Verification/validation of categorical methods – resolving rLOD, POD, LOD95 and other approaches

Analytical methods are classified as continuous (i.e., quantitative) or categorical (i.e., qualitative) based on their response. Categorical methods include those providing binary outcomes (yes/no, positive/negative, etc.)  such as screening tests for use in the analysis of food or oil products (including seeds of oils and food crops). These methods are also referred to as “qualitative” methods. Their characterization depends on the concentration of the measurand and whether the measurand is detected or not. A qualitative analytical method proposed as a standard assay, should be validated to prove its acceptability with respect to scope and accuracy. The type of validation experiment performed depends upon the availability of standard reference material, or another established or reference method against which the candidate method can be compared.

Throughout the world several international standards provide validation and verification guidance for categorical methods and the determination of equivalence, but these are not harmonized. Characteristics such as the rLOD, LPOD, LOD95 are used in contexts ranging from the beta-binomial distribution to the Poisson distribution to provide estimates of equivalence. In all cases, the potential of a Bayesian region of practical equivalence procedure has not been developed to provide a practical equivalence estimate in cases where results from similar matrices are compared. This presentation will review the goals of categorical testing, current standard methods for validation and determining equivalence and the possibilities for potentially harmonizing or interchanging results from them.