimating the overall impact of getting people to eat more or less red meat on diabetes markers. They found that there was no impact on any of the main factors, including different measures of blood sugar, pancreatic function, or insulin. In fact, they found a small benefit for red meat intake on one specific marker of blood glucose.
It’s almost impossible to square these results with a situation where red meat intake causes or cures diabetes. It’s hard to understand how red meat could cause diabetes diagnoses in observational research but have no major impact on blood glucose, pancreatic function, or other markers of diabetes in experimental studies. It’s possible that there’s some extremely difficult to uncover mechanism that’s only possible to see in massive datasets, but a much more likely explanation is that red meat just isn’t as bad for your health as many people say it is.
That’s not to say red meat is brilliant for you. It’s also certainly a bad thing for the planet, and there are ethical concerns with much of modern agriculture. If you don’t want to eat meat because it’s causing global warming (it is), or because you are concerned that farming practices are inhumane, go ahead.
But if you’re just looking for some advice on how to manage your health, just know that there are many different ‘healthy’ diets, and some of them contain red meat.
*Note: Technically, this isn’t quite accurate. The massive surge in nutritional epi in the 2010s and 2020s is more a phenomenon of the 90s than it is a modern thing - what actually happened was that a bunch of researchers set up these huge longitudinal cohorts in the mid-90s and early 00s. Those cohorts often asked people what they ate, and they now have decades of follow-up data especially with modern data linkage methods. That gives researchers the opportunity to apply to use that data to ask interesting questions, but it does mean that the vast majority of this type of study is done by people who didn’t actually collect any of the data themselves.
So sugar doesn't enter the picture? So, less red meat and more maple bars? Yes, I am being snarky here.
A good piece. However, I disagree with the assertion that red meat "is certainly a bad thing for the planet. I presume you think that because you see carbon dioxide as being a serious problem. Or do you think it's bad for the planet in some other way?