Dimension of model parameter space and operating characteristics in adaptive dose-finding studies

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacometrics

Background:

• Adaptive, model-based dose-finding methods are crucial in clinical trials.
• Debate exists on using parsimonious (simple) versus comprehensive (complex) models.

Purpose of the Study:

• To investigate the impact of parameter space dimensionality on adaptive dose-finding study performance.
• To determine if increasing model complexity improves or hinders study outcomes.

Main Methods:

• Theoretical analysis of parameter estimation and study behavior.
• Simulations to evaluate operating characteristics under different model complexities.

Main Results:

• Increasing model dimensionality often leads to counterproductive outcomes.
• Difficulties include inconsistent estimates, erratic escalation/de-escalation, and suboptimal dose targeting.

Conclusions:

• Parsimonious models are generally superior for adaptive dose-finding.
• Overly complex models can compromise study efficiency and accuracy.