Theory and Basic Criteria for Practical Sampling

1. Introduction to "Good Sampling Practice"

Mycotoxins are often distributed in food commodities in a very heterogeneous manner, because they are formed by moulds that occur either in isolated pockets in bulk materials, or in individual seeds or nuts. When checks are carried out to ensure compliance with statutory or guideline maximum limits, it is vital that the analytical result is a true value for the consignment. If this is not achieved, satisfactory consignments may be rejected or contaminated loads accepted. In addition to health and safety concerns, this can have serious financial implications for the trader and may lead to costly legal disputes.

The importance of sampling on the analytical result cannot be underestimated. As far as possible, tested and validated sampling plans that provide a final sample that is truly representative must be used, because of the potentially enormous commercial impact. This fact sheet demonstrates the importance of sampling from a practical perspective. Provision of mathematical or statistical background has been avoided.

Factors required to achieve "representative sampling" are summarized below.

- Samples should be taken by qualified, suitably trained, and motivated personnel, who understand the importance of sampling for obtaining an accurate test result.

- Appropriate, clean and well-maintained tools should be used. Adverse conditions such as poor weather, environment and factors such as cross-contamination should be avoided.

- Incremental samples have to be drawn from various spots randomly distributed throughout the entire lot or sub-lot. Any unplanned variations from this procedure should be recorded.

- Samples should be taken when all parts of the lot are easily accessible, which is normally the case when samples are moved or during discharge. Any part of a lot must have an equal chance of being selected.

- Transportation and storage facilities for samples must be adequate, and storage times must be as short as possible.

- Good sampling practice necessarily requires clear and secure labeling of samples, giving essential information such as name of sampler, place of sampling, lot number, conditions, applied tools, time, date and observations.

2. Definitions Relating to Sampling Plans

The European Union lays down sampling protocols for some mycotoxins in a range of products for statutory purposes. However, it is recognised that these are demanding procedures and the protocols include advice that alternative sampling protocols may be satisfactory for other purposes.

The following terms are used (Definitions are based on Commission Directive 98/53/EC dated July 16, 1998):

Lot: An identifiable quantity of a food commodity delivered at one time and determined by the official to have common characteristics, such as origin, variety, type of packing, packer, consignor or markings.

Sub-lot: Designated part of a large lot in order to apply the sampling method on that designated part. Each sub-lot must be physically separated and identifiable.

Incremental sample: A quantity of material taken from a single place in the lot or in the sub-lot.

Aggregate sample: The combined total of all incremental samples taken from the lot or from the sub-lot.

Laboratory sample: Sample intended to be used for the laboratory (= sub-sample).

Each single lot has to be sampled separately. Large lots should be sub-divided into sub-lot (if possible; - see definition for sub-lot). Aggregate samples result from united incremental samples, which are mixed sufficiently. After mixing of incremental samples it is important to ensure that the resulting aggregate sample contains portions of the entire lot or sub-lot.

3. The Problem of ‘Representative Sampling’ for Mycotoxins

A sample should be representative of the lot from which it is taken. ‘100% representative’ means that the analytical result obtained for a sample gives the true value of a parameter in a given lot. If the composition of a lot is absolutely uniform, any small sample could be regarded as a representative sample. However, a lot is not normally homogeneous, so that the distribution of a compound in a lot is uneven and concentrations vary from spot to spot over a broad range of values.

This can be demonstrated for aflatoxin if compared with the situation for protein in a "theoretical" lot of peanuts. The entire lot consists of 100 packets. Table 1 shows the amount of protein and of aflatoxin B₁ in each packet.

Table 1: Sample Lot (100 packets) with given values for protein and aflatoxin B₁ content in each packet

The difference in the situation for protein and aflatoxin B₁ content is clear with relative standard deviation values of only 3% for protein, but 636% for aflatoxin B₁. Any packet in the lot would give an acceptable representative value for the protein content, but for aflatoxin B₁, a representative value would only be obtained if the entire lot is analyzed as one sample.

The following questions then arise about sampling for aflatoxin B₁:

- How many incremental samples are necessary to give an acceptable representative sample for the entire lot?

- How does the number of samples influence the quality of analytical results?

3.1. Sampling variation

The influence of sampling on the analytical result is demonstrated using the "theoretical" peanut sample presented above.

With the values for each carton covered, individual samplers randomly marked 10 cartons resulting in a variety of patterns(see Fig 1 below)

Fig. 1 Sampling patterns as obtained from individual samplers

Three samplers repeated the random marking on ten occasions and the average of each set of 10 packets was calculated. The results are summarized in table 2.

Table 2: Calculated results for 10 sample sets (10 incremental samples each)

The content of aflatoxin B₁ calculated varied widely so that none of the samplings determined the true value of aflatoxin. In addition, nearly 50% of the samplings would have resulted in the lot being unjustifiably rejected due to high aflatoxin B₁ content (shaded data).

3.2. Influence of the number of incremental samples

In a second experiment the number of incremental samples taken per set and per sampler was doubled. The results are summarized in table 3.

Table 3: Calculated results for 12 sample sets (20 incremental samples each)

It is clear that doubling the number of incremental samples has a dramatic effect on the results. The aflatoxin B₁ content determined is much closer to the true value. In addition, the number of refusals due to high aflatoxin content dropped to about 25%.

4. Conclusions

A sufficient number of incremental samples is vital to obtain a representative sample. The greater the number of incremental samples, the closer the result will be to the true value. It is difficult to estimate the optimum number of incremental samples that enable a general recommendation to be made because it depends on the distribution and concentration of the contaminant, the size of the lot, the nature of goods and the value of goods under test. Thus sampling protocols must be a compromise between the cost and labour required to take the samples and the variability expected in the result.

The effort justified in acquiring a representative sample must also take account of the measurement uncertainty of the analytical method used for the sample. The sampling requirements for an analytical method with measurement uncertainty of 50% are less than for a method with 10% uncertainty.

Representative sampling is a time consuming procedure with considerable costs. However, it is the basis of a reliable test result and must be achieved. Better sampling will reduce the chance that a consignment meeting legislation will be unnecessarily rejected.

There remains a need for more experimental projects to develop robust sampling plans and protocols and to provide trade with clear and reasonable guidelines.

Suggested Bibliography:

[1] Schatzki, T.F. (2000), "Distribution of aflatoxin in pistachios. 7. Sequential sampling", Journal of Agricultural and Food Chemistry 48 (9) 4365 – 4368.

[2] Vandeven, M., Whitaker, T., Slate, A. (2002) "Statistical approach for risk assessment of aflatoxin sampling plan used by manufacturers for raw shelled peanuts", J. AOAC International 85 (4) 925 – 932

[3] Whitaker, T.B. (2003), "Standardisation of Mycotoxin Sampling Procedures: an urgent necessity" Food Control 14 (4) 233 – 237.

[4] Whitaker, T.B., Hagler, W.M., Johansson, A.S., Giesbrecht, F.G., Trucksess, M.W. (2002) "Sampling shelled corn for fumonisin" Mycotoxins and Phycotoxins in perspective at the turn of the millennium: proceedings of the 10^th International Symposium on Mycotoxins and Phycotoxins, Guaruja, May 2000 de Koe W.J., Samson R.A., van Egmond, H.P., Gilbert, J., Sabino, M., IUPAC Wageningen W.J. de Koe 2002 97 – 107.

[5] International Organization for Standardization (1999) "Cereals, pulses and milled products. Sampling of static batches. ISO 13690:1999 (E).

[6] Schatzki, T.F., deKoe, W.J. (1999) "Distribution of aflatoxins in pistachios: seller’s and buyer’s risk" Journal of Agricultural and Food Chemistry 47 (9) 3771 – 3775.

[7] Ministry of Agriculture, Fisheries and Food, RHM Technology, Burke, S., Smith, R., Kemsley, K. London MAFF, MAFF R&D and Surveillance Report No. 500.