Clinical relevance is often assessed considering the proportion of responders. The responder definition is often arbitrary, e.g. a 50% reduction from baseline. However, for patient reported outcomes empirically validated definitions of treatment responders can be derived using statistical methods.

We performed a blinded analysis from a placebo-controlled study of hot flushes in postmenopausal women. Patients recorded the daily number of hot flushes and their satisfaction with treatment.

A non-parametric discriminant analysis yielded the cut-off value between minimally improved and unchanged or worsened patients as well as between much improved and minimally improved. As a result, a responder was defined as having at least a minimal improvement at week 4 and a substantial improvement at week 12. This definition was agreed with the FDA and used as an endpoint in a pivotal trial.