On optimal treatment regimes selection for mean survival time

Area of Science:

• Biostatistics
• Clinical Trials
• Personalized Medicine

Background:

• Maximizing patient survival time is crucial in clinical studies with time-to-event endpoints.
• Patient heterogeneity necessitates personalized treatment strategies integrating clinical and genetic information.
• High-dimensional predictors pose challenges for reliable and interpretable optimal treatment regime selection.

Purpose of the Study:

• To propose a robust loss-based estimation framework for optimal treatment regimes and variable selection.
• To develop a model-free estimator for restricted mean survival time under optimal treatment.
• To assess the performance of the proposed methods in parameter estimation, variable selection, and treatment decision.

Main Methods:

• A robust loss-based estimation framework coupled with shrinkage penalties for estimating optimal treatment regimes.
• Asymptotic properties of the proposed estimators were theoretically studied.
• Development and asymptotic analysis of a model-free estimator for restricted mean survival time.

Main Results:

• The proposed framework facilitates simultaneous estimation of optimal treatment regimes and variable selection.
• Asymptotic properties of the estimators were established.
• Simulations demonstrated the method's empirical performance in parameter estimation, variable selection, and optimal treatment decisions.

Conclusions:

• The developed robust framework effectively addresses challenges in selecting important variables for optimal treatment regimes.
• The method provides a reliable approach for personalized treatment strategies, enhancing patient survival time.
• The application to AIDS clinical trial data illustrates the practical utility of the proposed methodology.