Statistics from Altmetric.com
In May 2007 the Food and Drug Administration (FDA) ordered that all antidepressant drugs carry an expanded black-box warning incorporating information about an increased risk of suicidal symptoms in young adults aged 18–24 years. The new warning was based on the results of a FDA meta-analysis that included 372 placebo-controlled antidepressant trials and nearly 100 000 patients.1 On the basis of this analysis the relationship between antidepressant drug treatment and the incidence of reported suicidal behaviour in clinical trials was strongly related to age: the risk associated with drug treatment relative to placebo was found to be elevated in subjects under age 25, neutral in subjects aged 25–64 (reduced if suicidal behaviour and ideation are considered together), and reduced in subjects aged 65 and older.
Paradoxically, these data have been interpreted as reinforcing and consolidating two opposing positions. On one side, it has been argued that antidepressant drugs do not kill people, the real killer is untreated depression with its high lifetime risk of suicide.2 On the opposite side of the divide it is argued that the well documented risks associated with antidepressant drugs largely prevail over uncertain benefits.3
Both these positions seem very far from the everyday experience of treating depression. Under ordinary circumstances doctors have always considered suicide symptoms as useful and pragmatic measures of clinical outcome. If, after a reasonable period of time, suicidal thoughts or plans occur, persist, or get worse, then treatment is considered insufficient. If we apply this pragmatic approach to the FDA Report, a high degree of coherence between its main findings and the literature investigating whether antidepressants are effective in alleviating depressing symptoms becomes evident.
One key finding of the FDA report is the protective effect exerted by antidepressants in older subjects. This is consistent with randomised evidence showing that antidepressants are very effective in elderly individuals with major depression, with a number needed to treat (NNT) of 4 for tricyclics (implying that approximately four patients would need to be treated to produce one that recovers that would not have done so if given placebo), 8 for selective serotonin-reuptake inhibitors (SSRIs) and 3 for mono-amine oxidase inhibitors (MAOI).4 Similarly, the FDA finding of a neutral, or moderately positive, effect of antidepressant treatment in adults aged 25–64 is highly consistent with findings from several re-analyses of clinical trial data, where modest differences in response rates between antidepressants and placebo in adults aged 18–65 years were highlighted.5 ,6 Another key finding of the FDA analysis is the increased risk of reported suicidal behaviour calculated in people aged below 25 years. Again, this is consistent with a systematic review that suggested that tricyclic antidepressants may not be useful in treating depression in pre-pubertal children, while in adolescents there was only marginal evidence to support the use of drugs.7 Similarly, there is only weak and controversial evidence that SSRIs are efficacious in this age group.8 ,9 In terms of response, a systematic review and meta-analysis carried out in children and adolescents showed that the NNT was of uncertain clinical and statistical significance for paroxetine (NNT 12 (95% CI NNT benefit 5 to ∞ to NNT harm 20)) and sertraline (NNT 10 (95% CI 5 to ∞)). For citalopram and venlafaxine efficacy data were limited and pooled estimates could not be calculated, while for fluoxetine the treatment response appeared to be clinically worthwhile (NNT 5 (4–13)).8 Recent trials that compared the SSRIs with psychological treatments confirmed the beneficial effect of fluoxetine, which was significantly superior to cognitive behavioural therapy (CBT) in adolescents with moderate to severe major depressive disorder,10 and showed that combined CBT and an SSRI yielded no advantage over an SSRI, in the context of usual care.11 ,12
It is of interest that this coherence is evident not only at a class level but also at an individual drug level. In terms of suicidal behaviour or ideation, the FDA analysis revealed that, in adults, fluoxetine and sertraline, but not other SSRIs, were associated with a statistically significant protective effect; in terms of suicidal behaviour, sertraline was again associated with a statistically significant protective effect, while paroxetine was associated with a statistically significant increased risk (table 1). These figures are in line with a growing body of evidence suggesting that differences between individual SSRIs may exist, and that sertraline may have the edge in terms of efficacy over other SSRIs.13 ,14
The coherence between the FDA findings and the literature investigating the efficacy of antidepressants has some important implications for clinical practice. While the current debate about the possibility that antidepressants increase the risk of suicide in depressed individuals continues, doctors need reliable information on how effective antidepressants are in reducing the most disabling symptoms of depression, including suicide symptoms. The FDA report clearly indicated that age must be taken into account as a moderator of treatment effect. Doctors treating depression have classically considered severity of depression as a key moderator of treatment effect, with a continuum from lack of, or at least limited treatment effect in mild depression to beneficial effect in moderate to severe depression.15 In practice, this continuum has created difficulties because the threshold for prescribing remains an elusive concept.16 Similarly, it is expected that the continuum from increased risk of suicide symptoms in young age to reduced risk in older age will create analogous difficulties, especially in young adults, where the evidence for efficacy is ambiguous.
The FDA analysis highlighted two fundamental questions that the scientific community should urgently address. The first refers to the fact that doctors had been prescribing SSRIs for more than 20 years before the results of this re-analysis of clinical trial data became available. Most of the trials included in the FDA report were conducted several years ago and, consequently, the data could have been synthesised earlier and the results provided at least a decade ago. The time lag between the availability of evidence and its use to inform clinical practice was highlighted long ago17 and organisations involved in research synthesis have already accepted the need and taken responsibility to carry out timely and regularly updated re-analyses of clinical trial data. However, the FDA analysis, and other recent meta-analyses of clinical trial data,18 ,19 clearly showed the advantages of basing re-analyses on individual-patient data. This leads to the problem of data accessibility. The FDA report demonstrates that it is possible to obtain full and systematic access to individual-patient data—but it is well known that individual-patient data meta-analyses are difficult. Given the clinical importance of this problem we need a global initiative that aims to make all trial data accessible to selected organisations involved in research synthesis. A first step in achieving this would involve the medicines regulatory authorities. Regulatory authorities should require, before reviewing applications for the licensing of new drugs, a statement in which the applicant agrees that all data must be accessible to specific organisations involved in research synthesis.20 Without this agreement, regulatory authorities should not accept applications for new medicines: this should encourage drug companies to participate. This policy should apply to published and unpublished data, thus mitigating the occurrence of publication bias. Scientific journals could adopt a similar stance, requiring that agreement before accepting reports of clinical trials. This would allow accessing data from clinical trials that are never submitted to regulatory authorities—for example, those carried out by independent institutions. This global initiative would create a repository for clinical trials that might be accessed to answer questions on the beneficial and adverse effects of drugs.
The second problem is that re-analyses of clinical trial data should be regularly updated. New antidepressant drugs have recently been marketed and others will soon be available. As in any other field of medicine, doctors need to know whether the latest insights on conventional drugs similarly apply to newer compounds. Possibly, this information should become available less than decades after first marketing.
Competing interests: None.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.