Randomized Interventional Effects in Semicompeting Risks, With Application to a Hematopoietic Cell Transplantation...

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Survival Analysis

Background:

• Intermediate events can modify the risk of primary (terminal) events, creating semicompeting risks in clinical studies.
• Decomposing treatment effects into direct and indirect pathways is crucial for understanding mediated effects.

Purpose of the Study:

• To extend the randomized interventional approach to time-to-event outcomes with intermediate and terminal events under right censoring.
• To develop methods for identifying and estimating direct and indirect treatment effects in the presence of semicompeting risks.

Main Methods:

• Utilizing a randomized interventional approach with a random draw for the intermediate event process.
• Presenting an identification formula for interventional effects and discussing assumption variants.
• Employing nonparametric maximum likelihood estimation and proposing a latent frailty sensitivity analysis.

Main Results:

• The study illustrates the method using hematopoietic cell transplantation data, analyzing death mediated by relapse.
• Graft-versus-host disease (GVHD) was considered as a time-varying confounder.
• Matched unrelated donor transplantation showed preferable survival rates compared to haploidentical donor transplantation for lymphoma patients, particularly with post-transplant cyclophosphamide (PTCy) GVHD prophylaxis.

Conclusions:

• The proposed method effectively decomposes treatment effects in semicompeting risks scenarios.
• Matched unrelated donor transplantation is a favorable strategy for improving survival in specific lymphoma patient populations undergoing hematopoietic cell transplantation.
• The findings highlight the importance of considering intermediate events and time-varying confounders in survival analysis.