News · Nutrition & Diet
Three additive types tracked about 45% more depression; five others tracked nothing
Ultra-processed food is usually condemned or defended as one block. Following 171,818 UK adults for a decade, only three of ten additive categories tracked depression, and they did so in proportion to how much was eaten.
- Only three additive categories tracked depression: flavorings, colorings and sweeteners.
- Each of the three rose with the dose, which is what a real signal looks like.
- Five other categories showed nothing, including sugars and modified starches.
- That cuts against treating all ultra-processed food as one thing.
- 171,818 UK adults over a decade, and diet was recorded by questionnaire.
The case against ultra-processed food is usually made wholesale. It is a category defined by how food is manufactured rather than by what is in it, which means a supermarket loaf and a fluorescent energy drink sit in the same box and get the same warning.
Writing in the European Journal of Nutrition, researchers took the box apart. Rather than counting how much ultra-processed food people ate, they counted specific additive markers: ten categories and 39 specific MUPs, assessed as a percentage of total food intake, across a cohort that included 171,818 participants followed for a decade.
Three categories tracked depression. Five did not.
What was actually measured
The distinction matters, so it is worth being precise about it.
A conventional study of this kind asks what share of someone’s diet is ultra-processed and compares the top of that range with the bottom. It cannot say which part of the processing is doing anything, because processing is not one thing.
This analysis instead identified the markers themselves: the flavorings, the colorings, the sweeteners, the emulsifiers, the modified starches, and grouped them. Each was measured as a percentage of total food intake, which allows the question to be asked as a dose rather than a category.
Over an average follow-up duration of 10.7 years, 5424 cases of incident depression occurred among the participants.
The three that moved
Flavor, coloring agent and sweetener behaved the same way as each other and differently from everything else.
All three were positively and dose-dependently associated with depression: the more of them in the diet, the higher the risk, in a steady progression rather than a jump at some threshold. At the high end against the low, each carried roughly 45 to 50% more depression than the lowest-risk point.
Dose-dependence is the part worth noticing. A single comparison of high against low can be produced by almost any confounding. A relationship that rises step by step across the whole range is harder to manufacture accidentally, which is why it appears on every checklist for judging observational evidence.
The five that did not
Flavor enhancer, processing aid, modified starch, varieties of sugar, and modified oil showed no significant association at all.
Sugar is the striking absence there. It is the additive most people would name first if asked what in processed food might affect mood, and across 171,818 people over ten years it did not separate from the background.
Two further categories, protein source and fiber, produced a U-shape, with risk raised at both the low and high ends. That is a different pattern again and probably a different explanation: very low intake of either marks a poor diet overall.
Why the split is the finding
If ultra-processed food were harmful because it is ultra-processed, the markers of it should all point the same way. They do not.
Three categories rise with the dose, five do nothing, two bend at both ends. That is not the pattern of a single underlying harm being detected through ten different proxies. It is the pattern of specific substances doing specific things, or of specific substances marking specific eating habits.
Which of those two it is, this study cannot say. The distinction it establishes is still useful: the interesting question is which additives, not how processed.
The objection that has to be faced
Depression changes how people eat, and it does so long before anyone gets a diagnosis.
Low mood pushes people toward food that is convenient, palatable and heavily flavored, which is a fair description of the three categories that came out positive. Someone in the early years of a depressive illness may be eating more of exactly these things because of the illness.
Diet here was recorded before the diagnosis, which is the right way round and not a solution. The prodrome of depression can run for years, and a single dietary questionnaire cannot be placed reliably outside it.
What the measurement can and cannot bear
The dietary data come from the Oxford WebQ, an online recall of the previous day’s eating, completed at least once. That is a thin slice on which to build a decade of exposure, and it assumes the day recorded resembles the rest.
Misreporting in that instrument is not random either. People under-report the things they feel judged for, which for this study means under-reporting precisely the sweetened and heavily flavored items being examined.
Where it leaves the advice
Not with a list of additives to avoid. The evidence is nowhere near that, and three associations in a cohort study do not identify a culprit.
Where it does land is on the framing. Public advice has settled on ultra-processed food as a single enemy, partly because that is a simpler message than a list of E numbers. This analysis suggests the category is doing real work as a heuristic and hiding real differences underneath.
For anyone trying to eat better, that is not a license and it is a small correction: the loaf and the energy drink are not equivalent, and the evidence has started to be able to tell them apart.
People also ask
What did the study find?
The categories flavor, coloring agent and sweetener were positively and dose-dependently associated with depression risk: flavor (HR 1.46, 95% CI 1.29-1.64), coloring agent (1.49, 1.30-1.70) and sweetener (1.45, 1.32-1.58) at high versus low intake. Two categories showed a U-shaped association and five showed none.
Which categories showed nothing?
Flavor enhancer, processing aid, modified starch, varieties of sugar, and modified oil. Two more, protein source and fiber, produced a U-shape, meaning risk was higher at both extremes rather than rising steadily.
Does this mean sweeteners cause depression?
No. This is an observational cohort, and people eating a lot of sweetened products differ from those who do not in income, weight, activity and much else. What the study adds is that the association is specific to certain additive groups rather than general to processed food.
Why does dose-dependence matter?
Because a relationship that strengthens steadily as intake rises is harder to produce by accident than a single threshold comparison. It is one of the standard reasons for taking an observational association more seriously.
Could depression be changing what people eat?
Yes, and that is the central objection. Low mood shifts people toward convenient, palatable, heavily flavored food, and depression develops over years before it is diagnosed. Diet here was measured before diagnosis, which helps, but does not settle it.
How was diet measured?
By the Oxford WebQ, an online questionnaire recording what a person ate over the previous day, completed at least once. That is a thin slice of a decade of eating and it depends on accurate recall.
What is the practical reading?
That the blanket advice to avoid ultra-processed food may be aiming at the wrong target, and that the interesting question is which additives, not how processed. Nothing here is a reason to change a diet on its own. This is general information rather than medical advice.