EXPRESSION OF THE REFERENCE VALUE FOR PROFICIENCY TESTS

Profficiency tests raise two types of problems: (i) the homogeneity of a sample must be carefully checked before using it as a reference material; (ii) it is necessary to define a conventional reference value (RV) in order to rank participants. Since these topics are poorly addressed in the literature, a reference material was specially prepared in order to propose a procedure for verifying homogeneity and to establish whether the algorithms classically used to compute the conventional true value have an influence on the conclusion of the test. The homogeneity of the sample was not perfect and univariate and multivariate techniques were used to demonstrate some analytes can be suspected as heterogeneous. However, this was considered adequate for introducing the sample in a large proficiency test organized between four laboratory associations which regularly perform such tests. Altogether more than 3000 measurements were collected and eight algorithms were applied to compute the RV on 14 analytes. Thus is was possible to demonstrate that some of these algorithms are better adapted than others. It is also obvious that some analytes are better suited to proficiency testing. It can be concluded from this work that some effort towards standardization would be profitable to check homogeneity or to compute the RV, especially since the economic weight of proficiency testing is becoming even more important with the development of certification accreditation