joineRML

joineRML models the joint distribution of multivariate longitudinal measurements and time-to-event (survival) outcomes to quantify associations between correlated repeated biomarkers and event risk.

Key Features:

• Multivariate Longitudinal Modeling: Implements a multivariate linear mixed-effects sub-model to analyze several correlated repeated measures over time.
• Time-to-Event Analysis: Integrates a Cox proportional hazards regression model with time-varying covariates to relate longitudinal trajectories to event timing.
• Association Structure: Models the link between longitudinal and survival components via a zero-mean multivariate latent Gaussian process.
• Fitting Algorithm: Estimates model parameters using a Monte Carlo Expectation-Maximization (EM) algorithm.
• Inference and Computational Efficiency: Provides approximate standard errors from the empirical profile information matrix, offers bootstrap-based estimation as an alternative, and employs computational strategies to improve fitting speed.

Scientific Applications:

• Clinical Research: Supports analyses where multiple biomarkers are measured longitudinally alongside patient survival or event occurrence, exemplified by studies such as primary biliary cirrhosis.
• Biostatistics: Enables development and evaluation of models that jointly handle complex multivariate longitudinal and time-to-event data structures.

Methodology:

Specifies a multivariate linear mixed-effects model for the longitudinal component and a Cox proportional hazards model with time-varying covariates for the event-time component; links components via a zero-mean multivariate latent Gaussian process; fits parameters with a Monte Carlo EM algorithm and obtains inference from the empirical profile information matrix with bootstrap as an alternative, alongside computational speed-enhancing techniques.