All Times EDT
Keywords: Synthetic control, robust causal estimates
We propose a clinical design and analysis procedure for comparing a new therapy with a synthetic control obtained using real world data including disease registry/electronic health record. The method is motivated by a trial to compare relapse-free survival (RFS) rate at 3 years between locally treated high-risk ocular melanoma patients on adjuvant combination immunotherapy versus a matched contemporaneous control population, where a randomized control trial is not feasible. Although causal inference procedure such as the doubly robust estimate (DRE) has been available to obtain unbiased estimate of the treatment effect if either the propensity score or the outcome regression model is correctly specified. However, it is known that DRE may fail and give estimates with large bias and variance, even when the propensity and/or outcome models are just mildly misspecified. We utilize the enhanced DRE recently developed to obtain causal estimate of treatment effect which is shown to have much enhanced robustness against model specifications, thus providing evidence for efficacy of the immunotherapy. This work is in collaboration with Ao Yuan, Suthee Rapisuwon and Michael B. Aitkins.