Mechanisms Underlying Suicide Risk: Integrating RDoC to Inform Novel and Personalized Intervention Research
The NIMH Division of Translational Research and NIMH Division of Services and Intervention Research sponsored a workshop titled, “Mechanisms Underlying Suicide Risk: Integrating RDoC to Inform Novel and Personalized Interventions.” With input from the National Institute on Drug Abuse (NIDA), the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), and the National Institute on Alcohol Abuse and Alcoholism (NIAAA), a multidisciplinary group of researchers convened to review current themes in suicide research and to consider how approaches consistent with the Research Domain Criteria (RDoC) framework could provide new insight on underlying mechanisms of suicide risk.
Several challenges exist in research on suicide including the low base rate, transdiagnostic nature of the behavior, variability across the lifespan, and potential sensitive periods of development. All of these factors contribute to the significant amount of heterogeneity seen in the expression of suicidal thoughts and behaviors, and increase the difficulty of this research. Given this heterogeneity, there are likely classes of individuals with similar biological constitutions that are associated with similar increased suicide risk or differential treatment response rates.
An opportunity exists with the RDoC framework which makes it possible to break down suicidality into its component parts and identify the constructs or sets of constructs that interact with each other to give rise to suicide ideation, attempts and deaths, thus illuminating the mechanisms of suicidality and potentially identifying modifiable targets for intervention. The overarching goal of the workshop was to inform personalized novel intervention development by generating a set of constructs measured at specific levels of analysis that could facilitate research on risk and protective mechanisms, and could be points of intervention for those at risk for suicidality. An additional goal was to facilitate coordinated research efforts to identify common data elements and measures, and to explore possibilities for future multidisciplinary collaborations that might address the challenges of research on suicidal thoughts and behaviors.
An Introduction to the RDoC Framework and Existing Suicide Models and Theories
The initial session of the meeting included a demonstration of the need for advancement in the field of suicide prediction research. A review of the past 20 years of suicide research revealed a pattern of studies that examined the same predictive factors, utilizing similar methods, and making similar conclusions. In spite of identification of potential population-based risk factors, we have not advanced as a field in our ability to predict suicide on the level of individuals. The meta-analysis was followed by an overview of the goals of the RDoC Initiative and study design principles, noting the advantages of the RDoC research framework for a transdiagnostic phenomenon like suicide. Common elements of suicide theories were identified (e.g. emotional pain, stress response, cognitive rigidity) as was their corresponding locations in the RDoC matrix, suggesting areas of traction for future research. Among the limitations of a number of current theoretical approaches noted was the lack of consideration of the dynamics of constructs—specifically how proximal causes of heightened risk or factors could explain why individuals move from ideation to attempt. The field would benefit from novel approaches that include multiple levels of analysis and ways to look at how factors combine to increase risk. A considerable impediment to progress has been variability in experimental measures and a lack of agreement on which constructs have greatest relevance for suicide risk.
Another challenge in this new approach to suicide research was to find ways to share negative findings, so that efforts could be focused on concepts and paradigms that are more likely to be fruitful. Because some of the negative findings could be due to lack of statistical power, workshop participants also felt that coordination across studies could also enable progress in several ways: Harmonization and standardization of measures; multi-site studies to increase sample size, formal infrastructure to support training activities; expanded opportunities to present research meetings, and federation of data collection and data sharing.
Incorporating RDoC Measures and Methodology into Suicide Risk Research
The second session was focused on specific examples of how research might adopt dimensional measures and RDoC concepts. A growing number of studies point to several key domains for suicide risk, including risky decision making, cognitive control, overgeneralized autobiographical memory and increased anhedonic symptoms. Neuroimaging studies implicate neural circuits involved in cognitive control, mood regulation and decision-making. An important goal is to understand why people transition from suicidal thoughts to suicidal actions. Research comparing individuals with suicidal ideation to individuals who attempt suicide reveals that some behavioral tests are capable of distinguishing between the two groups. For example, people who attempt suicide respond to rewards slower, show greater startle reflex, and are more distracted by irrelevant information than individuals with suicidal ideation. Brain activations during these behavioral tasks also show differences between people who have experienced suicidal ideation and those who have previously attempted suicide. Other research has shown that individuals who have difficulty redirecting and regulating their mood (termed maladaptive repair) are more likely to have current and future depression and are more likely to engage in suicidal thoughts and behaviors. RDoC domains that are relevant to mood repair and suicide risk that could be further researched include negative valence, cognitive control, arousal and regulatory systems, habit, and social processes.
Another topic addressed was the low base rate of suicidal behavior. RDoC-like approaches that incorporate social and environmental factors in population level interventions, while also considering predisposing factors such as genetic variation that may alter neurodevelopment and psycho-development, could present another way to advance our understanding as to why some individuals experience suicidal thoughts and behaviors. Given the low base rate and the interest in peripheral or genetic biomarkers, the banking of biosamples, including tissue banks for post-mortem molecular studies, is also needed.
Developing Suicide-Relevant Biomarkers
The third session centered on identification of neuroimaging and molecular biomarkers for suicide risk in the central and peripheral nervous system. The epigenetics of suicide behaviors has been investigated in the brain tissue of individuals who have died by suicide and in peripheral tissue of living individuals who have attempted suicide. The SKA2 gene has been identified as one that is biochemically modified (i.e., methylated) and associated with brain volumetric changes, and linked to glucocorticoid and serotonin receptor mechanisms. Neuroimaging research has found that among adults with bipolar disorder and/or major depressive disorder with a history of suicide attempt, volumetric, functional, and structural connectivity neuroimaging data implicate the frontal cortex.
The bio-behavioral mechanisms that predict a transition from non-suicidal self-harm to suicide would be particularly helpful in preventing youth at risk from transitioning to ongoing risk states, yet we don’t know much about the neural and physiological systems contributing to this risk class. Research from the field of metabolomics, particularly research focused on metabolic diseases resulting from inborn errors of metabolism that have psychiatric manifestations, could provide insight into mechanisms that confer risk for suicidality. For example, cerebral folate deficiency can produce treatment resistant depression, and treatment with folinic acid alleviates depressive symptoms. This suggests that metabolomics could be used as a screening tool for more severe cases, and may provide a path to mechanism discovery. Suicidal behavior is often preceded by an acute stressor and among some subgroups, the greater the stressor, the greater the risk for suicide. Understanding the association between peripheral inflammation, biomarkers in the hypothalamic-pituitary-adrenal axis, and changes in the central nervous system of people with suicide history, could better explain the course of risk.
Lifespan Developmental Trajectories, Substance Use, and Other Special Issues for RDoC Suicide Research
In Session 4, workshop participants turned their attention to individual variability and subgroups of risk. Participants discussed the multiple ways that acute alcohol use can increase imminent risk for suicide and identify which individuals are most at risk for suicide and identify times that are particularly high risk. For example, acute alcohol use may place individuals in an aversive state with high sensitivity to negative valence and rumination. At the same time, alcohol use may impair cognitive control and attention systems, making it difficult to inhibit impulses to act or shift attention away from negative stimuli.
Non-suicidal self-injury (NSSI), occurs in 9-16% of adolescents and incidence rates have been increasing over the last two decades. NSSI is of particular interest because it is associated with increased risk for a suicide attempt. More longitudinal studies that examine NSSI, pain-associated dopamine reward signaling, and affect regulation, could contribute to the identification of common and distinct predictors of these risky behaviors. Workshop participants encouraged the inclusion of NSSI measures in all research studies of suicide, and identified a need to determine how NSSI can be measured dimensionally. Data was presented that demonstrated stress reactivity as a central integrative construct that is predictive of suicidal ideation. This data focused on the threat and reward constructs and showed that adolescents at increased risk for suicidality have difficulties regulating stress response as well as blunted neural responses to rewards. The group recommended attending to gender differences and the timing of early life stress as strong influences in the trajectory of suicide risk over the life course. In late-life suicide research, computational modeling has been used to understand suicide ideation and behaviors, based on a diathesis-stress model of interacting predisposition vulnerabilities with stress from life experiences. One advantage of computational neuroscience is the potential to link many levels of analysis. Studies of older adults also reveal strong age by sex interactions with increased rates of suicide deaths particularly in older males who attempt suicide.
Implications for Treating Suicidal Thoughts and Behaviors
The implications of RDoC-driven suicide research for intervention was also discussed. One approach for testing RDoC type constructs is integrating research into psychiatric inpatient hospitalizations. Risk factors such as early life adversity, maltreatment, and DNA methylation profiles could be assessed, and modifiable targets (aspects of cognition, behavior, affect, and social interactions) can be identified and treated in the hospital, as well as through the transition to long-term outpatient care. One key to preventing suicidal actions in individuals with a reactive style is to gather continuous event-prompted assessments, harnessing technology such as cell phones and utilizing ecological momentary assessment (EMA). Improved interventions for maladaptive reactions could possibly pre-empt the need for inpatient care. Other treatments under investigation include ketamine, which is currently being offered off label and shows dramatic, rapid effects on suicide attempt, but its duration of effects and contra-indications are challenges for adoption for long-term suicide prevention. Currently, ketamine’s metabolites are being considered as safer, effective potential novel compounds. Two electrophysiological interventions were also considered at the meeting. Transcranial magnetic stimulation (TMS) aims to modulate electrical signaling in the brain. A randomized, sham-controlled trial of TMS on individuals who had been hospitalized with suicidal ideation produced significant reductions in suicidal ideation. Important future avenues for research include tracking relapse rates, better understanding of optimal stimulation parameters, and potential use in outpatient settings. Modern electroconvulsive therapy (ECT) was also discussed for suicide intervention; more recent advances in administration have reduced side effects while maintaining effectiveness for treatment resistant depression. However, there remains limited research on ECT as a treatment for preventing suicide.
Meeting participants suggested that a number of potential opportunities could be implemented in the near future, including: finding a venue for sharing null results, determination of common data elements (risk and protective factors as well as outcome measures) to provide cross-talk between studies, sharing/pooling data through the NIMH Data Archive, leverage existing studies to include suicide measures, and collection of biosamples. Increased focus on specific research areas were suggested and include early life adversity and its timing, culture, characterization of NSSI, alcohol use as imminent risk factor, access of care, and combination of treatment interventions. A significant gap identified in the current research is on social connectedness, and how can we provide the connectedness that past researchers provided one-on-one with research and clinical interactions in a more cost efficient way that leverages technology.
Medium-term goals include increased research on some under-represented domains and constructs within the RDoC matrix, investigation of the interactions among domains, more analysis of gender and age, fostering of advanced statistical methods (such as computation modeling of dynamic processes) for multi-level and cross construct integration. A need for increased cross-federal efforts in suicide research was also identified. A parallel set of initiatives was discussed that would aim to leverage existing datasets as well as serve as an incubator for novel prospective studies. The need for collaboration and continued discussion was identified and suggestions included webinars, phone conversations, additional workshop meetings, and meetings at related scientific conferences.
Longer term goals include team science approaches that would leverage current findings, utilize sophisticated machine learning techniques for theory development, and follow an enriched risk group prior to first suicide attempt. Establishment of cross-site training programs was identified as way to facilitate the multi-disciplinary research that needs to occur to advance this field of research. The workshop participants expressed an interest in continuing to share ideas and data in future virtual or in-person meetings.