Subgroup identification via Interaction Tree and Mixed Model for Repeated Measures with application to Alzheimer's...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Precision Medicine

Background:

• Subgroup identification is essential for personalized medicine and designing effective treatments.
• Longitudinal clinical trials require robust methods to analyze treatment effects over time.
• Existing methods may struggle with nonlinear interactions and regulatory compliance.

Purpose of the Study:

• To develop and validate a novel approach for subgroup identification in longitudinal clinical trials.
• To integrate the strengths of tree-based methods (Interaction Tree) with longitudinal data analysis (Mixed Model for Repeated Measures).
• To ensure the method aligns with regulatory guidelines for assessing treatment effects.

Main Methods:

• The proposed ITree-MMRM method combines Interaction Tree for capturing nonlinear interactions with MMRM for longitudinal data analysis.
• Parameter tuning options and bootstrap methods are explored for optimal tree pruning and to mitigate overoptimism.
• The approach is designed to handle heterogeneous treatment effects and provide reliable subgroup identification.

Main Results:

• Simulations demonstrate that the ITree-MMRM method outperforms existing subgroup identification techniques.
• The model successfully identified patient subgroups with differential long-term treatment responses.
• The approach maintains flexibility in detecting complex interactions while adhering to statistical standards.

Conclusions:

• The ITree-MMRM model offers a powerful and flexible tool for subgroup identification in longitudinal clinical trials.
• This method enhances the design of personalized treatments by uncovering patient subgroups with specific responses.
• Application to an Alzheimer's disease trial highlights its utility in real-world clinical research.