9
Introduction It is hard to imagine the anguish experienced by the parents, relatives, and friends of a child who has taken his or her own life. That such an event could be precipitated by a supposedly beneficial drug is a catastrophe (Editorial, The Lancet, April 24, 2004). Youth suicide is a major public concern: it is tragic and senseless. Although youth suicide is rare, it contributes to a large loss of potential life years and warrants both major scientific and social attention as well as efforts of prevention. When major depressive disorder presents in childhood and adolescence, it is often associated with substantial morbidity and ele- vated risk of suicide [1]. Treatment with antidepres- sants, especially selective serotonin reuptake inhibitors (SSRIs) is recommended in clinical guide- lines, as part of an integrated treatment effort for depressive disorders in childhood and adolescence [2]. However, psychopharmacological antidepressant treatment of depressed children and adolescents is controversial. First, even professionals might have difficulty in accepting children as being severely de- pressed and some might argue that antidepressant Lars Sønderga ˚rd Kajsa Kvist Per K. Andersen Lars V. Kessing Do antidepressants precipitate youth suicide? A nationwide pharmacoepidemiological study Accepted: 9 January 2006 / Published online: 23 February 2006 j Abstract The association be- tween treatment with selective serotonin reuptake inhibitors (SSRIs) and suicide in children and adolescents on the individual and ecological level were exam- ined in a nationwide Danish pharmacoepidemiological regis- ter-linkage study including all persons aged 10–17 years treated with antidepressants during the period 1995–1999 (n = 2,569) and a randomly selected control pop- ulation (n = 50,000). A tripartite approach was used. In Part 1, changes in youth suicide and use of antidepressants were examined. In Part 2, we made an assessment of youth suicide characteristics. In Part 3, we analysed the relative risk (RR) of suicide according to antidepressant treatment cor- rected for psychiatric hospital contact to minimize the problem of confounding by indication. The use of SSRIs among children and adolescents increased substan- tially during the study period, but the suicide rate remained stable (Part 1). Among 42 suicides nationally aged 10–17 years at death, none was treated with SSRIs within 2 weeks prior to suicide (Part 2). There was an increased rate of suicide associated with SSRIs (RR = 4.47), however, not quite significant (95% CI: 0.95– 20.96), when adjusted for severity of illness (Part 3). Conclusively, we were not able to identify an asso- ciation between treatment with SSRIs and completed suicide among children and adolescents. j Key words antidepressants – youth suicide – adverse reac- tions – adolescent psychiatry – pharmacoepidemiology ORIGINAL CONTRIBUTION Eur Child Adolesc Psychiatry (2006) 15:232–240 DOI 10.1007/s00787-006-0527-6 ECAP 527 L. Sønderga ˚rd, MD, (&) Prof. L. V. Kessing, MD, DMSC Dept. of Psychiatry University Hospital of Copenhagen Rigshospitalet Blegdamsvej 9 2100 Copenhagen Ø, Denmark Tel.: +45/35456237 Fax: +45/35456218 E-Mail: [email protected] K. Kvist, MSC Prof. P.K. Andersen, PhD, DMSC Dept. of Biostatistics University of Copenhagen Copenhagen, Denmark

Do antidepressants precipitate youth suicide?

Embed Size (px)

Citation preview

Page 1: Do antidepressants precipitate youth suicide?

Introduction

It is hard to imagine the anguish experienced by theparents, relatives, and friends of a child who has takenhis or her own life. That such an event could beprecipitated by a supposedly beneficial drug is acatastrophe (Editorial, The Lancet, April 24, 2004).

Youth suicide is a major public concern: it is tragicand senseless. Although youth suicide is rare, itcontributes to a large loss of potential life years andwarrants both major scientific and social attention aswell as efforts of prevention. When major depressive

disorder presents in childhood and adolescence, it isoften associated with substantial morbidity and ele-vated risk of suicide [1]. Treatment with antidepres-sants, especially selective serotonin reuptakeinhibitors (SSRIs) is recommended in clinical guide-lines, as part of an integrated treatment effort fordepressive disorders in childhood and adolescence[2]. However, psychopharmacological antidepressanttreatment of depressed children and adolescents iscontroversial. First, even professionals might havedifficulty in accepting children as being severely de-pressed and some might argue that antidepressant

Lars SøndergardKajsa KvistPer K. AndersenLars V. Kessing

Do antidepressants precipitate youthsuicide?

A nationwide pharmacoepidemiological study

Accepted: 9 January 2006 /Published online: 23 February 2006

j Abstract The association be-tween treatment with selectiveserotonin reuptake inhibitors(SSRIs) and suicide in childrenand adolescents on the individualand ecological level were exam-ined in a nationwide Danishpharmacoepidemiological regis-ter-linkage study including allpersons aged 10–17 years treatedwith antidepressants during theperiod 1995–1999 (n = 2,569) anda randomly selected control pop-ulation (n = 50,000). A tripartiteapproach was used. In Part 1,changes in youth suicide and useof antidepressants were examined.In Part 2, we made an assessmentof youth suicide characteristics. InPart 3, we analysed the relativerisk (RR) of suicide according toantidepressant treatment cor-rected for psychiatric hospitalcontact to minimize the problem

of confounding by indication. Theuse of SSRIs among children andadolescents increased substan-tially during the study period, butthe suicide rate remained stable(Part 1). Among 42 suicidesnationally aged 10–17 years atdeath, none was treated with SSRIswithin 2 weeks prior to suicide(Part 2). There was an increasedrate of suicide associated withSSRIs (RR = 4.47), however, notquite significant (95% CI: 0.95–20.96), when adjusted for severityof illness (Part 3). Conclusively, wewere not able to identify an asso-ciation between treatment withSSRIs and completed suicideamong children and adolescents.

j Key words antidepressants –youth suicide – adverse reac-tions – adolescent psychiatry –pharmacoepidemiology

ORIGINAL CONTRIBUTIONEur Child Adolesc Psychiatry (2006)15:232–240 DOI 10.1007/s00787-006-0527-6

EC

AP

527

L. Søndergard, MD, (&)Prof. L. V. Kessing, MD, DMSCDept. of PsychiatryUniversity Hospital of CopenhagenRigshospitaletBlegdamsvej 92100 Copenhagen Ø, DenmarkTel.: +45/35456237Fax: +45/35456218E-Mail: [email protected]

K. Kvist, MSCProf. P.K. Andersen, PhD, DMSCDept. of BiostatisticsUniversity of CopenhagenCopenhagen, Denmark

Page 2: Do antidepressants precipitate youth suicide?

treatment teaches children and adolescents to solveemotional problems pharmacologically rather than byemotional and social competence. Second, it is diffi-cult to evaluate the effect of antidepressant pharma-cological treatment in this group and this may haveled to contradictory findings [3–5]. Third, it is pos-sible that treatment with SSRIs worsens youthdepression, facilitates suicidal ideation and increasesthe risk of suicide. Case reports and a British databaseof adverse drug reactions used in pharmacovigilance[6] suggest such a relationship; however, otherresearchers have claimed that these data are severelybiased [7]. Further, results from studies on adultshave not strengthened the hypothesis that treatmentwith SSRIs increases the risk of suicide [8, 9].

In the UK, the Committee on Safety of Medicines[10] and in the US, the Food and Drug Administration(FDA) [11] have warned against treating children andadolescents with SSRIs because of the putative risk ofincreased youth suicide. These warnings have beenconsented by the National Boards of Health in severalcountries including Denmark. In a recent FDA anal-ysis of 24 clinical trials, risk ratios for pooled analyseswere found ranging from 1.7 to 2.2 with respect tosuicidal ideation in drug versus placebo groups,across different antidepressant medications [12]. Itshould be emphasized that no completed suicidesoccurred in these trials.

If treatment with SSRIs facilitates youth suicide, itwill be methodologically difficult to evaluate and nostudy has yet shown an increased suicide rate onindividualized data. Suicide is a relatively rare eventand when suicide occurs in children and adolescentstreated with SSRIs, suicide coincides with thedepression itself; i.e., analyses of the association be-tween SSRIs and suicide will be influenced by con-founding by indication [13].

In the present study, we aimed to assess the rela-tion between SSRIs and suicide among children andadolescents by using a tripartite approach in anationwide pharmacoepidemiological study includingdata from a 5-year period.

Methods

Part 1: We compared the annual purchase of antide-pressants among children and adolescents aged 10–17 years with the annual national suicide rate.

Part 2: Every suicide among children and adoles-cents treated with antidepressants initiated when aged10–17 years in Denmark from 1995 to 1999 wasidentified, and psychiatric hospital contact prior tosuicide, duration of treatment and other characteris-tics were assessed.

Part 3: In a cohort study, we compared the suiciderate among children and adolescents purchasing an-tidepressants first time when aged 10–17 years withthe rate among children and adolescents unexposedto antidepressants, correcting for gender, age andpsychiatric hospital contact.

j Data sources

We obtained data from four Danish registers.The Danish Medicinal Product Statistics [14, 15]

was established under the Danish National Board ofHealth auspice and information on every despatchaccording to prescriptions redeemed in Danishpharmacies has been collected since 1994. Data hasbeen released for research from 1 January 1995 asthe first year is regarded as a running in period.Since 1995, all children by law had issued their ownhealth insurance certificate enabling prescriptionsmade to children to be identified. In Denmark, allmedication for depression is dispensed on doctors’prescription only, and is purchased only at phar-macies. This register was used in all three parts ofthe study.

The Cause of Death Registry, the Danish NationalBoard of Health [16], records the causes and dates ofall deaths in Denmark, including suicide as entered onthe Cause of Death Certificates at the time of death.Suicide was identified when the cause of death wascoded as intentional self-harm (ICD-10 codes: X600–X840). This register was used in all three parts of thestudy.

The Danish Psychiatric Central Research Register[17] contains diagnoses and information aboutduration of contact to the psychiatric hospital ser-vices, including both inpatient and outpatient psy-chiatric services. This register was used in parts 2 and3 of the study.

The Danish civil registration system, authorized bythe Ministry of the Interior and Health [18] contains aunique personal identification number for all indi-viduals residing in Denmark. This identificationnumber enables encryption, retrieving and merging ofindividual data from different databases.

j Sampling

Part 1: When comparing the annual purchase of an-tidepressants during the period from 1995 to 1999with the exact annual national number of youth sui-cides, the latter was obtained from The Danish Na-tional Board of Health. We had no access toinformation (i.e., psychiatric diagnoses) in The Dan-ish Psychiatric Central Research Register for those

L. Søndergard et al. 233Antidepressants and youth suicide

Page 3: Do antidepressants precipitate youth suicide?

persons who were not included in the study sample asdefined beneath. We obtained information aboutpurchase of antidepressants from The DanishMedicinal Products Statistics.

Parts 2 and 3: All patients aged 10–17 years whohad purchased antidepressants according to one ormore prescriptions from 1 January 1995 to 31December 1999 were identified in the DanishMedicinal Product Statistics as exposed. Patientsincluded were all followed in time until end of study(31 December 1999) irrespective of age. The exposedpatients were divided into three categories accordingto major similarities in pharmacological mechanismof the purchased antidepressant. First, antidepres-sants were classified into newer and older antide-pressants. Second, newer antidepressants weresubdivided into SSRIs (citalopram, fluoxetine, flu-voxamine, paroxetine and sertraline) and newer non-SSRI antidepressants (mirtazapine, venlafaxine, ne-fazodone and reboxetine). The group of older an-tidepressants consisted mainly of tricyclicantidepressants, TCAs (amitriptyline, amoxapine,clomipramine, dosulepin, doxepin, imipramine, lo-fepramine, nortiptyline, opipramol, protriptyline andtrimipramine) and to a minor extent tetracyclic an-tidepressants (maprotiline and mianserin) andmonoamine oxidase inhibitors (isocarboxazid andmoclobemide).

An unexposed control population of 50,000 chil-dren aged 10–17 years was randomly selected on 1January 1995 (25,442 boys and 24,529 girls). Twentynine persons were excluded as they either died, emi-grated or reached an age of 18 on this date. Some ofthe remaining individuals later purchased antide-pressants, and thereby dynamically changed statusfrom unexposed to exposed individuals from thattime point.

Demographic data on emigration were obtainedfrom Statistics Denmark. Neither exposed norunexposed individuals contributed to time at risk ofsuicide when not resident in Denmark due to emi-gration.

When comparing the three parts of the study itneeds to be emphasized that Parts 1 and 2 representdifferent approaches. Part 1 concerns all children andadolescents aged 10–17 years who committed suicide.Part 2 concerns all children and adolescents who wereaged 10–17 years and treated with antidepressants atinclusion. Followed through a 5-year study periodsome reached an age older than 17 years (up to22 years). However, as all patients treated with an-tidepressants are included in Part 2, it is possible todeduce how many of them who committed suicide(Part 1) was treated with antidepressants, nationally.

The study was approved by the Danish Data Pro-tection Agency.

j Statistical analyses

Part 1 and 2: none.Part 3: In a cohort study, we investigated whether

antidepressant treatment predicted increased risk ofsuicide when adjusting for psychiatric hospital serviceby using Poisson regression analyses.

Poisson regression analyses are standard multipleregression models for incidence rates [19] in whichnumbers of suicides and person-years at risk arecomputed and analysed in subgroups given by cova-riates. In addition to treatment with SSRIs, the fol-lowing covariates were considered: gender, age group(10–17 vs. 18–22), and previous psychiatric admis-sions (yes/no). Censoring was done at death fromother reasons than suicide, emigration and at end ofthe study (31 December 1999).

Results

Part 1: From 1995 to 1999 the use of SSRIs amongchildren and adolescents increased substantially(Fig. 1). This pattern was the same for both boys andgirls (results not shown). The prescribing rate of an-tidepressants was found to be increasing with age andonly 23.3% of children and adolescents treated withantidepressants were in the age group of 10–13 years.

In the period from 1995 to 1999, 2,569 children andadolescents purchased antidepressants: 974 boys and1,595 girls, respectively. Among those purchasingantidepressants, 77.7% purchased SSRIs (Table 1).Nevertheless, there were only minor fluctuations inthe youth suicide rate during the study period (Ta-ble 2). Nationally among the 5.5 million inhabitantsin Denmark, 42 children and adolescents aged 10–17 years committed suicide from 1995 to 1999: 37boys and 5 girls, respectively (Table 2).

Part 2: In the study sample, as some of the unex-posed later changed status, we found totally 51,731

0

50

100

150

200

250

300

350

400

450

500

1995 1996 1997 1998 1999

FluoxetinCitalopramParoxetineSertralineFluvoxamine

Fig. 1 Annual number of children and adolescents purchasing SSRIs 1995–1999 in Denmark

234 Eur Child Adolesc Psychiatry, (2006) Vol. 15, No. 4ª Steinkopff Verlag 2006

Page 4: Do antidepressants precipitate youth suicide?

children and adolescents aged 10–17 years at inclusion.When this group was followed from 1995 to 1999 wefound that 19 persons had committed suicide: 14 boysand 5 girls (Table 3). The youngest boy was 15 yearsold and the youngest girl was 17 years old with amedian age of 18 and 20 for the two sexes, respectively.Among these 19 children and adolescents who com-mitted suicide, 5 had purchased antidepressants at leastonce during the study period. These five patients wereall girls. Four of these five girls had reached an age olderthan 17 year at the time of suicide.

Among those who committed suicide we foundonly citalopram, fluvoxamine, mianserin and par-oxetine represented; none had been treated with flu-

oxetine, any newer non-SSRI antidepressant, anyolder antidepressant or lithium. On average these fivegirls who committed suicide had purchased 163 De-fined Daily Doses (DDDs) [20] of any SSRI and anaverage of 866 DDDs of various non-antidepressantmedications (benzodiazepines, antihistamines, anti-asthmatics, antibiotics, steroids, estrogens and anti-convulsants). Assuming a prescribed dose of oneDDD a day and excluding DDDs intended to be usedafter the date of death, a cumulative average use of 95DDDs of SSRIs can be estimated (Table 3). For the 5girls who committed suicide, the average time fromfirst purchase of antidepressants to death was415 days (range: 99–1,029 days). The average time

Table 2 Youth suicide in Denmark 1995–1999

Year 1995 1996 1997 1998 1999

n r n r n r n r n r

Boys aged10–14 1 0.7 1 0.7 3 2.1 2 1.4 3 2.015–17 5 5.2 4 4.2 8 8.9 6 6.9 4 4.7Girls aged10–14 0 0 0 0 0 0 0 0 0 015–17 0 0 1 1.1 3 3.5 1 1.2 0 0

Note: n = number; r = suicide rate per 100,000

Table 1 Distribution according tosex and type of antidepressantmedication

Boys (n = 974) Girls (n = 1,595) Total (n = 2,569)

Purchase of antidepressantsSSRI 700 (71.9%) 1,296 (81.3%) 1,996 (77.7%)Newer non-SSRI 57 (5.9%) 111 (7.0%) 168 (6.5%)Older 264 (27.1%) 289 (18.1%) 553 (21.5%)

Note: The percentages in each column do not add up to 100% as patients may purchase more than one type ofantidepressants.

Table 3 Characteristics of youthsuicide Unexposed to SSRIs

(n = 14)Exposed to SSRIs(n = 5)

SexMale 14 (100%) 0 –Female 0 – 5 (100%)Age at suicide (median, range) 20 (15–21) 18 (17–21)Purchase of medicinea

SSRIs – 163 (56–429)SSRIs. Corrected for time of deathb – 95 (56–164)Any other 0 866 (8–2,144)Contact to psychiatric hospitalc, Cumulative days (mean, range)Inpatient 0.1 (0–1) 398 (119–730)Outpatient 1.9 (0–27) 53 (25–85)

aCumulative Defined Daily Doses, DDD (mean, range)bProvided one DDD a day, DDDs intended to be used after the date of death are not includedcPatients may have had both inpatient and outpatient contacts

L. Søndergard et al. 235Antidepressants and youth suicide

Page 5: Do antidepressants precipitate youth suicide?

from end of antidepressant treatment to death was118 days (range: 0–262 days) provided a dose of 1DDD a day and treatment to be continued underadmittance to hospital.

Even given such conditions, only two adolescentswere treated with SSRIs within 2 weeks of suicide inDenmark in this 5-year period. Both had reached anage older than 18 years in the follow-up period. Thus,no children in Denmark aged 10–17 were treated withantidepressants at the time of suicide in the 5-yearstudy period. Strikingly, none of the fourteen boys,found in the study sample, who committed suicide wastreated with any antidepressant or any other medica-tion, either for psychiatric or for somatic illness.

Among those who committed suicide, none died byintoxication with antidepressants.

Among those five girls treated with SSRIs, four hadbeen admitted to hospital within two months or lessof suicide and, in addition, three had had an outpa-tient course (Table 3). These four girls had on averagecumulatively been admitted to psychiatric hospital formore than a year from 1995 to 1999. The followingmain diagnoses were represented on discharge: or-ganic psychosis, schizophrenia, eating disorder andpersonality disorder; however, none had a diagnosisof depression, either as a main diagnosis or as anauxiliary diagnosis. Two persons had previously beendiagnosed with an attempted suicide, but were nottreated with SSRIs prior to these attempts. Among the14 boys, the use of psychiatric service was negligible.

When combining the results in Part 2 with Part 1 itcan be deduced how many of those who committedsuicide nationwide had been treated with antide-pressants at any time in the study period. Nationallywe found none of the 37 boys in Part 1 who com-mitted suicide when aged 10–17 years was treatedwith antidepressants from 1995 to 1999. Nationally wefound one of the five girls in Part 1 who committedsuicide when aged 10–17 years was treated with an-tidepressants from 1995 to 1999.

Part 3: A total of 2,311 children and adolescentsaged 10–17 years at inclusion purchased SSRIs duringthe follow-up period. A total of 1,326 (57.4%) had hada psychiatric hospital contact (in- or outpatient) atleast once in their life prior to their first purchase.Only 1.1% of the control persons had used a psychi-atric hospital service.

In the first model, we investigated the rate ratio ofsuicide by Poisson regression analyses and foundsignificant main effects of age, sex and treatment(Table 4, upper part). Those aged 18 or below had amarkedly and statistically significant lower suiciderate than those who were older. Further, girls had amarkedly and statistically significantly lower rate ofsuicide than boys. Those treated with SSRIs had ahighly statistically significant and strongly increased

rate of suicide compared to those not treated withSSRIs (19.21 (95% CI: (6.77–54.52)).

In the second model, we corrected for psychiatrichospital contact (Table 4, lower part). Those treatedwith SSRIs had still an elevated rate of suicide com-pared to those not treated with SSRIs. However, therate ratio had now decreased to 4.47 and it was nolonger quite significant (95% CI: 0.95–20.96).

Discussion

Part 1: Suicide is extremely rare in children youngerthan 14 years. The suicide rate increases with ageamong adolescents, and is higher among boys thanamong girls. For many years [21] and also during thepresent study period (Table 1) there have been onlyminor fluctuations in the Danish youth suicide rate,but the use of antidepressants in children and ado-lescents has steadily increased. There is no basis forthe assertion that other influencing factors on youthsuicide have changed during the follow-up period. Ifthe SSRIs are a precipitating factor in youth suicide,we would expect an increasing youth suicide rate withcalendar time, however, this was not demonstrated.The indication for antidepressant treatment has beenexpanded over time, presumably including personswith less severe illness. In the comparison in Part 1this extended indication might conceal an associationbetween SSRIs and suicide on the individual level,which was further explored in Parts 2 and 3 of thestudy.

Table 4 Rate Ratio’s (RR) Risk of suicide

Poisson regression Covariates RR (95% CI)

First modela Age group at suicide10–17 0.20 (0.07–0.57)18–22 1.00

Sexfemale 0.29 (0.10–0.82)male 1.00

TreatmentSSRIs 19.21 (6.77–54.52)No SSRIs 1.00

Second modelb TreatmentSSRIs 4.47 (0.95–20.96)No SSRIs 1.00

Psychiatric contactc

Contact 8.44 (1.97–36.18)No contact 1.00

aIncluding suicide as outcome and age, sex and treatment as covariatesbIncluding suicide as outcome and age, sex, treatment and psychiatric contactas covariatescDefined as psychiatric hospital contact

236 Eur Child Adolesc Psychiatry, (2006) Vol. 15, No. 4ª Steinkopff Verlag 2006

Page 6: Do antidepressants precipitate youth suicide?

We found only citalopram and sertraline to beused increasingly in children and adolescents. Forother SSRIs there were only minor changes in theprescription pattern from 1995 to 1999 (Fig. 1). Theuse of older antidepressants has been slightlydecreasing and the use of newer non-SSRI antide-pressants has been slightly increasing during thestudy period. However, these drugs were relativelyseldom prescribed to children and adolescents, asSSRIs were the main prescribed antidepressants. Thechoice of a certain SSRI might be influenced bymarketing mechanisms specific to the Danish marketin as much as no comparative study has shown clearsuperiority of any SSRI in treating children andadolescents.

The prescribing rate of antidepressants was foundto be increasing with age. This seems to reflect anincreasing prevalence of depression with increasingpost-pubertal age. Further, the youngest might havebeen treated with antidepressants to a higher degreeon indications other than depression [22].

In the present study, we have no direct knowledgeof the indication for those 2,569 children and ado-lescents treated with antidepressants, but from an-other Danish cross-sectional survey study coveringthe same time period it was found that the indicationfor antidepressant treatment in children and adoles-cents was depression in 55% of the girls and in 42%of the boys who were treated [22]. Additionally,depression was a more frequent indication amongadolescents than among children [22].

Part 2: In the present study, we have no knowledgeof free medications either despatched from psychiat-ric ambulatories or during hospitalizations. Due tothis, we extrapolated an ongoing treatment to becontinued during admittance to hospital. We cannotascertain in which dose a purchased antidepressantwas prescribed if the prescription was not regularlyrenewed; accordingly, we assumed that the taken dailydose equals the defined daily dose purchased in orderto estimate a maximum duration of treatment. De-spite these maximum estimates of treatment duration,only two persons were found, who had been treatedwith SSRIs within a period of two weeks before sui-cide during the five years of observation. Both per-sons started SSRI treatment before 18 years of ageand were older than 18 years at the time of death.Thus, we cannot exclude a possible association be-tween treatment and suicide for two persons; how-ever, other factors might be apparent. First, severediagnoses of mental illness such as organic psychosesor schizophrenia were confirmed from the DanishPsychiatric Central Research Register, among four ofthe five persons treated with antidepressants and whocommitted suicide. Although youth suicide is knownto be associated with depression, none of those trea-

ted with antidepressants and who committed suicidehad been admitted to hospital with a diagnosis ofdepression. Even though misdiagnosis is a possibility,it is less likely as the diagnoses have been made atrepeated psychiatric hospital admittances.

Second, we found a short interval between dis-charge from admittance to psychiatric hospital andsuicide, thus suggesting that suicide may have been aconsequence of an acute exacerbation of such a dis-integrating primary mental illness.

Third, the average period from first purchase ofantidepressants to death was rather long.

Final, among those five persons who purchasedantidepressants and who committed suicide, we foundthat all had purchased numerous drugs from otherpharmacological classes during the same periods asthey had purchased SSRIs. This sheds doubt on theputative suicidal effect of a certain drug.

Part 3: In Denmark, the National Board of Healthin 2000 issued guidelines recommended SSRIs to beinitially prescribed to children and adolescents, onlywhen the indication for SSRI treatment is made by achild and adolescent psychiatrist. Data used in thepresent study (1995–1999) precedes this precept, anda proportion of children and adolescents treated withantidepressants might have been treated by theirgeneral practitioner. Nevertheless, we found that57.4% of those treated with SSRIs and 1.1% of thecontrol persons had had psychiatric hospital contactin any period prior to the first purchase of antide-pressants. Further, private practicing specialists inchild and adolescent psychiatry may have treatedsome patients. These results indicate that antide-pressant treatment is related to a high degree ofmedical specialization, and further that those childrenand adolescents who are treated have a severe burdenof disease.

Previous psychiatric hospital contact can be re-garded as a proximate variable of the severity of apsychiatric disorder. Previous psychiatric hospitalcontact per se was found to predict a highly increasedsuicide risk (RR = 8.44, 95% CI: 1.97–36.18). As canbe seen from Table 4, the effect of treatment withSSRIs versus no such treatment on the rate of suicidewas not quite significant when the analysis was ad-justed for the effect of previous psychiatric hospitalcontact. Nevertheless, the rate ratio of 4.47 (0.95–20.96) was still over 1 and compatible with a stronglyincreased rate ratio up to 20.96 as indicated by thewide confidence interval. This raises the possibilitythat inclusion of an even larger population of controlsand patients might have resulted in a significantassociation between SSRIs and suicide. However, evenif the association was found statistically significantsuch an effect might be a consequence of the psy-chiatric illness itself, i.e., confounding by indication.

L. Søndergard et al. 237Antidepressants and youth suicide

Page 7: Do antidepressants precipitate youth suicide?

j Interpretation in relation to previous research

So far no other nationwide study has systematicallyinvestigated the association between treatment withSSRIs and completed suicide. In a nationwide Swedishregister study analysing all the forensic toxicologicalinvestigations of suicides, SSRIs had not been de-tected among any of 52 children under 15 years andamong adolescents aged 15–19 years, treatment withSSRIs was associated with a lower RR compared withother antidepressants [23]. Concordantly, we were notable to identify an association between treatment withantidepressants and completed suicide among chil-dren and adolescents.

Other studies have investigated the associationbetween treatment with SSRIs and suicidal ideationsor attempts among youth [7, 12]. Our finding is con-sistent with the findings in a recent study investigatingthe effect of combined treatment with fluoxetine andcognitive behavioural therapy among adolescents withmajor depressive disorder. Treatment with fluoxetineand cognitive behavioural therapy was associated withdecreased suicidal ideations [24].

In a pharmacoepidemiological analysis of datafrom a US prescription database, an inverse rela-tionship between regional change in use of SSRIs andsuicide in different ZIP code regions was found on theaggregate level and a protective effect of SSRIs onyouth suicide was suggested [25]. In the presentstudy, we combined ecological data (Part 1) andindividualized data (Parts 2 and 3). In a recent casecontrol study, treatment with SSRIs was found to beassociated with an increased risk of suicide amongadolescents [26]. As this study by Olfson was limitedto adolescents recently discharged from hospital witha diagnosis of depression it is possible that hetero-geneity exists, i.e., it is possible that SSRIs may inducesuicidal behaviour confined to a relatively small groupof severely ill patients and that in the aggregate levelantidepressants may have a protective effect on sui-cide as found in the previous study by Olfson andcolleagues [25].

From some randomized controlled studies, it hasbeen suggested that suicidal ideations and attemptedsuicide are precipitated in children and adolescentswhen treated with SSRIs. Such a putative increase insuicidal ideations and attempted suicide might notnecessarily be reflected in a proportionate increase insuicide rate even the evidence of a link betweencompleted suicide and attempted suicide in the younghas been found to be strong [27]. Nevertheless, amongadolescents, predictors for completed suicide differsubstantially from predictors for attempted suicide.Adolescents who complete suicide are more oftenboys [27] and the strongest risk factor for completed

suicide is mental disorders (in particular affectivedisorders, substance use disorders and antisocialbehaviours) [28]. Further, girls who commit suicideare more likely than males to have had an affectivedisorder [29]. In contrast, children and adolescentswho attempt suicide are more often girls and lessoften present with a mental disorder [30].

Selective serotonin reuptake inhibitors are pre-scribed to children and adolescents on several psy-chiatric indications other than youth depression. Ithas been argued that SSRIs promote suicide onlyamong those with a diagnosis of depression [5], i.e.,those who have pre-existing or latent suicidal idea-tions that can be worsened or triggered. Further, aFDA analysis found treatment of non-depressivechildren and adolescents with SSRIs also to be asso-ciated with a slightly higher risk of attempted suicide[12]. In such analyses, it cannot be excluded that thesuicidal ideations among children and adolescentswith depression might reflect confounding by indi-cation, i.e., those treated tend to have more seriousconditions.

Previous studies have found rates of depressionamong adolescents who committed suicide rangingfrom 35 to 76% [31]. Further, as low as 3% of teenagerswith major depressive disorder have been found to bein treatment with antidepressants at the time of sui-cide [32]. Strikingly, in the present study, none of the37 boys who committed suicide in Denmark when 10–17 years old in the time period 1995–1999 was treatedwith antidepressants. Boys have been found to denyfeelings of sadness and the only symptoms ofdepression might be irritability, boredom, social oreducational troubles [2]. Among the 14 control per-sons who committed suicide during follow-up, all wereboys and none had purchased any other medication.This indicates male youth health problems to be gen-erally widely neglected. Nevertheless, we cannot ex-clude the possibility, that a proportion of these boyswere not particularly mentally disturbed. Further,suicide among boys might involve a larger proportionof other diagnostic categories such as substance abuse.

Overall, a majority of those treated with antide-pressants were girls, and to some extent this reflectsdepression as being more prevalent in girls.

j Limitations

Youth suicide is a rare event and the question whetherantidepressant treatment precipitates youth suicide ornot qualifies for different approaches in elucidatingthe question.

As previously mentioned, we have no knowledge ofmedication given under admittance to hospital. It is

238 Eur Child Adolesc Psychiatry, (2006) Vol. 15, No. 4ª Steinkopff Verlag 2006

Page 8: Do antidepressants precipitate youth suicide?

unlikely that any of those in the study sample whocommitted suicide were taking antidepressants initi-ated and provided during hospitalization, but notrefilled during outpatient treatment as none werediagnosed either with depression or anxiety at dis-charge from hospital.

It is possible that some children and adolescentsactually committed suicide but their death was diag-nosed as an accidental death. However, as we wantedto include as valid data as possible, we did not includeinformation on accidental death as outcome. If suchinformation were included the risk of differentialmisclassification between deaths by suicide and deathby other reasons including somatic illness would re-duce the possibility of detecting a putative associationbetween SSRI and suicide. As the aim of the study wasto reveal an association between SSRIs and suicide, ifit exists in reality, we chose to include only a diag-nosis of suicide as outcome.

We only know psychiatric diagnoses for those whohad received psychiatric hospital service and there-fore the indication for antidepressant treatment areunknown for the greatest proportion of the persons inthe study.

It was not possible to perform statistical analysis(Part 3) on the association between other antide-pressants than SSRIs and suicide as no children andadolescents treated with older antidepressants, newernon-SSRI antidepressants or lithium committed sui-cide during the study period. As only few childrenand adolescents were treated with antidepressantsother than SSRIs no conclusions can be drawn fromthis on the risk of completed suicide.

In Part 3, we were not able to identify a definitelyincreased risk of suicide among children and ado-lescents treated with SSRIs. Comparing the risk ofsuicide among children and adolescents with thegeneral population involves the problem of con-founding by indication and those two groups are notdirectly comparable. Using prior psychiatric hospitalcontact to minimize the effect of confounding byindication can only be regarded as a rough proxyvariable.

If the study population had been larger, an in-creased risk of suicide among those treated with an-tidepressants compared to those not treated mighthave been significant. Repeating the present studyafter some years might strengthen the results. Even ifsuch a putatively significant increased risk was found

the problem of confounding by indication still per-sists, i.e., those treated may have a higher risk ofsuicide related to their mental disease than the generalpopulation. Further, it is possible that heterogeneityoccurs. Subgroups of children and adolescents mighthave increased risk of suicide when antidepressanttreatment is initiated, and other subgroups mightbenefit from treatment resulting in a decreased risk ofsuicide. Thus, a true increased effect of antidepres-sants might outnumber a true decreased effect ofantidepressants resulting in a non-significant associ-ation between antidepressants and suicide.

In most clinical trials, suicidal behaviour is anexclusion criterion and such studies are not useful inassessing risk of suicide. The present study is notlimited by such exclusions.

j Clinical implications

When deciding to treat a child or an adolescent with aSSRI the effect of treatment must be considered inrelation to potential side effects of the treatment. Theevidence of efficacy in response to SSRIs amongchildren and adolescents is not as documented asamong adults and with contradictory findings [7].Nevertheless, five antidepressants have been found tobe significantly more effective than placebo in at leastone trial each: fluoxetine, sertraline, paroxetine, cita-lopram and nefazodone [7]. Great vigilance must betaken when treating children and adolescents withdepressive disorder, as it is a life-threatening condi-tion. Therefore, the risk of completed suicide needs tobe assessed carefully in the treatment of youthdepression.

Analysing the association between treatment withSSRIs and completed suicide, we were not able toconfirm the casuistically reported risk of completedsuicide among children and adolescents treated withSSRIs. Based on findings from prior casuistically re-ports and the present study, not treating severelydepressed children and adolescents with SSRIs maybe inappropriate or even fatal.

j Acknowledgements The study was funded by The LundbeckFoundation and Slagtermester Max Wørzner og hustru IngerWørzners Mindelegat. The sponsors of the study had no role instudy design, data collection, data analyses, data interpretation orwriting the report.

L. Søndergard et al. 239Antidepressants and youth suicide

Page 9: Do antidepressants precipitate youth suicide?

References

1. Kovacs M (1996) Presentation andcourse of major depressive disorderduring childhood and later years of thelife span. J Am Acad Child AdolesPsychiat 35:705–715

2. Park RJ, Goodyer IM (2000) Clinicalguidelines for depressive disorders inchildhood and adolescence. Eur ChildAdoles Psychiat 9:147–161

3. Brent DA, Birmaher B (2004) Britishwarnings on SSRIs questioned. J AmAcad Child Adoles Psychiat 43:379–380

4. Garland EJ (2004) Facing the evidence:antidepressant treatment in childrenand adolescents. CMAJ 170:489–491

5. Whittington CJ, Kendall T, Fonagy PP,Cottrell PD, Cotgrove A, Boddington E(2004) Selective serotonin reuptakeinhibitors in childhood depression:systematic review of published versusunpublished data. Lancet 363:1341–1345

6. Committee on safety of medicines(2004) Selective Serotonin ReuptakeInhibitors (SSRIs): Overview of regu-latory status and CSM advice relatingto major depressive disorder (MDD) inchildren and adolescents including asummary of available safety and effi-cacy data. Available at: http://medi-cines.mhra.gov.uk/ourwork/monitor-safequalmed/safetymessages/ssriover-view_101203.htm

7. American College of Neuropsycho-pharmacology (2004) Executive sum-mary preliminary report of the taskforce on SSRIs and suicidal behavior inyouth, January 21, 2004 (2004) Avail-able at http://www.acnp.org/exec_sum-mary.pdf

8. Beasley CM, Jr., Dornseif BE, Bosom-worth JC, Sayler ME, Rampey AH, Jr.,Heiligenstein JH, Thompson VL, Mur-phy DJ, Masica DN (1991) Fluoxetineand suicide: a meta-analysis of con-trolled trials of treatment for depres-sion. BMJ 303:685–692

9. Tollefson GD, Tollefson SL, Sayler ME,Luxenberg MG (1994) Absence ofemergent suicidal ideation duringtreatment: a comparative, controlled,double-blind analysis employing sev-eral distinct antidepressants. Depres-sion 2:2–79

10. Duff G (2003) Selective serotonin re-uptake inhibitors: use in children andadolescents with major depressive dis-order. http://medicines.mhra.gov.uk/ourwork/monitorsafequalmed/safety-messages/cemssri_101203.pdf

11. FDA Public Health Advisory (2003)Reports of suicidality in pediatric pa-tients being treated with antidepressantmedications for major depressive dis-order (MDD)[Food and Drug Admin-istration Web site]. Available at http://www.fda.gov/cder/drug/advisory/mdd.htm. 27–10–2003.

12. CDER 2004 Meeting Documents (2004)September 13–14 Joint Meeting withthe Pediatric Advisory Committee[Food and Drug Administration Website]. Available at http://www.fda.gov/ohrms/dockets/ac/cder04.html

13. Jick SS, Dean AD, Jick H (1995) An-tidepressants and suicide. BMJ310:215–218

14. Medicinal Product Statistics 1998–2002Denmark (2003) The Danish Medi-cines. Agencym Copenhagen

15. Sørensen HT, Larsen BO (1994) Apopulation-based Danish data resourcewith possible high validity in pharma-coepidemiological research. J Med Syst18:33–38

16. Juel K, Helweg-Larsen K (1999) TheDanish registers of causes of death.Dan Med Bull 46:354–357

17. Munk-Jorgensen P, Mortensen PB(1997) The Danish Psychiatric CentralRegister. Dan Med Bull 44:82–84

18. Malig C (1996) The civil registrationsystem in Denmark: II VRS technicalpaper, no 66. International Institute forVital Registration and Statistics

19. Clayton D, Hills M (2005) StatisticalModels in Epidemiology. Oxford Uni-versity Press, Oxford

20. WHO Collaborating Centre for DrugStatistics Methodolgy (2003) ATC in-dex with DDDs WHO CollaboratingCentre for Drug Statistics MethodologyOslo

21. Nordentoft M, Bille-Brahe U, Morten-sen PB, Runge K, Frederiksen K, In-gerslev O (1998) Bilagsdel til Forslag tilhandlingsplan til forebyggelse af sel-vmordsforsøg og selvmord i Danmark.Sundhedsstyrelsen, Copenhagen

22. Buhl SC, Bohm JE, Thomsen PH,Dalsgaard S (2003) Indications for anduse of antidepressants in child andadolescent psychiatry—a cross-sec-tional survey in Denmark. Eur ChildAdolesc Psychiat 12:114–121

23. Isacsson G, Holmgren P, Ahlner J(2005) Selective serotonin reuptakeinhibitor antidepressants and the riskof suicide: a controlled forensic data-base study of 14,857 suicides. ActaPsychiatr Scand 111:286–290

24. March J, Silva S, Petrycki S, Curry J,Wells K, Fairbank J, Burns B, DominoM, McNulty S, Vitiello B, Severe J(2004) Fluoxetine, cognitive-behavioraltherapy, and their combination foradolescents with depression: Treatmentfor Adolescents With Depression Study(TADS) randomized controlled trial.JAMA 292:807–820

25. Olfson M, Shaffer D, Marcus SC,Greenberg T (2003) Relationship be-tween antidepressant medicationtreatment and suicide in adolescents.Arch Gen Psychiat 60:978–982

26. Olfson M (2005) Antidepressants andYouth Suicide: A Pharmacoepidemiol-ogists Perspective. New Clinical DrugEvaluation Unit Meeting. Boca Raton,Florida July 2, 2005

27. Beautrais AL (2003) Suicide and seri-ous suicide attempts in youth: a mul-tiple-group comparison study. Am JPsychiat 160:1093–1099

28. Beautrais AL (2000) Risk factors forsuicide and attempted suicide amongyoung people Aust New Zeal. J Psychiat34:420–436

29. Brent DA (1995) Risk factors for ado-lescent suicide and suicidal behavior:mental and substance abuse disorders,family environmental factors, and lifestress. Suicide Life-threat Behavior25(Suppl):52–63

30. Lewinsohn PM, Rohde P, Seeley JR,Baldwin CL (2001) Gender differencesin suicide attempts from adolescence toyoung adulthood. J Am Acad ChildAdoles Psychiat 40:427–434

31. Lecomte D, Fornes P (1998) Suicideamong youth and young adults, 15through 24 years of age. A report of 392cases from Paris, 1989–1996 J. For Sci43:964–968

32. Shaffer D, Craft L (1999) Methods ofadolescent suicide prevention. J ClinPsychiat 60(Suppl 2):70–74

240 Eur Child Adolesc Psychiatry, (2006) Vol. 15, No. 4ª Steinkopff Verlag 2006