|
Abstract:
|
It has become rare to conduct a large clinical trial without prespecified interim analyses. Each interim analysis should be identical to the primary analysis at the end of the trial. A problem arises when a model-based primary analysis includes a treatment x covariate interaction term. Does one include it in the interim analysis? What if it is highly significant? The possible courses of action are to ignore the interaction effect and base the decision to stop or continue solely on the main treatment effect or to do subgroup analyses. For the latter, the test is unlikely to be powered properly for each subgroup, and if inconsistent results are obtained among the subgroups, the decision regarding the study conduct becomes complex. Roundtable luncheon participants will discuss the pros and cons of each course and share their experiences.
|
