A0767
Title: Enhanced robust causal estimation of estimands in clinical trials
Authors: Ming Tan - Georgetown University (United States) [presenting]
Abstract: The central question in comparative clinical trials is how the treatment outcome compares to what would have happened to the same subjects had they received a different or no treatment, which is intrinsically a causal inference problem. In addition, after a patient is treated, intercurrent events, such as receiving rescue medicine, may depend on treatment conditions and can impact data collected to assess efficacy. The robustness of estimands of a relevant estimation, with respect to underlying assumptions, is then the key to the success of the trial, as reflected in regulatory guidance from ICH, FDA, and EMA. The aim is to present an enhanced doubly-robust estimation method utilizing semiparametric models with nonparametric monotone or concave link functions for both the propensity score and the outcome models. The models are estimated using an iterative procedure incorporating the pool adjacent violators algorithm. The asymptotic properties of the enhanced DREs are then studied. Simulation studies are performed to evaluate their finite sample performance. The benefit of this approach is explored for several causal estimands of interest. The method is then applied to analyzing several clinical trials.
