Ultra-Processed Foods Are Killing Us All
Again. According to the headlines. Sigh.
When I was a child, we were always warned about the dangers of junk food. While there was no specific definition of what “junk” meant, we all knew the general idea - high fat, salt, and sugar foods that you were only meant to eat at birthday parties or steal very occasionally from dad’s stash. That the stash moved regularly was a sign of how inventive my brother and I became at accessing hidden places that we could not easily reach or open.
But these days, there’s a new, scientific term: ultra processed foods. Rather than the generic idea of “junk”, we’ve got a reasonably specific classification system which determines how “processed” a food is. This has then been linked to any number of health outcomes, from heart disease and cancer to, more recently, all-cause mortality.
Yes, according to recent research my childhood forays into the chocolate cupboard were not just cheeky, they were shortening my lifespan by a measurable amount.
Except, the issue is far more complex than that. We’ve known for decades that eating too many high-fat/salt/sugar foods is bad for you, but the reality is that this evidence doesn’t add much to our understanding of what to eat. Let’s look at the data.
New Study
The new study is a meta-analysis of previous papers looking at ultra processed foods and mortality. The authors took seven different datasets of food intake questionnaires, plugged them all into a statistical model, and used this to estimate how much worse off people were due to ultra processed foods.
They found that for every additional 10% of a person’s diet that’s made up of these foods, there is a 2.7% increase in the relative risk of mortality. So compared to someone who eats no Froot Loops, a person who gets 40% of their calories from the cereal is on average 10% more likely to die over the next 30 years or so.
The authors also calculated what’s known as the population-attributable fraction of ultra processed foods on premature mortality. This is an estimate of what proportion of the early deaths are related to ultra processed foods. While the number was very low in some places - only 4% in Columbia - in high-income countries with very high consumption of these foods the numbers were as high as 14%. This implies that if no on in the country ate any of these foods, the premature death rate could drop by as much as 14%.
If you’ve been reading my work for a while, you’ve probably already identified that these numbers are not nearly as solid as you might imagine. The estimates of ultra processed food intake are based on food frequency questionnaires, which are notoriously unreliable. The correlation with premature mortality is based on various statistical models with numerous assumptions, some more reasonable than others. There are many uncorrected confounders that make the associations less likely to be causal than they could be. Overall, the link between ultra processed foods and mortality is, at best, a bit weak, and may not represent a real finding as much as it shows that people who eat a lot of cheap, low-quality foods are less healthy in many ways.
All that being said, I think it’s informative to ask: what if these results are real? What then?
The data shows that, at an individual level, you can reduce your risk of all-cause mortality by 2.7% by keeping your calories the same but switching 10% of your intake from ultra processed foods to minimally or simply processed foods. What would that mean, in context?
Well, 10% is probably achievable for most people. If you currently have a bowl of ultra processed cereal for breakfast during the week, you could switch that to home-made overnight oats. If you’re currently snacking on potato chips on a daily basis, you could instead cut up some carrots. Of course, keeping the calories the same will be a bit weird - one packet of chips is something like 3 whole carrots - but that’s kind of the point as well.
Above this it starts to get really challenging. If you are eating 30%+ of your calories as ultra processed foods, there’s probably a reason for it. The main way to get that much of these foods is if you’re very short on time and eat a lot of instant noodles, take-out, and frozen foods.
At this point, we’re asking people to almost completely change their diets. Not just the odd meal or two, but pretty much every meal they eat. Get rid of the sauces, start cooking from scratch, eliminate the take-out containers. It’s not impossible, but it’s not a simple ask either.
On top of this, it’s unlikely that we could realistically get people to eat no ultra processed food at all. People in the lowest quarter of ultra processed food intake still get about 16% of their calories from these foods. If we were to largely revolutionize our food system and make ultra processed foods much harder to access - not a bad idea in and of itself - we’d still be eating a reasonable quantity of them because they are just so convenient.
My favourite example of this is bread. If you buy packaged sliced bread from a store, it is almost always ultra processed. It lasts for 1-2 weeks on the counter, or up to a month in the fridge. If you instead buy fresh bread from a bakery, it is usually just processed - no “ultra”. Fresh bread usually lasts 3-5 days and goes stale quickly in the fridge. It is much tastier, but if you only have time to go grocery shopping once a week, fresh bread is going to be a sometimes rather than everyday food by default.
Bottom Line
It’s probably true that people who eat more ultra processed foods die a bit earlier than those who don’t. Personally, I doubt that the 14% number is accurate - population attributable fractions have well-known limitations - but society as a whole would probably be a bit healthier if the average consumption of, say, chicken nuggets went down a bit.
That being said, the issue is very complex. Substituting one food for another, particularly at a societal level, is very difficult. Even if we ignore the questions about how much of this association is causal, and whether ultra processed is a useful descriptor, the issue here is that everyone already knows the problem. If you got rid of everything that my mother told me as a 5 year old was “junk” food from your diet, you’d be close to 0% ultra processing in your life.
The reality is that whether we call it junk, or processed, or develop an entirely new name for it in the future, unhealthy food is more complicated than a simple recommendation. Yes, if you’re eating only frozen meals you might want to consider changing your diet, but you probably already knew that. When I was 18 and playing video games I didn’t eat a diet of mostly Korean snacks because I thought they were healthy, I did it because I was 18 and they went well with beer and Mountain Dew.
So yes, you should probably eat less unhealthy food. You can define that in many different ways, but generally if you eliminate the stuff with high levels of fat, sugar, and salt you’ll be on the right track. But the problem of ultra processed food is far more complex than any advice I can give you in, and involves all of society rather than simple dietary prescriptions for people like you and me.
But you know, it's not just this one study.
I was a UPF skeptic before I started digging into the literature. Above and beyond KD Hall's excellent and well controlled studies that totally control diet and switch between UPF and "real" food diets in both directions, there's a huge array of other studies with large effect sizes, that generally show a 1.3x - 1.7x all cause mortality hazard ratio. That's roughly a "smoking" size hazard, extremely significant.
Even after controlling for fiber and macros, the top quartile of UPF eaters by percent of diet have a 1.3 - 1.7x all cause mortality Hazard Ratio compared to the bottom quartile, and the amount of UPF eaten follows a direct dose-reponse curve in terms of the incremental all cause mortality.
“But wait,” you might say, “you can’t just control by macros, UPF is eaten by fat poor people, so they have many other negative health confounds!” And you’d be right!
Happily, many researchers thought of this. In fact, a 1.7x all cause mortality HR study (the SUN study in Spain) controlled for education level, SES, obesity, marital status, smoking status and historical smoking, age cohort, and a double handful of specific health conditions. Then because there could still be residual confounding, they did sensitivity analysis and calculated Vanderweele E values to get the minimum strength of association before calculating the final impacts.
A Swedish study with n=27k6 measured actual UPF metabolites in blood and found a 1.23x HR in all cause mortality for each additional sigma of UPF consumption.
They also controlled for BMI, marital status, smoking history and status, educational level, alcohol use, exercise, coffee drinking, familial health history, and more in this analysis. Their finding is in line with several other studies: “Our observed positive association with all-cause mortality was supported by a recent meta-analysis [40] and four subsequent cohort studies [17-19,21]”
I've left out another snippet I could have put here that literally runs down 10 additional studies, all with N in the tens to hundreds of thousands, showing similarly large effect sizes. I actually wrote a whole post about this here: https://performativebafflement.substack.com/p/processed-food-followup-ultra-processed?r=17hw9h
I'd be willing to bet most of your readers don't smoke. I think the evidence is probably stronger than you portray it to be in this post, and if the relatively conscientious and health-aware folk in your audience knew the true effect sizes, they might be willing to make some of those major changes in diet, even if difficult.
Sure, at the societal level it's hard to move the needle - but part of the reason to write is to positively influence your readers when you can.
Every food can be part of a healthy diet, or an unhealthy diet. People need to relax and enjoy their food. Even three Big Macs a day — and nothing else — is a pretty decent diet: https://skeptoid.com/episodes/4629
"Historically, the Inuit did perfectly fine on a diet consisting mainly of saturated fat, the Maasai on cow's milk and blood loaded with cholesterol, Paleolithic Europeans on a staple of starchy grains and tubers with a little of everything else sprinkled in."