What We (Don't) Know About Nutrition: Red Meat

While I was writing another article related to nutrition (coming out soon), I found myself plunging rather deep in a particular nutritional rabbit-hole. Namely: is red meat consumption bad for you?

The commonly accepted idea is that it is. But as we'll see, there actually is little evidence that it is the case.

I think my investigation of this is an excellent way to demonstrate quite a few interesting ideas:

It's incredibly difficult to link diet to health outcomes, and the impact variations in reasonable diets have is fairly small.

This was the original point I was trying to make with the red meat litterature example.

If you have no egregious nutrient deficiencies and you keep your bodyfat percentage in check, I believe the impact of nutrition on your long-term health will be a couple percents (at best) change in the chance to get or not get some kind of cancer, etc.

You should also sleep well and do some exercise, but those are non-nutrition-related.

You shouldn't necessarily trust the consensus in some branches of science (like nutrition).

However, that doesn't mean you should just stick to whatever opinion you had before, which is probably not based on facts, or merely on anecdote.

Instead, you should research the topic yourself (preferrably) or find people whose scientific judgement you trust (not easy, people tend to pick a camp).

It's hard to do good science.

Even the published litterature is rife with confounders and bias. And people forget that correlation doesn't imply causation.

There's also some bad faith going on in some higher instances.

Bonus: cool Slate Star Codex articles about confounders: one, two.

Absence of evidence is evidence of absence.

I stole this one from LessWrong, but basically it means that if you try to find evidence of X (e.g. red meat increases mortality), and you find none or very little, that's evidence that X is probably not true.

And last but not least:

Relax, that ribeye steak is probably not going to kill you.

So let's get right into it.

Anecdotes and Confounders

It's easy to find exemples of people who've follow a non-conventional diet (veganism, keto, carnivore) for a long time and seem to be doing quite fine. It's also easy to find the opposite, but if you dig, you'll often find other factors that can explain adverse health outcomes. The only way to link food and health outcomes is to do studies with fairly large populations... and to do them well. Your experience with your zombie-looking vegan neighbour and y our fat meathead friend don't prove anything.

To give an example how studies can go astray, let's talk about red meat. Red meat gets a bad rap as a cause of cancer and heart diseases, but no studies has ever found that red meat causes those. Instead, quite a few studies have found that high red meat consumption is correlated with these issues (and correlation does not imply causation).

This is still pretty bad... excepted if the study fail to control relevant variables. The first that they should look for is bodyfat percentage. If you eat more meat but you also eat more of everything, yes, your health outcomes are going to look much gloomier, no surprise. Maybe it means you eat less vegetables and your gut flora suffers, also an issue. Unfortunately, poor control of confounders is the norm.

An Outrageous Guideline

Something interesting occurred recently (November 2019). The prestigious "Annals of Internal Medicine" journal published some dietary guidelines for red meat consumption:

The panel suggests that adults continue current unprocessed red meat consumption (weak recommendation, low-certainty evidence). Similarly, the panel suggests adults continue current processed meat consumption (weak recommendation, low-certainty evidence).

The panel's recommendation was based on 5 meta-reviews (more properly "systematic review and meta-analysis"). Three of those (1, 2, 3) looked at "observational studies" which study meat consumption in the context of people's existing habits. These tend to be highly-confounded, for the reasons I explained above. In, in fact, the studies were found to be confounded:

Limitation: Inadequate adjustment for known confounders, residual confounding due to observational design, and recall bias associated with dietary measurement.

Limitations: Limited causal inferences due to residual confounding in observational studies, risk of bias due to limitations in diet assessment and adjustment for confounders, recall bias in dietary assessment, and insufficient data for planned subgroup analyses

Limitations: Observational studies are prone to residual confounding, and these studies provide low- or very-low-certainty evidence according to the GRADE criteria.

A fourth review just considered people's preference and concludes they really like meat, even when faced with potential health consequences.

The fifth meta-review (pdf) is more interesting, as it reviews randomized trials. Instead of observing people's existing diet, diets are assigned to people randomly, which reduces confounding with other factors. But unfortunately, there are very few studies who did this over a long period of time (which is necessary to assess health outcomes), for ethical reasons. The conclusion ends being up based on a single study, which found no adverse health outcomes. Unfortunately, it's not a low-red-meat study, but a low-fat-diet study in which the low-fat group ended up consuming less red meat than the control group: on average 1.4 less "servings" per week, which is not huge.

From Ignorance to Knowledge

All this may sound silly — do we actually know anything?

Well, yes we do! What this tells us is that there is no blatant evidence (or not even much evidence) for the hypothesis that red meat increases mortality.

From the data we have (and the goal of these studies is to be relatively comprehensive, including all studies whose data can be exploited which are not methodologically hopeless), we know that:

  1. It is highly unlikely that eating more red meat decreases mortality.
  2. It is unlikely that eating less red meat increases mortality.
  3. It is highly unlikely that eating more red meat significantly increases mortality.

(1) is already consensus. As for (2), it is certainly possible that red meat increases mortality (though (1) is possible too) but if that were the case, we'd expect to see much stronger evidence. Regarding (3), even if there is an increase in mortality, it is unlikely to be high.

To distinguish (2) and (3), let's think about them in another way. The paucity of evidence that we see means that either red meat doesn't increase mortality at all (2), or that it does increase mortality, but ever so slightly so that it's hard to distinguish the mortality of red-meat-eaters compared to that of non-red-meat-eaters.

It's important to note that we learned something: we went from a state of not having evidence to a state of having evidence, with the evidence favoring the null hypothesis: that red meat consumption matters very little.

If you draw a line representing a continuum of hypotheses, from -100 = "red meat consumption decreases mortality" to 0 = "red meat consumption doesn't change morality" to 100 = "red meat consumption increases mortality, the evidence we have puts us at around 1.

                 👇 evidence points here
-100 ---------- 0 ---------- 100
meat 👼      meat 🤷     meat ☠️ 

Controversy and The Shills

The good people at the Harvard School of Public Health weren't too happy with the Annals of Internal Medicine's guidelines, as evidenced by this article.

I should probably warn you, I'm going to say bad things about the good people of the Harvard School of Public Health.

In my opinion, their most egregious statement is the following:

The publication of these studies and the meat guidelines in a major medical journal is unfortunate because following the new guidelines may potentially harm individuals’ health, public health, and planetary health. It may also harm the credibility of nutrition science and erode public trust in scientific research. In addition, it may lead to further misuse of systematic reviews and meta-analyses, which could ultimately result in further confusion among the general public and health professionals.

You should the (pretty short) article yourself. Essentially it tries to convince you that these guidelines are a bad thing.

One of their better argument is that the recommendations are bad because it says environmental considerations are out of scope. That is a whole other can of worm, and I personally don't have an opinion on what the environmental impact of meat production is. This also veers political. But personally, I'm not very interested in the recommendation but rather in the question: "Is red meat bad for your health?".

It's here that Harvard sins, because they try to say that it is bad for your health. In the Q&A section of the article, they give a mostly accurate and fair criticism of the meta-reviews. But then they try to spin this by saying that the meta-reviews show that red meat consumption is bad for you. Except, as we already said, it shows very very weak evidence of that.

Let's update our graph from before:

                 👇 evidence points here
-100 ---------- 0 ---------- 100
meat 👼      meat 🤷    | meat ☠️ 
                       👆
                    where Harvard SPH says the evidence points

And that's why they're being disingenuous.

Let's also plug the reaction of Gwern Branwern on that article:

You don't see any problem with the Noble Lie? 'They should not publish these results showing our nutrition research is bullshit, because it may justifiably erode public trust to know that nutrition research is bullshit, and it might spark further research which would lead to "further confusion" as the bullshit is further exposed; we don't think there is any downside to lying to the public about meat consumption, damaging lifestyles and ignore the possibility that our own mistaken results are themselves damaging public health but will pretend it's a one-sided wager - "heads we win tails meat loses".'

By the way, remember I spoke about people whose scientific judgement you trust? Gwern is someone like that, he has great in-depth research up on his website.

Is Harvard really anti-meat? Well...

TMAO & The Red Scare

Harvard is a big entity, which is probably not anti-meat as a whole.

That being said, Harvard Medical School did publish this online article.

A substance called trimethylamine N-oxide, which is produced when your body digests red meat, may raise the risk of cardiovascular problems.

Finally! A causal link from red meat to harm. Hallelujah!

Well they only reference (and not even link) is to a paywalled editorial, but there is actually pretty solid research linking red meat with high TMAO. Case solved?

Well no, because the link between TMAO and cardiovascular problems is shaky, see for instance this paper and this quite nicely written article titled "How TMAO Fooled Us".

Apparently, fish increases TMAO even more than red meat, but everyone knows that fish is healhty! Some people really aren't bothered by apparent contradictions.


Well that's it for today. I hope this was entertaining/interesting.

Psst, if you're looking for the conclusions, they're actually at the top of the article!