Comparison of primary analysis strategies of randomized controlled trials with multiple endpoints with application to...

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Regulatory Science

Background:

• Single primary endpoints in randomized controlled trials (RCTs) are often impractical due to rare or heterogeneous events.
• Regulatory agencies permit multiple endpoints, but analytical strategies vary in addressing scientific questions and null hypotheses.

Purpose of the Study:

• To compare the statistical power of three primary analysis strategies for multiple endpoints in RCTs: composite endpoints (CE), multiple testing and multiplicity correction (MTMC), and generalized pairwise comparisons (GPC).
• To evaluate these strategies under various scenarios, including endpoint prioritization, correlation, and opposing treatment effects, particularly for time-to-event data.

Main Methods:

• The study employed clinical trial simulations to assess the power of CE, MTMC, and GPC strategies in two-arm RCTs.
• Simulations considered different scenarios reflecting endpoint characteristics and treatment effect heterogeneity.

Main Results:

• Global testing strategies (CE and GPC) generally demonstrated higher statistical power compared to MTMC for time-to-event endpoints.
• Global procedures can indicate statistical significance even with heterogeneous treatment effects across endpoints, necessitating component-wise evaluation.

Conclusions:

• Global analysis strategies for multiple endpoints in RCTs can enhance statistical power but require careful interpretation.
• Understanding the balance between statistical efficiency and interpretability is vital for selecting appropriate endpoint definitions and primary analysis strategies in future clinical trials.