Nonparametric machine learning for precision medicine with longitudinal clinical trials and Bayesian additive...

Area of Science:

• Biostatistics
• Clinical Trials
• Machine Learning

Background:

• Precision medicine aims to tailor treatments to individual patient characteristics, moving beyond one-size-fits-all approaches.
• Traditional clinical trial analysis assumes uniform treatment effects across populations, potentially missing subgroup-specific benefits.

Purpose of the Study:

• To develop an advanced statistical framework for precision medicine in clinical trials.
• To enhance the identification of treatment heterogeneity and individualized treatment rules.

Main Methods:

• Proposed mixedBART, an extension of Bayesian additive regression trees (BART).
• Incorporated random effects for longitudinal data and subject clustering.
• Introduced a novel interaction detection prior to identify treatment heterogeneity.

Main Results:

• Demonstrated the utility of mixedBART through simulation studies.
• Applied the mixedBART framework to real-world randomized clinical trial data.
• Showcased improved identification of patient subgroups with varying treatment effects.

Conclusions:

• MixedBART offers a powerful tool for advancing precision medicine in clinical trials.
• The framework facilitates the development of individualized treatment rules by detecting treatment heterogeneity.
• This approach can lead to more effective and personalized treatment strategies.