Ultra Processed Food Is Not As Addictive As Cocaine
Ultra Processed Food Is Not As Addictive As Cocaine
Why recent headlines about ultra-processed foods are wrong
Ultra-processed food is the latest in a long line of ways humans have thought up of trying to categorize things into easy groups of “healthy” vs “unhealthy'“. The idea is that foods which are produced in factories and come pre-packaged are generally unnatural and bad for your health, while things that haven’t been over processed are mostly fine to eat. There are many easy comparisons that make this intuitively seem right - a bag of corn chips vs a carrot, for example.
If you’ve been reading the news recently, this distinction has seemed more important than ever. Scientists have apparently shown that not only are ultra-processed foods bad for you, they are also as addictive as controlled drugs like heroine and cocaine. Specifically, the claim is that more than 1 in 10 (12%-14%) of people are currently addicted to these foods. For those of us who enjoy the occasional bag of frozen fries, this sounds like terrifying news.
Fortunately, the actual science isn’t nearly as scary as you might’ve heard. While ultra-processed foods are potentially not great to eat in large amounts, they certainly aren’t comparable to snorting a large pile of powder like you’re the lead in Scar Face.
The Science
The first thing to note about the most recent headlines is that they genuinely aren’t based on any specific evidence. The story here contains essentially no data showing that ultra-processed food is addictive in any real sense, and certainly nothing proving that 14% of people are addicted to these foods.
The main paper that has caused all of this ruckus is an analysis piece published in the British Medical Journal in October 2023. For those who don’t work in scientific publishing, an analysis piece isn’t actually analysis - it’s essentially an op-ed but more sciency.
Now, that’s not to say that the opinions of senior professors aren’t interesting or important. Analysis papers can be very useful to understand an issue. This one was written by very smart people who know quite a lot about the area.
But they aren’t evidence in a meaningful sense. They aggregate the opinions and beliefs of the authors, and authors can be biased, wrong, or just convey things in an unfortunate way. That means that you can’t cite analyses of this kind directly as evidence for a thing to be true.
This particular case is a perfect example of this really important issue. The key figures - the things that every single media headline quoted - were from the very first sentence of the analysis paper. This sentence reads:
“A recent analysis of two systematic reviews including 281 studies from 36 different countries found the overall pooled prevalence of food addiction using YFAS was 14% in adults and 12% in children”
This sentence then references these two papers in adults and children. But there’s a huge problem here that you can identify even from the sentence above - these reviews don’t look at ultra-processed foods! In fact, the Yale Food Addiction Scale (YFAS) doesn’t specifically mention processing at all. Some of the examples given in the YFAS as potentially addictive foods are minimally processed or unprocessed, such as steak and rice. Every single article that used these high figures for ultra-processed food addiction was by definition wrong.
The story here actually gets even worse when you look at those two studies. They’re both systematic reviews which also include meta-analyses, aggregating together studies looking at how many people report food addiction according to the YFAS score mentioned above.
But when you look at what makes up these estimates, it’s clear that they can’t be used for some kind of population-wide proclamation. The study looking at food addiction in children, for example, included 22 samples. Of these, 7 were specifically of overweight or obese children. While I haven’t checked every study, most of the others appear to be highly-selected convenience samples like this paper looking at secondary schools in Egypt, another paper surveying addictive behaviors in 14-21 year olds in the Netherlands, or this online sample of parents and their children from 2015 where parents were paid $0.5 for their participation. Regardless of your perspective on the issue, these samples just don’t represent the entire population, and so may give us completely different numbers than you’d see if the studies were more robust.
The adult sample is even more questionable. This paper looked at studies of food addiction, and categorized them into either “clinical” or “non-clinical” samples. The cited 14% figure appears to have come from the aggregate result of the “non-clinical” studies that were pooled together (oddly enough, the study actually notes that the estimate for adults is 24%, not 14%, but since this includes mostly studies of obese people and those undergoing bariatric surgery it’s easy to see why this wasn’t chosen as a population estimate in the BMJ analysis paper).
So what are these “non-clinical” studies that are aggregated together to form the 14% estimate? We’ve got this study of adult inpatients undergoing weight loss treatment. This paper looking at obese children. This study of Hispanic people in the USA who were looking to get bariatric surgery. This study of people with obesity. And on, and on. I’m not sure how all of these papers were classified as “non-clinical” by the authors, but it’s clear that the definition was not a useful one.
Moreover, a huge weakness of both of these studies is that they aggregated terrible samples together with good ones. They both used a standard inverse-variance meta-analysis, and with the high heterogeneity they found between the studies what that essentially means is that all of the numbers that they pooled together had the same weighting. In other words, while they did do a meta-analysis, it was basically identical to calculating the simple unweighted mean average that you learn in high school.
That means that huge, national samples such as this estimate of 134,000 older women in the US have the same weighting in the models as this tiny sample of 14 men with obesity-related hypogonadism. When you produce an aggregate estimate and a large proportion of that estimate is on useless samples, the overall number is itself a bit meaningless too.
In fact, I could not find a single sample in either of these massive reviews that took a representative, large sample of an entire population and estimated the proportion of people who experienced food addiction as defined by the YFAS. The closest I came was the above study in older women, which found that 5.8% of the people included had scores on the YFAS consistent with addiction, but even that’s a sample of older women and certainly not representative of the US population as a whole.
The point here is that these systematic reviews, while quite interesting, have produced figures that are at best pretty vague. It’s quite likely that the true rate of food addiction in the population is drastically lower than the 12% in children and 14% in adults from the quote above, because if you look at the larger studies with better samples the rate of food addiction drops substantially.
If you’re following along, this is a great example of the practice of citation laundering that I’ve talked about before - no one in their right mind would say that men with obesity-induced hypogonadism are representative of a population, but if you chuck that estimate in with a bunch of other ones then it’ll be cited in the media as trustworthy because no one can be bothered to check. I haven’t even checked these studies perfectly in this blog, because after you’ve gone through 20 references from a paper like this it’s fair to assume that there are issues with the next 100 or so as well.
Food “Addiction”
So, we know that the headlines were all wrong. The studies didn’t specifically look at ultra-processed food, and even within the context of food addiction you can’t use these estimates for the general population anyway.
But that brings us to the important debate about what the term food addiction actually means. The idea is very well-meaning - people have developed things like the YFAS to identify when individuals feel like their eating is out of control and causing them harm - but when it comes to food it’s obviously more complex than addictions like cocaine or nicotine. All food “addiction” is necessarily on a spectrum, because if you were to go cold turkey and quit entirely you’d starve to death.
Moreover, unlike most substances we classify as addictive, food is part of everyday life. While it’s theoretically possible that, say, fast food is “addictive” to some extent, it’s hard to differentiate that addictive element from the constant advertising, promotion, and integration into society that these products have. Imagine if you could easily order a line of cocaine for less than it costs to catch a cab, and it would arrive fresh and delicious at your doorstep within minutes. If cocaine was advertised at every sports game. The idea of “addiction” takes on a very different tone when you think about how much society has to do with which foods we eat and when.
If nothing else, ultra-processed food is so much easier than cooking your own. If you’ve got 2 kids and work long hours, it’s really hard to cook a healthy meal every day, and chicken nuggets and chips in the oven is ready in minutes and requires almost no cleaning up. It’s very easy for relatively wealthy, educated researchers to brand certain eating patterns as unhealthy, but it’s much harder to dramatically change the food environment such that healthy food is actually easy to access for everyone.
This debate was part of what the original paper in the BMJ that caused all the headlines was trying to address. Without getting too far into the specifics, the main issue here is that whether or not you can be “addicted” to food is itself the subject of much debate, and whether this is specifically caused by processed foods is even more questionable.
There’s also the arguments around harm. Addiction, like most psychological health constructs, is defined by negative impacts on your life. But it’s entirely possible to experience most of the things identified in the YFAS and be perfectly healthy, at least insofar as metabolic and physical wellbeing.
On the other hand, the question of food “addiction” is only made more complex when you look at the systematic review in adults I’ve discussed above. People with diagnosed anorexia, for example, were nearly twice as likely to experience food “addiction” as defined by the YFAS than obese people, which also complicates the construct. Ultra-processed food “addiction” may be the only type of addiction where people who have quit are more likely to receive a diagnosis than people who indulge regularly.
Ultimately, the data is pretty complex either way, and I’m not going to be able to settle the debate around the existence of food addiction as a construct in a relatively short article like this. What we can say is that pretty much all of the headlines about “ultra-processed” foods were completely wrong. There may be some portion of the population who are addicted to food, but whether that food is explicitly ultra-processed is more complex. In addition, the number of people impacted is likely to be far lower than the headlines suggested.