ABSTRACT
Large plate bioassays are normally assessed by analysis of variance. The British Pharmacopoeial approach requires a validation of the assay through a test for parallelism. This approach is shown to be impractical as it does not allow for improvements in technique which tend to reduce the error mean square value for the bioassay and hence increase the F ratio for parallelism. A new parallelism acceptance criterion is proposed in which a critical value for parallelism mean square is determined by multiplying the appropriate critical F value for parallelism by the value for error mean square which provides a limiting value for the fiducial limits. This approach allows for improved bioassay technique but does not lead to the acceptance of doubtful assay results. The results of 15 months of bioassays covering over 500 assays using the proposed parallelism acceptance criterion are discussed.
