Kavli Affiliate: Xiang Zhang
| First 5 Authors: Chenyin Gao, Xiang Zhang, Shu Yang, ,
| Summary:
Externally controlled trials are crucial in clinical development when
randomized controlled trials are unethical or impractical. These trials consist
of a full treatment arm with the experimental treatment and a full external
control arm. However, they present significant challenges in learning the
treatment effect due to the lack of randomization and a parallel control group.
Besides baseline incomparability, outcome mean non-exchangeability, caused by
differences in conditional outcome distributions between external controls and
counterfactual concurrent controls, is infeasible to test and may introduce
biases in evaluating the treatment effect. Sensitivity analysis of outcome mean
non-exchangeability is thus critically important to assess the robustness of
the study’s conclusions against such assumption violations. Moreover,
intercurrent events, which are ubiquitous and inevitable in clinical studies,
can further confound the treatment effect and hinder the interpretation of the
estimated treatment effects. This paper establishes a semi-parametric framework
for externally controlled trials with intercurrent events, offering doubly
robust and locally optimal estimators for primary and sensitivity analyses. We
develop an omnibus sensitivity analysis that accounts for both outcome mean
non-exchangeability and the impacts of intercurrent events simultaneously,
ensuring root-n consistency and asymptotic normality under specified
conditions. The performance of the proposed sensitivity analysis is evaluated
in simulation studies and a real-data problem.
| Search Query: ArXiv Query: search_query=au:”Xiang Zhang”&id_list=&start=0&max_results=3