Applications of Statistical Regression and Modeling in Fill–Finish Process Development of Structurally Related Proteins

• Published in: Journal of pharmaceutical sciences Vol. 100; no. 2; pp. 464 - 481
• Main Authors: Jiang, Ge, Thummala, Abhinaya, Wadhwa, Manpreet-Vick S.
• Format: Journal Article
• Language: English
• Published: Hoboken Elsevier Inc 01.02.2011; Wiley Subscription Services, Inc., A Wiley Company; Wiley; American Pharmaceutical Association; Elsevier Limited
• Subjects: Antibodies - chemistry; antibody; Biological and medical sciences; Chemistry, Pharmaceutical - methods; Drug Discovery - methods; drug product manufacturing; fill-finish; filter capacity; Filtration; General pharmacology; Immunoglobulin G - chemistry; Medical sciences; modeling; Models, Statistical; Pharmaceutical technology. Pharmaceutical industry; Pharmacology. Drug treatments; Protein Conformation; regression; Regression Analysis; Statistical applications; Viscosity; Vmax; viscosity; filtration; modeling; antibody; drug product manufacturing; regression; Vmax; fill–finish; Statistical applications; filter capacity; Drug; Viscosity; Filtration; Pharmaceutical technology; Antibody; Process; Modeling; Protein; Filter; fill―finish; Statistical model; Preparation; Manufacturing
• ISSN: 0022-3549; 1520-6017
• DOI: 10.1002/jps.22296

In order to increase efficiency and reduce cost, many biotechnology and pharmaceutical companies utilize platform approaches for discovery and development of structurally related therapeutic proteins. In the case of the monoclonal antibody modality, retention and reuse of prior development knowledge is especially useful to gauge risks, improve speed and reduce cost for developing similar molecules in the future. In this paper, we present two applications of statistical regression and modeling to help decision making during antibody drug product fill-finish process development. The applications are for estimating viscosity and filter capacity (Vmax) values. Experiments were performed to obtain relevant data sets of viscosity, protein concentration, density, and Vmax values for various candidate antibodies. Then, statistical models were developed and optimized to estimate viscosity and filtration Vmax values for new antibodies. Viscosity of protein formulations is an important physical property that impacts almost all manufacturing operations, as well as delivery or administration of drug products. Vmax is a critical parameter for filter size selection in manufacturing processes. Development and optimization of both models followed similar steps: identifying multicollinearity and interactions, removing unnecessary explanatory variables, performing appropriate data transformation, and evaluating different model options. We obtained 95% prediction limits for the mean and individual values from the models, and further verified the models by comparing predicted values with additional experimental data. These applications of statistical tools enabled leveraging prior knowledge for process development of new molecules belonging to the same class of structurally related proteins. Although the two specific models presented here may not be directly applicable for all proteins, the approach and methodology presented can be broadly useful for structurally related protein products during their development.