An increasing body of evidence is available suggesting that the Hamilton Depression Rating Scale (HAMD) is not a sensitive measure of treatment effect. In this investigation, we explore the sensitivity of the individual items of the Montgomery Asberg Depression Rating Scale (MADRS) and compare the consequences of selecting a different scale as primary endpoint in the analysis of efficacy. A graphical approach is proposed for the evaluation of the sensitivity of individual items to response, using data from randomised, placebo-controlled clinical trials in which HAMD and MADRS were measured concurrently. Subsequently, we illustrate the impact of differences in the sensitivity of the primary endpoint for the detection of statistical significance in treatment effect. In contrast to the HAMD, our item-by-item analysis of the MADRS reveals that all individual items are sensitive to response, irrespective of treatment type. However, some HAMD subscales still outperform MADRS in the detection of treatment effect. The selection of these subscales as primary endpoints in clinical trials could save over 1/3 in patients compared to the full HAMD whilst keeping the same statistical power.