Empirical Bayesian Modeling of Pharmacokinetic Data

• Activity Number: 664
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): David Bryant King*+ and Richard A Forshee and Robert Mitkus
• Companies: FDA/CBER and FDA/CBER and FDA/CBER
• Keywords: Pharmacokinetic ; Bayesian ; Hierachical ; Informative Censoring ; Emperical methods

Abstract

The use of Gibbs sampling to fit hierarchical Bayesian nonlinear models is an appealing approach to analyze data arising from experiments in the life sciences. One area for which Bayesian methods perform particularly well is in the analysis of pharmacokinetic (PK) data. In the following paper we will illustrate the pragmatic value of hierarchical Bayesian methods toward the analysis of PK experiments with the analysis of a particular dataset. We will discuss the use of empirical methods to help elicit prior parameters and how to adjust posterior intervals to account for prior empirical information. In the analysis we will also discuss how we have adapted our Bayesian model to contend with the issue of informative censoring or missing not-at random (MNAR) in PK data. In the particular dataset we analyze the animal subjects have an unequal number of repeated observations and were observed for different lengths of time. The analysis will also show that non- hierarchical models of the data will lead to "survivor bias" which is a form of bias where animals which have more observations for longer times have more influence on the shape of the PK curves.