Issues in the design and analysis of clinical trials with time-to-event outcomes
Designers of clinical trials today face a number of different challenges. In a number of disease areas several treatments become available at any one time with only a limited number of patients available for investigation purposes. Furthermore, in disease areas such as HIV and cancer the pressure is high to find effective treatments quickly. Due to the recent advances in the human genome project more and more interaction effects between a treatment under study and the genetic make-up of a patient may be successfully analysed. This thesis aims to evaluate existing tools for the design and analysis of clinical trials with a time-to-event outcome and provide extensions in areas where existing tools do not perform satisfactorily. Particular emphasis is placed on sample size calculations for multi-arm and multi-stage trials and other complex mechanisms such as loss to follow-up and patient withdrawal from allocated treatment. Furthermore, advances are made in the area of treatment-covariate interactions, particularly in terms of analysis tools for such interactions.