Bayesian optimal designs for pharmacokinetic models: sensitivity to uncertainty.

Area of Science:

• Pharmacometrics
• Mathematical Modeling
• Statistical Design of Experiments

Background:

• Bayesian optimal design is crucial for efficient pharmacokinetic (PK) studies.
• Understanding the impact of prior uncertainty on design selection is essential for robust study planning.

Purpose of the Study:

• To investigate the sensitivity of Bayesian optimal designs for PK models to the magnitude of prior uncertainty.
• To analyze how design characteristics change with varying levels of prior knowledge.

Main Methods:

• Utilized two- and three-parameter PK models with fixed and mixed effects.
• Applied two Bayesian optimal design criteria: Efficiency criterion (ED) and Approximate Profile Information (API).
• Examined different error weighting schemes and varied the magnitude of prior uncertainty.

Main Results:

• Increased prior uncertainty generally led to the addition of unique design points compared to local designs.
• The emergence of additional design points occurred gradually, forming distinct bifurcation patterns.
• These observed bifurcation patterns indicate a high degree of sensitivity in the design to the magnitude of prior uncertainty.

Conclusions:

• The number and placement of design points in Bayesian optimal designs are highly sensitive to the magnitude of prior uncertainty.
• Bifurcation patterns serve as a visual indicator of this sensitivity in pharmacokinetic model design.
• These findings highlight the importance of carefully assessing prior uncertainty in the application of Bayesian optimal design methods.