Augmented binary method for basket trials (ABBA).

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacological Research

Background:

• Traditional clinical trials often use responder outcomes, which dichotomize continuous measures, potentially reducing statistical power.
• Augmented binary methods have been proposed to leverage undichotomized data for improved power.
• Basket trials are increasingly utilized in clinical research to evaluate treatments across multiple subtrials.

Purpose of the Study:

• To extend an augmented binary method for use in basket trials, specifically for clinical areas employing responder outcomes.
• To enhance the efficiency of basket trials by enabling information borrowing on treatment effects between subtrials.
• To develop a novel statistical methodology within a latent variable framework using Bayesian hierarchical modeling.

Main Methods:

• Proposed an augmented binary method tailored for basket trials with responder outcomes.
• Employed a latent variable framework and Bayesian hierarchical modeling.
• Investigated methodology performance through simulations, comparing point estimates and high-density intervals against standard binary outcome analyses for basket trials.

Main Results:

• The proposed augmented binary method demonstrated a reduction in the 95% high-density interval of the posterior distribution for the log odds ratio.
• The method showed an increase in statistical power when treatment effects were consistent across subtrials.
• Simulations confirmed the enhanced efficiency and power compared to standard basket trial analyses.

Conclusions:

• The novel augmented binary method offers a more efficient approach for analyzing basket trials with responder outcomes.
• This methodology effectively borrows information across subtrials, leading to improved statistical power and precision.
• The approach was successfully illustrated using real-world data from rheumatology clinical trials.