Multiple Testing in Clinical Trials
Multiplicity problems caused by multiple analyses performed on the same dataset occur frequently in a clinical trial setting. The following are examples of multiple analyses encountered in clinical trials.
- Multiple comparisons. Multiple testing is often performed in clinical trials that involve several treatment groups. For example, most Phase II trials are designed to assess the efficacy and the safety of several doses of an experimental drug compared to a control.
- Multiple primary endpoints. Multiplicity can be caused by multiple criteria for assessing the efficacy profile an experimental drug. The multiple criteria are required to characterize accurately various aspects of the expected therapeutic benefits. In some cases, the experimental drug is declared efficacious if it meets at least one criterion. In other cases, drugs must produce significant improvement with respect to all endpoints (e.g., new therapies for the treatment of Alzheimer’s disease are required to demonstrate their effects on both cognition and global clinical improvement).
It is commonly recognized that failure to account for multiplicity issues can inflate the probability of an incorrect decision and could lead to regulatory approval of inefficacious drugs and increased patient risks. For this reason, regulatory agencies mandate a strict control of the false-positive (Type I error) rate in clinical trials ...