PMIL Mycotoxin Webinar: Part 2 Representative Sampling for Mycotoxins - YouTube

Hello my name is Andrew Slate.

This is part 2 of a 3 part presentation on sampling for mycotoxins.

Representative sampling is critical for valid mycotoxin testing.

Everyone involved with mycotoxin testing should understand representative sampling.

First of all lets look at some of the definitions of terms considered in representative sampling.

These definitions can be found at the Codex website and in the appendix at the end of this slide set.

First of all, what is a sampling plan?

How do you define a lot?

Do you need to define sublots?

What is an incremental sample?

What is an aggregate sample?

What is a laboratory sample?

And what is a test portion?

This is the diagram of the critical steps of sampling protocol.

From sample selection, to sample preparation and analysis.

Lets start with the lot.

Once you have defined your lot,

the first part of the sample selection process,

is to take many incremental samples from many locations in the lot.

Combine these incremental samples to form the aggregate sample.

The aggregate sample size is dependent on the lot size.

Large lots yield large aggregate samples, small lots yield small aggregate samples.

The aggregate sample should be at least as large as the required laboratory sample.

Using an approved divider the aggregate sample can be divided to produce the laboratory sample.

The laboratory sample size is specified by the sampling protocol, and is independent of lot size.

If the protocol calls for a 2 kilogram sample, then the laboratory sample should be 2 kilograms for any size lot.

The entire laboratory sample should then be comminuted in an approved grinder,

and ground as finely as possible.

And then from that ground material, the test portion should be selected.

Using again, an approved division method.

The test portion is then extracted and analyzed for the mycotoxin in question.

The resulting concentration should be compared to the maximum level for accept-reject of the lot.

The steps shown here produce a representative laboratory sample,

a representative test portion,

and a representative analysis of the test portion.

If all of these steps are executed properly, and representative sample test portion and analysis is done,

Then we have a valid sampling protocol.

However, the natural variability is not eliminated by representative sampling

And we know that most of the variability comes from sampling and sample preparation.

A small portion of the variability is due to the analytical step.

So the first part of the process is sample selection.

In summary, sampling is generally a large source of testing variability.

Variability statistics are only valid for representative samples.

Critical sampling components are the sample section and the laboratory sample size.

Again, aggregate sample size is dependent on the lot size.

However, the laboratory sample size is independent of lot size.

So for a representative sample,

every kernel in the lot must have an equal chance of being chosen for the sample.

This is accomplished by selecting many incremental samples, throughout many locations in the lot.

This selection process must be random.

And, even if the representative sample is selected, the inherent sampling variability is not eliminated.

When we select our sample, there are two situations, static and dynamic lots.

The minimum recommended increment size is generally about 200 grams.

And the maximum recommended sampling rate is 1 increment every 400 kilograms.

For dynamic lots where product is in motion,

collecting incremental samples is much easier and straight forward.

For instance, in processing lines, conveyors, elevators, packaging hoppers;

it's a simple matter to take a sampling cup, or an automatic sampler,

and get incremental samples at the proper intervals.

Here's an example of shelled groundnuts on a conveyor.

When the groundnuts leave the conveyor that's a going point at which an increment can be taken.

Here's a diagram of an in-line automatic sampler.

We have conveyor belt, product on the belt, and as you can see.

Kernels contaminated with mycotoxins may not be homogeneously dispersed.

Therefore, its wise to cut the entire stream when you take an incremental sample.

This is just a photograph of a commercially available in-line automatic sampler that cuts the entire stream.

For static lots, product not in motion.

Getting incremental samples is much more difficult.

Product can be in sacks, boxes, transport containers or retail packaging.

Here's an example of groundnuts in large super sacks.

In this case, each sack contains 1000 kilograms,

and it would be very difficult to get an incremental sample from each sack.

These are super sacks closer up on a truck.

Again, each sack contains 1000 kilograms.

How do you collect your laboratory sample from this shipment?

One way to collect samples from containers in static,

would be to use a vacuum probe type sampler.

This is the top view of one of the super sacks, showing various probing locations.

And for example, if each probe collected 2 kilograms,

and one probe was taken from each sack,

a 20 sack lot would yield a 40 kilogram aggregate sample.

Here's a pneumatic sampler for inshell groundnuts.

Where the farmers pull their wagon loads of inshell groundnuts underneath the sampler.

The sampler probes the wagons at many different locations from top to bottom,

to collect the aggregate sample.

In this case, shelled groundnuts in small bags are sampled using a hand trier, or manual probe,

to collect the incremental samples.

Once the aggregate sample is formed,

a good method for dividing the aggregate sample to get your laboratory sample

would be a rotary type divider as shown on the left here.

Or if the aggregate sample is small,

a simple riffle divider could be used to divide out your laboratory sample.

Once your laboratory sample has been collected and comminuted,

you must select a representative test portion from that ground laboratory sample.

To do that,

every comminuted particle needs to have an equal chance of being chosen from the laboratory sample.

The selection process should be random.

And again, even if a representative test portion is selected,

the inherent sample preparation variability is not eliminated.

To comminute the laboratory sample.

An AMS mill such as the one shown on the left, is a good solution.

This mill provides 2 sub-samples automatically.

And these sub-samples are truly representative of the laboratory sample.

If a mill is used that does not automatically provide sub-samples,

the comminuted laboratory sample may be divided using a riffle divider to get at the test portion.

Once the test portion is obtained,

use a representative analytical method to process, extract, and measure the mycotoxin in the test portion.

So that brings us back to the diagram, the critical steps.

We know that the performance of any testing plan depends upon:

the number and size of the laboratory sample,

the laboratory sample preparation method,

the size of the test portion,

and the analytical method used to analyze the test portion.

However, all of these 3 phases of testing protocol must be representative as defined here.

And we know that the error between the concentration measured in the test portion,

and the true lot concentration depends upon sampling error, sample preparation error, and analytical error.

We also know that sampling error and sample preparation error,

are the two biggest error components of the three.

Analytical error is generally small, 5 to 10 percent of the total error.

So the total natural variability of any testing protocol, depends upon the protocol itself.

Representative sampling must be performed in order to validate any sampling protocol.