New: ASCloud, our paperless food safety system. Full document FSMS from R3,800 a month, R638 plans from R499. See pricing

Allergen cleaning validation on a shared production line

Allergen cleaning validation is the evidence, gathered before you rely on a clean, that a written changeover procedure removes a named allergen from a named line to a level you have set and can defend. Verification is the check, every time afterwards, that the validated procedure was followed and still works. Many South African factories that share a line between allergen and non-allergen products have the second without the first.

That gap matters under South African law. Regulation 45 of R146 says that where there is a risk of cross contamination with a common allergen, due diligence must be exercised to prevent it and an allergen control policy must be implemented. Precautionary labelling may not be used to get round good manufacturing practice and an effective allergen control policy. Where it is used, the risk, how it was assessed and the steps taken to avoid it must be documented.

R146 does not use the word validation, but a validation study is the strongest evidence a site can show that its due diligence works. If you are the person who has to design the study, our Advanced Allergen Management and Validation Part 1 course teaches it step by step.

In short

  • Codex CXC 80-2020 says validation should be specific to the allergen, the process and the product matrix.
  • The target comes from a reference dose. For peanut, milk and egg the FAO/WHO expert consultation set 2 mg of total protein.
  • Action level in mg/kg equals the reference dose in mg times 1,000, divided by the reference amount in grams. Peanut with a 40 g reference amount gives 50 mg/kg.
  • ATP and total protein swabs are not specific to allergens. ELISA is the usual method for validation, and lateral flow devices suit routine verification.
  • FoodDrinkEurope’s guidance asks for three consecutive successful runs before a clean is called validated.

Validation and verification are different jobs

The Codex Code of Practice on Food Allergen Management for Food Business Operators, CXC 80-2020, draws the line clearly. Validation shows that a cleaning procedure is capable of removing the allergen to the required level. It is tied to what was tested, so a clean validated for milk powder on a blender is not automatically validated for peanut paste on the same blender.

Verification then shows, each time, that the procedure was carried out and worked. Codex expects visual inspection as a minimum, with testing where it is feasible and appropriate. Section 5.7.1 of the code lists validation data for allergen cleaning among the records a business could keep.

The certification schemes ask for both. FSSC 22000 requires validation and verification of allergen control measures in clause 2.5.6 of its additional requirements, with verification testing at a frequency based on risk. Version 7, published in May 2026 and audited from 1 May 2027, also requires the verification data to be trended. BRCGS Food Safety Issue 9 asks for allergen cleaning methods to be validated and their effectiveness verified routinely, and management of allergens is one of its fundamental requirements.

The target comes from a reference dose

A validation needs an acceptance criterion. “No allergen detected” is a weak one, because it depends on how sensitive the test is, and FoodDrinkEurope falls back on it only where no action level has been set. The better start is a reference dose, the milligrams of protein from an allergenic food expected to cause an objective reaction in only a small proportion of allergic people.

The FAO/WHO expert consultation on food allergen risk assessment published reference doses in 2022, based on the dose expected to trigger a reaction in 5 per cent of the allergic population. A later report added soy and celery. The VITAL 4.0 programme, released in August 2024, adopted the same values.

Allergenic foodReference dose (mg total protein)Action level for a 40 g reference amount (mg/kg)
Walnut, pecan, cashew, pistachio125
Almond (provisional), celery125
Peanut, cow’s milk, egg, sesame250
Hazelnut375
Fish5125
Wheat5125, but no more than 20 mg/kg gluten
Soy10250
Crustacea2005,000
Reference doses from FAO/WHO (2022 and 2023). Action levels calculated by ASC. The wheat dose covers wheat allergy, not coeliac disease, so VITAL 4.0 caps the action level for cereals containing gluten at 20 mg/kg.

The action level converts the dose into a concentration you can test against. The VITAL 4.0 guide gives the formula as the reference dose in mg, multiplied by 1,000 and divided by the reference amount in grams, which is the largest amount a person would normally eat at one sitting and never less than the declared serving. The answer is in mg of protein from the allergenic food per kg of product. A small reference amount gives a higher action level, and a large one a lower action level.

Check the units on every laboratory report before you compare a result with that figure. Kits measure a marker protein and report it in different ways. The Allergen Bureau‘s own example is that 10 mg/kg of peanut is about 2.6 mg/kg of peanut protein, and a milk kit may report casein, beta-lactoglobulin or milk powder, each with its own conversion. Ask the laboratory to report mg of protein from the allergenic food per kg, as the FAO/WHO experts recommend.

Sites that set their criteria some years ago may have used VITAL 3.0, whose values were ten times lower for peanut, milk and egg. There is no need to relax a stricter criterion, but the validation file should say which framework it used.

The same arithmetic now sits in the Codex guidelines on precautionary allergen labelling, adopted at the 49th session of the Codex Alimentarius Commission in July 2026, which reserve “may contain” for unintended presence above the action level after mitigation. Codex is not South African law and R146 sets no numerical limit, but it is where retailers and export customers are heading. Our article on the new Codex rules on “may contain” covers the labelling side.

Which test can actually see the allergen

The test has to measure well below the action level. The FAO/WHO report advises that the method’s limit of quantification should be about three times lower than the action level, to allow for real-world variation. For peanut with a 40 g reference amount, the report puts the desired limit of quantification at 16 mg/kg.

ATP and total protein swabs are fast and cheap, but they are not specific to allergens. A review of allergen cleaning studies and guidance published by the UK Food Standards Agency found that surfaces which look clean, and even give a negative ATP result, can still carry detectable allergen. Several of the guidance documents it reviewed say ATP and protein swabs are not acceptable for validating allergen removal.

The FoodDrinkEurope guidance (version 2, March 2022) sets out what each specific method is good for.

  • ELISA detects the allergen protein itself, down to low mg/kg levels, and is the method to use for validation.
  • Lateral flow devices give a yes or no answer in minutes. They can support routine verification but are not appropriate tools for validation, and their results should be confirmed by ELISA from time to time.
  • PCR detects DNA rather than protein. It is useful as a confirmation where ELISA finds more than about 10 to 20 mg/kg, and it is not suitable for egg or milk, which carry little DNA and which it cannot tell apart from chicken or beef.
  • LC-MS/MS is mostly a confirmatory laboratory method. It is less affected by processing than the antibody methods are.

Codex warns that the kit must be validated for the food matrix, and that a casein test will not find whey when whey is the allergen of concern. The Allergen Bureau adds that cleaning chemicals can destroy the proteins a kit is built to find. Take validation samples after the final water rinse and before any sanitiser goes on, and ask the laboratory for evidence that the kit works in your product. A protein the kit can no longer see may still cause a reaction.

Heat is the other trap. In one study summarised in the FAO/WHO report, every ELISA kit found egg at its claimed limit of quantification in raw cookie dough, but none did in cookies baked for 30 minutes. Baking changes the proteins and locks them into the matrix, so less is extracted and recognised. It does not make the food safe, because Codex notes that heat generally does not destroy allergenic proteins.

Where and how to sample

Sampling is where most validation studies are weakest. The FoodDrinkEurope guidance calls swabs semi-quantitative at best and says they should not be used alone. Pair them with rinse water or, better, the first product made after the clean. Codex makes the same point for push-through cleans, where the push-through material or the first product through the line should be tested.

For clean-in-place systems, Codex expects verification by rinse samples or swabs, and warns against reusing rinse water where it could carry allergen back into the system. FoodDrinkEurope adds practical rules for rinse samples. The final rinse should be between pH 6.0 and 8.0, and samples should reach the laboratory below 4 °C and be tested within 24 hours.

Before the trial, run a swab recovery study. Apply a known amount of the allergenic food to a clean piece of the line’s own surface and let it dry before you swab and test it. The percentage recovered shows how much the swab misses on that surface. Without it, a low swab result could simply mean the swab did not pick the residue up.

Choose the swab points by worst case, not convenience. Dead legs, gasket seats, the corners of hoppers, scraper blades, belt joints and the underside of guards hold residue long after the open surfaces are clean. Our allergen risk assessment and cross contact scoring guide explains how to rank them.

How many runs an allergen cleaning validation needs

One good result proves very little, because a clean can pass on a day when the line happened to be easy. The FoodDrinkEurope guidance asks for the acceptance criteria to be met on three consecutive runs. A failure part of the way through means finding out why, changing the procedure if needed and starting the count of three again.

Each run should follow the written procedure exactly as the cleaning team will on a normal night shift, with the normal chemical, concentration, contact time and tools. A validation done by the hygiene manager with a new brush and an extra half hour proves only that the hygiene manager can clean the line.

Revalidate whenever something that affects the clean changes, such as a recipe, the equipment, the chemical or the cleaning contractor, and after any verification failure. That includes new products. A cookie with more peanut, or peanut pieces instead of paste, falls outside the validated worst case. Module 3 of our Part 1 course works through worst case selection, swab recovery correction and revalidation triggers.

A worked example from a biscuit line

Take a bakery that deposits peanut butter cookies and then plain shortbread on the same depositor, oven band and cooling conveyor. The shortbread is sold without a peanut warning, and its reference amount is 40 g. It makes both every day, so the peanut to shortbread changeover is the one to validate. The doughs are mixed in separate mixers, otherwise the mixers would belong in the study too.

  1. Set the target. The peanut reference dose is 2 mg of protein, so the action level is 2 × 1,000 ÷ 40, which is 50 mg of peanut protein per kg of shortbread. The laboratory method should quantify about 16 mg/kg or less in baked shortbread and report as peanut protein.
  2. Fix the worst case. Validate after the longest peanut run, using the cookie with the most peanut protein and the stickiest dough. Swab points include the feed rollers and scraper blades, the nozzles, the hopper corners, the scrap return, two marked points on the band and one on the cooling conveyor. The clean strips the hopper, rollers and nozzles for washing, and dry cleans the band and conveyor with scrapers and brushes, then a vacuum, never compressed air.
  3. Write the acceptance criteria before the trial. Every point on the inspection list must be visually clean, with no crumbs or peanut pieces. Peanut protein must be below 50 mg/kg in the first dough deposits and in the first baked biscuits, by an ELISA kit shown to work in both. Dough results are divided by one minus the bake loss, so with a 10 per cent bake loss a dough result of 45 mg/kg counts as 50 mg/kg. Any positive swab triggers an investigation before release.
  4. Run the recovery study on the band, then carry out three changeovers on normal shifts. Take the first dough deposits before the oven, then baked biscuits at the packer infeed at 0, 1, 5 and 10 minutes after the first ones arrive, at least three samples each time. Hold all shortbread from each trial run until that run’s results are back.
  5. Read the results. Suppose the highest figure, as peanut protein and corrected for bake loss, is 22 mg/kg in the first biscuits of the second run, with everything else below the limit of quantification. All three runs are under 50 mg/kg, so the clean is validated. The 22 mg/kg run is still worth a look. Its only positive swab was at the scrap return, cleaned last and in a hurry. If the procedure is changed to fix that, the next three changeovers count as a fresh validation.
  6. Set routine verification. At every changeover, a pre-start inspection signed by someone other than the person who cleaned, and a lateral flow swab of the rollers, nozzles and scrap return before shortbread dough goes in, so a failure costs a re-clean rather than product. Then a lateral flow test of the first shortbread, with a kit shown to work in it, and everything made since the changeover held until it is read. A positive result stops the line for a re-clean and retest, the held shortbread is treated as containing peanut, and the clean is revalidated. Every quarter, an ELISA on the first product, with the results trended.

Had the second run come back at 70 mg/kg, the clean would have failed. The shortbread from that run, still on hold, goes to waste or peanut cookie rework, or every pack carries a “may contain peanuts” statement. The team finds the cause, changes the procedure and runs three new consecutive trials. Until then, the shortbread is made before the peanut cookies, or carries the statement as a documented, temporary decision under regulation 45. A warning label does not replace the cleaning control.

The validation file then feeds the label decision, but it is one input, not the whole answer. Residues must be reliably below the action level, with no visible peanut pieces, and the risk assessment must show no other route for peanut into the shortbread, such as rework, shared scoops and dough tubs, or a bought-in ingredient with its own peanut warning. Then the business has evidence that a “may contain peanuts” statement is not needed. The Allergen Management Risk Assessment Template (RA17) records that decision line by line, so the label wording is something you can show an auditor rather than a habit.

What auditors and customers ask to see

An auditor reviewing allergen cleaning asks for the allergen risk assessment, the written cleaning procedure, the validation protocol and its results, the routine verification records, and training records for the people who clean. Expect two more questions. How are verification results trended and fed into the review of the allergen plan, which FSSC 22000 Version 7 requires from 1 May 2027? And what shows the test kit works in your product? Many auditors will also watch a changeover clean and its sign-off before the next product starts.

Our complete guide to allergen management covers the wider programme. R146 lists egg, cow’s milk, crustaceans and molluscs, fish, peanuts, soybeans, tree nuts and the significant cereals as common allergens, but not sesame, which Codex and many export markets now treat as a priority allergen.

Why food businesses choose ASC

Rated 4.9 out of 5 from more than 400 Google reviews across our three offices.

Gqeberha Google reviews Johannesburg Google reviews Cape Town Google reviews

Our online courses are rated 4.9 on average across 2,586 course reviews, from more than 7,100 enrolments.

ASC is ISO 9001 certified, a SAATCA registered online training course provider (TC 065), a FoodBev SETA accredited provider (No. 587/00337/1900) and an Exemplar Global Recognised Training Provider.

Allergen validation training

Learn to set the limit, clean to it and prove it

Advanced Allergen Management and Validation Part 1 (FS98) is an advanced online course of five modules, about nine hours in all. You work through action level arithmetic on the Codex and VITAL 4.0 frameworks, validation protocol design with three consecutive cleans, swab recovery studies and how to read ELISA, lateral flow, PCR and LC-MS/MS results. It ends with a proctored, closed book final examination of 50 questions, with a 70 per cent pass mark and three attempts. Need the removal science and the audit side as well? The complete course covers both parts.

R4,999 online. The complete course, Parts 1 and 2 together, is R7,999, which is R1,999 less than buying the two parts separately.

Enrol in Part 1 WhatsApp us about allergen validation

Documenting the risk? Allergen Management Risk Assessment Template (RA17) (R690, instant download).

Training the team? Allergen Management for Food Manufacturers Courses Bundle (R2,843 online for three courses, R3,790 bought separately).

Questions we are often asked

What is the difference between allergen cleaning validation and verification?

Validation proves in advance that a cleaning procedure removes a specific allergen from a specific line to an agreed level. Verification checks, at each changeover, that the validated procedure was followed and still works.

Can ATP swabs be used to validate an allergen clean?

No. ATP and total protein swabs are not specific to allergens, and a surface can pass an ATP test while still carrying allergen. Use an allergen specific method such as ELISA for validation.

How many runs does an allergen cleaning validation need?

FoodDrinkEurope’s guidance asks for the acceptance criteria to be met on three consecutive runs. A failure means finding the cause, correcting the procedure and starting the three runs again.

Does a validated clean mean I can remove a “may contain” statement?

It gives you the evidence to decide, together with a risk assessment of every other route the allergen could take. R146 requires the risk, how it was assessed and the steps taken to be documented wherever precautionary labelling is used, and the Codex annex of July 2026 links the decision to the action level.

Leave a Comment

I accept the Terms and Conditions and the Privacy Policy

Exemplar Global Recognised Training Provider (RTP) badge issued to ASC Food Safety
International recognitionExemplar Global Recognised Training Provider (RTP)Exemplar Global, the international organisation that certifies auditors and recognises training providers, recognises 28 of our auditor, HACCP, GMP and ISO courses, including four lead auditor courses, whether we deliver them in class, in live virtual classes or online.
News & updates 5 new
4.9/5 what do you need today?