It has long been held that antidepressant drugs show a ‘delayed onset of action’ (as detailed in a NICE, 2004 guideline), a characteristic that encourages many prescribers to trial a particular antidepressant for weeks or months before judging whether it is truly effective or ineffective. A challenge to this ‘conventional wisdom’ (Tylee and Walters, 2007) emerged in the early 1990s, with Stassen et al. (1993) arguing that the phenomenon of a distinctly delayed onset interval was a myth and that the efficacy of an antidepressant drug can be predicted within the first week of treatment. One commonly suggested reason for the delayed-onset view is that most data sets evaluating the outcome of those receiving an antidepressant drug comprise nonresponders as well as responders, and that the inclusion of nonresponders in analysing aggregate data will artefactually distort and delay trajectories of improvement in true responders.
In their paper, Stassen et al. (1993) plotted Hamilton Rating Scale for Depression (HAM-D) scores for three groups – two receiving an antidepressant and one receiving a placebo. A trial ‘responder’ was defined by showing a 50% or greater reduction in their HAM-D score at the end of the 4-week trial, whereas ‘improver’ status was defined by a HAM-D score decrease of 20% or more at any stage of the trial. Trial ‘responders’ (whether receiving the active drug or the placebo) were more likely to have achieved ‘improver’ status in the first week, suggesting a very early onset of action. Subsequently, a number of studies have returned similar findings of ‘rapid onset of action’. For example, Taylor et al. (2006) analysed data from 28 placebo-controlled trials of selective serotonin reuptake inhibitors (SSRIs) involving nearly 6000 patients and reported that SSRI treatment was associated with ‘symptomatic improvement in depression by the end of the first week of use’. Thompson (2002) evaluated the dual-action antidepressant mirtazapine and described its onset of improvement as being evident in the first week.
In terms of mechanisms, Stassen et al. (1993) observed that, in ‘responders’, both the time course and the trajectory of improvement were exactly the same for those receiving the active drug as for those receiving the placebo, with the only difference across the two groups being that those receiving the active drug were more likely to be in the responder group. Those authors then conjectured that depressed patients have a biological predisposition to be a responder or not, and that antidepressant drugs may act to convert a percentage of intrinsic nonresponders into responders, thus triggering improvement.
We postulate another explanation for consideration. Most studies and meta-analyses pursuing the issue have analysed data from studies of major depression (which is likely to include a high representation of nonmelancholic depressed patients) and where the placebo response rate is high. For example, Walsh et al. (2002) extracted data from over seventy placebo-controlled studies of antidepressant drugs as treatments for major depression. They found a mean placebo response rate of 30% (range 12–52%) and 51% (range 32–70%) for the active drug, but also observed an increase in both placebo and drug response of some 7% per decade over the two preceding decades. Returning to the meta-analysis by Posternak and Zimmerman (2005), the percentage improvements of those assigned to the antidepressant drug and placebo groups, respectively, were 35 and 40% at week 1 and with almost equivalent additional weekly increases across the placebo and active drug groups from week 2 through to week 6. Thus, it is conceivable that those with major depression studied in drug trials have a high propensity for spontaneous and/or placebo remissions whether receiving active drug or placebo, and that the active antidepressant drug may simply ‘trigger’ such an improvement trajectory in many such study participants and, in others, may exert a ‘truer’ antidepressant effect more slowly – particularly in those with a more ‘biological’ depressive condition.
The extent to which studies include individuals experiencing a placebo response or spontaneous remission will compromise study effect size estimates and confound causality. Rapid nonspecific remission (whether individuals are in receipt of placebo or active drug) risks the individual meeting 20% ‘improvement’ and 50% ‘responder’ status simply as downstream consequences of the nonspecific response. In such instances, the predictor (improver) is not itself shaping the outcome (responder). As the percentage of study participants having such a non-specific response increases, the trajectory pattern of improvement will become increasingly more precipitous and so suggest a progressively briefer interval between initiation of medication and onset of ‘improvement’ in depression. In an extreme situation – for example, when almost all participants show such a pattern – ‘improvement’ status would be reached in a day or two and ‘responder’ status in less than a week, risking the conclusion that antidepressants work in a couple of days.
The general question (i.e. how rapidly do antidepressants start to work) may therefore require a quite different study paradigm if ‘work’ equates to ‘true’ efficacy. Study components might include studying a depressive condition showing a low placebo response rate and where any improvement is likely to reflect a specific antidepressant drug benefit. Melancholia is a key candidate, having longstanding historical ascriptions of primary biological underpinnings and a superior response to antidepressant medication than to psychotherapy (Parker et al., 1996) as well as a low placebo response rate quantified in the order of 20% (Peselow et al., 1992). Any such study is, however, somewhat dependent on the definition of ‘melancholia’. The Diagnostic and Statistical Manual of Mental Disorders, 4th ed. (DSM-IV) criteria for melancholia overlap distinctly with the criteria for major depression, thus risking DSM-defined melancholic patients actually having a nonmelancholic episode of major depression. As a consequence, the large-scale studies that have been carried out in the last decade and that have examined response patterns in subsets of DSM-IV-defined melancholia (McGrath et al., 2008) are not ideal for pursuing the hypothesis.
We have reported elsewhere (Parker et al., unpublished data) a study comparing the effectiveness of antidepressant medication compared with cognitive behaviour therapy (CBT) in a sample where patients fulfilled the DSM-IV criteria for melancholia at a stringent level and also fulfilled other clinically judged melancholia criteria. In brief, 18 patients received antidepressant medication whereas 11 patients received CBT for a 12-week period. At the 4-week review, those receiving antidepressant medication had shown significantly greater improvement than those receiving CBT, with little change reported in the CBT group. At 12 weeks, the response rates were again significantly higher in the patients receiving antidepressant medication than those receiving CBT (i.e. 67 vs. 37%). Thus, the lack of any distinctive improvement by four weeks in the CBT patients supported two of melancholia’s ascriptions (i.e. poor response to a psychotherapy and minimal placebo response rate) whereas the significant improvement at four weeks by those receiving the antidepressant supported the third ascription (i.e. that melancholia is selectively responsive to antidepressant medication). Although the sample size is very small, the lack of response over the first 4 weeks to psychotherapy suggests that recruitment ensured the exclusion of spontaneously improving patients and included those capable of a distinct antidepressant response. This makes the sample suitable to determine whether a rapid response to antidepressant medication is evident in those with a melancholic depressive subtype.
Patients and methods
Patients who fulfilled the criteria for melancholic depression over and above that required for a DSM-IV diagnosis were recruited by our facility’s research registry and through advertising. Of 167 patients screened [using a structured clinical interview schedule [Mini International Neuropsychiatric Interview (Lecrubier et al., 1997) and Quick Inventory of Depressive Symptoms>11 (Rush et al., 2003)] and clinical screening by the chief investigator], 33 patients were randomized to receive either the SSRI citalopram for 12 weeks, a broader antidepressant intervention strategy (initially commencing with desvenlafaxine) for 12 weeks or CBT for 12 weeks. The chief investigator was blinded to the randomization. Of the 32 patients accepting treatment assignments, 91% completed the study, allowing us to track improvement in 18 patients receiving antidepressant medication and 11 patients receiving CBT. In addition to the monthly administration of the HAM-D (Hamilton 1960), patients completed a daily measure of depression severity – the six-item Daily Rating Scale (DRS) (Parker and Roy, 2003), allowing us to quantify the severity of depression on a daily basis. We adopt the same paradigm advanced by Stassen et al. (1993) in defining improvement, responder status and determining the onset of action of an antidepressant drug. The mean score on the HAM-D before intervention was 17.5 (SD=5.6), with no significant differences before intervention between those who were allocated to medication or CBT groups or between those who were responders or in remission at the conclusion of the trial. The mean DRS score on day 1 was 32.1 (SD=10.8); again, there were no significant differences before the intervention between those allocated to medication or CBT groups, or between those classified as responders or in remission at the conclusion of the trial.
Of the 18 patients receiving antidepressant medication, there was a 61% improvement in the HAM-D measure over the 12 weeks, with 12 (67%) being ‘responders’ (in improving more than 50% by the end of the 12-week trial). We then sought to establish the interval from commencing medication to its recipients improving by 20%, examined in reference to DRS scores. Analyses aimed to quantify the percentage DRS score improvement from baseline to day 2 and then to each and every subsequent day so as to determine how many days elapsed before the marker of 20% improvement was reached. For the entire group of 18 patients receiving medication, the marker of 20% improvement was not achieved until day 12, whereas for the key group (the 12 formalized drug responders), the 20% improvement marker was achieved at day 10.
The current study is clearly a pilot (limited by a very small sample size) in testing the postulate that a ‘true’ antidepressant effect may not be as rapid as indicated in previous studies, and where their suggestion of an early onset of action may be artefactual, being determined and distorted by the representation of those showing a placebo response or a spontaneous response pattern. Although those receiving medication showed significantly greater improvement on the HAM-D compared with those receiving CBT at the first follow-up period after 4 weeks, the prediction of 12-week responder status from the ‘improvement’ status (equating to a 20% reduction in depression severity scores) was not achieved until day 10 by those receiving the antidepressant medication.
This study is limited by the size of the sample and the lack of a placebo arm.
The study is heuristic in the sense that it argues for an alternative strategy in applying the common paradigm of seeking to determine the interval required for antidepressant drugs to evidence ‘activity’. In essence, we recommend that studies seeking to identify the ‘true’ response pattern of antidepressant drugs and the interval before their demonstrable onset of action should be restricted to studies of those with a well-defined melancholic depression, and ideally comprising samples with placebo comparators and where those receiving the placebo medication showed either no improvement or minimal improvement over the study.
This study was funded by an Australian National Health and Medical Research Council Project Grant (630619). The olanzapine used in this trial was provided on request free of charge by Eli Lilly. Professor Parker is supported by a National Health and Medical Research Council program grant (510135) and a New South Wales Department of Health Infrastructure grant.
Conflicts of interest
The first author has served on advisory boards and given sponsored lectures for the manufacturers of the two trial antidepressant drugs. The olanzapine used in this trial was provided on request free of charge by Eli Lilly. The remaining authors have no conflicts of interest.
Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry. 1960;23:56–62
Lecrubier Y, Sheehan DV, Weiller E, Amorim P, Bonora I, Harnett Sheehan K, et al. The Mini International Neuropsychiatric Interview (MINI). A short diagnostic structured interview: reliability and validity according to the CIDI. Eur Psychiatry. 1997;12:224–231
McGrath P, Khan A, Trivedi M, Stewart J, Morris D, Wisniewski S, et al. Response to a selective serotonin reuptake inhibitor (citalopram) in major depressive disorder with melancholic features: a STAR*D report. J Clin Psychiatry. 2008;69:1847–1855
Depression: management of depression in primary and secondary care. National Clinical Practice Guideline 23. 2004 Great Britain The British Psychological Society, The Royal College of Psychiatrists
Parker G, Roy K. The development of a six-item daily self-report measure assessing identified depressive domains. J Affect Disord. 2003;73:289–294
Parker G, Hadzi-Pavlovic D, Boyce PParker G, Hadzi-Pavloic D. Issues in classification: II. classifying melancholia. Melancholia: a disorder of movement and mood. 1996 New York Cambridge University Press
Peselow E, Sanfilipo M, Difiglia C, Fieve R. Melancholic/endogenous depression and response to somatic treatment and placebo. Am J Psychiatry. 1992;149:1324–1334
Posternak M, Zimmerman M. Is there a delay in the antidepressant effect? A meta-analysis. J Clin Psychiatry. 2005;66:148–158
Rush AJ, Trivedi M, Ibrahim H, Carmody TJ, Arnow B, Klein DN, et al. The 16-Item quick inventory of depressive symptomatology (QIDS), clinician rating (QIDS-C), and self-report (QIDS-SR): a psychometric evaluation in patients with chronic major depression. Biol Psychiatry. 2003;54:573–583
Stassen H, Delini-Stula A, Angst J. Time course of improvement under antidepressant treatment: a survival analytical approach. Eur Neuropsychopharm. 1993;3:127–135
Taylor M, Freemantle N, Geddes J, Bhagwagar Z. Early onset of selective serotonin reuptake inhibitor antidepressant action. Arch Gen Psychiatry. 2006;63:1217–1223
Thompson C. Onset of action of antidepressants: results of different analyses. Hum Psychopharmacol. 2002;17:S27–S32
Tylee A, Walters P. Onset of action of antidepressants. BMJ. 2007;334:911
Walsh T, Seidman S, Sysko R, Gould M. Placebo response in studies of major depression. JAMA. 2002;287:1840–1847