|
Abstract:
|
In general, clinical trials are often conducted with stratified randomization to maintain the balance of treatment allocation with respect to the important prognostic factors. Per FDA ICH E9 guidance, the stratification factors used during the randomization should be adjusted in the statistical analysis. In such a trial with binary endpoint, the overall treatment effect is generally analyzed by common risk difference in stratified 2x2 tables. A clinical trial for non-inferiority assessment of an investigational treatment is determined by comparing the confidence interval (CI) with the pre-specified non-inferiority margin (NIM). We will review the various ways to calculate the 95% CI for treatment difference and their assumptions. The confidence intervals (CI) can be constructed from various approaches, such as (1) the Wald CI, (2) the Newcombe (hybrid score) CI and (3) Miettinen & Nurminen (score) CI. We examine the performance of these approaches through simulations by evaluating the coverage probability and length of the CI. Numerical examples from recently approved antimicrobial compounds are presented.
|
