Could a one-sided test be used in a superiority trial if this is decided in advance? That would be problematic if the effect turned out to go in the opposite direction, meaning that the new treatment had a poorer effect than the standard treatment. In this case, it would have to be attributed to chance, irrespective of how great the difference was. One-sided tests have greater statistical power in one direction, but exclude the possibility of claiming any effects in the opposite direction. This is illustrated in Figure 1. This and other arguments in favour of using two-sided tests are elaborated in (1).