Estimating scaled treatment effects with multiple outcomes.

Area of Science:

• Biostatistics
• Clinical Trials
• Epidemiology

Background:

• Classical study designs focus on single outcomes, but modern research often collects data on multiple outcomes.
• Patient-centered research necessitates evaluating treatment effects on various outcomes simultaneously, as importance varies among individuals.

Purpose of the Study:

• To propose scaled effect measures that standardize effects on multiple outcomes onto a common scale.
• To develop efficient, nonparametric, doubly robust methods for estimating these scaled effects and testing hypotheses about treatment effects on all outcomes.

Main Methods:

• Utilized potential outcomes framework for scaled effect measures, employing mean-variance and median-interquartile range standardizations.
• Developed nonparametric, doubly robust estimation methods for scaled effects and summary measures.
• Incorporated methods for assessing effect modification by covariates.

Main Results:

• Demonstrated efficiency gains in randomized trials and reduced confounding bias in observational studies with high-dimensional covariates.
• Illustrated the proposed methods through simulation studies and a real-world data analysis.
• Provided theoretical efficiency and asymptotic behavior of the proposed estimators.

Conclusions:

• The proposed nonparametric methods offer a robust approach for analyzing multiple outcomes in both randomized and observational studies.
• These methods facilitate patient-centered research by providing a unified scale for treatment effects across diverse outcomes.
• The approach effectively handles complex scenarios, including high-dimensional covariates, enhancing the reliability of treatment effect estimations.