On the estimation of total variability in assay validation.

In the pharmaceutical industry, an assay method is considered validated if the accuracy and precision for an assay meet some acceptable limits. This paper discusses the assessment of assay precision in terms of the estimation of total variability of an assay from a one-way random effects model which is often considered in assay validation. We propose a general class of estimators that includes the analysis of variance estimator and the maximum likelihood estimator. We derive the optimal estimator, in terms of smallest mean squared error, within this class and consider an approximate version of this optimal estimator. We report on a Monte Carlo simulation to study its finite sample performance. We also present two examples to illustrate the use of the proposed methodology.

Duke Scholars

Author Shein-Chung Chow Biostatistics & Bioinformatics, Division of Biostatistics