Defining 'response' in antipsychotic drug trials: recommendations for the use of scale-derived cutoffs

Abstract

Scale-derived cutoff points are frequently used to define 'response' in antipsychotic drug trials. This procedure is useful, because responder rates can be understood more intuitively than a difference in means of rating scales. As various definitions of response have been used, we examined original participant data to assess whether the choice of the Brief Psychiatric Rating Scale-based response cutoff had an impact on the results of seven (n=1870) antipsychotic drug trials in schizophrenia. We also analyzed whether the chronicity of the illness has an impact on the question of which cutoff is most sensitive in detecting differences between drugs. The results in terms of p-values and response rate differences varied substantially in dependence on the cutoff chosen. The use of response rate ratios attenuated the variability. In contrast to a widely held belief, low response cutoffs were not more sensitive in detecting differences between drugs than higher cutoffs. In more chronic, less responsive participants, there was a trend for higher cutoffs to be less sensitive in detecting differences between drugs than lower ones. The results of clinical trials depend considerably on the response cutoff chosen. Therefore, the cutoff should never be chosen post hoc, a large range of cutoffs should be presented and the a priori choice of the primary cutoff should be based on clinical relevance. The use of ratios rather than differences attenuates the variability. Cutoffs need to be calculated on the basis of 0-6 rather than on 1-7 scoring systems. We suggest a table presenting responder rates in 25 percent steps covering the whole range up to 100% reduction from baseline, which could be displayed together with recently presented criteria for remission.