A Bayesian multivariate joint frailty model for disease recurrences and survival

Area of Science:

• Biostatistics
• Medical Statistics
• Survival Analysis

Background:

• Cancer recurrence and survival are critical outcomes in oncology.
• Simultaneous analysis of multiple time-to-event outcomes is complex.
• Existing models may not adequately capture correlations between different recurrence types and death.

Purpose of the Study:

• To develop a multivariate frailty hazard model for joint analysis of local recurrence, distant recurrence (metastasis), and death.
• To investigate treatment effects on these correlated outcomes.
• To quantify the impact of local and distant recurrences on patient survival and their inter-correlation.

Main Methods:

• A multivariate frailty hazard model was developed for joint modeling of three time-to-event outcomes.
• The model was implemented within a Bayesian framework.
• A hybrid Monte Carlo algorithm, incorporating gradient evaluation and guided walk progress, was used for posterior distribution computation, allowing simultaneous updates of multivariate state vectors.

Main Results:

• The proposed joint frailty model successfully integrates the analysis of treatment effects, recurrence-specific mortality risks, and recurrence correlations.
• Simulation studies validated the model's performance and the computational algorithm's efficiency.
• The model provided insights into the complex interplay of disease recurrence and survival in the soft tissue sarcoma data.

Conclusions:

• The joint frailty model offers a unified approach to address key questions in cancer research regarding recurrence and survival.
• This methodology enhances understanding of disease progression and treatment efficacy by modeling correlated events.
• The Bayesian framework and hybrid Monte Carlo algorithm provide a robust computational solution for complex survival data analysis.