Bayesian variable selection and estimation in semiparametric joint models of multivariate longitudinal and survival...

Area of Science:

• Biostatistics
• Statistical Modeling
• Survival Analysis

Background:

• Joint modeling of longitudinal and survival data is crucial for understanding disease progression and patient outcomes.
• Existing models often rely on restrictive assumptions, such as normality of longitudinal data.
• The impact of longitudinal response history on dropout risk requires further investigation.

Purpose of the Study:

• To propose a novel semiparametric joint model for multivariate longitudinal and survival data (SJMLS).
• To relax the normality assumption for longitudinal outcomes and leave baseline hazard functions unspecified.
• To incorporate the effect of longitudinal response history on dropout risk.

Main Methods:

• Developed a semiparametric joint model for SJMLS.
• Employed Bayesian penalized splines to approximate unspecified baseline hazard functions.
• Utilized a Bayesian Lasso (BLasso) method, combining Gibbs sampler and Metropolis-Hastings algorithm, for parameter estimation and covariate selection.

Main Results:

• The proposed BLasso method effectively estimates parameters and selects important covariates in SJMLS.
• Simulation studies demonstrate the good finite sample performance of the developed techniques.
• The model successfully handles non-normal longitudinal data and unspecified baseline hazards.

Conclusions:

• The novel semiparametric joint model provides a flexible framework for analyzing complex longitudinal and survival data.
• The BLasso method offers a robust approach for simultaneous estimation and covariate selection in SJMLS.
• The methodology is validated through simulations and a real-world application in breast cancer research.