Thread by @GidMK

A lot of talk about processed foods at the moment, so I thought I'd put together a quick explainer 1/

This attention is because of two recent studies - one from France, one from Spain - both of which showed that "ultra-processed" food was linked to poor health 2/

You can find the studies here 3/

bmj.com/content/365/bm…
bmj.com/content/365/bm…

What both of these studies did was very similar. They took a large cohort of people - 20,000 and 100,000 - and looked at the association between what foods people said they ate and their long-term health 4/

Both of these studies were large, well-controlled epidemiological trials, done with precision and a lot of thought/consideration to try and make sure that the findings were robust 5/

(Basically, these were really good studies that most epidemiologists would be proud to have put out. Two of the best studies I've seen in the last 12 months) 6/

So what did they find?

Eating "ultra-processed" foods was associated with an increased risk of heart disease and death 7/

The first question is obvious - what are "ultra-processed" foods?

both studies used the NOVA definition, which looks like this 8/

Now, if this looks a bit arbitrary to you, that's because it sort of is. "Ultra-processed" is, in many ways, as much as judgement call as anything

For example, eating a steak is minimally processed, eating the same meat as meatballs is ultra-processed instead! 9/

The main idea, however, is to separate out foods that are high in traditionally unhealthy things - salt, fat, sugar - from ones that aren't 10/

It seems, based on these studies, that we should stop eating ultra-processed foods entirely!

But there are a few caveats to consider here 11/

The first thing to think about is absolute risk. If you look at these studies, the absolute risk difference between the highest and lowest consumers of ultra-processed foods was quite small 12/

To put this into context, the "62% increased risk" of death that a lot of people reported on relates to an absolute increase of about 0.01% in absolute terms 13/

To put it another way, this means that the number of people who die rises from 2/10,000 to 3/10,000 when they quadruple their intake of these foods

A bit less scary than a "62% increase" I think 14/

Similarly, the risk of heart disease of any kind increases by less than 1% per 5 years when you go from eating very few of these foods to eating a lot of them /15

The next thing to think about is residual confounding. As I said, these were excellent, well-done studies

However 16/

The problem with these sorts of trials is that it's impossible to entirely eliminate external factors

It's possible that something other than processed foods caused the observed relationships 17/

For example, we KNOW that social factors like poverty and other forms of disadvantage impact who gets heart disease and who dies in our society

These studies might have controlled for that enough to infer causality. They also might not 18/

I'm not saying they did anything at all wrong - quite the opposite. And if you read the studies themselves, the authors are appropriately cautious about what this might all mean 19/

That being said, the fact that a recent study demonstrated a convincing link between processed food and weight gain in a controlled laboratory setting lends weight to the argument that "processing" is a decent marker for how healthy foods are 20/

The other issue in all of this is that we know that people are terrible at self-reporting what food they eat

Both studies controlled for this as well, but it's hard to know if that was entirely effective 21/

Also, if you look at the statistical models, the risk only becomes really apparent at the highest intakes of ultra-processed foods

It's hard to say if there's a big impact of eating just a bit more of these foods 22/

Ultimately, what do these results mean?

Well, for governments, they are pretty impactful 23/

If you wanted to prevent heart disease and death in your country, it looks like one realistic way of doing that is discouraging people from eating ultra-processed foods 24/

(Whether this is possible is another question entirely, and personally I'd argue that social influences will always dictate what people get to eat far more than the health of foods themselves - you have to fix society to fix these issues) 25/

But to an individual? To you and me?

I don't think these results mean all that much 26/

If your diet is currently more than a quarter ultra-processed foods like fast food, prepackaged meals, and biscuits, it might be a good idea to cut back

You probably already knew that, though 27/

If, like most of us, you have the occasional sneaky cookie, you probably don't have to worry all that much 28/

This thread got a bit long, so to sum up:

- processed foods may be bad for health
- we aren't sure if it's the food or society
- the impacts are important to governments, not as much for individuals 29/

P.S. I forgot to mention, I did write about another almost identical study last year (using the same cohort of people as the French one) here medium.com/@gidmk/process… 30/

In this piece, I go through the same issues, but my conclusion is a bit different

As any scientist will tell you, evidence moves on! 31/

I'm still not entirely convinced that ultra-processed foods are independently bad - or indeed that this is always a useful category at all - but these studies and @KevinH_PhD's recent work has changed my mind a bit 32/

I would still argue that societal changes are almost certainly the bigger issue here, but I suspect that reducing our consumption of ultra-processed foods would be a good first step 33/

P.P.S. I think this point from @Botanygeek is also very important - "processed" and "unprocessed" are heavily culturally influenced, which is one of the biggest issues with talking about the results of these studies

https://twitter.com/Botanygeek/status/1133997826399186944

Pocast is now out on the studies in this thread - have a listen!

https://twitter.com/GidMK/status/1134628919938822145