Mediation Analysis of Path-Specific Effects in Randomised Clinical Trials With Repeatedly Measured Mediators and...

Area of Science:

• Clinical Trials Methodology
• Causal Inference in Medicine
• Pharmacodynamics

Background:

• Understanding drug mechanism of action (MoA) is crucial for scientific and regulatory bodies.
• Causal mediation analysis can elucidate MoA when it is not yet established.
• Existing methods may not fully utilize longitudinal data or adequately adjust for confounding.

Purpose of the Study:

• To present a general framework for causal mediation analysis in clinical trials with repeated longitudinal measurements.
• To enable direct establishment of drug MoA by analyzing path-specific effects.
• To provide a method that accounts for all longitudinal data and adjusts for post-baseline confounders.

Main Methods:

• Development of a causal mediation analysis framework inspired by time-to-event outcome approaches.
• Utilizing causal diagrams to define path-specific effects from longitudinally measured variables.
• Incorporating an estimation method designed for repeated measures and confounding adjustment.

Main Results:

• The proposed framework facilitates direct MoA determination.
• The method effectively utilizes the full longitudinal data.
• Proper adjustment for confounding variables is achieved.

Conclusions:

• The presented framework offers a robust approach to causal mediation analysis in longitudinal clinical trials.
• This method enhances the understanding of drug MoA by leveraging comprehensive data and causal inference.
• Provided SAS code allows for general application in similar research settings.