This thesis concerns the linking together of the currently distinct techniques of pharmacokinetic-pharmacodynamic and health economic modelling. These have limitations, with economic models being empirical and thus hard to extrapolate outside the evidence base where they are constructed, and challenging to implement early in the drug development process. Pharmacokinetic-pharmacodynamic models, by contrast, are mechanistic but produce a limited range of outputs, not generating all the values useful to inform decision making. Pharmacokinetic-pharmacodynamic-pharmacoeconomic models, by linking these approaches, have the potential to overcome these limitations. The feasibility, validity and applicability of such an approach are assessed through two case studies. The first contains both retrospective and prospective simulations of rituximab for the treatment of follicular lymphoma. Retrospective analyses allow simulated results to be compared with trial-based data, and show an acceptable degree of concordance between the two methods. The prospective simulation of a trial currently recruiting will enable comparisons with the results of the trial, when these become available. The second and larger case study uses anticoagulation and stroke prophylaxis for patients with atrial fibrillation as an example. The end result is a full, prospective simulation of genotype dosed warfarin compared with both standard clinical dosing and a number of newly available oral anticoagulants. To make such an analysis possible, necessary prerequisite work was undertaken with the construction of a discrete event simulation to extrapolate both trial and simulation results to a lifetime horizon and an indirect comparison of all available treatments, to ensure all possible alternatives are considered in the analysis. The modelling approach described has the potential to allow the calculation of earlier estimates of cost-effectiveness than are currently available, which can be used to inform and improve the efficiency of the drug development process, and enable better extrapolations of trial-based analyses to different patient populations or dosing regimens.