Leveraging External Data in Bayesian Adaptive Designs

• Activity Number: 272 - Statistical Innovations in Regulatory Science
• Type: Topic Contributed
• Date/Time: Tuesday, August 4, 2020 : 1:00 PM to 2:50 PM
• Sponsor: Biometrics Section
• Author(s): Alejandra Avalos Pacheco* and Steffen Ventz and Lorenzo Trippa and Brian Alexander and Andrea Arfe
• Companies: Harvard Medical School and Harvard School of Public Health and Harvard School of Public Health and Foundation medicine and Harvard Medical School

Abstract

In recent years there has been a growing interest in designs that incorporate non-experimental data, also known as real world data, or data from completed trials. I will discuss two uses of such data in the design and analyses of clinical studies. Firstly, I will introduce a novel Bayesian hybrid platform design that leverages external data via a non-parametric Bayesian model averaging approach, adjusts for confounding, and satisfies a set of required operating characteristics required by regulators. I will show the usefulness of this hybrid design using a collection of phase II and III trials of cancer immunotherapies for glioblastoma. Secondly, I will discuss validation techniques to quantify the accuracy of clinical outcome predictions obtained when leveraging external data, and compare their performance with other predictive clinical outcomes that do not incorporate such external information. I will illustrate the latter strategy with repositories of trial data.