A New Parametric Accelerated Failure Time Model for Semi-Competing Risks Data

Area of Science:

• Biostatistics
• Epidemiology
• Survival Analysis

Background:

• Cohort studies often focus on disease occurrence, requiring accurate modeling of competing risks like death.
• Semi-competing risks, where a terminal event (death) can censor a non-terminal event (disease), necessitate specialized statistical approaches.
• Illness-death models, tracking transitions between healthy, diseased, and dead states, provide a framework for analyzing such data.

Purpose of the Study:

• To introduce a novel statistical method using accelerated failure time (AFT) models for analyzing illness-death transitions.
• To provide an intuitive interpretation of results based on survival functions, facilitating communication.
• To develop a flexible parametric model accommodating semi-competing risks, left truncation, and interval censoring.

Main Methods:

• Employed accelerated failure time (AFT) models for each transition in an illness-death model (healthy-diseased, healthy-dead, diseased-dead).
• Proposed a trivariate parametric model using Weibull distributions for ages at events, incorporating random effects for intra-individual correlations.
• Addressed methodological challenges including left truncation and interval censoring for disease onset using maximum likelihood estimation.

Main Results:

• The proposed AFT-based illness-death model yields plausible results, consistent with existing methods in analyzing dementia onset and mortality.
• Simulation studies demonstrate promising accuracy and numerical robustness of the new modeling approach.
• The method effectively handles semi-competing risks, left truncation, and interval censoring inherent in cohort data.

Conclusions:

• The AFT-based illness-death model offers a valuable and interpretable alternative for analyzing complex survival data with semi-competing risks.
• This approach enhances the understanding of disease progression and mortality dynamics in cohort studies.
• The model's flexibility and performance suggest broad applicability in epidemiological and clinical research.