Shared frailty sieve estimation for dependent left truncated and interval censored data.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Analyzing left-truncated and interval-censored survival data presents significant challenges, especially when failure times and observation processes are dependent.
• Existing methods, such as copula models, may impose restrictive assumptions on complex interval-censoring mechanisms.

Purpose of the Study:

• To propose the first shared frailty model tailored for left-truncated, interval-censored survival data.
• To capture heterogeneity and dependency between failure time and observation processes in survival data analysis.

Main Methods:

• A sieve maximum likelihood approach is developed, utilizing I-splines and M-splines to approximate unknown baseline hazard and examination intensity functions.
• The model accounts for left truncation, interval censoring, and dependent observation processes simultaneously.

Main Results:

• The asymptotic properties of the proposed estimators are theoretically established.
• Extensive simulations confirm the method's consistency, efficiency, and robustness across diverse scenarios.
• The approach demonstrated its capability in a real-world AIDS cohort study.

Conclusions:

• The developed shared frailty model offers a flexible and powerful tool for analyzing complex survival data.
• This method provides reliable parameter estimates, improving the understanding of time-to-event data with truncation and censoring.
• The application to AIDS data underscores its practical utility in epidemiological research.