CRMs with high levels of cfp

In the certificate of analysis of a CRM, tables with 95% confidence limits of the certified value are given in relation to the number of capsules and the number of replicates. The certificate does not state the optimal number of capsules or replicates a laboratory should use. An example of a certificate of Bacillus cereus (BCR-CRM 528) is given in Annex 3.3. Statistical methods can be used to determine the required number of capsules and replicates that allow a good judgement of an experiment and that is realisable in practice. 

A true mean for the number of cfp, in one analytical portion, for the studied technique and for a particular laboratory, is obtained through experiments. To determine precisely, a very large number of capsules should be analysed. In practice, only a limited number of capsules can be analysed. With the results of this small number of capsules an estimate, called of the mean number of capsules for the laboratory can be computed. This value of mi f, is an estimate of the unknown /// /,. 

When a laboratory wants to know if its true mean, is similar to the certified value,i.e. the number of cfp in one analytical portion, the analyst compares with If /%/, is within the 95% confidence limits (C.I.), as can be found in the tables for users in the certificate, the laboratory concludes that /// /, is comparable to The 

9SVa C.l. means that the probability to conclude that is indeed the case in 

When nii f, is much larger or smaller than the conclusion will be that is different from It is possible that a wrong conclusion about jU/ is drawn from the results of the experiments When the conclusion that /// /, s drawn, but in reality Hi f, = p erf lyP I error occurs as shown in Table 3.3. is assumed representative for///, but in reality is not. The probability for a type I error is commonly called a i.e. the probability of concluding from the experiment that a significant difference exists between and Type I errors should be limited as much as possible, a 

---

For microbiological CRMs with high levels of cfp, tests are based on the geometric mean. The required number of capsules and replicates will depend on the level of differences the analyst wants to detect between the certified value and the laboratory geometric mean and on a and p. Of course, any difference is relevant and must be detected but this is unrealistic since the number of capsules to be used for such a level of acceptance would be infinitely large [13]. It is economically also unrealistic. 

---