Flexible Bayesian additive joint models with an application to type 1 diabetes research

Area of Science:

• Biostatistics
• Epidemiology
• Medical Informatics

Background:

• Joint modeling of longitudinal and time-to-event data is increasingly used to understand biomarker-event associations.
• Investigating the link between fluctuating disease markers and disease progression is crucial for understanding diseases like type 1 diabetes.

Purpose of the Study:

• To develop a flexible additive joint modeling framework for analyzing longitudinal biomarker data and time-to-event outcomes.
• To investigate the relationship between autoantibodies and the progression to type 1 diabetes.

Main Methods:

• Developed a general framework for flexible additive joint models incorporating smooth nonlinear, time-varying, and random effects.
• Utilized Bayesian P-splines to model nonlinear subject-specific biomarker trajectories and time-varying associations.
• Estimated the model within a Bayesian framework and implemented it using the R-package bamlss.

Main Results:

• The proposed framework effectively captures complex, nonlinear biomarker trajectories.
• Demonstrated the ability to model time-varying associations between biomarkers and disease event processes.
• Provided new insights into the progression of autoimmune diseases, using type 1 diabetes as a case study.

Conclusions:

• Flexible additive joint models offer a powerful approach for analyzing complex biomarker-disease event relationships.
• The developed methods provide deeper understanding of disease progression dynamics.
• The R-package bamlss facilitates the application of these advanced statistical techniques.