A Bayesian Location-Scale Joint Model for Time-To-Event and Multivariate Longitudinal Data With Association Based on...

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Within-individual variability in health indicators is crucial for understanding disease progression.
• Traditional methods using summary statistics for longitudinal data can lead to biased results in survival models.
• Existing joint models often assume constant variance, failing to capture dynamic changes.

Purpose of the Study:

• To develop and validate a novel joint model that incorporates within-individual variability of longitudinal markers.
• To accurately estimate the association between dynamic health marker variability and time-to-event outcomes.
• To address regression dilution bias inherent in standard survival analysis methods.

Main Methods:

• A mixed-effect location-scale model was employed to analyze longitudinal biomarkers, their within-individual variability, and correlations.
• A proportional hazard regression model with a flexible baseline hazard was used for time-to-event outcomes.
• The proposed joint model shares information from longitudinal biomarkers as a function of random effects.

Main Results:

• The novel joint model demonstrated superior performance compared to standard joint models with constant variance in simulation studies.
• The model effectively quantifies within-individual variability for longitudinal markers.
• The model successfully evaluated the association between lung function, malnutrition, and mortality in cystic fibrosis patients.

Conclusions:

• The developed joint model provides a robust framework for analyzing longitudinal data and time-to-event outcomes, accounting for within-individual variability.
• This approach offers improved accuracy in estimating hazard ratios and understanding disease progression.
• The findings have significant implications for clinical research and patient management, particularly in chronic diseases like cystic fibrosis.