Food, Politics, science

Correlation Is Not Causation: Let Them Eat Meat!

In the first pages of The China Study (aka the ‘vegan bible’ which blames the rise of chronic diseases on animal protein consumption), Dr. Colin Campbell presents the following graph:

Screen Shot 2015-02-19 at 09.40.52

What do we see? An almost perfect correlation between daily meat consumption and colon cancer incidence in women. The countries in which people eat the most meat have the highest incidence of the disease. Therefore, if we want to avoid colon cancer, we should eat less meat.


Not so fast.

I would like to walk you through the pitfalls of interpreting this kind of observational data, and explore perhaps the most controversial and ill-interpreted correlation to date: meat intake and chronic diseases.

Does the graph above really prove that meat causes colon cancer? Let me show you something.
Using some of the incredible data available on GapMinder, the data geek inside of me produced this stunning piece of science: (data source is the IARC and the World Bank) (I should also get props for copying Dr. Campbell’s graph style so well)
Screen Shot 2015-02-18 at 13.34.00

Tadaa! I have clearly just proven that since countries with a higher percentage of internet users have higher breast cancer rates, the internet causes breast cancer. Therefore women who wish to avoid breast cancer should stop going online.

(round of applause)

I hope that you are rolling your eyes right now. What have I actually proven? That there exists a correlation between internet use and breast cancer rates. Countries with a higher percentage of internet users also have higher rates of breast cancer. I have in no way proven that using the internet causes breast cancer. In the same way, Dr. Campbell has not proven that meat consumption causes colon cancer. He has merely pointed out a correlation.

Correlation does not imply causation.

Be skeptical towards media headlines which link a diet to a health outcome. If the data comes from an observational study, nothing has been proved. Just because two things vary in a similar way over time or across countries, does not mean that one is the cause of the other.

But we are human, and therefore, we are biased. We more readily extrapolate a correlation to be a causation when it reinforces a common perception (in our case, that meat is unhealthy). We aren’t fooled so much when observational data correlates two non-sensical variables.

Thank you, Mexican Lemons, for preventing car accidents.

Of course, there is value in observational studies. They help develop hypotheses which are then tested in randomised controlled trials or in clinical studies. The goal is to understand why this correlation exists.

A correlation result from an observational study (example: drinking coffee correlates to less incidence of Type 2 Diabetes) is not verified until the biochemical mechanism underlying it is found (coffee inhibits the formation of amyloids, preventing pancreatic beta cells from degrading, therefore decreasing diabetes risk).


In our case, meat has never been proven to cause cancer. However, there have been myriad observational studies displaying very robust and consistent data linking meat intake to increased incidence of disease. Perhaps a mechanism will be found one day which actually proves how specific compounds in red meat create tumors. In the meantime, we can only try to explain the correlation.

Sinha et al. looked at over 500,000 people, and divided them into groups, from those who ate the least meat to those who ate the most. The researchers observed the rates of death within these groups. They found that the more meat someone ate, the more likely the person was to die. Eating more meat = more death. But it wasn’t just death by cancer or heart disease (the usual correlations)- in fact, this observational study linked meat intake to all causes of death.

Screen Shot 2015-02-19 at 09.43.41
Screen Shot 2015-02-19 at 09.43.45

That’s right, an increase in meat intake was correlated to a higher incidence of cardiovascular disease and cancer, but was also a correlated to higher likelihood of death by car accident, lightning, injury, murder, gun wounds, you name it.

But why? And how? Can we really blame meat for car accidents? And by the same accord, can we really blame meat for cancer and CVD deaths? We can safely assume that there exists no biochemical mechanism through which eating meat increases your nemesis’ desire to kill, makes you a worse driver, or makes it more likely for you to injure yourself.


Sometimes, a correlation hides something else. Let’s say you find data which links hair color to skin cancer incidence; the blonder a population is, the higher their skin cancer rate. The media would say: ‘dying your hair a dark color will prevent skin cancer”. What’s probably going on is that hair color is an indicator of skin color: a fairer skin is more prone to sunburns, and skin cancer. Changing the color of your hair will have no effect, but protecting your skin will.

What if meat intake itself is not the culprit, but is merely and indicator of other risk factors?

Let’s have a little look at some possible explanations for this meat-disease correlation, courtesy of yours truly:

Quality: the vast majority of the meat eaten in our society is industrialised and processed. Cattle is fed an ill-adapted diet, is sick, lives in unsanitary conditions, and is pumped with antibiotics, growth hormones, pesticides, antidepressants… The meat we eat or find in most fast (and slow!) food restaurants isn’t healthy by any standard. Perhaps the compounds found in ‘modern meat’ and its inherent bad quality are what explains that a diet high in meat seems to have a harmful effect.

Accompaniments: When you think meat, you think french fries, hotdogs, hamburgers, with barbecue sauce or ketchup. Maybe even a milkshake and some sort of sweet desert. Let’s be real, we never eat meat by itself. In fact, it is safe to assume that the average high-meat diet is also a high-carbohydrate diet. Controlled trials have shown that cutting carbohydrates and replacing them with either fat or protein always reduces blood pressure and LDL cholesterol. The culprit for cardiovascular diseases could therefore be the carbohydrate consumption which correlates with meat consumption, and not the meat itself. If you want to keep your heart in check, cut the carbs.

Screen Shot 2015-02-19 at 09.15.49
Lifestyle : What does a diet high in meat indicate? A greater likelihood to engage in a lifestyle considered unhealthy: smoking, drinking, not getting much exercise, eating little amounts of fish, fruit, and vegetables. Because of the stigma against meat, people who make healthy choices and are conscious of their health (regular exercise, lots of fruit and vegetables) tend to also eat less meat. So the low-meat-eating group could be renamed the ‘health-conscious’ group. If someone eats more meat, they are in effect going against what society considers ‘healthy’- so they probably also engage in other correlated ‘risky’ activities. Maybe they are more likely to own a motorcycle, or a gun, or drive faster, too. All these factors could explain the increased risk of dying of ‘all other causes’ for high meat eaters.

Here’s the bottom line: beware of observational studies. The link between meat and cancer is nowhere near as clear as the media would like us to think.

And the takeaway: choosing grassfed/pastured meats and cutting refined carbohydrates will have more health benefits than cutting meat altogether.

Happy hacking!


9 thoughts on “Correlation Is Not Causation: Let Them Eat Meat!

  1. Presumably Campbell’s data set is a bit old. One feature of NZ society was, at least until recently, the vast amount of sweet baking done by women at home. This food would be shared around when their friends came to call. Bakeries selling sweet pastry were everywhere. And women- everywhere – don’t eat most of the meat. It goes to the men and the growing kids. The per caput meat consumption is NOT the consumption by women (NZ didn’t have food surveys to tell you what women ate specifically, but Gloria Steinem will tell you they didn’t eat their fair share of meat). Women were smoking by now in NZ – they didn’t smoke in Japan.
    So you have these women, their colons full of flour and sugar, sitting around smoking together. Anything else? New Zealand is a selenium-deficient country, this is partly fixed with imported food today. SHEEP will get colon cancer in parts of the country that are selenium-deficient. Sheep don’t eat meat.
    I rest my case.


  2. Sylvia Zohrabian says:

    This piece highlights one of the biggest issues with cancer research today. Cancer progression is due to various factors, and very few studies conduct multivariable analysis of these risk factors. Correlation is not Causation…Brillant!

    Liked by 1 person

  3. Pingback: Listen To Your Heart: Feed It Bacon | HEALTH HACKING

  4. says:

    This article is the epitome of a self serving retard making a fool of herself. The opening analogy for a correlation between breast cancer and the internet is the weakest analogy that has zero merit and the two said to be contributing subjects have zero to do with one another. Stopped reading after that….


    • But that’s precisely the point, don’t you see? There are an infinity of correlations between the most disparate of things, so why should it be any surprise that there are just as many correlations when it comes to food and health? The author picked an outlandish analogy precisely to demonstrate that it isn’t to be trusted.
      The only thing that correlation is useful for is indicating that it might be a good idea to do some research on the matter and figure out whether there is any causation between the two data sets or if there is an unknown third factor that could account for both, which, coincidentally, happens to be the case.

      You’re so set in your ways that even pure logic doesn’t seem to sway you anymore. That’s worrisome.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s