Inference on tree-structured subgroups with subgroup size and subgroup effect relationship in clinical trials

Area of Science:

• Clinical Trials
• Biostatistics
• Statistical Inference

Background:

• Classical multiple testing procedures in clinical trials often lack interpretability and efficiency.
• Existing methods fail to adequately consider subgroup size and effect relationships in simultaneous subgroup inference.
• There is a need for advanced statistical methods to handle complex subgroup structures.

Purpose of the Study:

• To propose a novel data-adaptive and interactive multiple testing procedure for clinical trial subgroups.
• To develop a method that incorporates subgroup size and effect relationships within a tree structure.
• To improve the interpretability and efficiency of statistical inference for prespecified tree-structured subgroups.

Main Methods:

• The proposed method is built upon the selective traversed accumulation rules (STAR).
• It employs a data-adaptive and interactive approach for multiple testing.
• The procedure is designed to work with prespecified tree structures and considers subgroup size and effect.
• Accommodations for post hoc identified tree structures are also discussed.

Main Results:

• The proposed STAR-based method offers a more interpretable and efficient inference for tree-structured subgroups.
• The method demonstrates its merit through re-analysis of the panitumumab trial data.
• It provides a practical solution for simultaneous statistical inference on multiple candidate subgroups.

Conclusions:

• The developed procedure offers a significant advancement in statistical inference for clinical trial subgroups.
• It addresses limitations of classical methods by incorporating subgroup characteristics.
• The method facilitates more robust and meaningful conclusions from subgroup analyses in clinical research.