Who checks that food additives are safe?

A medicine proves itself first. An additive often doesn't.

A new medicine must pass trials before anyone may sell it. A food additive in the US can take another road. The 1958 Food Additives Amendment requires approval from the Food and Drug Administration (FDA), unless experts "generally recognise" the substance as safe. That exception, called GRAS, is where most of the action is.

Under GRAS, a company can:

1. pick a substance it wants to use;
2. pay a panel of experts to review the evidence;
3. declare the substance GRAS; and
4. choose whether to tell the FDA at all.

The US Government Accountability Office reviewed this system in 2010. It found companies can decide a substance is GRAS "without FDA's approval or knowledge", that the FDA sees only the determinations companies choose to send it, and that the agency had not checked whether companies made their determinations properly.

Who sits on the panels

In 2011, researchers at the Pew Charitable Trusts mapped the US system and counted more than 10,000 substances allowed in human food. In 2013 the same group looked at the 451 GRAS determinations companies had sent to the FDA between 1997 and 2012. An employee of the manufacturer made 22% of them. An employee of a consulting firm the manufacturer chose made another 13%. Experts picked by an independent third party made none.

None of this proves any one substance is harmful. It shows who does the checking: the people paid by the seller.

How long a fix takes

When the system gets it wrong, the fix is slow. Two cases:

Substance	What was known	What the FDA did
Artificial trans fats (partially hydrogenated oils)	A 2006 review in the New England Journal of Medicine linked trans fat intake to coronary heart disease, sudden cardiac death and diabetes.	Declared them no longer GRAS in June 2015. Food makers had until June 2018 to take them out.
Red No. 3 (erythrosine), a pink-red dye	Caused thyroid tumours in male rats. The FDA banned it from cosmetics and skin products in 1990.	Revoked its use in food in January 2025. Food makers have until January 2027.

The Red No. 3 case needs a fair note. The FDA says the rat tumours come from a hormone mechanism that doesn't occur in people. It banned the dye because US law forbids any colour that causes cancer in animals, not because it found a risk to humans. Either way, the same dye stayed in food for 35 years after the same agency took it out of lipstick.

Same evidence, different answers

The European Union works the other way round. An additive can't go on sale until the European Food Safety Authority (EFSA) has assessed it and it is on an approved list. That produces different decisions from the same research.

• Food colours and children. In 2007 a randomised, double-blind trial in Southampton gave about 300 children drinks with a mix of artificial colours and the preservative sodium benzoate, or a placebo. Some children became measurably more hyperactive on the additive drinks. Since July 2010, EU foods containing any of the six colours tested must carry the warning "may have an adverse effect on activity and attention in children". The US and Australia require no such warning.
• Titanium dioxide (E171), a whitener in lollies and icing. In 2021 EFSA said it could not rule out that the additive damages DNA, and the EU banned it from food in 2022. Food Standards Australia New Zealand (FSANZ), the UK and Canada reviewed the same question and found no safety concern. It is still allowed here.

Neither side is always right. The point is that the rules, not the science alone, decide what ends up in the packet.

What Australia does

FSANZ approves additives before use. An additive goes into the Food Standards Code only if the evidence shows no harmful effects are likely, and FSANZ assesses both the chemical's hazard and how much of it people eat. That is a real check, and stronger than GRAS. The evidence, though, comes largely from the applicant, the company that wants the approval.

Newer evidence the old tests missed

Most additive tests ask whether a chemical is toxic at a given dose. Recent research asks different questions.

• Processing itself. In a 2019 US National Institutes of Health trial, 20 adults lived on a hospital ward for four weeks: two on ultra-processed food, two on unprocessed food, matched for calories, sugar, fat, salt and fibre on offer. On the ultra-processed diet they ate about 500 more calories a day and gained 0.9 kg. On the unprocessed diet they lost 0.9 kg.
• Emulsifiers and the gut. In a 2022 randomised feeding trial in healthy adults, the emulsifier carboxymethylcellulose (a thickener in ice cream, sauces and baked goods) changed gut bacteria and gut chemistry, and some people reported more stomach discomfort. The study was small and short. It is a signal worth following, not proof of disease.

Could the burden be reversed?

New Zealand tried it for a different product. Its Psychoactive Substances Act 2013 banned new recreational drugs ("legal highs") by default and made the maker prove a product posed a low risk before sale. The lesson cuts both ways. In 2014 Parliament barred approvals based on animal tests, and the regulator said a product probably couldn't show low risk without them. By 2015, researchers reported, no product had been approved. Reversing the burden works only if there's a workable way to meet it.

What you can check

• Read the ingredients, not the front of the pack. Additives appear by name or code number, such as 171 or 466.
• Look up any code on FSANZ's list of food additive names and code numbers.
• If a product is imported from the US, remember a GRAS ingredient in it may never have been seen by a regulator.
• Eating more food that's close to how it grew is the one choice the 2019 trial supports directly.