Joint analysis of recurrence and termination: A Bayesian latent class approach

Area of Science:

• Biostatistics
• Clinical Research
• Transplantation Medicine

Background:

• Transplant recipients face risks of non-fatal rejections and graft failure (death).
• Accurate modeling is crucial for understanding covariate effects on these dual event trajectories.

Purpose of the Study:

• To develop a semiparametric latent-class-based joint model for analyzing recurrent events and death in transplantation.
• To provide a Bayesian framework for coherent interpretation of covariate effects (e.g., race, gender) on event trajectories.
• To offer practical Bayesian methods for estimation and prediction, accommodating complete or limited prior information.

Main Methods:

• Developed a semiparametric latent-class-based joint model.
• Employed a fully Bayesian approach with complete prior specification for baseline functions.
• Derived a partial likelihood-based semiparametric Bayesian approach for analyses with limited prior information.
• Accommodated fixed and time-varying covariates.
• Utilized Markov Chain Monte Carlo (MCMC) tools implemented in publicly available software.

Main Results:

• The proposed joint model provides coherent interpretation of covariate effects on all relevant functions and model quantities.
• The Bayesian methods demonstrate practical advantages and improved performance compared to competing methods.
• The model and methods are applicable to transplant studies and general survival analyses with recurrent events and death.

Conclusions:

• The novel Bayesian joint model offers a robust and interpretable framework for analyzing complex event data in transplantation.
• The developed methods provide practical advantages for estimation and prediction, enhancing clinical decision-making.
• The approach is versatile, accommodating various covariate types and prior information scenarios.