Multi-arm multi-stage clinical trials for time-to-event outcomes

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacometrics

Background:

• Multi-arm multi-stage (MAMS) designs offer efficiency in clinical trials by reducing time and patient numbers.
• Controlling Type I error (statistical significance) is complex in MAMS trials due to multiple comparisons and stages.

Purpose of the Study:

• To investigate and propose methods for constructing efficacy and futility boundaries in MAMS clinical trials for time-to-event outcomes.
• To adapt existing statistical procedures for scenarios with consistent or differing intermediate and final outcome measures.

Main Methods:

• Proposing a generalized Dunnett procedure for MAMS trials with consistent outcomes to control familywise error rate (FWER).
• Modifying existing methods to control both pairwise error rate (PWER) and FWER for MAMS trials with differing intermediate and final outcomes.
• Exploring the performance of MAMS designs under violation of the proportional hazards assumption, including delayed treatment effects.

Main Results:

• The generalized Dunnett procedure effectively controls FWER when intermediate and final outcomes are the same.
• Modified methods extend PWER control to FWER for MAMS trials with different outcome types.
• Violation of the proportional hazards assumption, particularly with delayed effects, can lead to power loss in MAMS designs.

Conclusions:

• The proposed MAMS approaches provide robust statistical frameworks for efficient clinical trial evaluation.
• An alternative test statistic is suggested to mitigate power loss when proportional hazards assumptions are violated.
• These methods streamline the comparison of numerous therapies, accelerating the drug development process.