The European Medicines Agency recently granted approval to agomelatine (Valdoxan), the first antidepressant with selective affinity for melatonergic MT1/MT2 receptors and the 5-HT2C receptor. In the development programme of agomelatine, the initial relapse prevention study (study 1) lacked assay sensitivity because of an unexpectedly low relapse rate in the placebo arm. However, post-hoc analysis supported the potential of agomelatine in preventing relapse in a subgroup of patients with more severe depressive symptoms at inclusion. Moreover, the largest treatment effect relative to placebo in the acute studies was observed in patients with more severe levels of symptoms (Kennedy and Emsley, 2006; Montgomery and Kasper, 2007; Olié and Kasper, 2007). We were concerned that symptom levels alone are an insufficient measure of illness severity, and in any case, it is impractical to insist on increasingly higher scores on symptom scales for study entry. Our hypothesis was that the lack of assay sensitivity in the first trial required key modifications to ensure that the patient population showed adequate severity not just on ratings of symptom severity but also on measures of functional impairment.
The present paper reports in parallel the findings of the two relapse prevention studies carried out with agomelatine and describes the particular methodological innovations that were introduced in study 2 to improve the assay sensitivity (Goodwin et al., 2009). We anticipate that some of the methodological innovations reported here could contribute towards better quality recruitment of patients into placebo-controlled studies and in turn result in lower failure rates for future clinical trials in the field.
Materials and methods
The two studies, run in accordance with the principles of Good Clinical Practice E6 of the International Conference of Harmonisation (CPMP/ICH/135/95) and the Declaration of Helsinki, Finland (1964, 1996), were approved by the relevant local ethics committees and included only patients who had provided written informed consent.
Study 1 was carried out in 92 centres in three countries (France, Spain and Germany) from September 1999 to June 2002, and study 2 was carried out in 57 centres in five countries (Australia, Finland, France, South Africa and the UK) from February 2005 to February 2007.
Eligible patients were physically healthy male or female outpatients aged 18–65 years with recurrent depression and a current major depressive episode assessed as moderate or severe, according to the Diagnostic and Statistical Manual of Mental Disorders, 4th ed., Text Revision (DSM-IV-TR) criteria.
In study 1, patients were selected on the basis of a minimum of two episodes within the last 3 years or a minimum of three episodes within the last 5 years. No such criteria were requested for study 2.
The current major depressive disorder (MDD) episode, with or without melancholic features, was required to have started at least 8 weeks before selection and, in addition, patients of study 2 had to have been previously free of significant symptoms for at least 6 months. A seasonal pattern, psychotic features or postpartum onset were not allowed.
Exclusion criteria, including nonauthorized disorders and concomitant therapies, were identical for both studies, and have been described previously (Goodwin et al., 2009).
Concomitant medications were identical for both studies, except for benzodiazepines, which were permitted in study 1. Only zolpidem could be taken until week 2 in case of insomnia. In study 2, all benzodiazepines had to be stopped at the time of selection.
Key study differences
In addition to a Hamilton Depression Rating Scale (HAM-D) 17-item total score of at least 22, patients must have, in study 2, a sum of items (1+2+5+6+7+8+10+13) of HAM-D 17-item of at least 55% of HAM-D 17-item total score: this was to ensure a central contribution from core symptoms of the depression syndrome (Kasper et al., 2010). In addition, study 2 entry required a Clinical Global Impression (CGI) Severity of illness [(CGI)-S] of at least 4, a Hospital Anxiety Depression Scale depression subscore of at least 11. The Sheehan Disability Scale (Sheehan et al., 1996) had to be filled in by patients at selection. A summary of the main similarities and differences between patient selection in the two studies is presented in Table 1.
Designs of the two studies
The two specifically designed studies evaluated the efficacy of agomelatine in preventing relapse of depression either at a fixed 25 mg dose (study 1) or at a flexible 25–50 mg dose (study 2).
In study 1, after an 8-week open-label treatment phase, patients who achieved a HAM-D total score of 10 or less were randomized to a 6-month double-blind, placebo-controlled treatment period. Patients could continue for an optional 18-week double-blind extension period.
In study 2, following an 8- or a 10-week open-label treatment period, patients who achieved a HAM-D 17 total score of 10 or less and CGI-I of 2 or less were randomized to a 6-month double-blind, placebo-controlled treatment period. Patients could continue for a further optional 20-week double-blind extension period.
The primary analysis in both studies was carried out over the 6-month period and was reported for study 2 (Goodwin et al., 2009). Findings on extension periods are instructive as the duration of the trial corresponds to the usual treatment period to avoid relapse in patients who respond to an antidepressant drug.
Key study differences
The full randomization criteria were known by the investigator in study 1, whereas they were blinded to the investigator and the patient and applied centrally in study 2. In study 2, the agomelatine dose at week 2 was either maintained at 25 mg/day or increased to 50 mg/day in patients with insufficient improvement on the basis of blinded criteria. During study 1, all patients took one tablet orally once a day in the evening, irrespective of the treatment allocated, whereas during study 2, all patients took two tablets orally once a day in the evening – either 2× 25 mg agomelatine or 1× 25 mg agomelatine and 1× placebo.
Eligible patients were assigned to agomelatine or placebo treatment according to a balanced randomization with implicit stratification according to centre for study 1, and stratification according to the clinical centre and the randomization visit (week 8 or 10) for study 2.
A summary of the main similarities and differences between the two studies’ designs is presented in Table 1.
Primary efficacy assessment
In both studies, the main measure was the time to relapse, defined as the time between the date of the first randomized treatment administration and the date of the relapse (or censoring). In study 1, relapse was defined as a HAM-D 17-items (Hamilton, 1960) total score of at least 16 at one visit or a suicide/suicide attempt. In study 2, an additional relapse criterion was withdrawal for lack of efficacy according to the clinical opinion of the investigator [on the basis of HAM-D 17 and CGI (Guy, 1976) scores].
Secondary efficacy assessment
Secondary efficacy parameters included occurrence of relapse within the 10-month extension period, change in the HAM-D 17-item total score and CGI scores and response to treatment (decrease in the HAM-D 17 score from baseline ≥50%) over the open period, and a change in the HAM-D 17 total score and CGI scores over the 10-month double-blind period.
Safety and tolerability
The tolerability and safety evaluations in both studies were carried out on the basis of emergent adverse events (AEs), vital signs (supine systolic blood pressure and diastolic blood pressure, supine heart rate, weight), biochemistry and haematology parameters and ECG abnormalities. When premature discontinuation of treatment was because of an AE, the information related to the outcome of the event was collected.
The time to relapse over the double-blind treatment period was compared for agomelatine and placebo groups using a log-rank test (stratified for centre type and randomization visit for study 2). To estimate the hazard ratio (HR) of relapse on agomelatine compared with placebo, a Cox model associated with the likelihood ratio test was used (with adjustment for centre type and randomization visit for study 2). For study 2, a Cox model with adjustment for HAM-D 17-item total score at inclusion was also used for sensitivity analysis. A nonstratified log-rank test and an unadjusted Cox model were used.
All efficacy and safety assessments were carried out in patients entering the double-blind period and having taken at least one dose of study randomized treatment and, for efficacy assessment, in patients having at least one postbaseline efficacy assessment over the double-blind period.
Efficacy assessments were also carried out in the subset of patients with a baseline HAM-D total score of more than 25 plus CGI-S of at least 5 (study 1) and with a baseline HAM-D total score of at least 25 (study 2).
Statistical analysis was carried out on SAS software, version 8.2 (SAS Institute Inc., Cary, North Carolina, USA). The type I error was set at 5%. Descriptive statistics were provided for secondary efficacy parameters and for emergent AEs during the double-blind treatment period.
Patients at inclusion and during the open-label period
Five hundred and fifty-one patients entered the open-label period of study 1 and 492 patients entered the open-label period of study 2. Patients at inclusion had a mean number of 4.4±3.3 episodes in study 1 and a mean number of 3.7±2.1 episodes in study 2.
No notable differences were found at the start of the open treatment in the mean HAM-D total score (26.3 vs. 27.0), or CGI-S score (4.8 vs. 4.9) across the two studies; only a slightly greater proportion of severely depressed patients as defined by the usual cut-off score on the HAM-D of at least 25 was found in study 2 compared with study 1 (74 vs. 70%) (Table 2).
During the open-label period of both studies, the HAM-D total score and the mean severity of illness score of the included patients decreased similarly. In study 1, during the period W0–W8, the mean HAM-D total score in the included patients showed a steady decrease on agomelatine treatment from 26.3±2.9 at week 0 to 11.1±8.4 at the last postbaseline visit over this period (n=543). The rate of response to treatment by HAM-D was 71.5% over the open 8-week period (last postbaseline value).
The mean CGI-S total score on agomelatine treatment was reduced from 4.8±0.7 at week 0 to 2.7±1.6 at the last postbaseline visit over this period (n=544).
In study 2, the mean HAM-D total score decreased progressively from 27.0±2.7 at week 0 to 9.9±7.3 at the last postbaseline assessment. At the same time, 78.6% of patients were rated as responders to treatment. The mean severity of illness score on CGI decreased from 4.9±0.7 at week 0 to 2.4±1.3 at the last postbaseline assessment.
In study 2, for 109 patients (22.1% of included patients), the agomelatine dose was increased to 50 mg at W2.
Patients at randomization
There were no relevant differences in the demographic characteristics between patients treated with agomelatine or placebo in either of the studies (Table 3), and the level of severity was similar. Nevertheless, a somewhat greater proportion of severely depressed patients was observed in study 2 compared with study 1 (80 vs. 68%). For patients in study 2, the duration of the current episode was almost twice as long as those entering study 1.
In study 1, the results of the primary analysis showed no beneficial effect for agomelatine relative to placebo. In the intention-to-treat population, the proportion of patients who had relapse during the 6-month double-blind period in the agomelatine group (25.9%) was not significantly different from the proportion of patients having relapse in the placebo group (23.5%). The HR over time for agomelatine versus placebo at 6 months was 1.045 (Fig. 1 and Table 4). Over the 10-month period, the proportion of patients who had relapse in the agomelatine and placebo groups was not significantly different (Fig. 1 and Table 5).
In study 2, the proportion of patients who had relapse during the double-blind period in the agomelatine group (20.6%) was half that in the placebo group (41.4%). The proportion of relapse over 6 months was significantly lower with agomelatine compared with placebo (P=0.0001), and the risk of relapse over time was reduced by 54% for agomelatine-treated patients (Fig. 2 and Table 4). When calculated using an unadjusted Cox model, similar results were obtained [HR: 0.461 (0.306; 0.693)].
Over the 10-month period, the proportion of patients who had relapse in the agomelatine group was significantly lower with agomelatine compared with placebo (23.6 vs. 47.7%; P<0.0001). The risk of relapse over time was reduced by 56% for agomelatine-treated patients (Fig. 2 and Table 5).
Exploratory analyses showed no influence of visit of randomization (W8 or W10), dose and dose adjustment to the observed effects.
In both studies, during the double-blind period, the effect of long-term agomelatine treatment as measured by the HAM-D total score was stable over time. In study 1, the mean HAM-D total scores at randomization were 6.0±2.7 for patients on agomelatine and 6.2±2.7 for patients on placebo. The mean HAM-D total score was 5.8±5.3 in the agomelatine group and 6.2±5.5 in the placebo group after 10 months. In study 2, the mean HAM-D total scores at randomization were 6.1±2.6 for patients on agomelatine and 6.0±2.7 for patients on placebo. The mean HAM-D total score was lower after 10 months in the agomelatine group (7.8±7.4) than in the placebo group (11.5±8.6).
In study 1, both CGI mean scores were similar after 10 months in the agomelatine group (1.5±0.9 for severity of illness, and 2.9±1.6 for global improvement) and in the placebo group (1.6±1.0 and 3.0±1.6, respectively). In study 2, both CGI mean scores were lower after 10 months in the agomelatine group (2.1±1.3 for severity, and 3.7±1.7 for overall improvement) than in the placebo group (2.7±1.5 and 4.5±1.8, respectively).
Subpopulation of severely depressed patients
In both studies, agomelatine was superior to placebo in preventing relapse in the subset of severe patients. The relapse rates in the agomelatine group were 21.3 and 21.9% in studies 1 and 2, respectively, whereas the respective relapse rates in the placebo group were 31.3 and 45.1% over 6 months. In study 2, the survival analysis showed a statistically significant difference in favour of agomelatine (P=0.0001). The risk of relapse over time was reduced with agomelatine by 37% in study 1 and by 57% in study 2 (Table 4).
In study 1, the time-to-relapse curves over 10 months showed that the proportion of patients who had relapse was significantly lower with agomelatine (23.6%) compared with placebo (37.5%; P=0.046), and the risk of relapse was reduced by 43% for agomelatine-treated patients (Table 5 and Fig. 1). In study 2, the statistically significant difference in favour of agomelatine was maintained at 10 months (P=0.0001), as well as the strong reduction in the risk of relapse over time on agomelatine as compared with placebo (Table 5 and Fig. 2).
A Cox regression analysis over 10 months showed that there was no significant interaction between treatment and baseline severity in both studies.
During the open-label treatment period, 182 (33.9%) patients withdrew from study 1, among them 133 (24.1%) as a result of lack of efficacy, in part because patients who failed to fulfil the randomization criterion on W8 were also classified under this reason. In study 2, 153 (31.1%) patients withdrew during the open period, among them 44 (8.9%) as a result of lack of efficacy, 55 (11.1%) because they did not fulfil the eligibility criteria for the randomization phase.
A few patients withdrew as a result of AEs during the double-blind period: one patient (0.5%) in study 1 in the agomelatine group and three patients (1.6%) in the placebo group, and in study 2, four patients (2.4%) in the agomelatine group and one patient (0.6%) in the placebo group. In both studies, the majority of the AEs were mild to moderate and the percentage of patients with at least one emergent AE was similar (study 1, agomelatine: 33.2% and placebo: 37.2%; study 2, agomelatine: 51.5% and placebo: 52.3%). Patients reporting at least one severe emergent AE were comparably distributed between the agomelatine and the placebo group. In both studies, the most common emergent AEs reported in the agomelatine group were regularly the same as for placebo: headache, gastroenteritis (study 1), nasopharyngitis and back pain.
In both studies, there were no clinically relevant mean changes in laboratory parameters and vital signs. One death occurred in the agomelatine group during the double-blind period of study 1, but the event was considered by the investigator as being unrelated to the study treatment.
The comparison of two agomelatine relapse prevention studies, the second of which was modified specifically to remedy the factors we hypothesized to account for failure of the first, is, to our knowledge, the first such formal publication. In both studies 1 and 2, the rate of relapse with agomelatine (20–25%) in the entire population is similar to the 20–28% relapse rate consistently obtained with other antidepressants, including duloxetine (Perahia et al., 2006), fluoxetine (Reimherr et al., 1998), venlafaxine (Simon et al., 2004), mirtazapine (Nierenberg et al., 2004) and escitalopram (Rapaport et al., 2004; Kornstein et al., 2006). Thus, the failure to show a significant difference between the treatment groups in the entire population of study 1 appears to have been driven solely by an unexpectedly low rate of relapses on placebo (∼25%). This is in contrast to the average placebo patient-relapse rate in the above-mentioned studies of about 40% (Geddes et al., 2003).
The severity of depression has been reported to be a predictor of relapse (McGrath et al., 2006), and our subanalysis of study 1 suggested that inclusion of more severely ill patients with depression is predictive of a better response to antidepressants relative to placebo. There is an extensive literature on this phenomenon (Khan et al., 2002; Khan and Schwartz, 2005; Blom et al., 2007; Montgomery and Kasper, 2007; Kilts et al., 2009; Fournier et al., 2010), but the key is how this is best achieved. It is often assumed to be completely captured by the score on a single rating of symptoms; our experience suggests that this is only the partial answer. Indeed, agomelatine study 1 used a HAM-D threshold of 22 and failed. Simply pushing the inclusion threshold on HAM-D to increasingly higher levels will reduce study feasibility and generalizability (and tempt investigators to inflate symptom ratings).
To recruit a more severely ill population, we proposed that it would be necessary to use measures in addition to a HAM-D minimum entry. Our criteria in study 2 included, first, a minimum subscore calculated from specific HAM-D items (1, 2, 5, 6, 7, 8, 10 and 13) that take account of Bech items (Bech et al., 1975), and items 5 and 6, which are reported to be sensitive in detecting changes with antidepressant treatment (Lecrubier and Bech, 2007; Santen et al., 2008). These items probe six of the nine symptoms listed in the DSM-IV definition of major depression. Second, patients were required to have a CGI-S of at least 4 and, third, the self-rating questionnaire Hospital Anxiety Depression Scale depression was included to avoid discordance between the clinician evaluation and the self-perception by the patient (Fava et al., 2003; Kobak et al., 2007). The proposed cut-off of 11 is recognized as the threshold for defining a depression state (Zigmond and Snaith, 1983).
In addition, the DSM-IV definition of patients with MDD requires that the symptoms must have caused significant distress of impairment in social, occupational or other important areas of functioning, although this is not usually formally assessed. A disability scale, the Sheehan questionnaire (Leon et al., 1992), was used to assess these impairments.
There were other minor changes in the protocol of study 2: patients could enter the continuation phase either at week 8 or 10, but this did not influence subsequent relapse rates or randomization rates. In addition, randomization criteria were blind for both the investigator and the patient. This methodological factor, aimed at managing investigators’ overestimation of change (Kobak et al., 2007), cannot explain per se the different findings in studies 1 and 2. Actually, the absence of blinded criteria in study 1 should have favoured the randomization of patients prone to be sensitive to clinical procedures.
Our protocol changes probably contributed towards the exclusion of such unsuitable patients. What made the difference between studies 1 and 2? The increased weighting of the eight core HAM-D items per se may make little difference because, in study 1, almost nine out of 10 patients included fulfilled this criterion. Moreover, neither the HAM-D total score nor the CGI recorded at baseline (either at inclusion or at randomization) was different between the studies; thus, the level of severity of included patients in both studies was similar when assessed by these scales. Even the apparently increased duration of the index episode was not statistically significant, although it may reflect a more impaired patient population in study 2. Scores on standardized rating scales are commonly used as an index of severity, but the classification of severity greatly depends on the scale used (Zimmerman et al., 2012). In addition, this approach contrasts with how severity is measured in clinical practice, where the focus might be more on functional impairment. Rather than a specific threshold or measure, it appears much more likely that the increased demands for recruitment made the key impact to improve patient selection. These innovations probably ensured identification of the dysfunction that is a defining characteristic of major depression and aligned the recruited patient sample closer with those patients normally seeking treatment in clinical practice. It is notable that, compared with the population of patients included in study 1, the inclusion of better characterized patients in study 2 led to a higher proportion of patients responding to treatment, with lower HAM-D and CGI-S scores at the end of the open period.
Of course, the present comparison has some limitations as both studies were not primarily designed with this in mind, but to assess a drug’s efficacy. The variables we have focused on certainly are the most important, but we cannot rule out that other potentially significant differences in the design of both studies (e.g. differences in numbers of study sites and countries) could have also added to some variance.
Ensuring inclusion of the relevant sample of patients is important from an ethical as well as a scientific perspective, as it minimizes the risks of nonconclusive studies (Walsh et al., 2002; Khan and Schwartz, 2005). The example of agomelatine is explicit in this respect as the above-mentioned innovations have influenced the design and implementation of the successful agomelatine trials initiated afterwards and published in recent years (Kennedy and Emsley, 2006; Olié and Kasper, 2007; Kasper et al., 2010).
The authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Trial registration name: study 1: not applicable – study 2: a study to determine the maintenance of efficacy of agomelatine to prevent relapse in outpatients with major depressive disorder. A 8–10 weeks open-period treatment with agomelatine followed by a 24-week randomized double-blind period, placebo-controlled, parallel groups and 20 weeks of optional double-blind treatment period.
Trial registration number: study 1: not applicable – study 2: ISRCTN53193024.
This study was sponsored by Servier (Suresnes, France).
Conflicts of interest
G.M. Goodwin, has held grants from Sanofi-Aventis and Servier, has received honoraria for speaking from AstraZeneca, BMS, Eisai, Lundbeck, Sanofi-Aventis, Servier and for advice from AstraZeneca, BMS, Lilly, Lundbeck, P1Vital, Sanofi-Aventis, Servier, Wyeth. P. Boyer, has received honoraria from Servier and from Wyeth for advice and for speaking or organizing training. R. Emsley, has participated in speakers/advisory boards and received honoraria from AstraZeneca, Bristol-Myers Squibb, Janssen, Lundbeck, Organon, Pfizer, Servier and Wyeth. He has received research funding from Janssen, Lundbeck and AstraZeneca. F. Rouillon, has received honoraria for advice from Servier, Janssen, Lilly, GSK, has participated in advisory boards from Lundbeck, Lilly, Organon, and Sanofi-Aventis, has received honoraria from Wyeth, BMS and Biocodex for speaking or study coordination. C. de Bodinat is an employee of Servier.
Bech P, Gram LF, Dein E, Jacobsen O, Vitger J, Bolwig TG. Quantitative rating of depressive states. Acta Psychiatr Scand. 1975;51:161–170
Blom MB, Spinhoven P, Hoffman T, Jonker K, Hoencamp E, Haffmans PM, van Dyck R. Severity and duration of depression, not personality factors, predict short term outcome in the treatment of major depression. J Affect Disord. 2007;104:119–126
Fava M, Evins AE, Dorer DJ, Schoenfeld DA. The problem of the placebo response in clinical trials for psychiatric disorders: culprits, possible remedies, and a novel study design approach. Psychother Psychosom. 2003;72:115–127
Fournier JC, DeRubeis RJ, Hollon SD, Dimidjian S, Amsterdam JD, Shelton RC, Fawcett J. Antidepressant drug effects and depression severity: a patient-level meta-analysis. JAMA. 2010;303:47–53
Geddes JR, Carney SM, Davies C, Furukawa TA, Kupfer DJ, Frank E, Goodwin GM. Relapse prevention with antidepressant drug treatment in depressive disorders: a systematic review. Lancet. 2003;361:653–661
Goodwin GM, Emsley R, Rembry S, Rouillon F. Agomelatine prevents relapse in patients with major depressive disorder without evidence of a discontinuation syndrome: a 24-week randomized, double-blind, placebo-controlled trial. J Clin Psychiatry. 2009;70:1128–1137
Guy W ECDEU Assessment Manual for Psychopharmacology. US Department of Health, Education and Welfare publication ADM 76-338. 1976 Rockville, MD National Institute of Mental Health:217–222
Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry. 1960;23:56–62
Kasper S, Hajak G, Wulff K, Hoogendijk WJ, Montejo AL, Smeraldi E, et al. Efficacy of the novel antidepressant agomelatine on the circadian rest-activity cycle and depressive and anxiety symptoms in patients with major depressive disorder: a randomized, double-blind comparison with sertraline. J Clin Psychiatry. 2010;71:109–120
Kennedy SH, Emsley R. Placebo-controlled trial of agomelatine in the treatment of major depressive disorder. Eur Neuropsychopharmacol. 2006;16:93–100
Khan A, Schwartz K. Study designs and outcomes in antidepressant clinical trials. Essent Psychopharmacol. 2005;6:221–226
Khan A, Leventhal RM, Khan SR, Brown WA. Severity of depression and response to antidepressants and placebo: an analysis of the Food and Drug Administration database. J Clin Psychopharmacol. 2002;22:40–45
Kilts CD, Wade AG, Andersen HF, Schlaepfer TE. Baseline severity of depression predicts antidepressant drug response relative to escitalopram. Expert Opin Pharmacother. 2009;10:927–936
Kobak KA, Kane JM, Thase ME, Nierenberg AA. Why do clinical trials fail? The problem of measurement error in clinical trials: time to test new paradigms? J Clin Psychopharmacol. 2007;27:1–5
Kornstein SG, Bose A, Li D, Saikali KG, Gandhi C. Escitalopram maintenance treatment for prevention of recurrent depression: a randomized, placebo-controlled trial. J Clin Psychiatry. 2006;67:1767–1775
Lecrubier Y, Bech P. The Ham D(6) is more homogenous and as sensitive as the Ham D(17). Eur Psychiatry. 2007;22:252–255
Leon AC, Shear MK, Portera L, Klerman GL. Assessing impairment in patients with panic disorder: the Sheehan Disability Scale. Soc Psychiatry Psychiatr Epidemiol. 1992;27:78–82
McGrath PJ, Stewart JW, Quitkin FM, Chen Y, Alpert JE, Nierenberg AA, et al. Predictors of relapse in a prospective study of fluoxetine treatment of major depression. Am J Psychiatry. 2006;163:1542–1548
Montgomery SA, Kasper S. Severe depression and antidepressants: focus on a pooled analysis of placebo-controlled studies on agomelatine. Int Clin Psychopharmacol. 2007;22:283–291
Nierenberg AA, Quitkin FM, Kremer C, Keller MB, Thase ME. Placebo-controlled continuation treatment with mirtazapine: acute pattern of response predicts relapse. Neuropsychopharmacology. 2004;29:1012–1018
Olié JP, Kasper S. Efficacy of agomelatine, a MT1/MT2 receptor agonist with 5-HT2C antagonistic properties, in major depressive disorder. Int J Neuropsychopharmacol. 2007;10:661–673
Perahia DG, Gilaberte I, Wang F, Wiltse CG, Huckins SA, Clemens JW, et al. Duloxetine in the prevention of relapse of major depressive disorder: double-blind placebo-controlled study. Br J Psychiatry. 2006;188:346–353
Rapaport MH, Bose A, Zheng H. Escitalopram continuation treatment prevents relapse of depressive episodes. J Clin Psychiatry. 2004;65:44–49
Reimherr FW, Amsterdam JD, Quitkin FM, Rosenbaum JF, Fava M, Zajecka J, et al. Optimal length of continuation therapy in depression: a prospective assessment during long-term fluoxetine treatment. Am J Psychiatry. 1998;155:1247–1253
Santen G, Gomeni R, Danhof M, Della PO. Sensitivity of the individual items of the Hamilton depression rating scale to response and its consequences for the assessment of efficacy. J Psychiatr Res. 2008;42:1000–1009
Sheehan DV, Harnett-Sheehan K, Raj BA. The measurement of disability. Int Clin Psychopharmacol. 1996;11(Suppl 3):89–95
Simon JS, Aguiar LM, Kunz NR, Lei D. Extended-release venlafaxine in relapse prevention for patients with major depressive disorder. J Psychiatr Res. 2004;38:249–257
Walsh BT, Seidman SN, Sysko R, Gould M. Placebo response in studies of major depression: variable, substantial, and growing. JAMA. 2002;287:1840–1847
Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatr Scand. 1983;67:361–370
Zimmerman M, Martinez JA, Friedman M, Boerescu DA, Attiullah N, Toba C. How can we use depression severity to guide treatment selection when measures of depression categorize patients differently? J Clin Psychiatry. 2012 [Epub ahead of print Sept 4]