Regression analysis of failure time data with informative interval censoring

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Interval-censored failure time data occurs when subjects miss visits, preventing exact event time determination.
• Standard methods often assume censoring is independent of failure time, which may not always be true.
• Dependent censoring can bias results in survival data analysis.

Purpose of the Study:

• To propose a novel regression method for interval-censored failure time data.
• To address situations where the censoring mechanism is dependent on the true failure time.
• To model the dependence structure using latent variables.

Main Methods:

• Utilized a proportional hazards frailty model to account for dependent censoring.
• Employed the Expectation-Maximization (EM) algorithm for parameter estimation.
• Investigated the dependence structure between censoring variables and failure time via latent variables.

Main Results:

• Developed and presented a new method for regression analysis of dependent interval-censored failure time data.
• Examined the finite sample properties of the proposed estimators through simulation studies.
• Demonstrated the method's applicability using real-world data from an AIDS study.

Conclusions:

• The proposed method effectively handles interval-censored failure time data with dependent censoring.
• The proportional hazards frailty model and EM algorithm provide a robust estimation framework.
• The approach offers a valuable tool for analyzing complex survival data, as shown in the AIDS study example.