A comparison of procedures for testing noninferiority using proportion difference, odds ratio, and proportion ratio between two independent binomial samples
Testing noninferiority is frequently used in clinical trials to demonstrate that a new experimental treatment is not much worse than an active control treatment. If the new treatment is noninferior to the existing one, the new treatment can be a useful alternative treatment to the active control treatment when the former is less expensive, easier to administer, and has fewer side effects. We evaluate the performance of test procedures for testing noninferiority using proportion difference (PD), odds ratio (OR), and proportion ratio (PR) between two independent binomial samples. We further use data taken from a study of two antibiotics, clarithromycin and erythromycin, to illustrate the practical use of these test procedures, power function, and sample size calculation formulas. Using Monte Carlo simulation, we found that the proposed procedure based on the PR showed significant improvement in terms of a substantial increase in power and a significant reduction in the minimum required sample size.
