Mediation analysis with causally ordered mediators using Cox proportional hazards model

Area of Science:

• Biostatistics
• Epidemiology
• Survival Analysis

Background:

• Causal mediation analysis explores mechanisms linking exposures and outcomes.
• Limited research exists on mediation effects for survival outcomes, especially with multiple mediators.
• Existing methods often require rare outcome assumptions or specialized models.

Purpose of the Study:

• To propose a novel multi-mediation analysis for survival outcomes.
• To develop a method applicable to the Cox proportional hazards model without the rare outcome assumption.
• To investigate path-specific effects (PSEs) in multi-mediator settings.

Main Methods:

• Developed a counterfactual framework for identifying path-specific effects (PSEs).
• Derived closed-form formulas for PSEs on transformed survival time and survival probabilities.
• Established asymptotic properties for the proposed estimators and conducted extensive simulations.

Main Results:

• The proposed method effectively estimates path-specific effects in multi-mediator survival analysis.
• The approach is valid without the rare outcome assumption, unlike some previous methods.
• Simulations demonstrated good finite sample performance compared to existing techniques.

Conclusions:

• The novel Cox model-based mediation analysis provides a robust tool for survival data.
• This method expands the applicability of causal mediation analysis to complex survival scenarios.
• Demonstrated utility in a hepatitis study analyzing liver cancer risk.