The hardest thing to explain about aflatoxin in tree nuts is that it is not distributed. In a twenty tonne container of walnut kernels that fails a limit, the contamination is very often carried by a few dozen individual kernels. Everything else in the load is clean. That single fact drives the whole control system: it is why the sampling rules are written into law alongside the limits, why a small grab sample proves almost nothing, and why two honest laboratories can return contradictory results on the same lot without either of them making a mistake.
Buyers who understand this stop asking whether the supplier tests and start asking how the supplier samples. That is the shift this guide is written to produce.
Where the toxin comes from
Aflatoxins are secondary metabolites of Aspergillus flavus and Aspergillus parasiticus, moulds that are common in soil and orchard debris across warm growing regions. The mould itself is not the problem; toxin production requires a combination of a susceptible substrate, warmth and available water, and it is favoured by stress in the field and by any interruption in drying after harvest.
Four conditions repeatedly produce contaminated lots.
Drought and heat stress in the orchard. A stressed tree produces kernels with more shell splits and insect damage, and both are entry routes. Contamination in tree nuts is frequently a field event that processing then has to remove rather than a storage event that processing caused.
Insect damage. Navel orangeworm in almonds is the textbook case. The insect breaches the shell, the mould follows, and the resulting kernel is often visually distinguishable, which is what makes sorting effective at all.
Delayed or interrupted drying. Nuts arriving from the field carry field moisture. Every hour before drying reaches a stable kernel moisture is an hour in which the mould can grow. Rain during a windrow harvest is the classic trigger.
Rewetting in store. Nuts held above the moisture and water activity range at which mould growth stops, or exposed to condensation from an uncontrolled temperature swing in a container, will support growth after the drying step did its job. Aflatoxin does not go away once formed; it is heat stable through ordinary food processing.
The commercially important consequence is that toxin already present cannot be destroyed. It can only be removed with the kernels that carry it.
The instruments that govern this
Three separate pieces of Union law apply, and they are commonly confused with each other.
Maximum levels for aflatoxin B1 and for the sum of aflatoxins in tree nuts are set by the Union contaminants instrument, Commission Regulation (EU) 2023/915, which replaced Regulation (EC) No 1881/2006. The structure matters as much as the numbers: the regulation sets different lines for nuts placed on the market for the final consumer and for nuts intended to be sorted or otherwise physically treated before consumption. A lot that is legal as a raw material for a sorting operation is not automatically legal at retail, and a contract that does not state which line applies has left the most important term undefined.
Sampling and analysis are governed separately, by Commission Implementing Regulation (EU) 2023/2782, which replaced Regulation (EC) No 401/2006 and applies from April 2024. This instrument sets the aggregate sample size, the number of incremental samples by lot size, the sublot structure for large lots and the performance criteria the analytical method has to meet. Its transitional arrangement allows methods validated under the previous rules to continue in use for a defined period, so a certificate may legitimately cite either instrument during the changeover; what it may not do is stay silent about which sampling scheme was followed.
Increased official controls at the border are set by Commission Implementing Regulation (EU) 2019/1793, which lists particular products and origins for a raised frequency of identity and physical checks, and in some cases for an accompanying certificate and analytical report. The list is amended regularly. It is not a quality judgement on a supplier; it is an origin-level frequency setting, and a buyer importing from a listed origin should plan for the check and its cost rather than be surprised by it.
Alongside the law, Codex code of practice CXC 59-2005 sets out prevention and reduction measures for aflatoxin in tree nuts, and Codex standard CXS 193-1995 carries the international maximum levels. Neither is binding in the Union, but both are useful when a supplier outside the EU needs the control logic explained in its own vocabulary.
Read the limits in the consolidated text of the instrument on the day of shipment. Contaminant lines and control lists are amended frequently and a number copied into a specification two years ago is a liability.
Why sampling dominates
If contamination sits in a few kernels per tonne, then the probability of catching it depends almost entirely on how many kernels the sample contains and how well they represent the lot. This is why the sampling regulation prescribes large aggregate samples built from many incremental samples taken across the lot, and why it requires the aggregate to be comminuted and mixed before a test portion is drawn.
Three practical consequences follow.
- A grab sample from the container door is worthless. It is drawn from a single point, it is small, and it systematically under-detects. A certificate based on one is not evidence.
- Sample preparation carries as much variance as sampling. A coarse grind leaves the toxin concentrated in a few particles and the test portion may miss them. The regulation addresses this with slurry and fine-grind requirements, and a laboratory that cuts this corner produces low results reliably.
- Re-testing a failed lot with a fresh sample is not fraud detection, it is expected variance. Two compliant samples from the same lot can differ several-fold. This is why dispute resolution has to be agreed in the contract in advance, naming who samples, who witnesses and which laboratory is final.
What sorting actually achieves
Removal is the only intervention that works, and it works well when it is engineered properly.
| Step | What it removes | Practical limits |
|---|---|---|
| Visual and manual sorting | Obviously mouldy, discoloured, insect damaged and shrivelled kernels | Labour intensive, operator dependent, effective on gross defects only |
| Colour sorting | Kernels outside the accepted colour range, including many discoloured by mould | Cannot see internal defects; a clean-looking contaminated kernel passes |
| Multispectral optical sorting | The above plus defects visible outside the human range | Higher capital cost, requires calibration per variety and per crop year |
| X-ray inspection | Density anomalies, internal insect damage, shell fragments | Slower throughput, and it detects damage rather than the toxin itself |
| Hand pick line after sorting | Residual defects the machines released | The last few per cent, and the most expensive per kilogram |
No sorting line removes aflatoxin directly. Every one of them removes kernels that are statistically likely to carry it. That distinction is why an aggressive sorting programme reduces average contamination substantially but never guarantees a result, and why testing after sorting remains necessary.
The commercial corollary is that sorting intensity shows up in price and in yield. A supplier quoting a low price on a lot with a tight aflatoxin clause is either absorbing a sorting cost that the price does not support, or has not priced the rejects. Ask what the sorting reject rate is; the answer tells you more about the operation than the certificate does.
Testing routes and what they are for
ELISA is fast, inexpensive and well suited to screening at intake. It is quantitative within its working range but is subject to matrix effects and cross-reactivity, and a positive should be confirmed.
HPLC with fluorescence detection, usually after immunoaffinity clean-up, is the long-established confirmatory route and is what most trade certificates rest on.
LC-MS/MS gives simultaneous determination of several mycotoxins including ochratoxin A, better specificity and lower detection limits, at a higher cost per sample. Where a buyer runs a multi-mycotoxin panel, this is the sensible route.
Whichever is used, the method has to meet the performance criteria in the sampling and analysis regulation, and the certificate of analysis has to state the method, the limit of quantification, the measurement uncertainty and the sampling scheme applied. A certificate reporting only “not detected” without a limit of quantification is not usable, because a high detection limit can make a non-compliant lot look clean.
Writing a clause that holds
A workable aflatoxin clause contains six elements.
- Which regulatory line applies: product for the final consumer, or product to be sorted or physically treated before consumption. State it explicitly.
- The sampling scheme, by reference to Commission Implementing Regulation (EU) 2023/2782, with the party responsible for drawing the sample named.
- The analytical method and its limit of quantification, with measurement uncertainty reported and the rule for applying it agreed in advance.
- Testing frequency: every lot, or a defined skip-lot scheme with the trigger for reverting to every lot on a single non-conforming result.
- Dispute resolution: retained samples, an agreed referee laboratory, and who pays under which outcome.
- Consequence: rejection, re-sorting at supplier cost, or downgrade to a non-food channel, stated per outcome rather than left to negotiation after a failure.
What to check at intake
At goods-in, incoming inspection on a nut lot should cover the certificate against the six clause elements above, moisture and water activity on a fresh sample, a visual defect assessment against the agreed defect classification, container condition and evidence of condensation, and pack integrity. Retain a sample from every lot; without it the dispute clause is decorative.
The EU aflatoxin limits for nuts and how aflatoxin testing works answer the two questions this guide raises most often, and allergen management in mixed facilities covers the other control programme that a nut handling site has to run in parallel.