Quantile regression models with multivariate failure time data

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Mean regression models are limited for correlated failure time data.
• Multivariate failure time data with intracluster correlation is prevalent in biomedical research.
• Existing methods often fail to adequately address correlated survival data.

Purpose of the Study:

• To adapt quantile regression for right-censored correlated survival data.
• To develop an estimating equation approach for parameter estimation.
• To enhance efficiency using a weighted version of the estimating equation.

Main Methods:

• Utilizing an estimating equation approach under the working independence assumption.
• Implementing a weighted version for improved statistical efficiency.
• Employing nonparametric functional density estimation for variance estimation.
• Applying bootstrap and perturbation resampling for variance-covariance matrix estimation.

Main Results:

• Parameter estimates are consistent.
• Asymptotic distributions of parameter estimates are normal.
• The proposed methods demonstrate reliable performance in simulation studies.
• The approach is validated using clinical trial data for otitis media.

Conclusions:

• The developed methods effectively handle right-censored correlated survival data.
• The estimating equation approach provides a robust framework for quantile regression in clustered settings.
• The study offers a valuable tool for analyzing complex survival data in biomedical research.
• The findings are supported by both simulation and real-world clinical data analysis.