Sensitivity of the individual items of the Hamilton depression rating scale to response and its consequences for the assessment of efficacy.

The Hamilton depression rating scale (HAM-D(17)) has been the gold standard in depression trials since its introduction in 1960 by Max Hamilton. However, several authors have shown that the HAM-D(17) is multi-dimensional and that subscales of the HAM-D(17) outperform the total scale. In the current study, we assess the sensitivity of the individual HAM-D(17) items in differentiating responders from non-responders over the typical treatment period used in clinical efficacy trials. Based on data from randomised, placebo-controlled trials with paroxetine, a graphical analysis and a statistical analysis were performed to identify the items that are most sensitive to the rate and extent of response irrespective of treatment. From these analyses, two subscales consisting of seven items each were derived and compared to the Bech and Maier and Philip subscales using a linear mixed-effects modelling approach for repeated measures. The evaluation of two clinical trials revealed endpoint sensitivity comparable to the existing subscales. Using a bootstrap technique, we show that the subscales consistently yield higher statistical power compared to the HAM-D(17), although no subscale consistently outperforms the others. In conclusion, this study provides further evidence that not all items of the HAM-D(17) scale are equally sensitive to detect responding patients in a clinical trial. A HAM-D(7) subscale with higher sensitivity to drug effect is proposed consisting of the HAM-D(6) and the suicide item. This response-based subscale increases signal-to-noise ratio and could reduce failure rate in efficacy trials with antidepressant drugs.