Statistical methods of screening cut point determination in immunogenicity studies

Background: Currently, screening cut point (CP) calculated from an assay validation with replicates are applied to an immunogenicity study with nonreplicates, for which the antidrug antibodies rate is determined. IID treats the replicate of a sample as coming from another independent sample. AVE uses average results from each sample across runs but inter-assay variability is reduced. Therefore, we propose a random effect model (REM) for calculating CP. Materials & method: We investigate impact of noncompatibility design between validation and immunogenicity studies on CP and compare these methods. Conclusion: IID may not fit for use when replicates' variability dominates all sources of uncertainty. REM considers covariance structure of repeated measurements. CP by REM is smaller than that by IID but larger than that by AVE.