Statistical methods for the assessment of clinical relevance

Abstract

It is commonly accepted that the results of clinical trials to be important for medical practice do not only have to be statistically significant but also clinically relevant. While an elaborated and canonical methodology exists for statistical significance tests, there is no common sense so far on how to judge the clinical relevance of a medical finding. The assessment of the clinical relevance of a study result should provide quantified information about its practical importance. For this, both statistical procedures and appropriate effect measures on which the relevance judgment is based are required. The test for relevant superiority and the relevance assessment based on the observed effect are presented as two statistical approaches for the assessment of clinical relevance. The properties of these procedures are investigated and contrasted. Furthermore, an overview of effect measures used for relevance assessment is given and their characteristics are illustrated. Application of the methods is illustrated with a clinical trial example.