As new international guidelines seek to harmonise precautionary allergen labelling, Ben Cornwell examines the science underpinning thresholds for ‘may contain’ warnings and what it takes for manufacturers to make a scientifically defensible labelling decision.

Imagine two chocolate bars sitting alongside one another on a supermarket shelf. Both are made in facilities handling peanuts, yet only one carries a familiar warning: ‘may contain peanuts’.

Which is safer?

The manufacturer applying the warning may have identified a genuine cross-contact risk. The other may have evidence that exposure remains below a scientifically justified threshold. Equally, one may have added a warning on legal advice while the other has never undertaken a sufficiently robust risk assessment.

For someone with a peanut allergy, neither wrapper offers much insight into the work that has gone on behind it.

Dr Sung Poblete, Chief Executive Officer of Food Allergy Research & Education (FARE), says precautionary allergen labelling (PAL) remains voluntary and largely unregulated in the US, where her organisation has identified more than 45 different precautionary statements on supermarket shelves. She explains that inconsistent and widespread use of PAL can leave some consumers avoiding potentially suitable foods altogether, while others decide to take the risk.

The consequences of defensive allergen management became particularly apparent following the introduction of mandatory sesame labelling under the US Food Allergy Safety, Treatment, Education and Research (FASTER) Act in January 2023. Some manufacturers deliberately added sesame to bakery products that previously did not contain it, leaving sesame-allergic consumers with fewer choices.

International efforts to give precautionary warnings a consistent scientific basis reached a milestone in July 2026, when the Codex Alimentarius Commission adopted new guidelines as an annex to its General Standard for the Labelling of Pre-packaged Foods (CXS 1-1985).

Informed by scientific advice from the Food and Agriculture Organization (FAO) and World Health Organization (WHO), the guidelines link precautionary warnings to the risk that remains after manufacturers have taken appropriate measures to minimise cross-contact. Although their adoption does not automatically change national legislation, they offer businesses a framework for assessing their existing practices.

Consumers need to understand that under a standardised framework, a product without precautionary allergen labelling (PAL) has undergone a risk assessment, not that the manufacturer removed the statement without evaluating the risk.”

Dr Sung Poblete, Chief Executive Officer of Food Allergy Research & Education (FARE)

How much allergen is too much?

Even among people allergic to the same food, the amount of allergenic protein needed to trigger a reaction can vary considerably. Researchers use data from clinical food challenge studies to model these population-level dose-response relationships.

The eliciting dose (ED) is the amount of allergenic protein predicted to provoke objective allergic symptoms in a specified proportion of people allergic to that food. ED01 corresponds to one percent of that population and ED05 to five percent. Some individuals may experience subjective symptoms, such as an itchy mouth, at lower doses.

The FAO/WHO expert consultation recommended reference doses broadly based on ED05. Expressed in milligrams of total protein from the allergenic food, these values provide benchmarks against which manufacturers can assess potential exposure on a single eating occasion. For peanut, that figure is just 2mg of peanut protein. These remarkably small quantities are population-level risk management values and cannot guarantee protection for every allergic individual.

Allergen reference doses at a glance

Selected FAO/WHO recommendations, broadly based on ED05.

AllergenReference dose

Peanut

2mg

Milk

2mg

Egg

2mg

Sesame

2mg

Cashew

1mg

Walnut

1mg

Almond

1mg

Hazelnut

3mg

Values represent milligrams of total protein from the allergenic food and are not intended to define ‘allergen-free’ claims.

These values concern immunoglobulin E (IgE)-mediated food allergies. The Codex guidelines address gluten exposure in relation to coeliac disease through a separate reference dose.

Source: FAO/WHO expert consultation on allergen reference doses.

Professor Paul Turner, an allergy specialist involved in the FAO/WHO expert consultations, explains that the evidence supporting individual reference doses varies between allergens. Some have extensive clinical challenge datasets, while others have required more conservative assumptions because fewer data were available.

The experts rounded and grouped certain values for conservatism and simplicity, recognising that recommendations may evolve as further evidence emerges.

Finding the right cut-off

Choosing ED05 over ED01 might initially appear counterintuitive. In developing the reference doses, however, the FAO/WHO experts considered the severity of allergic reactions alongside their likelihood.

The FAO/WHO expert consultation concluded that selecting ED05 over ED01 would not be expected to increase rates of severe anaphylaxis. The Food Standards Agency (FSA), in a board paper published in June 2026, similarly considered evidence indicating that additional objective reactions at ED05 were generally expected to be mild to moderate.

ED05 does not mean that five percent of allergic consumers eating products without PAL will experience a reaction.

Professor Turner acknowledges that such a risk would be unacceptable to him, but emphasises the distinction between clinical sensitivity and actual exposure.

“We’re not talking about every single package having that risk. We’re not talking about every single patient having that risk. Patients vary in how sensitive they are,” he explains. A contaminated chocolate bar must first reach someone with the relevant allergy, whose likelihood of reacting then depends on the amount of allergenic protein consumed and their individual sensitivity.

Professor Turner takes some reassurance from Japan, where mandatory allergen labelling has operated with an analytical threshold of 10mg/kg of soluble allergenic protein for specified allergens since 2002. Although Japan’s concentration-based system differs from the FAO/WHO approach and does not permit conventional ‘may contain’ warnings, it offers a longstanding example of threshold-based allergen management.

There is also a practical question of whether manufacturers can reliably measure the concentrations they are being asked to control. For some priority allergens, the concentrations associated with ED01-based action levels are beyond the reliable quantification capabilities of available analytical methods.

“It’s about finding the right sort of cut-off so that it can work in industry,” explains Professor Turner. “If this suddenly costs loads and loads of money, it’s not going to work. Likewise, if it’s going to end up being something that everything just has a ‘may contain’ label on it still, that’s not going to work either.”

The FAO/WHO assessment of PAL concluded that a lower reference dose could slightly reduce the risk of allergic reactions but would result in more products carrying warnings, with little expected improvement in public health at a population level.

Businesses can still adopt more conservative internal operating limits. Professor Turner says some manufacturers already use ED01 as an operational buffer against an ED05-based threshold, allowing for process variability.

The FSA also notes that standardisation could prompt more businesses to use PAL and improve communication of cross-contact risks through supply chains. Some products may consequently acquire warnings as others lose them.

Agreeing on a reference dose is only part of the challenge. Manufacturers still have to translate that dose into a concentration they can assess and manage in their products.

From milligrams to manufacturing decisions

Consider a manufacturer producing milk chocolate and dark chocolate on shared equipment. Small amounts of milk protein may remain inside the machinery after a changeover. Whether that warrants a warning on the dark chocolate depends partly on how much of it someone might eat.

The Allergen Bureau’s Voluntary Incidental Trace Allergen Labelling (VITAL) programme provides an established methodology for translating reference doses into product-specific concentrations that inform PAL decisions. The reference dose is divided by a reference amount representing food consumption on a single eating occasion.

From reference dose to action level

Illustrative example of unintended milk protein in chocolate.

Reference dose: 2mg milk protein.

Action-level transition point (mg/kg) = Reference dose (mg) ÷ Reference amount (kg)

Reference amountCalculationTransition point (mg milk protein/kg food)

50g chocolate

2mg ÷ 0.05kg

40mg/kg

100g chocolate

2mg ÷ 0.10kg

20mg/kg

Doubling the reference amount halves the corresponding transition point.

For unavoidable, readily dispersible cross-contact, VITAL assigns:

  • Action Level 1: below the transition point; PAL is not required
  • Action Level 2: at or above the transition point; PAL is required where legally permitted

Illustrative calculations only. Labelling decisions require a full risk assessment, including separate consideration of particulate contamination.

Sources: VITAL 4.0 Food Industry Guide (p. 11) and FAO/WHO allergen reference doses.

Dr Simon Brooke-Taylor, Director and Principal Consultant at Brooke-Taylor & Co and a member of the Allergen Bureau’s VITAL Scientific Expert Panel, identifies choosing an appropriate reference amount as one of the more challenging decisions for manufacturers.

A manufacturer’s suggested serving size may differ considerably from what someone actually eats. Dr Brooke-Taylor illustrates the challenge using a familiar example.

“So if we’re talking about a Mars bar, that’s maybe one. But if you buy a multi-pack, that could be two,” he says. Recent discussions within the VITAL programme have considered whether single products and multipacks should have different reference amounts, although the question remains unresolved.

VITAL 4.0 recommends using the maximum amount eaten on a typical eating occasion, which should be no smaller than the declared serving size. The guide also acknowledges the FAO/WHO recommendation to use the 50th percentile of consumption data for a single eating occasion.

As the chocolate example illustrates, underestimating how much someone might eat could leave a manufacturer working to an insufficiently protective concentration.

The work behind the warning

Calculating a transition point of 20mg/kg of milk protein for a 100g reference amount is relatively straightforward. Establishing whether residues in shared manufacturing equipment could cause the finished product to reach or exceed that concentration can be considerably harder.

Professor Turner describes how milk residues trapped inside a small T-shaped junction in shared chocolate processing equipment can be particularly difficult to remove. Shutting down the line, wet cleaning and drying the equipment before restarting production could, in his example, take several days.

Artisan chocolate processing

Other cross-contact risks originate long before ingredients reach the factory. Dr Brooke-Taylor cites wheat grown in parts of Australia, which may encounter lupin through shared agricultural and transport equipment.

“There is a place for analytical testing, but it must not replace that first risk assessment,” he explains.

He advocates routine verification, supported by a documented allergen management strategy, to demonstrate that controls continue to perform as expected during normal production.

”If you’re doing testing, you should show that you’re consistently either below the level if you don’t have a PAL statement or you’re getting it as low as possible if you do have a PAL statement,” says Dr Brooke-Taylor

Professor Turner describes the reference dose as “not a permissive level”. Where an assessment identifies potential exposure above it, manufacturers should first investigate whether further mitigation is possible.

Under the FAO/WHO framework, PAL is justified where appropriate mitigation cannot sufficiently control cross-contact and a qualified risk assessment identifies possible unintended allergen exposure above the relevant reference dose. Where the assessment supports unintended exposure at or below that dose, PAL should be omitted. Mandatory declaration requirements for allergenic ingredients still apply.

Both the FSA’s updated allergen labelling guidance and the Codex Code of Practice on Food Allergen Management (CXC 80-2020) reinforce the importance of preventative controls. The responsibility to minimise cross-contact continues after a warning has been applied.

Dr Brooke-Taylor recalls instances in Australia and New Zealand where authorities found allergen levels in products carrying PAL that were inconsistent with good manufacturing practice. He explains that regulators can order recalls in such circumstances despite the presence of a warning.

What evidence should support a PAL decision?

  1. Cross-contact pathways
    Identify where unintended allergens could enter the product, including during sourcing, transport, storage and manufacturing.
  2. Effectiveness of controls
    Assess existing preventative measures and establish whether further mitigation is practicable.
  3. Residual exposure
    Estimate potential allergenic protein intake using an appropriate reference amount. Consider particulate contamination separately.
  4. Supporting evidence
    Document relevant assumptions and exposure calculations alongside evidence of control effectiveness. Account for sampling and analytical limitations where testing is undertaken.
  5. Ongoing verification
    Verify controls over time and reassess following relevant changes to ingredients, suppliers, equipment or manufacturing processes. VITAL also recommends reviewing assessments in response to allergy-related consumer complaints, new knowledge or at least every 12 months.

Summary of key risk assessment considerations. Applicable regulatory requirements should also be consulted.

Sources: FSA allergen labelling guidance and VITAL 4.0 Food Industry Guide.

What a negative test cannot prove

Imagine an occasional peanut fragment finding its way into a chocolate bar through shared equipment. Hundreds of individual bars might contain no detectable peanut, while one contains a fragment delivering enough protein to exceed the reference dose.

A handful of negative tests could miss that contamination entirely. As the Allergen Bureau’s guidance on agricultural cross-contact illustrates, combining material into composite samples may also dilute localised allergen concentrations. An average concentration may therefore poorly represent what an individual consumer actually ingests.

Even when a sample is representative, the composition of the food itself can influence how reliably an analytical method measures the allergen. The FAO/WHO experts identified a need for improved reference materials and a better understanding of these matrix effects.

Laboratory analysis of chocolate

Variability between production runs further complicates the design of a risk-based food testing programme. Manufacturers need to consider how representative their results remain over time.

VITAL 4.0 distinguishes readily dispersible cross-contact from discrete allergen particles. Where unavoidable particulate contamination could deliver at least the relevant reference dose, the programme assigns Action Level 2, requiring PAL within its framework where legally permitted.

Repeated negative tests alone cannot justify removing a warning where such a risk remains. Equally, a consumer’s previous experience of eating a product without reacting offers little reassurance where contamination is sporadic.

“You might have eaten the product 20-30 times quite safely, because you’ve ignored it, and then one day you get a reaction,” says Dr Brooke-Taylor.

There is a place for analytical testing, but it must not replace that first risk assessment.”

Dr Simon Brooke-Taylor, Director and Principal Consultant at Brooke-Taylor & Co and a member of the Allergen Bureau’s VITAL Scientific Expert Panel

When the warning disappears

A consumer who has avoided a particular chocolate bar for years may understandably be cautious when its precautionary warning disappears. Manufacturers adopting more rigorous risk assessments may need to explain what has changed.

“Consumers need to understand that under a standardised framework, a product without precautionary allergen labelling (PAL) has undergone a risk assessment, not that the manufacturer removed the statement without evaluating the risk,” explains Dr Poblete.

FARE’s research, she says, indicates that consumers would welcome fewer, more meaningful warnings. The organisation advocates a single standardised statement naming the specific allergen, such as ‘may contain almonds’ rather than ‘may contain tree nuts’. Consistent placement immediately after the ingredients list or ‘Contains’ statement would also help shoppers know where to look.

Even among allergy experts, how that remaining risk should be communicated is unresolved. In an August 2026 World Allergy Organization (WAO) consensus, 70 percent of participating experts supported informing consumers about residual risk below reference doses, but there was no consensus on how best to do so.

Dr Poblete believes education and discussions with allergists will help individuals understand what that residual risk means for them. The FAO/WHO recommendations advise avoiding foods carrying PAL for the relevant allergen, with any exceptions discussed with a healthcare professional.

“Families would no longer have to wonder whether PAL was added just for legal protection, or pass on a food because it says ‘made in a facility with’,” says Dr Poblete.

Making a warning mean the same thing

Consistent wording will only go so far if the evidence behind those warnings differs between manufacturers. Even with internationally agreed reference doses, businesses may differ in their assumptions about how much of a product someone might consume and how cross-contact is evaluated.

Professor Turner draws reassurance from comparisons of peanut-allergy challenge data across several countries, which he says have not identified substantial differences in sensitivity at the low exposure levels relevant to ED05. He notes differences at higher doses, although these comparisons alone cannot establish whether similar patterns exist for other allergens.

With PAL remaining voluntary in the UK, US, Australia and New Zealand, much of the immediate progress will depend on industry. The Allergen Bureau’s VITAL programme provides an established example of quantitative risk assessment in practice, and Dr Brooke-Taylor encourages businesses to take the initiative rather than wait for further regulation.

“In short, PAL needs to move from ‘litigation protection’ to ‘consumer protection,’ and FARE believes the strongest version of this framework would be mandatory,” says Dr Poblete.

Professor Turner argues that more meaningful cross-contact assessments could provide ingredient purchasers with clearer information about the risks entering their own production systems. Reducing that uncertainty could also help businesses avoid costly recalls.

Return to those two chocolate bars on the supermarket shelf. Under a consistently applied risk-based framework, their labels should differ for a reason. One manufacturer has identified a residual cross-contact risk that warrants a warning; the other has evidence to support leaving it off.

The challenge for industry will be demonstrating that both decisions deserve the same degree of confidence.

Further reading

The following FAO/WHO scientific summaries, published in 2024, provide an accessible overview of the evidence and recommendations that informed the new Codex guidelines on precautionary allergen labelling.

Regulatory and industry guidance