Dwalingen in de methodologie. XXXIII. De ethiek van de schatting van de steekproefomvang: Minder proefpersonen nodig bij eenzijdige dan bij tweezijdige statistische toetsing

In sample size calculations for clinical trials, two-sided statistical testing is the usual starting point. Compared with one-sided testing, this option requires the inclusion of a larger number of study subjects, and a larger number of study events will be needed before a conclusion can be reached. It is therefore ethically relevant to consider in which situations one-sided testing should be the preferred option. One-sided statistical testing is preferable when the research hypothesis is one-sided (is intervention A better than intervention B?), or when only a clearly better result for A compared to B will have clinical consequences (e.g. if A is more cumbersome or more invasive to the patient than B). For the estimation of the number of study subjects needed, one-sided testing should then be assumed. If a new intervention is to be tested by comparison with placebo treatment or with absence of treatment, one-sided testing is an adequate starting point for sample size estimation.