Antidepressant signal detection in the clinical trials vortex.

J Clin Psychiatry 2015;76(5):e657–e659 (doi:10.4088/JCP.15com09934). © Copyright 2015 Physicians Postgraduate Press, Inc. V (Lu AA21004) is a novel antidepressant medication approved at doses of 10–20 mg/d for the treatment of major depressive disorder (MDD) by the US Food and Drug Administration (FDA) in September 2013 and by the European Commission in December 2013. This issue of the Journal includes reports of 2 clinical trials conducted exclusively in the United States that evaluated the efficacy and safety of vortioxetine versus placebo for the treatment of MDD. The 2 trials were similar in design and used 3 treatment arms: placebo and 2 vortioxetine doses (10 and 20 mg/d in the study by Jacobsen and colleagues1 and 10 and 15 mg/d in the study by Mahableshwarkar and colleagues2). In the Jacobsen et al trial, the 20-mg/d dose achieved superiority (a finding replicated in a second, as yet unpublished, US trial),3 but in the Mahableshwarkar et al trial, neither dose statistically separated from placebo on the primary outcome. Two prior published US trials comparing a 5-mg/d dose versus placebo also failed to demonstrate superiority of vortioxetine.4,5 These results contrast sharply with prior trials conducted outside the United States, in which 5and 10-mg doses demonstrated efficacy over placebo in adults with MDD,6–8 and 1 trial in which vortioxetine 10–20 mg/d proved superior to agomelatine in treatment-resistant MDD.9 The Mahableshwarkar et al and Jacobsen et al trials did not include an active comparator arm. Thus, we cannot conclude whether the failure of vortioxetine at 10–15 mg/d to separate from placebo represents a “failed trial” or a “negative trial.” Four prior published trials of vortioxetine for MDD used an active comparator (venlafaxine XR or duloxetine), resulting in 1 negative US trial,5 1 failed ex-US trial,7 and 2 positive trials.6,10 The conflicting results of the vortioxetine trials raise again an important question for researchers, regulators, and the pharmaceutical industry: Why do trials using the same drug and similar designs fail to agree with one another? Roughly half of the trials of compounds receiving FDA approval for antidepressant use failed to demonstrate superiority over placebo.11 Although some of these negative trials employed doses later determined to be below the minimum efficacious dose, the problem remains salient. These inconsistent results greatly increase the cost of drug development and diminish the public’s confidence that psychiatric treatments are any better than placebo.12 The sources of variability in outcomes for clinical trials of psychiatric medications can be divided into 3 broad categories: trial design characteristics, study participant characteristics, and quality of study conduct. Other than for dose differences, the Jacobsen et al and Mahableshwarkar et al trial designs were similar to all prior published placebocontrolled vortioxetine treatment trials for MDD: a 6–8 week treatment duration in adults with recurrent, at least moderately severe, depression who had failed to respond to no more than 1 antidepressant in the current episode. Although there was some minor variability in the baseline minimum symptom severity thresholds required for inclusion, this factor did not appear to affect trial outcome. Thus, it is very unlikely that the observed differences in outcomes can be explained by study design characteristics. Rather, differences in outcomes must arise from either (1) the conduct of the study procedures or (2) characteristics of the enrolled study participants. Potentially important sources of inconsistency of study procedures between sites include the level of rigor in application of inclusion/exclusion criteria, quality and consistency of rating scale assessments, and the degree of emotional support and advice provided to participants by the study staff.13 Attempts to improve the accuracy and precision of severity ratings include the use of centralized raters, who are remote from the site and masked to the design of the trial, as well as the use of computer ratings and interactive voice response technology to rate symptom severity.14,15 Some data support the use of these measures in improving signal detection.16,17 Although centralized raters were used in the Mahableshwarkar et al trial, the vortioxetine arms did not separate from placebo. Historically, the greatest concern about study participant samples has been their degree of responsiveness to placebo. Most negative and failed trials of proven antidepressants have occurred in studies that produced a high rate of placebo response; strong response to placebo creates a ceiling effect that leaves the active drug with little room to demonstrate efficacy. A meta-analysis of antidepressant trials revealed that studies with placebo response rates ≤ 30% provide the best chance for demonstrating superiority of an investigational drug.18 However, in the Mahableshwarkar et al and Jacobsen et al trials, in which the 10-mg/d dose failed to demonstrate superiority, the placebo response rates were only 28.4% and 32.9%, respectively. In contrast, in the positive Alvarez et al ex-US trial, the placebo response rate was 45%.6 What can explain these poorer vortioxetine outcomes in trials with lower placebo response? The most likely source of variability in trial outcomes lies in differences in the sample populations enrolled. MDD is a heterogeneous illness, but there is little reason to think that the populations with MDD in the United States are any more or less likely to respond to antidepressant medications than those from outside the Th is w or k m ay n ot b e co pi ed , d is tr ib ut ed , d is pl ay ed , p ub lis he d, re pr od uc ed , t ra ns m itt ed , m od ifi ed , p os te d, s ol d, li ce ns ed , o r u se d fo r c om m er ci al p ur po se s.