Aflatoxin testing has three stages: draw a large aggregate sample from many points across the lot under the scheme in Commission Implementing Regulation (EU) 2023/2782, comminute and homogenise it to a fine slurry or powder, then analyse a test portion by ELISA for screening and by HPLC with fluorescence detection or LC-MS/MS for confirmation.
Why it works this way
Contamination is extremely unevenly distributed. In a failing lot the toxin is typically carried by a handful of kernels per tonne while everything around them is clean. That makes the variance of the sampling step far larger than the variance of the analysis, which is why the law prescribes the sampling in detail and leaves the choice of method to performance criteria.
Three practical points follow from the distribution.
Sample size and sample points are not negotiable. The regulation sets the number of incremental samples by lot weight, the aggregate sample mass and the sublot structure for large consignments. A grab sample from the container door under-detects systematically.
Sample preparation carries real variance too. A coarse grind leaves the toxin concentrated in a few particles that the test portion may miss, which is why slurry preparation is specified.
Two compliant results can differ several-fold on the same lot. That is expected variance, not laboratory error, and it is why dispute resolution has to be agreed before shipment.
What follows in practice
Require the certificate of analysis to state the sampling scheme, the method, the limit of quantification and the measurement uncertainty. A result reported only as “not detected” is unusable without a limit of quantification behind it. Aflatoxin management in tree nuts sets out the full programme and the EU limits explain the structure of the maximum levels.