Reconciling Validation and Routine Decision Rules

The basic aim when applying prestudy and in-study validation procedures to a measurement method is to reconcile the objectives of the two validation phases. When the tolerance interval approach is used for prestudy validation and the 4 6 2 rule is used during in-study validation, the common objective is to control the proportion n of measurement results (X — jiy) that fall within the acceptance limits [—2,+ 2]. 

The ability to reconcile the prestudy and in-study rules and risks depends then on the adequacy of the parameters chosen. These parameters should ensure that a 

The alignment of risk between the prestudy and in-study validation phases can be envisaged in two ways, as shown by Boulanger et al. [29]. On the one hand, if the number of QC samples, n, to be used and the minimum, 5, of QC samples within the acceptance limits in the s n X rule are fixed (e.g., 4 6 15), the value of 7imin should be chosen so as to ensure that if the method remains valid, the s n X rule is accepted in most cases (e.g., with a minimum probability ymin). On the other hand, for a given prestudy validation scheme and X fixed), the value of 5 QC samples within the acceptance limits (for a given n) should guarantee that most of the runs will be accepted if the method remains valid. 

Let us now consider the particular case of the 4 6 X rule. As stated above, a good prestudy validation rule should be based on a value of 7rmin that ensures acceptance of a routine test in most cases (say y = 90%) if the method is valid, that is, 

For a given s and n, this is obtained by inverting the binomial (n, n) distribution function, as shown in Fig. 5.3. 

---