Pharmacometrics to improve evaluation and individualization of anticancer drug treatment

Abstract: The success rate in clinical development of newer anti-cancer molecules is the lowest compared to other major diseases. Improving the success rate and better early evaluation of efficacy of anti-cancer drugs remain challenging obstacles. Population modelling and model informed drug discovery (MIDD) has been utilized as a fundamental component in facilitating drug development and for decision making in the past decade. The aim of this thesis was to develop pharmacometric approaches to analyze various types of data collected in oncology trials to facilitate drug development and explore the value of model-based dose individualization to improve anticancer drug use.The developed models for the longitudinal tumor size data illustrated that three-dimensional measurements can be more sensitive than current standard - unidimensional measurements - at predicting progression free survival and overall survival. A framework for tumor lesion modeling was developed which allows for quantification of inter-lesion and inter-organ variabilities in tumor dynamics. A new mechanism–based population modelling approach for tumor dynamics model describing sensitive, quiescent and resistant tumor parts was able to characterize the variable tumor response patterns.A new methodology for analyzing survival data in oncology – a parametric multistate model – was developed that can describe the intermediate events and jointly characterize the outcome events including both PFS and OS. In the multistate model frame work, the predictors were evaluated in a prospective manner to not introduce immortal time bias. Furthermore, we applied the multistate model framework to assess the confounding effects of second line treatment in an OS analysis and illustrated that the multistate approach can delineate the impact of second line therapies on survival. Identification of responders and non-responders early after therapy initiation is crucial to trigger treatment modification whenever needed. Highly sensitive methods for tumor response quantification, that correlate with clinical outcome, are therefore required. A simulation study illustrated that the tumor follow-up duration can influence the accuracy in the model derived metrics and thereby impacting the prediction of hazard of death for individual patients. Further a model-based framework was developed that can be used to simulate the potential of biomarker-based dose individualization in sunitinib treatments. Toxicity adjusted dosing and biomarker-based dose adjustment algorithms increased median overall survival as compared to a fixed dose schedule and therapeutic drug monitoring-based dose adjustments, without markedly raising the risk of intolerable toxicities. Model-based dose individualization was suggested to provide a rapid, economical and safe approach to compare various dosing strategies, and guide dose individualization.In summary, the developed modelling approaches provide a better understanding of the relationships between drug exposure, short-term tumor response, and long-term clinical outcome. Such model-based approaches could be applied to improve the efficiency in clinical drug development, and may be used as a support for selecting dose and therapy for individual patients.