ORIGINAL ARTICLE Hamilton scale and MADRS are interchangeable in meta-analyses but can disagree at trial level Lorenzo Guizzaro a,b, * , David Dickinson Vaughan Morgan c , Andrea Falco d,e , Ciro Gallo a a Statistica Medica, Universita della Campania Luigi Vanvitelli, Napoli, Italy b European Medicines Agency, Domenico Scarlattilaan 6, 1083 HS Amsterdam, the Netherlands c Pharmaceutical Medicine Group, King’s College London, London, United Kingdom d Department of Science and Technology (DST), Universita’ del Sannio, Benevento, Italy e Oxon Epidemiology, Madrid, Spain Accepted 29 April 2020; Published online 6 May 2020 Abstract Background and Objective: Major depressive disorder is a multidimensional disease, in which demonstrating the efficacy of treatments is difficult. The Hamilton Rating Scale for Depression (HRSD) and the MontgomeryeAsberg Depression Rating Scale (MADRS) cover different domains but are used interchangeably as primary measures of the outcome in trials anddwith standardized measuresdin meta-analyses. We aimed at understanding (i) whether the choice of the outcome measurement tool can influence the outcome of a trial, and if so, (ii) whether one systematically outperforms the other, and (iii) whether using standardized measures of the effect in meta-analysis is justified. Methods: Short-term randomized trials in patients with major depressive disorder that used both the scales were systematically searched and the results were collected. To quantify the differences in the resultsdboth in terms of the standardized mean difference (SMD) and odds ratio (OR) for responsedand their range, data were analyzed and plotted with the BlandeAltman method. Results: 161 comparisons from 80 studies were included, involving a total of 18,189 patients. Neither of the two scales appears sys- tematically more sensitive to the treatment effect than the other in terms of SMDs (P-value 5 0.06, 95% CI 0.044 to 0.001) or ORs (P- value 5 0.15, 95% CI 0.25 to 0.04). However, the variability of differences between the HRSD and MADRS largely depends on the number of patients included in the comparison. Conclusion: No systematic differences between the two scales were found supporting the use of standardized measures in meta- analyses. However, the same trial may give very different results with either scale, especially in small trials. Further research is needed to understand the causes of this variability. Ó 2020 Elsevier Inc. All rights reserved. Keywords: Methodology; Major depressive disorder; Hamilton depression rating scale; Montgomeryeasberg depression scale; Meta-analyses; Outcome mea- surement tools; BlandeAltman 1. Introduction Major depressive disorder (MDD) is an intrinsically multidimensional clinical entity [1] that creates a significant burden on individuals and societies [2]. The evaluation of pa- tients with MDD is difficult [3], and a panoply of scales has been proposed. Among these, the Hamilton Rating Scale for Depression (HRSD) [4]dparticularly its 17-items version (HRSD17)dand the MontgomeryeAsberg Depression Rat- ing Scale (MADRS) [5] are accepted [6] as primary outcome measurement tools in phase 3 trials. Drawing conclusions from studies using either tool implicitly assumes that they measure the same construct, i.e., that they measure in the same way the severity of depression. This assumption is made explicit when the magnitude of the effect of different interventions is compared in meta- analyses that use standardized measures of the effectd standardized mean difference (SMD) and odds ratio (OR) of responsedindependently of which the outcome mea- surement tool originated the data. The use of these mea- sures has come to be the widespread practice, despite early warnings that these should only be relied on when different scales are on the linear transformation of the other [7]. The less frequently used conversion tables ([8]) make a similar assumption at the patient level, although not the same distributional assumptions. Conflicts of interest statement: The authors have no conflicts of inter- ests to disclose. * Corresponding author. European Medicines Agency, Domenico Scar- lattilaan 6, 1083 HS Amsterdam, the Netherlands. Tel.: þ31 (0)88 781 7857. E-mail address: lorenzo.guizzaro@ema.europa.eu (L. Guizzaro). https://doi.org/10.1016/j.jclinepi.2020.04.022 0895-4356/Ó 2020 Elsevier Inc. All rights reserved. Journal of Clinical Epidemiology 124 (2020) 106e117