Clinical Equipoise and Adaptive Clinical Trials

Ethically permissible clinical trials must not expose subjects to risks that are unreasonable in relation to anticipated benefits. In the research ethics literature, this moral requirement is typically understood in one of two different ways: (1) as requiring the existence of a state of clinical equipoise, meaning a state of honest, professional disagreement among the community of experts about the preferred treatment; or (2) as requiring an equilibrium between individual and collective ethics. It has been maintained that this second interpretation makes it mandatory to minimize the number of patients receiving the treatment that will eventually be shown to be inferior by the trial. This requirement has led to the development of adaptive trials, i.e., trials in which treatment allocation is determined by data accumulated during interim analysis. Many statisticians argue that in some circumstances—typically with potentially high benefits, as in the much discussed ECMO trial—adaptive design is the only ethically permissible experimental design. Nevertheless, some proponents of clinical equipoise argue that adaptive trials are neither ethically required nor permissible. More specifically, they argue that clinical trials using adaptive designs fail to meet the moral requirement of clinical equipoise, since these trials presuppose an epistemic state that is incompatible with a physician’s duty of care to her subjects. This paper emphasizes that the debate is to a large extent resting on an epistemological confusion. Specifically, I argue that this response conflates two different conceptions of statistical evidence (i.e., frequentist and Bayesian), and that recognizing this distinction elucidates an epistemological framework in which adaptive trials are both consistent with and recommended by the moral requirement of clinical equipoise.