The R.O.A.D. to precision medicine.

Area of Science:

• Biostatistics
• Translational Medicine
• Computational Biology

Background:

• Randomized clinical trial (RCT) subgroup analysis has limitations.
• Observational data analysis often suffers from unobserved confounding.
• Precision medicine requires robust methods for personalized treatment selection.

Purpose of the Study:

• To develop a novel framework for analyzing subgroup data from both RCTs and observational studies.
• To transform observational data for use in a manner akin to randomized data.
• To enable more accurate and intuitive treatment recommendations for patient subgroups.

Main Methods:

• A two-step process to adjust predicted outcomes and counter unobserved confounding in observational data.
• Utilizing adjusted predictions to train decision trees for optimizing treatment assignments.
• Applying the framework to gastrointestinal stromal tumors (GIST) and extremity sarcoma datasets.

Main Results:

• The framework's treatment recommendations outperformed current guidelines for GIST in an external cohort.
• Identified a subset of extremity sarcoma patients who may not require treatment, contrary to initial trial indications.
• Recommendations were successfully validated in external patient cohorts for both GIST and sarcomas.

Conclusions:

• The proposed framework effectively addresses limitations in subgroup analysis for both RCT and observational data.
• This approach facilitates precision medicine by providing optimized, data-driven treatment recommendations.
• The methodology shows potential for improving patient outcomes by identifying optimal treatment strategies for specific patient subgroups.