Regression analysis of additive hazards model with sparse longitudinal covariates

Area of Science:

• Biostatistics
• Survival Analysis
• Longitudinal Data Analysis

Background:

• Additive hazards models complement proportional hazards models for failure time data.
• Analyzing these models with time-dependent longitudinal covariates is challenging due to data requirements and potential biases from common estimation methods.

Purpose of the Study:

• To develop a novel statistical method for unbiased regression coefficient estimation in additive hazards models with intermittently observed longitudinal covariates.
• To address the limitations of existing methods like last value carried forward and joint modeling.

Main Methods:

• Proposed a weighting approach based on the distance between covariate observation times and failure times.
• Established theoretical properties including consistency and asymptotic normality of the proposed estimators.

Main Results:

• The proposed weighting method yields unbiased regression coefficient estimation.
• Simulation studies confirmed the theoretical findings, demonstrating numerical support.

Conclusions:

• The developed methodology offers a practical and statistically sound approach for analyzing failure time data with longitudinal covariates.
• The method's utility is demonstrated through an application in an Alzheimer's disease study.