Do antidepressants make children and adolescents suicidal?

Antidepressants are widely used to treat a range of mental illnesses experienced by children and adolescents. Treatment guidelines vary on the place of antidepressants in the treatment hierarchy for children and adolescents with depressive disorders but all agree that antidepressants have a place in the prescriber's armamentarium.1-3 Case studies identified that high doses of fluoxetine were associated with suicidal behaviour in adolescents dating back to 1991.4 Investigation at this time with available data suggested no link between antidepressants and treatment-emergent suicidal ideation or behaviour.5 In May 2003, GlaxoSmithKline advised the Food and Drug Administration (FDA) in the USA that paroxetine use was associated with suicide-related adverse events in paediatric patients in a clinical trial.6 At the end of 2003, the Medicines and Health Products Regulatory Agency, the UK's counterpart to the FDA, sent a letter to all doctors in the UK to warn them against using all but one of the antidepressants (fluoxetine) in children and adolescents. In October 2004, the FDA reached a split decision (15 yes, eight no) ordering that pharmaceutical companies add a ‘black box warning’ (BBW) on all antidepressant medication warning of the heightened risk of suicidal thoughts and behaviours.7 For a review of the history of the BBW, see Licinio and Wong.8 The Royal Australian and New Zealand College of Psychiatry, the Royal Australian College of General Practitioners and the Royal Australian and New Zealand College of Physicians released a joint statement in response in 2005 arguing for the continued access of children and adolescents to antidepressants but for their use within the context of a comprehensive management plan and careful monitoring mindful of the increased risk of suicidal thinking and behaviour.9 The FDA expanded its BBW about increased suicidal ideation and behaviour to include young adults 18–24 at the start of treatment (first 1–2 months) in 2007.10 Prior to the BBW, two-thirds of adolescents diagnosed with depression were treated with an antidepressant.11 Following the BBW, the rates of antidepressant dispensing in the USA fell to 58% lower than would have been expected from pre-BBW trends, with a decrease in the new diagnoses of paediatric depression.11 In the USA, paediatricians and primary care physicians demonstrated the greatest reduction in prescription of antidepressants following the introduction of the BBW.11, 12 This paper provides a narrative review of the link between suicide and antidepressant use in children and adolescents and reports on the breadth of evidence since the BBW. It is the aim of this paper to provide a range of reports and views in order to inform practitioners about the risks when prescribing. We identified the relevant studies by searching the Ovid Medline and Cochrane Reviews using the search terms ‘Adolescent Psychology/’ or ‘Child Psychology/’ or ‘Child Psychiatry/’ or ‘Adolescent Psychiatry/’ or ‘adolescent’ or ‘child.mp’ or ‘youth.mp’ AND ‘suicide.mp’ or ‘suicidal.mp’ or ‘suicidal ideation.mp’ or ‘suicide, attempted.mp’ or ‘Self-Injurious behavior.mp’ or ‘exp Self-injurious Behavior/’ AND ‘antidepressive agents’ or ‘Serotonin Uptake Inhibitors.mp’ or ‘SSRI.mp’ or ‘escitalopram.mp’ or ‘citalopram.mp’ or ‘fluvoxamine.mp’ or ‘paroxetine.mp’ or ‘sertraline.mp’ or ‘venlafaxine.mp’ or ‘mirtazapine.mp’ or ‘nefazodone.mp’ or ‘bupropion.mp’ AND ‘meta-analysis.mp’ or ‘Case-Control Studies/.mp’ or ‘Case-control.mp’ or ‘Epidemiological Methods/’ or ‘Epidemiological Studies.mp’ or ‘Epidemiological Studies/’ or ‘clinical trial.mp’ or ‘Clinical Trial/’ or ‘Toxicology/’ or ‘toxicology.mp’. The databases were searched for studies from December 1991 to December 2013. We also consulted bibliographies of other reviews. The focus of this narrative review was as follows: (i) meta-analyses of randomised controlled trials (RCTs); (ii) pharmacoepidemiological; and (iii) toxicological studies. The age of interest was 6–18 years. Studies that spanned across this age but overlapped with older cohorts of participants were considered for inclusion. The search of the databases revealed 153 papers. From these, the authors removed the duplicates, non-English papers and those that related to adult studies. Only those meta-analyses that assessed for placebo-controlled RCT were included. Meta-analyses that assessed for efficacy but not suicidal thoughts and behaviours were also excluded. From those remaining, 11 meta-analyses of placebo-controlled trials, 17 pharmacoepidemiological studies, one meta-analysis of observational studies and six post-mortem studies are reported. A meta-analysis of RCTs allows for the pooling of suicidal thoughts and behaviours across a number of different antidepressant trials resulting in an improved ability to determine a difference between antidepressant and placebo groups. Eleven meta-analyses are reported in Table 1. The meta-analyses varied in the RCTs that were chosen and available for inclusion. Whittington et al. in their analysis assessed each agent individually for adverse events.13 Mosholder and Willy assessed drug manufacturer data as provided by the FDA.14 However, Hammad assessed the FDA data after the suicide events had been reclassified by Columbia University owing to the poor and inconsistent classification of suicidal and self-harm events.6 Five RCTs (MDD) fluoxetine, paroxetine, sertraline and venlafaxine 21 RCTs (14 MDD, seven non-MDD- OCD, GAD, SAD) paroxetine, sertraline, venlafaxine, fluoxetine, citalopram, mirtazapine, nefazodone and fluvoxamine 20 RCTs (MDD, OCD, GAD, SAD) citalopram, fluoxetine, fluvoxamine, mirtazapine, paroxetine, sertraline and venlafaxine 22 RCTs (MDD, OCD, GAD and social anxiety) SSRIs and SNRIs 4 RCT (MDD, OCD) sertraline 24 RCTs (MDD, OCD, anxiety, ADHD) citalopram, fluvoxamine, paroxetine, fluoxetine, sertraline, venlafaxine, mirtazapine, nefazodone and bupropion 16 RCTs (MDD) fluoxetine, sertraline, citalopram, paroxetine, venlafaxine and mirtazapine 27 RCTs (MDD, OCD, non-OCD anxiety disorders) SSRIs, nefazodone, venlafaxine, mirtazapine 10 RCTs (depressive disorder) paroxetine, fluoxetine, sertraline, citalopram 17 RCTs (MDD) citalopram, escitalopram, fluoxetine, mirtazapine, paroxetine, sertraline and venlafaxine 35 RCTs (depression and other indications) SSRIs and atypical antidepressants While suicide was not reported in any of these RCTs, overall, the meta-analyses all reported an increased risk of suicidal thoughts and behaviours compared with placebo in the early stages of treatment in children and adolescents on SSRIs and other antidepressant treatments. However, the studies varied in their specific findings. Bridge et al. and Julious reported this increase in suicide risk when all antidepressants were pooled but not when sub-analyses by psychiatric condition were undertaken.20, 23 One study showed an increase in suicide risk when a more liberal fixed effects estimate was undertaken but not when a more conservative random effects regression was done.19 When individual antidepressants were considered, the risk ratio was consistently lowest for fluoxetine. Venlafaxine was the antidepressant most consistently found to be associated with suicidal thought and events.13, 15, 22 Children may be at higher risk of suicidal thoughts and behaviours when treated with antidepressants than adolescents.17 Those children and adolescents treated with antidepressants for anxiety disorders did not have an increased risk of suicidal thoughts and behaviours in three of the studies.16, 17, 20 The most recent Cochrane review (2012) included 19 published and unpublished RCTs (n = 3335 participants) that compared newer generation antidepressants with placebo in children and adolescents (6–18 years) who were diagnosed with a depressive disorder.22 This Cochrane review used two outcome measures: (i) suicide-related outcomes (thoughts and behaviours) as reported in Hammad6 and (ii) suicidal thinking as measured by the Suicidal Ideation Questionnaire-Junior High School Version (SIQ-Jr).22 For the suicidal-related outcomes, there was a 58% greater risk for those children and adolescents on the newer antidepressants versus placebo. When the data were further stratified by age, there was no statistical difference on suicide-related outcomes between antidepressant and placebo for adolescents (13–18 years).22 The finding of an increased risk of suicide-related outcomes was not significant for any individual antidepressant except for venlafaxine, which had a very wide confidence interval.22 Meta-analysis of SIQ-Jr data found that there was no difference in proportion with suicidal ideas between antidepressant and placebo.22 The risk difference between antidepressants and placebo across the meta-analyses varied with the FDA study of Hammad, which was at the upper end of the risk estimate at 2%,15 compared with Bridge et al. of 0.7%20 and Julious at 0.9%.23 This suggests that there is an increased risk of antidepressant-related suicidal thoughts or behaviours of up to 20 events per 1000 patients compared with placebo. Epidemiological studies aim to detect relatively rare outcomes, such as suicide, in large samples exposed to variables of interest (e.g. antidepressant treatment).24 By contrast, observational studies link administrative and clinical databases, allowing the investigation of possible links between antidepressant prescriptions and suicide attempts or suicide in very large samples from real-world clinical practice.24 Seventeen pharmacoepidemiological studies that included children and adolescents in ecological, case-control and cohort-designed studies are summarised in Table 2. Antidepressants or SSRI prescription were inversely related to the suicide rate.24-28 Five observational studies reported a gradient of suicide attempt risk. The highest risk was found to be in the month before starting the antidepressant and the next highest risk in the weeks and month after starting the antidepressant, declining thereafter.29, 33, 34, 36, 38 Other studies reported varying findings regarding suicide attempts. One study found an increase in suicide attempts and suicide in children and adolescents who had been admitted and treated with antidepressants for depression.30 Another study found an increased risk of suicide attempts but not death for antidepressants other than paroxetine,32 while another found no link between SSRIs and suicide.31 SSRIs were found to have the same suicide risk as tricyclic antidepressant (TCA),39 a higher risk than TCA35 and a lower risk compared with TCA.26 A compelling finding from the observational and case-controlled studies is the meta-analysis of Barbui, Esposito and Cipriani.41 Barbui, Esposito and Cipriani conducted a random effects meta-analysis of observation studies involving those patients treated for moderate or severe depression with SSRIs.41 They included eight studies of which five had child and adolescent participants.41 Barbui, Esposito and Cipriani found that exposure to SSRIs doubled the risk for suicide or attempted suicide in children and adolescents compared with those not exposed to any antidepressants (odds ratio 1.92), while for adults, the risk was decreased.41 For individual agents, paroxetine and venlafaxine increased the risk of suicide and suicide attempts, while citalopram, fluoxetine, fluvoxamine and sertraline were not significant.41 If antidepressants cause adolescents to become suicidal, it has been reasoned that it is very likely that antidepressants will be present in toxicology assays of adolescent suicide victims (see Table 3). Six toxicology studies assessed for the presence of antidepressants in adolescents who were assessed by the coroner to have died by suicide. Collectively, these findings suggest that it is reasonably uncommon for adolescents who have died by suicide to have been taking newer antidepressants, such as SSRI at therapeutic doses. The infrequent presence or absence of antidepressants at autopsy suggests either the adolescent was not prescribed the antidepressant or was not taking it in the days prior to their suicide. The meta-analyses described have consistently indicated that there is an increased risk of suicidal thoughts and behaviours in the order of between seven and 20 incidents per 1000 of those treated with an antidepressant compared with placebo. These findings occur most consistently when the data are pooled. With the exception of venlafaxine, subgroup analyses do not provide an association between specific antidepressant agents and suicidal thoughts and behaviours, a finding that may be related to small sample sizes resulting in analyses that are underpowered to detect a difference. It has been accepted by the FDA that a 2% increase in risks of suicidal thoughts and behaviours for children and adolescents enrolled in an RCT can be extrapolated to an increased risk of suicide of a depressed adolescent medicated in the community.7 It is however difficult to know how this suicide risk in clinical trials translates in office-based prescribing. The participants from predominantly pharmaceutical industry-sponsored RCTs used for the meta-analyses are different from those adolescents treated in the real world. In many of the RCTs, those adolescents who were suicidal at assessment or who had serious psychiatric comorbidity were excluded from the studies.48 Depressed adolescents enrolled in an RCT have been estimated to have half the rate of suicidal thoughts and behaviours of depressed youth in the community.49 There are methodological issues in pooling RCTs for a meta-analysis. The RCTs are predominantly studies in which suicidal thoughts and behaviours were not assessed prospectively. A review of a number of the controlled trials for the treatment of depression with antidepressants has found that the RCTs are quite heterogeneous with respect to recruitment, inclusion and exclusion criteria, study design, and outcome measures.48 Gibbons has argued that pooling of intent-to-treat person-level longitudinal data is a superior statistical method than meta-analysis.50 When this was done for RCT studies using fluoxetine, there was no increase in youth suicide risk over placebo.50 The main outcome measure used in the meta-analyses, the composite outcome of suicidal thoughts and behaviours, may not be the best single measure for assessing future risk for death by suicide. It has been argued that suicidal thoughts alone are a poor proxy for suicide. Baldessarini, Pompili and Tondo have opined that ‘suicidal ideation, as ascertained incidentally in trials, is a convenient but probably unreliable, nonequivalent, and potentially misleading surrogate for suicidal behavior (attempts or suicide)’ p. 246.51 Suicidal thoughts in older adolescents poorly identify those adolescents who will die by suicide. Mann et al. in a review estimated that for older adolescent boys, the ratio of suicidal thoughts to death by suicide is 9000:1, and for older adolescent girls, it is 19 000:1.5 Adolescents contemplate suicide more than adults. It remains unclear what the risk of dying by suicide is for those adolescents who are medicated for depression and who have suicidal thoughts. The ratio of suicide attempt to suicide in the older adolescent group was estimated to be 400:1 for boys and 3000:1 for girls.5 Those meta-analyses that have pooled suicidal thoughts and behaviours may have inadvertently mixed different risk factors (by an order of up to 20 times) for suicide. Further, if suicidal thoughts and behaviours are increased by antidepressant use by 20 per 1000 compared with placebo, then it may be very difficult to detect if antidepressants are increasing child and adolescent suicides. A problem with the meta-analyses described is that they use data from studies of up to 19 weeks.14 Valuck, Orton and Libby have reported on antidepressant risk for a suicide attempt noting that initiation of antidepressants refers to the first 55 days and early maintenance refers to days 56–179.38 Given that the pharmacoepidemiological studies suggest that the risk of suicide is highest prior to and following initiation of treatment, it is possible that the findings of increased risk from RCTs might have overestimated the risk during treatment that occurs beyond 19 weeks of antidepressant treatment. Studies that compare the treatment of depression in children and adolescents using medication and cognitive behavioural therapy have longer observation periods. The most recent Cochrane review comparing psychological therapies with antidepressants did not combine the suicidal-related adverse-related events in a meta-analysis owing to the diversity of assessment instruments used.52 However, with respect to suicidal thoughts, there was a higher level of suicidal thinking in the antidepressant-treated group compared with psychological-treated group following treatment and at 6–9 months of follow-up.52 It is unclear what the significance of this means other than to possibly reinforce the message of psychological interventions to be considered as first-line treatment for adolescent depression over antidepressant medication. Outcomes of the ecological studies suggest that there is an inverse effect with increased antidepressant prescription rates associated with a declining rate of suicide at a population level. The inference however is that antidepressants are not increasing suicide rate as would be expected if these medications indeed have suicide-inducing qualities. However, antidepressants are prescribed for a multitude of indications, and it cannot be assumed that all those antidepressant prescriptions are for depressed children and adolescents. The prescriptions might be in the main for use in conditions with a lower suicide risk than depression such as chronic fatigue or autism, thereby muddying the findings. Given that suicide is a final common pathway for a myriad of diagnoses, interacting risk factors, compliance with medication and individual dosing, it is a challenge to know how to extrapolate the finding of an inverse relationship between antidepressants and suicide rate at a population perspective to explain a given individual's risk of treatment with an antidepressant. The meta-analysis of observational studies found an increased risk of suicide and attempted suicide of nearly twice that of those not on an SSRI.41 This doubling of the risk is the same as the finding by Hammad from the FDA data.15 While these findings are reasonably convincing of an increased suicide risk, it must be remembered that there are limitations with observational studies. Unlike RCTs, patients in observational studies are not randomised. Those adolescents in observational studies with higher levels of depression are more likely to be given an antidepressant medication.24, 53 While toxicological studies have not found, or infrequently found, antidepressant medication at autopsy, there are a number of problems with the post-mortem toxicology studies. Not all those who died by suicide had toxicology assays. In some of the studies, they did not exclude those who had an injury to death period of less than 3 days.46 Antidepressants with a half as fluvoxamine or may been if the injury to death is more than Further, it has been suggested that antidepressant may be a that suicidal It can also be argued that those who are on antidepressants are less likely to die if they on their prescribed antidepressants, given the of newer antidepressant A further the of of children and adolescents who may be more to treatment-emergent suicidal There is some evidence that those children and adolescents who had more severe suicidal behaviours at assessment or with or were more likely to suicide-related which of children and adolescents are more at risk is a challenge for the There are from the on suicidal effects of antidepressants in children and adolescents for the The up the risks and before starting an antidepressant. The child or adolescent and their be in and with the risk of increased risk of to them prior to an antidepressant. Venlafaxine and paroxetine have been more than the other newer antidepressants and are not as a first-line treatment. The in medication mindful that the highest risk is likely to be in the early of treatment and when the is or Suicidal thoughts and behaviours be prior to starting the medication and early in the treatment. The UK guidelines suggest medication review for the first month and at review for the first 3 The likely by which antidepressants might increase suicidal thoughts and behaviours are summarised in Table for increased and in to suicidal thoughts and behaviours be conducted at each the medication to be then to is A study the monitoring of suicidal events is in have argued of the same available studies regarding the risk of suicidal events during antidepressant treatment. of the challenge is that suicide, the adverse of is relatively and each of the methodological used to the risk have their the suggests that there is a small but increased risk of suicidal thoughts and behaviours when using antidepressants in children and early in treatment. This review the monitoring of an adolescent over the weeks after an antidepressant and about the of suicidal thoughts and The joint of the of Psychiatry, General and Physicians on the use of antidepressants of the risks with treatment, early monitoring for suicidal thought and behaviours, and use within the context of comprehensive management plan remains a very relevant is to a who with a history of and social further a history of a history of poor of with suicidal thoughts but no suicide or attempts. There is a history of depression in and diagnoses with depressive and social anxiety and refers to a for cognitive behavioural therapy after eight of While there is some is with of and suicidal suggests a of fluoxetine for the treatment of and after the response to over has a with and about the of further therapy or a The risks and of each are and to the fluoxetine, as has had a response in the to this antidepressant. the common and rare effects of fluoxetine with and and for increased suicidal thoughts and that the risk of this is is at 10 in with further after a and reports and effects that had after 4 days a in At this there are no suicidal thoughts or not have any or increased or the to 20 and the end of the At this reports that anxiety has to to to for two per There is also some in and the fluoxetine at the review to is after weeks and to is for the next 12 when a the fluoxetine is The fluoxetine is and the are at over the