Comparing methods for estimating patient-specific treatment effects in individual patient data meta-analysis

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Translational Medicine

Background:

• Individual patient data (IPD) meta-analysis synthesizes data from multiple trials, enabling personalized treatment recommendations.
• Selecting relevant treatment-covariate interactions is crucial for accurate patient-specific effect estimation but remains challenging.

Purpose of the Study:

• To compare the performance of shrinkage methods against standard and stepwise regression for selecting treatment-covariate interactions in IPD meta-analysis.
• To evaluate methods for improving the estimation of patient-specific treatment effects.

Main Methods:

• Simulation study comparing standard IPD meta-analysis (all interactions) with stepwise regression and five shrinkage methods (LASSO, ridge, adaptive LASSO, Bayesian LASSO, SSVS).
• Evaluation of methods for continuous and dichotomous outcomes across various scenarios.
• Application to real-world datasets in cardiology and psychiatry.

Main Results:

• Shrinkage methods demonstrated strong performance for both continuous and dichotomous outcomes.
• These methods generally yielded lower mean squared error for patient-specific treatment effects compared to standard and stepwise approaches.
• Stepwise regression was found to be less effective.

Conclusions:

• Shrinkage methods are recommended for IPD meta-analysis aiming to estimate patient-specific treatment effects using multiple effect modifiers.
• Avoid stepwise regression due to its suboptimal performance in this context.
• These findings support the use of advanced statistical techniques for enhanced precision in personalized medicine.