Using decision lists to construct interpretable and parsimonious treatment regimes

Area of Science:

• Biostatistics
• Clinical Informatics
• Personalized Medicine

Background:

• Treatment regimes are key to personalized medicine, aiming to optimize patient outcomes and reduce healthcare burdens.
• Developing effective treatment regimes requires close collaboration between statisticians and clinical scientists.
• Interpretability of treatment regimes is crucial for clinical adoption and scientific validation.

Purpose of the Study:

• To propose a flexible and interpretable class of treatment regimes.
• To develop a robust statistical method for estimating optimal treatment regimes within this class.
• To facilitate the integration of clinical science into statistical modeling for treatment regime development.

Main Methods:

• A novel class of treatment regimes represented as if-then statements was proposed.
• A robust estimator for the optimal treatment regime within the proposed class was derived.
• The method's performance was evaluated using simulation experiments and real clinical trial data.

Main Results:

• The proposed class of treatment regimes offers immediate interpretability.
• The derived estimator demonstrated effective finite sample performance in simulations.
• The method was successfully illustrated using data from two distinct clinical trials.

Conclusions:

• The proposed if-then statement-based treatment regimes are practical for clinical application.
• This approach supports the collaborative development of interpretable and effective personalized medicine strategies.
• The method provides a valuable tool for statisticians and clinicians to refine treatment strategies.