Semiparametric regression analysis of failure time data with dependent interval censoring

Area of Science:

• Biostatistics
• Survival Analysis
• Clinical Data Analysis

Background:

• Interval-censored failure-time data are common in clinical studies, where exact failure times are unknown.
• Standard methods often assume independence between examination times and failure times, which may not hold in practice.
• Patient health status can influence both examination frequency and failure risk, leading to dependent interval-censoring.

Purpose of the Study:

• To develop a statistical model for dependent interval-censored failure-time data.
• To address the challenge of unobserved latent variables affecting both failure time and examination processes.
• To propose a robust method for analyzing complex clinical data where patient health influences follow-up schedules.

Main Methods:

• A joint frailty model incorporating a shared gamma frailty term is proposed.
• The model links a Cox model for failure time and a proportional intensity model for the recurrent visiting/examination process.
• A semiparametric maximum likelihood estimation approach is used to estimate model parameters.

Main Results:

• The proposed joint frailty model effectively accounts for the association between failure time and the visiting process in interval-censored data.
• Asymptotic properties, including consistency and weak convergence of the estimators, are established.
• The model demonstrated utility in analyzing a bladder cancer dataset.

Conclusions:

• The joint frailty model provides a powerful tool for analyzing dependent interval-censored failure-time data, particularly when unobserved factors are present.
• This approach offers improved accuracy in survival analysis for clinical scenarios with complex patient monitoring.
• The methodology is applicable to various fields requiring analysis of time-to-event data with dependent censoring mechanisms.