|
Abstract:
|
It is commonplace for late-phase clinical trials to include multiple endpoints to fully characterize the benefit of the studied treatment, potentially across efficacy measures, doses, and populations. Statistical methodologies that account for the multiplicity introduced by testing these endpoints simultaneously must be utilized in order to satisfy regulatory requirements and control the familywise type I error rate at the pre-specified level of significance. The graphical testing framework (Bretz et al. 2009, 2011) provides an intuitive representation for a multiple testing procedure, while offering flexibility and transparency in how the endpoints are evaluated. These features greatly facilitate the incorporation of key information such as likely range of effect sizes, medical value of individual hypotheses to be tested, and correlations among the endpoints; while at the same time maintaining a balance between statistical power and the complexity of the multiple testing procedure. In this presentation, I will discuss methodologies for designing and optimizing a graphical testing scheme to maximize the value of a late phase clinical trial.
|
