LAUSR

Randomized non-inferiority trials are designed to study whether an experimental treatment (or diagnostic test) is not unacceptably worse in terms of efficacy compared to the standard of care, whilst providing specific advantages over the standard of care justifying this comparison [1–3]. The magnitude by which the experimental treatment is acceptably less efficacious than the standard of care is commonly referred to as the non-inferiority margin. This pre-specified margin allows for constructing the null hypothesis stating that the experimental treatment is statistically significantly worse (i.e. inferior) than the standard of care. Rejecting this null hypothesis permits a conclusion of non-inferiority. The aforementioned specific non-efficacy advantages over the standard of care are referred to as non-efficacy benefits, and can for instance come in the form of lower probability of adverse events or complications, convenience of use, or lower costs. The non-inferiority design lends itself particularly well to trials on de-intensifying or withholding treatment with known adverse effects. Elegant examples of such trials present both the implied non-efficacy benefit as a primary outcome (tested for superiority) and the potential loss in efficacy as a co-primary (or hierarchical secondary) outcome (tested for non-inferiority), hence aiming to demonstrate that de-intensified treatment comes with less adverse events balanced by acceptable losses in efficacy [4–7]. For example, consider a trial aimed to study whether withholding one of two antithrombotic medications prescribed after transcatheter aortic valve implantation, would result in less bleeding complications (primary endpoint tested for superiority; inferred non-efficacy benefit of de-intensifying antithrombotic treatment) without leading to an unacceptable higher number of thromboembolic events (secondary endpoint tested for non-inferiority) [5]. Although not so common, meta-analyses of trials have also employed non-inferiority designs to analyze aggregate data in order to answer noninferiority questions with greater precision [8]. In this issue of the Eur J Epidemiol, Dr Acuna and colleagues make a case for ways to improve the reporting of non-inferiority trials by putting greater emphasis on nonefficacy benefits [9]. The work by Dr Acuna and colleagues provides a strong theoretical foundation with clear practical recommendations on how to implement non-efficacy benefits and non-inferiority margins into everyday reporting of non-inferiority trials. However, contemporary practices on this matter are less well documented. Hence, we aimed to evaluate the current scientific practice regarding the use and reporting of non-efficacy benefits and non-inferiority margins in clinical cardiology—a field of medicine with an ever increasing number of non-inferiority trials [3]. We focused on recent trials in high-impact journals as these generally serve as focal point for future studies to be modelled on.