A Bayesian approach to joint analysis of multivariate longitudinal data and parametric accelerated failure time

Area of Science:

• Biostatistics
• Neurology
• Clinical Trials

Background:

• Parkinson's disease (PD) causes progressive, multidimensional impairment affecting sensory, functional, and cognitive domains.
• Assessing PD progression is challenging due to its complexity, requiring multiple longitudinal outcomes.
• Clinical trial endpoints are often complicated by terminal events like death or dropout, which can depend on patient measurements.

Purpose of the Study:

• To develop and evaluate a joint random-effects model for correlated longitudinal outcomes and time-to-event data in Parkinson's disease.
• To address the limitations of single outcome measures and the dependence of event times on longitudinal disease markers.
• To provide a robust statistical framework for analyzing complex clinical trial data in neurodegenerative diseases.

Main Methods:

• A multilevel item response theory model was employed for multivariate longitudinal outcomes.
• A parametric accelerated failure time model was used for time-to-event analysis, accommodating violations of the proportional hazard assumption.
• The models were integrated via shared random effects, with Bayesian inference implemented using Markov Chain Monte Carlo (MCMC) in 'BUGS'.

Main Results:

• The proposed joint modeling approach effectively handles correlated longitudinal data and time-to-event outcomes in the context of Parkinson's disease.
• Simulation studies demonstrated the validity and performance of the statistical methodology.
• Application to the DATATOP study provided insights into the progression of Parkinson's disease and treatment effects.

Conclusions:

• The joint random-effects model offers a powerful tool for analyzing complex, multidimensional data in Parkinson's disease clinical trials.
• This approach enhances the ability to accurately assess treatment efficacy by integrating diverse outcome measures.
• The methodology is broadly applicable to other complex longitudinal studies in medicine and public health.