If an additive is banned in Europe, why is it still allowed elsewhere?

Usually because the two systems set different thresholds for acting when the evidence is incomplete, not because one of them is concealing harm. European regulation leans on the precautionary principle, which permits restriction when a risk cannot be excluded; several other systems act only when a risk is demonstrated at realistic exposures. Titanium dioxide is the clearest example: EFSA concluded in 2021 that safety could no longer be assumed because genotoxicity could not be ruled out, and the EU banned it, while other regulators reviewing the same data concluded that harm had not been shown. There are also differences in approval routes, in the age of the original assessments, and in consumer and industry preference. Divergence is a signal to look at the reasoning, not proof that something is being hidden.

Do food colours cause hyperactivity in children?

There is a small real effect, and the regulatory response has been proportionate to it. Two rigorous randomised double-blind trials at Southampton gave children mixtures of artificial colours with sodium benzoate and found small but measurable increases in hyperactive behaviour, in children from the general population rather than only those with diagnosed ADHD. EFSA judged this insufficient to change the permitted intake levels, partly because the mixtures made it impossible to identify which components were responsible. The EU nevertheless requires products containing six specific colours to state that they may adversely affect activity and attention in children — and most manufacturers reformulated rather than carry the label. So the honest answer is: a small effect, real enough to justify labelling, not large enough to justify prohibition.

Are emulsifiers a genuine concern?

This is the one area where the science is genuinely open and moving. Studies in mice have shown that common emulsifiers such as carboxymethylcellulose and polysorbate-80 can thin the protective mucus layer lining the intestine, allow bacteria to come into closer contact with the gut wall, change the microbiota and produce low-grade inflammation and metabolic changes. A small controlled feeding study in humans found microbiota and metabolomic changes with carboxymethylcellulose. That is a coherent and specific mechanism, which is more than most additive concerns can claim. It is also early: the animal doses may not reflect human exposure, the human studies are small and short, and there is no evidence yet linking emulsifiers to clinical disease. It deserves serious research and does not yet justify avoidance.

Are nitrates in food dangerous?

It depends entirely on what they come with, which makes this one of the most instructive examples in food chemistry. Around eighty per cent of dietary nitrate comes from vegetables, and vegetable nitrate is associated with lower blood pressure and cardiovascular benefit, because it is converted to nitric oxide, a vasodilator — this is the basis of beetroot juice research. The same ions in cured meat contribute to the formation of N-nitroso compounds, which are established carcinogens and part of the mechanistic case against processed meat. The difference is the surrounding matrix: vegetables supply vitamin C and polyphenols that block the nitrosation reaction, while cured meat supplies haem iron and amines that promote it. So neither "nitrates are toxic" nor "nitrates are fine" is a usable statement.

Should I choose products with shorter ingredient lists?

Not as a safety strategy, though it may be a reasonable proxy for something else. A shorter list does not mean safer: preservatives prevent microbial growth and mould toxin formation, and removing them shortens shelf life, increases food waste and can raise genuine safety risk. Natural colour replacements are often less stable and used in larger amounts. Some replacement substances have smaller toxicological datasets than the long-established additives they replace. What a short ingredient list often does indicate is that the product is less heavily processed, and that is where the actual outcome evidence lies — but the marker should not be mistaken for the mechanism. Reformulating a sweet or a snack with natural colours does not change what it is, and the sugar, salt, energy density and portion size matter far more than the additive list.

How does GMJ News cover food additives?

GMJ News covers why regulatory divergence reflects different thresholds for acting under uncertainty rather than concealed harm, the acceptable daily intake framework and its conservative safety factors, the Southampton colour studies and the EU labelling response, the titanium dioxide precautionary ban and the reasoning on both sides, emulsifiers and intestinal mucus layer research as the genuinely open question, the nitrate matrix effect distinguishing vegetables from cured meat, the GRAS approval route and its structural criticisms, and the clean label problem in which additive removal carries its own trade-offs. Part of PHIG's contested science and food safety programmes.

Home › Topics › Food Additives and Colours

Food Additives and Colours

GMJ News knowledge hub · last reviewed September 2026 · Georgian Medical Journal

Summary Statistics Glossary GMJ news FAQ Documents Organizations Research

Food additives occupy a space where regulatory divergence between jurisdictions is routinely misread as evidence of concealed danger, when it more often reflects different regulatory philosophies and different tolerance for uncertainty — the European decision to remove titanium dioxide from the permitted list in 2022 on the grounds that genotoxicity could not be excluded, alongside continued authorisation elsewhere, is a disagreement about how to act under uncertainty rather than a dispute about a demonstrated harm (WHO). On synthetic colours, the evidence is more nuanced than either side concedes: the Southampton studies found small effects on hyperactivity in children in the general population, sufficient for the EU to require warning labelling but not to prohibit use, which is a defensible middle position that satisfies nobody. The emerging and more scientifically interesting question concerns emulsifiers and the intestinal mucus layer, where mechanistic animal work is suggestive and human evidence remains preliminary — an area where honest uncertainty is the only accurate summary.

Key messages

REGULATORY DIVERGENCE IS NOT EVIDENCE OF CONCEALED HARM

The most common inference in this area is also the least reliable: that if a substance is permitted in one jurisdiction and restricted in another, the permitting jurisdiction must be hiding something. Divergence more often reflects different thresholds for acting under uncertainty, different legal frameworks — the EU precautionary principle versus risk-based US regulation — different historical approval routes, and consumer and industry preference. The EU decision to remove titanium dioxide in 2022 rested on inability to exclude genotoxicity, not on demonstrated harm, which is a defensible position and a different claim.

SYNTHETIC COLOURS: a small real effect, and a proportionate regulatory response

The Southampton studies found small but statistically detectable increases in hyperactive behaviour in children in the general population following mixtures of certain artificial colours with sodium benzoate. EFSA judged the evidence insufficient to change acceptable daily intakes but the EU nonetheless requires products containing the implicated colours to carry a warning that they may adversely affect activity and attention in children. That is a genuine middle position — acknowledging a signal without prohibition — and it satisfies neither those who want a ban nor those who consider the effect negligible, which is often what an honest regulatory response looks like.

THE MECHANISTICALLY INTERESTING QUESTION: emulsifiers and the intestinal mucus layer

This is where the science is genuinely moving. Animal studies show that common emulsifiers including carboxymethylcellulose and polysorbate-80 can thin the protective intestinal mucus layer, permit bacterial encroachment on the epithelium, alter microbiota composition and promote low-grade inflammation and metabolic syndrome. A small controlled human feeding study of carboxymethylcellulose found microbiota and metabolomic changes and reduced mucus thickness. This is preliminary, the doses and human relevance are debated, and no clinical outcome evidence exists — but it is a specific, testable hypothesis of exactly the kind that deserves serious investigation rather than either dismissal or alarm.

NITRITES AND NITRATES: source determines effect

A useful case study in why chemical identity alone is insufficient. Nitrites used in meat curing contribute to N-nitroso compound formation and are part of the mechanistic basis for the processed meat classification. Yet vegetables are the dominant dietary source of nitrate by a wide margin, and vegetable nitrate is associated with lower blood pressure and cardiovascular benefit through the nitrate-nitrite-nitric oxide pathway. The difference lies in the matrix: vegetables supply vitamin C, polyphenols and other reductants that inhibit nitrosation, while cured meat supplies haem iron and amines that promote it. The same ion, opposite consequences.

THE CLEAN LABEL PROBLEM: removing additives is not automatically an improvement

Consumer pressure for shorter ingredient lists has produced reformulations that are not always favourable. Preservatives prevent microbial spoilage and mycotoxin formation, and removing them shortens shelf life, increases waste and can raise food safety risk. Natural colour replacements are frequently less stable and used at higher concentrations. Some replacements are less studied than the substances they replace, since long-standing additives typically have larger toxicological datasets. The point is not that additives are beneficial but that removal has trade-offs, and a shorter ingredient list is a marketing signal rather than a safety measure.

WHAT ACTUALLY DRIVES THE HARM: the food, not usually the additive

The strongest evidence in this space concerns the products additives appear in rather than the additives themselves. Ultra-processed foods are associated with adverse outcomes, and the presence of multiple additives is a defining marker of that category — which makes additives an excellent indicator and a poorly evidenced cause. This distinction matters practically: reformulating a confectionery product to use natural colours does not make it a health food, and regulatory attention focused on individual additives can displace attention from energy density, sugar, salt and marketing, where the evidence is far stronger.

Key statistics

Not evidence of harm

regulatory divergence usually reflects different thresholds for acting under uncertainty

EFSA/FDA/JECFA

Warning, not ban

EU requires activity and attention labelling for Southampton colours rather than prohibition

EU Regulation 1333/2008

EFSA 2021

titanium dioxide no longer considered safe as a food additive — genotoxicity could not be excluded

EFSA 2021

Emulsifiers

mucus thinning and microbiota change shown in animals and one small human feeding study

Nature/Gastroenterology

Vegetables

the dominant dietary nitrate source, associated with benefit — matrix determines the effect

EFSA/AJCN

Marker not cause

additives are a defining marker of ultra-processed food, which is where the outcome evidence sits

Editorial

Food additives — where the disagreement actually lies

Source: Bars show strength of supporting evidence. Most additive safety is well established; the emulsifier question is the genuinely open one.

Glossary of key terms

Acceptable daily intake and the safety factor
ToxicologyThe ADI is the amount of a substance that can be consumed daily over a lifetime without appreciable health risk. It is derived by identifying the no-observed-adverse-effect level in animal studies and dividing by an uncertainty factor, conventionally 100 — tenfold for interspecies extrapolation and tenfold for variation between humans, with further factors applied where data are limited. Exposure assessments then compare estimated population intake with the ADI, usually finding realistic intakes well below it. Two consequences follow. First, exceeding the ADI occasionally does not imply harm, because the safety margin is deliberately conservative. Second, the ADI is not a threshold of toxicity, and describing it as the level at which something becomes dangerous, as commonly happens, misrepresents the entire framework.

The Southampton studies and EU labelling
Regulation/EvidenceTwo randomised double-blind placebo-controlled trials at the University of Southampton administered mixtures of artificial colours with sodium benzoate to children in the general population and found small but statistically significant increases in hyperactive behaviour. EFSA reviewed the findings and concluded they provided limited evidence of a small effect, insufficient to alter acceptable daily intakes given uncertainties about which components were responsible in the mixtures. The EU nonetheless legislated a mandatory warning statement for six specific colours — sunset yellow, quinoline yellow, carmoisine, allura red, tartrazine and ponceau 4R. Many manufacturers reformulated to avoid the label rather than carry it, achieving a substantial reduction in use without prohibition — an instructive example of labelling as a regulatory lever.

Titanium dioxide (E171)
RegulationA whitening and opacifying agent used in confectionery, bakery products, sauces and chewing gum. EFSA concluded in 2021 that it could no longer be considered safe as a food additive, on the specific grounds that concerns about genotoxicity could not be excluded — chiefly because a proportion of the particles are in the nanoscale range and the available data could not establish absence of DNA damage. The EU ban followed in 2022. It remains permitted in several other jurisdictions, where regulators concluded the available evidence did not demonstrate risk at realistic exposures. Both positions follow logically from different starting points: precaution when uncertainty cannot be resolved, versus action only on demonstrated risk. Neither implies dishonesty by the other.

Emulsifiers and intestinal barrier function
Emerging scienceEmulsifiers stabilise mixtures of fat and water and are among the most widely used additive classes in processed food. Work from Chassaing and Gewirtz showed that carboxymethylcellulose and polysorbate-80 administered to mice altered microbiota composition, reduced the thickness of the inner mucus layer, allowed bacteria to approach the epithelium, and induced low-grade inflammation, metabolic syndrome and, in susceptible strains, colitis. A subsequent randomised controlled feeding study in humans found that carboxymethylcellulose altered microbiota composition and metabolomic profile and reduced mucus-epithelial separation, though clinical effects were minimal over the short study duration. The mechanism is coherent, the human data are early, and the appropriate description is a promising hypothesis requiring outcome studies.

Nitrate, nitrite and the food matrix
Food chemistryAround 80% of dietary nitrate comes from vegetables, particularly leafy greens and beetroot, and this intake is associated with lower blood pressure through reduction to nitrite by oral bacteria and then to nitric oxide, a vasodilator — the basis for beetroot juice supplementation research. The same ions in cured meat participate in N-nitroso compound formation, which is carcinogenic. The determining factor is what accompanies them: vegetables provide ascorbate and polyphenols that inhibit nitrosation, while cured meat provides haem iron, amines and the conditions of the stomach and colon that favour it. This is the clearest available illustration that chemical identity does not determine biological effect independently of context, and it undermines both "nitrates are poison" and "nitrates are harmless" as general claims.

GRAS and the regulatory approval routes
RegulationIn the United States, substances Generally Recognised As Safe are exempt from formal pre-market approval, and since 1997 manufacturers may determine GRAS status themselves and notify the FDA voluntarily, or not notify at all. Critics including the US Government Accountability Office have argued this creates a route by which substances enter the food supply without independent review, and that conflicts of interest in expert panels are inadequately managed. The EU system requires positive authorisation with an E number following EFSA assessment, and maintains a re-evaluation programme for previously approved additives. This structural difference explains part of the regulatory divergence between jurisdictions, and it is a legitimate criticism of process rather than a demonstration that any specific additive is harmful.

Latest GMJ coverage

Articles will appear here as the archive grows.

Food Additives and Colours

Key messages

Key statistics

Glossary of key terms

Latest GMJ coverage

Frequently asked questions 12 Q&A — structured for Google featured snippets and AI discovery

Knowledge hub: guidelines, conventions and reports

Organizations working in migration and health

Related health topics

Company