Selection of hypothesis weights and ordering when testing multiple hypotheses in clinical trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Confirmatory Phase III clinical trials often involve multiple objectives, necessitating careful selection of multiple testing procedures.
• Free parameters, such as hypothesis weights and ordering, significantly impact the interpretation and power of these procedures.

Purpose of the Study:

• To discuss the selection of free parameters for multiple testing procedures in Phase III clinical trials.
• To identify classes of multiple testing procedures with varying interpretations of hypothesis weights and ordering.
• To provide practical guidelines for choosing hypothesis weights and ordering.

Main Methods:

• Classification of multiple testing procedures into single-step (e.g., Bonferroni) and stepwise (e.g., Holm, fallback, chain) methods.
• Examination of procedure behavior with unequally weighted and a priori ordered hypotheses.
• Illustration using case studies from clinical trials with multiple endpoints, dose-placebo comparisons, and patient populations.

Main Results:

• Different classes of multiple testing procedures offer distinct interpretations of hypothesis weights and ordering.
• Stepwise procedures allow for reweighting hypotheses during the testing process, potentially increasing power.
• Guidelines are provided for selecting hypothesis weights and ordering based on trial characteristics.

Conclusions:

• The choice of hypothesis weights and ordering is critical for the validity and power of multiple testing procedures in complex clinical trials.
• Understanding the behavior of different procedure classes aids in optimizing trial design and interpretation.
• Practical guidance is offered to statisticians and researchers involved in Phase III clinical trial design.