Biostatistics Seminar Series

Biostatistics Seminar

Thursday 2/4 3:30PM - 4:30PM

A115 Crabtree

Marnie Bertolet, PhD Pitt Public Health Assistant Professor of Epidemiology; the Clinical and Translational Science Institute; and Biostatistics Thursday, February 4, 2016 3:30 pm, A115 Crabtree Hall Response-Adaptive Trials: The Reality can be Messy Abstract. The Statins for Pulmonary and Cardiac Complications of Chronic HIV (SPARC) pilot clinical trial utilized a drop-the-loser urn response-adaptive design to allow participants higher probability of being randomized to the more effective study arm. A number of real-life complications made this design challenging including: 1) The use of a surrogate outcome measured at week 4 to adapt the randomization probabilities; 2) Some randomizations that occurred prior to the determination of all the previous participants’ surrogate outcomes; and 3) Participants with visits outside the window for the surrogate outcome measure. The theory for these designs assumes that the adaptive randomization is based on the final outcome. This provides two additional challenges: 1) SPARC has multiple final outcomes five months after the surrogate is measured; and 2) The surrogate was not so strongly related to the outcome as expected. Dr. Bertolet will describe how these challenges were addressed and the impacts on the implementation, analysis and results of the SPARC trial. 

Last Updated On Monday, March 21, 2016 by GSPH Webmaster
Created On Monday, March 21, 2016