BAGS: A Bayesian Adaptive Group Sequential Trial Design With Subgroup-Specific Survival Comparisons.

Area of Science:

• Biostatistics
• Clinical Trials
• Statistical Methods

Background:

• Randomized trials often involve patient subgroups with potential treatment-interaction effects.
• Standard analyses may lack power or efficiency when subgroup heterogeneity is present.
• Adaptive designs can improve statistical power by leveraging subgroup information.

Purpose of the Study:

• To propose a Bayesian group sequential design for survival comparisons in randomized trials with potential treatment-subgroup interactions.
• To enhance within-subgroup power by adaptively combining homogeneous or splitting heterogeneous subgroups.
• To incorporate baseline covariates for improved subgroup identification and basing comparative tests on the average hazard ratio.

Main Methods:

• A Bayesian group sequential design framework is developed.
• A latent subgroup membership variable allows adaptive subgroup management.
• Incorporation of baseline covariates for subgroup identification is included.
• Guidelines for calibrating prior hyperparameters and design parameters are provided.

Main Results:

• Simulations demonstrate the design's robustness across various scenarios.
• The proposed method shows substantially greater power than ignoring subgroups or using separate subgroup tests when subgroups are homogeneous.
• The design effectively controls the overall Type I error rate.

Conclusions:

• The proposed Bayesian group sequential design offers a powerful and adaptive approach for subgroup analyses in clinical trials.
• This method is particularly advantageous when treatment effects vary across distinct patient subgroups.
• The design provides a flexible framework for optimizing statistical power and improving subgroup identification in survival studies.