Primary Submission Category: Randomized Studies
Designing target trial emulations for COVID-19 pharmacotherapy effectiveness studies – challenges and implications of assigning time zero
Authors: David Bui, Kristina Bajema, Kristin Berry, Lei Yan, Yuan Huang, Yuli Li, Nallakkandi Rajeevan, Stephanie Argraves, Valerie Smith, Matthew Maciejewski, Amy Bonhert, Denise Hynes, Mihaela Aslan, George Ioannou,
Presenting Author: David Bui*
Real-world pharmacotherapy effectiveness (PE) studies should be carefully designed using target trial emulation (TTE) principles for valid causal inference. Designing target trials of outpatient COVID-19 PE presents unique challenges, especially in determining time zero—the point in a trial at which a patient is determined eligible, randomized to a treatment group, and starts follow-up. In TTEs, defining time zero can be difficult when only one group, namely treated persons, has a clear time zero because there is no obvious time zero for comparators not initiating an alternative treatment. There are several approaches for defining time zero (index date) in observational studies of COVID-19 pharmacotherapies: 1) using the date of positive SARS-CoV-2 test in both treated and untreated persons, 2) using the date of treatment initiation in treated persons and test-positive date in untreated persons, or 3) using the date of treatment initiation in treated persons and index date corresponding to the same number of days following the test-positive date in untreated persons. We describe how well each of these approaches emulates a randomized trial of oral antiviral COVID-19 treatment versus no treatment. Using real-world data from the U.S. Veterans Health Administration, we conduct a set of TTEs using each time zero approach, describe relative strengths, implications, and limitations for causal inference.