A simple and robust model for enrollment projection in clinical trials

Area of Science:

• Statistics
• Clinical Trial Management

Background:

• Enrollment projection in clinical trials is increasingly important.
• Existing statistical methods can be complex to implement.

Purpose of the Study:

• To implement a simple, robust empiric Bayes Poisson Gamma model (PGM) for practical clinical trial enrollment projection.
• To evaluate the PGM's performance in simulations and real-world oncology trials.

Main Methods:

• Utilized an empiric Bayes approach with a Poisson Gamma model (PGM).
• Assumed constant, site-specific enrollment rates from a common Gamma distribution.
• Employed data-driven prior parameter selection.

Main Results:

• The PGM demonstrated satisfactory performance in simulations and oncology trials.
• Compared to a nonparametric model, the PGM yielded narrower credible intervals due to its parametric assumptions.
• Model predictions may be less accurate when assumptions are substantially violated.

Conclusions:

• The proposed Poisson Gamma model (PGM) is a practical and robust tool for clinical trial enrollment forecasting.
• The PGM's parametric nature provides precision but requires careful consideration of its underlying assumptions.
• Further research may explore model adaptations for scenarios with violated assumptions.