What we all need to know about ultra-processed foods – but which science can’t tell us yet
I can’t stop hearing about “ultra-processed foods”. Perhaps that’s better than being unable to stop eating them, since by all accounts they’re extremely bad for you. So-called “UPFs” are the category of foods that are said to be driving the obesity epidemic, making us and our children overweight and unhealthy.
Well, actually, not by all accounts. The exact health effects of UPFs—which apparently make up around half of Western diets—and in fact even the very definition of “UPF” itself, are questions that remain scientifically controversial.
Let’s start with the definition. What does UPF actually mean? In an article this week promoting his currently chart-topping book Ultra-Processed People, the celebrity doctor Chris van Tulleken defined it this way: “if you’re trying to work out if something is a UPF, a good rule of thumb is that it’ll be wrapped in plastic and contain an ingredient you don’t find in a domestic kitchen.”
Hmm. Not particularly useful, given I’ve just come back from the shops with some Parmesan cheese (wrapped in plastic; contains rennet that I don’t usually have in my kitchen; probably not considered ultra-processed).
What about a more scientific definition? A UN document written by the originators of the “UPF” label contains this rather circular one: “the practical way to identify if a product is ultra-processed is to check to see if its list of ingredients contains at least one item characteristic of the ultra-processed food group”.
In fairness, the researchers do provide a list of the sorts of ingredients they deem characteristic of UPFs: flavour enhancers, sweeteners, thickening agents, preservatives, bulking agents. The idea is that these ingredients have been added during various “processes” to make the food tastier (“hyper-palatable”, to use the jargon), easier to eat, and have a longer shelf-life.
Do experts agree on this definition? Not fully. In a study published last year, a group of more than 150 nutrition experts were asked to classify a list of different foods into four categories depending on their level of processing, from “unprocessed” to “ultra-processed”. Even after providing the experts with lists of ingredients, the researchers reported, they still very commonly disagreed with each other about how to classify about a quarter of the foods. The current criteria, they wrote, “do not currently allow foods to be unequivocally defined as ultra-processed”.
That’s something to bear in mind as we go on: different experts, and different nutrition researchers, might define somewhat different foods as ultra-processed. But for the sake of argument, let’s accept that there’s some consistent, if fuzzy, set of foods that can be defined as ultra-processed.
As we’ve seen, those foods are more delicious, easier to eat, and last longer before going out of date. My first thought would be: “sign me up!” But of course, there’s a catch: according to quite a lot of observational research studies (where people report how many UPFs they tend to eat, and then are followed up to check their health), eating more UPFs is linked to higher rates of all kinds of health problems: obesity, diabetes, and earlier death.
And it’s not just observational studies, which of course are less-than-optimal because they just tell you about correlations rather than causes. The most famous study of UPFs is a beautifully-done randomised controlled trial from 2019, exactly the sort of thing we need to see more of in nutritional research. In the study, 20 people signed up to stay in a research centre for a month, eating only the foods provided to them by the researchers. Of course, this is an artificial setting, but studies like this are really the only way we can have enough control of people’s diets to properly understand what causes what.
10 of the participants ate an unprocessed diet, with three large meals per day, plus snacks, provided to them made of healthy ingredients like fresh fruit and veg, fresh eggs, grains like quinoa, and grilled meats like chicken. They could eat as much or as little as they liked.
The other 10 had an ultra-processed diet, with meals and snacks containing the same number of calories and amounts of sugar, fat, and other nutrients as the unprocessed group, but consisting of foods from tins or packets, as well as burgers, chips, crisps, and sausages. Again, they didn’t need to eat the full amount presented to them, but could if they wanted.
After 2 weeks, the groups switched over to the other type of diet, so that by the end of the month everyone had had a fortnight of processed and a fortnight of unprocessed food.
The results? On average, the participants gained nearly a kilogram during the ultra-processed fortnight, and lost around the same amount when eating unprocessed foods. They’d eaten substantially more calories when on the ultra-processed diet, even though they could in theory have consumed the exact same number in both groups.
Now, you might think—as has been argued by the health economist Christopher Snowdon—that it’s quite simple: isn’t it just that the participants preferred the taste of the ultra-processed foods, so ate more of them? We already know that people will eat more of the foods they find delicious, so what have we actually learned here? “We don’t need”, says Snowdon, “to demonise a vast range of foods to make this point, nor do we need to pretend that there is some magical process taking place in factories that makes ultra-processed food fattening per se”.
The researchers, though, point to a questionnaire filled in by the participants during the study: they found that there was no statistically significant difference in the participants’ rating of the pleasantness of the foods whether in the processed or unprocessed groups. That is, they liked both types of food equally.
Of course, people often do “preference falsification” – reporting that they love eating healthy foods because that’s what they feel socially pressured to say (everyone knows that a healthy meal is the “right”, “socially desirable” option). Then again, if that was true, you might expect to see people rating the unhealthy UPF meals as less pleasant than they really were: “oh no, I don’t actually like eating all those burgers and cookies – honest!”.
It’s hard to guess which way it would go, which illustrates how difficult it is to use self-report measures and take people at their word. Remember, though, that part of the very definition of UPFs is that they’re “hyper-palatable” compared to unprocessed food, which would predict that the participants should have said that the ultra-processed foods were tastier. The fact that they didn’t is, in fact, a bit of an anomaly.
Perhaps it’s not specifically to do with taste and more to do with ease of eating. To my knowledge the only other randomised controlled trial of UPFs found that if a food was soft-textured and a UPF (like instant potato mash), as opposed to hard-textured and unprocessed (like boiled rice), participants tended to eat more of it, and thus consume more calories (though in that study the participants did report finding the UPFs significantly tastier, so Snowdon’s straightforward explanation could apply).
It seems like everyone agrees that having a diet that includes more UPFs (and remember, there are also some disagreements around the edges about what exactly counts as a UPF) will likely make you put on more weight. But there’s no consensus about how it happens: is there something special about the processing or the ingredients – something addictive, or something that interferes with our body’s hormonal systems – that makes these foods particularly dangerous? Or is it a much more banal story: that UPFs are just really tasty, so people tend to eat them in larger quantities?
Clearly we need more evidence – specifically, we need more of the randomised trials where the participants have their entire diets controlled. It would be particularly useful if there was a study comparing not just a generic unprocessed meal to a generic ultra-processed one, but comparing the exact same foods in processed versus unprocessed form (say, homemade hamburgers versus ready-made, preservative-filled ones from a plastic packet). That’s a much trickier study to do, but it would be a better test of the question we all want answered: what is it about UPFs that makes people eat more of them?
It also seems likely that, since the definition of UPFs is so expansive, covering so many different types of food, so many “processes”, and so many ingredients and additives, that it’s only some of the component parts of UPFs that really affect our health. Only with further, more detailed research will we be able to work out which are the specific ones we should be worried about.
Even though it was perhaps a little premature to write a book on a topic where the evidence is still so unclear and ambiguous, the success of Chris van Tulleken’s Ultra-Processed People shows us just how much interest there is in the health effects of UPFs. But we shouldn’t let the discussion of UPFs get too far beyond the rather limited data we have. Let’s hope scientists—and, more importantly, research funders, who need to pay for many more high-quality randomised trials—are listening.