Abstract:
|
The designs of current clinical trials often involve interim analysis for assessment of futility (low conditional power) or for sample size adjustment to maintain desired conditional power. In a single-arm study, Visual Analog Scale improvement from baseline was measured at three post-baseline visits for each patient and the average of these post-baseline VAS scores was the primary efficacy endpoint. The primary analysis compared the mean of this endpoint to a pre-specified performance goal using Mixed Model Repeated Measures. A simulation method is proposed for assessing conditional power in this setting, conditioned on the interim observed data and under the assumption that the observed interim sample characteristics are the true population characteristics. Specifically, once the interim data are observed, multiple post-interim datasets are simulated from a population with the same characteristics as the interim observed data. Complete simulated data sets are then composed of the observed interim dataset appended to each simulated post-interim data set. The proportion of complete simulated data sets for which the null hypothesis is rejected is the simulated conditional power.
|