Design and Analysis of Pharmacokinetic and Pharmacodynamic Trials

Pharmacokinetic (PK) and Pharmacodynamic (PD) experiments play an important role in drug development. Statistical design determines the treatment allocation and time points for taking PK/PD measurements. A good design will reduce the cost and suffering of humans/animals used in the experiment and ensure valid statistical inference. PK/PD modelling uses mathematical and statistical models to describe the time-concentration profile and the relationship between drug concentration in the blood and the effects. It also provides appropriate information for dosing, dose regimen selection and labelling. We investigated the asymptotic properties of two stage procedures and developed a two-stage procedure based on the subject by period model for the analysis of cross-over PK/PD data. This procedure does not rely on normality assumptions and can be enhanced by robust procedures such as the bootstrap. We proposed the use of generalised nonlinear mixed models (GNLMM) to analyse data in which the normality assumption is problematic. We proposed ways of extending current approaches for the nonlinear mixed model for fitting and hypothesis tests in GNLMMs. We also proposed an exact generalised estimating equation procedure to fit GNLMMs. This procedure is less computer intensive than the exact maximum likelihood and Bayesian approaches, yet results in consistent estimates of the PK/PD parameters. We investigated the asymptotic properties of different procedures for GNLMMs, particularly when the model is misspecified. We considered optimal cross-over designs for PK/PD models and found that dual balanced AB/BA designs are optimal and that a uniform design is optimal among all dual balanced designs for more than two treatments and periods. We developed algorithms to construct optimal designs in different practical scenarios such as optimal designs under constraints, optimal designs for multiple objectives, Bayesian designs and other design problems which are not ordinary design problems such as the optimal design for Tmax and Cmax.

Name dc:type.qualificationname

PhD

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

De Montfort University

Year dc:date.issued

1998

Author dc:creator

• Wang, Jixian

Licence

• https://dora.dmu.ac.uk/bitstreams/8c8f3200-c9f3-4fb0-8ffc-0b5deb7425f4/download