Assessing Importance of Biomarkers: a Bayesian Joint Modeling Approach of Longitudinal and Survival Data with...

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Oncology

Background:

• Longitudinal biomarkers, including patient-reported outcomes (PROs) and quality of life (QOL), are crucial in cancer clinical trials.
• Integrating PRO/QOL data with survival analysis offers deeper insights into symptom-survival relationships.
• Existing models may not fully capture the complexities of disease progression and overall survival.

Purpose of the Study:

• To develop advanced trajectory-based models for joint analysis of longitudinal PRO/QOL and survival data in cancer.
• To specifically address disease progression and overall survival, accounting for factors like treatment switching.
• To assess the contribution of longitudinal data to the overall model fit using Bayesian methods.

Main Methods:

• Proposed a class of mixed-effects regression models for longitudinal measures.
• Incorporated a cure rate model for disease progression time and a Cox proportional hazards model for overall survival time.
• Utilized a semi-competing risks framework, with disease progression as a nonterminal event and death as a terminal event.
• Employed Bayesian inference to derive decompositions of DIC and LPML for model fit assessment.

Main Results:

• Developed methods to assess the fit of longitudinal and survival components separately.
• Introduced ΔDIC and ΔLPML metrics to quantify the impact of longitudinal data on survival model fit.
• Demonstrated the utility of the proposed models in a head and neck cancer clinical trial context.

Conclusions:

• The proposed joint modeling framework effectively integrates longitudinal PRO/QOL data with survival outcomes.
• The developed Bayesian metrics provide valuable tools for model evaluation and data contribution assessment.
• These models offer a robust approach for comparative assessment of patient-reported changes and their impact on survival in cancer research.