A Joint Model for (Un)Bounded Longitudinal Markers, Competing Risks, and Recurrent Events Using Patient Registry Data

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Survival Analysis

Background:

• Joint models for longitudinal and survival data are popular but struggle with complex data structures like recurrent and competing events.
• Existing models often assume Gaussian distributions for biomarkers, which is unsuitable for bounded markers, leading to biased results.
• Handling multiple bounded longitudinal markers alongside recurrent and competing events in a single model is challenging.

Purpose of the Study:

• To propose a novel Bayesian shared-parameter joint model.
• To accommodate multiple (possibly bounded) longitudinal markers, recurrent events, and competing risks simultaneously.
• To improve the analysis of complex survival data and biomarker associations.

Main Methods:

• Developed a Bayesian shared-parameter joint model.
• Utilized the beta distribution for bounded longitudinal markers.
• Incorporated recurrent event processes and competing risks.
• Modeled various association forms, discontinuous risk intervals, and gap/calendar timescales.

Main Results:

• A simulation study demonstrated superior performance compared to simpler joint models.
• The model was applied to the US Cystic Fibrosis Foundation Patient Registry.
• Quantified associations between lung function, BMI, and pulmonary exacerbations, accounting for competing risks of death and transplantation.

Conclusions:

• The proposed model effectively handles complex survival data with multiple bounded markers and competing risks.
• It offers more precise insights into disease progression, as shown in the Cystic Fibrosis Foundation Patient Registry analysis.
• The efficient implementation in the R package JMbayes2 facilitates complex analyses.