Food additives to avoid: the two with real trial evidence

Only two threads in the food-additives-to-avoid conversation have trial-grade human evidence behind them. The first is synthetic dye mixtures and short-term hyperactivity in some children. The second is the gut effects of the emulsifier carboxymethylcellulose (CMC). Almost everything else on the popular avoid-lists, including most of what gets said about nitrites, nitrates and sweeteners, rests on mouse studies or on cohort data that can show a pattern without proving the additive caused it.

That distinction matters more than the lists themselves.

What counts as a food additive to avoid

Consumer avoid-lists group additives by public concern rather than by a shared regulatory finding of harm. An ingredient can carry FDA GRAS status or an EFSA acceptable daily intake and still land on several of these lists.

That is not a contradiction. It just means “on a list” and “shown to be harmful in a trial” are two different claims, and the lists rarely tell you which one they are making.

• Regulatory approval means an agency reviewed the toxicology and set a safe intake level. It does not mean no one has ever raised a concern.
• Appearing on an avoid list means enough consumers or writers flagged it. It does not mean a trial tested it in humans.

If you want to check where a specific ingredient on your shelf falls, a scanner app such as EatMatrix reads the label directly rather than relying on someone else’s list.

Why food additives to avoid became a search term

Two separate evidence threads drive most of the public concern, and they are not equally strong.

The stronger one is behavioral: a body of randomized human trials on synthetic dye mixtures and children’s behavior, discussed below. The weaker one, in terms of what it can prove, is a set of large observational analyses.

A NutriNet-Santé analysis of about 92,000 French adults found that higher intake of food additive emulsifiers, one category tracked across the wider cohort of more than 100,000 participants, was associated with increased cancer risk in this observational data^[1]. That is cohort data: it can flag a pattern worth investigating, but a cohort cannot rule out that something else about people who eat more emulsified processed food explains the difference.

Clean-label marketing has amplified interest in additive avoid-lists well beyond what the underlying regulatory and trial record shows.

What the human and animal literature contains

The best-documented human evidence in this whole category comes from one line of research: artificial colors and behavior.

A 2007 randomized, double-blind, placebo-controlled crossover trial tested mixes of artificial colors and the preservative sodium benzoate in 153 three-year-old and 144 eight- and nine-year-old children^[2]. One of the two tested mixtures produced a statistically significant increase in a global hyperactivity score compared with placebo in the three-year-olds (effect size 0.20, p=0.044), and both mixtures did so in the eight- and nine-year-olds once the analysis was restricted to children who drank at least 85 percent of the challenge drinks^[2].

That is a real, replicated signal across two age groups and two independent mixtures.

The emulsifier evidence runs the opposite direction: strong in mice, thin in humans. Carboxymethylcellulose, a common thickener, had mostly been studied for gut effects in mice and in cell models. Human data were still described as “poorly characterized” when the first small controlled feeding trials in people were published^[3].

That gap between animal and human evidence is common in this category, and it is exactly where avoid-lists tend to overreach.

How researchers dosed additives in these trials

Consumer advice rarely says what dose a trial used. Here is what these two lines of research used.

27 clinical trials reviewed on synthetic food dyes and children's behavior PMID 35484553

• The dye trials: an earlier Southampton-group trial gave preschool children a daily challenge drink containing 20 milligrams of artificial colorings and 45 milligrams of sodium benzoate, amounts the researchers modeled on typical intake from sweets and soft drinks^[4].
• The CMC trial: healthy adults ate an identical diet enriched with 15 grams a day of carboxymethylcellulose, or an emulsifier-free version of the same diet, for eleven days^[3].

15g CMC given daily to healthy adults in the controlled feeding trial PMID 34774538

The mouse studies behind the emulsifier concern used a different design entirely. The foundational 2015 study found that continuous dietary emulsifier exposure altered gut microbiota and promoted colitis and metabolic syndrome in mice, using ongoing exposure rather than the on-and-off pattern of typical human eating^[5].

The dose problem: a mouse getting an additive in every meal for its entire short life is not the same exposure as a person eating packaged food a few times a week. Neither figure is wrong. They just answer different questions.

The same mismatch shows up outside food additives entirely. Rodent studies that push megadoses of vitamin A to find liver damage let regulators set a cautious upper intake limit for people. They do not prove that a daily multivitamin is dangerous. High-dose animal toxicology is a tool for finding a ceiling; it does not preview what a normal diet does.

What is established, and what is extrapolated

Here is the honest split, thread by thread.

Established by multiple human trials:

A systematic review identified 27 randomized crossover challenge trials on synthetic food dyes and children’s behavior. Just over half found some association with adverse behavior, and about half of those reached statistical significance, which the reviewers concluded supports a real link between dye mixtures and short-term behavioral effects in some children^[6]. The same review found that current FDA acceptable daily intakes rest on older studies that were not designed to catch behavioral effects at all^[6].

Established by observational cohorts, which show association but cannot prove causation:

A NutriNet-Santé cohort study of 101,056 French adults found that higher dietary nitrite and nitrate intake, including from food additive sources, was associated with increased cancer risk^[7]. That is an observational association. It cannot by itself prove the additive caused the disease, only that the two track together in a large population after statistical adjustment.

101,056 French adults in the NutriNet-Santé nitrite and nitrate cancer-risk cohort PMID 35303088

Mostly extrapolated from mice, with one human data point:

The claim that common emulsifiers cause gut inflammation in humans at real-world doses rests mainly on mouse studies. The one controlled human feeding trial found altered gut bacteria and metabolites, and in two of seven CMC-fed participants a marker of gut inflammation, without any participant developing a diagnosed inflammatory disease^[3].

Two out of seven showing a biological marker is worth watching. It is not the same as a trial showing CMC causes inflammatory bowel disease in people.

Extending a single mouse study or mechanism finding straight into “avoid this additive” advice skips the steps regulators use: what dose, in what population, over what duration. That shortcut is where most avoid-list overreach happens.

Bottom line: synthetic dye mixtures have the strongest human trial evidence in this category, though the size of the effect and the FDA’s willingness to act on it remain contested. Carboxymethylcellulose has the only controlled human feeding data on emulsifiers, and it shows real but modest changes well short of diagnosed disease. Nitrite, nitrate and sweetener concerns rest on large, well-adjusted cohorts that show association without proving causation.

Where to read the primary evidence

Four papers cover most of what matters here, each doing a different job:

• For the single best paper on dyes and behavior, read the 2007 randomized trial that tested artificial color mixtures against placebo in preschool and school-age children^[2].
• For the fuller picture on colors, the systematic review of 27 clinical trials is the most complete synthesis available^[6].
• For the cohort side, the NutriNet-Santé nitrite and nitrate cancer-risk analysis is the largest and most cited human observational work connecting additive mixtures to disease risk^[7].
• For emulsifiers, the controlled carboxymethylcellulose feeding trial is the one to watch as researchers extend the same design to other emulsifiers; it is currently the only trial of its kind in healthy humans^[3].

None of this makes every additive on the popular lists equally suspect, and it does not clear any of them either. A shopper deciding what to do is better served by knowing which claim in front of them is trial-grade and which is a cohort association wearing a trial’s confidence.

If you want to check that distinction ingredient by ingredient rather than list by list, see how the leading scanner apps compare on the evidence or how much of a competing app’s scoring is proven.

This article summarises publicly available regulatory assessments and peer-reviewed research about food ingredients. It is general information and is not intended to diagnose, treat, cure, or prevent any disease, and it is not personalised dietary or medical advice. Regulatory positions change; the EatMatrix app reflects the current assessments in our database. If you have an allergy, a medical condition, or are pregnant, talk to a qualified clinician or dietitian before changing what you eat.