Modeling Enrollment with Random Staggered Site Start up Times

• Activity Number: 66
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Bradley Ferguson*+ and Vladimir Anisimov and Valerii Fedorov
• Companies: North Carolina State University and Quintiles and Quintiles
• Keywords: enrollment ; projection ; clinical ; clinical ; trial ; forecast

Abstract

With rising competition between pharmaceutical companies and increasingly strict drug regulations, running clinical trials efficiently and effectively is crucial for the drug development process. Enrolling patients is a vital and costly part of that process and being able to accurately model and predict enrollment is of utmost importance. Many current enrollment models are deterministic and fail to account for various sources of variability in the enrollment process. This could lead to missed deadlines and unmet enrollment goals. In this paper, we propose a predictive model in which patients arrive at sites according to Poisson processes with site specific enrollment rates and start up times. This approach has advantages over global predictive models as it provides the opportunity to account for differences in site start-up (SSU) times and enrollment rates. We treat the enrollment rates and SSU times as possibly random quantities and provide closed form expressions of enrollment moments which avoid Monte-Carlo simulation. Our approach also allows us to perform re-projections using a Bayesian re-calculation of the posterior enrollment rates.