On Clinically Relevant Differences and Shifted Nullhypotheses

Abstract

SummaryThe currently usual one value for the judgement of the clinical relevance of therapeutic effects frequently does not suffice to adequately formulate the problems of clinical studies, and the statistical standard procedure (the testing of the classical nullhypothesis) fails to take this value duly into account.Therefore, it is proposed to judge the clinical relevance and importance by means of four values, fixed in discussions with the clinician before commencement of the study, and to proceed by testing non-zero nullhypotheses (shifted nullhypotheses) where the “clinically relevant difference” is the shift parameter.Methodical problems resulting from the shifting of the nullhypothesis are discussed, and other possibilities to take into account the clinically relevant difference (introduction of criteria of success) are considered.