A Bayesian Up-and-Down Design for Clinical Trials with Multiple Adverse Events in Opposite Directions

• Activity Number: 609
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Yiyi Chen*+ and Zunqiu Chen and Matthew Halsey
• Companies: Oregon Health & Science University and Oregon Health & Science University and Oregon Health & Science University
• Keywords: Clinial Trials ; Bayesian ; Adaptive Design ; Up-and-Down

Abstract

Displaced supracondylar humerus (SCH) fractures are common in children, typically treated by closed reduction percutaneous pinning (CRPP) if displaced. While the timing of pin removal is critical for treatment success, no guideline or recommendation on this is available. This is partially because it is not ethical to randomize patients to removing pins at a time that is different from convention. Three potential "adverse events" are associated with the timing of pin removal for SCH fractures in children: Infection, Displacement and Stiffness. The risk of having infection and/or stiffness increases with longer durations of pin placement while early pin removal may lead to displacement. We propose a two-stage Bayesian group sequential design to walk up-and-down based on the most up-to-date observations from the study until concluding in the optimal time for pin removal. Five candidate time windows (19-21, 22-24, 25-27, 28-30, 31-35 days) are considered, and a weighted loss function is used to determine the next stochastic walk movement: window escalation (E), de-escalation (D) or staying (S) at the current window. The trial is allowed to stop early once "equilibrium" window has been found. Simulation study is conducted to find the expected sample size and the operating characteristics of the design.