Evaluation of results derived from the analysis of certified reference materials--a user-friendly approach based on simplicity.

Certified reference materials (CRMs) have now been in regular use for several decades. Their production and certification are regulated by international standards. But, even today there are no agreements on procedures for evaluating results obtained by the users. As a consequence, the way CRM results are treated in the literature leaves a lot to be desired. A statistical evaluation is rarely, if ever, described in published reports. The most common approach is to compare the found mean and/or range with the certified range and then state if the mean falls within the certified range, or if the two ranges overlap. If this happens, the analyst is usually satisfied. In addition, usually no regard is paid to the fact that the certified interval is based on a 95% confidence interval (CI) and the found interval on standard deviation and that this evaluation has little, if any, statistical relevance. Long-term evaluation of a CRM often consists in nothing more than producing a control chart, which relates the found results to the certified mean and CI. This paper is an attempt to improve the situation by providing a set of easy-to-use guidelines for evaluating results from CRMs. During the process we have identified different areas in which there is a need for such guidelines: 1. short-term evaluation of a single, or multiple, determination at one or several specific times; 2. identification of systematic and random errors; 3. evaluation of CRMs when used in a collaborative trial of a method; and 4. long-term evaluation for monitoring an analytical process over extended periods of time. It is important that the guidelines do not require expert competence in statistics from the analyst. Such obstacles would probably render most guidelines unused.