What is the threshold for a clinically relevant effect? The case of major depressive disorders.

BACKGROUND: Randomized trials can show whether a treatment effect is statistically significant and can describe the size of the effect. There are, however, no validated methods available for establishing the clinical relevance of these outcomes. Recently, it was proposed that a standardized mean difference (SMD) of 0.50 be used as cutoff for clinical relevance in the treatment of depression.
METHODS: We explore what the effect size means and why the size of an effect has little bearing on its clinical relevance. We will also examine how the "minimally important difference," as seen from the patient perspective, may be helpful in deciding where the cutoff for clinical relevance should be placed for a given condition.
RESULTS: Effect sizes in itself cannot give an indication of the clinical relevance of an intervention because the outcome itself determines the clinical relevance and not only the size of the effects. The "minimal important difference" (MID) could be used as a starting point for pinpointing the cutoff for clinical relevance. A first, rough attempt to implement this approach for depression resulted in a tentative clinical relevance cutoff of SMD = 0.24. Using this cutoff, psychotherapy, pharmacotherapy, and combined treatment have effect sizes above this cutoff. DISCUSSION: Statistical outcomes cannot be equated with clinical relevance. The "MID" may be used for pinpointing the cutoff for clinical relevance, but more work in this area is needed.