Joint modelling of longitudinal response and time-to-event data using conditional distributions: a Bayesian...

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Joint models for longitudinal and time-to-event data are crucial in clinical research.
• Existing models often use random-effects (frailty) to link these outcomes.
• There is a need for models that more directly capture the dependency between longitudinal measurements and event occurrence.

Purpose of the Study:

• To propose a new joint model integrating a linear mixed-effects model for longitudinal data and an accelerated failure time model for time-to-event data.
• To establish a direct link between longitudinal measurements and time-to-event outcomes via a latent random process.
• To develop and evaluate a Bayesian estimation method for this novel joint model.

Main Methods:

• A joint model combining a linear mixed-effects model and an accelerated failure time model.
• Linking the two sub-models through a latent random process to model dependency.
• Bayesian estimation using standard priors, with computations performed in OpenBUGS.
• Simulation studies for model evaluation and comparison with conditional and locally independent models.

Main Results:

• The proposed joint model effectively captures the direct dependency between longitudinal measurements and time-to-event outcomes.
• The Bayesian estimation method provides a viable approach for model parameter estimation.
• The simulation study demonstrates the model's performance and calibration properties.
• The model was successfully applied to analyze data from Duchenne muscular dystrophy and AIDS patients.

Conclusions:

• The novel joint model offers a more direct way to understand the relationship between longitudinal changes and event times.
• The Bayesian framework and OpenBUGS implementation facilitate practical application of the model.
• This approach enhances the analysis of complex clinical data, as shown in the DMD and AIDS patient studies.