A Novel Bayesian Nonparametric Method to Use Real-World Data in Clinical Trials

• Activity Number: 332 - Synthetic Clinical Trials Design to Accelerate FDA Approvals
• Type: Invited
• Date/Time: Thursday, August 12, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Noirrit Kiran Chandra* and Peter Mueller
• Companies: University of Texas at Austin and University of Texas Austin
• Keywords: Bayesian non-parametric; clustering; clinical trial; real world data; cancer; glioblastoma

Abstract

Randomized clinical trials (RCT) are the gold standard for approvals by regulatory agencies. Trials span therapeutics, vaccines, medical devices, and behavioral change regimens. RCT protocols involving human subjects are spread over four phases from safety and dose finding in phase I to large-population trials consisting of patients with afflictions for which therapies are designed. Clinical trials are increasingly time consuming, expensive, and laborious with a multitude of bottlenecks. The advent of electronic health records (EHR) can open opportunities for digital healthcare innovations involving mostly observational data and potentially expedite clinical trials dramatically. In this project we propose to use data from EHR as the synthetic control group in a clinical trial experiment instead of using real patients who randomly receive placebo/existing treatment. We also propose a novel Bayesian clustering method where we find equivalent population classes from the synthetic control group and actual treatment group. We show equivalence of the two classes by standard supervised classification algorithms. We demonstrate the method in Glioblastoma patients.