Wang, Pin-wen. Two-stage winner designs for non-inferiority trials with two to three experimental treatments and an active control. Retrieved from https://doi.org/doi:10.7282/T3BK19X8
DescriptionIn drug development, two-stage winner design can be cost-effective when the best treatment is to be determined from multiple experimental treatments. In this design, an interim analysis is devised to select the best treatment to avoid high cost, long-term trial conduction, and exposure to ineffective treatments. When an existing effective treatment is available, but new experimental treatments have advantages such as less cost, easier delivery, less invasive and fewer side effects, etc., including placebo in the trial may be considered unethical. It is desirable to conduct a non-inferiority trial that directly compares new treatments with the existing treatment (active control), and show that the new treatments are not less effective than the active control by a certain amount (so called non-inferiority margin). In this dissertation, we extended the framework of the two-stage winner design of Shun et al. (2008) to conduct non-inferiority tests. Specifically, we considered designs of trials with two or three experimental treatments and an active control. Our methods include superiority hypotheses as a special case, but the hypothesis setting is more general than that studied by Shun et al.. We studied the distribution of test statistics, cut-off values, sample size and power calculations using exact distribution of the test statistics as well as using normal approximations. Theoretical justifications and extensive numerical assessments were conducted to calculate the design parameters and evaluate the performance of our methods.