NHLBI ICTR Webinar #2, Arm dropping and response adaptive randomization

Presented by Kert Viele

Multiple arm studies, including comparative effectiveness, factorial designs, dose ranging, and platform trials, allow for the exploration of complex questions within a single study. Allocating large numbers of patients to each of multiple arms, however, can quickly become expensive. Fortunately, early signals in the data can be used to eliminate poorly performing arms and focus resources, allowing efficient identification of the better arms while simultaneously improving the outcomes for patients within the study.

In this webinar Dr. Kert Viele will discuss arm dropping and adaptive randomization. In arm dropping, arms are completely eliminated from the study when they fail to meet criteria such as being unlikely to beat control or being unlikely to be the best arm. Adaptive randomization is less stringent, adjusting allocation probabilities up or down depending on an arm’s performance.

He will present examples, illustrate several criteria used for arm dropping and adaptive randomization, discuss auxiliary variables such as the timing of interim analyses and aggressiveness of adaptation, and discuss methods for avoiding pitfalls such as adapting too early and “locking in” erroneous early signals. At the end of the webinar, attendees should understand the motivation behind each adaptation, the potential efficiency gains, and methods for exploring alternatives and obtaining a productive design.

Originally broadcast on Thursday, November 15, 2018 at 3pm ET