RDoC Webinar: Suicide Intervention Research and Treatment
>> TIMOTHY FOWLES: Welcome, everyone, to the third Delaware Project RDoC ABCT webinar. Today's webinar will address the topic how can RDoC inform suicide intervention research and treatment. I'm Tim Fowles. I'll be moderating our discussion today, and I'll introduce you to our panelists in a minute, but first I want to tell you a little bit about the Delaware Project vision, and why we're tackling this problem from this particular angle. The Delaware Project is really aimed at lifting the burden of mental illness, and that's felt nowhere more keenly than in the case of suicide. And the primary way that we do that is by translating mechanisms of pathology into active treatment ingredients and integrating intervention science through all of the stages: basic science to intervention development, efficacy trials to effectiveness, all the way to dissemination implementation. And then communicating amongst these different, formerly siloed areas of intervention science to inform the larger picture and help, again, lift the burden of mental illness. And not only are we trying to do this as scientists, but we're trying to train the next generation of intervention scientists to do the same.
Today we have three panelists who will focus on this kind of translational vision in the case of suicide prevention intervention. Our first speaker will be Dr. Stacia Freedman-Hill from National Institute of Mental Health. After which we'll hear from Dr. Matthew Nock from Harvard University, and finally from Dr. David Brent from University of Pittsburgh. Each of these eminent scientists deserves their own lengthy introduction. For the sake of time, I'm going to let their projects and their work speak for themselves. Keep in mind that there is a Q and A feature on the webinar. You'll see that on your screens, and we encourage you to type in questions as we go along. We're not going to take questions on the fly, but we will reserve about 20 to 30 minutes at the end of the webinar to address those questions the best we can. Without further ado, Dr. Freedman-Hill.
>> STACIA FRIEDMAN-HILL: Thank you. So I'm Stacia Freedman-Hill. I'm a program director in the division of translational research at the National Institute of Mental Health. Research on suicide is a priority for NIMH because suicide is a leading cause of death in the United States. In 2017 more than 47,00 Americans died by suicide. Suicide is the 10th leading cause of death overall in the United States. It's the 4th leading cause of death in individuals age 35 through 54, and shockingly it's the 2nd leading cause of death in individuals age 10 through 34.¬†In addition to deaths by suicide, suicide has an impact even greater. 1.4 million adult Americans in 2017 attempted suicide. 3.2 million adults made suicide plans, and more than 10 million adults had serious thoughts about suicide.
More alarming are the trends in suicide rates over time. Since 1990 suicide rates have increased by 33%. Now, as a program officer at NIMH, one of my responsibilities is to identify research which can make significant impact. A typical R01 grant funds five years of research. Well, in the last five years suicide rates have increased by 10%. So it's very sobering to realize that as NIMH staff identifies research for investment, and as investigators conduct that research, that the ground is continuing to shift. So obviously we need to speed the rate of discovery of new interventions and improvements to existing interventions. Given this urgency, NIMH staff thought about ways that research on suicide might be guided by the NIMH research domain criteria framework, or RDoC.
So RDoC is a research framework, and its goal is to identify psychological and biological processes which are the underlying causes of mental health and mental illness. It is organized into six functional domains and it encourages investigation across multiple levels of investigation, for example, from genes and molecules to behavior and self-reports. The RDoC framework also acknowledges the influence of neurodevelopment and the environment. Now it's not set in stone. The RDoC framework is something which is a tool. It's still evolving. An example of that is the sensory-motor domain was just added a few months ago. It's also not intended as a schema for clinical diagnosis or to replace the DSM.
Within the six functional domains, experts have identified from existing research literature constructs that are associated with each domain. And if you drill down further into the RDoC matrix you can see, for example, for the construct of acute threat or fear, that there are units of analysis. These are associated with different levels of investigation. So in a way, you might think of this as a menu that you could use to design experiments, to gain understanding into the neural and psychological mechanisms of mental illness.
One way that RDoC has been used is to gain better schemas for classification of patients. So I'm going to show you-- here's a figure from just one study that we have called BSNIP. This happens to be a study of psychosis. It used a transdiagnostic approach, recruiting subjects who have bipolar disorder or schizophrenia and then employing RDoC. There were objective measurable units of analysis for neurocognition, neuro measures such as EEG or neuroimaging, genomic analysis. By measuring all of these things a classification emerged which is closer tied to the ideology of the disorder. It eliminates some of the heterogeneity that's seen when you classify based on symptom lists. And because it was based on the psychological and biological processes it's closer to understanding the biological basis of disorders.
We think this could be useful for suicide research because we know that people who have suicidal thoughts and behaviors probably represent multiple ideologies and pathways, and RDoC might help us identify biotypes of suicide risk. But we want to go beyond classifying patients defense or understanding mechanisms to actually being able to treat patients by modifying those mechanisms. So RDoC is hand in hand with an emphasis on experimental therapeutics. This is a guideline for discovery of interventions or improvement of interventions. So in traditional clinical trials, subjects are often selected on the basis of clinical symptoms. At the end, we would evaluate whether or not the symptoms change by the intervention. But the problem with that approach is that it doesn't elucidate how the treatment actually works. And it doesn't really give you a blueprint for what you should do next. Instead, a few years ago, we began to emphasize experimental therapeutics and research which is staged.
So in the initial stage, we emphasized target investigation. So the targets which you might call a mediator or a mechanism were proposed from the RDoC framework. They do not necessarily have to be biological. They can be psychological processes. They could be behavioral or cognitive or even interpersonal processes.
The second stage is once we have identified targets that we think are involved in a disorder, we need to show that our proposed intervention engages the target. So will proposed intervention show a change in the target? And can we identify and quantify the parameters for optimal engagement?
And then lastly, in the third stage, we have a clinical trial. Well, now that we have identified a target, we've shown we can engage it and what the optimal parameters are, we're going to test it to find out does it actually change clinical outcomes and is it relevant to the clinical problem. The benefits of using RDoC and experimental therapeutics approach is that it's streamlined, efficient discovery. We learned whether our results are positive, they provide us information and optimal dosing and timing. We also learn from negative findings. It rules out mechanisms. We just don't go down that pathway again. It also is the basis of precision medicine. The understanding that we get through this approach gives us better diagnostic tools and also gives us a better understanding of which interventions work for which patients.
So you might be thinking, this just takes a long time. This can be a slow process. I've already told you that there's a sense of urgency. Don’t we have interventions which we know already work. So that's true. We do in fact have interventions for suicide which are evidence-based. For example, our toolkit could include cognitive behavioral therapy. So previous studies have shown that suicide attempters who engaged in CBT for 18 months had reduced reattempts by 50%. We could also employ dialectical behavior therapy which has also been shown to reduce reattempts. An older form of intervention is caring contacts which follows patients after a release from treatment. In our context mean either through letters or phone calls or emails. In some studies, this has also been shown to be efficacious and particularly low-cost.
So we do have interventions which are effective but these were developed before NIMH emphasized experimental therapeutics. And as a consequence, we may not actually understand what the targets of these interventions are or what the mechanisms of action are. And so what I want to hope to convince you all is that it matters, and there's some reasons why we might to understand the mechanisms of action so for starters, if we understood the mechanisms of action, it would facilitate precision medicine. We can use deep phenotyping to identify for a given patient, "What is the best treatment option?" This may also, understanding the mechanisms may tell us about timing. "When's the best time to intervene? How can we sequence with potentially other treatments as well?" We may find that there are multiple targets, and we may need multiple interventions. And the precision medicine approach can tell us maybe something about how we might combine different therapies. We can also use it to improve existing treatments. The whole process of target identification and showing that we can engage target also gives us objective measures of progress. And that's important because sometimes patients might not be aware of changes, and if our objective measures are maybe potentially more sensitive, it can prevent the situation where we abort an intervention prematurely. It might be working and we just weren't aware yet. With objective measures, we'll know that. The objective measures also allow us to compare different types of interventions and whether we can engage the same targets in a different way. And they also help us understand people who are not responding may be one reason why they're not responding. And then it can also help us to improve delivery. So one potential example of that would be if you could use, for example, physiological monitoring via biosensors. You might know when was the best time to send a caring letter or when a therapist might want to initiate, in the moment, coaching. So I wanted to show how you might go about designing research studies which could test the mechanisms underlying existing treatments. I'm going to focus just on DBT, and I'm going to focus actually on just one type of skill which was worked on in DBT. So distress tolerance is one of sort of four areas that DBT skills training is focused on. So let's consider just distress tolerance and let's look at in the light of the RDoC framework. So I've put up the domains and constructs and identified some things which we might hypothesize are important for distress tolerance. So for example, acute threat or arousal are things which seem reasonably to be involved in distress tolerance, but there are other things such as attention, maybe reward processes. And in blue, I've shown you some other areas which we suspect interact. So social communication, for example, we know that social interactions might be the trigger for distressing situations. I'm going to focus only on acute threat in the interest of time, but David Brent and Matt Nock are going to talk about some of these other RDoC domains in their talk. So if we drill down into acute threat we see a list of measures. And again, we can use this to guide some research on how we might look at the mechanisms by which DBT skills affect change. So things that we could measure from here would be cortisol. We might look at the hypothalamus or the amygdala. We might measure heart rate, and we might want a laboratory task such as the Trier social stress task.
So here I have a figure from a just published review of stress response. And I thought this was a great image that really shows a model of stress-response systems, and it also has a great mapping onto RDoC. So in this model the HPA axis that hypothalamus, pituitary, and adrenal glands are key to stress responses. There's a cascade of hormones, and they act on things such as heart rate, inflammation. And the model also acknowledges that there can be influences from interpersonal stress, there can be distal risk factors; for example, on childhood adversity, and also importantly, the role of neurodevelopment. This is a system where these circuits are changing particularly over adolescence. So if we go back, if we look at-- now we have some targets. We have neuro areas of the brain which we can look at. We have things we can measure, like heart rate or cortisol, and we can also decide whether to do this in a naturalistic environment or to induce stress in a laboratory environment, which will allow us to control some of the experimental factors.
So the Trier social stress task is one that has been used in a lot of studies to look at distress tolerance through lots of variations of this, but they have common elements. Involves a patient or a participant coming into the lab. There's a baseline period of rest instructions. This is followed by a stressful situation. Generally, it's a job talk in front of a panel who can be neutral or offer negative feedback. And it's followed by a recovery period. And the key to this paradigm is that you can take measurements before and after the stressful period. So for example, typically you can collect saliva to measure cortisol. You can measure heart rate, maybe pupil dilation, you can do some cognitive tasks, maybe a Stroop task or a task of cognitive control during this resting period. And then you can repeat all these measures during the recovery period, and so you can see what is the effect of the stressful event. And so here's an example of a study that used the Trier social stress task and collected measures of cortisol from saliva.
What I want to bring your attention to is that the study included healthy controls. So participants whose parents had a mood disorder-- so the healthy controls, their parents did not have a mood disorder. And with the offspring of the parents who had mood disorders, those offspring could either be non-suicidal - they may have a history of a suicide attempt - or they may have had some suicide-related behaviors. For example, they may have made an aborted attempt or had severe ideation which resulted in an ER visit. So the key things here that I want to show you is that in the healthy controls here on the non-suicidal offspring, over here are the points measured before the stressful event. These are the points here that are from the recovery period. You can see that their similar patterns of increases in cortisol during the recovery period. Similarly, if you look at the offspring who had behaviors that-- not attempts. When you look at the recovery period, they look similar to the other two groups. But one thing here is that if you look at the offspring who had made previous attempts, you see that overall there's a reduced level of cortisol during baseline and during the recovery period. So this measure can identify individuals who have made previous suicide attempts, maybe at risk for suicide, and it can even distinguish them from those who maybe have ideation but not have made attempts. Here's a similar study, again, using the Trier social stress task measuring cortisol. This was able here to find a different pattern between individuals who had brief periods of ideation versus those whose ideation tended to be longer and continuous. And then you might be thinking, well, in a typical clinical setting it might be difficult to measure cortisol. So I wanted to be sure to show you that there are other physiological measures which may also be biomarkers. So there's also a study of females who had a history of immune disorder and this looked at heart rate variability. And here you see that individuals who had a previous suicide attempt, there's a difference in their heart rate variability versus non-attempters.
So I want to bring this all together as I start to complete my talk. Let's consider how these translational experiments might tell us something about the mechanisms by which DBT or other interventions work. We consider just distress tolerance and just acute threat, but we identified the first targets. So the cortisol measurements-- we identified a target, which was brain lesions and mechanisms for acute fear. We had units of analysis such as the cortisol measures or the heart rate measures, which in turn allowed us to quantify. And potentially you could combine this with an intervention and see how the intervention affected these units of analysis which are measurable. By measuring these, we could potentially also personalize medicine. So a precision medicine approach would, for example, look at how much distress tolerance was an issue for a particular patient. And if we looked at multiple constructs, we could then start to put together patient profiles to see what are the most important areas or the most pressing areas that we want to work on with our interventions.¬†These biomarkers for the stress response can also be used to track progress on the overall settings, and even identify high-risk states even prior to a patient's awareness that they're in a high-distress state. And they allow us to potentially consider whether combining therapies and quantify the boost you might get from combining pharmacological interventions with DBT or maybe how we want to-- what's the optimal way to sequence learning DBT skills.
So I focused in this talk on how we can use RDoC in an experimental therapeutics approach to understand our existing interventions. But more importantly, we can use RDoC to discover new targets, new ways of engaging targets for novel interventions. So I want to close by letting you know that NIMH has multiple grant mechanisms which are tailored for different stages of experimental therapeutics. And the program staff is happy to help you identify which grant mechanism is the best for the research that you might be doing. And then lastly, I'm hoping that I can show you that the new approaches to science that NIMH has been emphasizing are intended to speed the pipeline from science to service. And hopefully, that by speeding this up we can begin to actually turn the tide and reduce suicide rates. So I want to thank you for listening, and I want to thank my colleagues for helping me put together this talk and for all that they did every day. Thank you.
>> TIMOTHY FOWLES: Thank you very much, Dr. Friedman-Hill. We'll next hear from Dr. Matthew Nock, and I do want to mention that this webinar will be archived and posted online. So you'll have access to it later on.
>> MATTHEW NOCK: All right. Thank you so much. Thanks to everyone for joining. As mentioned, my name is Matthew Nock. I'm a professor of psychology here at Harvard University. And with the brief time that I have, I want to talk a little bit about some ideas about why people kill themselves and, in particular, I want to talk about the potential role of RDoC in the development of better explanatory models. And by better, I simply mean better for our field. Suggesting that our current explanatory models are insufficient and inaccurate as understanding, predicting, and preventing suicidal behavior. So before getting started, I want to briefly thank the-- of course, the funding supported some of the work I'm going to present. Most importantly NIMH because they are here listening. I also want to note I have no conflicts of interest, and I want to acknowledge just a few of the really important people who contributed to some of the work I'm going to share today. So in the brief time that I have, I'd like to talk to you about four things. First, I want to go very briefly through some about suicide, the history of suicide in about one minute, and then talk about where we are currently. I then want to identify some key gaps as I see them in our understanding of suicidal behaviors. I then want to talk about what I see as the role of RDoC in potentially helping to better understand suicidal behavior. And I'll conclude with some ideas about needed direction. So again, in about one minute, I'll fly through early conceptualizations of suicide. The thing I want us to draw out here is that from our earliest models of suicide, or I should say our earliest models of suicides we’re really interested in how really strong psychological explanatory power. If you think back to work done by Durkheim over 100 years ago, he described suicide as resulting from a breakdown in a person's social integration, for instance. We later saw theories about suicide as resulting from anger at another person turned inward leading one to destroy one's own body, suicide resulting from problems of cognitive rigidity, if you can't think of the solution to a problem, you might be more likely to choose suicide as an option. And Shneidman famously in our field talked about suicide as an escape from intense psychological pain. The idea that people who want to die by suicide don't necessarily want death, they're trying to escape from some perceived intolerable circumstance, in the same way that someone might jump out of a window to escape a burning room. So these early models made a lot of sense, I think, help clinicians understand-- and helps others understand how and why these other people may be struggling, although they were admittedly overly simple, and they're focused only on one variable or two variables at a time. With the development of DSM III, as we all know, and the incorporation of the research diagnostic criteria, which are much greater reliability in our diagnostic system, but many argued that that came at the expense of understanding. I think that the implement affected the study of suicide as well and our understanding of suicide. And over the past few decades, we've come to see people, I think, in large part because of studies showing that about 90 to 95% of people who died by suicide have a prior diagnosable mental disorder that has led many people to think about suicide as a complication or a consequence of psychiatric or mental disorders or a specific disorder. It is true that mental disorders have a strong association with suicide, but that doesn't help us really explain how or why people become suicidal or how a lot of people with mental disorders become suicidal. He and others have argued that there's focus-- there's a lot of-- somehow myopic focus on the DSM has created academic blinders that's really impeded our progress. And I'm in another type of depth that this indicates for the understanding of mental disorders. I would argue that this has really limited this focus, and this intense focus on the DSM has limited our understanding of suicidal behaviors more broadly. Some may disagree with that perspective and you are, of course, welcomed to do so. As W. Edward Deming said, In God We Trust, all others must bring data. And so I brought some data to see whether this is the case in our field. And I want to do this briefly by pointing to a recent meta-analysis done by Joe Franklin, an assistant professor at Florida State University. What Joe and friends did was look at every study that could be found trying to predict suicidal outcomes over the past 50 years. And when you look across the past 50 years of research, and you bin studies from those probably different decades, you've been in the first two decades together, what you see is if you look at where the top five types of risk factors examine, it turns out that the top five risk factors in the research literature are the same for the past 50 years. We look at sociodemographics, internalizing and externalizing different symptoms, prior self-injurious often behaviors and major life events. And those top five are the same five over and over and over again, and in fact, in about 75 to 80 percent of all prediction cases, meaning for all analyses done over the past 50 years in the literature, we've been looking at those five types of perspectives. What has the outcome been? If we look across our effect sizes over the past 50 years, and we plot it here. On the Y-axis, on the vertical axis, this is odds ratio. So we would expect higher odds ratio, meaning stronger predictive power – are our odds ratios, increasing from left to right? Are we getting better and better at identifying predicting factors for suicide attempts and suicide death? Sadly, we are not. This is a pretty flat set of bars. If anything, it looks like maybe falling slightly. And this shouldn't be surprising, right? if we're looking at the same predictors, and we're using the same method over and over again, it's perhaps not surprising that we're getting the same results. So we need to do things a little bit differently.
So this is where we're currently. Not a glorious state of affairs, but it's where we are in the study of suicidal behaviors. So one way forward might be to think about what are the gaps. This is one quick slide. I apologize it's quite busy. I'll walk through it very quickly just to highlight what I see are some of the key gaps here. The first is we need novel risk factors, as I just described. The next four key gaps, as I see them, are we need to better understand what factor is with the transition from thought to action. Only about a third of people who think about suicide actually make an attempt. And most of the risk factors that we've identified are actually predictive of suicidal thinking, but not of who makes the transition from thought to action. We need to get much better at predicting that transition. We also need to better understand how we combine information about different risk factors. That metanalysis that I mentioned shows that the vast majority of studies look at one risk factor at a time. We need more complex models. We also need a better understanding across levels of analysis. Many of us focus on subjective mental states, looking at self-reports or questionnaires or interviews. Others look at physiological responses. Others - brain circuitry. Others - genetic association. These are all important, but we need to start working across these different levels to put together a deeper picture of why people think about suicide, and why they make suicide attempts. And for predictive and clinical purposes, we do need objective markers of risk, and we need to better understand short-term prediction of risk. The vast majority of our research has a prediction period of a year or longer. We need to have a better understanding of how or why or when a person is going to transition to a suicide attempt short-term, which clinically is what we really, really need.
What framework could possibly help us address these different gaps? What I would argue is that the RDoC framework that they Stacia just did a really wonderful job walking through and describing is really well-suited to address many of these gaps. You are - if you're tuning in - likely familiar or getting familiar with the Research Domain Criteria, and you now know, if you didn't know, there's a new domain in town, the sensorimotor domain. So we have now six domains here. And the goal in this area of the field, as I see it, is to use these domains we construct to better the occurrence of suicidal thoughts and the transition to suicidal behavior. So how can we do that? How can we impose some structure here? Well, if we think about all the theories of suicide that I described earlier and subsequent ones up until today-- if we look at all these different models, theoretical models, they start to look like the models have some common elements to them. And many models that are more simple have one factor or two factors or three factors. One may even think it of as random selections from the pool of common themes, common themes throughout models of suicide, and this hopefully will resonate with the clinicians in the room, people who are suicidal tend to express that they're in psychological pain, like they have trouble experiencing pleasure they're hopeless or isolated, have poor arousal and regulate their arousal, problems with sleep, problems with agitation. These are pretty big fuzzy constructs that actually map on quite well. If you think about it to constructs within the negative valence system, positive valence system, cognitive systems, taking hopelessness for instance, having negative expectations about the future. There's some really interesting basic science work on prospection. The ability to think about what seems like the future by Dan Schachter, Dan Gilbert and Randy Buckner and others showing up that the ability to prospect, to think about the future relies on memory and flexibly recombining information from a person's path to generate thoughts about the future. Hopelessness may result from problems in these more basic areas for which we have a good underlying understanding. Social isolation reflects perhaps problems in social processes and arousal and hesitation regulation map on pretty cleanly to these other domains. And so if we study RDoC constructs as more basic components that make up some of these pieces of the theory, this might help us get better traction in our understanding. So how do we move forward? How do we create better explanatory models, more effective, more accurate, more predictive, more powerful models. RDoC, I would argue to help us do that by building a framework for hypothesis testing. But more importantly for model building. And through this lens, we might benefit from thinking about suicide as the results of the interaction, of dysfunctions in multiple evolutionarily adaptive domains. With these RDoC constructs representing adaptive constructs, adaptive things a person might do, to survive. But maybe there are extremes or deviations and that can lead to suicidal behavior. And I'll talk more about that in the next slide. And so what I would encourage those who are psychological scientists to do, many of us have a tendency to think about RDoC as an inherently biologically reductionistic system. And it goes right down to genes into cells. I would argue it doesn't need to. If you look at the domain, you look at the constructs under the domain, these reflect what's happening right now in psychological science and basic science and basic psychological science. So we might think of this as a map for better connecting with our basic science colleagues. So what do I mean by this earlier statement? Much of what we do as humans, if not the vast majority of what we do is aimed at adaptive functioning at surviving and thriving in our world. We need things like a strong stress response to prepare us for flight or for taking off. We need to have implicit cognition to reason about and understand that it was about the world. We need to experience psychological pain from an evolutionary perspective. If you're part of a group, you're part of a tribe and your new break away or isolated from that group, some kind of alarm bells should go off in your head saying, "This is uncomfortable. This hurts. Get back into the group. Otherwise, you're going to die. You're going to starve. You're going to be eaten alive by a dinosaur. We should have a tendency to escape from aversive situation. If a room and burning, we should want to jump out of the front. We should mind get out of there. We should be able to think flexibly. We should be able to simulate the future. It can help us survive. We should be able to make effective decisions. We should have a fear response. And in some instances, we should be able to override that for your response. All of these things are as active and we all experience these on a continuum. We all varied. Things aren't dimensional, right? What I'm arguing, what I'm suggesting is if we're too high in some of these dimensions, some of these traits and too low in others of them, this might lead us to have an increased risk of suicidal thinking, of suicide attempts, and suicide death. I'm not proposing that these are *the* eight, or that any one or two or three of these parts that are the right interactions or rather that we think about these constructs, the psychological traits as ones that are connected to basic psychological science, and from which we can draw from and study to better understand suicidal outcomes. For each of these constructs I mentioned, there are well developed and validated objective measures for these constructs. And for some of them, thinking across units of analysis or levels of analysis, we have, in the case of decision making, for instance, connections to an underlying neurocircuitry, which itself is connected to an underlying genetic structure. So we can start to move across into analysis with this approach.
So what might this look like and how can we move forward in this direction? One thing that we can do, and we should do, is look to see well, what is our starting point? What do we know about RDoC constructs and suicide? And I am part of a group with Cathy Glen and Egan Kleinman and Christine Shaw and others that recently received a small amount of funding from NIH, who have generously supported this mission of trying to better understand RDoC and suicide. And one thing that we've been doing is going through the past literature to identify quantitative studies that have measured RDoC constructs - going to cut to the chase, there aren't a lot of them - and quantify the links with suicidal outcomes to ultimately develop a searchable database that people can use to visualize these associations and all scientists can use to study them. As a first step, we took data from the earlier meta-analysis by Joe Franklin and friends that I mentioned and said, "Well, what prospective studies do we have that measure RDoC constructs and predict suicide ideation, attempts, and deaths?" To be brief, I'm going to show you what we found for predictors of attempts.
So these are our five RDoC domains. One thing to notice is the number of studies. This is a small number of studies from the hundreds or thousands that have been done. Only a handful of studies made the list in terms of looking at RDoC constructs. Most of them focused on negative valence and social processes, not many in these other domains. In terms of effect sizes, negative valence has a strong odds ratio, as we expect. The others are a little bit below. But we see a pretty strong signal for arousal and regulatory systems. Small number of studies, but fairly large odds ratio. This is the place outside the lamp post light that we should be looking.
And the conclusion of the paper, I'll just read because it really nicely sums it up, the RDoC framework provides a novel and promising approach to suicide research but relatively few studies of suicidal behavior have fit within this framework. So future studies have to go beyond the usual suspects of the risk factors that I described earlier to understand psychological processes that combine to lead to this deadly outcome. And I want to conclude by talking about how we might do that, by developing and evaluating objective measures of psychological constructs of interest from the RDoC construct list or outside it, as Stacia importantly mentioned. The RDoC framework is a living, breathing, thriving list. And then to develop treatments that target the construct that seem to have important associations with suicidal behaviors.
So I want to give one quick example. I know my time is just about up. I want to focus on the idea of suicide as escape. Shneidman had famously said and I think I really like the quote and his approach, because myself, having interviewed thousands of suicidal people, it resonates really strongly. Shneidman describes suicide as being pushed by pain in every instance he's seen and that suicidal thoughts and actions represent efforts to escape that pain. Well, how do you study escape, other than to ask people who are suicidal or made a recent suicide attempt say the reason they did it, very often, is to escape from intense situations. Well, how do you study that? I'm going to briefly describe some really beautiful work done by Alex Milner who's a research scientist here at Harvard. He's been interested in the study of quantifying escape and what he did was develop a brief behavioral task that measured people's decisions to escape from aversive states. He did this by drawing on a really rich, deep bias literature based on reinforcing learning, and he developed a task in which people are exposed to many trials. They learn from feedback they get. They have to make subsequent choices, and what he did was use computational modeling approach to quantify how people make choices, how they make decisions moving through time, space, and their context. The hypothesis here is that it's normal. It's adaptive to act to escape from an aversive¬†condition. If you were in a burning room, it's normal to act to try to get out of there. But this escape bias might be especially strong in people who are suicidal. My screen went blank. Organizer, am I still on? Do I have you?
>> STACIA FRIEDMAN-HILL: Yes.
>> TIMOTHY FOWLES: You're still on. We'll stay tuned while you work on technical stuff.
>> MATTHEW NOCK: Can you see my screen?
>> TIMOTHY FOWLES: Yeah.
>> MATTHEW NOCK: Okay. Everything disappeared for a moment, and I don't know why. I know especially maybe it's like the Oscars -- cut me off. So in the task, a person has to make decisions to act or not to act, to press a spacebar, or go or not go, and they do this, seeing four types of images. And in some conditions, there's a really aversive noise playing, and the way to shut that noise off is either to go on some trials or to not go, to do nothing. In other instances, there's no noise playing, and if you go, that noise doesn't come on, and in other instances if you don't go, that noise then will come on. And to get through this trial, you know the right thing to do, but you don't learn to go or to not go. What he found is that in conditions when the noise is on, normal-- normal, non-suicidal, people without mental disorder have a tendency to go to escape. If there's aversive noise playing, most will have a tendency to want to push the escape bar, to do something to escape. If there's no noise playing, most of us have a tendency to no-go. You learn fast that if things aren't broke, don't fix it, don't act to escape. He then wondered why suicidal people show this stronger go to escape from an aversive context. So we compared non-suicidal people to suicidal people. The bars is on the right I will ask you to to ignore for now. Those are the same as the previous bars. What I want to highlight is among the people-- sorry, the one on the right with psychiatric control participants, ones of the last psychiatric patients who are also suicidal, those folks show the strong go by. When they're in an aversive context, they have a really strong bias to act, to escape that aversive context, much more so than non-developing psychiatric patients. And for suicidal people, for who they had a neutral state, they have a tendency not to go. They struggle to learn to do something to change that state. So this is good support for the escape theory of suicide and more importantly, it has a strong link to basic science on learning and decision-making computational modeling, providing us traction for working in this area. The next step in this area is to test this escape tendency in situ – out in the world -- with smart phones, to see if peoples’ escape tendencies as it increases their suicidal thoughts, and likely their suicidal behavior increases. And then to develop and test training that targets this escape tendency, and see if we can drive it down. So in conclusion, what we need to do is move beyond tests of DSM disorders, as predictors. Consider the RDoC framework as a way of addressing many of these significant gaps. Develop and test objective measures of common psychological elements, like, escape, decision-making, and then once we quantify these associations, target these with novel interventions to try and drive behaviors down. Thank you so much for your time. I look forward to any questions people might have.
>> TIMOTHY FOWLES: Thank you very much, Dr. Nock. We'll next hear from Dr. David Brent. And again, we are collecting questions via the Q&A so please go ahead and post those and we've reserved time at the end to address those. We've already gotten a few but looking forward to a great discussion. Go ahead Dr. Brent whenever you're ready.
>> DAVID BRENT: Can you see this now?
>> TIMOTHY FOWLES: Yes, yes. We see it. Looks good.
>> DAVID BRENT: All right. Now you can hear me. So I'd like to first thank NIMH for inviting me to participate and for my co-presenters for really elegant talks. I know I learned a lot. And what I'd like to do now is just try to take apart one really central intervention that's used in almost all evidence-based treatments for the prevention of suicidal behavior which is a safety plan. And to talk about some convergent evidence in alteration of social processes related to self and suicidal behavior and then finally to talk about how better clarification of these targets can support the search for more effective interventions. And you can think about safety planning as trying to reset the balance between distress and restraint. And you heard from Matt and also from Stacia about the importance of arousal and agitation and mental pain as a contributor to suicide.
And in the short run, the goal of a safety plan is to help people avoid acting on those experiences. Simply, a safety plan involves looking at what are the triggers for suicidal behavior, figuring out ways to avoid those triggers if possible, and if you can't, you figure out ways to cope with those triggers. And three of the mechanisms that we're going to talk about are distress tolerance - what Stacy also discussed - reviewing reasons for living, and reaching out. And on the right-hand side of the slide, you can see some work of Barbara Stanley that a brief safety planning intervention - a one session intervention - significantly cut the rate of suicide attempts in follow-up in people seen in emergency rooms. So you can see that this element of intervention is effective.
So one of the things that occurred to me as I began to think about this is there's a lot of little different domains and constructs of our job that you have to activate for even the simplest component of intervention. So first, you have to know that you're upset, distressed and that involves a construct under the social process domain. Then you need to access working memory to figure out a strategy to respond. You need to go into cognitive systems to select the right response and hopefully, that response will have an effect on the negative valence system in arousal which in turn, feeds back to the sense of distress. And reasons for living. This is where you think about reasons why you wouldn't want to kill yourself and so again, you would go into working memory. You would then update and consider some different options. And one of the most common reasons that people give for reasons for living are number one: they don't want to hurt their family, which may activate the social process construct related to affiliation and attachment. But they also have moral objections to suicide and may fear some negative consequences and that may actually activate the negative valence system.
And then, in addition, you're having a situation where you're considering reasons why you want to live. There may be things that you're looking forward to which may activate the positive valence system which in turn, together can reduce your suicidal risk. And finally, the issue of reaching out to others. Again, you begin with the sense of distress. You select a response which in this case is to attach to somebody else and that, in turn, provides feedback to the negative valence system and arousal and hopefully reduces suicidal risk.
I'd like to transition to another aspect of RDoC which has to do with the concept of self under the social domain. In specific, something referred to as self-referential thinking. It's the tendency for a person to think about material or an idea in relation to one's self. So, for example, if you ask a suicidal person to think about funeral, they're much more likely to say, "My funeral," or think about their own funeral as compared to somebody who's not suicidal who may think about funerals in the abstract or some historical experience thereof. And there's evidence on multiple levels of an alteration and self-reference and self-association in suicidal behavior and it ranges from the way people use language and in this study by (name of researcher), the very elegant use of natural language processing to identify more common use of self-referential language and people who were moving from groups that discuss mental health issues to groups that discuss suicidal-related issues. And the more self-referential the language, the more likely they were to make that transition. And work that Matt has been involved in is identified associations between self and death that are stronger in suicidal people than they are in non-suicidal people. This is another example of an altered association between self.
And finally, there's a recent study by Alercon that looked at adolescents, some of whom were suicidal or had made suicide attempts, while they were looking at faces that morphed from themselves to somebody else and back. And they found that while they were processing this, that there was a greater association between the amygdala and the anterior cingulate cortex which is one of the centers of self-reference. I'd like to share some work that we're doing where we're looking at individuals who are suicidal and non-suicidal while they're in a fMRI scanner and ask to think about concepts related to suicide and emotions. Marcel Just is the neuroimager, a cognitive psychologist who really has developed and pioneered these techniques. Matt Nock and Christine Cha have helped us to select the actual words and Dana McMakoi and Lisa Pan helped us to develop the protocol. And so in a pilot study, we examined 17 young adults with suicidal ideation and 17 healthy controls, and we have them think about 30 words, each one for about three seconds. 10 related to suicide, 10 related to positive emotions, and 10 related in negative emotions. And then we use machine learning to discriminate between the groups looking at their activation patterns. We wanted to see whether we could discriminate ideators from controls. Within the ideators, we wanted to see whether we could identify which of the people had made a suicide attempt and we wanted to see whether there were distinct emotional signatures associated with those words that discriminated between groups. And this is a list of the words so you can see that, for example, the suicide words are things like death, fatal, funeral, suicide. In fact, we were able to discriminate ideators from controls with a pretty high degree of accuracy. The words were not not just related to suicide although the single strongest discriminator was death. And the areas of the brain that were differentially activated tended to be those that are involved in self-reference. And we also look within the ideated group. So again, this is quite a small sample but we were able to discriminate those ideators who had made an attempt from those who had not. And again, “death” was the leading candidate. And it was a subset of the same brain regions that were differentially activated. And here you can see two suicide-related words and the pattern of activation in two brain areas that are associated with self-concept -- with a greater activation to “death” and to “lifeless” in the attempters compared to the other ideators. And this just shows you where in the brain these different regions were differentially activated.
We also found that the degree to which a person deviated from the average of the healthy controls activation pattern-- that was correlated with the severity of suicidal ideation. And I came across this study that examined people being treated with CBT for depression. And to my knowledge, it's one of the only studies that's directly studied self-referential thinking. And they asked them to consider emotion related words and asked whether they referred to them or not. And what they found was that effective CBT resulted in a decrease in the association between the medial prefrontal cortex and the anterior cingulate cortex which is the opposite of the finding that Alarcon had, showing that greater connectedness in this area was related to suicidal ideation and behavior.
And so in this case, a decline in connectivity related to self-referential thinking was related to a decline in depression. And so what we can see is that there's a very robust relationship between self-association or self-preoccupation and suicidal risk that's really seen at multiple levels, from the way people use language to changes in connectivity and activation on the fMRI to using the implicit association test. And it raises the question about whether most of the therapies that we use indirectly affect this. But we rarely target this directly.
One exception to that is the intervention that's a game developed by Joe Franklin, who at the time was a graduate student of Matt Nock’s, called Therapeutic Evaluative Conditioning, in which the game was to have somebody get points when they paired words related to self to images that were positive and words or pictures related to suicide with pictures that were disgusting. And actually, in three separate clinical trials, they demonstrated that while people were engaging in this game, they showed a dramatic reduction in self-harm, suicidal ideation, and suicidal behavior. And so I think what we can see is that there's really a complex interplay between cognitive social system arousal and negative valence systems as targets and markers of improvement. And I'd like to say that I think the current state of the art research is to try to identify mechanisms of action and therapy that are via a change in connectivity or activation. But it seems to me that the psychotherapy techniques that we're using now involves a sequence of activation of domains and constructs compared to one static target. And therefore, if we want to study these interventions we need something with good temporal resolution. The flip side may be maybe we need interventions that don't rely on activating so many different subsystems. Because if you think about it, if you're 90% effective in eight different subsystems you're going to end up with a 40% response rate. And so I think both for treatment development and for taking apart and understanding the mechanisms of the treatments that we're now doing can be quite useful. And one common theme that I believe we need to think about more is the salience of self-referential thinking particularly with respect to association between self and death or suicide as one promising target. Thank you for your attention. I'd like to thank my collaborators on the study that I presented with Marcel and Matt Vickerly and NIMH for funding us. And for Becca Price, Jay Fournier, and Danny Pine who provided some feedback about these slides. Thank you.
>> TIMOTHY FOWLES: Thank you so much Dr Brent. We're now going to go ahead with questions. And I appreciate everyone who's shared questions. We're going to start. We don't have time to do all of the questions. But we'll try to get as many as we can. And there are some that kind of touch on themes we can do in various ways throughout. So bear with us. And we'll go until 1:45 about. The first question is-- and I'm going to start with you, Dr Friedman-Hill. You suggested that biomarkers could be used to predict whether a person was at risk for suicide or whether a given treatment would be effective for him or her. Can you comment on how group findings permit predictions at the individual level as you've suggested?
>> STACIA FRIEDMAN-HILL: So in precision medicine we are hoping to emphasize risk factors for individual. But that doesn't mean that the tradition of studying groups of subjects is still not informative. So for example, by stratifying and looking at comparisons between groups in our translational research it gives us clues, the between-group differences gives us clues to what we might want to look at in individual study participants or patients. But as Matt Nock pointed out, the key about risk is that it's not continuous or stable. It's a dynamic process. And what we're really interested in is not just identifying patients who are at risk for making a suicide attempt. But we're interested in knowing patients who are at high risk, when is their risk greatest. So we want to understand proximal risk. Not just who's at risk. But when they might be at the most risk. This is actually a great opportunity for me also to point out that inspired by the review analysis from Matt Nock and his colleagues and a workshop that we held a few years ago, we actually did we released this year a funding opportunity to look at the arousal system and how it contributes to proximal risk. And those applications have been received, we've had review and we're looking forward to making some decisions about the funding for those applications. So to just sum up, we're still interested in between-group comparisons but ultimately we need to identify risk processes that over time the temporal dynamics will identify individuals who are at high risk.
>> TIMOTHY FOWLES: Thank you. Dr. Nock, do you have anything to add along those lines meaning the same question?
>> MATTHEW NOCK: Yeah. Sure. I agree with everything Stacia said. There are people more interested in this, there's some great new work being done by Ron Kessler and Rob De Rubeis and others taking data from existing treatment trials and trying to understand who is going to respond best to that treatment and to those treatments. And to give a concrete example that Ron Kessler has used, Stacia talked about caring letters and caring text. People perhaps who feel more isolated from others might respond better to being reached out to in caring text intervention where those who don’t feel isolated and their about suicide are driven by other factors might not respond as well. So as she described precision medicine approaches that are trying to identify what factors in which people present at baseline – when a person presents for treatment -- might tell us which is the best treatment for them. Some concrete examples that are happening in the literature right now. And her second point, as I heard it, I think an equally important one, which is these factors might change over time. So we have to better model risk and risk factors dynamically. Some things are going to vary. Some things like social connectedness are going to vary over time. So it's which interventions work for which people at which time point. And I think with advancement in technology and monitoring we can get better and better at tailoring that precision.
>> TIMOTHY FOWLES: Thank you. Dr. Brent, anything to comment along those lines going from group level to the individual level?
>> DAVID BRENT: Just that the idea with-- we're talking about looking at moderators of treatment response. And once you are able to identify those you can use those to develop a profile for who's going to respond best to what. And as Matt said, as well as when.
>> TIMOTHY FOWLES: Thank you. Okay. Next question. This is a very applied question and I'll start with directing this at you Dr. Nock but others can chime in. This attendee writes that there is a counselor and they've noticed that social media and video games are a big part of their student's lives and this came to mind as Dr. Nock you were talking about escape mechanisms. And the person wondered if these kinds of social media, video games can be a type of escape mechanism and how does how does RDoC relate to this context? Any thoughts?
>> MATTHEW NOCK: Yeah, it's a great question. I think the escape process so the desire, the bias, to try and escape from adverse situations is one that drives a lot of forms of psychopathology. I don't know exactly where or how I would see this in the current RDoC matrix but I do think it's a very important process. In our work so far seems to be an important driver for things like not only suicide but non-suicidal self-injury, eating pathology, alcohol substance use. Again some people had an increased tendency to feel bad. Have these really aversive thoughts and feelings. And the engaging behaviors can alleviate that. Whether it's through cutting, or burning, or drinking, or drugging, or controlling their eating. Or maybe through using social media, or video games, or something else, as a means of escape. I think we have to better understand it, and try and better understand the mechanism. Once we see that it's involved, how is it involved? How is it working? Is it through escape? Is it through creating a connectedness to others? Relieving the bad feelings, generating positive ones? I think a lot of important questions in there that can help give us traction, and I think the way to go, as I've tried to highlight in my talk is, doing some more basic science to understand, well what are the potential factors involved? How do we study this in a more concrete way, and then bring that into the field once you identify some levers and start turning them to see, are there healthier thing we can have people do, rather than cutting and burning, and using social media and video games in a harmful way, to try and drive down aversive states and experiences.
>> TIMOTHY FOWLES: Thank you. Dr. Brent, we'll go to you next. Do you have any thoughts on the same question? By the way, I'll go to each of you in turn, but you don't necessarily have to comment if you don't have anything to add. Dr. Brent, anything on the social media, video game escape question?
>> DAVID BRENT: So I don't know about the escape component. I think one of the problems that has plagued research on social media is that most of its cross sectional. And so it's difficult to tell whether the difficulty in mental health that people have is causing them to use it more, or vice versa. But the longitudinal data is just, such as it is, seems to suggest that a greater use is associated with more psychopathology. And the mechanisms: one is that kids are cyberbullied. Another is that they're using it at a time where it's interfering with their sleep. And then kids engage in social comparison, and presenting a persona that's not accurate. People have found that the more that their true persona deviates from that which they're presenting, there's more difficulties that they have. And interestingly, the area that seems to moderate the impact of social media on kids' mental health is parental monitoring. That the degree to which parents are involved and set limits and maybe help kids interpret what they're experiencing, is protecting against the effects of -- negative effects of -- social media such as we understand them now.
>> TIMOTHY FOWLES: Thank you. Dr. Friedman-Hill. Anything to add?
>> STACIA FRIEDMAN-HILL: Sure, I want to actually get back to something Matt said about escape. So just thinking off the top of my head about the RDoC matrix, I think some of the constructs that could help us unpack what's involved in escape would be in the popular valence, for example, reinforcement learning. Understanding reward valuation. Additionally, for behaviors, actually, like cutting, there's a component of habit learning. So the new sensorimotor domain may be helpful as well. I also think that escape really underscores the fact that these domains, and these constructs really interact. So again, distress, tolerance, stress responses, interact with reward and reinforcement learning, and habit learning. And so it's really necessary to use, look at multiple constructs to understand and unpack these phenomena.
Thank you. We probably only have time for one more question. We got a lot of great questions and it's hard to choose. But let's field this one. And then, again, the webinar will be archived. So you can watch it offline. This last question deals with the translation from identifying through screenings to treatment. And the question is: If we have biomarkers or we have other means like you've all discussed to identify people who are at higher risk for suicide in ways other than their self-report, what happens if you get people who are identified at high-risk but aren't reporting any ideation? What do you think we'd do in that case? And Dr. Nock, why don't you-- would you mind starting and giving a shot at that question?
>> MATTHEW NOCK: I'd be happy to start, but I'm not going to have a good answer because we really don't know, in my mind. We're actually doing a study right now to try and find this out. It's a project I'm doing with Jordan Smoller and Ben Reis at Harvard Medical School and others. Basically, in this, we're taking information from people's electronic health records and using machine learning algorithms to identify people at risk for suicide items. And I can identify right now about a third to 40% of attempters a few years before they make their attempt. Great. It turns out that we have a huge percentage of those folks who are false positives. They never go on to make a suicide attempt in the time that they are monitored. So it brings up a really ethical challenge. Or a sort of: What do you do? Do you tell someone that risk when really they might not be? And it might be that they are not yet thinking about suicide but they are going to. So in that case you want to intervene, I think. It might be that they're never going to think about suicide. And so we wonder, might there be some kind of iatrogenic or harmful effect to telling someone that specifically they're at risk for suicide when really they're not? And so we're doing studies now to try and figure out exactly-- to take a more empirical approach to figure out what things can we do and what is the outcome of the things that we tried in terms of informing clinicians about this risk, informing patients themselves. So hopefully in the next few years, we'll have a better answer.
>> TIMOTHY FOWLES: Okay. Thank you. Dr. Friedman-Hill, do you want to take a stab at that question?
>> STACIA FRIEDMAN-HILL: So I think that's a really interesting question. And, like Dr. Nock, I think I would need more time for a very thoughtful answer. But some initial thoughts are that in mental health care we are usually directing care at those that are help-seeking. But one thing that NIMH has recognized is that there are opportunities to ask questions and potentially identify individuals who may be seen in a medical setting that wasn't related to a suicide attempt. So we have research studies and protocols now for questionnaires that can be administered in the emergency room, not just for individuals who may have come in because of a suicide attempt. We also need to recognize that there are suicide attempters, some of them are already being seen. They're already in the healthcare system. So we have some efforts that are mining electronic health records to also identify risk factors. So we do have opportunities. And now we do have some tools for surveys that we can use that maybe will help identify individuals who might not necessarily before be help-seeking.
>> TIMOTHY FOWLES: Sounds great. Dr. Brent, you get the last word? Anything to add on this question?
>> DAVID BRENT: Well, I think we're working on a paper now that may give us some answers –- that Matt's also involved in. I'm working with Cheryl King and others in a large screening project in pediatric emergency rooms. And one of the things that we did is we administered the Implicit Association Test to around 2,000 adolescents that we have follow-up on. And the paper we're temp calling Implicitly Suicidal. Usually, my titles don't really get to the final stage. The editors don't seem to have much of a sense of humor. But what we want to do is identify-- there's going to be a subgroup of people that screen positive on that IAT but deny suicidal ideation. And so we want to know, well, what are the characteristics of those people? And what happens to them compared to people who are discordant in the other way or concordant? And so, I would say, within the next few weeks we should have some interesting information and some interesting clues about exactly what those people are like.
>> TIMOTHY FOWLES: Thank you. We are, unfortunately, out of time. But I want to thank, again, our panelists: Dr. David Brent, Dr. Stacia Friedman-Hill, and Dr. Matthew Nock. As I said, this webinar will be archived later on. So you can check back on the NIMH website. Also, we're very excited to keep this conversation going. One of our goals with the Delaware Project is to facilitate communication, again, in the interest of integrated intervention science. And I think this has been a huge success in that regard. So thank you, panelists. Thank you, audience. We'll see you next time.
>> STACIA FRIEDMAN-HILL: Thank you.
>> DAVID BRENT: Thank you.
>> DAVID BRENT: Bye-bye.
>> STACIA FRIEDMAN-HILL: Bye-bye.
>> MATTHEW NOCK: Thanks