Adjusting for intercurrent events using Bayesian joint models for longitudinal outcomes in clinical trials.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Longitudinal Data Analysis

Background:

• Intercurrent events complicate clinical trial endpoint interpretation and measurement.
• Terminal events can preclude complete longitudinal outcome assessment, leading to biased estimates if not properly handled.
• A robust methodology is essential for managing outcome-related terminal intercurrent events.

Purpose of the Study:

• To propose and evaluate a Bayesian joint modeling approach for handling terminal intercurrent events in clinical trials.
• To improve the accuracy and reliability of treatment effect estimation in the presence of incomplete outcome data.
• To provide a principled methodology for both the design and analysis phases of clinical trials.

Main Methods:

• Developed a Bayesian joint model analyzing longitudinal outcomes and terminal events simultaneously using shared random effects.
• Employed multiple discrete-time survival submodels to accommodate diverse event types.
• Conducted extensive simulations mimicking clinical trials with competing risks (e.g., recovery and death).

Main Results:

• The proposed Bayesian joint modeling approach demonstrated superior statistical power compared to methods ignoring intercurrent events.
• Power increased by approximately 15% in scenarios with substantial bias from intercurrent events.
• Joint modeling effectively reduced bias caused by terminal intercurrent events.

Conclusions:

• The Bayesian joint modeling approach is effective for addressing terminal intercurrent events in clinical trial design and analysis.
• Explicitly accounting for event-related truncation of longitudinal follow-up enhances precision and reliability of treatment effect estimates.
• This methodology improves the interpretation of clinical trial outcomes when measurements are incomplete due to terminal events.