Structural Nested Cumulative Failure Time Models to Estimate the Effects of Interventions.

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Standard failure time analyses can be biased by time-varying confounders affected by prior treatment.
• Existing methods like the parametric g-formula and marginal structural models have limitations.

Purpose of the Study:

• To propose a novel method, g-estimation of structural nested cumulative failure time models (SNCFTMs), for causal effect estimation.
• To address challenges of time-dependent treatments, informative censoring, and time-dependent confounders.
• To enable calculation of unconditional cumulative risks under static treatment regimes.

Main Methods:

• G-estimation of structural nested cumulative failure time models (SNCFTMs).
• Modeling the ratio of counterfactual cumulative risks under different treatment scenarios.
• Utilizing inverse probability weights to adjust for informative censoring.
• Developing a procedure for estimating unconditional cumulative risks under static treatment regimes.

Main Results:

• The proposed SNCFTM method provides a robust approach to causal inference in the presence of complex time-dependent factors.
• The method correctly adjusts for time-varying confounders affected by prior treatment and informative censoring.
• A procedure is presented to estimate static treatment effects from dynamic treatment models.

Conclusions:

• SNCFTMs offer a powerful tool for estimating causal effects in longitudinal studies with time-dependent treatments and confounders.
• The method enhances the reliability of findings in epidemiological and clinical research.
• Accurate estimation of treatment effects is crucial for understanding disease etiology and prevention strategies.