Probability-Guided 3+3 Design in Phase I Dose-Escalation Study

• Activity Number: 31
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yung-Seop Lee*+ and Jung Wook Park
• Companies: Dongguk Univ. and Astellas Pharma Global Development, Inc.
• Keywords: exponential distribution ; weibull distribution ; 3+3 design

Abstract

In recent years, there has been emerging interest in improving the performance of Phase I (P1) dose escalation study design to accurately estimate maximum tolerate dose (MTD), particularly. However, majority of the P1 dose escalation studies were designed based on traditional algorithm-based design such as 3+3 design. Algorithm based design tends to underestimate true MTD since it does not use the information from the previously cleared cohorts. A model-based design such as continual reassessment method (CRM) under Bayesian framework has gained a lot of interest but it requires appropriate prior information, which may not be available in first time in human study, and substantial computation. Therefore, many Phase I studies still use 3+3 design for its operational simplicity. Here we introduce simple frequentist approach to improve dose escalation decision based on probability distribution in 3+3 design setting in order to maintain its operational simplicity. We use exponential and weibull distribution with maximum likelihood estimation to guide the dose escalation decision. Simulation study will show the performance improvement over the traditional 3+3 design.