Nonparametric Path-Specific Effects on a Survival Outcome Through Multiple Time-to-Event Mediators.

Area of Science:

• Causal inference
• Biostatistics
• Epidemiology

Background:

• Human diseases often progress through sequential, time-to-event milestones.
• These sequential events, like hepatitis progression to death, are subject to censoring and may influence each other.

Purpose of the Study:

• To develop a causal mediation model for time-to-event outcomes with multiple sequential mediators.
• To define and estimate interventional path-specific effects (iPSEs) for complex disease pathways.

Main Methods:

• Utilized a causal mediation framework with intermediate and terminal events.
• Derived counterfactual hazard expressions using a counting process model under sequential ignorability.
• Employed composite nonparametric likelihood estimation for maximum likelihood estimation of counterfactual hazards and iPSEs.

Main Results:

• Proposed estimators demonstrate asymptotic unbiasedness, uniform consistency, and weak convergence.
• Hepatitis B-induced mortality is predominantly mediated by liver cancer and/or cirrhosis.
• Hepatitis C-induced mortality may be mediated through extrahepatic diseases.

Conclusions:

• The developed causal mediation model effectively analyzes sequential time-to-event data in disease progression.
• The findings highlight distinct mediation pathways for mortality risk associated with hepatitis B and C infections.