Pharmacometrics in early clinical drug development

Abstract

Pharmacometrics, the science of quantitative clinical pharmacology, has been recognized as one of the main research fields able to improve efficiency in drug development, and to reduce attrition rates on the route from drug discovery to approval. This field of drug research, which builds heavily on the sciences of mathematics, statistics, pharmacology and biology, provides a tool for studying relationships between drug exposure and drug effects. This thesis concerns the application of pharmacometrics in the early clinical development of drugs, and has a strong focus on oncology. The thesis shows that the use of population modelling & simulation greatly facilitates the appropriate handling of all available data, including data which may be difficult to handle using standard approaches (e.g. missing, censored or baseline data). It is shown here that, especially when faced with such limitations or challenges, the use of modeling techniques instead of reliance on the more conventional statistical techniques, allows the most appropriate and data-efficient analysis of data. Of course, this is especially relevant when data is sparse or associated with high variability, which is often the case in (early) clinical studies. The thesis shows that the use of pharmacodynamic biomarkers in a PK-PD analysis can greatly support the early clinical development of new drugs. Several pharmacokinetic-pharmacodynamic (PK-PD) analyses in early drug development are presented in the thesis, incorporating a PD biomarker for efficacy or toxicity. One of the chapters presents the construction of a model for the occurrence of hypertension and proteinuria toxicity during treatment with a novel anticancer drug, based on early clinical data. Subsequent simulations using this model provided support for the hypertension intervention scheme proposed for Phase II clinical development, and showed that intra-patient dose escalations using hypertension as a biomarker may be feasible without inducing considerable toxicity. In another analysis, preclinical and early clinical biomarker and tumour growth data were combined to allow an early assessment of anticancer activity of another novel drug. In two other chapters, investigations into the predictive ability and schedule-dependency of a PK-PD model for haematological toxicity during treatment with anticancer drugs are described. When modelling early clinical or preclinical data, the modeller is often confronted with technical difficulties concerning incomplete data. In the final section of the thesis, investigations are presented into two such problems: data below the limit of quantification (BLOQ) and incomplete covariate datasets. For the former problem, it was shown that the use of extrapolated BLOQ concentration data provides superior performance over established methods. For the latter problem, an approach in which an extra parameter for covariate effect was estimated for the group in which data on the covariate was missing, provided the easiest and most robust performance to analyses such datasets. The thesis also describes the construction of a graphical user interface, Piraña, which facilitates the use of the modelling software NONMEM and improves workflow for pharmacometric scientists.