Antidepressants do work after all

In the current issue, Andrade and colleagues describe naturalistic outcomes following antidepressant therapy in 808 people with major depressive disorder (MDD) in India. The authors adopt a prospective observational design, in which they compare the change in depressive symptoms from initiation of first antidepressant therapy through a 6-week follow-up. With direct-to-psychiatrist care being common in India, the authors were able to use the Montgomery–Åsberg Depression Rating Scale (MADRS) – a clinician part-rated assessment that is putatively sensitive to changes during treatment (Montgomery and Asberg, 1979) – within a representative, primary care cohort. Further, by recruiting 205 different psychiatrists across multiple regions of India, with each asked to recruit the same number of patients, the authors ensured that the diversity of the sample was considerable.

Of the 98.8% of patients who were followed up at 6 weeks, the vast majority received treatment with desvenlafaxine, vilazodone or escitalopram. The primary clinical outcome (50% reduction of MADRS score) was met in 66.1% of the sample, whilst MADRS remission (score <10) was recorded in 36.7% of patients. In terms of tolerability, patient-rated adverse events were mostly described as absent (32.3%) or minor (46.0%), whilst only 0.9% reported that the treatment was worse than the illness. Noted limitations of the study include its observational design, lack of systematic recording of specific adverse events, non-random recruitment of the sample, and a comorbid mental health diagnosis (mostly anxiety disorders) for 14.9% of the sample.

Despite these limitations, clinicians and patients should be encouraged by the key finding that two-thirds of patients responded to first-line antidepressant therapy when used in a routine, primary care setting. Indeed, clinical response might have been even greater if follow-up had been extended beyond 6 weeks and if no patients with comorbid anxiety disorders had been recruited. Moreover, desvenlafaxine – the most commonly prescribed antidepressant in the study – has relatively poor efficacy compared with other antidepressants (Cipriani et al., 2018), suggesting that different choices of antidepressants could improve outcomes even more.

The Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial was carried out in the United States and is frequently cited as a benchmark for response rates to antidepressant therapy (Rush et al., 2006). In STAR*D, 47% of patients responded to first-line therapy with citalopram, followed by more modest response rates to second-line therapies. Compared with the present (MDD-1) study, the more modest outcomes in STAR*D could in part be explained by the differences in study populations. Whereas MDD-1 recruited a naturalistic, primary care sample receiving their first ever antidepressant, STAR*D was a more tightly defined clinical trial population, most of whom had recurrent or resistant depression. Further, Andrade and colleagues speculate that depressed patients in India have a more favourable stress–support ratio and better trust in the medical system, which may help to explain the more promising outcomes from their study.

A criticism of the MDD-1 study is that, lacking a control arm, we cannot say that any benefits of antidepressants were not strongly driven by ‘placebo’ effects. Critics of antidepressants have asserted that pharmacological benefits of antidepressants are overestimated due to non-scrutiny or non-publication of negative clinical trials. In a 2008 meta-analysis of 35 clinical trials submitted to the US Food and Drug Agency, Kirsch and colleagues reported that drug–placebo differences increased as a function of initial severity, rising from almost no difference at moderate levels of initial depression to a relatively small difference for patients with very severe depression (Kirsch et al., 2008). Further, the relationship between initial depression severity and antidepressant efficacy was attributed to decreased responsiveness to placebo among very severely depressed patients, rather than increased responsiveness to antidepressants.

More recently, however, this negative view of antidepressants has become difficult to sustain. A 2018 network meta-analysis of 421 studies – including 86 unpublished studies from trial registries and pharmaceutical company websites – found that all antidepressants were more efficacious than placebo in adults with MDD. The overall effect size of 0.3 is comparable to accepted interventions in chronic physical diseases, such as chronic obstructive pulmonary disease and multiple sclerosis (Leucht et al., 2012). Subsequent to this 2018 meta-analysis, a randomised controlled trial (RCT) of 655 patients in primary care showed that sertraline was more effective than placebo at reducing Beck Depression Inventory-II and Generalised Anxiety Disorder-7 scores after 12 weeks (Lewis et al., 2019). This was notable because patients (only 54% of whom had ICD-10-defined depression) were specifically recruited where there was clinical uncertainty regarding any benefit likely from antidepressant therapy. This suggests that a greater proportion of patients might benefit from antidepressants than previously assumed.

A key consideration when assessing antidepressant efficacy is the primary outcome measure used. In the meta-analysis by Kirsch and colleagues, the primary effect parameter was change in total score on the Hamilton Depression Rating Scale (HDRS). Now over 50 years old (Hamilton, 1967), the HDRS presents several difficulties when used for this purpose and these were well known to Max Hamilton. In diagnosing MDD, all symptoms are not equal. Whereas one of the hallmark symptoms, low mood or anhedonia, must be pervasive to meet diagnostic criteria for MDD in the DSM-5 classification, no other specific symptoms – if there is a total of five – need be present. The HDRS, by contrast, scores 1 item for depressed mood alongside 16 other items, many of which (including hypochondriasis, somatic anxiety and gastrointestinal symptoms) are not even diagnostic criteria for MDD. Furthermore, several further items on the HDRS, including weight loss, headache and gastrointestinal symptoms, are common side-effects of antidepressants. As such, when using the total HDRS score as an outcome, improvement in core symptoms of depression will be diluted by other items, leading to probable underestimation of clinical benefit.

When core symptoms of MDD are specifically examined, the benefits of antidepressants become much clearer. In a patient-level, post hoc analysis (n = 6669 patients) of 32 drug–placebo comparisons in industry-sponsored trials of selective serotonin reuptake inhibitors, Hieronymus and colleagues compared the outcome of HDRS-sum score against the single item outcome depressed mood score (Hieronymus et al., 2016). While 18 out of 32 comparisons (56%) failed to separate active drug from placebo at Week 6 with respect to reduction in HDRS sum, only 3 out of 32 comparisons (9%) were negative when depressed mood was used as an effect parameter. This suggests that the frequent use of the HDRS sum as the primary outcome has maligned the reputation of antidepressants and potentially even hampered the development of novel compounds.

With a consensus from hundreds of clinical trials that antidepressants are more effective than placebo for acute treatment– particularly for the core symptoms of MDD – criticism of antidepressants has moved later into the antidepressant course. Critics have speculated that the potential of antidepressants to prevent relapse in clinical trials is over-stated, as observed relapse of depression following switch to placebo for a follow-up phase could in fact be an antidepressant discontinuation syndrome (so-called withdrawal confounding) (Hengartner, 2020). However, similar relapse rates are seen for various antidepressants against placebo (Hansen et al., 2008), regardless of their differing individual risk of discontinuation symptoms (Rosenbaum et al., 1998). Moreover, in an RCT of 241 people whose depression had responded to cognitive therapy – thereby negating the potential for pharmacological discontinuation – relapse rates were lower when patients were randomised to fluoxetine than placebo over 8-month follow-up (Jarrett et al., 2013).

In sum, Andrade and colleagues’ study adds to the compelling case that antidepressants do work after all, both in trials and in the clinic. Future studies should build on their findings through longer-term follow-up of the antidepressant course, including switches to second- and third-line therapies. Likewise, studies should examine naturalistically whether different treatments for depression (including both antidepressants and psychological therapies) are associated with differing longer-term response and relapse rates. In so doing, more patients may be enabled to achieve a timely remission and sustained recovery from depression.