Setting Specifications of Correlated Quality Attributes through Multivariate Statistical Modelling

Specifications of drug products are quality standards that provide assurance that the drug is fit for its intended use. Establishment of specifications limits relies on data collected during the life cycle of a product, including those from preclinical and clinical trials, stability studies, and process validation, and often requires application of appropriate statistical methods. While quality attributes may be correlated, specifications are often determined without taking into account possible interdependence. Such practice may potentially result in specifications that are narrower than acceptable ranges of product performance. Through a case study, this paper discusses how this issue can be coped with effectively by using multivariate statistical modelling.

Specifications of drug products are quality standards that provide assurance that the drug is fit for its intended use. Specifications are conventionally set up separately for each of the quality attributes of interest, without accounting for interdependence among the quality attributes. Such practice may potentially result in specifications that are narrower than acceptable ranges of product performance. Through a case study, this paper discusses how this issue can be coped with effectively by using statistical modelling.