91 Effective treatments for suicidal youth: Pharmacological and psychosocial approaches

There are few empirically tested treatments for suicidal adolescents. In this chapter, the relationship between adolescent depression and suicidal behaviour is reviewed, as well as the risk factors for suicidal behaviour in depressed adolescents. The impact of psychosocial and pharmacotherapeutic interventions on depression and on suicidal outcomes is reviewed, including recent reports that SSRIs may be associated with an increase in spontaneously reported suicidal adverse events. We then review the few empirically tested psychosocial interventions for the treatment of suicide-attempting adolescents, including treatments that target family interactions, cognitions, social and emotion regulation skills, and the social ecology (either social network or school) of the suicidal adolescent. Studies have also examined the impact of different aspects of clinical management, such as the development of a safety plan, providing open access to an inpatient unit, and augmenting usual care with occasional postcards from clinical staff. The possible impact of lithium, divalproex, neuroleptics, and SSRIs in the reduction of impulsive aggression and suicidal risk is also discussed. Recommendations for further research are advanced.

This chapter reviews proven and promising treatments for suicidal youth, beginning with the management of depression, the single most common psychiatric disorder accompanying suicidal behaviour in children and adolescents. Subsequently, the interventions that are known to target suicidal behaviour are reviewed. On the basis of this review, recommendations about best practice for the treatment of suicidal youth and for further research are delineated.

Depression is the single most important psychiatric risk factor for child and adolescent suicidality. Over 80 per cent of adolescent suicide attempters, and 60 per cent of adolescent suicide completers have at least one major mood disorder (Brent et al. 1993; Kerfoot et al. 1996; Lewinsohn et al. 1996; Shaffer et al. 1996; Gould et al. 1998; Bridge et al. 2006). Since depression conveys a 10–50-fold increased risk for suicidal behaviour, it follows that improved identification and treatment of depression should be an important element in the reduction of suicidal risk. The risk for suicidal behaviour in depressed children and adolescents is particularly elevated in those whose depression is characterized by early onset, greater severity, and chronicity. Comorbid conditions, particularly substance abuse, conduct disorder, impulsive aggressive personality traits, as well a history of abuse, parent–child discord and criminality and a family history of suicidal behaviour, also contribute to an increased risk for suicidal behaviour (Lewinsohn et al. 1996; Brent 1997; Fergusson and Woodward 2002; Brent and Mann 2005). Because suicidality in depression is the product of the interaction between depression and personal family and social contextual factors, prevention of suicidal behaviour in depressed, suicidal individuals may require both treatment of depression as well as interventions that target these associated contextual factors.

Both antidepressants and specific types of psychotherapy (cognitive behaviour therapy and interpersonal therapy) have been shown to be efficacious in the treatment of adolescent depression (Lewinsohn et al. 1990; Wood et al. 1996; Brent et al. 1997; Clarke et al. 1999; Mufson et al. 1999; March et al. 2004; Mufson et al. 2004). In Chapter 92 individual psychotherapy techniques are described. Less is known about the direct impact of these treatments for depression on suicidal ideation and behaviour, because suicidal subjects are frequently excluded from clinical trials, and suicidality is often not assessed or reported as an outcome.

In one study, cognitive behaviour therapy (CBT) was found to be superior to both family and supportive therapies for the treatment of adolescent depression, yet all three treatments produced comparable reductions in suicidality (Brent et al. 1997) in patients with current or lifetime suicidality (attempt or ideation with a plan). CBT was markedly superior to supportive therapy for the treatment of depression (Barbe et al. 2004). Interpersonal therapy (IPT) was found to be superior to clinical management for adolescent depression, but there was no differential impact on suicidal ideation or attempts (Mufson et al. 1999). In the multisite Treatment of Adolescent Depression Study (TADS) in which fluoxetine, cognitive behaviour therapy (CBT), and the combination were compared to each other and to placebo, both fluoxetine and the combination of medication and CBT produced substantial improvements in depression relative to placebo (and to CBT alone), but only the combination treatment produced a more robust reduction in suicidal ideation compared to placebo (March et al. 2004). In the TADS trial, fluoxetine was associated with more suicidal events than placebo, and over the 9-month continuation and follow-up, participants who stayed in their original treatment showed a lower rate of suicidal events in the combined treatment versus fluoxetine alone (Emslie et al. 2006; The TADS Team 2007). However, two other large clinical trials with participants with greater severity of depression, longer duration, or higher levels of intake suicidal ideation did not find a protective effect for the combination of CBT plus medication against the occurrence of suicidal events (Goodyer et al. 2007; Brent et al. 2008).

One quality improvement study based in primary care compared the provision of specialty mental health care on-site (CBT, medication or both) versus usual care for adolescent depression (Asarnow et al. 2005). The quality improvement condition showed a greater reduction in depression, as well as a near-significant trend towards a reduction in the incidence of clinically significant suicidality.

Many of the predictors of poor response to either SSRIs or cognitive therapy are also associated with suicidal behaviour: chronicity, severity, comorbidity, high hopelessness, family discord, history of abuse, and parental depression (Clarke et al. 1992; Brent et al. 1998, Emslie et al. 1998, Lewinsohn et al. 1996, 1998; Rohde et al. 2001; Barbe et al. 2004). The TADS trial confirmed several of these observations, but found one result that was in the opposite direction of several of these studies: greater cognitive distortion was associated with a larger effect size for the combined treatment, and also CBT worked best in participants that came from homes with higher incomes (Curry et al. 2006). Interventions that address these co-occurring difficulties may improve treatment outcomes for depression and may also provide relief of suicidality. Pharmacogenetic studies hold promise for personalizing treatment. The more fun-ctional allele of the serotonin transporter promoter gene (long allele) has been shown to predict better response of depressed children and adolescents to citalopram (Kronenberg et al. 2007).

A meta-analysis including over 4300 child and adolescent subjects treated with antidepressants found that new-onset or worsening suicidality occurred about 1.8 times more often in the drug condition than in placebo, with an average of 4 per cent of those on medication developing suicidality versus 2 per cent on placebo. A later re-analysis of all paediatric clinical trials using second-generation antidepressants that included newer studies and used random- rather than fixed-effects assumptions in estimating risk differences found only a 0.9 per cent risk difference in the rate of suicidal events between antidepressant and placebo overall, and within the depressed subsample (Bridge et al. 2007). In contrast to the FDA report, this meta-analysis compared benefit to risk, and found that over 11 times more individuals were likely to benefit from an antidepressant for the treatment of depression than were likely to experience a suicidal event, which many would consider a favourable risk-to-benefit ratio. The risk to benefit ratio was also more favourable for adolescents versus children, except for trials of fluoxetine, in which the effects were similar. Effect size was inversely proportional, and the placebo response rate was positively correlated with the number of sites, suggesting that studies with a large number of sites may not screen out potential placebo repsonders and therefore may be underestimating the true effect of antidepressants on paediatric depression (Bridge et al. 2007). This is important because it means that the ratio of those who benefit from the drug versus those who become suicidal may underestimate the benefit to those with at least moderately severe depression.

The mechanism by which antidepressants may increase the risk for suicidal behaviour is not known, but possible explanations include increased irritability and hostility, akathisia, disinhibition, withdrawal symptoms induced by non-compliance, or development of a mixed state in those with a bipolar diathesis. The latter condition is much more likely to be induced by SSRI treatment in younger versus older patients (Martin et al. 2004). Data from both psychotherapy and pharmacotherapy trials show that self-reported ideation at intake is the single best predictor of a suicidal event in the clinical trial (Apter et al. 2006; March et al. 2006).

As a result of the concern raised about antidepressants and suicidality, the FDA is now requiring that clinical trials use a standard classification of suicidal events, the Columbia Classification Algorithm of Suicide Assessment (C-CASA), which has demonstrated reliability and validity (Posner et al. 2007). Consequently, using this system, the US Food and Drug Administration (2003) found an increase in risk for suicidal events in patients taking anti-epileptic drugs versus placebo (0.43 per cent versus 0.22 per cent; http://www.fda.gov/CDER/drug/infopage/antiepileptics/default.htm). There is also a strong push for rapid publication and access to data from industry-funded studies, which traditionally have shown a strong publication bias towards publication of positive results.

The detection of a small but palpable signal for suicidality associated with SSRI treatment must be placed in a larger public health perspective. The adolescent suicide rate in the United States and in many developing countries has been declining for over a decade, a period that coincides with the introduction of SSRIs (Brent 2004). Although pharmacoepidemiology studies are correlative and evaluate group rather than individual effects, they do show a relationship between an increase in SSRI prescriptions and sales and a decline in the overall and adolescent suicide rate (Olfson et al. 2003; Gibbons et al. 2005; Ludwig and Marcotte 2005). Subsequent to the FDA warning about antidepressants, there has been a decline in the number of prescriptions of SSRIs for youth without any concomitant increase in referrals for psychotherapy, and a decline in the rate of diagnosis of adolescent depression in the United States (Libby et al. 2007). This decline in SSRI prescriptions has also been noted in the Netherlands and in the United Kingdom. In the United States and the Netherlands, the decline in the use of antidepressants was associated with an increase in the adolescent suicide rate in the year following the FDA announcement (Gibbons et al. 2007). In contrast, the suicide rate in adolescent males and females continued to decline, and the rate of admissions for suicide attempts continued to increase, even after a dramatic decrease in the rate of SSRI prescriptions (Wheeler et al. 2008). Nevertheless, even a finding of no relationship between SSRI prescriptions and either suicide or suicide attempt rates supports the view that SSRIs do not significantly increase the risk for suicidal behaviour.

Rotheram-Borus et al. (1996) compared a six-session family CBT plus a one-session family psychoeducation session delivered in the emergency room to family CBT alone for a largely Dominican, female adolescent sample of 140 suicide attempters (Rotheram-Borus et al. 1996). The psychoeducation intervention increased compliance with the intervention, so that the study compares family CBT plus psychoeducation to a lower dose of family CBT (5.7 versus 4.7 sessions attended). The educational intervention trained emergency room staff to be more supportive and less negative towards adolescent suicide attempters and their families and provided a videotape for families to explain emergency room procedures and the rationale for continued treatment. Also, the family received one emergency family therapy session to develop a coping plan for potential suicide-inducing situations and provided liaison to the outpatient staff to insure continuity of care. The family cognitive behavioural therapy was a six-session programme that focused on developing positive family problem-solving, communication, and mutual reinforcement.

At the end of treatment, the enhanced CBT group showed lower adolescent suicidal ideation and lower maternal depressive and general symptomatology, greater maternally reported family cohesion, and more positive attitudes towards treatment (Rotheram-Borus et al. 1996). At 18-month follow-up, the rates of attempts and of significant suicidal ideation were not different between the two groups although there was a non-significant trend for a lower re-attempt rate in the experimental group (8.7 versus 14.6 per cent) (Rotheram-Borus et al. 2000). The enhanced intervention had the greatest impact on maternal emotional distress and family cohesion in families of more symptomatic adolescent attempters. Due to the design, one cannot tell to what extent these findings are attributable to the emergency room intervention itself or its impact on adherence.

Harrington and colleagues compared a 5-session home-based family intervention plus routine care to routine care alone for 162 adolescent who made suicide attempts by overdose in a randomized clinical trial (Harrington et al. 1998). The experimental intervention targeted family conflict, parental distress, adolescent suicidality, and hopelessness.

While satisfaction and compliance were higher in the home-based family treatment, the experimental intervention was no better than routine care overall for reducing ideation or re-attempt. However, among the non-depressed subgroup, the home-based treatment reduced suicidal ideation more so than did treatment as usual. The family intervention did not produce any significant changes in family climate relative to routine care, contrary to the intent of the treatment. However, there were fewer out-of-home placements in the family treatment (Byford et al. 1999).

A study compared a group therapy, termed developmental group therapy, plus usual care to usual care alone for sixty-three adolescents who had engaged in at least two episodes of deliberate self-harm within the past year. The experimental treatment focused on family conflict, problem-solving, interpersonal relationships, anger management, school problems, depression and hopelessness.

On average, participants had engaged in four prior episodes of self-harm, around half had experienced definite or probable sexual or physical abuse, and most came from socio-economically disadvantaged backgrounds. The experimental treatment consisted of a median of 8 group sessions, and a median of 2.5 individual sessions by the same therapists. Routine care was used much less often by those who were assigned to the group treatment compared to those who had routine care alone (median number of sessions 1 versus 4) and antidepressants were prescribed for a minority of participants in both treatments.

The experimental treatment, compared to routine care showed a reduction in episodes of self-harm (RR = 0.6) and a much lower rate of having two or more episodes (RR = 0.16). The time to first repetition was also longer in the experimental arm (11.9 versus 7 weeks). School attendance was improved in those subjects assigned to the experimental treatment, and the rate of behavioural disorder was lower as well (RR = 0.3). There was no differential treatment effect on depression, suicidal ideation, or global outcome.

A study compared a skills-based therapy (SBT) that consisted of training in problem-solving and emotion regulation versus supportive relationship therapy (SRT) for thirty-nine adolescent suicide attempters randomly in a randomized clinical trial. Treatment in both groups consisted of 6 individual sessions, and one adjunct session for over the first 3 months, followed by 3-monthly maintenance sessions. At the therapist's discretion, up to two additional family and two additional individual sessions were also available. About half of subjects had a history of two or more attempts, more than two disorders, and comorbid marijuana abuse. The randomization was balanced, except that a higher proportion of those assigned to supportive treatment had a disruptive behaviour disorder.

Both groups showed similar improvements in depression and ideation, with a trend towards better problem-solving in the SBT group. Subjects showed a non-significant trend for a higher rate of re-attempts in SBT group versus SRT at 3 months (4/15 versus 1/16) and 6 months (4/15 and 2/16). The majority of subjects in both treatments were in the non-clinical range on the Suicide Ideation Questionnaire by the end of treatment and upon 3- and 6-month follow-up. The dropout rate was higher in the experimental treatment (6/19 versus 2/19). Although the sample is too small for any definitive conclusions, this study suggests that SRT might be at least as efficacious as the experimental treatment.

Multisystemic therapy (MST), a flexible, family-based treatment that involves case-management and admixtures of individual and family treatment was compared to psychiatric hospitalization and usual care in 156 youths presenting to the emergency room with suicidal ideation, suicide attempt, homicidal ideation, psychosis, or other threat of harm to self or others (Huey et al. 2004). MST empowers families to communicate with, monitor, and discipline their children effectively, help promote prosocial activities for children and helps children disengage from antisocial peers. In the context of suicidality, MST focuses on helping parents to identify, contain, monitor, and diminish suicidal risk. Intervention typically lasts 3 to 6 months, but is intensive, with up to daily contact, delivered in the patient and family's natural ecology—often the home and the school. A large proportion of the MST group also were hospitalized (44 per cent) and this was not accounted for in the analyses, making the results difficult to interpret. At intake, according the suicide attempt item from the Youth Risk Behaviour Survey, 31 per cent of those assigned to MST had attempted suicide versus only 19 per cent of those who were hospitalized. At post-treatment and 1-year follow-up, the rates of re-attempt in MST were 14 per cent and 4 per cent versus 9 per cent and 4 per cent, respectively, which was reported as significantly different in the paper, although no treatment-by-time interactions were reported. There were no treatment effects for suicidal ideation, depression, internalizing symptoms, or hopelessness. Parental-rated control increased in MST relative to the hospitalization group at the end of treatment, but these effects did not persist upon 1-year follow-up.

A study evaluated the efficacy of augmenting the connection between suicidal adolescents and individuals in their social network whom the suicidal adolescent identified as being significant and supportive. This intervention, termed Youth-Nominated Support Team (YST-1) provides psychoeducation to individuals nominated by suicidal adolescents, and facilitates weekly contact between those nominated individuals and the suicidal adolescents. Therapists encourage support people to facilitate activities with the suicidal individual in the service of overall treatment goals. A group of 289 psychiatrically hospitalized adolescents who were admitted for significant suicidal ideation or an attempt were randomized to YST-1 plus usual care or usual care alone. Subjects in both conditions as part of routine care received psychotherapy and nearly all received some psychotropic medication. Attrition was greater in the experimental intervention (24.2 versus 12.9 per cent) because of unwillingness or inability for some adolescents to involve at least two supportive individuals. This also resulted in a low (35 per cent) acceptance rate into randomization. Adolescents in the YST-1 condition most commonly nominated parents, non-parental relatives, non-adult relatives, peers, or teachers.

Suicidal ideation declined in both groups, but among girls, the YST-1 group showed a greater decline in suicidal ideation in completer analyses, although no differences were found in the proportion who had suicidal ideation at a level below clinical cut-offs. No difference was found for the 6-month rate of suicide attempts. In both intent-to-treat and in completer analyses, girls experienced greater improvement in mood-related functioning in the YST-1 condition. The effect sizes for ideation and mood-related function reported in girls were in the small-to-moderate range.

To enhance protective factors against suicide, a school-based intervention, consisting of a one-semester ‘Personal Growth Class’ to enhance attendance, self-esteem, sense of personal control, and association with a prosocial peer group was implemented in two of four high schools (Eggert et al. 1994). The intervention groups showed greater improvement in drug-related problems, self-esteem, grade-point average, and involvement with prosocial peers. A subsequent study compared this intervention plus a screen for suicidality conducted by an interviewer to a more intensive version of this class (two semesters) and to screening alone (Eggert et al. 1995). In all three intervention groups, suicidal youth were assigned to a case manager at the school. Surprisingly, the brief assessment intervention was as efficacious as the more extended intervention in reduction of suicidal ideation and associated risk factors, with one exception—the Personal Growth Class resulted in greater improvements in self-rated personal control. The results of this study suggest that a brief screening and contact with a case manager may be an effective and sufficient intervention in schools to reduce suicidal risk, although this awaits replication in comparison to a control intervention.

Another proposed prevention approach is on-site screening of youth for depression and suicidality, using the Columbia Suicide Screen (Shaffer et al. 2004). This self-report questionnaire showed a sensitivity 0.75 and specificity of 0.83 against a structured interview, with a positive predictive value of 16 per cent. The TeenScreen appears to identify at-risk youth who are not already recognized by school professionals: of adolescents identified as thinking of suicide, having attempted suicide, or being depressed, 74 per cent, 50 per cent, and 69 per cent of these youth were previously unrecognized by school personnel (Shaffer and Craft 1999). There has been no published evaluation of the TeenScreen on referral or outcome. However, there do not appear to be iatrogenic effects of screening youth for suicidal ideation (Gould et al. 2005). See also Chapter 68 in this book.

The Signs of Suicide (SOS) programme combines self-screening and an educational intervention (video and discussion) designed to improve high school students' efficacy in recognizing suicidal behaviour and depression and in seeking help. In five high schools, a total of 2100 students were randomly assigned to the intervention or a control condition; all students were surveyed 3 months after the intervention with regard to attitudes and knowledge about depression and suicide, help-seeking, and rate of self-reported suicide attempts. In the intervention condition, students were screened using the Columbia Depression Scale, a brief scale derived from the Diagnostic Interview Schedule for Children (Aseltine and DeMartino 2004). If they scored 16 or higher, they were instructed to seek help immediately. Upon 3-month follow-up, students were provided with a list of sources for clinical care. Students in the intervention group were 40 per cent less likely to have made a suicide attempt (3.6 versus 5.4 per cent). showed greater knowledge about depression and suicide, and more adaptive attitudes towards these problems, but without significant changes in help-seeking behaviour. The program effects on suicide attempts may have been mediated through improvements in student knowledge and attitudes.

A one-session combination assessment and intervention using contracting and teaching simple emotion regulation techniques for youth staying in runaway shelters judged to be a high suicidal risk was evaluated using a quasi-experimental design (Rotheram-Borus and Bradley 1991). In youth judged to be at high suicidal risk by a standard assessment, youth were asked to engage in five behavioural tasks:

In a pre-post evaluation, the number of attempts in shelters in the three months prior to implementation was 9, as compared to two attempts in the 18 months after the programme was implemented.

Dialectic behaviour therapy (DBT), a treatment that focuses on the development of emotional regulation, distress tolerance, and interpersonal social skills, has been shown in personality disordered adults to reduce recurrent suicidal and self-harm behaviour (Linehan et al. 1991, 1993, van den Bosch et al. 2002; Bohus et al. 2004). DBT has been modified for use for suicidal adolescents, by shortening the treatment from 1 year to 12 weeks, reducing the total number of skills taught, incorporating family members into treatment, and offering an optional 12-week patient consultation group (Miller et al. 1997). Their modification of DBT emphasizes the following skills: mindfulness, emotion regulation, distress tolerance, and interpersonal effectiveness (Miller et al. 2007). Identity problems, impulsivity, emotional instability, and interpersonal problems were targeted by cultivating mindfulness, distress tolerance, emotion regulation, and interpersonal effectiveness. An open pilot study of 27 14–19-year-old suicidal adolescents showed improvements in all target areas and a trend towards a lower rate of suicidal behaviour on follow-up (Miller et al. 2000).

Rathus and Miller (2002) then conducted a quasi-experimental study in which 111 suicidal adolescents with borderline personality disorder features received either DBT (N = 29) or treatment as usual (TAU) (N = 82). the latter of which consisted of twice-weekly supportive-psychodynamic therapy plus weekly family therapy (Rathus and Miller 2002). At intake, the DBT group had more previous attempts, was more depressed, and had had more previous hospitalizations. Although not statistically significant, fewer subjects made suicide attempts in the DBT condition (3.4 versus 8.6 per cent). fewer subjects were hospitalized, and the completion rates for treatment were higher in the DBT group. Within-treatment examination of the impact of DBT showed significant reductions in suicidal ideation, general symptomalogy, and borderline symptoms. However, changes in symptomatology were not compared to the TAU group. This study supports further work in the application of DBT to the treatment of suicidal adolescents.

In a subsequent study, sixty-two adolescents with suicidal ideation or attempt were admitted to one of two psychiatric units, one of which offered inpatient DBT (Katz et al. 2004). Although subjects were not randomly assigned, admission was based on bed availability rather than any clinical indicators. Subjects were evaluated at intake, discharge, and 1 year later. The DBT condition involved ten daily DBT skills training sessions, and twice-weekly individual treatment, delivered over two weeks. Routine care consisted of a daily psychodynamic therapy group, individual psychodynamic therapy at least once a week, and a psychodynamic milieu. Both groups received pharmacotherapy and were discharged to community mental health clinics.

Both groups showed similar substantial reduction in self-reported depression, suicidal ideation, and hopelessness. There were also no differences in re-attempts, re-hospitalization or adherence to outpatient treatment. Fewer incident reports were generated on the DBT unit. This study, as the authors note, is limited by lack of randomization, blind ratings, formal fidelity checks for DBT, and relatively small sample size. While the findings do show that DBT can be delivered on an inpatient unit, results in high retention and satisfaction, and produces a reduction in targeted symptoms, this pilot study does not provide support to the hypothesis that DBT is superior to standard inpatient care.

Brown and colleagues (2005) have developed a cognitive behaviour model for adult suicide attempters that focuses almost entirely on cognitions that are associated with suicidal behaviour, which in turn is derived from a careful chain analysis of the attempt (Brown et al. 2005). Important components of the treatment include active case-management, development and monitoring of a safety plan, development of a hope kit and other coping strategies for suicidal crises. Treatment targets included improved problem-solving with increased adaptive use of social support, and improved compliance with medical, psychiatric, and substance abuse treatment. A group of 120 adult suicide attempts with high rates of depression and substance abuse were randomized to either CBT plus routine care or case management plus routine care. CBT-treated subjects were half as likely to re-attempt as those in the enhanced usual care group over the 18 month follow-up (24.1 versus 41.6 per cent) and showed great improvements on self-reported depression and hopelessness on follow-up. This approach has been adapted for adolescent suicide attempters and is now being pilot tested in five sites in a project funded by the National Institutes of Mental Health.

For a description of different techniques used in different individual therapies, see Chapter 92. One of the standard interventions for individuals judged to be at high risk for suicide is psychiatric hospitalization. The risk of suicide and re-attempt is extremely high after discharge from the hospital (Kjelsberg et al. 1994). One study has found that written contact after discharge from the hospital was associated with a lower risk for suicide in adults at high suicide risk who were non-compliant with outpatient treatment (Motto and Bostrom 2001). A replication of this intervention, ‘Postcards from the Edge,’ was conducted on adolescent and young adult self-poisoners who were medically hospitalized (Carter et al. 2005). A total of 772 aged 16 and older subjects were randomized to receiving 8 postcards over 12 months plus treatment as usual versus treatment as usual alone. There were 192 episodes of self-poisoning in the control condition versus 101 events in the experimental intervention group (RR = 0.55). Patients randomized to the experimental arm used 110 fewer hospital bed-days. However, the proportion of those who re-attempted was not different between groups. This intervention appears to be an inexpensive method for reducing repetition of hospital-treated suicide attempts. However, the proportion of subjects who reattempted was not different between the experiment and control groups (15.1 versus 17.3 per cent).

Cotgrove and colleagues conducted a randomized trial of post-hospital care of adolescent suicide attempters (Cotgrove et al. 1995). Subjects (N = 105) were randomized to either immediate, unquestioned readmission to hospital plus usual care, or usual care alone, and followed up one year later. Of those with open access, 6 per cent reattempted suicide versus 12 per cent of those in usual care, which, while lower, was not statistically significant. A post-hoc analysis seemed to indicate that open access was protective only for those at low or moderate risk for suicide re-attempt. Of those given open access, 11 per cent requested hospitalization.

There have been no pharmacological trials that target adolescent suicidal behaviour, and very few in adults. We review studies in those agents that show efficacy in targeting impulsive aggression and/or suicidal behaviour.

Lithium has not been evaluated for suicidal behaviour in a double-blind controlled trial in either children or adults, although one is now ongoing in adults (Lauterbach et al. 2005). Naturalistic studies comparing those treated with lithium compared to those on either no treatment or alternative mood stabilizers (carbamazepine, divalproex) find a protective effect against suicide (Rifkin et al. 1997; Baldessarini and Tondo 2003; Goodwin et al. 2003). In a population-based sample of 20,638 members of a health plan aged 14 and older with a diagnosis of bipolar disorder and at least one filled prescription of lithium, divalproex, or carbamazepine, the rates of suicide attempts and completions were compared. After adjustment for age, sex, comorbid medical and psychiatric conditions, the risk of death by suicide was increased 2.7-fold, and the risk for attempted suicide resulting in hospitalization was increased 1.7-fold (Goodwin et al. 2003). In a similar type of study, 12,662 Medicaid patients with bipolar disorder were compared with respect to risk for suicide and attempted suicide as a function of their prescribed medication (Collins and McFarland 2008). The risks for attempted suicide were 2.7 times higher in those using divalproex, and 1.6 times higher for those using gabapentin than in those using lithium. Similarly, the risk compared to those on lithium for completed suicide was 1.5 times higher in those on divalproex (not significant) and 2.6 times higher in those on gabapentin.

A similar analysis found that lithium was protective against suicide for patients with recurrent unipolar depression (Guzzetta et al. 2007). In 329 depressed patients, the risk for suicide and suicide attempts was compared while on (1149 person-years) and off (1285 person-years) lithium. The risk for completed and attempted suicide were 8.7 and 6.8 times higher, respectively, in patients during periods of non-exposure to lithium.

While the results of these studies are of interest, because patients are not randomly assigned to their medications, the results are difficult to interpret. However, a meta-analysis of clinical trials with lithium in mood disorder patients provides a compelling case for reduction of suicide risk associated with the use of lithium. In 32 trials, 1389 patients were randomized to lithium and 2069 were assigned to other compounds or placebo. Those who received other compounds were four times more likely to die by suicide than those who received lithium, and five times more likely to either die by suicide or make a suicide attempt (Cipriani et al. 2005).

Lithium has been used to target aggression in children and adolescents without mood disorder (Campbell et al. 1984; Malone et al. 2000). Most, but not all, find a positive effect in the reduction of aggression. These results support further exploration of the role of lithium in the management of suicidal behaviour.

Two studies have found efficacy for divalproex in the management of impulsive aggression, mood lability, and behavioural symptoms in child, adolescent, and adult samples (Donovan et al. 2000; Hollander et al. 2003; Steiner et al. 2003). Divalproex has not been experimentally evaluated for management of suicidal behaviour. However, in naturalistic studies, lithium appeared to be superior to divalproex in preventing suicidal behaviour in bipolar adults (Goodwin et al. 2003; Collins and McFarland 2008).

Three experimental studies in adults support the use of neuroleptics in the treatment of suicidal outcome in adults. One large study of schizophrenics who were judged at high suicidal risk because of suicide attempt or current suicide ideation, clozapine was found to be superior to olanzapine in reducing suicidal behaviour over a 24-month period (Meltzer et al. 2003). Consistent with these results, an open trial of clozapine in adolescent schizophrenic patients showed a reduction in aggressive behaviour (Kranzler et al. 2005). In a national sample of adult schizophrenic patients who had been hospitalized for a suicide attempt, both antipsychotic use and combined antipsychotic and antidepressant use was associated with a lower mortality due to suicide (Haukka et al. 2008). The combination of olanzapine/fluoxetine (OFC) was compared to lamotregine (LMG) for the treatment of bipolar depression in a 7-week trial (Brown et al. 2006). Patients assigned to OFC showed greater global improvement, reduction in depression, and a lower incidence of suicidal and self-injurious behaviour (0.5 versus 3.4 per cent, p <0.04). Another, unpublished study found an injectable neuroleptic to be superior to placebo in preventing recurrent suicidal behaviour in adults with a history of repeated attempts (Montgomery et al. 1979). While the use of neuroleptics for the management of youthful suicidal behaviour has not been evaluated, atypical neuroleptics have been shown to be efficacious in managing aggressive behaviour (Findling et al. 2000; Schur et al. 2003). The most significant drawback of this approach is the problem of significant weight gain found even in the short-term use of these medications.

Fluoxetine has been shown to diminish both impulsive aggression in personality disordered patients, and the intensity and frequency of ‘anger attacks’ in patients with depression (Fava et al. 1993, 1996; Coccaro and Kavoussi 1997). However, SSRIs do not appear to prevent recurrent suicidal behaviour in adults without mood disorder (Montgomery et al. 1994; Verkes et al. 1998). No studies have evaluated the impact of fluoxetine on impulsive aggression in paediatric samples, nor has this domain been assessed as an outcome in paediatric clinical trials involving fluoxetine or other SSRIs.

Efficacious treatments have been demonstrated for depression among youth, but it has not been conclusively demonstrated that either antidepressants or psychotherapy reduce suicidal ideation or risk for suicidal behaviour in depressed adolescents compared to control treatments. This may be in part due to exclusion of the most suicidal individuals from clinical trials, lack of careful assessment of suicidal indicators in most trials, and inadequate sample size to detect effects. In one clinical trial, the combination of CBT and medication was the only treatment superior to placebo in reducing suicidal ideation. Epidemiological evidence seems to suggest that there is a relationship between increased use of SSRIs and a decline in the adolescent suicide rate and conversely between a decline in SSRI use and an increase in adolescent suicide. While there is some evidence to support a relatively rare but palpable increase in suicidality with the treatment of SSRIs, many more youth appear to benefit from, than are harmed by, antidepressants.

Subsequent clinical trials in patients with mood disorder and other conditions at high risk for suicide should consider including suicidal subjects, improving the assessment of intake suicidal risk factors (ideation, past attempt, impulsive aggression, hopelessness) and the effect of interventions on these risk factors and on suicidal ideation and behaviour. In a recent clinical trial with treatment-resistant depressed adolescents, almost 60 per cent of participants showed clinically significant suicidal ideation upon entry, supporting the feasibility of the recruitment, management, and protection of suicidal participants in a clinical trial (Brent et al. 2008). This would permit a more accurate estimate of the impact of treatment on suicidal risk. Future work should also focus on the identification of predictors of treatment response and adverse events in those treated with SSRIs.

Treatment studies of adolescent suicide attempters are few, often with small samples and modest effects. The most promising results to date have been those of which found a reduction in recurrent deliberate self-harm in a pilot study of a brief intervention. On the basis of the results in adult studies, treatment trials to evaluate the efficacy of DBT and CBT for the treatment of suicidal adolescents should be developed. The role of simple after care interventions like ‘Postcards from the Edge’ should also be pursued (Carter et al. 2005).

There are several pharmacological agents with promise to reduce suicidal risk and precursors therefore, such as impulsive aggression, namely, lithium, divalproex, and atypical neuroleptics. Because each of these medications has several side-effects that make them less than desirable for adolescents (e.g., weight gain), studies should first be conducted on those for whom other reasonable psychosocial interventions have failed and where the patient has engaged in recurrent suicidal behaviour.

While there is a great deal known about the risk factors for suicidal behaviour, and the precursors thereof, there has been relatively little attempt to prevent the development of suicidal behaviour in at-risk youth. Given the association of core protective factors and low risk for suicidal behaviour, the role of interventions to strengthen protective factors to prevent onset of suicidality and other health risk behaviours should be assessed in the context of both universal and indicated prevention (Resnick et al. 1997; Borowsky et al. 2001).

Despite suicide being the leading and most severe complication of psychiatric disorder, there are a paucity of studies that target individuals at high suicidal risk. Investigators have avoided these populations because of the difficulty in engaging such subjects in treatment and because of concerns of a suicide occurring in a clinical trial. On the other hand, systematic exclusion of high-risk subjects from clinical investigation truncated our knowledge about how to best treat these patients. Therefore, ethical guidelines and safeguards should be developed that allow for these studies that include suicidal individuals; in order to improve treatment for these vulnerable people.