Conducting Research on Adolescent Suicide Attempters: Dilemmas and Decisions

Abstract

Research on effective treatments for adolescent suicide attempters is urgently needed. However, there has been a lack of research in this area. This is likely a result of the multiple challenges faced by investigators working with individuals at high risk of suicide. Based on our experiences conducting a large, randomized clinical trial with adolescent suicide attempters, in this article, we review ways to address these challenges in order to facilitate needed research on suicide prevention in adolescents.

Suicide is the third leading cause of death among 10 to 24 year-olds in the United States (Centers for Disease Control and Prevention, 2010). Recent statistics from a nationally-based survey of high-school students in the United States showed that 15.8% had seriously considered attempting suicide in the past year, 12.8% had made a plan about how they would attempt suicide, and 7.8% had attempted suicide one or more times (Eaton et al., 2012). Among 15-24 year-olds, there are approximately 100-200 suicide attempts for every completed suicide (Goldsmith, Pellmar, Kleinman, & Bunney, 2002) and prior suicide attempts are one of the strongest predictors of subsequent suicide attempts and suicide deaths in both adolescents and adults (e.g., Harris & Barraclough, 1997; Lewinsohn, Rohde, & Seeley, 1994; Shaffer, et al., 1996).

Currently, there are no treatments specifically targeting suicide attempts in adolescents that meet criteria for a “well-established” empirically supported treatment (American Psychological Association, 2006). As a result, guidelines for managing and treating these high risk adolescents are based on a combination of “expert opinion” and a small number of randomized and non-randomized intervention trials (Asarnow & Miranda, in press). There are only eight randomized controlled trials (RCTs) of treatments for adolescent suicide attempters that targeted reduction in reattempts as their primary outcome. Only four of these trials yielded significant results. These interventions consisted of: 1) group therapy including both cognitive-behavioral and psychodynamic techniques (Wood, Trainor, Rothwell, Moore & Harrington, 2001); 2) multi-systemic therapy (Huey et al., 2004); 3) mentalization-based treatment (Rossouw & Fonagy, 2012); and 4) integrated CBT for co-morbid alcohol abuse disorders and suicidal thoughts or behaviors (Esposito-Smythers, Spirito, Kahler, Hunt, & Monti, 2011). The group therapy approach failed to be replicated in two subsequent follow-up trials (Green et al., 2011; Hazell et al., 2009) and the other three studies have yet to be replicated. The four trials that did not yield significant decreases in suicide attempts included: 1) a green card offering rapid, no questions asked hospital admission if requested (Cotgrove, Zirinsky, Black & Weston, 1995); 2) a brief home-based problem solving intervention (Harrington et al., 1998); 3) a skills-based approach targeting problem-solving and affect management (Donaldson, Spirito, & Esposito-Smythers, 2005; and 4) a youth-nominated support team (plus a second trial using a slightly modified version of the approach; King et al., 2006, 2009). It is clear that further research is urgently needed.

Research on suicide attempters presents multiple challenges for investigators, which likely accounts for the lack of needed research in this area (Iltis et al., 2013; Linehan, 1997; Pearson, Stanley, King, & Fisher, 2001). Challenges include management of the significant anxiety associated with working with suicidal individuals, perceived liability risks for investigators, the need for sufficient expertise and resources to monitor and treat suicidal subjects, and the large sample sizes needed for sufficient statistical power to detect between-group differences in suicide-related outcomes (Pearson et al., 2001). In this article, we discuss ways to address these issues based on our experiences conducting the Collaborative Adolescent Research on Emotions and Suicide (CARES) Study, the first RCT of Dialectical Behavior Therapy (DBT) that specifically targets adolescent suicide attempters with current high suicide ideation. Our goal is to facilitate additional research in this understudied area by offering suggestions that reduce the stressors and concerns associated with studying highly suicidal adolescents. First, we provide a brief description of the CARES Study. Next, we review multiple roadblocks that are likely to be encountered when working with this population and strategies for addressing them.

Collaborative Adolescent Research on Emotions and Suicide (CARES) Study

The CARES Study is a multi-site RCT being conducted at the University of Washington, Seattle Children’s Hospital, Harbor-UCLA Medical Center, and the University of California, Los Angeles. A total of 170 adolescents will be enrolled in the study across sites. Inclusion and exclusion criteria are shown in Table 1. Adolescents who meet study inclusion criteria are randomly assigned to receive six months of either DBT or Individual and Group Supportive Therapy (IGST). Outcome assessments are conducted at 3, 6, 9, and 12 months. The primary outcome variable is suicide events (suicide, suicide attempts, or emergency department visit or inpatient hospitalization for suicidality). Assessments also incorporate a number of domains that are associated with increased risk of suicide attempts, including multiple measures of psychopathology (e.g., mood and anxiety disorders, PTSD, psychosis, substance abuse, and borderline personality disorder traits), difficulties with emotion regulation, impulsivity, social adjustment, coping skills, and family functioning. Potential mediators of treatment outcomes, such as increased emotion regulation and decreased family conflict, will also be examined. At present, we have enrolled approximately two-thirds of the sample.

Table 1.

Inclusion Criteria	Exclusion Criteria
Elevated suicide ideation within the past month	Acute psychiatric or medical symptoms (e.g., traumatic brain injury, substance dependence requiring inpatient detoxification) that would interfere with the adolescent’s ability to participate in outpatient psychotherapy and/or study assessments
History of at least one life-time suicide attempt	Adolescent is court ordered to treatment
Recurrent intentional self-injury: History of at least 3 intentional selfinjuries, one within 12 weeks of referral to the study	IQ score less than 70
Presence of at least 2 BPD criteria besides the recurrent intentional self-injury criterion
12-18 years old
At least one family member or responsible adult agrees to participate in assessment and treatment

Research on Suicidal Adolescents: Dilemmas and Decisions Selection of an experimental condition

Two factors are needed in selecting an experimental treatment for study. First, the treatment to be studied needs to have enough preliminary evidence to warrant a RCT. Second, there has to be a need for another study, i.e., the study must be designed to provide new information. DBT was selected due to its known efficacy with suicidal adults (Koons et al., 2001; Linehan, Armstrong, Suarez, & Allmon, 1991; Linehan et al., 2006; Verheul et al., 2003. However, no RCTs on DBT have been conducted with adolescents selected for high suicidality. This is a problem due to the fact that DBT is already being widely provided to adolescents in clinical settings in the absence of data on efficacy. A number of pilot trials of DBT adapted for adolescents have been conducted demonstrating the feasibility and promise of DBT for the adolescent population (Fleischhaker et al., 2011; Katz, Cox, Gunasekara, & Miller, 2004; Rathus & Miller, 2002; Woodberry & Popenoe, 2008). What is missing is a sufficiently powered RCT of DBT for adolescents selected due to previous and current high suicidality. In sum, based on the strength of the data demonstrating the efficacy of DBT with suicidal adults, the promising results obtained in small studies of DBT with suicidal adolescents, and the wide-spread dissemination of DBT for adolescents in response to clinical need, without support from a RCT, it is clear that a RCT of DBT with adolescent suicide attempters is justified and is a critical next step in research on adolescent suicide prevention.

Selection of a control condition

An optimal control condition needs to be safe, potentially effective, and desirable to participants. In one large-scale study of treatment for adolescent suicide attempters, researchers were unable to conduct an RCT as planned due to youth and parents’ unwillingness to be randomized to the study conditions (which included CBT, medication, and CBT plus medication; Brent et al., 2009). Because we were interested in maximizing internal validity, we used an active treatment control condition in which we could control for as many aspects of treatment delivery as possible. As there currently are no evidence-based treatments for suicidal adolescents, there was no clear choice of a control treatment (Spirito, Stanton, Donaldson, & Boergers, 2002). We selected IGST based on prior studies showing that supportive therapy led to decreases in suicidality (defined as suicidal ideation with a plan or a suicide attempt) equivalent to CBT and systemic behavior family therapy in a sample of depressed adolescents (Brent et al., 1997) and was equivalent to CBT in decreasing suicidal ideation and attempts in a sample of adolescent suicide attempters (Donaldson et al., 2005). Supportive therapy techniques were also shown to be the most commonly reported elements of TAU in a sample of adolescent suicide attempters (Spirito et al., 2002). Client-centered therapy has also been used as a comparison group in multiple RCTs that examined trauma-focused CBT with traumatized youth (Cohen, Deblinger, Mannarino, & Steer, 2004; Cohen, Mannarino, & Knudsen, 2005). In order to enhance internal validity, IGST is designed to control for key treatment elements such as hours of treatment provided, treatment modalities provided (e.g., both individual and group therapy), therapy drop-out policies, therapist expertise and availability of supervision.

Recruitment of high risk adolescents

Although suicide is a leading cause of death among adolescents, it occurs at a relatively low base rate in the general population. Hence, large samples are needed for sufficient statistical power to detect between group differences in suicidal behaviors. Moreover, it is important that researchers use a sample at high risk for suicide so that enough suicidal behaviors occur during the study to compare groups on suicide-related outcomes (Linehan, 1997). In the CARES Study, the need for a large sample size was addressed by conducting a multi-site study. In order to insure that we recruited a sample at high risk of engaging in suicidal behavior, we based our inclusion criteria on documented risk factors for suicide and suicide attempts in adolescents (see Table 1). Finally, over time, we established strong referral networks with settings that were likely to treat highly suicidal adolescents, such as inpatient units, residential treatment programs, emergency rooms, psychiatric mobile response teams, and community-based clinics. To the best of our knowledge, we have recruited one of the highest risk samples of suicidal adolescents to date.

How to safely manage suicide risk

Working with such a high risk sample requires responsible suicide risk management protocols for both experimental and control conditions. However, the use of intensive risk protocols across study conditions must also be balanced with the scientific concern of reducing power to detect between group differences (Pearson et al., 2001). Further complicating the matter is the lack of ability to accurately predict which individuals will ultimately die by suicide. Taking into account these multiple concerns, in order to ensure responsible risk management that was consistent with the two treatments provided, we created separate, detailed risk management protocols for each condition. Given that there are no standard suicide risk management protocols that are used uniformly across clinical settings in the United States, both risk management protocols utilized in this study are likely to be superior to treatment as usual, provide manualized safety monitoring and risk management, and are consistent with ethical and legal requirements for the protection of human subjects.

Clinicians providing IGST follow the risk management procedures outlined in the American Academy of Child and Adolescent Psychiatry’s Practice Parameters for the Assessment and Treatment of Children and Adolescents with Suicidal Behavior (Shaffer & Pfeffer, 2001). They are provided with extensive training on how to assess suicide risk and on the standard safety precautions to be reviewed with both youth and parent (e.g., removal of lethal means, increased parental monitoring, provision of telephone numbers of local emergency services) that are recommended in the Practice Parameters. They are also provided with guidance on how to perform these procedures while staying within the non-directive, client-centered approach of the treatment model. Youth and parents are also given the telephone numbers of local and national suicide hotlines that are available for 24-hour, 7 day per week crisis management if needed.

In the DBT condition, therapists follow DBT assessment and treatment protocols for suicidal individuals, including the Linehan Risk Assessment and Management Protocol (LRAMP; Linehan, 2009). The LRAMP is a semi-structured assessment checklist that guides clinicians through an extensive list of risk factors for imminent suicide and enables him/her to conduct and document a comprehensive risk assessment that addresses liability concerns. It also assists the clinician in determining and justifying his/her course of action (e.g., recommending hospitalization or not) and in creating a safety plan. The LRAMP is completed in the first session with a new client and subsequently completed at any time during treatment when the client reports self injury, a suicide attempt, an increase in suicidal urges, or threatens suicide. As part of the standard DBT protocol, therapists also provide clients and parents with 24-hour, 7 day a week telephone coaching within limits, with the goal of learning to use DBT skills in both suicidal and non-suicidal crises. In the absence of any data that hospitalization is an effective treatment for suicidality (for reviews, see Bridge et al., 2006; Gould et al., 2003), DBT has a strong preference for avoiding hospitalization for suicidal individuals and rarely suggests inpatient care, although it is not prohibited if it is needed. Although DBT promotes the use of coping skills instead of psychotropic medication (replacing pills with skills) to manage negative emotions, it also includes a rescue medication protocol. In particular, the DBT therapist recommends immediate treatment with medications in the following two instances, both of which have been shown to be predictors of imminent suicide: 1) severe insomnia combined with escalating agitation or suicide ideation (Bernert & Joiner, 2007; Fawcett, 2013; Linehan, 1981) and 2) a severe psychotic episode (Hawton, Sutton, Haw, Sinclair, & Deeks, 2005; Hor & Taylor, 2010).

Detailed safety protocols are also utilized during study assessments across both study conditions. Assessment interviewers utilize the Linehan Risk Assessment Protocol (LRAP; Reynolds, Lindenboim, Comtois, Murray, & Linehan, 2006). The LRAP includes an assessment of suicide and self-injury risk pre- and post-assessment, strategies to decrease distress and related suicidal and self-injurious urges, and procedures for when to increase the level of response (e.g., escorting the subject to the hospital). Of note, there is no evidence that assessment of suicidal behavior (whether for treatment or research purposes) “primes” vulnerable individuals and leads to increased suicide risk or risk of non-suicidal self-injury (Biddle et al., 2013; Gould et al., 2005). Given that self-harm and suicidal behaviors are inherent risks in a study that recruits expressly for highly suicidal people, and the importance of protecting our participants, the LRAP is administered as a standard part of each assessment battery. The LRAP includes a protocol for calling in a supervisor to speak with the subject before s/he is allowed to go home if the other elements of the LRAP do not sufficiently reduce distress.

Several additional steps have been taken to enhance and manage safety. The Principal Investigators (PIs) of the study are experts in working with suicidal clients, as well as in conducting large scale clinical trials. Study PIs and clinical supervisors are available to study staff at all times for consultation regarding safety concerns. As described above, study therapists and assessors receive extensive training on risk assessment and management protocols. As required by NIH for all intervention trials, the study has a Data Safety and Monitoring Board that meets on a quarterly basis to evaluate the safety of the trial. There is also a study ombudsman designated at each site who is available to independently evaluate whether or not a subject needs to be removed from the study protocol. Because there are no evidence-based treatments for suicidal adolescents, and no data showing that hospitalization or residential treatment are superior to outpatient care (Bridge et al., 2006; Gould et al., 2003; Van der Sande et al., 1997; Waterhouse & Platt, 1990), there is no strong rationale for pulling subjects out of the study treatment just because they become more suicidal during the study. However, if at any time, any individual involved in the adolescent’s treatment (e.g., the therapist, the adolescent, the parent, the supervisor, the PI) feels that he/she is not benefiting from the study treatment or is getting worse, and there is reason to believe that an alternative treatment exists that has a greater likelihood of addressing the client’s needs, a meeting with the ombudsman and the family is automatically initiated. The ombudsman makes the final decision as to whether or not the youth should be removed from the study protocol.

How to manage anxiety

Finally, and perhaps most importantly, working with highly suicidal adolescents creates a great deal of anxiety among therapists and investigators. Indeed, the thought of a child dying by suicide is difficult to bear and the assessment, management and treatment of suicidal clients are among the most stressful tasks facing clinicians (Jobes, 1995). It is critical that this anxiety is adequately addressed and managed in order to prevent it from interfering with the implementation of appropriate safety procedures (Pearson et al., 2001). For example, therapists’ fears may compel them to either under or over-assess suicidality, or to deviate from study protocols, which could lead to suboptimal risk management. It is critical that the study is led by investigators and clinical supervisors who can tolerate the anxiety associated with working with suicidal adolescents and model this for others. In order to address anxiety in our research teams, we have: a) provided ongoing training and education about risk management procedures, b) provided education about the limits of therapists’ ability to predict and prevent suicides, c) had therapists meet regularly with clinical supervisors and in teams to provide each other with support and guidance, d) had PIs and clinical supervisors who are available 24/7 for consultation, e) given therapists small caseloads to prevent burnout and allow time for careful management of cases, f) provided detailed safety protocols to be followed, g) conducted regularly fidelity monitoring of therapy and assessment sessions, and h) emphasized the critical importance of the work research team members are doing and the potential for their work to save lives in the future.

Conclusion

Adolescents who have attempted suicide are at high risk for subsequent suicide attempts and death by suicide and evidence-based treatment approaches are urgently needed. However, at present, there is a relatively small amount of treatment research that has been conducted on this population and no interventions meeting criteria for a “well-established” empirically supported treatment (APA, 2006). The lack of research studies in this area is likely due to the multiple difficulties encountered in working with a sample of highly suicidal individuals. In light of our experiences conducting the CARES Study, a large RCT examining the efficacy of DBT with adolescent suicide attempters, we discussed ways to address the issues that deter researchers from conducting this research. In particular, we discussed how to select scientifically-sound treatment and control groups, recruitment of high risk adolescents, safety protocols, and managing anxiety. We hope that this article will be instructive for investigators considering doing research on this topic and will facilitate additional research.