Suicide-related events among child and adolescent patients during short-term antidepressant therapy


Teizo Kuba, MD, Department of Neuropsychiatry, Faculty of Medicine, University of the Ryukyus, 207 Uehara, Nishihara, Okinawa 903-0215, Japan. Email:


Aims:  Antidepressants have been of limited use for adolescent subjects with depression because of drug-induced suicide-related events (SRE). Therefore, we investigated actual suicidality and its risk factors during antidepressant therapy among child and adolescent patients in clinical settings.

Methods:  The risks of SRE, consisting of suicidal ideation, self-mutilation and suicide attempt, were prospectively monitored among 70 child and adolescent patients (15.4 ± 2.8 years) during the first 3 months of antidepressant therapy.

Results:  The proportion of SRE decreased from 47.1% to 22.9% after the treatment. Subjects with persistent risks of SRE were significantly characterized by female sex (P < 0.05), psychotic features (P < 0.001), borderline personality disorder (P < 0.01), previous SRE (P < 0.001), and such baseline psychopathology as anhedonia (P < 0.005), irritability (P < 0.005) and hopelessness (P < 0.001). Discriminant analysis showed that baseline severity of SRE, borderline personality disorder and psychotic features were closely associated with SRE during antidepressant therapy. Total scoring using those three pretreatment factors predicted risks of SRE with sufficient sensitivity (81%)/specificity (98%) as well as high positive likelihood ratio (43.9).

Conclusions:  These findings suggest that deteriorated risk of SRE in child and adolescent patients receiving antidepressants should not be overestimated while some pretreatment characteristics may be useful to predict the outcome of SRE after antidepressant therapy.

BECAUSE OF ESTABLISHED clinical efficacy and high tolerability, newer antidepressants, including selective serotonin reuptake inhibitors (SSRI) and serotonin noradrenaline reuptake inhibitors (SNRI), have been widely used in the treatment of such psychiatric diseases as depression and several anxiety disorders, which may be occasionally accompanied by suicide-related events (SRE), especially in severe cases. Overall, the suicide rate has gradually decreased due to increased prescription of SSRI and other second-generation antidepressants in the USA,1 and a similar tendency has been observed in Japan.2 Ludwig and Marcotte3 have found that an increase in SSRI doses is negatively correlated with suicide rates.

About 2% of school-aged children and 8% of adolescents appear to experience depression at any one time,4 and some of them may need pharmacological intervention. In 2003, however, the Food and Drug Administration (FDA [USA]) and the Medicines and Healthcare Products Regulatory Agency (MHRA [UK]) have warned of increased risks of SRE among the younger generation receiving antidepressants. Since then, the Ministry of Health, Labor and Welfare in Japan has also taken similar measures. Some meta-analyses have indicated that increased suicidal risk outweighs the therapeutic benefits of treating child and adolescent patients with antidepressants.5–7

According to a large cohort study, antidepressants are generally associated with an increased risk of attempted suicide, irrespective of the types of drugs, while the risk of completed suicides and total mortality are rather reduced by antidepressant therapy.8 Especially focusing on these problems in children and adolescents, it has been reported that these subjects are at higher risk of SRE than adults after starting antidepressant medication, although the risks are relatively rare, that is, one in 3000 treatment for completed suicide and one in 1000 for serious suicide attempt during acute-phase antidepressant therapy.9 Meta-analyses of placebo-controlled studies also indicate that antidepressants may cause a significant but small and short-term risk of self-harm or SRE in child and adolescent patients with major depressive disorder, although no completed suicides were observed in any trials of their meta-analysis.10,11

After the temporary contraindication of SSRI and/or the following special warning for antidepressant use in younger patients, an increasing concern of antidepressant-induced SRE deters clinicians from prescribing antidepressants to child and adolescent patients who may truly need medication for symptom alleviation. Apter et al.12 suggested that such specific factors as diagnosis, sex, age, previous suicidal behaviors and baseline suicidal ideation could affect future occurrence of SRE in child and adolescent patients, and that paroxetine simply amplified these risks by their placebo-controlled comparative study. Therefore, it is essential to determine the backgrounds and characteristics of younger patients at high risk of SRE during antidepressant therapy together with evidence-based risk/benefit assessments. It is also noted that newly developed or increased risks of SRE associated with antidepressant treatments should be distinguished from persistent risks of SRE before and after the treatment, which tend to be easily confounded in clinical situations.

Therefore, we aimed to examine how antidepressants actually affect the severity of suicidality in child and adolescent patients during antidepressant treatment. As suggested by the report by Simon et al.,9 the highest risks of SRE are concentrated on the period during the first few months after initiating antidepressant treatments. The risk of suicidal behavior is especially increased in the first month after starting antidepressants.13,14 Thus, in the present study, the risks of SRE, consisting of suicidal ideation, self-mutilation and suicide attempt, were prospectively monitored in child and adolescent patients during the first 3 months of antidepressant therapy.



The subjects were 70 child and adolescent outpatients (32 boys and 38 girls), who received antidepressant therapy for at least 3 months in our clinic, from April 2004 to February 2009. Their age ranged from 6 to 19, with 24 subjects under 15 years of age and 46 subjects aged 15 or more (mean ± SD: 15.4 ± 2.8). Primary clinical diagnosis according to the DSM-IV-TR consisted of major depressive disorders (MDD: n = 27), adjustment disorder (n = 10), borderline personality disorder (BPD: n = 6), obsessive–compulsive disorder (OCD: n = 5), schizophrenia (n = 5), dissociative disorder (n = 4), social anxiety disorder (n = 3), schizoaffective disorder (n = 2), dysthymic disorder (n = 2), pervasive developmental disorder (PDD: n = 2), trichotillomania (n = 2), post-traumatic stress disorder (n = 1) and panic disorder (n = 1). Twenty-five subjects also had a comorbid psychiatric diagnosis, such as BPD (n = 4), mental retardation (n = 4), PDD (n = 4), MDD (n = 3), dysthymic disorder (n = 2), eating disorder (n = 2), epilepsy (n = 2), panic disorder (n = 1), trichotillomania (n = 1), attention deficit/hyperactivity disorder (n = 1) and steroid psychosis (n = 1).


The baseline demographics and clinical characteristics of participants (sex, age, major depressive episode, psychotic features, BPD, conflict with family, social maladjustment, psychopathology [such as anhedonia, irritability and hopelessness], and baseline suicidality) were recorded before starting antidepressants. The risk of SRE, consisting of suicidal ideation, self-mutilation and suicide attempt, were especially focused on and were prospectively monitored during the first 3 months of antidepressant therapy. Finally, subjects were divided into two groups: 16 subjects with persistent risks of SRE (SRE (+) group) and 54 subjects safely treated without risks of SRE (SRE (−) group) during the 3-month treatment.

This study was conducted as an open trial with flexible dose schedule design under usual conditions of general clinical practice. Thus, the type of antidepressant was selected according to severity and psychopathology associated with depression/anxiety symptoms, where daily dose was flexibly modified to obtain optimal efficacy and avoid unwanted effects in each patient. Antidepressant drug doses were expressed as imipramine equivalence, using equivalent conversion of psychotropic drugs reported by Inagaki et al.15 Mood stabilizers, antipsychotics and benzodiazepine anxiolytics were concomitantly administered to necessary cases. All the patients received non-specific supportive psychotherapy weekly or every 2 weeks during the study period. Treatment responders were defined as subjects achieving Clinical Global Impression of Improvement scale scores of 1 or 2 at the end of the 3-month treatment. Adherence to medication was confirmed by patients' family and clinicians.

In principle, the study protocol was naturalistic in design, and all the data were obtained from subjects and their family. All of the results were recorded as anonymous and grouped data after encoded or numbered, adhering to privacy policy to protect personal information. Written informed consent was obtained from each patient and their parents. This study was approved by the Ethics Committee of the University of the Ryukyus.

Statistical analyses

The χ2-test, Fisher's exact test and the Kruskal–Wallis test were used for comparison of subjects' backgrounds and treatment profiles between SRE (+) and SRE (−) groups. A comparison of antidepressant dose was made with the Mann–Whitney U-test. Prediction for persistent risks of SRE using total scoring of baseline risk factors was assessed with Fisher's exact test. The two-tailed test was used for all of these analyses. Discriminant analysis was conducted to assess contribution of independent variables (sex, age, psychotic features, BPD, familial maladjustment, social maladjustment, psychopathology [such as anhedonia, irritability and hopelessness] and baseline SRE) to the persistent risks of SRE during antidepressant treatments as the dependent variable. A P-value of less than 0.05 was regarded as significant. spss 11.0 for Windows (spss Japan Inc., Tokyo) was used for these statistical analyses.


At baseline, 33 (47.1%) out of 70 patients were at risk of SRE. Among these subjects with high risk of SRE, 72.7% (24/33) exhibited self-destructive behaviors (self-mutilation or suicidal attempt). After the 3-month treatment, the proportion of SRE decreased from 47.1% to 22.9% while two cases (3%) exhibited a worsened grade of SRE, probably not due to the adverse effects of antidepressants but based on the patient's own vulnerability in psychopathology and their limited capacity for self-regulation. Three other cases (one boy and two girls) exhibited transient antidepressant-induced hypomania without suicidality (4.3%).

Pretreatment demographic and clinical characteristics were compared between the SRE (+) group (n = 16) and the SRE (−) group (n = 54) during 3-month treatment (Table 1). The SRE (+) group was significantly characterized by female sex, age, psychotic features, BPD, conflict with family, social maladjustment, baseline psychopathology (such as anhedonia, irritability and hopelessness) and previous SRE compared with the SRE (−) group (Table 1).

Table 1.  Pretreatment characteristics in child and adolescent patients with and without suicide-related events (SRE) during 3-month antidepressant therapy
SRE during treatmentPresent (n = 16)Absent (n = 54)P
  • Suicidality plus included suicide attempt, deliberate self-injury and suicide ideation.

  • NS, not significant.

Age (≧15 years)87.5%59.3%<0.05
Major depressive disorder56.3%38.9%NS
Psychotic features37.5%3.7%<0.01
Borderline personality disorder56.3%1.9%<0.01
Conflict with family56.3%25.9%<0.05
Social maladjustment87.5%59.3%<0.05
Baseline psychopathology   
Baseline suicidality   
 Suicidality plus100%31.5%<0.001

As a result, the SRE (+) group needed significantly larger doses of antidepressants and more concomitant use of mood stabilizers and/or antipsychotics. However, their treatment response was much poorer compared with those without risks of SRE (Table 2). No significant difference was found in the variety of types of antidepressants used in the 3-month treatment between the two groups (Table 2).

Table 2.  Treatment profiles in child and adolescent patients with and without suicide-related events (SRE) during 3 months
SRE during treatmentPresent (n = 16)Absent (n = 54)P
  • Maximal doses during 3-month treatment as imipramine equivalence.

  • NS, not significant; SNRI, serotonin noradrenaline reuptake inhibitors; SSRI, selective serotonin reuptake inhibitors.

 Dose (mg/day)146.3 ± 72.285.0 ± 50.9<0.001
 SSRI13/16 (81.3%)48/54 (88.9%)NS
 SNRI0/16 (0%)2/54 (3.7%)NS
 Tetracyclic5/16 (31.3%)6/54 (11.1%)NS
 Tricyclic3/16 (12.5%)5/54 (9.3%)NS
Concomitant use   
 Mood stabilizers9/16 (56.3%)7/54 (13.7%)<0.001
 Antipsychotics8/16 (50.0%)8/54 (14.8%)<0.01
 Anxiolytics6/16 (37.5%)18/54 (33.3%)NS
Treatment responder2/16 (12.5%)32/54 (59.3%)=0.001

In association with the risks of SRE during treatment, significant risk factors as baseline characteristics (sex, age, psychotic features, BPD, conflict with family, social maladjustment, anhedonia, irritability, hopelessness and baseline suicidality) were enrolled in discriminant analysis. Among these factors, psychotic features, BPD and baseline SRE had strong correlations with SRE after treatment (Table 3). By using these three factors, risk assessment for SRE during antidepressant therapy was conducted. One point was allotted to each factor, yielding a total score of 3 as a perfect score. By setting the cut-off point at 2 (≧2 as risks of persistent SRE), this scoring system predicts the persistent risks of SRE (Fig. 1) with sufficient sensitivity (81%)/specificity (98%) and high positive likelihood ratio (43.9).

Table 3.  Standardized coefficients of discriminant analysis among contributing factors to suicide-related events
  1. Canonical correlation was 0.821.

  2. Wilks's lambda was 0.327, and its significance was P < 0.001.

Psychotic features0.595
Borderline personality disorder0.772
Conflict with family0.026
Social maladjustment−0.080
Pre-treatment suicide-related events0.575
Figure 1.

Total scores of baseline risk factors as a predictor for suicide-related events (SRE) during antidepressant therapy in 70 patients. One point was allotted to each item of psychotic features, borderline personality disorder or baseline suicidality, yielding the total score of 3 as a perfect score. Provided that the cut-off value was 2-points, 13 out of 16 subjects with persistent SRE (SRE (+) group) scored 2 or more while 53 out of 54 subjects without SRE (SRE (−) group) scored 1 or less. The sensitivity, specificity and positive likelihood ratio were 81%, 98% and 43.9, respectively.


In general, child and adolescent subjects are still in the process of developing psychological maturity. They tend to have narrow scope cognition, dichotomous thinking, low tolerance to stress and anxiety, lower self-esteem, emotional instability/dysregulation, tendency to escape from stressful reality and low threshold of acting out, which may easily lead to self-destructive behaviors in depressed/anxious youths.16,17 Indeed, a high proportion of our subjects showed self-destructive behaviors before treatment, indicating the necessity of immediate intervention for crisis by both pharmacological and non-pharmacological measures.

Although it has been reported that antidepressant therapy may increase suicide-related risks to some extent in the younger population than in adults,9–11,18 these small risks do not necessarily warrant a great disadvantage for antidepressant use in depressed/anxious youths. Bridge et al.19 conducted a meta-analysis of randomized controlled studies to assess the therapeutic efficacy and suicide-related risk of antidepressants in the treatment of pediatric subjects with MDD, OCD and non-OCD anxiety disorders. In pediatric subjects, 11.0% of MDD, 19.8% of OCD and 37.8% of non-OCD anxiety disorder patients favored antidepressants over placebo, while differences in increased suicidal risk (antidepressants vs placebo) were only 0.9% of MDD, 0.5% for OCD and 0.7% for non-OCD anxiety disorder subjects.19 They suggested that the benefits of antidepressants for pediatric patients appear to be much greater than the risks from suicidal ideation/suicide attempt irrespective of psychiatric diagnosis.

In the present study, more than half of our subjects with a high risk of baseline SRE actually benefited from acute treatments, including antidepressants without any unequivocal drug-related deterioration of suicidality. Furthermore, it has been shown that continuation of treatment with SSRI is clearly superior to placebo in preventing relapse and increasing the remittance period in child and adolescent patients with major depression.20,21

These results suggest that worsened risks of SRE in child and adolescent patients receiving antidepressants should not be overestimated in usual clinical settings except for youths showing hypersensitive reaction to antidepressants. However, it is noted that female sex, age of 15 years or more, subjects who have psychotic features, BPD, specific psychopathology (such as anhedonia, irritability and hopelessness) and baseline suicidality were strongly associated with future occurrence of SRE in the present study (Table 1). Several reports have indicated that a history of multiple suicide attempts is the most influential risk factor for later suicidality with intense crisis.22,23 In addition, severity of self-rated and persistent depressive symptoms predicted suicidal ideation emergence of suicidality during treatment.14 Therefore, our data on baseline suicidality as a primary predictor for persistent risks of SRE is completely consistent with these previous findings.

Suicidal ideation/behaviors and repeated self-mutilation are criteria of BPD according to the DSM-IV and ICD-10.24,25 Cluster B personality disorders in connection with impulsivity and aggressive behavior have also been pointed out as strong predictors for completed suicide in male adults with major depression.17,26 Thus, it is not surprising that BPD as a risk factor of SRE is reproduced in the present study. A comprehensive review on personality traits as correlates of SRE has pointed out that hopelessness is consistently regarded as a useful manifestation in screening for risk of suicidal behaviors.27 Other previous reports have suggested that high inward irritability and anhedonia are significantly associated with suicide-related outcomes in youth.28,29 Therefore, these specific psychopathologies should be focused on when differentiating subjects with persistent risks of SRE from those without them.

Baseline psychotic feature was a moderate distinguisher in assessing persistent risks of SRE in the present study. This finding may be supported by previous reports suggesting psychoticism as a temperamental vulnerability in attempted suicide and dramatically increased suicidal attempts in psychotic depression as compared with non-psychotic depression.30–32 Although current major depressive episode did not have any significant effects on SRE, therapeutic intervention, including antidepressant therapy, might have masked the risks of SRE in this study. Nevertheless, it is generally regarded that child and adolescent patients with major depression are at relatively higher risk for SRE even during antidepressant therapy than those with anxiety disorder.12,14 There was a trend in the present study that subjects with persistent risks of SRE were more affected by environmental factors as conflict with family and social maladjustment. Psychological distress from these factors can be a latent promoter for suicide-related acting out in the younger generation.17,33 Therefore, these less serious risk factors should be also taken into consideration as having supplementary effects on correct prediction for the risk of suicidality.

Multivariate analysis demonstrated that factors that were likely to predict the persistent risks of SRE were psychotic features, BPD and baseline suicidality. In a clinical setting, physicians may often be reluctant to administer antidepressants to patients with these three factors at baseline. As shown in the present study, these three items may be helpful for sensitive and specific prediction for the persistent risks of SRE in child and adolescent patients, even after antidepressant therapy. Thus, our scoring using three items can be conveniently applied to child and adolescent patients as a routine baseline assessment to differentiate high-risk subjects from others, although continuous and cautious monitoring of suicidality is still necessary in any child and adolescent patients throughout the period of acute treatments.

Meanwhile, treatment outcome of high-risk patients for SRE was characterized by larger doses of antidepressants and more concomitant use of mood stabilizers and/or antipsychotics, despite poorer treatment response (Table 2). These results suggest that intensive pharmacotherapy does not necessarily lead to success in the treatment of younger subjects at persistent risks of SRE, who may rather need more comprehensive treatments, including modest and combined pharmacotherapy, together with other non-pharmacological treatment, such as dialectical behavior therapy,34 adding cognitive behavior therapy,21 interpersonal psychotherapy for adolescents,35 family therapy and environmental rearrangement.

This study has some limitations due to a relatively small sample size, a wide variety of diagnoses and the naturalistically designed open trial. Because most of the subjects were treated with flexible dosage SSRI, the effects of the types and doses of drugs on the risks of suicidality under antidepressant therapy could not be extracted. In addition, severity of suicidality and specific psychopathology were roughly assessed through simple dichotomy (i.e. present or absent) without using established scales, even though the overall scoring of risk factors was of high predictive value with a strong positive likelihood ratio (43.9). Further studies with a larger number of subjects will be needed to confirm the present findings.

The present findings suggest that deteriorated risk of SRE among child and adolescent patients receiving antidepressants should not be overestimated, as treatments including antidepressant therapy had the advantage of reducing such risks in this small study. Patients with psychotic features, BPD and baseline suicidality were most likely to develop future SRE.