joineRML: a joint model and software package for time-to-event and multivariate longitudinal outcomes

Area of Science:

• Biostatistics
• Statistical Modeling
• Medical Informatics

Background:

• Joint modeling of longitudinal and time-to-event data is increasingly important.
• Existing statistical software is often limited to single longitudinal outcomes.
• There is a need for methods handling multiple longitudinal outcomes.

Purpose of the Study:

• To extend classical joint models to accommodate multiple longitudinal outcomes.
• To propose a practical algorithm for fitting these multivariate joint models.
• To introduce the R package joineRML for implementing these models.

Main Methods:

• A multivariate linear mixed model for longitudinal data and a Cox proportional hazards model for time-to-event data.
• A multivariate latent Gaussian process to link the longitudinal and event time sub-models.
• Monte Carlo Expectation-Maximisation algorithm with empirical profile information matrix for fitting.

Main Results:

• Demonstration of a multivariate joint model applied to primary biliary cirrhosis data.
• Illustration using three repeatedly measured biomarkers.
• Comparison of Monte Carlo EM with bootstrap estimation for model inference.

Conclusions:

• An open-source R package, joineRML, is now available for fitting multivariate joint models.
• The package incorporates computational speed enhancements.
• Facilitates advanced statistical analysis of complex health data.