Interim recruitment prediction for multi-center clinical trials

Area of Science:

• Clinical trial operations
• Statistical modeling
• Biostatistics

Background:

• Existing time-homogeneous models for multi-center clinical trial recruitment often yield overly optimistic and narrow prediction intervals.
• There is a need for more accurate and robust methods to monitor, model, and predict patient recruitment in complex trial settings.

Purpose of the Study:

• To introduce a general framework for monitoring, modeling, and predicting patient recruitment in multi-center clinical trials.
• To address the limitations of existing models by accounting for decaying recruitment rates and model uncertainty.

Main Methods:

• Development of two tests for detecting decay in recruitment rates.
• Introduction of an inhomogeneous Poisson process model with a monotonically decaying intensity, adaptable to various parametric curve-shapes.
• Utilizing Bayesian model averaging for predictions, incorporating parameter and model uncertainty.

Main Results:

• The proposed framework demonstrates improved prediction accuracy compared to existing methods in simulated and real oncology trial data.
• The methodology effectively identifies unexpected changes and patterns in patient accrual.
• The tests for recruitment rate decay show reliable performance in power studies.

Conclusions:

• The new framework provides a more realistic and robust approach to predicting clinical trial recruitment, particularly for multi-center studies.
• The methodology enhances the ability to monitor trial progress and detect deviations from expected recruitment patterns.
• This work offers valuable tools for optimizing clinical trial planning and management.