Food Standards Agency
Friday 3 July 2009
Safer food better business banner
AZ-Directory
What's NewRSS
What is RSS?Listen
Listen to this siteThursday 7 October 2004
This research project aims to develop a proposal for an inexpensive and widely applicable method for assessing uncertainty associated with sampling from bulk foods.
Study Duration : April 2001 to October 2001
Contractor : ADAS Consulting Ltd
Sampling is the most important initial stage in any analytical process, but the concept of sampling variability is often neglected. In order to arrive at a correct decision on compliance against quality specifications, control authorities need information on the total variance associated with any results gained. The total variance is made up of two major contributions: one from sampling and one from the measurement of the samples. Any result of a measurement is accompanied by a degree of uncertainty, and it is incumbent upon the analyst to estimate the magnitude of this uncertainty and ensure that it is appropriately small. Laboratories can check that they meet 'fit-for-purpose' requirements by implementing a program of quality assurance measures. Laboratories routinely employ ongoing quality control procedures and record the outcome on quality control charts.
The measures, put in place by a laboratory to monitor uncertainty only conveys part of the story, however. There is another important aspect. If the laboratory sample was not representative of a bulk consignment of goods, even the best analytical result could be misleading. Often the laboratory has no control over this as a third party submits the samples. Ideally then there should be some level of control in enforcement situations where sampling should have been undertaken in accordance with a sampling plan. Even so, there will be variation in the composition of the product throughout a batch and, consequently, variation in the composition of samples collected according to the plan. In order to make a reliable assessment of the quality of a consignment the user needs to know the uncertainty introduced by this sampling variation as well as the measurement uncertainty.
The aim of this research is to develop a proposal for an inexpensive and widely applicable method for assessing uncertainty associated with sampling from bulk foods that would be suitable for use in a quality control style. The project aims to assist in ensuring fitness-for-purpose of food surveillance programs and to provide a basis for long-term assessment of sampling precision. The estimates of sampling precision should allow end users of analytical data to estimate the combined uncertainty of results, a prerequisite to interpreting data correctly.
A number of analyte/matrices combinations were used in developing the concept of routine quality control of sampling (QCSAM). Namely:
This study finds that that it is easy to set up and use quality control of sampling (QCSAM) charts for the quality control of sampling precision, when the laboratory sample is a composite of a number of increments taken at random from the lot or consignment of material.
Data collected from a relatively small number of sampling events can be used to set up control limits for the split absolute difference (SAD). As always in quality control, the initially derived limits need updating from time to time as more data characterising the sampling and analytical methods was accumulated. In addition, the accumulating data provides a credible estimate of the overall sampling standard deviation, based as it is on results stemming from a number of different lots of the target material sampled under realistic conditions.
In this study there were relatively few outlying results, and they were handled by standard outlier tests, and zero-weighted in the calculation of the variances and thence the control limits. Generally it is important to discount outliers or to use robust methods to estimate the variances that define the control limits. Inclusion of outlying results in the calculations would tend to make the control limits too wide and the chart too insensitive to excursions.
In the majority of the analyte/test material combinations used in this study, the sampling variance was smaller than the analytical variance. This is perhaps characteristic of manufactured or processed materials as opposed to primary raw materials or unprocessed food items. For those materials the control chart represented mainly the analytical variation. However, that does not detract from their utility in relation to sampling. If the sampling variance for a particular material was unduly large, the fact would become apparent as an outlying point on the control chart. At that stage it would not be possible to distinguish between a sampling excursion and an analytical excursion: the chart would simply warn that one or other of the actions was out of control and the resultant measurement possibly not fit-for-purpose. Additional measures would be needed to see whether it was the analysis or the sampling that had gone out of control. Routine use of a homogeneous control material would act as a warning of analytical problems, or alternatively, the use of duplicate analysis of each of the splits (as in this study).
In many other circumstances, the QCSAM chart will represent sampling variations alone, and in the absence of counter-indications, a 'non-compliant' SAD value will indicate a problem with sampling: either the material under test is more heterogeneous than usual or the sampling procedure is flawed. While it may require some extra work to distinguish between these two possibilities, there could be little doubt that the sampling event as a whole was suspect and might need to be repeated under more rigorous conditions.
The QCSAM method has been demonstrated as both practicable and easy to implement under a much wider range of conditions than previously reported. The next stage in the demonstration of its capabilities would be to introduce the method into a number of real-life situations in which foodstuffs (or feedstuffs) are sampled on a routine basis. We recommend that such an investigation should be undertaken, preferably with some analytes and materials selected from known problem areas in sampling.
Farrington, D., Jervis, A, Shelly, S., Damant, A., Wood, R. and Thompson, M. (2004). A pilot study of routine quality control of sampling by the SAD method, applied to packaged and bulk foods. Analyst, 129, 359-363.
The final report is available from the FSA Library and Information centre.
To obtain a copy, please contact the Enquiry Desk, Dr Elsie Widdowson Library and Information Services, Food Standards Agency (tel: 020 7276 8181/8182 or email:
library&info@foodstandards.gsi.gov.uk
)
Contact
: For any enquiries concerning this research project, please contact the relevant Programme contact or email
science@foodstandards.gsi.gov.uk
Find out what our other sites have to offer