Up-Front Matching: An Ongoing Recruitment Method for Prospective Observational Studies That Mimics Randomization for Selected Baseline Covariates

• Activity Number: 385 - Biomarkers, Endpoint Validation and Other Topics
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 12:00 PM to 1:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Ibrahim Turkoz* and William H Olson
• Companies: Janssen Research and Development, LLC and WHO Statistical Consulting, LLC
• Keywords: Comparative effectiveness research; Enrollment strategy; Stratified frequency matching; Prospective observational studies

Abstract

For prospective observational studies that are designed to compare treatments this presentation will introduce a recruitment method, called “up-front matching”, that will result in balance for selected baseline covariates that mimics what one would see with randomization. Using the propensity score (PS) based on baseline covariates in a historical computer database, up-front matching (1) enrolls only patients whose PS is in the common support; and (2) performs frequency matching based on PS score strata. Up-front matching is compared to random sampling in a simulation. Across simulated studies, each with 200 patients per group, the distribution of the standardized mean difference for each of the covariates with up-front matching is well approximated by the normal distribution one would see with randomization (N(0, 2/200)), and with random sampling it is well approximated by N(std.diff, 2/200), where std.diff is the standardized difference in means. Use of up-front matching recruitment has the potential to (1) significantly enhance the scientific validity and statistical efficiency of observational studies, and (2) reduce the cost by following only patients in the common support.