How dangerous is ultra-processed food really?

Several media outlets are today reporting on a new, large "umbrella study" of new meta-analyses on the health effects of ultra-processed food. The study reports frightening negative effects in several areas, and therefore warns against ultra-processed food. But is this really true, and what does this study actually tell us?

Ultra-processed food is the new hype in the eternal debate about diet. While professionals and scientists are often sceptical about using this term in such a context, food influencers and the media benefit from simple terms that can create fear and sell simple solutions.

Unfiltered Perception is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

Store Norske Leksikon describes ultra-processed food (UPF) as follows:

Ultra-processed food is a term used to describe food where the raw materials used as ingredients have been significantly altered, usually by industrial processing. A characteristic of ultra-processed food is that it is difficult to recognise the raw materials from which the food is made. There is no clear definition or clear delimitation of what constitutes ultra-processed food.

The last sentence, which I have italicised, is worth noting. UPF is a rather woolly term, and I'll come back to the consequences of that soon.

A new and frightening mega-study

The new study is presented by NRK as follows:

A diet consisting of a lot of ultra-processed food can be harmful to many of the body's systems.

This is the opinion of researchers who have now conducted 45 meta-analyses of 14 major articles. All have been produced within the last three years.

The new umbrella study includes almost 10 million people and is said to be the largest and most thorough ever conducted.

The findings show that increasing the intake of ultra-processed food also increases the risk of dying from cardiovascular disease by as much as 50 per cent.

Overall, the researchers found that consumption of such food was associated with an increased risk of 32 undesirable and serious health conditions.

First and foremost, it is important to remember that these are associations, and we cannot know with certainty whether there are causal effects here.

In other words: Are there more cases of type 2 diabetes in people who eat a lot of ultra-processed food, or is it the case that people with low income, who are not physically active, are genetically predisposed to obesity, are not very concerned about diet and health in the first place and thus already have an elevated risk of lifestyle diseases choose to eat a lot of "junk food" because it is more readily available, faster to obtain, less expensive, and tastes good?

Perhaps, although we should not ignore the associations made here based on a very large data set.

Uncertain results

But perhaps a more important point of criticism is the low degree of confidence one should have in these conclusions. Take a look at these tables from the research article:

In the far right column, we see the GRADE score each endpoint in the study has. GRADE is a standardised way of assessing the credibility or quality of research included in a meta-analysis. It has four levels: Very Low, Low, Moderate and High.

As we can see from the table, 19 endpoints are graded as Very Low, 23 as Low, and only 4 as Moderate. None have a GRADE score of High.

So what does this mean? Here is a description of the different GRADE levels:

Very Low means that there is very little confidence in the effect size. It is likely that the actual health effect is significantly different from what the data shows.

For Low, confidence in the effect size is limited. The actual health effect may be significantly different from what has been found in the study.

And Moderate means that there is moderate confidence in the effect size. The actual health effect is probably not too far from the result obtained in the meta-analysis, but it is possible that there is still a large discrepancy here as well.

It is therefore difficult to put much faith in most of these negative health effects. Especially when several of the results are not statistically significant.

The only ones of at least moderate quality are that UPF may increase the risk of total mortality, prostate cancer (although not statistically significant), overweight/obesity and type 2 diabetes.

In meta-analyses, an important concept applies: SISO, also known as Shit in, Shit out. If you do a meta-analysis on bad studies, you will get a bad or unreliable result from the meta-analysis as well. The process of performing a meta-analysis does not magically turn bad studies into solid data1.

Unfortunately, the researchers behind the study, as well as the media and others who hype this study, communicate this point far too poorly.

A meaningless term

In many ways, this is bad enough, so why am I so negative about the study? Well, as I mentioned in the introduction, the term "ultra-processed food" is quite vague.

UPF is only about the degree of processing, i.e. the extent to which the final product is industrially processed. But is it this degree of processing in itself that has negative health effects, or is it the components of the food that matter? Shouldn't we also look at the nutritional content of the food? Yes, of course you should, otherwise the term will confuse more than it enlightens.

As I mentioned in the introduction, nutritionists are often unfavourable to the use of UPF in discussions about food and health. In 2022, for example, British nutritionists came together to discuss how the term "ultra-processed food" should be used in communication about food and health to the population. Their conclusion was, among other things:

The concept of UPF should not feature in dietary advice

They go on to write that foods that are classified as UPF and are less healthy are often quite obvious based on the ingredients they contain, e.g. high sugar, salt, fat etc. They also point out that it can be stigmatising to demonise UPF as many people rely on such affordable and simple food to feed themselves and their families. Not everyone has the luxury of being able to buy fresh produce and spend two hours making dinner from scratch for the family every day.

Ultra-processed food can simply be an easy and affordable way for many people to eat relatively healthy food, as long as they choose the right products.

In addition, they point out that there are many other important factors to consider when choosing food, such as price, nutritional content, sustainability, animal welfare and the ability to recycle packaging. If the authorities are to provide good dietary advice, these factors should often weigh more heavily than the degree of processing of the food.

In Sweden, we recently saw the same thing when Swedish nutritionists reviewed the research on processed food:

- "Science, public health and the public debate would benefit if we stopped using inadequate terms such as 'ultra-processed' and instead described which foods are meant," says Cecilia Nälsén, researcher at Örebro University.

[...]

- "The term 'ultra-processed' only appears in the classification systems and is not an established term in food science research. "Our report shows that there are too many ambiguities with the term and that it risks confusing and creating concern instead of being useful for consumers or researchers," says Cecilia Nälsén.

The interesting thing is not whether the food is “ultra-processed” or not, but what the food actually consists of. And perhaps the most interesting thing in this context is a large study published last year in The Lancet. The study followed 266,666 otherwise healthy people from various European countries for more than ten years.

The results showed that higher intake of UPF slightly increased the risk of cancer and cardiometabolic diseases. However, when they looked at food subgroups, they found that these negative effects were primarily linked to the consumption of animal products, as well as artificially sweetened2 and sugary soft drinks.

Interestingly, they also found no negative health effects associated with e.g. cereals, bread, ready meals, desserts or plant-based meat alternatives, even though these are all classified as ultra-processed foods. And that some foods that are considered unhealthy were not actually linked to negative health effects3 at all:

However, several other major types of UPF previously seen as harmful: sweets and desserts, ready meals, savoury snacks and plant-based alternatives to meat products also got the all-clear. They are "not associated with risk of multimorbidity", said the authors.

Babies and bathwater

This is one of the main problems with the idea of ultra-processed food. If you look at the health effects of UPF overall, as the new umbrella study has done, you find (weak/unreliable) negative health effects for a lot of things. However, as UPF includes so many different foods, it would be throwing whole cohorts of babies out with the bathwater to conclude that foods classified as UPF are by definition unhealthy.

To illustrate this, we can look at this study that looked at the risk of type 2 diabetes in relation to UPF intake and found that higher UPF intake increased the risk of diabetes, as did the new umbrella study.

However, when looking at subgroups of ultra-processed foods in this study, the picture becomes more nuanced:

Here we see that ultra-processed foods such as cereals, wholemeal bread, fruit-based products, some sweets, as well as yoghurt and dairy-based desserts actually reduced the risk of type 2 diabetes. However, because animal products, ready-made dinners and some other foods increase the total risk so much, the overall effect is also negative.

In other words, it may seem that UPF as a whole is problematic, while in reality it is only some subgroups of UPF that increase the risk of various diseases.

Therefore, if you simply look at the conclusion and stop eating wholemeal bread and yoghurt because it is UPF and you think it increases your risk of type 2 diabetes, you have done yourself a disservice.

The problem is not ultra-processed food per se, but specifically what food you eat and its nutritional content. Not least, it's about the balance in your diet.

What's healthy and unhealthy here we've pretty much known for a long time already. To suddenly introduce the term "ultra-processed" adds nothing new in this context.

Conclusion

So what have we learnt from this new, large umbrella study of new meta-analyses? Not very much.

The results are generally very unreliable, with a few exceptions. It probably just confirms what we already know about diet: Consuming a lot of processed red meat, saturated fat, sugary drinks and too much salt is unhealthy. But mainstream dietary advice already warns against this.

So rather than choosing food based on whether it's "ultra-processed" or not, you should follow the standard dietary advice that has been with us without major changes for many decades already.

In addition, you should consider more than just ingredients or degree of processing when choosing food. Environmental aspects and animal welfare are just as important. And for most of us, price, time and availability will also play a role.

Ultra-processed food is not inherently dangerous. It's always about quantity and balance in the diet. What can be dangerous is having too one-sided a diet, or too much of the unhealthy foods consistently over a long period of time.

So by all means: Don't stop eating high-fibre wholemeal bread from the shop just because it's technically ultra-processed food. Don't steer clear of fruit-based desserts or yoghurt because they're ultra-processed. Instead, look at the amount of added sugar, how much saturated fat and salt it contains, and the total calorie content, because whether the food is ultra-processed or not, it's ultimately these factors that matter to your health.

But we already knew that.