Some issues in predicting patient recruitment in multi-centre clinical trials

Area of Science:

• Clinical trial methodology
• Statistical modeling
• Biostatistics

Background:

• Previous models by Anisimov and Fedorov used Gamma-Poisson mixtures (negative binomial) for forecasting trial recruitment and completion times.
• These models effectively link frequency and time domains, correlating negative binomial distributions with Type VI Pearson distributions.
• Forecasting is crucial not only for ongoing trials but also for those not yet initiated.

Purpose of the Study:

• To extend existing models for forecasting clinical trial completion times by introducing a higher hierarchical level.
• To develop a method for forecasting completion times for trials that have not yet started.
• To present a practical approach using an orthogonal parameterization of the Gamma distribution.

Main Methods:

• Utilized a hierarchical modeling approach with an orthogonal parameterization of the Gamma distribution.
• Modeled the two parameters of the Gamma distribution separately.
• Applied the method to data from 18 multi-centre clinical trials.

Main Results:

• The proposed hierarchical model provides a framework for forecasting trial completion times, particularly for new trials.
• Demonstrated the application of the orthogonal parameterization of the Gamma distribution in this context.
• The method was successfully illustrated using real-world clinical trial data.

Conclusions:

• The developed hierarchical model offers a valuable extension for predicting clinical trial durations.
• This approach enhances the ability to forecast completion times for trials at any stage, including those yet to commence.
• The method provides practical insights for improving clinical trial planning and resource allocation.