Informative Priors for Modeling Immunogenic Responses of Biopharmaceuticals

• Activity Number: 140
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Rinke Klein Entink*+ and Babs O. Fabriek and Geertje van Mierlo and Frans Tielen and Esther Reefman
• Companies: TNO and TNO and TNO and TNO and TNO
• Keywords: Bayesian ; informative priors ; multilevel modeling

Abstract

To reduce animal testing, a model-based approach was developed to obtain a first indication of the risk of immunogenicity of biopharmaceuticals. A set of 14 substances was selected for which data was available from 91 human clinical trials on the prevalence of immunogenicity in the population. Between-study variability was modeled using multilevel logistic regression, including covariates on the third level of modeling to explain between-substance variability. As covariates, physical/chemical characteristics, in vitro test results and animal test data were available. A Bayesian analysis of the model was performed because of the low number of available substances and to allow for the propagation of uncertainty at different levels of the model. Expert knowledge on immunogenicity was translated into informative prior distributions on the model coefficients, using mixtures of truncated distributions to restrict the parameter space and account for uncertainty in the assumptions. A sensitivity analysis and first validation of the prior assumptions was performed. The influence of different prior choices and derived practical knowledge for implementing informative priors are discussed.