Statistical Procedures for Bioassays When the Condition of Similarity Does Not Obtain

Abstract

Well-established statistical procedures are available for the analysis of dilution parallel line or slope ratio assays for which the condition of similarity obtains. Research scientists have long been aware that this condition is commonly violated, in log-dose response assays, for example, divergent rather than parallel lines may be obtained. The deviation cannot always be traced to deficient experimental techniques. In fact, as indications of differences in the response processes of the standard and test preparations, or as indications of impure test preparations, such findings may provide the most important inference from an assay experiment and immediately suggest further investigation into the causes of the differences. Statistical procedures have been developed to describe the phenomena in quantitative terms and, especially, to permit potency comparisons. The procedures may also have merit even in dilution assay situations where the condition of similarity may apparently be violated if appreciable differences between the responses of the standard and test preparations result from poorly matched doses.