Abstract:
|
Two-period, two-treatment crossover designs are commonly used in clinical trials. These trials can include censored outcomes, such as for a novel coagulant treatment to increase clotting time or a novel sleep aid to decrease time to falling asleep. For continuous endpoints, it has been shown that baseline measurements collected before the start of each treatment period can be useful in improving the power of the analysis. Methods to achieve the same gain for survival endpoints in this setting have not been studied. In this project, we will examine a method that utilizes period-specific baseline observations of a failure time outcome in a crossover trial. Censored outcomes in cross-over trials require special consideration. Brittain and Follmann (2011) proposed a hierarchical rank test that focuses on the idea that preventing an event is clinically more important than delaying an event. Others have considered a frailty method approach. We examine whether methods can be improved by incorporating a baseline measurement of the failure time of interest. Properties of different approaches are examined through simulation. We will also present analysis of a real data example.
|