Borrowing from Historical Data in Cancer Drug Development: A Cautionary Tale and Experience with Bayesian Adaptive Platform Designs

• Activity Number: 15 - Incorporating Knowledge from Previous Clinical Trials into the Design of New Trials
• Type: Invited
• Date/Time: Monday, August 3, 2020 : 10:00 AM to 11:50 AM
• Sponsor: Biopharmaceutical Section
• Author(s): James Normington* and Connor Jo Lewis and Somnath Sarkar and Jiawen Zhu and Federico Mattiello
• Companies: University of Minnesota and Securian Financial Group, Inc. and Flatiron, Inc. and Roche-Genentech and F. Hoffman-La Roche, Inc.
• Keywords: adaptive design; Bayesian analysis; historical data; platform design

Abstract

Complex innovative trial designs (CIDs) sometimes borrow information from similar but already-completed studies. We begin with a colorectal cancer case study in which relying solely on historical control information erroneously identifies a significant treatment effect. We then catalog situations where borrowing historical information may or may not be advisable. We find that even after accounting for variations in study design, baseline characteristics, and standard-of-care improvement, our approach consistently identifies Bayesianly significant differences between the historical and concurrent controls under a range of priors. Next, we propose a Bayesian adaptive platform design that uses commensurate prior methods at interim analyses to borrow adaptively from the control group of an earlier-starting trial. Our approach performs well via simulation across settings with varying degrees of commensurability and true treatment effects, and compares favorably to an adaptive "all-or-nothing'' ad-hoc frequentist borrowing approach. We also consider a three-drug extension where a new imaginary intervention joins the platform, and again show our procedure to perform well via simulation.