MYCOTOXINS

The Peanut Aflatoxin Code Rewritten: CXC 55-2004 After Twenty Years, and What South African Groundnut Handlers Must Change

By Mthokozisi Nkosi, Food Safety Specialist & Lead Auditor, ASC Food Safety · 15 min read

CXC 55-2004, the Codex Code of Practice for the Prevention and Reduction of Aflatoxin Contamination in Peanuts, was revised at CAC48 in Rome, 10 to 14 November 2025, its first revision in over twenty years. It added a peanut growth stage table for harvest timing, a scope extension to feed made from peanut by-products, and a section on roasting. The Chemical Contamination Risk Assessment Template (RA16) is R690.

What is CXC 55-2004 and what did the 2025 revision change?

CXC 55-2004 is the Codex Code of Practice for the Prevention and Reduction of Aflatoxin Contamination in Peanuts. It was revised at CAC48, the 48th Session of the Codex Alimentarius Commission, in Rome, 10 to 14 November 2025. That was its first revision in over twenty years. It added a peanut growth stage table for harvest timing, scope extended to feed made from peanut by-products, and a section on roasting.

Those three additions are the whole of what changed, and I am going to be careful about that boundary for the rest of this article. A code of practice is not a maximum level and it is not a specification. It is Codex setting out, for governments and for industry, the practices that keep a contaminant out of a commodity in the first place. Maximum levels for contaminants live in a different text, CXS 193-1995, the General Standard for Contaminants and Toxins in Food and Feed, and the legal limit that binds a South African consignment lives in South African law and in your customer’s specification.

The revision is worth your attention for a reason that has nothing to do with legal force. A code of practice that has stood untouched for two decades is a document that quietly stopped being read. When Codex reopens one, buyers, certification bodies and importing authorities read the new version, and questions that were never asked before start arriving in supplier questionnaires. If you shell, roast, blanch, crush, blend or import groundnuts in South Africa, expect the growth stage table and the by-product feed scope to appear in a customer document within the next twelve months.

The other correction worth making while you are here is about the general hygiene text. CXC 1-1969, General Principles of Food Hygiene, did not change. The published text still reads adopted 1969, amended 1999, revised 1997, 2003, 2020, 2022, with editorial corrections in 2011. A revision is in progress at the Committee on Food Hygiene and has not been adopted. There is no 2026 edition, whatever your search results tell you. Our full list of every Codex change from CAC48 and CAC49 sets out the whole year in one place.

How Aspergillus flavus gets into a groundnut crop

Aflatoxins are produced by Aspergillus flavus and Aspergillus parasiticus, soil fungi that are already present in most warm agricultural soils, including South African groundnut soils. They are not an infection that arrives from outside. The question is never whether the organism is in the field. It is whether the crop gives it the conditions and the entry route it needs to colonise the kernel and produce toxin.

That distinction changes how you think about control. There is no point in a supplier questionnaire asking whether a grower has ever had Aspergillus in the soil. The useful questions are about water, temperature, pod integrity and time, because those are the four variables the grower can influence and the four that decide whether a fungus that is always present becomes a fungus that has produced toxin.

Aflatoxin B1 is the member of the group that drives the risk assessment. It is a genotoxic carcinogen, which is why regulators everywhere set limits at the lowest level that is reasonably achievable rather than at a level derived from a tolerable daily intake, and why there is no safe quantity you can trade against a commercial pressure. That is also why the discipline in this article is prevention and rejection rather than treatment. There is no processing step you can bolt on that makes the problem go away.

The fungus is always in the soil. What you control is whether the crop hands it an open door and enough time to walk through it.

Why harvest timing decides your aflatoxin result

Harvest timing matters because drought stress before lifting, and delayed lifting after the crop is mature, are the two conditions that let Aspergillus flavus colonise the kernel. A moisture stressed plant loses the physiological resistance that keeps the fungus out, and a mature crop left in hot dry soil keeps accumulating toxin every day it stays there. The 2025 revision added a growth stage table for this decision.

Take the mechanism in order, because it is worth understanding rather than memorising. A groundnut fruits underground. The pod develops in the geocarposphere, the narrow zone of soil around the pegs and pods, and that zone is where the contest happens. While the plant has water, the pod shell stays intact, the kernel fills, and the plant’s own defences hold. When water runs short in the last weeks of pod fill, several things happen at once. Soil temperature in the pod zone rises. The pod shell dries, becomes brittle and cracks. Kernels shrink away from the shell and the testa loses integrity. The plant’s capacity to resist invasion falls off. Aspergillus flavus tolerates low water availability better than most of its soil competitors, so the same conditions that weaken the plant also remove the fungus’s competition.

Delayed lifting adds the fourth variable. A crop that has reached maturity is no longer building anything. If it stays in warm dry soil past that point, the pods continue to weather, the shells continue to break down, and any colonisation already under way continues to produce toxin. Lifting too early has its own cost in yield and in immature kernels, which is precisely why a growth stage table is the practical tool. The decision is not early or late in the abstract. It is whether this crop, at this stage, in these soil conditions, is better in the ground or out of it.

For a South African buyer this reframes what you should be asking a groundnut supplier. Not “do you test”, which everybody says yes to. Ask what the lifting decision was based on, whether the crop went through moisture stress in the final weeks, whether irrigation was available, how long the windrow sat and what the weather did while it did. Those answers tell you what your test result is likely to look like before the sample is drawn, and they belong on the supplier file rather than in a phone call nobody wrote down.

Drying, storage and transport: where a clean crop goes wrong after lifting

A crop that comes out of the ground with acceptable aflatoxin can still fail at sale, because Aspergillus flavus keeps growing wherever moisture, temperature and time allow. Slow drying, rain on a windrow, kernels binned above the agreed moisture, condensation in a container and moisture migration in a bulk store are all post-harvest opportunities. The toxin that appears in a warehouse was not necessarily made in a field.

The controlling variable after lifting is water activity rather than headline moisture percentage, because moulds respond to available water and not to a bag average. That has three practical consequences. Drying has to be quick enough that the crop does not spend days at intermediate moisture, which is the worst possible condition because it is wet enough for growth and dry enough that nobody is worried. Drying has to be even, because a bin that averages a safe moisture can still hold a wet core. And storage has to hold that condition, which means managing temperature gradients, ventilation, condensation and the length of time a lot sits.

I am deliberately not printing a target moisture percentage or a water activity figure in this article. The number that governs your intake decision must come from your specification, the standard your customer names, or the commodity requirement that applies to your market, and it must be written down in your specification with its source. A number lifted from an article and typed into an intake procedure is a finding waiting to happen, and it is also how sites end up rejecting perfectly good crop or accepting crop they should not.

Peanut by-products in animal feed, and why a dairy cares what a crusher does

The 2025 revision extended the scope of CXC 55-2004 to feed made from peanut by-products. That matters commercially because the by-product stream is where the contamination concentrates. Press cake, screenings, sorted out kernels, hulls and shelling fines are the fractions a human food operation deliberately removes, and they are the fractions that historically go to animal feed. The sorting that protects one supply chain loads the other.

This is the point in the article South African groundnut businesses should read twice. Colour sorting, density separation and hand picking work because aflatoxin is concentrated in discoloured, shrivelled, immature, insect damaged and broken kernels. Removing that fraction genuinely lowers the toxin in what remains. But the removed fraction has not been destroyed. It has been collected, and if it leaves the site as a feed ingredient without its own specification, sampling and rejection rule, the risk has been sold rather than controlled.

Aflatoxin B1 eaten by a lactating animal is metabolised and a proportion is excreted in milk as aflatoxin M1. That carry over is the reason the Codex code of practice now reaches feed made from peanut by-products, and it is the reason a dairy processor two or three steps down the chain has a legitimate interest in what a groundnut crusher does with its press cake. The dairy cannot remove M1 from milk. Pasteurisation does not destroy it. Blending across a milk intake spreads it rather than removing it. The dairy’s only real control is upstream, in the feed specification and in the approval of whoever supplies the feed mill.

There is a certification dimension to this as well. Feed and animal food is a distinct food chain category under FSSC 22000 Version 7, published in May 2026, which builds on ISO 22000:2018 plus a sector prerequisite programme standard from the ISO 22002 series plus the FSSC additional requirements. The series was restructured in July 2025, with most parts moving from Technical Specification to full International Standard, and ISO 22002-6:2025 is the feed and animal food part. If your business sits on both sides of the line, selling food grade kernels and feed grade by-products, you are running two chains under two sets of prerequisite requirements from one site.

Does roasting destroy aflatoxin in peanuts?

Roasting reduces aflatoxin. It does not eliminate it, and it is not a control measure you can validate as a kill step the way you validate heat against a pathogen. The 2025 revision added a section on the part roasting plays in reducing aflatoxin, and the operative word is part. Roasting is one contribution to a lot that was already acceptable. It is not a rescue treatment.

The reason the analogy with a thermal kill step breaks down is worth setting out properly, because this is where I see commercially motivated thinking creep into technical documents. When you validate a heat process against a vegetative pathogen, you know the organism, you know its thermal resistance under defined conditions, you can measure the delivered process, and you can express the outcome as a reduction from a defined starting load. Every one of those conditions fails for aflatoxin.

• Aflatoxin is a chemical, not an organism. It is heat stable enough to survive normal roasting conditions in substantial part. Killing the fungus does not remove the toxin the fungus already made
• The starting level is unknown and heterogeneous. You cannot express a reduction from a number you never had, and the sampling problem means you probably never had it
• The extent of degradation varies with temperature, time, moisture and the initial level, so the same roaster running the same programme on two lots does not deliver the same proportional reduction
• The reaction products of partial thermal degradation are not fully characterised, so a reduction in the measured parent compound is not automatically a proportional reduction in risk
• There is no meaningful critical limit to monitor. A time and temperature that produce the product you want are not a time and temperature that produce a defined toxin outcome

What follows from that is a specific instruction. Do not write roasting into your HACCP plan as the control measure for aflatoxin, and do not tell a customer that a roasted product is compliant because it was roasted. The control measure for aflatoxin is raw material control: specification, supplier approval, sampling at intake, sorting, and rejection. For aflatoxin, roasting sits alongside the process controls as a contributor rather than as a control measure, and it is honest to record it as one.

Read that sentence narrowly, because it says nothing about roasting’s status for other hazards on the same line. Roasting a low moisture product is very often a validated critical control point for Salmonella, with a defined thermal process, measured critical limits and validation behind it. Keep that critical control point exactly where it is, with its limits and its monitoring intact. What you may not do is let a hazard the roast was validated against carry a hazard it was not.

Why the sampling plan, not the laboratory, decides your aflatoxin result

Aflatoxin is famously heterogeneous. It concentrates in a very small proportion of the kernels in a lot, so two honest samples drawn from the same consignment can return very different results without anybody doing anything wrong. The uncertainty introduced by drawing and preparing the sample is usually far larger than the uncertainty of the analytical method. With this contaminant, the sampling plan is the whole game.

The practical consequence is that a certificate of analysis quoting a low number tells you almost nothing unless you know how the sample behind it was taken. A single grab sample from the top of a container, comminuted badly and analysed by an excellent laboratory, produces a precise number about a sample that does not represent the lot. A well designed plan with many incremental samples drawn across the lot, combined into a large aggregate sample, ground to a fine particle size and subsampled properly, produces a less flattering number that actually describes what you bought.

1. Define the lotA sampling plan applies to a defined quantity of product of one origin, one variety, one harvest and one consignment. If your lot definition is “whatever came off the truck today”, every result that follows describes something you cannot trace.
2. Take many increments, across the lotIncremental samples must be drawn from throughout the lot, not from where the sampler can reach. This is the step that gets cut when a truck is waiting, and it is the step that matters most.
3. Build a large aggregate sampleThe aggregate sample has to be big enough to have a realistic chance of containing the rare contaminated kernels. A small aggregate sample is a lottery ticket, not a measurement.
4. Comminute and subsample properlyGrinding the aggregate sample to a fine, uniform particle size before subsampling is what converts a few hot kernels into a signal the test portion can see. Poor comminution undoes good sampling.
5. Use a named method and a competent laboratoryRecord the method, the laboratory, the accreditation status and the reported measurement uncertainty. CXS 234-1999, Recommended Methods of Analysis and Sampling, was restructured and republished with a cover reading amended 2026, and it is the Codex reference to work from rather than a plan you invented.
6. Write the accept and reject rule before you testDecide in advance what result triggers acceptance, what triggers rejection, and what triggers resampling. A rule written after an inconvenient result is not a rule.

Two more things belong in the procedure. First, define who may authorise a resample and on what grounds, because “test it again” is the most common way a failing lot becomes a passing lot. Second, keep the retained sample. When a customer or an authority disputes a result months later, the retained sample from a documented plan is the only thing that lets you defend or revise your own conclusion.

What is the maximum level for aflatoxin, and where do I find it?

Establishing the limit is the first thing to do, and it is not something to take from an article, including this one. The figure that binds your consignment depends on the commodity, its intended use, whether it goes for further processing or direct consumption, and the market you sell into. ASC will confirm the position for your commodity in writing and cite the instrument it comes from.

I am stating that plainly rather than printing a number because of how these numbers behave. Maximum levels differ between the commodity as harvested and the commodity ready for consumption. They differ between total aflatoxins and aflatoxin B1. They differ between food and feed. They differ between South African law, the Codex general standard and the requirements of the market you are exporting into, and your customer’s specification is frequently tighter than any of them. A single figure quoted without those qualifiers is how a site ends up rejecting saleable crop or shipping product it cannot defend.

What you should do is straightforward. Establish the applicable legal limit in South Africa for your exact commodity and use. Establish the limit in the market you export to. Establish your customer’s contractual limit. Take the tightest of the three as your operating specification, record which instrument each came from, and put a review date on it. The ASC South African legislation register indexes 413 instruments across 42 parent Acts and is where to start looking, and our South African food legislation guide explains how the parent Acts fit together.

What your groundnut specification must say about aflatoxin

A groundnut raw material specification that names a mycotoxin limit and stops there is not a specification, it is a wish. The specification has to say what the limit is, which instrument it came from, how the material will be sampled, what the accept and reject rules are, what documentation must accompany each consignment, and what the supplier must tell you when something changes.

• Aflatoxin parameter stated as total aflatoxins and as aflatoxin B1 separately, each with its source instrument named next to the figure and a review date on the specification
• Origin, variety, harvest season and lot definition, because a specification that cannot define a lot cannot define a sample
• Moisture or water activity at delivery, taken from the standard or specification that governs your commodity rather than from a general article
• The sampling plan that will be applied at intake, named in the specification so the supplier knows the basis on which their consignment will be judged
• Documentation required with each consignment: certificate of analysis against a named method, the sampling basis behind it, batch identity and traceability back to the grower or the shelling lot
• Change notification obligations covering origin, grower base, drying or storage arrangements, sorting equipment and any change of the site that processes the material
• Damaged, discoloured and immature kernel tolerances, because that fraction carries the contamination and is the fraction sorting removes
• Where by-products are sold on, the specification, sampling and batch identity that goes out with the by-product stream
• Allergen declaration in line with regulation 43 of R146 of 2010, which lists peanuts among the allergens that must be declared on a South African label

On the supplier side, groundnuts belong in your highest risk tier and the evidence should match. Clause 3.5.1 is one of the twelve fundamental requirements of BRCGS Global Standard Food Safety Issue 9, published 1 August 2022, and a failure against a fundamental ends the audit. If you are certified to a GFSI recognised scheme and buying a commodity with a known mycotoxin risk, the auditor will open that file first. Our BRCGS Issue 9 implementation guide covers the wider system, and importing food into South Africa covers the documentation side for imported consignments.

What to do with a groundnut load that fails an aflatoxin maximum level

Decide the disposition rules before a load fails, because the pressure to be creative arrives with the failure. A failing consignment goes to a defined quarantine location, with a named person authorised to release it and a written list of permitted dispositions: reject and return, destroy, or divert to a lawful, documented use for that level. Roasting it, blending down or resampling until it passes are not on that list.

Segregation is the part sites get wrong physically rather than procedurally. Held material needs a location that is not the same bay as released material with a sign on it. Groundnuts move as bulk, in totes and in bags, and a held lot sitting next to released stock in the same aisle is a mix waiting to happen on the busiest day of the week. Give hold status a physical location, a system status and a label, and audit that the three agree.

1. Hold on the first resultPlace the lot on hold on the first out of specification result rather than waiting for a confirmatory test. If the confirmatory test clears it you release it, having lost nothing.
2. Decide whether a resample is legitimateA resample is legitimate where the original sampling deviated from the plan. It is not legitimate because the result was inconvenient. Record the technical justification, and have it authorised by someone who does not carry the commercial consequence.
3. Trace forward and backwardEstablish what else came from that lot, that grower, that store or that container, and whether any of it has already been used or dispatched. One failing lot is rarely a lone event.
4. Decide the disposition against the written listReject and return, destroy, or divert only where that use is lawful and documented for that contamination level. Record who decided, on what basis and on what date.
5. NotifyTell the supplier, and tell your customer where product may already have moved. Where an imported consignment is involved, remember that Port Health inspects imported foodstuffs at points of entry and can detain a consignment.
6. Do the root cause workA failure that ends at rejection teaches you nothing. Take it back to growing, lifting, drying, storage or transport, and change the supplier agreement or the intake rule accordingly.

The commercial argument against all of this is always the same, that the load is worth a great deal and the customer is waiting. Section 55 of the Consumer Protection Act 68 of 2008 gives the consumer a right to safe, good quality goods, and section 61 imposes liability on producers, importers, distributors and retailers for harm caused by unsafe goods, irrespective of negligence. The value of the load is not a defence, and it is not a number that appears anywhere in that section.

Where the aflatoxin maximum levels are actually set

CXC 55-2004 tells you how to keep aflatoxin out. CXS 193-1995, the General Standard for Contaminants and Toxins in Food and Feed, is where Codex sets maximum levels and the principles behind them. Its cover reads adopted in 1995, last amended in 2025, because CAC48 added lead maximum levels of 2.5 mg/kg for the commodity Codex calls spices, dried bark and 2.0 mg/kg for culinary herbs, dried, and removed outdated radionuclide guidance.

Those two texts work as a pair, and reading only one of them is how sites end up with a limit and no method of achieving it, or a set of good practices and no target. If you handle spices or dried herbs alongside groundnuts, which many South African dry commodity businesses do, the lead amendment is the other half of the same year’s work and we have set it out in full in the new Codex lead limits for spices and herbs.

Two further reference points are worth knowing. CXS 200-1995, the Standard for Peanuts, is the Codex commodity standard for the product itself and was last modified in 2026. And CXS 234-1999, Recommended Methods of Analysis and Sampling, was restructured and republished with a cover reading amended 2026. Between the code of practice, the commodity standard, the contaminants standard and the methods and sampling standard, Codex has covered the ground, and all four are free downloads from FAO. Nobody should be selling them to you.

Frequently asked questions

Published by ASC Food Safety, South African food safety and quality consultants. This article is general guidance and not a substitute for certification-specific advice.