Methodology of PKPD modelling and analysis for myelotoxicity #
Rebecca Rumney
12:10 Wednesday in 2Q50/51.
Part of the Epidemiology and statistical learning session.
Abstract #
Neutropaenia, anaemia and thrombocytopaenia are common side effects of chemo-therapeutic drugs that can lead to severe consequences. Pharmacokinetic and pharmacodynamic (PKPD) models can be used to predict these adverse effects and to optimise dosing regimens to avoid them. These models, however, require specialist knowledge to formulate them and to analyse their outputs, with many choices and assumptions made in the process. This study aims to help standardise the analysis methodology, to ensure these models are robust and the modelling choices transparent. This will allow the development of re-usable workflows, enable a clearer analysis for criticism, provide better comparability between studies, and make the training of analysts in this area easier.
I propose a set of questions that need to be answered in analyses in PKPD models, as well as a set of methods that can be used to answer these questions. I then apply these methods to a non-linear mixed effects PKPD model for myelotoxicity. I use Bayesian inference to estimate the parameters of the model. I explore parameter identifiability methods to determine whether the parameters can be identified. I also determine how model selection techniques can be used to make model choices.