A two-level copula joint model for joint analysis of longitudinal and competing risks data.

Area of Science:

• Biostatistics
• Clinical Data Analysis
• Survival Analysis

Background:

• Clinical data often involves multiple longitudinal outcomes and event-times with competing risks.
• Accurate modeling of the complex dependencies within such data is crucial for reliable inference.
• Existing methods may struggle with the joint analysis of disparate longitudinal and time-to-event data.

Purpose of the Study:

• To propose a flexible two-level copula joint model for analyzing clinical data with multiple longitudinal outcomes and competing risks.
• To incorporate conditional dependence between longitudinal outcomes and event-times.
• To accommodate skewed data and analyze covariate effects on outcome quantiles using linear quantile mixed models.

Main Methods:

• A two-level Gaussian copula model was developed to link submodels for longitudinal data and event-times.
• Linear quantile mixed models were used for continuous longitudinal outcomes to handle skewness and quantile-specific effects.
• Bayesian inference with Markov Chain Monte Carlo (MCMC) sampling was employed for model estimation.

Main Results:

• The proposed copula joint model demonstrated superior performance over conventional methods assuming conditional independence in simulation studies.
• The new model achieved smaller biases and better coverage probabilities for Bayesian credible intervals.
• The method was successfully illustrated using clinical data from renal transplantation.

Conclusions:

• The proposed two-level copula joint model provides a robust framework for analyzing complex clinical data with multiple longitudinal outcomes and competing risks.
• This approach effectively captures the intricate dependencies between longitudinal processes and time-to-event outcomes.
• The model offers improved accuracy and reliability for clinical data analysis, outperforming methods that assume conditional independence.