Forecasting Enrollment in Clinical Trials When Site-Level Accrual Rates Vary with Time

• Activity Number: 318
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yannis Jemiai*+ and Nitin Patel and Suresh Ankolekar and Pralay Senchaudhuri
• Companies: Cytel Inc. and Cytel Inc. and Maastricht School of Management and Cytel Inc.
• Keywords: Enrollment ; Forecasting ; Recruiting ; Gamma-Poisson ; Time varying enrollment rates ; Bayesian model

Abstract

We have studied the performance of the Gamma-Poisson model of Anisimov and Federov (Stat Med, 2007) in forecasting enrollment using historical data from several Phase 2 and Phase 3 trials. We have focused on its ability to improve forecasts of trial completion time as the trial progresses. We find that while it is often adequate, there are several instances when the assumption of a constant Poisson rate for each center leads to substantial errors. We extend the model to accommodate time varying rates using a state-space model proposed by Harvey and Fernandes (J Bus Econ Stat, 1989) and report results from using this model on the historical data. We also examine a simpler approach using tracking errors to trigger user intervention to improve the forecasts.