I’m sure we’ve all heard at one point or another that red meat is bad for our health. It’s one of the dogmas that’s been spouted by government health authorities for decades. Back in 2015, the WHO declared red meat to be a carcinogen. Here in Sweden, the Public Health Authority recommends that people limit intake of red meat to 500 grams per week. Personally, I probably eat at least twice that, so I guess I’m in big trouble.
The problem with these recommendations is that they are based on little to no evidence, mainly very low quality observational studies that show a marginally increased risk of cancer with increased red meat consumption, and debunked hypotheses, such as the cholesterol hypothesis (a.k.a. the diet-heart hypothesis), which states that saturated fat and cholesterol in the diet causes heart disease.
If saturated fat and cholesterol in the diet cause heart disease, then foods that contain a lot of these things, like red meat, should logically cause heart disease. But we now know that the cholesterol hypothesis is nonsense, as I’ve written about previously (although it lives on as a kind of zombie-hypothesis, in accordance with the principle that science advances one funeral at a time).
I think you can already guess what my personal biases are on the red meat issue. I’m inherently skeptical of the idea that red meat is unhealthy, for the simple reason that it has constituted a major part of our diets for at least the last couple of million years. Evolution generally doesn’t produce animals that become sick from the main components of their diets.
But maybe evolution decided to make an exception when it comes to humans. Luckily, three articles were published in the Annals of Internal Medicine in November 2019, all by the same group of researchers, that help shed light on whether red meat is bad for our health. The researchers received no specific funding and reported no conflicts of interest. Just to be clear before we begin, in case anyone is uncertain of the definition, red meat is meat that comes from mammals, in other words, cows, pigs, sheep, and so on.
The first of these articles was a systematic review and meta-analysis of cohort studies looking at the link between red meat consumption and cancer. A cohort study is a type of observational study in which you get two groups of people that are similar except for a single variable that you want to study, in this case red meat consumption, and then follow them over time to see what happens. It’s not as good as a randomized controlled trial, because no matter how hard you try to get rid of confounding variables, you’ll never be able to get rid of them all. But it’s the next best thing, and often it’s the only realistic option, since most people aren’t willing to be randomized to either a high- or low- red meat diet.
There is one very big confounder that affects all observational studies of red meat, that we need to be aware of before we even begin to look at the data, because it is virtually impossible to compensate for. It is the fact that we’ve all been told for decades that red meat is bad for us. This means that people who care about their health will tend to limit their intake of red meat, while people who don’t care so much, won’t. The people who care will also do a lot of other things that the people who don’t care won’t do. They’ll exercise more, eat more vegetables, smoke less, drink less alcohol, meditate more, and probably do a thousand other things differently from the people who don’t care so much. No matter how hard we try, it is impossible to correct for all these different confounders. For that reason, we shouldn’t accept small positive effects of reduced red meat consumption as real.
Let’s clarify this point by looking at smoking. There has never been a randomized controlled trial that shows smoking causes lung cancer, yet we accept this as true. Why? Because the observational studies show an enormous increase in risk of lung cancer in people who smoke. The relative risk of lung cancer in smokers is around 2,000% more than in non-smokers. In other words, there is a twenty-fold increase in lung cancer.
In order to accept a difference as real in an observational study with lots of scope for confounding, the difference should really be at least a doubling or a halving. If the difference is less than that, then the relationship is most likely not causal.
Anyway, let’s dig in to the data. As mentioned, the first review looked at cohort studies that compared intake of red meat and cancer incidence. In order to be included in the systematic review, studies had to have at least 1,000 participants. 118 studies were included in the analysis, with a total of six million participants.
So, what did they find?
A decrease in intake of red meat of three portions (roughly 360 grams) per week was correlated with a 7% reduction in the relative risk of cancer. However, the result was not statistically significant. When the researchers broke the data down by different types of cancer, they still were not able to find a statistically significant correlation between red meat and any type of cancer. Red meat did not correlate significantly with breast cancer, or prostate cancer, or gastric cancer, or esophageal cancer, or pancreatic cancer.
Red meat didn’t even correlate with colorectal cancer, which I was taught in medical school is definitely caused, at least in part, by eating red meat. In fact, when it comes to colorectal cancer, there wasn’t even a hint that red meat might increase the risk. The change in relative risk was exactly zero percent.
What can we conclude?
As mentioned earlier, anything less than a halving or doubling of relative risk in a study of this type isn’t a meaningful result. And here we have a piddling 7% relative risk reduction, that isn’t even statistically significant, in spite of the huge data set. So I think it is safe to say that the weight of evidence available from cohort studies is telling us that red meat does not increase the risk of cancer. Someone should tell the WHO.
Let’s move on to the second systematic review, which was looking to see if there was any correlation between intake of red meat and heart disease, as well as with overall risk of death. This review was also looking at cohort studies, so all the previous comments about confounding and being skeptical of small effects still apply. As in the previous review, studies were required to have at least 1,000 participants in order to be included in the analysis.
A total of 61 articles, with altogether four million participants, were included in the analysis. Here’s what they found.
A reduction in intake of red meat of three portions per week was correlated with a 7% relative reduction in mortality over the course of follow-up, which averaged eleven years. Just as with the previous review, the relationship was not statistically significant. There was also a 5% reduction in the relative risk of cardiovascular disease, which again, was not statistically significant.
There was a 10% relative reduction in risk of type 2 diabetes and a 6% relative reduction in risk of stroke. These two last ones were statistically significant according to the reviewers, but only because they’ve forgotten to adjust for the fact that they’re looking at a whole bunch of relationships at once. As I’ve written about many times on this blog, a p-value of 0,05 (equivalent to a confidence interval of 95%) only applies when you’re looking at a single relationship. If you look at multiple relationships, you have to use stricter criteria for statistical significance, or you increase your risk of false positive results enormously. Unfortunately no-one who does medical research seems to be aware of this, and none of the people who do peer-review for the major medical journals seem to be aware of it either. After adjustment for looking at multiple variables simultaneously, these relationships also lose their statistical significance.
We already know that red meat doesn’t cause type 2 diabetes because type 2 diabetes is non-existent in primitive hunter-gathering societies, which have a much higher intake of red meat than even the most carnivorous western societies. Yet the moment people from these hunter-gathering tribes switch to our westernized lifestyle, with a much lower intake of red meat (and, incidentally, a much higher intake of carbohydrates), the prevalence of type 2 diabetes explodes.
To me, the fact that pretty much any bad thing you look at is connected with a small 5-10% increase in relative risk when you increase your consumption of red meat by a few portions shows that it’s not the red meat itself that is the cause. It defies belief that the same single factor would cause whatever negative health outcome you care to look at. Rather, reduced intake of red meat is correlated with other healthy behaviors, because everyone “knows” that red meat is bad, and it is these other healthy behaviors that are confounding the results. By telling everyone that red meat is bad for decades, authorities have created a self fulfilling prophecy, where any bad health outcome you care to look at will be correlated with red meat intake. At least that will be the case as long as you rely on observational data.
Which is why we need data that isn’t observational. We need randomized controlled trials. Luckily, the third article by the group of researchers is a systematic review of randomized controlled trials of the health effects of red meat consumption.
In order to be included in the review, the difference in red meat consumption between the intervention group and the control group had to be at least one serving per week, and participants had to be followed for at least six months. Twelve trials were identified that fulfilled these criteria, with the smallest including only 32 participants and the largest including 49,000 participants.
Unfortunately, only two of the trials provided data on how many people died during follow-up, and on how many people had serious cardiovascular events during follow-up. One of these had 49,000 participants (the famous Women’s Health Initiative study), while the other only had 600 participants and was stopped early (a big no-no, as any regular reader of this blog knows, because it invalidates the results). The researchers therefore very reasonably decided to ignore the study with 600 participants and base their conclusions on the study with 49,000 participants.
An advantage with this study, other than the enormous size, was that it followed patients for over a decade, which is a fantastically long follow-up period for a randomized controlled trial.
The main disadvantage was that it only included post-menopausal women, which makes it harder to draw firm all-encompassing conclusions. Will men and women be affected differently by eating red meat? Will pre- and postmenopausal women be affected differently? Who knows. My guess is probably not, but without data it’s impossible to say.
Another disadvantage is that the study wasn’t really looking at what happens when you reduce red meat. It was looking at what happens when you reduce fat intake, and the reduction in red meat was an indirect consequence of reducing fat. What this means is that confounders are being let in through the back door. If you do find a difference between the groups, is it due to the red meat reduction, or the fat reduction, or some other indirect dietary change? It’s impossible to know.
Randomized controlled trials of diet interventions often have this problem, because people will always compensate for changes in their diet. If people are told to eat less of one thing, they will usually eat more of other things. And foods are not like drugs, that usually only contain a single active substance. Even simple foods contain a huge number of different substances, all of which have the potential to confound the results.
Obviously this is a big problem – if this study did turn out to show a reduction in bad things with a reduction in red meat intake, it would be impossible to say whether it was due to the red meat or to some confounding factor, so in that sense, this data is not much better than the observational cohort study data we’ve already discussed.
On average, the participants in the intervention group reduced their intake of red meat by 20%, which worked out to 168 grams per week. So, what happened?
In the intervention group there was a 1% reduction in the relative risk of death. However, it was not even close to being statistically significant. Considering that this was a study with almost 50,000 people, we can be pretty sure that increased red meat consumption does not increase your risk of pre-mature death to any meaningful extent, at least if you’re a post-menopausal woman.
Data was also provided on death from cardiovascular disease, stroke, breast cancer, and cancer more generally. These also all showed marginal relative risk reductions with reduced intake of red meat, none of which were anywhere close to being statistically significant.
So that’s it. This is the sum of evidence that exists on red meat and health outcomes. I think the data we’ve discussed in this article illustrate quite clearly that the official dietary recommendations to limit intake of red meat are not supported by the evidence. These studies constitute the best available evidence that exists on red meat and health, and none of them can show that red meat results in any negative health outcomes. The evidence that does exist is extremely weak and rife with confounders, the effect sizes are tiny, and there is no statistically significant relationship anywhere you look.
What can we conclude?
The best available evidence suggests that there are no negative health consequences associated with eating red meat. If there are any negative consequences, the effect sizes are so tiny that they’re really not even remotely worth bothering about.