High-Dimensional Inference for Personalized Treatment Decision.

Area of Science:

• Statistics
• Biostatistics
• Medical Informatics

Background:

• Personalized treatment decisions increasingly rely on high-dimensional regression models.
• Interactions between treatments and patient covariates are crucial but often lack clear statistical interpretation.

Purpose of the Study:

• To develop an asymptotically unbiased estimator for interaction coefficients in personalized treatment models.
• To provide statistically interpretable results for covariate effects in treatment decisions.

Main Methods:

• Utilized Lasso regression solutions to estimate interaction coefficients.
• Derived the limiting distribution of the estimator for unknown or misspecified baseline functions.
• Developed confidence intervals and p-values for inferring covariate effects.

Main Results:

• The proposed estimator is asymptotically unbiased.
• The method demonstrates accuracy and robustness against model misspecification in simulations.
• The approach was successfully applied to the STAR*D study for major depressive disorder.

Conclusions:

• The developed statistical method enhances the interpretability of personalized treatment decision models.
• This approach offers a robust framework for identifying relevant covariates in clinical practice.
• The findings have implications for optimizing treatment strategies in major depressive disorder and other conditions.