|
Abstract:
|
Statistical analyses for primary endpoints in confirmatory clinical trials need to be clearly specified in the study protocol, before any data is collected. This includes, of course, power and sample size calculations for the specified methodology. Although this is mostly not an issue when simple methods are used, it can lead to difficulties when more complex approaches are considered. We will discuss issues involved in sample size and power calculations when a linear mixed-effects model is used to analyze a longitudinal clinical endpoint. These include, among others, considerations about drop-outs, uncertainty about underlying covariance structures, and sensitivity to misspecification of prior values for the model parameters. A real clinical trial on an neurological disease indication will be used to motivate and illustrate the methods presented.
|
