2026-10-13T05:11:34Z https://www.scidb.cn/oai

10.57760/sciencedb.j00253.05370 2026-05-29T16:45:37Z

2026-05-29 Estimation of Marginal Interventional Effects for a Time-to-Event Mediator in a Multi-State Model Framework and Its Application doi:10.57760/sciencedb.j00253.05370 en Objective To propose a marginal interventional mediation effect estimation method based on a multi-state model framework for follow-up data in which both the mediator and terminal outcome are time-to-event variables and a semi-competing risk structure is present.Methods A marginal interventional multi-state time-to-event mediation method (MIMS-TEM) was developed. Semi-parametric multi-state Cox models were used to characterize the state transition process, and marginal standardization was applied to estimate the total effect, direct effect, and indirect effect on the risk scale at the population level. Simulation studies were conducted to evaluate the performance of the proposed method. The method was further applied to the Henan HIV/AIDS Antiretroviral Therapy Cohort to analyze the mediating effect of CD4 recovery in the association between timing of ART initiation and all-cause mortality.Results Simulation studies showed that MIMS-TEM had low bias and stable confidence interval coverage across different truncation levels and effect scenarios, and generally outperformed simplified approaches that excluded pre-mediator death, treated pre-mediator death as censoring, or coarsened the mediator into a binary variable. Compared with the CMAverse method, MIMS-TEM showed more stable performance in estimating population-level marginal effects. In the application study, compared with ART initiation >30 days after diagnosis, ART initiation within 7 days reduced the 9-year cumulative mortality risk, with a total effect of -0.046 (95% CI: -0.070 to -0.029), a direct effect of -0.043 (95% CI: -0.069 to -0.028), and an indirect effect of -0.002 (95% CI: -0.005 to -0.001). ART initiation within 8–30 days also showed a protective effect.Conclusion MIMS-TEM can estimate population-level marginal interventional effects for a time-to-event mediator under a semi-competing risk structure, providing methodological support for evaluating public health intervention effects and investigating underlying mechanisms. The application results support the continued implementation of rapid ART initiation strategies in HIV/AIDS prevention and control, and suggest that the protective effect of timely ART initiation may be mainly achieved through pathways other than CD4 recovery. Multi-state model; Time-to-event mediator; Marginal interventional effect; Semi-competing risks; HIV/AIDS Antiretroviral therapy Nana.Wang Xuezhong.Shi Xiaocan.Jia Yali.Niu Xiaoran.Geng Yongli.Yang PUBLIC https://creativecommons.org/publicdomain/zero/1.0/ dataset Science Data Bank