Antidepressants, safety warnings and suicide risk in young people


There are some very emotive issues in clinical psychiatry, but few evoke comment as strongly as the prescribing of psychotropic medication to young adults or adolescents.

In 2003, the US Food and Drug Authority (FDA) issued a report recommending that antidepressant drugs, licensed for use in children and adolescents, carry a black box warning to highlight a possible increase in suicidal behaviour. The research findings underpinning this report were published in 2006 and showed, through meta-analysis, an increased relative risk of 1.95 for suicidal acts in children receiving antidepressants in comparison with placebo (Hammad et al, 2006).

This report was controversial and many have argued that it led to an under-recognition and under-treatment of adolescent depression with serious adverse effects. Seeking to explore this in further detail a recent study has been published in the BMJ and been the subject of much discussion and criticism (Lu et al, 2014).


The authors sought to investigate:

…if the warnings and related media coverage were associated with changes in antidepressant use and suicidal behaviour.

They then generated their hypothesis on the basis of two assumptions:

  1. Depression is an independent risk factor for suicidality
  2. Treatment with antidepressants is effective in reducing depressive symptoms


…we hypothesized that decreasing rates of overall antidepressant treatment after the warnings would be associated with a net increase in suicide attempts among young people.

Is there evidence that a reduction in teenage antidepressant use has led to an increase in suicide attempts?

Is there evidence that a reduction in teenage antidepressant use has led to an increase in suicide attempts?


In order to assess their hypothesis, the authors made use of data gathered from insurance providers and organised in a “Virtual Data Warehouse” that was collated at various research sites, before being combined in a final analysis at the main research site.

Sample cohorts

Their sample population was divided into three parallel cohorts:

  1. Adolescents (10-17 years old)
  2. Young adults (18-29 years old)
  3. Adults (30-64 years old)

No diagnostic criteria were applied to the cohort samples.

Outcome measures

In order to measure changes in prescribing and other behaviour the authors looked for varying rates of outcomes/markers in their cohort population:

  1. Antidepressant use: based on the number of dispensed antidepressant prescriptions
  2. Suicide attempts: the authors used a surrogate measure here (the number of incidents of poisoning by psychotropic medications)
  3. Completed suicides: Number of deaths with suicide returned as the cause of death

Statistical analysis

The cohorts were followed up over an approximately 10 year period from 2000 to 2010. Three time windows were identified by the authors:

  • Pre-warning period: 1st quarter 2000 to 3rd quarter 2003
  • “‘Phase in period”: 4th quarter 2003 to end 2004
  • Post warning period: 2005 to 2010

The “phase in period” was excluded from their analysis in order to prevent influence by “anticipation” of the warning and to allow clinicians to adapt their prescribing practice.

A statistical regression model was then used to estimate changes in the rate of outcome measure events.

The authors looked for antidepressant prescriptions and psychiatric medication overdoses. Image: Emuishere Peliculas CC BY-ND 2.0

The authors looked for antidepressant prescriptions and psychiatric medication overdoses. Image: Emuishere Peliculas CC BY-ND 2.0


The identified study cohorts consisted of approximately 1.1 million adolescents, 1.4 million young adults and 5 million adults per quarter.


  • Absolute decrease in antidepressant prescription rates of -0.70% (95% confidence interval (CI) -0.76 to -0.63)
  • Increase in rate of psychotropic drug poisonings of 0.002% (95% CI 0.0007 to 0.0033)
  • No change in rate of completed suicide

Young adults

  • Absolute decrease in antidepressant prescription -1.22% (95% CI -1.29 to -1.14)
  • Increase in psychotropic drug poisoning 0.0040% (95% CI 0.0033 to 0.0047)
  • No change in rate of completed suicide


  • A decrease in rates of antidepressant prescription -1.62% (95% CI -1.83 to -1.41)
  • No statistically significant change in psychotropic poisoning rate 0.0005% (95% CI -0.0006 to 0.0016)
  • No change in completed suicide rates
While overall suicide rates remained constant the authors claimed there was an increased rate of suicide attempts

While overall suicide rates remained constant, the authors claimed there was an increased rate of suicide attempts


The authors concluded:

Safety warnings about antidepressants and widespread media coverage decreased antidepressant use, and there were simultaneous increases in suicide attempts among young people. It is essential to monitor and reduce possible unintended consequences of FDA warnings and media reporting.


This paper has received a great deal of media attention – for example coverage by the BBC. The authors are quoted in this report as finding the results “disturbing”. The question then is whether the study can be interpreted in such a strong manner, I’m personally not so sure.

Surrogate outcome measures

Surrogate clinical measures are used when the event of interest is too rare or occurs too slowly to be detected within clinical research, unfortunately the use of surrogate measures has led to some dramatic errors in research practice (Yudkin and colleagues 2011).

In this case, the authors have selected psychotropic poisoning as a surrogate for attempted suicide – is this justified? The authors cite a paper that looks to identify suitable surrogate measures and algorithms to replace clinical codes of attempted suicide, given the poor reporting of this measure. In this paper, Patrick and colleagues identify that psychotropic drug poisoning detects approximately 40% of attempted suicides and have a positive predictive value of approximately 67% (i.e. one third of cases will be missed when using drug poisoning as a surrogate for attempted suicide). Rates of other acts, that may be suicidal in nature (for example cutting of wrists and upper leg) occur at lower rates but are not uncommon in adolescent populations. Overall it is not entirely clear, given these limitations, that the use of psychotropic poisoning as an outcome measure can support such strong conclusions.

Surrogate outcome measures and some findings contradicting previous research make the implications of this study unclear

Surrogate outcome measures and some findings that contradict previous research make the implications of this study unclear

Other measures of suicide attempts in adolescents

The authors of this paper purportedly demonstrate an increasing rate of attempted adolescent suicide throughout their period of observation. This finding is not in keeping with other epidemiological surveys however, which have suggested that attempted suicide rates may actually be decreasing in the US, for example as shown by the Youth Risk Behaviour surveillance system.


Through the use of a surrogate outcome measure, the authors have demonstrated a behavioural change following the introduction of the FDA black box warning, that is contradictory to previous epidemiological findings. The authors rightly highlight the importance of getting this information right as it is clearly of crucial importance that the risks of antidepressant prescription are adequately and correctly communicated (Geddes, Cipriani and Horne 2014).

Some commenters on the BMJ website have called for this paper to be withdrawn owing to its methodology and concluding message. I am not certain this is warranted, but I think that its findings certainly require careful interpretation and replication in other cohorts as they do not appear to be so clear cut as the authors’ concluding paragraph would suggest. This study cannot provide evidence for antidepressant efficacy or their impact on suicidality in adolescents, but can provide some indication of behaviour change.

The question is whether we think this, indirect, evidence of behavioural change should change our thinking in relation to antidepressant prescription?

We need clear clinical evidence in order to adequately support the decision to prescribe medication such as antidepressants

We need clear clinical evidence to prescribe antidepressants to young people

If you need help

If you need help and support now and you live in the UK or the Republic of Ireland, please call the Samaritans on 116 123.

If you live elsewhere, we recommend finding a local Crisis Centre on the IASP website.

We also highly recommend that you visit the Connecting with People: Staying Safe resource.


Lu CY, Zhang F, Lakoma MD, Madden JM, Rusinak D, Penfold RB et al. (2014). Changes in antidepressant use by young people and suicidal behavior after FDA warnings and media coverage: quasi-experimental study. BMJ, 348 (jun18 24), g3596–g3596. doi:10.1136/bmj.g3596.

Hammad TA, Laughren T, Racoosin J. (2006). Suicidality in pediatric patients treated with antidepressant drugs. Archives of General Psychiatry, 63(3), 332–339. doi:10.1001/archpsyc.63.3.332 [PubMed]

Yudkin JS, Lipska KJ, Montori VM. (2011). The idolatry of the surrogate. BMJ, 343, d7995. [PubMed]

Patrick AR, Miller M, Barber CW, Wang PS, Canning CF, Schneeweiss S. (2010). Identification of hospitalizations for intentional self-harm when E-codes are incompletely recorded. Pharmacoepidemiology and Drug Safety, 19(12), 1263–1275. doi:10.1002/pds.2037 [PubMed]

Geddes JR, Cipriani A, Horne R. (2014). Communicating the harmful effects of medicines. BMJ, 348(jun18 26), g4047–g4047. doi:10.1136/bmj.g4047 [PubMed]

Share on Facebook Tweet this on Twitter Share on LinkedIn Share on Google+