Two-week delay in onset of action of antidepressants: new evidence

Summary Many sources purport that antidepressants have a delayed onset of action, measured in weeks rather than days. Recentdata using weekly or daily mood ratings demonstrate that maximumimprovement occurs during the first 2 weeks, with some improvementwithin the first 3 days. Methodological issues may underliethe delayed-onset hypothesis.

INTRODUCTION

TOP

INTRODUCTION

The notion that onset of action of antidepressants takes severalweeks is widely quoted in clinical guidelines (Anderson etal, 2000; National Institute for Clinical Excellence, 2004)and has formed the basis for considerable biological and pharmacologicalresearch (Blier, 2003). It is also a notion that has beenthoroughly translated into clinical practice, where cliniciansregularly tell patients that the antidepressant is likely totake 2 to 4 weeks to start to work (Garfield et al, 2004).Yet this was not always the accepted view, and new data suggestthat this statement may be at best misleading and at worstinaccurate.

THE DELAYED-ONSET HYPOTHESIS

TOP

THE DELAYED-ONSET HYPOTHESIS

Conceptually, any drug that has a delayed onset of action wouldbe expected to show weak or negligible early efficacy but superiorefficacy in the medium term. This was examined in three antidepressanttrials by a group from Columbia University, New York; Quitkinet al (1987) observed that patients who achieved a sustainedresponse tended to do so after several weeks’ delay and,furthermore, those who responded early often failed to continueto full remission. Yet in the past 15 years this observationhas been challenged, because a delayed response has not beenreplicated by a number of independent groups (Dew et al, 2001;Szegedi et al, 2003). Posternak & Zimmerman (2005) recentlyconducted a meta-analysis of 47 double-blind placebo-controlledantidepressant trials encompassing 5100 patients on activemedication and 3400 on placebo. Across all studies when outcomewas defined as a reduction in Hamilton Rating Scale for Depression(HRSD) score, the authors showed that 23% of all differences between drug and placebo groups were already apparent by week1, and that 57% of possible differences had appeared by week2. Close inspection of responders in both arms for the exacttiming of their improvement revealed no differences in theantidepressant ${nu}$ placebo response rates during any week of treatment.Examination of antidepressant improvement trajectories in isolation was also valuable. It was found that 60% of improvement thatoccurred across all trials with antidepressants took placeduring the first 2 weeks. To look at this another way, in 80%of all trials the response was greater in weeks 1 and 2 thanit was either in weeks 3 and 4 or in weeks 5 and 6. On the basisof these data there seems to be little doubt that the largestimprovement per unit time produced by antidepressants occurswithin the first 2 weeks of treatment. Yet we still do notknow exactly how early within this period antidepressants beginto work – that is, whether there is any clinical delayin the onset of action.

MEASURING INITIAL ONSET OF THERAPEUTIC ACTION

TOP

MEASURING INITIAL ONSET OF...

If we are to address initial onset of action – that is,how long it takes a drug to begin to work – we shouldnot be satisfied with studies that use standard definitionsof response or remission, as neither of these accurately gaugesfirst onset. Although some authors have defined onset of actionas a 20% or 33% reduction in baseline severity scores, thisapproach conflates a true continuous variable (degree of sufferingfrom depression) with a pseudo-linear one (Parker et al, 1997).In addition, most clinical trials have not considered whatproportion of improvement is due to natural (untreated) remission,because few studies leave cohorts completely untreated. Of the handful of studies that have used a waiting-list arm (mostlypsychotherapy studies), the typical rate of spontaneous remissionof major depression (many studies allow waiting-list patientsto receive naturalistic treatment) is approximately 10% permonth, with a mean episode duration of 6 months (Posternak & Miller, 2001).This indicates that although both drugand placebo arms offer substantial benefit, some early remissionwould have occurred without intervention, particularly if patientswere recruited at peak depression severity.

When trying to measure initial onset of therapeutic action,a major methodological problem is finding an appropriate outcomemeasure (Leon et al, 2001). An ideal measure should have inherentlinearity across a broad range of psychopathological domainswhich translates into sensitivity to early change (Maier etal, 1988). This hypothetical tool should also be applied earlyand frequently to increase temporal resolution. The last pointrequires clarification. If there is even a possibility thatantidepressants begin to work within 2 weeks, how could thisbe detected if the first measurement takes place 2 weeks after starting the drug? The same argument holds for 1 week or even1 day. So although there are definite disadvantages of measuringmood too often in a long-term study, if a study is concernedwith early onset of action, measurement at 1 week (currentlyan unusually prompt measurement) is certainly too late. Thisapproach is not new. Studies of the response to electroconvulsivetherapy (ECT) often measure mood after each application two or even three times weekly (Husain et al, 2004). As an aside,it would be interesting to see how much of the variance inrapidity of onset of action of ECT compared with pharmacotherapyor psychotherapy is explained by differences in the frequency of assessment alone. Scales designed for regular daily moodratings have already appeared in other medical specialties (Peters et al, 2000). Although these have often been in theform of simple visual analogue scales or mood diaries, onegroup has used factor analysis to develop a daily mood scale for depression (Parker & Roy, 2003). Daily or even twicedaily manual or electronic mood diaries, although currentlyunfashionable, appear to be easy to use and reliable (Sherliker & Steptoe, 2000).Subjective self-rating methods are anequally important and powerful method of assessing first onset,and provide different information to classical objective scaleswhen compared head to head (Moller et al, 1996).

LARGE-SCALE STUDIES

TOP

LARGE-SCALE STUDIES

Some clinicians will remain unconvinced by evidence extractedfrom multiple small trials, albeit in the form of a meta-analysis (Posternak & Zimmerman, 2005). Support for this argumentwould be strengthened by at least one really large study withmeasures providing sufficient resolution in time. Three suchstudies have in fact been published involving 429 (Stassen et al, 1996),1277 (Stassen et al, 1997) and 369 patients (Parker et al, 2000) respectively. The results mirror thoseof the meta-analysis mentioned above. The Zurich group useddaily depression ratings on the HRSD and Zung Self-Rating DepressionScale. They found that regardless of which antidepressant thepatient was taking, there was a measurable early effect on day1. By day 3, 20% of patients had shown some improvement, andby day 7 50% had improved. Furthermore, 90% of those who showedany response during the first 3 weeks went on to become fullresponders. Drug–placebo differences (where apparent)could be detected as early as day 5. In the third study, Parker’sgroup examined responders and non-responders treated by 27Australian and New Zealand psychiatrists (Parker et al, 2000). Patients were requested to complete a self-report mood ratingevery third day. All patients showed a decrease in depression(and anxiety) within the first 3 days, but with little furtherimprovement in non-responders from days 4 to 6. Again, earlyimprovement within 1 week was a strong predictor of responder status.

Given these findings, it is pertinent to ask why the view that antidepressants have a delayed onset of action is so commonlyheld. There are two possibilities. The first is imprecisionconcerning the use of the term ‘onset of action’when we really mean ‘time to substantial remission’.This is understandable if we assume that patients want to knowwhen they will feel much improved rather than when they willbegin to feel a little better. The second is a failure to distinguishinitial therapeutic benefit, which occurs within days of startingan antidepressant, from the concept of drug ${nu}$ placebo separation,which accrues more slowly. In the Posternak and Zimmerman meta-analysisof antidepressant ${nu}$ placebo responses, the active arm showedmodest but definite early efficacy, but this was difficultto separate from a significant placebo response until a weekor more of cumulative benefit had occurred. These limitationsare amplified by relying on relatively blunt instruments appliedinfrequently by observers. The implication for research isthat sensitive daily measures are required to elucidate whetherthese early patterns of response are dependent on biological correlates of antidepressant or placebo treatment (Mayberg et al, 2002).The implication for clinical practice is thatcurrent evidence suggests it would be more accurate to sayto patients that in 90% of cases substantial improvement occurswithin the first 2 weeks, but that benefit continues to buildup over several weeks. (In the meta-analysis by Posternak & Zimmerman (2005), 60 out of 66 study cohorts on active medicationshowed a reduction in HRSD score of 50% or more within 2 weeks.)In those who have shown no response by 2 weeks there appearsto be a law of diminishing returns, which suggests that it maybe pertinent to re-examine another commonly quoted recommendation– that an antidepressant trial must be at least 6 to8 weeks before switching drugs.

REFERENCES

TOP