Simultaneous clustering and joint modeling of multivariate binary longitudinal and time-to-event data.

Area of Science:

• Biostatistics
• Medical Statistics
• Clinical Trial Analysis

Background:

• Joint modeling of longitudinal and time-to-event data is crucial in medical research.
• Heterogeneous populations necessitate clustering for robust statistical inference.
• Existing methods may not adequately address complex patient subgroupings.

Purpose of the Study:

• To develop a Bayesian joint modeling framework incorporating clustering for multivariate binary longitudinal outcomes and time-to-event data.
• To analyze a clinical trial dataset from cancer patients to identify distinct subgroups.
• To assess the impact of identified clusters on relapse prediction.

Main Methods:

• Utilized Bayesian data-augmentation for latent continuous outcomes from multivariate binary longitudinal data.
• Employed Bayesian consensus clustering to identify patient subgroups.
• Performed cluster-specific joint analysis using generalized linear mixed models and proportional hazards models.
• Applied the method to a cancer clinical trial dataset with biomarker measurements and relapse times.

Main Results:

• Identified three distinct latent patient clusters.
• Demonstrated substantial differences in covariate effects and median non-relapse probabilities across clusters.
• Simulation studies confirmed the effectiveness of the simultaneous clustering and joint modeling approach.

Conclusions:

• The proposed Bayesian framework effectively clusters heterogeneous patient populations within joint modeling.
• This approach enhances statistical inference and provides more precise predictions for time-to-event outcomes.
• The findings have significant implications for personalized medicine and clinical trial design in oncology.