Semiparametric additive time-varying coefficients model for longitudinal data with censored time origin

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Clinical Trial Methodology

Background:

• Longitudinal data analysis is crucial for modeling treatment effects on biomarkers over time.
• Challenges arise in studies, like HIV vaccine trials, where the time origin for biomarker measurements is often unknown or censored for some participants.

Purpose of the Study:

• To develop a statistical model for analyzing longitudinal biomarker data with a censored time origin.
• To investigate a semiparametric additive time-varying coefficient model to capture dynamic treatment effects.
• To apply the developed methodology to HIV vaccine efficacy trial data.

Main Methods:

• Developed weighted profile least squares estimators with kernel smoothing for longitudinal data.
• Employed the expectation-maximization approach to handle censored time origins.
• Utilized failure time regression models (e.g., Cox model) for event time distribution estimation.

Main Results:

• Derived asymptotic properties for parametric and nonparametric estimators, including variance estimators.
• Proposed a two-stage estimation procedure for weight selection to enhance efficiency.
• Simulation studies confirmed the proposed estimators' effectiveness and the method's applicability.

Conclusions:

• The developed statistical methods effectively analyze longitudinal biomarker data with censored time origins.
• The proposed model and estimation techniques are suitable for complex clinical trial settings, including HIV vaccine studies.
• The efficiency of the two-stage procedure is influenced by the distribution of longitudinal error processes.