Statistical methods for the analysis of multiple time to event endpoints
Clinical studies are conducted to show the efficacy or efficiency of a drug, agent, or treatment. Thereby, time to death, time to ischemic stroke, or time to hospitalization is often of primary interest. However, the time until such an event is observed can be quite long and hence the study duration extends unrealistically. To overcome this problem one can consider composite endpoints. Thereby, the occurrence of several different events is of interest. In a primary analysis it is neglected that several events can be experience by a patient, because only the first observed event is considered. If all events would be included, this could further decrease the study duration. Furthermore, in the common composite endpoint analysis it is assumed that all components of the composite have the same clinical relevance.
Some methods that included all observed evets per patient are considered the clinical relevance of the events were already described. However, this approaches were only considered separately. A combination of these method would be of interest to describe a disease process holistically. Hence, one part of the research project will the development of the combined methods.
For the analysis of (multiple) time-to-event data semi- or non-parametric methods are usually applied. However, parametric models could be of more interest for the analysis of multiple endpoints because thereby the model development and estimation can be easier. Thus, in the research project a bivariate accelerated failure time model will be developed.
A further problem of studies where multiple time-to-event endpoints are of primary interest, is the study planning/sample size calculation. Often no adequate sample size calculation is given for the described analysis method. Hence, in this research project methods for study planning and sample size calculation for multiple time-to-event endpoint will be developed and recommendations for the applied researcher will be given.
- Prof. Dr. Annika Hoyer (University Bielefeld)
- Prof. Dr. Oliver Kuß (Heinrich-Heine-University Düsseldorf, German Diabetes Centre)
The project builds on the following projects: