Marshall, Scott F.
Application of nonlinear mixed effects modelling in the early phases of drug development.
PhD thesis, University of Glasgow.
Full text available as:
It has been proposed that the use of cross-over or dose escalation designs for dose ranging studies in combination with more informative analysis could lead to a better characterisation of the dose response relationship. In example presented, the dose response relationship for the 3-hydroxy-3-methylglutaryl Coenzyme A inhibitors (HMG COA) inhibitor, simvastatin, was estimated from a cross-over study which covered the current recommended dose range (10 to 40 mg). Analysis using nonlinear mixed effects modelling approach demonstrated that the selected doses only covered 20% (70% to 90%) of the upper part of the estimated dose response relationship. It was concluded that a lower dose strength would be required to allow adjustment within the log-linear portion of the dose response relationship. The clinical implications of potential relationships between the pre-treatment cholesterol level and the model parameters were explored through prediction and simulation. On simulating the relationship between dose and the percentage of patients who would achieve reductions to below a recognised target concentration, it was found that a different set of dosages may better optimise clinical response.
Where strict experimental design is invalidated by study design or restricted recruitment, the resulting data can be unbalanced and not easily analysed by standard statistical methods. In the example presented, the number and size of doses of dofetilide used to test for PK/PD differences between patients with ischaemic heart disease (ISH) and healthy volunteers were different. A population PK/PD modelling approach was implemented, and no difference between the two groups could be detected. The Cmax and peak QTc ranged were predicted to be narrower following a fixed dose regimen in comparison to a dose per kilogram regimen. However, after incorporating the PK/PD variability, this was not predicted to manifest into an overall increase in the risk of Torsades de Pointes.
Actions (login required)