D optimal designs for three Poisson dose-response models.

Area of Science:

• Biostatistics
• Pharmacometrics
• Clinical Trial Design

Background:

• Phase II dose-ranging studies are critical for selecting optimal doses for Phase III trials.
• Count data is common in clinical research and can be modeled using Poisson distributions with random effects.
• D optimal designs aim to maximize information gained from experiments, crucial for efficient dose selection.

Purpose of the Study:

• To identify and evaluate D optimal designs for three Poisson dose-response models.
• To assess the performance of these designs across various parameter scenarios.
• To provide a framework for optimizing dose selection in Phase II studies with count data.

Main Methods:

• Investigated three Poisson dose-response models with increasing complexity.
• Employed an E(max) model to characterize drug effects.
• Determined D optimal designs and assessed performance using parameter precision and %CV of ED(50).

Main Results:

• D optimal designs were consistent across models and scenarios, typically including placebo, maximum dose, and a dose near the ED(50).
• Design performance varied substantially, with %CV for ED(50) ranging from 1.4% to 91% in a 1,000-subject simulation.
• Performance improved with higher baseline counts, reduced random effects, and larger E(max) values.

Conclusions:

• The identified D optimal designs provide a robust starting point for Phase II dose-ranging studies.
• Careful consideration of model parameters and patient variability is essential for optimizing study performance.
• This framework aids in designing efficient trials for count data endpoints.