Matching Within a Hybrid RCT/RWD: Framework on Associated Causal Estimands

• Activity Number: 476 - Beyond Propensity Scores: Novel Techniques in Utilizing Real-World Data in Clinical Trials
• Type: Topic Contributed
• Date/Time: Wednesday, August 10, 2022 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Guanglei Yu* and Margaret Gamalo and Junjing Lin
• Companies: Eli Lilly and Company and Pfizer Inc. and Takeda Pharmaceuticals
• Keywords: estimand ; augmented control; causal inference; multiple treatment propensity score; matching; real-world evidence

Abstract

Real-world data incorporation techniques in trial conduct and analysis has seen increasing interest in different stages of drug development. Of particular interest is leveraging external control data to augment the control arm in a concurrent randomized controlled trial, where patients are enrolled in both investigational treatment and control arm. However, very little discussion focuses on delineating what should be matched and what is actually being estimated in a hybrid trial setting. In general, external control can be matched in four different ways: (1) matching with the intersection between investigational treatment and concurrent control, (2) matching with the union of concurrent investigational treatment and concurrent control, (3) matching with concurrent control alone, and (4) matching with investigational treatment alone. In this presentation, the formulation of estimands for different matching schemes will be discussed alongside the matching methods. Simulation studies are also conducted to evaluate the performance characteristics under different matching schemes, estimation methods, effect size assumptions, and missingness of confounders.