Methodology for Nonlinear Mixed-Effects Modelling (Population PKPD) of preclinical models of cancer

Abstract

The use of pharmacokinetic and pharmacodynamic (PKPD) modelling is less common in preclinical studies than in clinical studies. The general aim of this thesis was to use PKPD modelling in a range of preclinical scenarios to improve the design and analysis of studies. Most of the studies were based in cancer, looking at either efficacy through tumour growth or toxicity such as myelosuppresion. The first chapter assessed alterations to the design of studies of myelosuppression following chemotherapy which allowed the characterisation of the pharmacokinetics from the main study animals rather than requiring an additional parallel group of satellite animals. The new "compact design" was tested in a simulation study, and was found to perform equally well as the traditional design, whilst reducing the number of animals required. In the second chapter, methods to account for drop out due to a tumour burden limit in xenograft studies were developed and tested on a simulated example. This informative drop out was found to lead to underestimation of the dose-response, however by using methods to account for drop out in the analysis the dose-response could be recovered. The preferred methods were then applied to a number of case studies in the following chapter. Next the use of patient-derived xenografts in mouse trials to detect drug-gene interactions was investigated and a number of methods to improve power were tested. The design was found to be unreliable, as the results differed greatly depending on the method used, which may largely be due to small sample sizes. Mixed-effects modelling was found to greatly increase the power to detect interactions versus the comparison of final tumour sizes. Finally, data from intravenous glucose tolerance tests and glucose clamp tests were used to improve the glucose input for tests of insulin sensitivity, with respect to parameter estimation in the minimal model. A new simpler test design was found to improve estimation of the parameters, reducing the correlation in key parameters of interest such as insulin sensitivity. The new design was shorter, and resulted in less glucose being administered overall. Although PKPD modelling is not commonly used in preclinical studies, this thesis presents a number of examples in which it can be used to improve both the design and analysis of preclinical studies.

Date of Award	31 Dec 2017
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Leon Aarons (Supervisor) & Kayode Ogungbenro (Supervisor)