Predictive Biomarker Graphical Approach (PRIME) for Precision Medicine

Area of Science:

• Biomarker discovery and validation
• Translational medicine
• Pharmacogenomics

Background:

• Precision medicine relies on biomarkers for patient stratification in drug development.
• Accurate biomarker cutoffs are crucial for identifying patients who will benefit from specific therapies.
• Conventional methods for determining biomarker cutoffs often use p-values from dichotomized data.

Purpose of the Study:

• To introduce PRIME, a novel predictive biomarker graphical approach for evaluating biomarkers on a continuous scale.
• To enable the incorporation of clinical significance into biomarker evaluation.
• To develop a method for identifying optimal biomarker cutoffs for patient enrichment.

Main Methods:

• Adapted a treatment selection approach and extended it using G-computation to account for covariates.
• Developed a model incorporating the interaction between a biomarker and treatment to predict risk.
• Utilized graphical displays of predicted risk to delineate biomarker-outcome relationships and identify cutoffs.

Main Results:

• The PRIME approach allows for continuous biomarker evaluation from a predicted risk perspective.
• Graphical displays facilitate the identification of biomarker cutoffs for patient stratification.
• PRIME incorporates features for comparing biomarkers (net gain) and assessing model fit (calibration).

Conclusions:

• PRIME offers a robust graphical method for evaluating biomarkers and determining optimal cutoffs in a continuous manner.
• The approach can accommodate various outcomes and covariates, enhancing its applicability in precision medicine.
• An R package has been developed to facilitate the implementation and demonstration of the PRIME approach.