Keywords: Bioassay, Potency, Weighting
The nonlinear modelling of bioassay data is a common practice in the biopharmaceutical industry for estimating biological activity. The 4-parameter logistic (4PL) model is frequently used for this purpose. Various weighting methods to account for heterogeneity in responses have been proposed. The design of the bioassay is also an important consideration and may yield multiple estimates of potency, possibly non-independent. Combining these individual estimates into a single pooled estimate is a required statistical practice. Acknowledging the uncertainty in the original nonlinear model is controversial from various perspectives. Finney (1978) and Bliss (1952) have given extensive discussions on combining potency estimates. More recently, the USP and EP have also included a section on combining estimates. We present a summary of these methods and propose an approach within the framework of a mixed model accounting for possible non-independence. A comparison of the standard and regulatory approaches is also summarized in the form of a case study.