A flexible B-spline model for multiple longitudinal biomarkers and survival

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Jointly modeling longitudinal and survival data is crucial in many fields.
• Traditional linear models often fail to capture complex longitudinal trends.
• Nonlinear patterns in longitudinal markers can significantly impact survival outcomes.

Purpose of the Study:

• To propose a novel joint longitudinal and survival model accommodating nonlinear longitudinal trajectories.
• To effectively link nonparametric longitudinal markers to survival data.
• To provide a robust method for analyzing complex biomedical data.

Main Methods:

• Utilized a nonparametric model for longitudinal markers employing cubic B-splines.
• Incorporated a proportional hazards model to connect longitudinal measures with the hazard rate.
• Employed Markov chain Monte Carlo (MCMC) for model fitting.
• Selected optimal model complexity using Conditional Predictive Ordinate (CPO) and Deviance Information Criterion (DIC).

Main Results:

• The proposed cubic B-spline model demonstrated superior fit to longitudinal data compared to parametric models.
• The joint model successfully elucidated the relationship between viral load, CD4 count, and time to event in AIDS clinical trial data.
• Simulation studies validated the method and model selection approach.

Conclusions:

• The nonparametric joint model offers a flexible and accurate approach for analyzing longitudinal and survival data with complex underlying patterns.
• This method enhances understanding of disease progression and treatment effects by integrating dynamic biomarker changes with event times.
• The approach is particularly valuable for HIV/AIDS research and other areas with similar data structures.