What is best practice for micro sampling imported food powders?

Hi. Am new to this forum and am hoping for some advice.

We import dried vegetable and fruit powders from India and Africa into the UK. Although we send samples for micro upon arrival, the sample size is not related to the volume. In other words, the same size sample is taken from a shipment of 500kg as from a shipment of 10 tonnes.

Is there an accepted ratio of samples taken to volume of product? A product may share the same batch number but still be spread across a dozen pallets. Is one sample from one pallet acceptable?

Thanks.

Hi. Am new to this forum and am hoping for some advice.

 

We import dried vegetable and fruit powders from India and Africa into the UK. Although we send samples for micro upon arrival, the sample size is not related to the volume. In other words, the same size sample is taken from a shipment of 500kg as from a shipment of 10 tonnes.

 

Is there an accepted ratio of samples taken to volume of product? A product may share the same batch number but still be spread across a dozen pallets. Is one sample from one pallet acceptable?

 

Thanks.

Hi nick,

Welcome to the Forum !

the answer(s) may depend somewhat on the micro. item(s) sampled for and associated specification, etc, etc.

Theoretically detection limits are only related to sample size, not lot size (this assumes various factors such as lot size >= x (maybe 10) times sample size, random mix, etc).

In practice tabled sampling sizes may be adjusted for lot size for other reasons, eg economic value.
 

 

A product may share the same batch number but still be spread across a dozen pallets. Is one sample from one pallet acceptable?

Again it may depend on the objective / specification.

If the composition / contamination was 100% identical between pallets, any 1 sample would have same properties.

For a given level of confidence in the result, the theoretical sampling requirement depends on the variation within and between pallets.

One sample is unlikely to be very meaningful for a heterogeneous distribution if a negative detection result is specified and obtained.

The typical approach is to get sufficient data by increased sampling of >1 lot so as to see what the actual statistical variation(s) is like. then you can do some predicting.

On the other hand (rightly or wrongly) agreed routine sampling procedures may be standardised for your business and agreed as such. ?

You might see in the literature that many published accept/reject  micro. specs are based on 5 samples. this is of dubious "accuracy" for a heterogeneous mix but (a) attempts to achieve an "average" evaluation and (b) is a  practical max. handling capacity for labs doing loads of products (probably overload for many labs). In the case you mention, one would logically sample 5 pallets, remove the prescribed sample size and analyse each sample separately. Or sometimes combine the 5 samples, eg for salmonella.

PS - the typical problem in practice is that one does not know if a specific batch number is related to same production or a combination of sub-lots.

Hence the existence of generic sampling plans.

You might also see the parallel thread to this one -

http://www.ifsqn.com...d-dry-products/

Thanks for the prompt and detailed response Charles.

Where we're trying to get to is having a simple company policy for all incoming food powders. Your point about batch numbers not necessarily being related to a single production is well noted.

I guess we could either:

Take an individual sample for every pallet

Take an individual sample for every tonne (or more)

Take samples from every pallet and then aggregate into one sample

So many variables!

Am intrigued as to how other businesses handle this issue.

Thanks for the prompt and detailed response Charles.

 

Where we're trying to get to is having a simple company policy for all incoming food powders. Your point about batch numbers not necessarily being related to a single production is well noted.

 

I guess we could either:

 

Take an individual sample for every pallet

Take an individual sample for every tonne (or more)

Take samples from every pallet and then aggregate into one sample

 

So many variables!

 

Am intrigued as to how other businesses handle this issue.

Hi nick,

Yes, so many variables. Sampling can be a probabilistic nightmare. Best carried out in as simple a way as possible IMO.

IMEX, the sampling undertaken / lot for (non-regulatory)  accept/reject purposes is dictated by the contracted product specification, typically 5 samples.

The capabilities for routine (informative?) sampling, are usually dictated by logistics / product value / lab. cost. And sometimes by customer demand where finished product is involved.

For routine container sampling / salmonella, I have used variations between 1 per container and 1 per ton (maximum of 5). Strictly rule-of-thumb and depending on factors like the product  risk  / value / contract / history.

To approx.  illustrate the theoretical limitations -

13 samples of 25 g are taken from a lot and tested for salmonella. if all are negative, one can claim with 95% confidence that the contamination in the lot (visualized as an equivalent number of 25g units) does not exceed 20% !

60 negative samples are required to reduce 20% to 5%.