Statistical inference in stability analysis

Statistical inference for drug shelf-life in stability analysis is considered. We propose a model that takes into account the batch-to-batch variation in assay results. Tests for batch-to-batch variation as well as the effects of some covariates such as time, package type, and dosage strength are proposed. When there is no batch-to-batch variation, we show that the method currently used in the pharmaceutical industry for determining the labeled drug shelf-life is statistically valid. In situations where batch- to-batch variation is present, we propose a method of determining the labeled shelf-life which ensures, with given statistical assurance, that the future batches of the drug product have shelf-lives longer than the labeled shelf- life. An example concerning a new drug application stability analysis conducted in a pharmaceutical company is discussed.

Duke Scholars

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