Quasi-experimental design for external control arm studies alongside single arm trials for regulatory purposes

Abstract

To accelerate new drug, biologic, and medical device development and to improve efficiency of delivery of the latest breakthroughs of innovative, life-saving treatments to millions of patients, single-arm trial (SAT) applications of rare diseases or conditions supplemented by their external control arm (ECA) evidence for regulatory approvals have been surging since 2016. However, there have been increasing concerns over potential biases threatening the internal validity of these applications from regulatory authorities, payers, and research community. There are two main sources of potential biases. The first source is heterogeneity between two arms at the level of patients, and the second one at the level of systems (e.g., two entirely different sets of hospitals from which patients in a SAT and patients in an ECA are drawn separately). The currently commonly used study design is a post-intervention measurement only design that though mitigating the first source of bias, is utterly unable to control for the second one. This perspective article will propose a quasi-experimental design as an alternative that may mitigate the second source of bias, aiming to improve the internal validity of SAT and ECA studies. We will start summarizing the two main sources of biases that may impede the causal inference of these studies. Two approved therapies supported by SAT and ECA studies will be used as an example to illustrate these biases in detail. We will then introduce the intuition of the quasi-experimental design, underlying assumptions and data requirements, and empirical strategies for estimating interventional effects. We will conclude this article by discussing caveats of applying this alternative design for SAT and ECA studies.