|
Abstract:
|
With uncertainties in the sign of between group difference and variability, the outcome of a non-inferiority trial could be clinically superior, statistically superior, clinically non-inferior and numerically superior, clinically non-inferior but numerically inferior, statistically inferior, or clinically inferior. By evaluating the likelihood of each potential claim and the associated benefit and risk based on interim data, we propose a sample size re-estimation strategy so as to optimize the outcome of the trial. Simulation study will be conducted to evaluate the inflation in Type I error, the impact on study power and the bias in point estimate.
|
