Nonparametric analysis of dependently interval-censored failure time data

Area of Science:

• Biostatistics
• Survival Analysis
• Longitudinal Data Analysis

Background:

• Observational cohort studies frequently encounter irregular intermittent observations, leading to interval-censored failure times.
• Dependently interval-censored failure times arise when observation times correlate with an individual's failure time.
• Existing inverse-intensity-of-visit weighting methods have been adapted for longitudinal data and parametric failure time analysis.

Purpose of the Study:

• To develop nonparametric estimation methods for failure time distributions with dependent interval censoring.
• To address challenges in analyzing irregularly observed longitudinal data in survival studies.
• To provide a methodology applicable to real-world cohort studies, such as the Toronto Psoriatic Arthritis Cohort Study.

Main Methods:

• Utilized weighted generalized estimating equations for nonparametric estimation.
• Employed monotone smoothing techniques to enhance estimation accuracy.
• Conducted simulations to evaluate the finite sample performance of the proposed estimators.

Main Results:

• The proposed nonparametric estimators demonstrate reliable performance in simulations.
• The methodology effectively handles dependent interval-censored failure times.
• Successful application of the methods to the Toronto Psoriatic Arthritis Cohort Study.

Conclusions:

• The developed weighted generalized estimating equations and monotone smoothing offer a robust nonparametric approach for failure time analysis with dependent interval censoring.
• This methodology improves the analysis of complex longitudinal data in observational studies.
• The findings have significant implications for survival analysis in clinical and epidemiological research.