Multiplicity Issues in Clinical Trials With Multiple Objectives

Abstract

Modern clinical trials for evaluating efficacy and safety of new treatments frequently include multiple objectives with questions of varying clinical importance. Answering them generally requires performing a number of statistical tests and analyses which raise multiplicity of tests issues. These issues can be complex and multidimensional in nature. For example, one dimension may relate to the assessment of the effects of the treatment on multiple endpoints, the other to the effects of multiple doses of the treatment, and yet another to the type of the tests (e.g., superiority or noninferiority-type tests). Also, the trial may seek claims for treatment benefits either for the total patient population or for targeted subgroups. In addition, there may be interest in finding whether or not a certain consistency of results persists across certain multiple endpoints; in some situations this may be measuring different but critical features of a disorder, or in other situations may be measuring the same underlying pathophysiology of the disorder. Addressing such problems in clinical trials for the purpose of controlling the Type I error rate requires the use of advanced statistical test strategies and methods, some of which have only recently appeared. Actually, the last decade has witnessed a number of novel methods as well as innovative extensions of old methods for addressing complex multiplicity problems in clinical trials. The main purpose of this article is to present—at the conceptual level—how multiplicity issues of confirmatory clinical trials that include multiple objectives can be addressed by using some of these new statistical methods that use α-propagation and gatekeeping concepts. One additional goal of this contribution is to address some issues that often arise in the use of coprimary and composite endpoints in clinical trials.

Key Words:

• Alpha-propagation
• Coprimary and composite endpoints
• Gatekeeping methods
• Multiple comparisons
• Primary and secondary endpoints
• Type I error rate control

Acknowledgments

Disclaimer: This article reflects the views of the authors and should not be construed to represent the FDA views or policy.