Equivalence and non-inferiority trials in the evaluation of non-pharmacological interventions: rationale, challenges and recommendations.

Abstract

Direct comparisons of non-pharmacological interventions are becoming increasingly common. Many randomised clinical trials are designed to assess whether the effect of one intervention is either equivalent to, or not inferior to, that of a criterion standard. This article provides an accessible introduction to such equivalence and non-inferiority designs. Topics covered include the choice of hypothesis and comparator, the choice of equivalence or non-inferiority margin and the benefits and drawbacks of common methods of analysis. While equivalence and non-inferiority trials offer unique possibilities, there are also challenges. Some of these are particularly pronounced in the non-pharmacological setting, such as the difficulty in establishing margins informed by placebo effects and structural barriers to achieving sufficient statistical power. Under-recognised general threats to the equivalence and non-inferiority designs are also discussed, including the threat of poor intervention delivery leading to washed-out effects, the especially pronounced threat of rhetorical 'spin' in the reporting of inconclusive findings and the threat of increasingly indirect evidence due to the sequential change of criterion standards. Implications for the non-pharmacological field are brought to the reader's attention, and suggestions are given to improve methodological rigour.