Objectives:We review the statistical considerations for superiority trial to design appropriate statistical hypothesis and decision-making.
Methods:We plan superiority design especially for placebo-controlled trial or combination drug trial. Many previous studies employed two-sided hypothesis testing and misuse of significance level, and this resulted in wrong sample size and confidence interval. Therefore we compared sample size estimation and confidence interval according to one-sided and two-sided hypotheses.
Results: There are differences in sample size estimation and confidence interval according to one-sided or two-sided hypotheses with the pre-determined power at the significance level,Conclusions: For superiority trial in the confirmatory stage, one-sided hypothesis is advisable. We can estimate a sample size from this hypothesis and recommend to use a confidence interval for decision-making.