Equivalence Testing Instead of Null-Hypothesis Testing

Abstract

A negative study is not equal to an equivalent study. The former cannot reject the null-hypothesis of no effect, while the latter assesses whether its 95 % confidence interval is between prior boundaries, defining an area of undisputed clinical relevance. Equivalence testing is important, if you expect a new treatment to be equally efficacious as the standard treatment. This new treatment may still be better suitable for practice, if it has fewer adverse effects or other ancillary advantages. For the purpose of equivalence testing we need to set boundaries of equivalence prior to the study. After the study we check whether the 95 % confidence interval of the study is

• 1. entirely within the boundaries (equivalence is demonstrated),

• 2. partly within (equivalence is unsure),

• 3. entirely without (equivalence is ruled out).