Using Model-Based Recursive Partitioning for Treatment-Subgroup Interactions Detection in Real-World Data: A...

Area of Science:

• Biostatistics
• Pharmacogenomics
• Health Informatics

Background:

• Treatment effects can differ significantly across patient subpopulations based on baseline characteristics.
• Identifying these treatment-subgroup interactions is crucial for advancing precision medicine and personalized patient care.
• Evaluating treatment effectiveness heterogeneity is essential in real-world data settings.

Purpose of the Study:

• To propose an analytical framework for detecting treatment-subgroup interactions.
• To evaluate treatment effectiveness heterogeneity in real-world data.
• To aid clinical decision-making in personalized medicine.

Main Methods:

• Utilized Model-based Recursive Partitioning Analysis (MOB) for subgroup identification.
• Employed filter-based confounder selection for confounding reduction.
• Applied multivariate logistic regression for treatment effectiveness assessment.
• Validated the framework using the China Acute Myocardial Infarction (CAMI) registry data.

Main Results:

• The framework successfully identified meaningful patient subgroups exhibiting treatment-subgroup interactions.
• Demonstrated the ability to assess treatment effectiveness variations across identified subgroups.
• The analysis involved evaluating the effects of 15 drugs in myocardial infarction (MI) patients.

Conclusions:

• The proposed analytical framework effectively detects treatment-subgroup interactions.
• This approach aids in understanding treatment effectiveness heterogeneity.
• The findings support improved decision-making for personalized medicine strategies.