West Coast Ballroom
Building and Validating Real-World Control Arms for Randomized Controlled Trials with Small Concurrent Control Arms: Challenges toward Establishing Best Practices (306704)
Douglas E Faries, Eli Lilly and Company
Huzhang Mao, Eli Lilly and Company
Xiang Zhang, Eli Lilly
Keywords: real world data, control arm, randomized controlled trial, bias control
Real-world data as external control arms (RWCA) for randomized controlled trials (RCT) with small or under-powered concurrent control arms (CCA) are increasingly used in drug development. This research aims to develop best practices for building and validating the RWCA to supplement the underpowered RCT control arms. We introduce processes covering from selecting target patients, minimizing covariate differences, improving covariates overlap, and addressing covariate imbalance, to performing validation test and treatment comparison. Steps and methods for assessing the importance of concurrency of RWCA data and augmenting the CCA with RWCA to form a hybrid control arm are discussed. A case study with RCT data and Flatiron Health electronic health records-derived database in lung cancer is used to illustrate the introduced approach. Methods of entropy balancing, inverse probability weighting and propensity score matching are applied to control for covariate imbalance prior to validation test and treatment comparison. We conclude the validity of RWCA for RCT could be established after adequate target patient selection, covariate matching and balancing, and validation test.