Primary Submission Category: Randomized Studies
Fusing efficiency: A review of data fusion methods with application to PIONEER 6 case study
Authors: Xi Lin, Jens Magelund Tarp, Robin Evans,
Presenting Author: Xi Lin*
Integrating real-world data (RWD) and randomized controlled trials (RCTs) is becoming increasingly important in advancing causal inference in clinical research. This fusion holds great promise for enhancing the efficiency of average treatment effect estimation, thereby reducing the required number of trial participants and expediting drug access for patients in need. The FDA and EMA have recognized the complementary nature of these data sources and their integration to improve the quality of evidence. Despite the multitude of available data fusion methods, choosing the most suitable one for a specific research question is challenging. This difficulty arises from the diverse assumptions, associated limitations, and implementation complexities.
Our project aims to systematically review and compare data fusion methods, focusing on efficiency gain in average treatment effect (ATE) estimation. Through extensive simulations mirroring real-world scenarios, we identified a qualitative behaviour demonstrating a common risk-reward tradeoff across different methods. We investigate and interpret this tradeoff in various scenarios, providing important insights into understanding the strengths and weaknesses of different methods.
This presentation offers a comprehensive overview of available methods, highlights key findings from simulation studies and presents a real-world case study where PIONEER 6 trial is augmented with a US medical claims database for a more powerful ATE estimation.