Skip to main content

Transforming the understanding
and treatment of mental illnesses.

Celebrating 75 Years! Learn More >>

Sleep and Suicide Prevention: Advancing Innovation and Intervention Opportunity-Day 1


DAVID LEITMAN: We currently have 1,000 registered attendees representing 47 countries across 6 continents. For those who could not attend all or part of this workshop, this meeting is being recorded. The transcript and summary as well as the actual recordings, will be available online at the NIMH events website as soon as possible.

Our schedule for the workshop is as follows. We will begin with some brief remarks from Dr. Joshua Gordon, the Director of NIMH and an introductory presentation on state of the science in sleep medicine and suicide prevention by my co-chair, Dr. Rebecca Bernert.

After these remarks, today’s agenda will consist of three sessions, each containing several 15-minute presentations, and ending with 20 minutes for a question-and-answer period. Participants will be muted, in listen-only mode and cameras will be turned off. Those of you listening are welcome to post your questions into the Zoom Q&A dialogue function. We will address as many of these questions as time permits during the session Q&A.

If you have technical difficulties hearing or viewing the webinar, please note these in the Q&A box and our technicians will work to fix the problem. Alternatively, you can email your problem to the email address on the screen currently.

Before I turn things over, I would like to thank my co-chair, Rebecca Bernert, for her herculean effort she has undertaken in conceiving this workshop during a very challenging period. In paraphrasing Hillary Clinton, I would like to say that it takes a village to raise a workshop. I would like to thank the members of the workshop committee: Laura Rowland, Clare Stevens, Aleksandra Vicentic and Andrea Goldstein-Piekarski. I would also like to thank Jane Pearson for her help and from the Division of Translational Research, Sarah Morris and Holly Lisanby, for their support for the workshop. Finally, I would like to the take time and thank Jonelle Duke, TaRaena Yates, Kayla Baker, and Andrew Nawrot, who have produced today’s workshop.

I will now turn this meeting over to Dr. Joshua Gordon, the director of the National Institute of Mental Health for some brief introductory remarks.

JOSHUA GORDON: Welcome. I am Joshua Gordon, director of the National Institute of Mental Health. It is really my pleasure to open this workshop on the Mechanisms of Sleep and Suicide Prevention.

First of all, I want to thank the chairs and the Planning Committee for organizing this excellent workshop. David Leitman and Rebecca Bernert chaired the organization of the workshop and Laura Rowland, Clare Stevens, Alexandra Vicentic were on the Planning Committee. I also want to thank support staff Andrew Nawrot, Jonelle Duke and TaRaena Yates.

Now, you might ask why we are having this workshop. Part of it is pretty obvious. Suicide has consistently been the tenth leading cause of death in the United States and there has been a steady increase in suicide rates for pretty much most of the past two decades.

The National Institute of Mental Health is committed to bending the curve of suicide in the US. Together with the National Action Alliance for Suicide Prevention, we pledged a few years ago to reduce the suicide rate by 20 percent by 2025.

Now, the latest figures from 2019 and early data from 2020 suggest that we might be doing just that. The age-adjusted suicide rate dropped ever so slightly from 2018 to 2019. But this still, this decrease of 20 percent we are trying to achieve – this remains an aspirational goal. That goal has guided our suicide prevention research agenda for the past five years, which has emphasized risk detection, screening, and intervention, particularly in health care settings.

What is the link to sleep? Though suicide is rarely caused by a single factor, we know that we need to take every opportunity we can to reduce risk factors where we find them. Abnormal sleep or sleep-related behavior has been found to be consistently associated with suicide risk. For example, the relative risk for suicide associated with sleep disturbances is approximately two to three-fold.

Sleep is also tightly connected with mental illness as a symptom, as a risk factor, and as a potential moderator. Understanding the role of sleep in suicide risk represents both the challenge and an opportunity.

The challenge is in understanding the nature of the link between sleep and suicide. What are the proximal or distal psychological or biological mechanisms by which sleep elevates suicide risk? Which aspects of sleep confer elevated risk for suicide?

Finally, might sleep be a mediating factor in disparities and risk for suicide?

These challenges need to be answered and they will be the topic of several talks over the next few hours.

The opportunity in sleep and suicide lies in the potential to harness sleep to reduce risk. Here, we need to ask what are the promising therapeutic interventions for the modulation of sleep and do they reduce suicide risk? This opportunity cannot be understated. We know that sleep can modulate mental illness symptoms and we hope that modulating sleep can also reduce this risk for suicide.

These questions and more will be asked throughout the course of the workshop. It will be my pleasure to follow along through the talks and the discussions that ensue. It will inform our approach to these questions from a research perspective in the future.

Thank you for joining us. Thank you to all the speakers. Thank you again to the organizers and I am really looking forward to this workshop.

Agenda Item: State of the Science in Sleep Medicine and Suicide Prevention

REBECCA BERNERT: Hello everyone and welcome. I would like to thank Dr. Gordon and the NIMH for the opportunity to organize this incredible workshop here with you all today and tomorrow and to my co-chair, Dr. David Leitman.

My name is Rebecca Bernert. I am a co-chair of this conference and our workshop today with, as I said, Dr. David Leitman. I could not be more honored to present the state of the science overview for you today on what is now being recognized as an emerging sub-field in suicide prevention.

My hope is to introduce our mission and workshop speakers who are all world-renowned experts in the fields of either suicide or sleep fields, suicide prevention or sleep fields.

I do want to take just a moment to recognize the extraordinary nature of our times during the current COVID crisis and the unprecedented circumstances we continue to face and in being here. Thank you.

We invite your engagement across the next two days. Please do send questions and we hope to join together to advance understanding in this promising area.

I’m going to begin with a brief introduction to the field of suicidology and inherent challenges in suicide prevention, highlighting a few areas of innovation to frame our discussion. I will provide some history of where we began, early studies of poor sleep, proposed – and potential impact to bridge these areas.

I will end with a spotlight on proposed explanatory pathways at the intersection of risk and findings marking its unique promise toward innovation.

As we know, suicide represents a preventable public health problem and really a global disease burden and accounts for nearly 800,000 deaths annually or approximately one life lost every 40 seconds. This generates a cumulative death toll higher than all homicides and war-related deaths combined.

The Institute of Medicine additionally estimates that 25 suicide attempts or up to 200 for youth occur for every death by suicide. And finally, beyond the inestimable cost at the individual community and societal levels, suicide currently represents an economic cost of over 93 and a half billion dollars adjusting for underreporting, underscoring its significance as a public health emergency.

Regarding nomenclature, suicide is considered the tragic outcome of medical illness and a dynamic interplay of multiple risk factors and diverse conditions that interact to carry risk.

Importantly, suicidal behaviors exist on a vast continuum of risk, ranging in severity from suicidal ideation to suicide attempts to death by suicide and an array of symptoms in between.

We will be discussing CDC-defined self-directed violence today, which is any behavior that deliberately results in self-injury and which distinguishes non-suicidal self-injury without suicidal intent from behaviors where suicidal intent is present.

Regarding worldwide prevalence, we say the highest rates of suicide in Eastern Europe with the largest numbers occurring in Asia, accounting for upwards of 60 percent of all suicide deaths.

We also see regional differences in this country with considerably higher rates in mountain states where rural status and reduced access to care and higher gun ownership coincide.

Due to time limits, I will not be able to review epidemiology and prevention, but I do want to emphasize the way that suicide rates vary dramatically according to method and demographic factors, and draw specific attention to the measurement of suicidal symptoms where we have screening tools to examine the lethality of suicide attempts, for example, which is among the strongest known risk factors for suicide and for the measurement of suicidal ideation symptoms - where intense, pervasive, uncontrollable symptoms with a high degree of intent and planning differentially predict risk in comparison with suicidal desire and ideation symptom dimensions.

Importantly, the World Health Organization suggest the following public health model for suicide prevention, starting with surveillance of the problem to identify risk and protective factors and to develop and test interventions for implementation as well as the scaling up of those programs and policies.

Unfortunately, there are inherent challenges that impede advancements in each of these four areas, many of which may not be necessarily obvious or intuitive.

I have highlighted these in a table for our review here briefly. This includes probability challenges insofar as suicide occurs rarely in the population to guide prediction without large-scale data capture, which is further challenged by intractable and even increasing rates.

Ethical issues include confidentiality and liability concerns, which may limit accurate detection of risk where a lack of uniform nomenclature reduces comparability and research despite emphasis by both the CDC and FDA.

Given distinct methodological challenges too, surveillance demands and infrastructural need that is truly substantial and extremely cost prohibitive be it for high-risk clinical trials or epidemiological studies in suicide. To this end, very few indicated treatments exist worldwide for suicide in comparison with overwhelming public health need.

To address such barriers, suicide prevention strategies prioritize study of evidence-based risk factors in interventions to advance what is known as a field and factors that are of course visible, proximal, modifiable, and that stand alone to predict risk - that is - above and beyond the influence of depression, which we know is present in over 90 percent of cases, takes obvious precedence here.

Preventive frameworks, which include the following platforms, such as primary interventions - These focus on prevention of suicide incidence whereas secondary frameworks, which may occur at any stage of risk, generally focus on early detection to aid intervention. Selective interventions test the use of new treatments among those at elevated suicide risk such as high-risk clinical trials, whereas universal approaches aim to reduce risk at the population level.

I would like to briefly highlight some emerging findings that center on addressing critical gaps in suicidology, reviewing a few examples that may be relevant to our discussion today. We’ll focus on lethal means restriction, harnessing technology, and treatment innovation.

First, lethal means restriction is a universal strategy that limits access to lethal means for making a suicide attempt to prevent its occurrence in the general population. It is among the most potent approaches in suicide prevention globally that also however remains underutilized and poorly understood as a public health strategy. It encompasses a wide range of methods, including carbon monoxide emission controls, blister packaging – and safe storage protocols and suicide deterrent systems. Essentially, if you place time and space between the thought and access to means to act on that thought, you prevent death.

The research comes here from the observation largely of what is called a reduction to zero finding at numerous international hotspots, particularly bridge sites following implementation of a suicide deterrent system. These effects are both large and immediate, and evidence does not support suicides occurring at other locations or by other methods. Likewise, research shows suicides increase upon barrier removal, which again reduce to zero following re-installation.

I wanted to highlight just a few findings here, including some cost effectiveness research, evaluating suicide barriers and other prevention strategies on bridge sites and lethal means counseling in either high-risk settings or samples where clinical trials are showing strong efficacy in the adoption of firearm safety practices, for example, among military personnel where gun ownership tends to be high.

We also have seen developments in the use of advanced technologies, including risk detection, training, and health care systems. This includes artificial intelligence to guide suicide risk prediction, which historically shows relatively poor sensitivity and specificity in the prediction of suicide risk.

By comparison, recent machine-learning models use electronic health records or user data and those are showing remarkable accuracy.

Regarding training, we also see calls for universal screening and mandated trainings and risk assessment, which we know enhance risk detection, motivating a number of states or health care systems to standardize these practices.

A few other examples include social media, crisis service use of real-time suicide prevention tools to enhance triage and learning health care systems to impact prevention at the population level.

Regarding treatment, as I said, few indicated interventions exist worldwide for suicidal behaviors and suicide prevention specifically. And especially problematic is our historical reliance on a treatment-by-proxy approach, where we treat the underlying psychiatric disorder and test whether it improves suicide risk downstream of intervention. This is despite elevated need, where half will refuse treatment following a suicide attempt and of those who do pursue it, 75 percent will drop out within the year.

We also see underuse of psychosocial interventions with known efficacy, creating urgency to develop novel therapeutic agents or approaches for anti-suicidal symptom response.

The selection here includes novel investigation of ketamine, rTMS, and those reducing help-seeking stigma, as well as clinical trials testing novel use of sleep-focused treatments for suicidal behaviors, which myself and my colleagues will be speaking about later in the workshop.

This provides introduction to sleep as a risk factor for suicide and how we came to be here. Despite sleep complaints being listed among the top ten warning signs of suicide by SAMHSA for some time, the relationship had not begun to be systematically investigated until 2005, where we were able to show in an outpatient sample that sleep disturbances, even controlling for depression severity, continue to predict elevated risk for suicidal ideation based on self-reported insomnia and nightmare symptoms.

I found this remarkable and proposed to replicate these findings, using more rigorous methodology, where we also proposed a study of mechanisms potentially underlying the relationship here. Namely, mood regulation given longstanding relationships between it and suicidal behaviors as well as known effects on sleep and mood.

We hypothesized at this time that poor sleep may perhaps fail to provide an emotional refuge for distressed individuals and as a barometer of our well-being, initiating a cascade of neurobiological or cognitive changes and thereby lowering the threshold for suicidal behaviors.

However, we were concerned about a host of methodological problems, including retrospective analyses and single time point designs, inadequate assessments, but in particular, a failure to control for the confounding influence of depression severity.

A critical step given that disturbed sleep and suicidal symptoms are each diagnostic criteria for depression, underlying the need to evaluate poor sleep as a valid stand-alone risk factor versus a mere corelate of greater depression severity.

I was fortunate to build out a program of subsequent investigations, supported by early NIH funding to address these issues systematically across diverse populations, designs, measurements, and suicide outcomes.

I am going to highlight just a few of these to show our approach here with the first being a study of poor sleep quality as a longitudinal risk factor for death by suicide among a large sample. This was drawn from over 14,000 community elders, where even controlling for baseline depressive symptoms, poor subjective sleep quality predicted increased risk for death by suicide ten years later.

Up until that time, there had not been a population-level study of poor sleep or sleep quality, in particular, predicting risk for suicide death independent or above and beyond the influence of depression.

Next was an objectively-assessed study of sleep as an acute indicator of risk. This was among a longitudinal investigation of high-risk young adults, using actigraphy sleep data over a very brief window of time - so days, weeks - and among an incredibly high-risk sample. These were youth with either a suicide attempt history or multiple suicide attempts by only age 19.

Using a multi-method approach, we saw that sleep or the findings revealed that – we saw it severely delayed, restricted, and highly variable sleep overall in this group, where objectively assessed sleep variability, that is, variability in the timing of sleep onsets and offsets over this brief period of time and relative to all other actigraphy parameters predicted increases in suicidal symptoms just one and three weeks later. This was even after controlling for baseline depression, as well as suicidal symptoms.

This effect was dramatic, looking more like the sleep of shift workers, for example. And actigraphy aligned with also subjective sleep disturbances to predict risk.

We also found that mood variability, which was evaluated as a proposed mechanism, predicted similar increases in suicidal symptoms and also increased sleep disturbances.

To our knowledge, this was the first simultaneous investigation of objective and subjective sleep indices and as an acute independent indicator of suicidal ideation risk.

Regarding diverse samples in populations, we also found that poor sleep similarly outperformed risk cross sectionally and longitudinally for suicidal ideation and attempts one month later, outperforming more traditional risk factors of depression and hopelessness in a military sample.

We later confirmed these effects, evaluating sleep architecture, conducting a methodologically-focused systematic review building in our work and others here today. And last, evaluating the therapeutic impact of the behavioral insomnia treatment in an uncontrolled study on reducing suicidal ideation post-treatment.

In summary, these findings supported sleep as an independent risk factor across diverse populations, designs, measurement techniques, and outcomes of risk. And we also found that modification, importantly, of this risk factor therapeutically impacted risk.

In our case, this provided strong rationale to test the use of sleep-focused treatments for suicidal behaviors based on funding we were fortunate to receive from both NIH and DoD sponsors.

This was also informed by clinical rationale, where unlike many other suicide risk factors, such as a past suicide attempt, for example - Disturbed sleep is uniquely modifiable, visible in the weeks and months preceding death (a key finding from our colleagues in 2008), non-stigmatizing, and highly treatable. This represents a low-risk strategy for prevention of a high-risk outcome, where the brevity of treatment highlights additional promise as a rapid action intervention and where existing treatments remain either unacceptable to patients based on attrition or inaccessible to those in need.

Finally, the transdiagnostic nature of both sleep and suicide, which each cut across psychiatric conditions, suggest a shared a neurobiology - promising insight into both the pathogenesis of risk and underlying mechanisms that we hope to understand better today.

Regarding military rationale, suicide rates have surged in recent years across service branches - And stigma is well-documented in the military as a barrier to health care utilization.

By comparison, sleep problems are overrepresented among military samples, with alarming rates up to 100 percent among veterans, for example, which also predict poor prognostic outcomes for these disorders. This has resulted in a number of initiatives or work with military agencies over the past decade, focusing, for example, on clinical practice guidelines and other calls regarding its utility as a screening or treatment target.

In our case, and our colleagues, this led to treatment development of several suicide prevention clinical trials, testing efficacy of different behavioral and pharmacological insomnia treatments for suicidal behaviors, including mechanisms-focused clinical trials for analysis of risk alongside efficacy testing, which I will be talking about more in depth tomorrow.

And a long list of exciting projects that we will be hearing more about today and tomorrow across the workshop. This includes ketamine treatments, zero burden monitoring or proximal risk across age groups and high-risk samples, artificial intelligence to guide clinical frameworks and new technologies, treatments to reduce stigma or stigma toward treatment seeking, inflammatory profiles, use of social media as well as learning health care models for population-level prevention. All highlighting exciting areas that we will be hearing more about throughout the workshop.

A number of theoretical frameworks have been proposed to guide discussion today, ranging from neurobiological to interpersonal theories to resiliency and self-healing models. I look forward to discussing all of these different pathways according to the research that is presented with our colleagues here today.

In closing, we know that inherent challenges in suicide prevention require multi-disciplinary collaborations to advance suicide prevention science with major public health impact.

Over the next two days, our workshop aims to share cutting-edge findings across fields to identify underlying mechanisms and advancements and produce what I hope may be a foundational roadmap to bridge best practices with innovation and perhaps produce a white paper or a document regarding a research agenda.

Importantly, we know that intractable, increasing suicide rates highlights critical need for integrative strategy, where any singular approach will be insufficient to prevent suicide on a grand scale.

Because, suicide is preventable, but remains a silent killer, and yet even one suicide - as we know - is too many, if that individual is our loved one.

I would like to thank the incredible support of our funders, sponsors, collaborators, and all of our speakers here within the workshop and thank you for your attention.

Now, I would like to introduce our next session and introduce our session chair for the panel, Dr. Kimberly Van Orden from the University of Rochester.

Session 1: Lifespan, Timing Mechanisms, Risk and Resiliency

KIMBERLY VAN ORDEN: Great. Thank you so much, Dr. Bernert, for inviting me. I am delighted to be here today. Our first session is talking about lifespan, timing mechanisms, risk, and resiliency. We will have several speakers, starting with Dr. Alex Crosby from the CDC, followed by Brant Hasler from the University of Pittsburgh, Helen Burgess from the University of Michigan, and Dr. Peter Franzen from the University of Pittsburgh. I will also be speaking I believe third on late-life suicide.

I am delighted to turn the virtual mike over to Dr. Alex Crosby, who will kick off this session.

Agenda Item: Health Disparities in Suicide Prevention, Risk, and Resiliency

ALEX CROSBY: Good afternoon to all of you. I have heard that there are a number of folks that are from around the world. It may not be afternoon where you are. Some of you it may be still good morning. It may be good evening. It may be good night wherever you are. Thank you so much for joining us today.

I wanted to thank the organizers for the invitation to come and to talk with you today and to be able to share a little bit about what is going on and the planners also for helping to put all of this together and thank you very much for being able to come and to present a little bit about what I will be saying today.

I am going to talk about the epidemiology of health disparities in suicidal behavior and maybe a little bit about prevention. It is not a surprise to any of you that the burden of suicidal behavior affects certain communities more than others. That you do not see an even distribution of suicidal behavior whether you are talking about suicidal thoughts or whether you are talking about non-fatal or fatal suicidal behavior evenly distributed across a population. It makes it incumbent upon us to try to take a look at how this distribution and disparities, how they manifest themselves because if we are moving towards the idea of trying to reduce suicides by 20 percent by 2025, we really have to be able to see how these patterns affect suicidal behavior across different populations so that we can try to affect that and try to change what is going on.

What I will do today is I will try to talk a little bit about why addressing health disparities is important, a little bit about the public health approach and how that fits into health disparities and how health disparities follow along that route and then tell you a little bit about what those health disparities –

Again, thank you very much for the opportunity to come and to share this with you. What I will do is I will address the rationale for correcting health disparities, describe that public health approach, and then show how the public health approach can help us address some of those disparities.

Why address health disparities at all? Just utilizing kind of four different rationales for why to address disparities that inequalities or disparities are unjust. There are ways that when we try to take a look at this that we should be able to address some of these things and try to adjust for those things that disproportionately affect certain populations. Inequities while it might seem like you are talking about a specific community or a specific demographic or a specific segment of the population, they really affect everyone. It does make sense for us to try to address disparities – health of our whole population better. Inequalities are avoidable. There are ways that we can take a look at them and try to decrease those disparities and then interventions to reduce inequalities are cost effective. There are ways in which we can try to address those.

Here is a public health approach. I am going to depict it kind of as four different steps and talk about these four different steps a little bit just to kind of give you the description. But actually, when I talk about disparities, I really will be focusing primarily on the green box, assessing that problem and give you the who, what, when, where of any particular issue, defining what it is, and then addressing a little bit about the trends, understanding a little bit more about the problem.

The next part of the public health approach is really identifying the causes and trying to look at the risk factors, the protective factors, looking at the etiologic agents, answering the why question, and then moving on to the blue is developing the programs and the policies and talk a little bit about that today, trying to identify what works and what does not, and then disseminating and implementation, trying to spread the word about what we know about the particular problem. How do you understand what you have talked about in terms of the first three steps and then how do you implement programs that you may have tried out in a pilot phase? We want to answer some of the questions and try to use this kind of scientific understanding to try to move along in regard to the problem.

Just a little bit about giving you a description of what it looks like in regard to suicidal behavior. One thing that I also want to mention is you will probably see a few – I do not know if I want to call it exactly repeats, but you will see a few of this kind of information coming back with some of the different presenters and hopefully, it will reinforce some of the things that we are saying just in terms of whether it is associated with a logical model or some of the trends in regards to suicide. You may see that a couple of different times here.

Suicide was the tenth leading cause of death, I think, as Dr. Gordon mentioned. 2019 is the most recent that we have got for the whole United States. Over 47,000 people died as a result of suicide during that year, which averaged out to about 1 suicide about every 11 or 12 minutes in the United States.

But even though it was overall the tenth leading cause of death, when you start to look at the different age groups, for example, you see that suicide disproportionately affects young people. That from the tenth leading cause of death overall, when you look here at those who are 10 to 14, 15 to 19, those in their 20s, those in their 30s, it is the second leading cause of death for those particular populations. When you look at those in their 40s, the fourth leading cause of death. Already you can start to see how there are disparities just in regard to age groups and suicide.

Another way of trying to take a look at that is racial and ethnic differences in regard to suicide. And here, you can see, again, one of the points that suicide has been increasing over about the past 20 years or so that you can see looking at the left side that suicidal behavior and deaths due to suicide among different racial and ethnic groups in the United States comparing 1999 with 2019 when you look at non-Hispanic whites, non-Hispanic blacks, Hispanics, non-Hispanic Asian Pacific Islanders, and non-Hispanic American Indians and Alaskan Natives that race increase across all of those different groups. But you can also see that rates are not the same especially in regard to looking at the furthest right among the males that the non-Hispanic American Indians and Alaskan Natives have the highest rates among those particular racial and ethnic groups in 2019 shift over to looking at the bar charts on the right side that you can see again, non-Hispanic whites, non-Hispanic blacks, Hispanic, Asian Pacific Islanders, and then American Indians and Alaskan Natives. Higher rates and variability between each of the different groups and especially American Indians and Alaskan Natives.

Here is looking at some of that similar data in regard to by age group and males versus females. Again, you can see that age groups. Rates are not exactly the same and variability across the different groups from 10 to 14, 15 to 19, those young adult into middle-aged adults, those in older adults by males and by females, rates very different across the different populations and across the different age groups.

Here is a map of the United States. You just saw a picture of it. This is just to show you that there is high variability across the different states in the United States. As you saw and can see in this particularly slide that along the Rocky Mountains tend to have the highest suicide rates in the United States. When we look at this year by year, it tends to be that those states kind of drawing that line from Canada down to Mexico from Montana, Idaho, Wyoming, Nevada, Utah, Colorado, Arizona, New Mexico, that that those states tend to rank in the top 10, top 15 every year in the United States and various reasons why that it is. We will see a little bit later on some of the different observations that have been made regarding those different states.

Another thing is to look at suicide by ethnicity and method and part of what you see here is that there are some very different methods that are accounted for with regard to suicide. When you look at the first group on the left side, black and non-Hispanics, you can see that firearms are the leading method following by suffocation and then by poisoning. Similar kind of pattern although rates very different when you look at white non-Hispanics. When you shift over to Hispanics kind of in the middle group there that you can see that suffocation is actually the leading method in that particular population and then firearms number two among American Indian and Alaskan Natives. You an see that suffocation oftentimes hanging is that method that is much – the leading method there similarly when you look at Asian Pacific Islanders.

Here are the rates when we look at suicide rates by the five leading racial and ethnic groups in the United States and some very different patterns here when we look at what is going on in regard to high suicide rates in particular age groups. You can see from the top line kind of in an aqua color, the American Indian, Alaskan Native. High rates among those who are adolescents and young adults whereas you look at the next line, which is the dark green with the diamonds in it, white, non-Hispanics, high rates there among middle age populations and some different patterns. When you look at African Americans, black non-Hispanics, tends to have high rates among adolescents and young adults. Rates decline after that when you look at Hispanics and then Asian Pacific Islanders have rates among adolescents and young adults. It drops down and then comes back up again – very different in regard to disparities among various groups.

Here, looking at suicide rates by county urbanization. When you look at the very top line, you can see that rural areas tend to have the highest suicide rates in the United States followed by lower and lower urbanization areas. The thing that stands out when you look at this particular slide is that rates in all of the different levels of urbanization have been going up. But the spread, the difference between the core, the big cities, and the rural areas tends to have gotten bigger. Rates in rural areas going up higher while the rates in urban areas also going – but not quite as high. Again, the disparity between rural areas and core city areas has been exacerbated over the past 20 years.

The annual rate of looking at suicides versus suicide attempts in regards racial and ethnic groups that you can see that there are some very big differences in looking at a ratio between suicide rates and non-fatal suicidal behavior that the ratio is very different when you look at white non-Hispanics versus black non-Hispanics into Hispanic populations.

Here is looking at some non-fatal information from Youth Risk Behavior Survey, which are adolescents that are reporting on a survey. And, here, you can see again some differences by racial and ethnic groups. These are looking at rates by sexual orientation. And you can see that those who identify as lesbian, gay, bisexual, much higher reports of suicide attempts than those that are heterosexual or those who are unsure.

Also, when you look at higher occurrence of suicidal behavior of those with specific learning disabilities. This one just shows learning disabilities, but it also – that same pattern is manifested for those with epilepsy, traumatic brain injury, eczema, or autism. Some of these neurological disorders, but some of them not.

We can look at little bit at the causes. You have seen the social ecological model and you can see where some of these risk factors are demonstrated in regards at the individual level, peer and family level, community level, and then societal level. Some of the causes for disparities may have to do with socioeconomic status, insurance coverage, health status, disease severity, availability of services, discrimination and it occurs at the system level, at the provider level, and also cultural perceptions.

What about some of the policies and what might work. This is really where it calls for comprehensive approach to try to take a look at suicidal behavior. The CDC released a technical package that identified seven different strategies oftentimes combining several of these strategies. They put communities in the best light to be able to reduce suicidal behavior. It really is a comprehensive approach to try to incorporate multiple strategies, multiple programs to make a difference.

And then lastly, implementation and dissemination and I will wrap this up is you can see that there are areas in which various documents ask for looking at disparities and try to eliminate disparities from Healthy People 2010 to 2020 to 2030 also as part of the surgeon general’s call to action to try to engage groups and support high-risk and underserved populations.

Patterns exhibit some similarities between groups, but also some differences. Risk and protective factors may also have some differences and some similarities. But there are limited programs to address specific communities. But more information is needed and there is a broad responsibility.

Thank you very much and I know questions will be coming a little bit later. I will be turning it over to Dr. Van Orden to go over and give you the next presentation.

KIMBERLY VAN ORDEN: Go ahead, Dr. Brant Hasler. I will turn it over to you.

Agenda Item: Sleep, Development, and Circadian Timing Mechanisms of Mood Dysregulation

BRANT P. HASLER: Good afternoon everybody. Thanks to the organizers for inviting me to participate. It is going to be a whirlwind tour today, but trying to give you some relevant findings we have been thinking about and how they might be relevant to suicide. No conflicts of interest to disclose.

I know it is a broad audience, so just a brief introduction about circadian rhythms. These are rhythms that are about 24 hours in length thus circa dia, around a day. And the purpose of them is to organize temporally these processes so they are optimally interacting with the environment and with respect to one another and with the idea that temporal order is essential for health.

We know from the past 15 years or so, there has been an explosion in circadian research and that the molecular clock is essentially in every cell of the body. Perhaps the best metaphor to think about is that we have an orchestra of clocks throughout the brain and the periphery and other organs with the central clock, our suprachiasmatic nucleus, the SCN as the conductor. We are hard-pressed to think of a process going on in the body where circadian rhythms are not relevant.

That certainly applies to reward function, something we have been particularly focused on. These are processes involving wanting or pursuing or anticipating rewards as well as liking or consuming rewards. They are relevant to a wide range of psychopathology. We know from work by our basic science colleagues that the circadian system is heavily implicated in reward circuitry.

And thus, when we are trying to understand longstanding evidence that sleep and circadian disturbances go hand in hand with mood and substance use disorders and trying to understand the mechanisms that links them, reward function is a plausible candidate.

And thus, today I would like to have a few take-home points, one, that circadian rhythmicity is present in affect, behavior, and brain function related to reward and then potentially relevant to suicide.

As we heard from the last talk, adolescents are not only highly vulnerable to suicide, but also subject to what we call chronic circadian misalignment. And the circadian misalignment is associated with a wide variety of processes and outcomes related to reward across levels of analysis and again potentially relevant to suicide.

Starting with that first one, in the sleep research world, there is a longstanding model and there is the two process model that basically says that the timing, the duration, and the continuity of sleep are largely dictated by these two interacting processes, one, a homeostatic process that increases with time spent awake and dissipates with sleep, notice the sleep drive, as well as that interacts then hand in hand with the circadian clock, which puts sleep propensity on a 24-hour rhythm. It turns out these two processes are relevant beyond sleep propensity. Back in ’97, Diane Boivin and colleagues did what is called a forced desynchrony protocol, a very rigorous, in-lab procedure where you are able to separate out the endogenous circadian rhythm from any effects of clock time or clock since awake.

As you can see on the left, the upper left, self-reported happiness demonstrated this endogenous rhythm that closely paralleled the core body temperature rhythm. And it also showed on the right that there is a time awake effect.

Greg Murray back in 2009 extended this work and he used both the positive affect, as measured by the PANAS, as well as a simple reward task, where they used heart rate as an outcome measure and found that both these measures, both self-reported positive affect in this heart rate measure of reward. Again, both showed a robust circadian rhythm that paralleled the core body temperature rhythm on the left and both showed this time since wake effect on the right.

And the main point here is that when we see time of day effects on reward-related processes out in the natural world, we know that they are not purely driven by environmental effects or sociocultural factors, but there is an endogenous rhythm that is present.

Now, as there are, of course, daily rhythms that we see under natural conditions and positive affect has been published on extensively as showing a robust daily rhythm, more mixed effects or mixed evidence about negative affect and I will come back to that in a moment.

We see daily rhythms in socializing. We see it in alcohol use. My postdoc, Garrett Hisler, recently published evidence that it is in self-reported alcohol craving. Greg Murray’s group used the Balloon Analogue Risk Task as a measure of reward motivation or wanting and again found this daily rhythm or daily pattern and reward motivation. And in most of these studies, they all suggested the peak and reward function occurs some time in the afternoon or early evening.

A few years back, we looked into the brain to see if we could see parallels there and a simple design. We were just looking at morning scans and afternoon scans. I should mention that Peter Franzen was involved in this work as well. We found a rough parallel to what we saw with those other measures that in the afternoon, the ventral striatum showed a stronger response to monetary reward than it did in the morning.

How is this relevant to suicide? We know from decades of work that suicide tends to have a time of day or vary by time of day. And in the past, a lot of this has suggested that it peaks somewhere during the day, often in the afternoon. More recent work has paid attention to wakefulness or accounting for whether people are sleeping or not. When you do that as shown here in the red then that peak vulnerability time seems to shift to nighttime hours.

This is just speculation but work by Jonathan Emens and colleagues again using one of these forced desynchrony protocols last year I thought might be relevant. And Jon showed that negative affect in contrast to his naturalistic study did show a circadian rhythm and it roughly was the reverse or mirror image of positive affect so negative affect peaking at night with positive affect troughing at night.

And Jon suggested that perhaps what may happen is that if people are awake late into the night when they should be sleeping – exposed to higher levels of negative affect that they normally would be sleeping through with potential harmful effects. That goes along with something that we had been talking about if folks are asked to get up too early and thus exposed to low levels of positive affect they would not normally be exposed to and perhaps that might have negative effects, including perhaps sort of bending that trajectory that would normally be a large increase in positive affect across the day and perhaps flatten.

Moving on then, adolescents and young adults are subject to chronic circadian misalignment. What do we mean by this? This is a phrase that is used in a variety of ways, but perhaps one of the most widely used definition and the one that we focused on is circadian misalignment is a mismatch between the timing of the behavioral sleep-wake schedule, that is the schedule that you choose or is imposed upon on, and that of the internal clock. With shift work and jet lag perhaps being the most obvious examples.

In adolescents, this is particularly relevant. As circadian timing and preferred sleep timing shift later or what is called delay post-puberty with a peak somewhere around age 20 that they peak and what we call being an evening chronotype. And then after age 20, there is this long slow shift towards being more of a morning lark over the rest of the lifespan.

But during this period of time, during adolescence, there is this delay that leads to a mismatch with early school start times such that if the preferred sleep/wake timing is something like midnight to 9:30. On school nights, they may be trying to go to sleep earlier and suffering insomnia as a result, having to get up much earlier than they prefer in the morning and thus suffering from sleep loss and then also trying to function at a time when their brain would prefer that they were asleep.

Then on the weekend, they stay up late to do the things that teens do. They sleep in extra long to make up for the sleep that they lost during the school week, but all good things come to an end. When Sunday night rolls around, they are forced to essentially travel back multiple time zones to get up for school on Monday morning and thus what has been termed social jet lag.

Now, we can operationalize circadian alignment using the behavioral sleep wake schedule and how that relates or aligns to the circadian clock. In humans, we most often measure the circadian clock using melatonin. Melatonin is a hormone of darkness. It goes up during the night. The onset secretion what we call the dim light melatonin onset so-called because we have to measure it under dim light conditions because melatonin is suppressed by light typically occurs around two to three hours before sleep onset and thus six and ten hours between midsleep and sleep offset. And these phase angles then we can use to quantify the relative alignment in individuals.

I first got interested in this idea with relevance to adolescent substance use back in 2008. I worked with Dick Bootzin, University of Arizona, where we were studying adolescents coming out of substance abuse treatment programs. And when we looked at them at baseline, we found that when we divided those between early circadian timing or early DLMOs and late circadian timing, late DLMOs, they essentially had the same sleep schedules. What really differentiated them was the alignment between their circadian timing and their sleep timing and these shorter phase angles in the late group were associated with greater substance abuse independence.

Thus, if so-called normal circadian alignment looks something like this then folks with delayed phase, use shorter phase angles were representing the misalignment.

Then why did this matter? We find that it is associated with a variety of outcomes particularly related to reward across levels of analysis. Most of our work is focused on particular phase angle in part for the sake of parsimony between DLMO and midsleep.

Across these levels or these outcomes, shorter phase angles were associated with generally worse function. We see shorter phase angles associated with lower positive affect. We see them associated with more weekend alcohol use and that is a published finding. We see them associated with altered decision making on a risky decision-making task. That is about to go in first for review. And we see it when we look in the brain so shorter phase angles associated with a lower striatal response to reward.

Now, we are not the first to be thinking about such things. And indeed, I think Al Lewy was probably the first to use this measure in the context of seasonal affective disorder we published back in 2006 in a large sample of individuals with SAD. A portion of them were characterized and perhaps a majority of them were characterized by later delayed phase. And in particular when you looked at the phase alignment, the circadian alignment, those shorter phase angles were associated with more symptoms of SAD.

Perhaps – not to say that this is relevant to all individuals with mood disorder, but perhaps again there is a portion of individuals where this is important. That goes along with work by Rebecca Robillard and colleagues where in adolescent and young adults with unipolar depression – they did a cluster analysis and identified those with conventional circadian timing and those with delayed circadian timing. And those with the delayed circadian timing were also characterized by more depression and more symptoms –

The strategy of starting with the sleep circadian phenotype and looking at the outcomes I think is a good one. A parallel work was done by this group looking at the sleep/wake disorder so those who had extreme delayed phase. And what they noticed is that when they took a bunch of – a sample of individuals that had been diagnosed with the disorder and then assess DLMO on them, only about half of them actually were misaligned and actually had delayed phase. But among those with this true, if you will, circadian misalignment, those – were also characterized by higher depression than those with the circadian misalignment, again, suggesting that there may be something particularly important about misalignment with respect to mood regulation.

And then ending with a bit of neuroimaging data. We have good results including work by Peter Franzen in humans showing that sleep deprivation alters reward circuitry function. We have lots of evidence from rodent models. The evidence that circadian effects and reward circuitry function in humans have largely been cross sectional to date.

We tried to address that in the context of a sample of 26 healthy adolescents. And basically, we imposed circadian misalignment on these individuals so we could see how they functioned within aligned conditions when they are sleeping on a schedule that fit with their internal timing versus misaligned condition where they are forced to sleep on schedule four hours earlier.

As we published just this year, what we found is that under misaligned conditions, the ventral striatum showed a lower response to winning monetary rewards and that included after accounting for how much sleep they had gotten the night before and how alert they were so using the psychomotor vigilance task.

And we saw parallel effects with the go/no-go task with the right inferior frontal gyrus also showing lower response during the no-go task and also after accounting for the sleep effects.

Very quickly, because I know my time is up. We know that – some evidence that if we misalign people, it messes them up. But what about if we fix their misalignment? We have two ongoing studies, one at 11th and 12th graders. It has been going on for a few years. One that is just starting in seventh and ninth graders in the context of a center grant led by Colleen McClung. In both cases, we are using sleep scheduling and bright light in the morning and blue blockers in the evening to try to correct misalignment and see if we can change any of these outcomes including in the brain.

In conclusion, we see circadian rhythms in reward-related processes that may be relevant to time-of-day effects on suicide risk. Circadian misalignment is particularly common among adolescents and young adults, who we know are at risk for suicide and perhaps this is relevant. We have some evidence of causal effects that affects the brain and so what happens if we correct it.

A list of my colleagues and funding sources and thank you for your attention. I will also turn it over to the next speaker, Dr. Kimberly Van Orden. Thank you.

Agenda Item: Suicide Risk in Late Life

KIMBERLY VAN ORDEN: Wonderful. Thank you, Dr. Hasler. You are a tough act to follow. I will do my best. I am going to be talking about the other end of the age spectrum. I will be talking about suicide prevention in later life. I am going to focus on older people. I am going to talk about some risk factors, some protective factors, and some things we can do to prevent suicide in later life.

No financial disclosures. I just want to mention some NIH funding and of course the best science is team science. I have some wonderful colleagues in the Center for the Study and Prevention of Suicide I want to acknowledge as well as some amazing mentees, who helped me do this work.

I am going to talk broadly about the work done in suicide and late life, not just my own work, though I will mention that here and there. I am going to start with the significance of the problem, picking up on what we learned from Dr. Crosby about the epidemiology of suicide and then I will talk about risk and resilience factors. As we do that, I am going to go through some of the prevention strategies that address each of those. And then I will end with some challenges and opportunities to hopefully be some things we could discuss later in the workshop.

The significance of the problem, as Dr. Crosby talked about, is that older adults have high rates of suicide. And the thing to place in context there is that older adults are the most rapidly growing segment of the population so population aging worldwide. We are going to have many more older people as we move forward. And around the world, suicide rates to tend to increase with age. In the US, the highest rates are among older men and in particular, older white men.

We also know there is some very interesting research looking at cohort effects. When we think about age, we also want to think about cohorts or generations. We know that in successive cohorts, we have a signal that rates may increase. We may see over time an even greater problem in terms of high rates of suicide at among older people.

Another key part of the epidemiology of suicide in late life is that older people when they attempt suicide are more likely to die on that attempt. Most older people die on their first attempt sort of highlighting the idea that we really need to emphasize more upstream prevention with older people as well as treating those who have suicidal thoughts.

We will move on to some of the risk factors. Here is the orientation. I am going to orient you to some of the things I am going to talk about today. One of my mentors, Dr. Yeates Conwell, and I worked on this. It is kind of a way to conceptualize the constellation of risk factors for suicide in later life. We will talk about each of these in turn: depression, physical illnesses, accesses to lethal means, social disconnection, and disability. Of course, we know that risk is greater when all of those come together.

I am going to start with depression. For each of these, I am going to talk briefly about what we know, kind of the rock-solid findings. And then I also want to talk a little bit about some of the newer findings so kind of an update on the science.

Mood disorders, especially depression, are the most common mental disorder among older people. Estimates range from about half to up to 90 percent. At the same time, while those disorders are present, that does not mean they will have been detected nor will older people likely have received treatment.

The combination of depression and medical illness is especially important to consider in later life.

And then another key part to think about when we think about depression in older people is how older people present especially salient for this workshop today is that sleep is one of the presenting problems for older people in the context of depression. Older people are less likely to spontaneously or primarily report sad or depressed mood. But they are likely to report lack of pleasure in things, apathy. And then very importantly, describe the somatic symptoms that go along with depression, including sleep. Sleep problems can be a way to detect risk in older people and it is something that is very salient and important to older people at risk for suicide.

More recent findings are looking at the relationship between depression and then other risk factors for suicide. We know they do not occur in isolation and that they may play off of each other so thinking about pathways to depression in later life that can inform prevention. A recent study looking at social isolation, loneliness, and depression. And then importantly, how those effects are bidirectional.

Some other things to highlight given this time that we are all in together is that older adults may actually be pretty resilient in terms of the COVID-19 pandemic. Just as how we know that with age, rates of depressive disorders actually go down. Same thing with what saw with the pandemic. That older adults are reporting less distress than younger people and that showed up as well in terms of suicide ideation.

And then in terms of treatment, we know that collaborative care for depression treatment in older people is effective, including with suicide ideation. Though important to keep in mind that the effects are relatively small.

Physical illness. A common part of later life. This is not going to be a risk factor that is very specific. I have a lot of data on there that could be – if people want to look at these slides later. But we do know that older people who die by suicide are characterized by physical health problems, in particular, multi-morbidity so many illnesses. Poor sleep quality, as Dr. Bernert – this is one of her studies as she spoke about. We know primary care is essential.

We also do not exactly which illness. There is inconsistency there – key takeaway – multiple illnesses together. And potential mechanisms include neurobiology of stress, of illness, as well as functional impairment.

Disability or physical functional impairment is linked to all the indicators of suicide risk in later life both instrumental activities of daily living as well as activities of daily living such as the more basic things like being able to dress and toilet. Sensory impairment as well is linked with suicide in later life. And then if we think about the conditions under which disability is especially linked to suicide, it is that needing assistance with ADLs, being in long-term care, as well as potential personality characteristics that might make it challenging for people to cope with disability like a high need for control or autonomy. We know that being able to be flexible in how we cope is essential.

There is some interesting work coming out about internalized ageism so believing these ideas that suicide is an expected thing or that depression is an expected thing with aging, not good for our health and also associated with decreased will to live.

There is a pretty complex association between suicide and cognitive functioning that we need to learn a lot more about. To the extent that there is a link between suicide and dementia, it is likely soon after the diagnosis of dementia. And then there is also some really important work done by my colleagues at the University of Pittsburgh, Dr. Kata Szanto and Alex Dombrovski, looking at cognitive control deficits and how those might increase risk for suicide attempts in later life.

Some recent findings are looking at specific illnesses such as veterans with a history of stroke. There is new data coming out on a meta-analysis in terms of adults with HIV. An interesting study in terms of data looking at ICU survivors relevant to the COVID pandemic that that is associated with increased risk of suicide.

We are also learning some about the timing of increased risk of suicide with regards to cancer diagnosis with the highest risk in the first six months after diagnosis.

Some new data also came out about increased risk for non-fatal suicide attempts soon after MCI as well as dementia. And, again, as has been said, it does not seem as if suicide rates have risen in the US overall through the pandemic, but perhaps in some populations. It may be that older adults have faired better. I would love to spend more time talking about all of the wonderful factors that can promote emotional well-being in aging. I certainly do not have the time. But there are a lot of factors such as improved emotion regulation, satisfaction with relationships, other things that are good to look forward to as we grow older.

Social disconnection is a risk factor. We know that social connections that create a sense of caring, contributing, and community have a range of benefits for health and well-being. In contrast, isolation and loneliness and all of the different indicators of that are linked to ideation attempt and death.

We know that the three intervention studies that have shown effects on suicide death in later life all involve promoting social connection. But the studies were not designed to test that as a mechanism. We do not directly know if reducing disconnection will prevent suicide.

Some recent findings, however – there was a recent study by Namkee Choi, Renee Pepin, and Marty Bruce that looked at behavioral activation for lonely, older adults and found some really positive effects on loneliness, satisfaction with social support, and increased social interactions. That is a very promising study.

I recently published a study with my lab mates, looking at engaged psychotherapy, which is a variant to problem-solving therapy that is very easy to deliver. These were lonely, older adults at risk for suicide. We found reduced depression in the context of loneliness and increased social/emotional quality of life. We will be hopefully running a follow-up study to really dig into the mechanisms there.

Peer companionship and friendly calling. A recent study that came out during the pandemic found that friendly calling during the pandemic was effective in reducing loneliness. This would be a community program where volunteers reach out and call an older person. That was found to be effective over just a few weeks.

My colleague, Dr. Conwell, and I published a study on face-to-face peer companionship before the pandemic. Older adults who felt lonely or like a burden on others and this was a randomized trial. We found that being matched with a peer companion was associated with reduced depression and anxiety and feeling like a burden. Likely some effects in terms of changing that trajectory towards better health in later life.

Dr. Bernert mentioned in the beginning the importance of mean safety so firearms. We know that having a firearm in the home is linked to increased suicide risk. I have some resources here for any of you who would like to learn more about safety planning with older people, including addressing firearms.

Some promising directions there that firearm safety counseling holds promise and that older adults find it acceptable. We also know that veterans with – my colleagues and looked at this. And they are less likely to be screened for suicide, less likely to be provided with interventions such as safety planning and mean safety so really important, promising direction to get that done more among our older people.

Very briefly to try to pull some of this together, one of the things I am really excited about with this workshop here is that we are bringing all of these really smart people doing great science together where often we are in these kinds of silos and we are able to think together about how sleep and suicide go together and then hopefully at some point how all these other mechanisms come together as well. This is a figure that my amazing mentee, Julie Lutz, put together with help from others in the field. What we are putting up here is potential pathways whereby social disconnection in later life can put into play all of these cascades of other risk factors that may then lead to suicide risk.

I am looking forward to discussions where we can start to bring all of our work together and really be comprehensive in how we address suicide risk in later life.

One last thing to end with as well is what I alluded to, which is that older people on the whole do really well as they age. In fact, emotion regulation gets better. Well-being increases. There are all of these resilience factors that are present as we age such as coping and more positive emotions.

When we think about the increased suicide rates in later life, I like to think about it as the developmental trajectory where the most common trajectory is one towards health and well-being, but some people get off that trajectory. There are all of these risk factors. If we can address those, we can get them back on the healthy aging trajectory.

Some opportunities for innovation I hope we can speak about. One of our most potent interventions for suicide prevention is caring contacts, but that has not been addressed comprehensively in older people. Again, looking at mechanisms, how do social connections link to suicide risk in terms of decision making? We need to know what works for men. Most of the intervention studies that have been effective for older people were shown to be effective for women.

And then we really want to think about a portfolio of interventions. Some people may want to address sleep directly. Some may like to look at social connection and then have improvements in sleep. We want to be able to personalize that.

Obviously, you do not need to read these references. These are here for you all for anyone who would like to look at the slides later. I know that I am out of time. I want to thank my colleagues in the HOPE Lab -- helping older people engaged. I am very excited to take some of your questions later.

I am going to stop sharing my slides and I am going to turn it over to our next speaker, Dr. Helen Burgess. Take it away, Helen.

Agenda Item: Circadian Physiology and Light Treatment in Psychiatric Illness

HELEN J. BURGESS: Great. Thank you so much. Thank you very much to the organizers for inviting me to speak on circadian rhythms and light treatment in psychiatric illness.

I do have two conflicts of interest. I do consult for two separate companies. One makes melatonin and one introduces workplace step programs. But I will not be talking about those today – I will be talking about is supported in the literature.

What I am hoping to do in just 15 minutes is talk briefly on how light treatment can impact circadian timing and also sleep and mood. I will talk briefly about some of the studies we have done looking at testing light treatment in populations with chronic pain and traumatic stress. And then I did do a quick dive into the literature to see what was out there in terms of light treatment in suicidal ideation.

In terms of how light treatment can affect circadian timing in the field, we talk about this pretty complicated concept of the light phase response curve, which I am not going to go into today due to lack of time. But I will hit you with the main take home messages associated with this.

But I will start by referencing something that I think both Rebecca and Brant sort of touched upon in their talks, which is that later circadian timing or what in the circadian field we sometimes call a phase delay or people who sleep late or evening time, we know that they tend to have higher rates of depression. There is a lot of cross-sectional data available. There is a smaller amount of data, but certainly some data to suggest that if you do phase delay people or shift their circadian time later that even healthy controls will report lower levels of well-being. Certainly, being on the late side appears to increase risk for depression and suicide.

In terms of the effects of light on the circadian system, we know – we can see here – think of this as your typical nighttime sleep episode. Light in a few hours before sleep and certainly in the early first half of sleep lead to phased delayed and will shift the clock later. And Brant kindly mentioned the melatonin rhythm in his talk and how that is the main way currently that we measure the – or infer the timing of the central clock. And just to kind of show you what a phase delay looks like, you can see a baseline melatonin profile in black here. You can see the whole rhythm shifting later by a few hours. We would expect to see similar shifts in other circadian rhythms in the body.

Now, morning light will actually phase advance or shift the clock earlier, and that looks a little bit like this. Again, you can see this baseline melatonin profile in black and you can see the whole rhythm just shifting earlier in time.

Now, a really key concept that I hope you take home from this is that there is something called the crossover point, and what this means is two to three hours before you normally wake up is a really significant point in time because light prior to that point in time will be interpreted by your brain as late nighttime or evening light and it will phase delay you. But once you hit two to three hours before your normal wake time, the switch flips and your brain now interprets that light as morning light and it shifts you earlier. It is really critical that with morning light treatment we do not start the light treatment too early. You could imagine if I thought I am going to get up four hours earlier and I am going to get my light treatment and this is really going to push my clock earlier, you can see that if I did that, I would actually risk starting the light treatment too early and I could end up – the timing of light treatment really has to be tailored to each individual’s timing.

In terms of the effect of light on sleep and the mood centers, we know that the primary circadian photo receptors in the retina – it is sort of a mouthful, but the so-called intrinsically photo sensitive retinal ganglion cells. They are most sensitive to blue wave lengths of light and they do receive inputs from the currents – as I mentioned, when you follow the projections of these cells, certainly the signal is sent to the circadian system. But we can also see these direct projections to the sleep centers and also to some of the mood centers. I should say this elegant work comes from Sidney Hataz’(phonetic) lab at NIMH.

From a clinical point of view, we do have one meta-analysis out there, suggesting that morning light treatment can yield pretty large effect sizes in terms of improving the subjective sleep quality symptoms or reducing insomnia symptoms. And even when sleep is objectively assessed, we see medium effect sizes for improvements in sleep with morning light treatment.

There is even stronger evidence for the effect on mood. And I believe what I have cited here is the third meta-analysis now, showing morning light treatment to be a pretty effective antidepressant. The effect size is similar to pharmacological antidepressants and typically, if you want to see improvements in mood above and beyond a placebo, you will be doing a light treatment for about two weeks. But there is also evidence for pretty immediate improvements in mood even after just one session of light treatments.

Just how is light treatment improving mood? It might be affecting serotonin levels. Catecholamines are also mentioned. But I think the best evidence so far is for like increasing levels of serotonin in the brain. This is one study in humans. It might be one of the best studies we have actually in which they brought people into the lab. These were healthy young women. And they had them drink a tryptophan-deficient amino acid mixture. That is well recognized to lead to reductions in serotonin because tryptophan is a precursor for serotonin and then – recognized to lead to decreases in mood. When they repeated this study – they had women sit in front of light boxes. What they were able to show is that mood did not change. From that, they indirectly inferred that light might be increasing bubbles of serotonin.

And that is actually supported by nice research coming out of Lily Yan’s lab in Michigan State. She has a really nice diurnal rodent model. And she had these rodents in either a month of bright days or a month of dim days. I am showing you here what she found in the dorsal raphe nucleus, but actually she looked at many different brain areas and essentially found the same thing, which was that this history, the bright days and a lot more light during the day led to increases in serotonin in the brain.

And there is also some seasonal data out there, suggesting that most of us have high levels of serotonin associated with longer day lengths so longer photo periods of the summer versus the shorter day lengths and shorter photo periods in the wintertime.

In terms of how we have been testing morning light treatment in folks, we started off with some chronic pain studies and this was one study we did. This was an open-label study in veterans with chronic low back pain. Each veteran had 13 days of morning light treatment. And in this particular study, we used light boxes and very importantly we had the light start at the average wake time or up to an hour earlier and that is the earliest we went, wanting to make sure we were giving them light at the right time to facilitate those advances. And we measured circadian timing at baseline and then also after treatment and some other variables as well.

We had a heat pain test. And what we were able to show is that the temperature at which people started to report at rest feeling painful, people were actually able to tolerate higher levels of heat. Pain threshold increased, but the temperature at which they needed to stop the test was the same.

Self-reported pain intensity reduced over the 13 days of morning light treatment. Pain behaviors came down. It was not significant, but pain interference came down. And self-reported physical functioning also improved.

We did not actually see large changes in depression or anxiety in this group, but they really did not have high levels. But interestingly, and I will go back to this in some future slides, we did see that PTSD symptoms reduced during the treatment.

In accordance with the literature, we did see improvements in self-reported sleep quality. That is a decrease. And we also saw a reduction in insomnia symptoms. We were able to phase advance the clock. While this was an open pilot, it certainly suggested that we should continue to look at morning light treatment as a possible adjunctive treatment for pain. In this study, the improvements in pain were associated with the improvements in sleep and also the circadian phase advance.

We have moved onto a wearable, light device that makes light treatment a lot easier. This is commercially available, and we have created a placebo by dimming down the light levels. We like this device because we are getting light around 500 nanometers, which is close to the heat sensitivity of the circadian temperature. And we can put a monitor on this device to also get a sense of treatment adherence because people are doing this at home.

And we have the study, a randomized controlled trial in people with fibromyalgia, that we will be wrapping up in the next month or so.

Coming back to the reduction in PTSD symptoms that we saw in association with my colleague, Alyson Zalta at UC Irvine, we ran a small internal pilot study, 13 people with probable PTSD. They had experienced a Criterion  A trauma and also had some PTSD symptoms and eight received the bright re-timer for a month and five received the placebo, the dim light. And we were able to show nice reductions in PTSD symptoms with a bright light and also a nice reduction in depression symptoms on the PHQ-9.

We are funded by NIH now to try and dig into basically help elucidate more about how light is soliciting changes in the brain. And we have decided to focus on the amygdala at this point. This is a transdiagnostic study in the sense that we are recruiting people with a Criterion A trauma and some symptoms of depression, anxiety –-

We have three groups. One group gets just 15 minutes of light every morning for a month. The other group gets 30 minutes every morning and the other group one hour every morning. We are looking to see if this can reduce amygdala activity to – in the fMRI and hopefully show a dose response relationship to kind of help support this idea that light treatment is systematically affecting amygdala reactivity to – stimuli.

Just quickly, last couple of slides. In terms of light treatment and suicidal ideation, when I dug into the literature, I found that there were initially a few case reports of suicide that had occurred during light treatment for winter depression. Lam and his group in 2000 published a study from an open trial with 191 depressed patients who were receiving two weeks of morning light treatment, 67 percent were clinical responders, and pretty importantly, they showed that 45 percent showed improvement on the SIGH-SAD suicide item. And only 3 percent showed slight worsening. There were no suicide attempts and there was no stopping of light treatment due to emergent suicidality. The conclusion from this study was that morning light treatment actually relieves suicidal ideation in winter depression and the emergence of suicidality during light treatment is very uncommon.

And then lastly, I just wanted to highlight something that comes up in the literature. It is often referred at triple chronotherapy. This is where you are combining morning light treatment with also some sleep deprivation and then also an advance or shifting earlier of the sleep episode. Just to show you, here, you have a baseline. People sleeping 10 to 7. And then a whole night of sleep deprivation, some morning light treatment, and then people are allowed to sleep essentially, but at a much earlier time. That is slowly delayed, but then you can see that their resting sleep episode is remaining on the early side between 10 to 5.

What they found was that this was pretty effective antidepressant and also on the Columbia Scale, they were able to show reduced suicidal ideation.

I will just thank everyone back at University of Michigan and the lab. I will stop sharing my screen and hand it over to Dr. Peter Franzen at the University of Pittsburgh.

Agenda Item: Suicide Risk in Youth and Emerging Adulthood

PETER L. FRANZEN: Hello everybody. It is my pleasure to be here today. First off, I just wanted to say that I will be talking about sleep and suicide risk in youth and emerging adulthood. I really just want to give a shout out to my colleague, Tina Goldstein, who I have been researching this area with. We run these studies jointly. I will provide a little bit of a review about what we know about sleep and suicide risk in youth and then talk a little bit about some findings from a recently completed study where we looked at the prospective association, using both objective and subjective measures in youth.

And why might we be interested in studying sleep as a proximal risk factor in youth? I think this article from the New York Times back in 2018 sort of highlighted this. They gave two cases of young college students who died by suicide. And in both instances in the chain of events that had led up to the death by suicide, there were problems with sleep. The first gentleman had struggles to sleep and erratic sleep habits in the days before he died by suicide and this woman also who died by suicide had barely been sleeping due to worried about a class. Really, I think this is a nice highlight of two examples where problems of suicide were a recognized contributor to the suicide.

And as was highlighted earlier, rates of suicide had been steadily increasing among youth, faster in youth than other age populations. It is now the second-leading cause of death in both children and adolescents. We can ask the question, why are suicide rates increasing if there has actually been a lot of research on understanding some of the factors that contribute to it. And of course, this is a really complex puzzle. As it was mentioned earlier, it is not going to be a single factor. It is obviously multifactorial. And researchers come a long way in identifying risk factors for suicidal behavior in a lot of domains, but a lot of these are just risk factors, for example, being a sexual or gender minority is associated with greater risk. But that is a distal risk factor as opposed to a proximal risk factor.

Although it is important to know about distal risk factors, which tell us a little bit about who is at risk for suicide so, for example, having certain psychiatric disorders or family history of suicide. That does not tell us very much about when people might become suicidal. For example, a couple risk factors include experiencing loss or the immediate post-hospital discharge period is known as a high-risk period.

Experts have said that the optimal targets for suicide prevention are going to be those that fit these three criteria. They are proximal. They occur in time. They are dynamic and they are modifiable, and sleep fits all three of these.

What do we know about the sleep association in youth? My colleague, Tina Goldstein, published this paper back in 2008 where they did a psychological autopsy study, comparing youth who had died by suicide versus community controls and rates of sleep disturbances specifically insomnia or hypersomnia or having any type of sleep problem were much higher in the youth who died by suicide.

We know, for example, with insomnia – the people with insomnia had much higher risk. In a particular study, we know that rates of insomnia start to go up a lot during adolescence, really with puberty. We start to see higher rates in girls versus boys after the onset of puberty. Lifetime prevalence of maybe around 10 percent. But that really does not account for the high rates of sleep loss that we see in youth.

From the Youth Risk Behavior Survey, nationally representative data, the majority of kids so 73 percent are getting 7 or fewer hours of sleep. Now, you might say, is seven enough. I think there is some controversy about that. But even if you take who gets six or fewer hours of sleep, that is 44 percent.

Why is short sleep so common? I think Brant Hasler did a nice job covering this, but there are a couple of biological changes. We know that the clock is being delayed so circadian rhythms shift later. If you look across pubertal development, what you see is both a later onset of the melatonin rhythm as well as overall less expression of melatonin.

In combination with these changes in circadian rhythms, we also see a lightning of homeostatic sleep drive. This is longitudinal data, showing very steep declines in slow wave sleep. Really, this is leading to our preference for you build up less sleep drive and your clock is later so it leads to preference for later bed and wake times. And of course, there are all these social and environmental factors at play during adolescence as well – decreases in parental control. There are more activities, maybe a job, exposure to – most kids have electronics in their bedroom now. I think the really big factor is early school start times. You get the school sleep squeeze where the biology is pushing people to go to bed late. It results in, as Brant nicely illustrated, lots of reductions in the amount of sleep as well as a lot of variability or instability in sleep-wake times.

We know from a number of studies now that have linked sleep duration to suicidal ideation. Now, this is cross-sectional data, but it was in a large sample of almost 30,000 youth where they looked at three separate outcomes. You can view these as increasing in severity. Having in the past year significant hopeless thoughts versus serious considering making a suicide attempt and then having actually made a suicide attempt.

And what you can see is a dose response relationship such that for each less hour of sleep, you are seeing increasing odds in these kinds of outcomes. But then if you also notice, we see this sort of J shaped or hockey stick distribution where see increases in long sleep, which might be due to hypersomnia in the context of depression. But there are less people who were experiencing it and the risk was less than for the short sleepers.

Again, this is a finding that has shown up in the literature multiple times. In this particular study by Liu and colleagues, youth who got short sleep were three times as more likely to have made a suicide attempt. But in that same study, it was not just sleep duration, but other sleep factors, in this case, nightmares that also showed greater risks of suicide. If you look at what – from a number of reviews, this really growing body of literature linking the sleep suicide association, it is really all over the place from short sleep duration or long sleep duration, insomnia, having an evening chronotype. Yes, growing evidence.

But we could critique the literature in a couple of ways. One, the vast majority of these studies relied on self-report measures of sleep. Many failed to control for depression. That is important because we know that depression and sleep are tightly linked. Do we know that the association between sleep and suicide exists over and above the effects of depression?

Most studies are also retrospective or cross sectional. Really, to better understand what the mechanisms are, developmental timing, we really need prospective and longitudinal studies.

And we might benefit from taking a more sleep health approach. This is the sleep health framework that Dan Buysse talked about in his 2014 paper, where you could imagine a complement of sleep health dimensions that sleep is not just about how much we sleep or even when we sleep. Regularity is a core component as well as how satisfied or sleep quality, your daytime alertness, your timing, your efficiency, and then also of course your duration. And maybe we could be relying and looking at multiple dimensions say with either actigraphy data shown here on the right or with sleep diary to better look at what aspects of sleep might be contributing to proximal suicide risk.

And another reason that our co-organizer, Dr. Bernert, has shown that using actigraphy variability in sleep onset timing was associated with prospective increases in a college sample population.

Just really briefly, I want to share some findings from our study that I conducted with Tina Goldstein, where we looked over a period, using an intensive longitudinal design over up to three months in kids who were very high risk. They were patients in an intensive outpatient program. They were in treatment three times a week for three hours at a time. This is a diversion program for either kids who just got out of the hospital or to prevent kids from going into the hospital.

We studied up to 12 weeks as long as they were being seen in the clinic because otherwise, we were concerned about managing safety risks. For up to this three-month period, we had participants wear daily actigraphy as well as complete a morning sleep diary report and an evening rating about their suicidal ideation. We had pretty amazing compliance. We had a total sample of 59 youth on average. People completed 91 percent of their actigraphy days and 84 percent of both the morning and the evening diary reports.

I will just spend a few minutes just sharing some of the findings. What we would do was send a link to participants in the evening and we asked about the continuum of suicidality - so from passive death wish through suicidal ideation and plans, including any suicidal intent, attempt, or non-suicidal self-injury. If they did indicate they were experiencing any of these things, we reminded them about some places they could call if they were feeling unsafe.

And then just to share some of the initial outcomes, we looked at regularity and variability and sleep duration and sleep timing. We did this by looking at coefficient of variation across weeks in the study, which is just standard deviation and adjusted by the mean, and in sleep time and sleep duration.

On these plots here on the X-axis, we are looking at the coefficient of variation for total sleep time or sleep duration with higher numbers on this scale, indicating more variability. And then on the Y-axis here on the left panel, we are looking at rates of suicidal ideation. On the right panel, if they indicated they were suicidal, we asked them on 1 to 100-point scale how intense were the feeling of suicidal ideation. And what you can see in both cases, there was a steep rise as variability went up in both rates and intensity of suicidal ideation. This is, I should mention, looking at sleep duration within the – and rates of suicide within the concurrent week so at the same time, adjusting for gender, age, and average depression for the week.

If we are looking at midsleep timing, we found similar increases with variability. In this case, we are looking at the higher numbers mean higher coefficient of variation or more variability in midsleep. And we saw, again, as variability goes up, much higher rates of suicidal ideation. But it was not significant in this case for the intensity of suicidal ideation.

In addition to looking at the variability versus concurrent week of suicidal ideation, we were curious whether variability across the week would predict the following week’s ideation and that was true for midsleep timing so again similar relationships where higher variability and midsleep was associated with increasing rates of next week’s rates of suicidal ideation as well as intensity of suicidal ideation.

To begin to look at some of the mechanisms involved and some new analyses just conducted by Drs. Jessica Hamilton and Aliona Tsypes, we were looking at whether reactivity to positive and negative social events might be one of the pathways by which poor sleep gives rise to increasing suicidal ideation. In our evening diary, we asked a couple of questions on visual analog scales, which was really stressing both positive and negative events that occurred in some kind of social context. We trained people to know that this was events that might involve somebody else. We asked how positive their positive event was and how upsetting or negative their most negative event was.

In these analyses, they person-centered these data so zero means they are average level of reactivity to negative events here on the right panel and positive events on the left panel. Higher numbers mean on a day that they were having more positive reactivity to events or more negative reactivity to events. We saw strong associations that on days where there was higher reactivity to negative events or blunted reactivity to positive events. Here, blunted reactivity was associated with increases in rates of suicidal ideation and then what they were finding is that our sleep variables were – that these reactivity to positive and negative events were mediating the association between sleep the night before and any suicidality experienced the following day.

I think next steps here are using both standard analytic approaches and potentially machine learning processes to real specific actigraphic profiles that might be increasing suicidality. We were pleased that we got funding from NIMH to do a bigger study and hopefully will be well powered to do this.

Also, I think it is important to understand what are the mechanistic pathways? What is it that sleep might be doing, which might give us new ideas for intervention? And of course, is this is a promising strategy for suicide prevention? Can we do real-time detection either through smartphones or activity trackers such as Fitbit to either provide alerts to parents and providers or even prompting the team to engage in safety plans.

Just a shout out to the study team who helped with the SPOT study. It was a heroic effort to get through this intensive longitudinal – I am grateful for their efforts. I will turn it back over to our moderator.

Agenda Item: Q&A

KIMBERLY VAN ORDEN: Wonderful. Thank you so much everyone for those amazing presentations. I will invite all of our speakers from this panel to go ahead and join us and turn your videos back on. We have lots of great questions from those in the audience. I picked some. I picked one for each of us. I will go in order of presentation so you know when you are going to be on the hot seat.

I will go ahead and start with Dr. Hasler. The question we had for you was a really interesting one. Someone asked about whether there are race and ethnicity differences in circadian cycles and whether that might be related to some of the epidemiological data that Dr. Crosby presented for us on race and ethnicity differences.

BRANT P. HASLER: Thank you for that question. It is a really great one. The first thing I would say is that we need to know a lot more about this. It is an important future direction. Something we started thinking about, but to some degree, it is hard to get a signal with the lack of research in the area.

What I can tell you is there has been a few good studies done largely by Charmane Eastman and colleagues, looking in the laboratory, whether there are differences between European Americans and African Americans in terms of circadian rhythms. They largely found that at least in adults, it is what is called a shorter circadian period. Our circadian varies. Some people are longer. Sometimes they are shorter. They are all around 24 hours, but rarely exactly 24 hours. They tend to be shorter in African Americans based on this research, which in general suggest that they would have earlier timing whereas having a longer circadian period tends to go with later timing. That actually goes along well with some large epidemiological work done by the UK Biobank that I think two papers have come out from them, suggesting that African – African Americans, but folks with African descent are more likely to be more in chronotype. Now, that has all been done in adults again. They have not replicated those differences in adolescents. It is a little harder to say what is going on with adolescents.

But it might suggest that per the question, do suicide rates decline as age increases in African Americans. Perhaps it has something to do – if there is this circadian component to it so they tend to see to have earlier timing could very speculatively be part of that story. I will leave it at that.

KIMBERLY VAN DOREN: Great. Thank you so much. Since I was next, I will briefly a neat question that I got from the audience about suicide and late life. Someone asked about the ideas of identity and usefulness and hearkening back to Erikson’s generativity versus stagnation stage and does that maybe play a role in suicide and late life. I will say absolutely yes.

There is an example that we talk about a lot in Rochester. George Eastman, the founder of Kodak, who died by suicide, his suicide note said my work is done. Why wait? It is a poignant example of how those ideas show up.

I will say as well, some of the research that points to that is research looking at things like personality characteristics, so someone who is low in openness to experience, that is linked to suicide risk.  You can imagine how someone who is kind of low on that might sort of struggle with taking on the transition from retirement, finding new ways of meaning and purpose. Really great question and thank you for that one.

And then I will go ahead and turn to Dr. Burgess. And the question that I have for you is talking about bright light therapy that you talked about. I am wondering if there are any contraindications for that and if any of those contraindications are things that are characteristics common in those at risk for suicide?

HELEN J. BURGESS: That is a good question. There are a few. Light devices, I should say, do not normally emit ultraviolet light. There are medications out there that are considered light sensitive. But you have to dig into them a little bit because often light sensitizing is referring more to UV light. With these light devices, we do not have to be concerned about UV light.

But there are some medications out there that can sensitize people to more of the blue wave lengths of light – that is something to look out for.

We also know that light can be somewhat proinflammatory in the eye. For the healthy eye, this is such a low level of inflammation. It is not really a concern. But we do get concerned when there is preexisting eye damage or eye disease. We sometimes are a little cautious about adding light treatment to that. In that situation, we usually refer people back to their ophthalmologist to talk about whether or not is suitable for them to start a light treatment.

And then the third group that comes to mind would be people with bipolar disorder. There is some evidence that people with bipolar disorder that do morning light treatment can sometimes develop mania. But the approach to that – in the literature, which seems to be working pretty well is people can still do light treatment. But instead of in the morning, they just move it to the middle of the day around midday. They can still see some nice antidepressant effects from that light treatment in the middle of the day. Those would be the three main ones that come to mind.

KIMBERLY VAN ORDEN: Perfect. Thank you. I will turn to Dr. Franzen. You are up next. And the question we have for you is whether you could talk to us a bit about the role of both objective and subjective measures of sleep and how we think about how those go together and link to suicide risk.

PETER L. FRANZEN: I think there is a place for both types, both objective and subjective. What is nice about getting objective data is it is free from retrospective biases or if you have a particularly bad night, having that over color your perceptions as well as potentially using it in a real-time risk way. But, of course, it is not just about objective patterns and sleep. People’s perceptions about their sleep are really critical as well. I think – data might tell us different sets of things as well as – it might also give clues on what might be the more important factors to be intervening as well. But I also just have a bias for loving to collect objective data.

KIMBERLY VAN ORDEN: Great. Thank you. I will go back around to some of our other questions that we have since we have a little bit of time. I will come back to you, Dr. Hasler, and ask you about – you presented on circadian factors and reward function. I was wondering if you could speak to sleep loss or disturbance in addition to circadian factors and how that might be linked to reward function.

BRANT P. HASLER: Another good question. What I would say is that we do our best to methodologically separate these different factors. We do our best to methodologically separate these factors. We do our best to conceptually separate these factors. But I think that you really cannot think about one without thinking about the other in the real world.

When I think about circadian misalignment, I am really thinking, and I try to convey this on that slide about the adolescents that it is really a constellation of things that are happening with sleep that they may be having trouble falling asleep. There is that sort of sleep disturbance insomnia process. At least during the school week, they are getting less sleep. There is that sleep depravation effect. And then of course, there is the adverse circadian phase aspect to it.

I think to a large degree, we see that each of these things whether it is sleep disturbance or sleep loss, as Peter has done a lot of work in or circadian effects, all seem to impact reward function. In a lot of ways, it seems like it makes it worse. But it is a really complicated question. I think there are also development differences at place.

Just to give an example of that, and Peter can correct me if I am wrong and characterizing his data, but there is some evidence that I have seen in our work trying to isolate the circadian factors that in younger adolescence, we see ventral striatal response go down in response – in the context of circadian disturbance whereas in later adolescence and young adults, it tends to go up. I believe and Peter, again, correct me if I am wrong. He has seen parallel effects with sleep loss. There may be circadian misalignment, again, which it is hard to separate out, that sleep loss effect, maybe having some parallel effects and perhaps either additive or synergistic effects with the sleep loss part.

PETER L. FRANZEN: I would add on to that. One thing though is hilarious. Brant and I will be talking and I will highlight the sleep thing first and he will highlight the circadian thing first. You can see where our clear biases are coming from.

You cannot really separate these processes. They are inextricably linked. It is part of the whole sleep health component. I do not know that we can really ascribe effects being – more to one than others. I can say that in my experimental work where we do sleep restriction studies where we have done sleep restriction studies in adolescents, we see more reactivity to negative emotional stimuli. And in looking at reward paradigms, blunted reactivity to rewarding stimuli.

I think that ties in with the kinds of findings we are showing that just asking for self-reports about reactions to positive and negative events seems like that might be a strong – it might be strongly related. I think that dovetails with the fact that we can see exaggerated reactivity to negative things and blunted reactivity to positive things. That might be some of the mechanistic pathways. I think that that is going to exist both in the context of sleep loss and in circadian alterations.

KIMBERLY VAN ORDEN: Great. Maybe continuing a little bit with those mechanisms that may be crossover a bit.

I was wondering if any of you want to speak to the idea of impulsivity. We know that impulsivity is such a strong risk factor for suicide attempts. Would anyone like to jump in and comment on that?

BRANT P. HASLER: I will say something quickly and then Peter may have something to add. I put that one slide in on the end that we saw in the scanner on an impulse control task, the go/no-go, that we were seeing this difference. I think this is not the place for it. There is a lot of work that needs to be done.

But there is a host of studies that look at self-reported evening-ness or chronotype or circadian preference depending on what you want to call it that is associated with greater self-reported global impulsivity. That has been shown time and time again.

I think that we are understanding impulsivity as a more multidimensional construct these days. There are some improved measures like the UPPS-P I think is probably the best example of that that helps us sort of parse that a bit.

We have a paper about to go in, showing that greater evening-ness again was associated more impulsivity at a global level. And then my colleague, Sarah Peterson, created an EMA version of the UPS where she – they were filling out every day over the course of I believe 17 days. I may have that wrong. Getting sort of trait levels of impulsivity as well. And we saw even stronger associations – and the state-level impulsivity. But it varied by subdimension.

But coming back to suicidality, one of the relationships that was relatively strong was negative urgency so that acting rashly under conditions of negative emotion, which I think is – we were paying more attention to the substance abuse angle that has been strongly related to substance use, but I think it seems highly relevant to suicide as well. This is another area of work where more work needs to be done and we have been thinking about how impulsivity may influence both these positive effect and negative effect pathways and further exacerbate those effects.

KIMBERLY VAN ORDEN: Great. We got an interesting question about the fact of adolescents who are in more acute treatments like dealing with being in an IOP, but also dealing with school schedules and that is basically changing all of their rhythms and the times that they are able to sleep. One person was asking if we are trying to do CBT-I for these adolescents - and we know that variability and sleep is going potentially be linked to variability and suicide ideation - Do you have any suggestions for how they might manage that?

PETER L. FRANZEN: I just saw that pop up in the Q&A. It is a good question because I do think this variability issue is really important. I did not mention, but we just started doing something very similar. We are taking these IOP patients. In this case, we wanted to be broader than just CBT-I intervention. We went with Allison Harvey’s transdiagnostic sleep and circadian intervention, which really does – one of the core components is targeting this regularity. That is one of the first things we are doing is really trying to teach adolescents about the two-process model, sleep-wake regulation, and then starting to understand why we would ask them to try to maintain a regular schedule.

Highlighting Helen’s work, we also are adding bright light treatment to that. This is just an open small pilot. We agree that this issue of adjusting regularity is both difficult to do, but probably really important in not just adolescents, adults too, but adolescents who particularly seem to struggle with that.

HELEN J. BURGESS: I was just going to highlight for folks that maybe are not in the sleep and circadian field. From a circadian point of view, the sleep variability results and the fact that sleep variability increasing risk for depression and suicide when you are asleep shifts around a lot. Your light exposure also shifts around a lot. Essentially, your circadian clock is having to shift around a lot. That is why there is so much focus on sleep regularity as a way of trying to stabilize circadian timing. It is not just whether you are an early bird or a night owl, but also in many ways on a daily basis how much your clock is having to shift around in response to your sleep schedule. That is how I think of it anyway.

PETER L. FRANZEN: And that morning time is really a key anchor both for – largely for setting the clock.

BRANT P. HASLER: The one thing I was going to add and Helen covered a lot of what I was going to say. The one thing I was going to add is with adolescents, it is particularly challenging because you could stabilize their sleep at the school day wake-up time. But then in some ways, it is just ensuring that they are spending a chronic amount of time on a schedule that is not good for them. It does not necessarily solve the sleep loss effect unless you are doing it like Peter is doing in his study, adding that light therapy to help their circadian system shift to the correct schedule.

My personal belief is that these innovative approaches are outstanding. I think we are not really going to get traction until we have the more system-level changes such as delaying school start times rather than trying to treat the individual where we know we have insufficient sleep medicine expertise out there to meet the need.

PETER L. FRANZEN: Thank you for saying that, Brant. The school start actually is a real leading issue. I think that is one of the social policy ways we can be helping – help kids from actually developing depression and ideally suicidal thinking as well if we can just get them to get more sleep in accordance with their biology.

KIMBERLY VAN ORDEN: Another biological need is one for taking a break and standing up and resting. I believe that we are about to our break time. I just want to thank you all for joining us for all those great questions, for our presenters doing those great presentations. I think our break is until 2:30. Thank you.


Agenda Item: Session 2: Neurobiology and Neuroplasticity

MARIA A. OQUENDO: Good afternoon or good morning, depending on what time zone you are on. My name is Maria Oquendo, and I am the chairman of the Department of Psychiatry at the University of Pennsylvania. It is a great pleasure to be able to lead this panel today. I am delighted to let you know that we have four speakers. The first one is Craig Heller from Stanford University. The second one will be myself. The third will be Chiara Cirelli from the University of Wisconsin, and the fourth one will be Dr. Victoria Arango from the National Institute of Mental Health.

I am going to turn it over to Dr. Heller. Dr. Heller, please.

Agenda Item: The Neurobiology of Sleep and Sleep Regulation

CRAIG HELLER: Hello everybody. It is my pleasure to join you for this interesting symposium. But I must say I don’t have the valid credentials for participating. When David asked me to participate, I protested and said I do not work on suicide. I do not even work on people. I am a small rodent doctor. He said, yes, but you know a lot about neurobiology of sleep and sleep regulation. That satisfied me so I shut up.

Then when the agenda came out, I suddenly had this realization that talking about the neurobiology of sleep and sleep regulation in 15 minutes, that will not work. I came up with an alternative plan and that was to tell you about a couple old studies and from those old studies, I think there may be some suggestions about how to improve the efficacy of cognitive behavioral therapy for PTSD.

One thing we know about PTSD is it characterized by short sleep and disrupted sleep. This wonderful graphic from a paper by Straus et al. in 2015 brings that message home. Here, we have actigraphy records for a group of PTSD patients, a group of primary insomnia patients, and controls. What you see so vividly here is the incredible disruption of sleep in the PTSD patients, even more so than in the primary insomnia. And the total amount of sleep is much reduced in the PTSD group due to this fragmentation.
The question then is is the PTSD causing this fragmentation or is the fragmentation in some way exacerbating the PTSD. Whether there is some positive feedback loop here that is making the PTSD even worse because of the disruption of sleep.

In order to ask that question, what you might want to do is think about what aspect of sleep, what function of sleep if disrupted could play a role in exacerbating PTSD. I guess we would all agree that sleep is necessary, and it must be necessary for the brain. We do not know what the function of sleep is. But there are a number of hypotheses, which are in active investigation these days. I list them here: memory consolidation and Gina will tell you more about this tomorrow, brain energy reserves, synaptic downscaling, which I am sure Chiara will tell you about as well as neuroplasticity, and the newest function of sleep, the glymphatic clearance of metabolic waste products.

What putative function of sleep could exacerbate PTSD so that it could be a contributing factor in the incidence of suicide? If you could identify what that function of sleep might be then you could ask the question of can it be manipulated. Can it be therapeutically manipulated in order to reduce the severity of the PTSD and reduce the risk of suicidal actions?

I pick for the consideration, declarative memory consolidation. Why? Because memory consolidation involves a very specific role of sleep and that role of sleep is the consolidation of declarative memories.

During that process of consolidation, the memories are consolidated with an emotional valence. For example, you are more likely to have a long and lasting memory of warm and fuzzy moments just as you are likely to have a long and lasting memory of traumatic events.

And then there is the old conclusion that recalled memories must be reconsolidated. Then the question is whether during that reconsolidation, the emotional valence can be altered through some therapeutic means and therefore would the memory that is of concern be able to be recalled without such extreme stress to the individual.

Now, the other thing to realize is that if we are going to use sleep for treating any particular condition through its effects on memory consolidation, the consolidation requires minimal quantum of continuous sleep. If we can improve the sleep, if we can improve its continuity, we are likely to have more effective benefits.

The story of memory during sleep, declarative memories go back to the discovery of place cells by John O’Keefe and then grid cells by the Mosers and that work led to a Nobel Prize. I think it was 2014. And that work has continued in a number of labs, notably, initially, Bruce McNaughton’s lab and then Matt Wilson, who was his student, continued that at MIT. Gina is going to tell you about her work in this area, I presume.

But here is the basic idea. If you put in a bundle of electrodes into the hippocampus, you do not necessarily see any pattern of firing of these many cells that you are recording from. But if the animal does something spatially - like running a maze, you see develop a sequence of firing, which is therefore presumably representing the memory of that event, the locations of the animal in that maze. And then when you study the animal sleeping, you see the recurrence of that pattern. And you see it recurred at a different timeframe, different rate. It is much faster. But nevertheless, it is a representation of the experience that the animal had during wake.

The advantage of the place cells is that it allows you to make a real correlation between a neural firing pattern and behavior and the behavior you can quantify easily by video tracking.

We cannot put bundles of electrodes in the hippocampus of humans, but nevertheless there is evidence for this memory replay, this declarative memory replay during sleep. This comes from several labs: Kim Paller’s lab, Bjorn Rasch’s lab. And what, essentially, they have been doing is forming declarative memory so individual, for example, in this case, it is memorizing where this pair of cards occurs on the screen. And while he is studying that pair of cards, he is exposed to a particular odor.

During sleep, the odor for half of the pairs of cards are reintroduced multiple times during sleep. Then the next day the retest occurs in which one of the pairs, one of the pair of cards is shown and the individual has to say where the other one would appear or should appear.

The data show that if indeed the odor that matched those pairs was replayed during night, that particular pair was more likely to be remembered. This is taking this evidence that the experience was replayed during sleep, stimulated by the presence of the conditioning stimulus, the odor.

What about evidence for altering the valence of a memory during sleep? For this, we go back to rodents. The experiment is quite simple and straightforward. The animal is fear conditioned with foot shock and the foot shock is paired with the conditioning stimulus and odor. That occurs at the early part of the sleep phase for the animal and then the animal is followed through sleep. During sleep, the odor is reintroduced or a controlled odor, which was not matched with the foot shock.

And then either one day or two days after the initial experience, the animal is retested in a different context. We eliminate the context dependency of the memory. In a different context, the cue is – the animal is re-exposed to the cue, to the conditioning stimulus. The question is then what is the frequency of freezing of the animal. Freezing indicates whether or not it remembers.

Here are the results. If in the testing we reintroduced the control odor, we have a certain amount of freezing. The animals always show a certain amount of freezing in a new environment. But if during prior sleep either one day or two days before the animal had experienced a conditioning stimulus then the amount of freezing is much greater. What this has done is it has intensified the negative emotional valence of the memory.

The opposite is true. You can actually reduce the emotional valence of the memory. In this set of experiments, there is the fear conditioning and then there is the odor reapplication during sleep. But prior to sleep, the animal has received bilateral injections of a protein synthesis inhibitor into the basal lateral amygdala. And then it is tested with the condition odor in a different context the next door.

And here what we see is that the control odor has not reduced the frequency of freezing on the subsequent day. And whether you add anisomycin, the protein synthesis inhibitor, it has no effect on the frequency of freezing.

But if the anisomycin has been delivered to the animal that is receiving the paired odor, the conditioned odor, the valence of the memory is much reduced.

Now, the interesting thing is if you repeat this same experiment, but you expose the animal to the paired odor during wakefulness, it does not alter the subsequent level of fear.

Possible applications of sleep replay to reduce PTSD. These may be naïve ideas, but at least they might be worth considering. Exposure therapy has proven beneficial in treating PTSD. But the benefits can be to a certain extent context dependent, dependent on the safe therapy site and not necessarily be that effective out in the real world.

Based on the work that I just showed you, would it be of interest to couple exposure therapy with a conditioning stimulus such as an odor and then re-expose the patient to that odor intermittently during sleep? Would that reactivate memory of the therapy session in a context-free environment during sleep in a context-free environment? And might that generalize the benefit of the therapy sessions?

Could pharmacological intervention during sleep replay and reduce the emotional valence? Post-trauma treatment with beta adrenergic reduces emotional valence in PTSD. Therapy session conditioned stimulus used to trigger replay during the first quarter of the post-session sleep might be more effective if combined with an adrenergic blocker.

And then finally, studies demonstrating the efficacy of a noradrenergic inhibitor for the treatment of nightmares or daytime hyperarousal are consistent with these findings.

Just quickly, I would like to talk briefly about the consequences of sleep fragmentation because the benefits of cognitive behavioral therapy might be compromised by the sleep fragmentation we see in PTSD.

We know that sleep fragmentation due to obstructive sleep apnea causes cognitive impairment and PTSD is similarly marked by sleep fragmentation.

Here are some animal studies in which sleep has been fragmented by optogenetic stimulation of hypocretinergic cells. And these stimuli are delivered at particularly intervals such as 60 seconds, 120 seconds, and so forth.

And what you find is that with the stimulation, you have fewer long bouts of wakefulness and more short bouts of wakefulness. Here we see short bouts of increase and long bouts have decreased.

Now, we see the effect of that on memory. This is novel object recognition. And after the training, the animals are then subjected for four hours to the fragmentation of sleep and then tested 24 hours later.

And what you see is that animals normally respond very well to this training session. They remember the familiar object. And if they are totally sleep deprived, there is no memory. And if they are fragmented at 30 seconds or 60 seconds, there is no significant memory. But if the sleep is fragmented at 120 seconds, the memory is normal. This would indicate that there are minimal quanta of continuous sleep necessary for the consolidation of declarative memories. Is this true in humans? We do not know.

The conclusions that I come to is that using conditioning stimuli to enhance sleep replay of cognitive behavioral therapy sessions might improve their effectiveness. Secondly, promoting better sleep in PTSD patients may improve their responses to cognitive behavioral therapy.

Thank you. I am sorry I have gone over.

Agenda Item: A Biosignature for Suicide

MARIA A. OQUENDO: Thank you very much, Dr. Heller, for that excellent presentation. I am going to go ahead and share my screen.

Good afternoon. I am going to give you a very brief overview of a biosignature for suicidal behavior. It is hard to do it justice in 15 minutes. I would like to encourage you to think of this more as a conceptual talk rather than details about the science supporting it.

But I wanted to just start with sharing an idea that perhaps you have struggled to. One of the things that we observe especially in the lay press and literature is that suicide is very often viewed as a reaction to a stressor. Something terrible happens to the person and the person responds with suicidal behavior. In fact, that is rarely the entire story. In fact, most of us are exposed to stress, sometimes very terrible stress. All we have to do is think about the pandemic, and we do not engage in suicidal behavior. Clearly, something else is going on. I hope that my talk gives you some ideas about the contributors to risk for suicidal behavior.

These are my disclosures. I am not going to be covering any of these items in my talk today.

Let me start by presenting a model for understanding suicidal behavior, the stress-diathesis model. This model has been around for a couple of decades. And the idea is that an individual can go from a state in which they have no suicidal ideation to in the context of the diathesis that has genetic components, adverse childhood experiences, and physical and physiological abnormalities and functioning of the hypothalamic-pituitary-adrenal axis, inflammation, serotonergic function, and polyunsaturated fatty acids. But in the context of that diathesis or vulnerability, they can develop suicidal ideation and a plan when they are exposed to a stress. And the stressor can be a life event or it could be the onset of a psychiatric episode. When those two things come together, the individual crosses this threshold and makes a suicide attempt or death.

An important issue with this model is that it does give us two different ways of intervening. We can intervene on these risk factors, impacting the diathesis, or we can try to do things like prevent recurrence of psychiatric episodes. Of course, the environment is very important and restricting access to lethal means is critical as is making sure that people have access to care.

Let me just start by saying that suicidal behavior is at least partly genetic. Although the studies that have been conducted thus far using GWAS approaches have been unpowered, more and more evidence is accumulating of the importance of genetics. But certainly, from a heritability point of view, we have lots of data.

This is just a slide from a study that we conducted many years ago where we showed – we followed depressed individuals who were suicide attempters and their offspring and depressed individuals who had never made a suicide attempt and their offspring. And what we found is that the offspring of non-suicidal probands had – yes, they had some suicidal behavior, but their rates were relatively low, especially compared to the offspring of depressed individuals with suicidal behavior. You can see that the percentage of individuals – the cumulative percentage - is much greater in those individuals.

However, if the parent who was depressed and had a suicide attempt also had a sibling who had made a suicide attempt, you can see here that the risk was as high, but it happened much earlier during life - again, supporting this idea that the heritability of suicidal behavior is really critical.

But of course, fortunately, our genes are not everything. And we know that events during development are extremely important and have impact on epigenetics, which will impact the phenotype of course and that these can be things such as childhood abuse or neglect, which I have already mentioned, but also environmental effects. We know that there are impacts of effects, environmental effects that can be perinatal or even intrauterine. This is one of the many reasons why genetic markers are so difficult to identify, but also explains why early childhood events have such important impact.

In fact, childhood adversity has been reported to predict later depression, which we know is related to suicidal behavior, impulsivity, and aggression, which is also related to suicidal behavior and that was mentioned during the previous session, and suicidal behavioral itself.

Interestingly, in animals, for example, maternal deprivation in monkeys, which is a model of early neglect, resets the serotonergic system so that there is lower serotonergic function. This persists into adulthood and is associated with more impulsive and aggressive behavior in adulthood.

In addition to childhood adversity, we know that the hypothalamic-pituitary-adrenal axis has a very key role in suicidal behavior. This is a cartoon, if you will, of how this works. We know that early childhood adversity tends to increase the release of cortisol and ends up desensitizing receptors in both the CNS and peripheral nervous system, but also extensively throughout the body. This has consequences for in the CNS, neurogenergic and serotonergic transmission and peripherally in terms of changes in inflammation and the release of proinflammatory cytokines.

This is critical because we know that individuals who have dysregulated HPA axis function are at greater risk for death by suicide. This is a study that was conducted many years ago in which patients who had failed to suppress cortisol secretion in response to dexamethasone, which is an external steroid, which normally should mostly suppress morning cortisol. Individuals who failed to suppress cortisol were 14 times more likely to die by suicide relative to individuals who had not, a very powerful effect.

I have already alluded to the importance of the serotonergic system in suicide. I know that Dr. Arango recognizes this slide and will probably tell you a lot more about the findings in post-mortem studies later. But I wanted to share with you some of the findings from a more clinically oriented study. I like the study even though it is very old because it was conducted with individuals who had made a very serious suicide attempt and ended up in the ICU.

In this study, they measured cerebral spinal fluid levels of 5-hydroxyindoleacetic acid, which is the primary metabolite of serotonin. And what they found in this study was that the individuals who had low levels of metabolites indicating that their serotonergic function was dampened were much more likely to die by suicide within a year of the hospitalization relative to those who had higher levels of serotonin metabolites in their CSF.

This is a study that we conducted with depressed individuals who had made suicide attempts. We compared individuals with high lethality to individuals with low lethality suicide attempts. And what we found is that those with higher lethality suicidal behavior showed these decreases in response to fenfluramine in prefrontal and dorsal lateral prefrontal cortex. And that the magnitude of the abnormality was related to both impulsivity and the suicide intent. As you can imagine, the degree to which the person is determined to die by suicide and their level of impulsivity will have an influence on the lethality of the actual behavior.

There has been tremendous interest in the impact of neuroinflammation on behavior and suicidal behavior is no exception. In the last decade or so as we have learned a lot more about CNS inflammation, some of the studies that have been I think very instructive include studies, for example, of the administration of interferon alpha. Interferon alpha, as you know, is a proinflammatory cytokine that is used for the treatment of hepatitis and some cancers. And what it does in addition to its intended effects on hepatitis and cancer is that it shunts tryptophan, which is the primary building block for serotonin. It shunts tryptophan away from being made into serotonin and instead indolamine dioxygenase creates kynurenine from tryptophan and that in turn is metabolized to quinolinic acid, which has some neurotoxic effects.

Interestingly, 70 percent of people who are exposed to interferon alpha develop depression. And in those individuals who developed depression, you can see it here in this little infographic - the ones who developed depression have higher level of ACTH, adrenal corticotropin-releasing hormone, which as you know is related to hypothalamic-pituitary-adrenal function and they also show higher levels of circulating cortisol. Again, this connection between inflammation and serotonergic function.

That is in depression. What about suicide? In suicide attempters, we see that attempters tend to have increased levels of proinflammatory cytokines such as IL-6 and also tumor necrosis factor alpha. And they have decreased levels of IL-2, which is an anti-inflammatory cytokine relative to non-suicidal depressed patients. Controlling for depression is critical because I already mentioned that depression has an impact on – is related to abnormalities in proinflammatory levels of cytokines.

Interestingly, violent suicide attempters relative to less violent attempters, that is, individuals who use methods such as gunshots or jumping, tend to display the highest levels of proinflammatory cytokines relative to individuals who use methods such as overdose and also compared to healthy controls.

Here, you can see this is a study of 24 teenagers who died by suicide of all different diagnoses relative to normal controls who had died from other causes. Here, you can see again the differences in proinflammatory cytokines, tumor necrosis factor alpha, interleukin-1 beta, and interleukin-6.

We talk about all of these systems as if they were unrelated, but in fact, we know that all of them or most of them interact very significantly. This is a cartoon from a paper that my team wrote a few years ago where we showed – we reviewed the literature on suicide and all of the underlying biological factors. And what we found is that they were – we found very strong support for the involvement of the hypothalamic-pituitary-adrenal axis represented here. And we also found a very important role for the serotonergic system, again, represented here. I already told you about the shunting of tryptophan to kynurenine and quinolinic acid, which has a negative impact on neuroplasticity. We know that serotonin has pro-neuroplastic effects and that lower cortisol levels also have pro-neuroplastic effects.

This is just to remind us of the importance of keeping in mind while for simplicity, oftentimes we study these systems and discuss these systems as if they were isolated. They are in fact very closely related to each other and interact and we should keep that in mind as we conduct research on these topics.

With that, I am going to stop sharing my screen and turn it over to Dr. Cirelli. Cirelli, please.

Agenda Item: Sleep and Neuroplasticity

CHIARA CIRELLI: Thank you for the opportunity to tell you a little bit about our efforts to try to understand why we need to sleep and especially as was mentioned before why the brain needs sleep. This work is done in collaboration with Giulio Tononi and is funded by NIH and the DoD.

We know that we need sleep. If we do not sleep, there are very broad negative effects on cognition from attention, learning, and memory to the ability to appreciate humor, reward, and risk. The question is why the brain needs sleep.

To try to address this question, you really have to start from the very definition of sleep, from the very feature that distinguishes sleep from any other state. And that feature is partial sensory disconnection. When we are asleep, by definition, we lose the ability to respond promptly to potentially very dangerous situations. There must be at least one very important function if not more, that is better performing sleep than in quiet wake. If that were not the case, evolution would have found a way to perform that function in quiet wake because quiet wake, again, is a less risky behavior than sleep.

We think that function is to maintain synaptic homeostasis. I do not have to tell this audience of course that synapsis is the basis of how the brain works and synaptic activity accounts for the bulk of the brain energy budget.

Now, in mammalian brain, the majority synapsis are glutamatergic synapses that act as Axo-spinous synapsis. They look like the ones that are depicted here. This is what I am going to focus mainly on my talk.

Here is the idea that we are trying to test. The synaptic hypothesis is that when we are awake, we are always learning something new and that is what we call ongoing learning. It is not just when we go to a lecture, to a class. And because most, although not all, but certainly most forms of learning are mediated through potentiation, the end result of this process by the end of the wake day is an increase in synaptic strength in mammalian brain circuits that you see here. And then the job of sleep would be to allow broad comprehensive and yet selective renormalization of synaptic strength.

Now, I think most - really the great majority of people do agree that learning happens through strengthening and that because of that, you need the process of renormalization and weakening of synapsis to maintain networks stable. But many people would think that perhaps the brain can do this job at any given time. It can strengthen some synapses and depress others. Really the core of this idea as you see here is that there is a time to go up and show a net increase in synaptic strength, which is during wake when you are exposed to the environment, and the arousal systems, like the locus coeruleus, are very active promoting learning. And there is a time sleep when these systems are less active when we are disconnected and that is crucial. We are no longer slave of the here and now. But the brain is active and therefore there can be a comprehensive sampling of synaptic weights. And the job of sleep can be done, meaning this down selection.

Of course, in 15 minutes, I do not have time to go through the details, but we can discuss later about what we know already about the specific electrophysiological and molecular mechanisms by which this process can happen, because this process we think can account for many if not all the function of sleep that Dr. Heller mentioned before. The cellular benefits in terms of saving energy, for instance, and bringing synapses away from saturation, but of course the systems levels benefit in terms of learning and memory.

Measuring synaptic strength is quite difficult in vivo and that is why we have been using many measures over the years in many species because we start from the assumption that if this function is correct, it should be true in flies as well as in humans.

I will just show you a few examples so using perhaps the most direct measures that we have. One is the molecular marker, very well established. The expression of AMPA receptors at the first correlation level increases with potentiation and increases with depression. On the left, you see our quite older by now study, which we took synaptoneurosomes from rat cortex of the hippocampus. We measured these markers after several hours, six or seven hours in the rat of wake compared to six or seven hours of sleep. You see that these markers on average are 30 to 40 percent higher after wake than after sleep.

On the right, you see a more recent study by Rick Diering’s lab. This in mouse forebrain in postsynaptic densities. Basically, the conclusion is the same.

Now, over the last several years we actually thought that perhaps the most stringent way to test the general claim about the imbalance between sleep and wake and synaptic strength was by measuring literally the size of synapsis. Because if synapses get stronger in wake and get bigger and vice versa, they should shrink during sleep. Of course, synapses are small, so you need very high spatial resolution. That is why we are using electron microscopy that allows you very effectively and automatically to require large stacks of images. 

In our case, the cortex and hippocampus. Unfortunately for us, once you have these stacks, the process of reconstruction, segmentation, and quantitation has to be manual.  That is why it is still so time consuming. In any case, we reconstructed the dendritic branches, the spines, the synapses, the external boutons. And what we really are interested is this red area, the axon-spine interface because that is an established structural marker of synaptic strength.

That area as well as the post-synaptic density is directly correlated with the amount of AMPAR receptors that are expressed in the synapse and the currents that are mediated by these receptors.

Here, is the take-home message basically. That indeed we see that after several hours of sleep on average, synapses get smaller than after wake. Here is the regional study that we did in the primary sensory cortex. We tested more than 7,000 synapses. Each dot here is one of these axon-spine interfaces. And you see that, overall, they decrease by 18 percent or so in the majority of synapses in sleep relative to two wake conditions, continuous wake and sleep deprivation.

Now, hippocampus. So far, we have looked in the  striatum and we see CA1 synapses. And basically, the story is the same. Actually, in the hippocampus, when we compare sleep with sleep deprivation, we even see with sleep a decrease not just in the size, but also in the number of some synapses.

Now, what about development? These are two-week mouse pups. Here, when we compare the motor cortex again after sleep and sleep deprivation, we still see that sleep brings about decreasing synaptic strength. Actually, on average, the decrease is even bigger. It is 34 percent rather than 18 percent. The other animals were more mature with one-month adolescent mice.

Now, we think that this overall process of renormalization brings about again cellular benefits because it allows to save energy and brings synapses far away from saturation. But what about the systems level benefits? Can we in a mouse model see how sleep could do the job, both sides of the story?

Very recently, this paper was just published. We tried to do so using mice and to follow photoimaging. We are again looking at synapses in the motor cortex. We are using SEP-GluA1 which is another synaptic marker that goes up with strengthening, goes down with depression.

We studied this in mouse before and after baseline sleep. The next day after training in a motor task and then we divide the mice. Some are allowed to sleep and some are sleep deprived. And then we image again on then the next day.

Now, we use this complex task that is sleep dependent. We know, we have shown before that mice can learn to stay on the wheel and run faster. And then after the first session if they are allowed to sleep, the next day they will start again at the good performance. But if they are sleep deprived that performance is degraded. We can say that sleep promotes the offline consolidation of this task.

What happens to the markers of synaptic strength? This is the overall behavior in all the spines, 1,500 or so that we tested in all mice. As expected, based on what I told you before, with normal sleep, at the end of baseline sleep, the marker goes down. The normalized difference is negative.

Then after training, it goes up because this task is involving the potentiation of synapses in the motor cortex. Then what you see here is, if the animal is allowed to sleep still, overall the net effect of sleep is depression.  Instead, this does not happen if there is sleep deprivation.

What about the link between performance and this process of renormalization?  So here we can follow each single synapse.  So we are focused on what we call the Max spine synapses, that show the largest increase immediately after training.  And we ask what happens to them depending on whether the animal can sleep or not.  

Here is what happens to these spines -- these Max spines are only the minority, 15 percent or so, and these are the other spines, the great majority of the spines.  You see that these spines grow a lot by definition with training.  And even then 24 hours later they are still as expected, stronger than at minus 24, before learning.  But you also see here that the sleep deprivation group, they behave in the same way.  

So sleep is not providing any specific advantage to the learned Max spine in itself.  However, here is the difference because in sleep all the other spines of the majority lose some weight.  This does not happen in sleep deprivation.  So when we compare the difference between Max and other in the two groups we see this significant interaction. And we also see that the extent to which the other spines are weakening, are having a negative normalized difference after sleep, correlates with how better these animals are the next day in their performance.  

So with this example we can, for the first time I think prove that this process of renormalization can account for both cellular and systems level benefits.  Cellular benefits because if overall you have weaker synapses, they cost less energetically.  But also at the performance level - these animals, if they sleep, are better off.  And performance is improved because there is an increase in the signal-to-noise ratio. And this increase is not due to an increase in the signal, but Max spines are not treated differently in sleep and sleep deprivation.  It is due to a decrease in the noise due to the weakening of the other spines with sleep.  I think I stop here and would be happy to take your questions later on.  Thank you. 
Agenda Item: Post-Mortem Investigations of Suicide

VICTORIA ARANGO: Good afternoon. My name is Victoria Arango. By the way, I am at NIMH. Some of you may have noticed that somehow my affiliation may have been misstated in some of the – agenda.

I have no conflicts of interest to report.

I am going to follow the leads of Dr. Heller and Dr. Oquendo because I feel that the amount of available data, postmortem data in suicide – that I decided to provide a history of when and how certain findings occur that had been pivotal in the process of initiating data in a particular field that has led to the accumulation of postmortem data in suicide.

I am going to start by talking about the Monoamine Hypothesis of depression that was written in 1965 by Schildkraut, which really refers essentially to catecholamines, not necessarily to all monoamines. It was later that Coppen emphasized the potential role of serotonin.

Now, relevant to this meeting because there are so many serotonergic findings in the brain of people who die by suicide that there is a very long controversy between the relationship of sleep to serotonin finally seems to have reached resolution. The serotonin is necessary for sleep at least in zebra fish and mice. This was written up in Nature a couple of years ago.

Many, many years, the research on suicide completion was concentrated on these neurotransmitters. Many investigators have reported a lot of changes in monoamines and their receptors and also on multiple neurotransmitters like trophic factors, peptides, neurons, and glia. Because of the very labor-intensity and difficulty in carrying out this type of work, the sample sizes have tended to be small and they have been difficult to replicate by different groups.

Then there were serious attempts that tried to separate depression from suicide. This began in the ‘60s. Originally, all the work was assumed that if you died by suicide, you have to be depressed, but that is not the case. It is important to recognize that there were great attempts trying to indicate these differences.

There were a lot of investigators in the ‘60s and ‘70s who used homogenates of brain to be able to look at neurotransmitters in the source nuclei and also later there were homogenates of brain that were used to be able to look at the effects of different or finding of many receptors being affected in the brain of suicide.

Even until now, there had not been any psychiatric characterization of the cases that were studied. That became very important because with the event of psychological autopsies, which was done in 1974, the Barraclough study of 100 suicide decedents and he found – he and his group found that 93 had at least a psychiatric diagnosis. This has been continued to be replicated today to the same degree so that between 90 and 96 percent of people who died by suicide have a diagnosable psychiatric disorder even if it has not been diagnosed.

The cholesterol story is very interesting. For a while there in the end of the ‘70s, there was a lot of implication of cholesterol and suicide and cholesterol and violent death. That means low cholesterol was implicated in violent – dying by violent death even more than the benefit that it provided cardiovascular in having the low cholesterol.

Now, this is still controversial. It has continued to be studied. There were – which actually showed in the CFS that the most aggressive monkeys were also the ones that had the lowest THIA and these were monkeys that also had the lower cholesterol because they were participating in an NIH-funded study to look at the effects diet and cholesterol.

Now, later on there were the adoption of the twin studies that were at the beginning of an indication, like Dr. Oquendo mentioned, that there were some genetic influences in suicidal behavior and in suicide completion.

Unfortunately, the problem – even though there is definitely the indication that there are many – that there is a genetic transmission of suicide or at least the heritability of suicide and psychiatric disorders. This has not been able to be shown with the new genetic methods. I will get back to this a little bit later.

One of the original studies that showed suicide had a genetic basis was the classic study by Egeland and Sussex in 1985 that showed that there was in the Amish County, which is a very isolated community -- they studied a number of suicides for 100 years between 1880 and 1980. They found that the great majority of the people who had died by suicide were diagnosed with a major affective disorder, actually like 94 percent of the cases.

They were situated in – they were concentrated only in four pedigrees. It showed that even though they were exposed to perhaps the same environmental stressors, there was something that was being transmitted within families.

Dr. Oquendo spent quite a bit of time on the stress responses in suicidal behavior and completion. I am just going to say very quickly right now that one of the pathways that have been considered for suicide risk is early life adversity. Although many people have been working on this problem, including my former group of co-investigators before I came to NIMH, one particular group that has really gotten a lot of results has been the Turecki group in Quebec.

Dr. Oquendo also spoke about neuroinflammation and lipid metabolism and stress. Stress, which is – suicide is not a normal response to stress. It is really a very abnormal and extreme response to stress. But yet the HPA axis is very affected and not only the HPA axis, but serotonin is very affected by stress. We spent quite a bit of number of years looking at this. I am just going to give you a summary of some of the effects of stress on the serotonergic system.

Right now, this is a rat that was subject to immobilization for two hours. This is a section through the brainstem. This is the dorsal raphe nucleus. It is stained for tryptophan hydroxylase with an antibody for tryptophan hydroxylase. You can see when we blow up this part of the dorsal raphe that stress increases the amount of tryptophan hydroxylase immunoreactivity following subject to stressors.

Not only does it happen in rats. It also happens in humans. When you look at a normal control at three different magnification levels, looking at the dorsal raphe nucleus, the source of all serotonin for the rest of the brain you can see when there are suicide decedents, how much darker the immunoreactivity is.

We spent a number of years at first just looking at the level of – of the darkness of the immunoreactivity, which is (indiscernible) way to be able to do this. 

There is a very large body of literature that shows that inflammation – there is a lot of inflammation in suicide, not only in suicide behavior, but in suicide completion. Part of that stress that is early life adversity or childhood adversity. Using a very large number of human brains in a study that -- we used about 250 brains quantified by several receptors, neurogenic and serotonergic receptors, one of the findings that we found is that 5-HT1A receptors of which this is another radiogram of 5-HT1A receptors in the human brain, were increased in the ventral prefrontal cortex of people who were exposed to early childhood adversity.

The same thing with another receptor in the serotonergic system, the 5-HT2A receptor, we found that suicide decedents who were also exposed to childhood adversity that appear in red had many more receptors for 5-HT1A binding than any one of the other groups.

It is very hard to do this work in humans because the dorsal raphe nucleus of which there is here a 3D reconstruction is about 25 millimeters in length. Here, you have tryptophan hydroxylase immunoreactivity activity shown from the front of the raphe, which is large, to the coil raphe.

And going back to the original studies that we are showing that there was less serotonin in the brain of people who had died by suicide, we actually did not find that. We found and this goes back for many years that people who died by suicide had a higher number of tryptophan immunoreactive neurons and it had a higher density of those neurons.

When we looked at a large number of suicides and controls, like 120 people looking at in situ hybridization of --

MARIA A. OQUENDO: Dr. Arango, I am sorry to interrupt you, but I just was hoping you could wrap up so that we could have time for Q&A. Thank you.

VICTORIA ARANGO: This is the last one. Okay. Anyway, it was actually more serotonin, unfortunately. This was continued when we did HPLC. Every millimeter across the raphe, we found that the suicides had more 5-HIAA and more 5-HT which is contrary to what we had found.

To give another one of these summary slides, there seems to be a disconnect between all of the markers that had been found in the forebrain to be hyperserotonergic and the hyperserotonergic brainstem.

There is a lot of people who contributed to this research. I thank you very much.

Agenda Item: Q&A 

MARIA A. OQUENDO: Thank you, Dr. Arango for that excellent presentation. We do have time for a few questions. The first question is for Dr. Heller. Dr. Heller, do you think protein synthesis inhibitors could be used to treat PTSD in humans?

CRAIG HELLER: Yes, I think that definitely – I think so because there are a number of antibiotics that are protein synthesis inhibitors that are freely used in humans. It would be worth a trial of combining some protein synthesis inhibitor with the recall hypothesis, the replay hypothesis.

MARIA A. OQUENDO: Thank you very much.

Dr. Cirelli, do we know anything about the specific molecular mechanisms responsible for sleep dependent synaptic renormalization?

CHIARA CIRELLI: Yes. We know there at least a couple of genes, one being 1A(?) and another one actually GSK(?) that have been linked directly to the ability of a synapsis to undergo synaptic depression - and they have been linked to how the brain is being used during wake. Basically, we think that, based on activity, some of these markers can work as positive ties that protect the synapses from the renormalization and others, perhaps the ARC being one of them, can actually select the synapses that have been used the least and making them more sensitive to the renormalization. Because at the end what we know is at least for the glutamatergic synapses that I focused on, the weakening at the end always involves the interpolation of these receptors. It must be the machinery that does the endocytosis that these are being recruited. There is actually quite a lot of candidates. My bet would be not just one or two. There must be several also because depending on the brain areas probably they change.

I just alluded to the fact that unfortunately the Yen(?) studies are extremely time consuming. It would be nice to measure whether these imbalance in the synaptic strength allow sleep, for instance, in the ventral striatum or in the amygdala. These areas, which it was mentioned this morning is so important to have a balance for mental health. That is hopefully something we can do in the future.

MARIA A. OQUENDO: Thank you very much. It is very helpful.

Dr. Arango, is there enough evidence to indicate that there is a biological component that imparts, A, if not B, vulnerability to suicide?
VICTORIA ARANGO: Yes, actually. I think that without diathesis that you mentioned or without the vulnerability, it would be very hard to really take your own life. We have a lot of data from individuals who are exposed to the same stressors or have the same difficulties, the COVID example. And yet, there are resilient individuals and there are susceptible individuals to suicide.

MARIA A. OQUENDO: Thank you very much. Another question for Dr. Heller. Why do we seem to know so much less about the roles of REM sleep and learning and memory relative to the roles of non-REM sleep?

CRAIG HELLER: That is an interesting question. I think it points to our lack of – a misperception about REM sleep. We just assume that REM sleep like non-REM sleep is correcting some imbalance or deficit caused during waking. But we published a different hypothesis a number of years ago with some pretty good evidence and that is that REM and non-REM are in a homeostatic balance. If you start looking at the function of REM as related to non-REM rather than as related to waking, you will take very different approaches.

MARIA A. OQUENDO: Thank you very much. Very helpful.
Dr. Cirelli, more non-REM questions. Do non-REM sleep and REM sleep have differential effect on sleep-dependent synaptic renormalization?  Excellent follow-on.

CHIARA CIRELLI: Yes. Excellent. We do not know. It is very difficult to study in animals. All the data that I presented are comparing total sleep, the major sleep period in a rat, in a mouse six, seven hours of total sleep, including of course non-REM and REM relative to wake. We do not know. It is very difficult to tease them apart because of REM, especially in mice, is very short and it is very difficult to do selective deprivation of non-REM and REM.

Having said that, my bet would be that actually they both contribute to this overall renormalization. The reason I am saying this is that they share at least one fundamental feature, which is the low level of most arousal systems. And that is what biases the system towards depression rather than strengthening. But it is my gut feeling. I do not have any direct evidence based on the data.

MARIA A. OQUENDO: Thank you very much. There was a question for me about the small percentage of individuals who died by suicide who do not have a diagnosis. This is a very interesting issue. One of the limitations, and I will be very interested in Dr. Arango’s thoughts about this too. One of the limitations of the psychological autopsy method, which has been shown to be extremely reliable incidentally using studies – in studies where a family member was asked to respond to the questions of a psychological autopsy and then the person about whom the questions were being asked was asked the questions directly. There is a tremendous amount of concordance. But without the benefit of being able to examine the person, it is very difficult to be 100 percent confident.

It is interesting too because in other parts of the world, for example, in China, about 50 percent of people who die by suicide reportedly do not have a diagnosis that can be detected by psychological autopsy. That may be because there are cultural influences on the manifestation of psychiatric disorders in other settings or it may be that it is just a different phenotype. There is a lot to learn there, I believe.

We have a question for Dr. Arango. What do psychological analyses of suicide survivors regarding the time around the event tell us about the stresses on these individuals and their emotional states at the time that they commit the attempt?

VICTORIA ARANGO: First of all, because you are a family member of the person who just died by suicide, you are at risk already. You have the vulnerability that will make you more prone to be able to die by suicide later.

MARIA A. OQUENDO: If I might just add to that as well. One of the most perplexing results that we had in some of our work not around suicide death, but around suicide attempts was that the impact of life events relative to the impact of recurrence of depression is quite small. For example, in a study that we conducted with about 400 individuals where we followed them over time, we found that a recurrence of depression increased the risk for suicidal behavior 13-fold whereas the occurrence of a life event increased the risk by 10 percent, a very small fraction.

It is interesting that from a cultural point of view, the perception is that the precipitants are the overwhelming precipitant for suicidal behavior. But the data suggests that that is not accurate. In fact, when we first saw this, we wondered whether part of what was happening was, if you will, an attempt to deal with the cognitive incongruity that here a person has tried to kill themselves or is dead by their own hand. If any of us think back to the last three months, we could come up with something very stressful that happened to us and that it might just be sort of a way of trying to explain the behavior retrospectively when people are asked. But, of course, unfortunately, we do not really know the answer to that.

I am not seeing any other questions. I am thinking that perhaps we can – we have one more minute, but I am going to give people another minute and we will have an 11-minute break instead of 10 minutes. We will see you back here at 4 o’clock. Thank you very much for your participation.


Agenda Item: SESSION 3: Hyperarousal and Inflammatory Markers of Risk

THOMAS C. NEYLAN: Welcome to session 3, that will be hyperarousal and inflammatory markers of risk.  Dr. Aric Prather from UC San Francisco will lead off, followed by Madhukar Trivedi, at UT Southwestern.  I'll give a presentation in the third slot, and then at the end, Steve Woodward from the National Center for PTSD and VA Palo Alto Health Care System will present.

I will go ahead and turn this over to Dr. Aric Prather.

Agenda Item: The Role of Stress, Inflammation, and Immune Function in Sleep

ARIC A. PRATHER: Thank you, Tom.  It is such an honor to be included with such esteemed colleagues.  This is such obviously an important topic, and hopefully we can discuss ways to move forward throughout the next two days.  What I'm going to be talking about is the role of stress, inflammation, and immune function in sleep.  A lot to cover in 15 minutes, but I think there's some interesting synergy around all of these pieces where we can kind of begin to think about what that means for depression as well as suicide more specifically.

First, the idea that sleep is related to immune function, it is.  Sleep is actually intimately tied to how our immune system works.  Both sleep as well as circadian biology help regulate the two key branches of our immune system, both our innate immune system as well as our adaptive immunity.  And you can see that here.  This is individuals that had serial sampling of immune cell number over the course of two days and nights and underwent sleep deprivation during this period, and the changes aren't necessarily exactly meaningful in the context of this talk, but certainly where those stars are, are differences in increase or decreases in what is observed in the periphery under sleep deprivation.

But you can still see that despite that sleep loss, they still show the circadian variation, suggesting that both are kind of working in regulating how any cells, in this case, are in the periphery or in lymphoid tissue, if not in the periphery, and this also has impact on function.  So there's been a lot of work on using experimental sleep loss in humans, but also looking at habitual short sleep duration as well, to look at how that is related to how the immune system is regulated.  And I've highlighted down below there an excellent review by Luci Besedovsky and colleagues that is incredibly comprehensive, so for people who just really want to know more, I would point you towards that source.

But in general, things that we see when individuals experience acute sleep loss or report habitual short sleep in some epidemiologic settings, we find a downregulation in how well our natural killer cells work.  These are important cell types that help surveil the system and kill antigens, particularly viruses.  We see a reduction in T cell proliferation when challenged in vitro, so T cells of course, as we know, as we're learning everything about how vaccines work, T cells are critical in mounting that immune response and creating an army to clear whatever is in the system.  

So when people don't get the sleep they need, they often don't build that army as effectively.  There's an impairment in antigen presenting cell function, so critical in how -- if an antigen is brought into the system, that information needs to be transmitted to the rest of the immune system, and so there's an impairment there.  

In studies that we've done and other people have done, have found that individuals that don't get a sufficient amount of sleep, either in the laboratory, or again, measured by wrist actigraphy and averaged across time, and then vaccinated, tend to mount fewer antibodies to that vaccination.  And then finally, in studies where we look at susceptibility to infectious illness, if we measure people's sleep over time and get an estimate of how much they're sleeping generally, and then actually expose them to a live virus, individuals in our hands, at least, have found the people who sleep six or fewer hours a night are about four times more likely to get sick when exposed to a known quantity of virus, in this case the rhinovirus, compared to people who sleep seven or more hours per night.

But in addition to those aspects of the immune system, one area where sleep seems to play an important role in regulation is with respect to inflammation.  We've heard a little bit about inflammation already and its role, and we'll hear more in subsequent talks, but the idea is that alterations in sleep can drive changes in the canonical stress pathways that regulate the immune response.  So this is sympathetic nervous system activity and the hypothalamic pituitary adrenal axis, can drive changes in gene expression, which ultimately lead to elevations in the secretion of proinflammatory mediators.  

So this can be interleukin-6, tumor necrosis factor alpha, and then from the liver, C-reactive protein.  These proinflammatory cytokines can actually then work up on the brain, via seeping in through the blood-brain barrier, but also primarily through its action on the afferent arm of the vagus nerve.  So in general, when we look at sleep disturbance, we often see an increase in the proinflammatory mediators of again IL-6, TNF, IL-1, and CRP.

There have been a variety of studies that have looked at this, so this is just one older one, looking at inflammation in people with insomnia and compared to controls, and again, serial sampled across the day, so the white boxes are the people with insomnia, and you can see that there is this circadian variation, but there's this uptick in people with insomnia in the early evening, having more of that, in this case, interleukin-6, IL-6, in the periphery when measured.  

Under cases where people undergo sleep deprivation, you find kind of this exactly what I was mentioning before.  This is a study from Monica Haack, where they actually randomized people to either 12 days of eight hours of sleep or 12 days where they only got four hours of sleep.  In this case they were only allowed to sleep from 11 p.m. to 3 a.m., so kind of getting up really early.  What they found is compared to these eight-hour sleepers that there was this significant increase over time in interleukin-6 and C-reactive protein, and again, as I mentioned, this also is seen at the genetic level.  

In a study where it was just one night of partial sleep deprivation compared to baseline, out of Mike Irwin's lab at UCLA, they found that individuals who underwent this partial sleep deprivation -- in this case, these individuals were not allowed to go to sleep until 3 in the morning, so having that late night.  And then they had their blood measured in the morning upon awakening, and they see this increase in mRNA expression of IL-6 and tumor necrosis factor.

Now there's been a ton of studies that have looked at this relationship between sleep and sleep disturbance and inflammation, so we have a really large meta-analytic review to look into, and here Irwin and colleagues did a meta-analysis -- now it's five years old -- but they were able to look at 72 studies and over 50,000 participants, and on the whole, individuals who report more sleep disturbances, so this is in symptoms or based on questionnaires, show an elevation in systemic levels of inflammatory mediators, in this case C-reactive protein, and the same is true for IL-6.

Interestingly, studies of acute sleep loss are a little bit more mixed, so in a meta-analytic review in this case there wasn't a significant effect size that supported the idea that taking people's sleep away acutely drove increases in some of these proinflammatory mediators, but that there's a lot of heterogeneity in how those studies were done, and many people do find these elevations under acute settings.

Sleep disturbance is key.  It's key to how we go through the day and how we feel generally and certainly with relation to our health and our levels of inflammation based on these reviews, and of course a lot of things contribute to sleep disturbance.  So if someone's sick, that absolutely can impact it, fatigue, pain, medications, and substances, and then stress, of course.  There's a whole insomnia literature based on the idea that stress drives more sleep disturbance.  And mental health, as we've already heard -- sleep disturbance, insomnia, hypersomnia, are often baked into the syndromes of psychiatric illness, including depression.

So this is relevant because inflammation, which is increased during sleep loss or sleep disturbance, also plays a regulatory role in how well we sleep.  So you can see here that at low levels it can lead to increases in non-REM sleep, a suppression in REM sleep as inflammation increases, but also can really fragment sleep in general, sleep architecture more generally, when inflammation is really high.  And this is relevant because inflammation has been implicated in the development of depression, as we all know, and from my view started with the macrophage theory of depression published in the early 1990s in Medical Hypotheses, but now there's a model that we've already seen today focused on this kind of interrelationship between the brain, the nervous system, and its impact on immunity, which then drives increases -- so in addition to stress, I think poor sleep can also contribute to this kind of recursive pathway.

We see where all these nodes are interrelated, but we do have the opportunity to test these relationships and one of the ways in which you can do that is via medications.  So we heard about this earlier today, and actually Peter Franzen, who spoke on a different topic has also looked into this, these relationships, looking at what's called cytokine-induced depression.  One way in which to do this is to treat people with IFN-alpha therapy.  As mentioned, this is a treatment that's often used in treating hepatitis C and some cancers, and it routinely leads to people developing clinical depression.

When I was at the University of Pittsburgh, now at UCSF, we recruited and treated 95 patients free of clinical depression at baseline and then tracked them over the course of their treatment, measuring levels of inflammation, measuring their global sleep quality and depressive symptoms, to try to disentangle some of the temporal relationships that might be occurring and the role that sleep may be playing in their development of depression, which would seem to have some implications for suicide risk.  

If we looked just at the role of inflammation in these patients, you can see that as we tracked inflammation out over four months, we could see that the inflammation routinely got higher in the people that ended up developing depression, right?  So that this IFN-alpha therapy drives the increases in inflammation, and those are the people that get depression, as expected.  And then people who had higher inflammation at the beginning of the study were also the ones that ended up having the worst sleep over the course of this treatment.  So those things were interrelated.

But then we were able to do some time-lagged models to look at this relationship, and I'm just going to show you kind of a picture of the significant effects that we found.  First, we found that systemic inflammation over the course of this four-month study predicted how bad someone's sleep was and how high their depressive symptoms are the next month, controlling for how they were doing the month prior.  So it seemed like systemic inflammation was driving these associations in a temporal fashion.

At the same time, poor sleep quality, their sleep quality, predicted how badly they were experiencing depressive symptoms the following month.  So though systemic inflammation seemed to be driving the poor sleep quality, poor sleep quality seemed to be predicting the depressive symptoms in this temporal fashion.

And in the same way, there was people who had more depression or their depressive symptoms predicted how much higher their inflammation got the following month.  So you can see how someone can get caught in the cycle and from my point of view, as a clinician, there's a lot of opportunities for intervention, right?  So we can kind of work on treating different aspects of this to improve the health and wellbeing of these individuals.

And there is some evidence that sleep is a vulnerability factor.  So another model for doing this is actually treating people with endotoxin, which we can do in healthy adults, and we can see that this increases IL-6 in sleep disturbed and not disturbed people, but depressed mood only goes up in individuals with sleep disturbance, despite everyone getting the endotoxin, all getting IL-6.  So obviously the story is still complicated.

And interestingly, there was also in that particular study a stronger effect in women.  So that IL-6 was related to higher levels of depressive mood, but only in -- but it differed by sleep disturbance, but it was only true in the women in the sample.

So what about sleep, inflammation, and suicide?  I know I only have a couple minutes here, but we have tried to look at this using a dataset curated called the Netherlands Study of Depression and Anxiety, where Brenda Penninx is the PI.  We are able to look at 2,300 people, all of which had either current depression or anxiety disorder or had remitted in terms of depression/anxiety.  We had a subset of course that had suicide attempts, and suicidal ideation.  They had blood measures to look at inflammatory markers and sleep duration and insomnia.

This is work done by Michael Dolsen who now actually is a postdoctoral fellow with Tom Neylan.  So it's good to keep it all in the family here.  What do we find?  We find that in people that were reported long duration actually -- that longer sleep duration was associated with suicidal ideation, which is something that we've heard about already today in adolescents.  Insomnia was associated with higher rates of suicidal ideation.  Higher inflammation, at least with respect to IL-6, was associated with suicidal ideation.  And then short sleep was actually associated with suicide attempts.

To the extent that we could, we did a mediational model and found that there was a significant indirect effect of longer sleep duration associated with higher levels of IL-6, which then predicted suicidal ideation.  You know, the why long sleep duration is an interesting thing that maybe we can chat about later, but also it seems to be coming up in the literature just with respect to suicide more generally.

So what can we do?  I mentioned we could all kind of try to tackle this complex thing, and one way to do this is with anti-inflammatory interventions, and there's growing evidence that there may be something called an inflamed depression, and so this was one trial that got a lot of press with respect to using infliximab, which is a TNF-alpha inhibitor, where they found no improvement in individuals relative to placebo in the depressive symptoms, except in people that already had elevated levels of inflammation, in this case CRP above 5.

So when they broke it down that way, based on this classifier, they saw that infliximab did lead to reductions in depression in that group, and more interesting relative to this is they also saw a significant improvement in suicide, suicidal ideation, when stratified this way.

And so this more recently, using the same dataset, they've also seen that it seemed to have a specific improvement on sleep.  So these are people in treatment resistant depression.  A subset of them actually underwent polysomnography, and they find that there's an improvement in WASO, wake after sleep onset, and a reduction in stage 2 sleep, and the extent to which they had an increase in WASO and an improvement in sleep efficiency -- or decrease in WASO and improvement in sleep efficiency seemed to be associated with how much change they saw in the inflammation that they were specifically treated with the infliximab.

So in summary, sleep plays a critical role in regulating immunity.  It's associated with elevated levels of inflammation, including proinflammatory mediators implicated in depression.  There's some evidence to link sleep, inflammation, and suicide, but more work needs to be done.  It's a very small literature.

And there is this idea that maybe anti-inflammatory treatments are part of the story.  It would be great to see a study that looks at anti-inflammatory treatments in concert with, say, cognitive behavioral therapy for insomnia or something like that, but to actually reduce some of these depression, as well as suicidal ideation.

So with that, I really thank you for your attention, and I am now going to turn the stage over to Dr. Trivedi from UT Southwestern.

Agenda Item: Inflammatory Profiles in Adolescent Suicidal Behaviors

MADHUKAR TRIVEDI: Thank you very much, Aric, and I'm going to talk a little about the inflammatory profiles in adolescent suicidal behaviors.  I have been an advisor consultant to these companies, but I will not be talking anything about this work.

So the real issue is clearly very clear, that the rates of suicide are increasing.  There is a link between insomnia and suicidal ideation, and there are promising biomarkers, especially immune dysregulation, and I think I'm going to focus primarily on the idea that we know there is inflammation associated with both insomnia and depression and suicidal ideation, but I think we have as a field, need to be now working on figuring out much more clearly the mechanisms and the real basic approaches to trying to understand which particular cell types, and I'll describe the work we are just about in the middle of right now.

So everybody agrees that suicide is increasing, especially in teens, and that is a major problem, but what we are lacking, and that's a different set of work, and that is the precise prediction of timing of suicidal behavior.  So as people have already mentioned repeatedly today, there are very well-identified risk factors, but they are not proximate risk factors.  So therefore if somebody has a risk factor, gender, et cetera, we can't suddenly change that.  So our work is cut out for us to try to figure out what we can do, by doing some objective testing that can help us predict immediate future events so that we can actually gear up to do something.

I'm going to talk briefly about the clinical aspects and then really go to the immune profiles itself.  We have been looking at this idea of insomnia, suicidal ideation, across many studies.  We have done large studies called EMBARC and CO-MED, but the two questions we asked were: is insomnia associated with suicidal ideation in the same visit, as in is it concurrent, and more importantly, so to speak, a modifiable risk factor, do early changes in insomnia predict subsequent levels of suicidal ideation?

In both these studies, repeated measures, mixed model analysis, we evaluated this association for concurrent association, and we find even that accounting for age, sex, race, and ethnicity, that insomnia was significantly associated with suicidal ideation in the same visit in both studies, but more interestingly, when we looked at the change in insomnia between baseline to week 2 and looked at their association with levels of suicidal ideation from week 2 to week 8, you find that this reduction in insomnia in these first two weeks is clearly associated with lower levels of suicidal ideation at subsequent visits in both these studies.  So this is really showing the temporal relationship with changing improving insomnia early on that could have payoff even within the following six-week timeframe.

We then looked at irritability and the association with suicidal ideation, and you see actually that irritability has a strong correlation with suicidal ideation in three large studies conducted over the years, and anxiety, insomnia, and suicidal ideation, depression and suicidal ideation.  Actually, in our studies at least in these three studies, and this was published last year in Neuropsychopharm, irritability and suicidal ideation, even for accounting for depression severity, continues to have a strong correlation, and this relationship is actually stronger for irritability than all the other factors we looked at.

So the other issue with, same thing like we did earlier for insomnia, when we looked at the change in irritability over the short timeframe from week 0 to 2, and looked at the relationship with changes in the suicidal ideation week 2 to week 8, we see the same pattern in all three studies, and the early improvement in irritability is strongly associated with eventual improvement in suicidal ideation in the subsequent six weeks.

So both of these clinically are beginning to show us in addition to the data we have from irritability, and these results, at least in relationship with irritability and suicidal ideation, we were able to show even after controlling for changes in anxiety and insomnia.

This is not unique to depression.  We've seen this even with stimulant use disorders, that study we completed where the relationship persists even in people who are brought in with stimulant use.

So finally, moving to inflammatory profiles, there is now good evidence that there are some very promising biomarkers in depression at least that have looked at reward circuits, emotion processing circuits, inflammation, metabolic and gut microbiome.  I am going to focus for the next five minutes, on this inflammatory markers with depression and suicidal ideation.

There is actually good evidence for inflammatory biomarkers that show high rates of depression in those treated as we've seen with interferon and LPS injection leads to depression behaviors.  High CRP levels associated with greater risk for not only depression but hospitalization and suicide, and these anti-inflammatory medications could improve, although I have to confess, the humbling reality is that all the efforts, including the data shown from Chuck Raison, that is a very small sample that they looked at as a secondary analysis from the infliximab study.  All the other attempts at really using randomized controlled trials with anti-inflammatory medications have not paid out, I think partly because we have not actually explained the exact nature of the inflammatory dysregulation that needs to be sorted out.

We have seen and published on this from our CO-MED data showing clearly that if we -- we have randomized people to an SSRI alone, escitalopram or bupropion plus escitalopram, and if the people -- if we had in hindsight done this by matching them with the CRP levels, you can see the difference.  Our overall remission rate was 31 percent.  If we had just matched them with CRP level and in our hands 1 milligram is a pretty good tertial, then our remission rates actually from the sample would have been 20 percent higher if we had just matched them with just CRP levels, although CRP as we all know is a nonspecific marker of inflammation.

So in order to tease this further, we looked at insomnia and immune dysregulation together in a population we have from a high-risk adolescent -- so suicidal teens who attempted suicide and then as was mentioned earlier in the diversion program, they are into an intensive outpatient program.  And we created to canonical correlation so to speak, a proinflammatory cytokine and chemokine score from these selected cytokines and chemokines, and we find a very strong correlation between this composite score and insomnia score.  So as the inflammation goes up, their insomnia is seen to be high.

In a similar sample comparing teens who have attempted suicide with those who are healthy controls and those at risk for depression, that is they have a family history of depression, we compared their immune profiles in the 40-plex assay and found IL-4 as a key marker for distinction between suicidal kids at risk and healthy controls.  This initial study from our SPARC study actually showed these recent suicidal behavior had markedly lower levels of IL-4, a key cytokine for T helper, T cell helper types, in the immune response, and this underscores to us that immune dysregulation and suicidality may involve dysfunctions in adaptive immune response, and IL-4 levels may signify an exaggerated Th1 immune response and potentially also an autoimmune phenomenon.  We measured autoantibodies in this sample, and they were also affected.

So based on these, we have now embarked on this large big sky grant from AFSP to try to tease out the exact nature of the immune dysfunction and there we are really trying to match the idea that we can look at the aberrant immune cells themselves so that we can eventually find therapeutic targets to monoclonal antibodies.  We will look at autoimmunity and also look at the chronic immune dysregulation.  The autoimmune function actually is very interesting because that can be used as a diagnostic test.

Obviously, all the other components that have been mentioned all day, including stress, early life trauma, et cetera, does impact suicidality.  So this ongoing study is really bottom line is we are going to study adolescents with suicidal behavior, at risk adolescents, and healthy adolescents, and develop a very elaborate immune profile to cell phenotyping, immune phenotyping, cytokines, autoantibodies, and an array of clinical measures that will define and we would follow them for 12 months, and this study is ongoing.  We have extensive assessments of these.

The most interesting part of this is, unlike the sort of global immune profiles that we have seen so far in general with obviously exceptions, we are using a mass cytometry profile using intracellular cytokine staining and/or cell surface proteins on PBMCs.  We will look at -- this is a variation in flow cytometry.  These are used to determine types of cells, T and B cells, Th1 versus Th2 versus Th17.  The antibodies are labeled with heavy metal ions rather than fluorochromes, and the metal signatures are then analyzed with time-of-flight mass spectrometer.

This basically is a quantitative way of assessing the immune dysfunction in people from PBMCs, and a very early pilot data example of this with cytometry, mass cytometry, can show for six healthy controls and 12 teens who have attempted suicide in the recent past, these allow us to visualize clusters of cells.  So if you can see, you can see that between healthy controls and suicidal teens, you have similar cell distribution with some degree of difference in density, but still they are there.

But in healthy controls, you see the cell types that are present that are absent in teens who attempted suicide.  So these can be used then to measure cell counts and identify druggable targets from this.  So that is sort of the work that we are engaged in right now.  This is early part.  We have 12 suicidal teens and 6 healthy controls, but more sample is getting accumulated.

Finally, we also looked at the idea that this ultimately also reflect in functional circuits, and so this is from 72 subjects who are in our longitudinal monitoring study.  We have for deep -- for endophenotyping that included 53 females out of 72, and we used a suicide assessment instrument in order to map neurocircuits onto these teens, and we find that this functional connectivity map where we used a similar approach we used in our American Journal paper for the full sample of depressed population, and here you see that network between striatum and the dorsal media network are -- they form a network, really is the more dominant network, and this is really associated with rumination in our sample here, and otherwise also and could again be a target for modification with something like trans-magnetic stimulation.

So I think I'm going to try to finish here in the interest of time.  It is very clear that inflammation plays a role, but I think we have to be really identifying the exact nature of this dysfunction.  In the future, we will use youths with depression with no history of suicidal behavior, and the idea is to develop precision approaches rather than just settle on saying there is inflammation, which sounds like it is obvious.

I'm going to stop here.

Agenda Item: Sleep, Hyperarousal, Posttraumatic Stress Disorder

THOMAS C. NEYLAN: Thank you, Dr. Trivedi, for getting us back on time.  It's a real pleasure for me to be presenting to everyone here, and I thank the organizers for inviting me.  I am going to really talk about how sleep can be an early target to prevent the development of PTSD and other post-trauma disorders.  The broad area of PTSD and hyperarousal and PTSD and sleep, you can get -- there's a really nice review article by my colleague Dr. Anne Richards in Neuropsychopharmacology, and I'll join the chorus of people subtly complaining about 15 minutes being too short.

But the key thing I want you to see in this slide is that, first of all, sleep disturbances are multivariate, you know, short sleep duration impairments in REM sleep as well as non-REM sleep, a key thing here is that sleep disturbances and anxiety and fear can have a bidirectional relationship.  It could be that because fear and anxiety recruit additional systems that affect arousal systems, for example, corticotropin releasing factor or immune cytokines that can impair sleep, and once you've got a feed forward loop, you've got a model of the engine of chronicity, and it suggests that interfering in that feed forward loop is actually a key thing to stave off the prolonged or disturbed biology and the accrual of comorbidities that so frequently happen to people who've been traumatized.

In the setting of the sort of feed forward cycle, you can see how various facets of brain health also add to the chronicity.  For example, frontal deficits from impaired sleep can impair efficient emotion regulation, and accurate appraisal of fear and threat, as well as impairing memory, which is critical for the recovery function.  You know, sort of adequate extinction of learning.

And there's a variety of possible mediators, some involving the known catechol and indolamine systems and orexin systems that are part of normal healthy wake arousal, as well as the additional recruitment of additional factors such as corticotropin releasing factor or cytokines.

But what I want to show you here is a model that's been very influential in the sleep field.  Everyone in the sleep field recognizes this four-factor model, originally started by Art Spielman, and it's essentially a stressed diathesis model that Dr. Oquendo referred to also in a different setting, and that is that prior to some precipitating event, there's variability in people's vulnerability to developing sleep problems.  In this case, the model here is insomnia.  

And then something happens in the environment, something happens in the person's life, there's a precipitating stress, and then what emerges are these perpetuating factors that can help become engines of chronicity that in the world of insomnia can be behavioral changes, spending more time in bed or using more caffeine or changing the timing of sleep, and then over time, there can be additional conditioning factors such as learning that the sleeping environment is associated with frustration and arousal, and that becomes kind of conditioned over time.  And the focus of behavioral therapy for sleep often goes after these perpetuating factors.  

Well, this same model also can be used to try to understand how sleep can be a driver of the accrual or the development of post-trauma psychiatric disorders such as PTSD and major depression, as well as other disorders, and that is that you can, in the setting of trauma, that's the obvious precipitating event, and then there can be these perpetuating factors that are more than behavioral, but they can be sort of the recruitment of these additional mediators such as inflammatory cytokines or corticotropin releasing factor that help maintain or perpetuate the disorder.

Of course in the theme of the conference is that with the cascade of development of PTSD and depression, that downstream we expect that there are high rates of suicide, which of course is what we're hoping to prevent by interfering with this process very early.  So in order to try to understand that, it's really helpful to have careful longitudinal data.  The NIMH has funded this U01 consortium grant called AURORA, which is recruiting, hopes to recruit, 5,000 emergency room patients seen in a consortium of emergency departments across the United States, and they recruit people who have had a traumatic event happen to them that are not so injured that they're going to be hospitalized.  

They're going to be discharged from the emergency department, and they provide consent in the emergency department.  They give some initial retrospective measures of mood and PTSD symptoms, as well as sleep symptoms.  They're given a wearable device, as well as a smartphone app where they get queried periodically for symptom distress, and the idea is to follow these people over a year, take a subsample in for deep phenotyping with imaging, functional imaging and startle testing, and again, the idea is to try to get a very sort of granular longitudinal view what develops over time, and the focus is on not just our psychiatric disorders, posttraumatic stress disorder, depression, post-concussion pain, but also RDoC constructs, sort of that thinking that people won't fit in these neat bins, can we characterize the -- are there clusters of people who show differences across these different continuous domains that are highlighted in RDoC.

So for the purposes of this meeting, I want to present some data, initial data from AURORA that was recently published, where the focus is on that retrospective report of sleep problems reported in the emergency department, and the development of PTSD and depression at two weeks and eight weeks.  The method here is that we actually restricted the sample to people who had had a motor vehicle accident.  The primary outcomes were PTSD and major depression at both two weeks and eight weeks post-emergency departments, and our primary predictors were, one, insomnia; two, nightmares; sleep duration; sleep stress reactivity, a measure developed at the Henry Ford Hospital where they asked people to what degree do they normally have sleep problems when they're faced with life stressors, and then finally daytime sleepiness.

There was going to be a host of covariates or control variables, including pre-emergency room or pre-motor vehicle accident depression and PTSD as well as acute peritraumatic distress, as measured by the peritraumatic distress inventory, and then demographics including socioeconomic status, and then very detailed characteristics of the motor vehicle accident itself, including were you injured, were there other people in the car, were they injured, were there people who were killed in the accident.  These were all entered in as control variables.

And to show you the primary results, one is that we found very high rates of PTSD and depression at two weeks and eight weeks.  We used the same measure, the PTSD checklist, PCL, at two weeks and called it acute stress disorder, and the prevalence of acute stress disorder was 41 percent at two weeks and PTSD at 42 percent at eight weeks, and depression was 30 percent at two weeks and 27 percent at eight weeks.

And the key take-home point that I want you to see from this data is that the conditional probability, if you had one of these, if you had two-week major depression, your conditional probability of having the others at two weeks or at eight weeks is extremely high, double digits odds ratios for the most part.

And then finally what I want to show are sort of the punchline results from this study.  So again, the predictors are insomnia, nightmares, sleep duration, daytime sleepiness, and the sleep stress reactivity.  We created a composite variable of all five of those, and what you can see is that three of them were important for predicting eight-week PTSD.  That include insomnia, nightmares, and sleep stress reactivity, not sleep duration or daytime sleepiness.

It was visible or apparent at two weeks, acute stress disorder, where the odds ratio were essentially the same, and then when you add peritraumatic distress to the models, again, all the models have pre-depression, pre-motor vehicle accident, depression and PTSD demographics and trauma characteristics.  When you add peritraumatic distress, it really didn't change the relationship of pre-MVC sleep problems to an eight-week outcome.

And then when you -- that's true if you add peritraumatic distress at two weeks or eight weeks.  But if you add two-week acute stress disorder to the model, this last model that you're seeing right over here, M5, the relationship -- I'm sorry, I'm going to back up, the relationship disappears.  So what it suggests is that the relationship of pre-ED sleep problems occurs very quickly in the first two weeks, and then from two weeks to eight weeks, there's no -- once you account for the two weeks level of distress, there's no relationship of these preexisting sleep problems to eight-week PTSD.

But if you go to depression, it's a little different.  The predictors that showed to be significant were nightmares as well as sleep stress reactivity, and interestingly enough not insomnia, as has been shown in a host of other studies, and I'll come back to that.

But if you look at two-week major depression, there was no effect of these sleep variables that significantly predicted higher rates of depression, but it emerged at eight weeks, and what emerged there was nightmares and sleep stress reactivity in the composite, and then once you put in two-week MDE, the composite variable survives.  There's a weak effect of longer sleep duration on protecting people from eight-week major depression.  You see that at eight weeks but not at two weeks.

But the key thing here is that the relationship of pre-trauma sleep problems didn't affect the early phase response for depression, but the transition from two-week, eight week, the development of the major depression after that acute two-week period.

So what do we do with this?  So we then took all of this data and developed these population attributable risk proportions models, which was modeled to predict the proportion of observed cases of major depression and PTSD that we saw at eight weeks that were driven by these sleep predictors, and if they were all causal, if they all had a causal role and you could normalize them, the PARP model showed that you could actually reduce major depression and PTSD at eight weeks by one third.

So that's really exciting, but it's still observational data, and it's also based on the assumption that all these sleep variables have a causal role.  So where we're going with this is to say that we are -- now we want to see if we can prevent eight-week PTSD and depression by an early intervention for sleep, and we're going to use Dan Buysse's Brief Behavior Therapy for Insomnia, and I can see that Dan is actually an attendee in this conference.  We'll have a control group with progressive muscle relaxation.  We are starting to pilot this, but the goal here is to sort of say let's deliver the proof that insomnia and sleep problems have a causal role in the development of these disorders.

We can't determine that through observational data; let's do it through an experimental intervention and see if it becomes an effective secondary prevention strategy for preventing post-trauma PTSD.  Again, the idea is to try to interfere with that early sort of feed forward cycle, and to stave off the development of these severe disorders.

So I'm going to stop there.  I just want to thank numerous colleagues that I have worked with over the years, including Sabra Inslicht and Anne Richards, Shira Maguen, and Aoife O'Donovan, and then the last on this list is Dr. Steve Woodward, who actually is giving our next talk.
Agenda Item: Nightmares, Heart Rate Variability, and Hyperarousal

STEVEN H. WOODWARD: Thank you, Tom.  Good afternoon, everyone.  My name is Steve Woodward.  I want to thank Drs. Leitman and Bernert for inviting me to participate in this workshop.

I work in the National Center for PTSD, where I've been concerned with nightmares and autonomic regulation during sleep, and how to measure these outside of the laboratory in the home for extended periods, because like suicidality, nightmares have a low base rate.  I will be discussing nightmares and heart-rate variability, but I want to briefly describe my journey to this workshop, which began with a single case.

This slide plots sleep heart-rate obtained from a study participant over the course of a three-month inpatient hospitalization for PTSD.  This veteran was participating in a study of a completely different topic, and had provided consent for the recording and use of his sleep data for the duration of his hospitalization.

The very last data point represents the night prior to a serious but thankfully unsuccessful suicide attempt.  The first data point is the night of his admission to the program.  At that time, this slender athletic man exhibited the sleep heart-rate of about 57 beats per minute.  Two nights before his attempt, however, his sleep heart-rate had risen to 81 beats per minute.  On the night immediately prior to the event, it then fell to 61 beats per minute.

Along the way, his sleep heart-rate exhibited a series of jumps, plateaus, and inclines, none of which could be explained by prescribed medications, nor did he ever fail a tox screen or gain or lose significant weight.  This was a completely serendipitous observation.  It was only when reviewing study data later that we saw this striking pattern and following up learned that it was from the veteran who had earlier attempted.

Needless to say, these data made a big impression on us, suggesting first that cardiac autonomic regulation might be altered in suicidality and, second, that our efforts to use long-term low-burden sleep recording to capture nightmares might have applications in the area of suicide.  But this is just a single case.  What's the literature tell us?

So I'm going to propose that we wade into the literature considering nightmares, PTSD, and suicide risk by considering their overlaps, what we might put into the center of this Venn diagram and why.

The why is because nightmares, often called the hallmark of PTSD, are also a major risk factor for suicidality associated with something like a fivefold increase in risk for suicidality, independent of the risks conferred by insomnia, anxiety, and PTSD.  In considering possible overlaps, let's start with risk factors.  It's well-known that early adversity represents an enormous risk factor for both adult PTSD and for suicidal ideation and behavior, as has been discussed in prior talks.  I'm not going to spend much time on these associations, because they're really ironclad.

It was eye-opening for me to find that Tore Nielsen, whom I regard as the leading researcher on nightmare disorder working today, has proposed that early adversity is also a risk factor for nightmare disorder, which is something like the red-haired stepchild of sleep disorder taxonomy.  But importantly, the adversity that Tore points to occurs prior to age three-and-a-half, roughly the age before which experiences do not make it into declarative memory and become reportable by the adult.  So I'm going to propose that we insert early adversity as the first anchor of our overlapping entities.

Now Tore invokes this stress acceleration hypothesis of Callaghan and Tottenham, which I'm going to put in a quick plug for.  This hypothesis references all the usual suspects constituting the core fear-generating and regulating circuitry, the amygdala, hippocampus, anterior cingulate cortex, et cetera, but then goes on to propose that early adversity results in accelerated but prematurely terminated maturation of these systems.  To me, this is a beautifully parsimonious framework for organizing developmental traumatology.  End of plug.

If you'll allow me, I'm going to focus on the anterior cingulate cortex as the point structure of the larger fear-regulatory network that is impacted by early adversity.  I think this is supported by both the neuroanatomy, which includes monosynaptic connections between anterior cingulate and inhibitory interneurons in the amygdala, and by well-replicated functional connectivity finds.

We can then ask: are there shared alterations of anterior cingulate cortex across PTSD, suicidality, and nightmare disorder?  The first of these are pretty easy.  In fact, reduced anterior cingulate cortex thickness and/or volume have been more consistently replicated even than smaller hippocampal volume in PTSD.  In suicidality as well, as reviewed by Auerbach et al, a majority of studies have reported smaller volume in this brain structure.
What about nightmare disorder?  There have been only two very recent studies addressing this question, but the results are in line with this framework.

Shen et al examined resting state fMRI in 15 persons with nightmare disorder and 15 controls and found a focus of excess regional homogeneity of BOLD fluctuations in anterior cingulate cortex.  ReHo, as it is sometimes called, is a relatively new measure, and excess ReHo has not been unequivocally associated with any neuropathological signs, but it has been observed in motor cortex in Parkinson's and in association with excess tau deposition in a rodent model.

Just this year, Marqui et al replicated this finding of elevated ReHo in anterior cingulate cortex in nightmare disorder.  So I'm going to add anterior cingulate impairment to the overlap region among nightmares, PTSD, and suicide risk, provisionally.

Now, an enormous literature has tried to define exactly what the anterior cingulate cortex does.  Many of us in PTSD think of it primarily as regulating the amygdala, but there is a much broader perspective from cognitive neuroscience.  Depending on the paradigm, the ACC appears to play a key role in error detection, in emotion regulation, or in arbitrating among competing responses.  A superordinate construct that has been proposed to organize these aspects is cognitive control.

Cognitive control is defined as the ability to maintain adaptive cognition in the face of interference from exogenous or endogenous distractors.  The latter would include strong emotions.  The most common of psychological tests employed in this domain is the Stroop task in which a person is asked to report the color of a printed word while ignoring what the word says or means or the feelings it evokes.  In emotional Stroop tasks, the word might say ambush, or Charlie, or rape, or funeral.  Interference with color naming is then indexed by response delay.

So is increased interference on emotional Stroop tests common across these entities?  There have been a lot of studies in PTSD patients using the Stroop, and while not monolithic, this literature supports increased interference with cognition in PTSD.  Similarly, in suicidality, Richard-Devantoy et al have reached the same conclusion.  I know that Dr. Zuromski will be considering this general area in much greater detail tomorrow.  While there has been only one study in nightmare disorder, it came to the same conclusion.  Cognition in persons with this diagnosis exhibit excess vulnerability to emotional distractors. 

So let's provisionally add reduced cognitive control to our overlap region.  Let me also point out that Calati et al, 2020, have recently suggested that impaired cognitive control plays a central role in suicide crisis syndrome.  

To summarize, I've talked about overlaps among PTSD, suicidality, and nightmare disorder as regards shared risk factors, shared brain structural compromise, and shared neuropsychological signs, but I have not talked about sleep.  How does this line of argument bridge to sleep?  I'm going to propose that the bridge we need here has been provided by Thayer and Lane's neurovisceral integration model.  This model posits that the cingulate-based cognitive control is indexed by high-frequency heartrate variability.

High-frequency heartrate variability is inter-beat interval variation that is coherent with the respiratory cycle.  Heartrate goes up when we breathe in, down when we breathe out, and the magnitude of this variation is a putative signal, index of vagal tone.

One very impressive finding, to me, in favor of an actual relatively direct neural bridge between the heart the brain comes from Ter Horst and Postema.  These researchers were interested in how emotional upset could trigger heart attacks in persons without known heart disease.  They injected pseudo-rabies virus into the hearts of rats.  Pseudo-rabies virus is a retrograde tracer which crosses synapses, so these are researchers were able to track viral infection up the vagus, into the brainstem, the midbrain, the hypothalamus, and finally into the anterior cingulate cortex, where they found an excess of targets in the right hemisphere.  Below that reference, I've plotted results from our lab in which we found positive correlation between right, but not left, anterior cingulate cortical volume and amplitude of high-frequency heartrate variability in a sample of PTSD patients studied at rest.

So I'm going to add reduced high-frequency heartrate variability to our overlap region.
Now, this has two important consequences.  First, this overall trip through the data, to me, reinforces Nielsen's proposal that nightmares of nightmare disorder have more in common with trauma-related nightmares than has been previously appreciated.  Second, we now have support from the literature for an objective index relevant to suicidality that we can measure during sleep over days, weeks, and months.  

Sleep heartrate variability is a relatively low-burden measurement involving, maximally, a couple of ECG electrodes.  We'll be talking about less burdensome measures.  It is relatively inexpensive.  It is easier to measure during sleep than during waking because of reduced movement artifact, and within subject designs can sidestep the heated debates around the optimal quantification of high-frequency heartrate variability.  Going back to the case I presented at the beginning, there was a concurrent reduction in the amplitude of high-frequency heartrate variability over the three months leading to his attempt.  

So, what have studies of sleep high-frequency heartrate variability shown?  Do they support this avenue of research?  Studying PTSD patients, Miller at al contrasted nights that were or were not followed by morning nightmare reports, and found that the amplitude of high-frequency heartrate variability was lower on nights following such reports.  Studying nightmare disorder patients, Paul et al contrasted spontaneous awakenings from REM sleep according to whether or not they were associated with a nightmare.  They also found reduced high-frequency heartrate variability in the pre-nightmare samples of REM sleep.

Now, what has not been studied yet is high-frequency heartrate variability during sleep in suicide, suicidality.  There is a literature on waking high-frequency heartrate variability in suicidality that's been reviewed by Kang and is generally found to be lower.  However, some studies have found such effects only under conditions of challenge.  So this is an area crying out for more work, I believe.

I hope I've made a case and gotten some people excited about measuring this index for extended periods of time in persons at risk.  I'm going to change gears now and focus on some feasibility issues.  

The first is that high-frequency heartrate variability declines substantially with age.  By age I mean 40.  This means that the range of variation both between and within older individuals will be compressed, which means that groups looking at this in adolescents and young adults will have an easier time of it.

Second, this places a premium on precise measurements.  In the lab, ECG may be sampled at 500 to 1000 hertz and artifacts excluded using manual review, but this methodology will not scale to large samples measured over long periods.

So how good is photoplethysmography, or PPG for short, the technology that Fitbits and garments use.  The field is currently dominated by consumer-grade systems designed to minimize cost, memory usage, and power consumption, and none save the raw PPG data for review or postprocessing.  At best, the PPG waveform is also much less well-defined in terms of its peak than is ECG, making interval timing measurements much noisier.  Nevertheless, measuring high frequency heartrate variability during sleep retains some of the advantages of low movement artifact relative to waking and high data volumes.

There's a growing literature validating consumer grade PPG, but we need to be informed consumers.  Most of the validation studies in the literature have used smartphone cameras, actually.  This particular attempt to validate the Apple Watch used a Polar strap rather than ECG as the ground truth, which is problematic.  

The Oura watch gained a lot of press for its role in a number of large scale COVID-19 studies.  This paper did use ECG as ground truth and provided astounding correlations between the Oura and ECG.  I'm sorry, this is a ring, right?  Between the Oura and ECG base heartrate and high-frequency heartrate variability.  But note that the study was executed by the manufacturers.

Lastly, I'd like to draw your attention to a technology that we may look to in the future, which has some major advantages.  These are electromechanical strips that go into the bed, under or on top of the mattress.  First and foremost, they are zero burden.  Once installed, they can record sleep in the home indefinitely with no user interaction, and sending data to the web nightly.  They can transduce both heartrate and respiratory movements with some precision, depending on how much you're willing to pay.  But this technology also has a serious problem, which is that you don't always know whom you are recording.  People sleep with partners, they spoon, making their signals difficult to distinguish.  Though this problem can be overcome with advanced signal processing, it requires three or more sensors to separate the couple's signals.  No manufacturer has taken on this problem.  As of today, with a watch or ring, you know whom you're recording.

To sum up, I believe overlaps across PTSD, suicidality, and nightmare disorder, combined with the Thayer and Lane model, support recording high-frequency heartrate variability during sleep in persons at risk for suicidal behavior.  Given current technology, certain caveats must be observed, but others may be overcome in the not-too-distant future.

Thank you. 

Agenda Item: Q&A

THOMAS C. NEYLAN: Thanks, Steve.  Let's go right into Q&A.  Madhukar and Aric, could you turn your cameras on and unmute? 

Here is a question possibly for Aric or Madhukar.  The relationship between immunity and insomnia and hypersomnia, does that help us understand the differences between those two phenotypes?  Very different.

MADHUKAR TRIVEDI: That's an interesting question.  We looked at mainly insomnia.  The samples are too small to look at hypersomnia, so I don't really have much data on the hypersomnia, but Aric may have some.

ARIC A. PRATHER: It's such an interesting question.  I think in the epidemiologic literature often it's focused on long sleep duration, versus hypersomnia, clinically diagnosed, and I think that's probably an area where we could begin to gain some ground if we had really strong clinical phenotyping.  The problem with long sleep duration is it's often -- it comes out constantly in the literature being related to elevated levels of inflammation, mortality, what-have-you, but just because someone reports that they'd been in bed for 10 hours doesn't mean that they're sleeping, so I think just relying on these one-item measures probably creates so much noise that it's hard to get a toehold on how to how to move the field forward in that way, though it is so consistent that maybe there is something there and maybe we just need to do some better measurement.

THOMAS C. NEYLAN: Great. Actually, another question for Aric. Are there other aspects of the immune system that could help us understand the relationship – differentiate those with depression and suicide ideation? If so, does sleep have a role in that?

ARIC A. PRATHER: Yes. I think this came out in Dr. Trivedi’s presentation. I think we have been relying on a few biomarkers of inflammation of the immune system as a way of trying to understand this, but of course, the immune system is incredibly complex and oftentimes these studies have not embraced advanced immunologic methods.

A recent paper that came out in the Neuroimmunology, Mood Disorders, and Alzheimer’s Disease Consortium, that was published in Biological Psychiatry last year, began to do kind of deep immunophenotyping of individuals with inflamed depression – is what they were looking at. But I think using the methods that we heard about today like time of flight, mass spectrometry, kind of intercellular cytokine staining, will help illuminate some of the complexities and potentially refine some of the signals and understanding different phenotypes of depression. Then if we have strong measures of sleep, we may be able to see where it fits as a mechanism in driving those differences.

So I think it is really about better characterizing the immune system beyond just things that are easy to measure, which tend to be things like C-reactive protein, IL6, things like that. 

THOMAS C. NEYLAN: Thanks, Aric. This is a general question for anyone who might know the answer. Given the role that inflammation has during pregnancy and childbirth, is there any thought of targeting immunity or sleep to help new mothers. Protect them from having, for example, postpartum depression and later suicide ideation?

ARIC A. PRATHER: Jennifer Felder had recently, in collaboration with her, at UCSF, had just completed a trial of digital CBTI in women during the perinatal period, to address this. What she found was compared to a wait list control, it strongly improved insomnia symptoms in these women.

Surprisingly, and work that she is moving forward, actually reduced depression and anxiety and in the follow-up period, which goes out in her study - I think - to 24 weeks - she is still looking at the data, but it looks like it actually potentially plays a preventative role compared to individuals who were in the wait list control with respect to depressive symptoms and potentially probability of major depression.

Of course, postnatal depression is such an important problem to tackle, that at least in the tools that she was using, it seems like directly targeting sleep and insomnia specifically, seemed to be effective.   

THOMAS C. NEYLAN: Thank you, Aric. There is a question for Dr. Trivedi. It is a very challenging question. What is the biggest challenge and prediction about suicide behavior?

MADHUKAR TRIVEDI: Predicting immediately. So we have a real problem, right, we do not actually have any markers of approximate prediction. We are very decent enough with longitudinal prediction. But I think if someone is in your clinic already or the emergency room, whether we can predict anything for the next 24 hours to seven days – we don’t have any as good enough markers. At least I am aware of, that you can objectively use. 

That is one of the reasons to pursue this immune response and then clarify the exact profile for any given individual. That is the procedure we are trying to do.

THOMAS C. NEYLAN: Thank you. A question for Dr. Woodward. How much added value do you get combining HR and HRV measures with accelerometry?  

STEVEN H. WOODWARD: I wish we knew. I think that is a great way to go. I think the promise of both these low burden measures responsive to the question that was just asked to Dr. Trivedi. Are there markers that we can track over time that will anticipate changes in suicidal risk? 

THOMAS C. NEYLAN: Thanks, Steve. There is a question for me. Why is it that insomnia wasn’t a predictor of two-week and eight-week depression in the Aurora Study, because it has been shown in a variety of other studies? 

I think the difference is that we used pre-emergency room depression and PTSD as a covariate. We actually took those covariates out and indeed, insomnia was a robust predictor of nuance than MDE and PTSD. I think it was because we have really rigorous covariates in that study.

THOMAS C. NEYLAN: Aric, the question directed to you. Disruptive sleep and upregulated immune response are putative contributed to delirium and suicidology. Should acute suicidology be considered a manifestation of delirium?

ARIC A. PRATHER: I don’t know if I have anything to answer that question. That is a great question and I know that those things are true. I know very little about delirium, though. In fact, Tom, it seems like something that you probably know something about.

THOMAS C. NEYLAN: It is something, obviously, it becomes an issue with aging and sleep, so we are interested in that. Sleepiness is a risk factor for conversion to Alzheimer’s. Of course, what happens is people become more at risk for that. But the suicide piece to that, I can’t connect the dots there. 

MADHUKAR TRIVEDI: Acute suicidal patients don’t really have too – if you consider their wanting to attempt suicide as a cognitive distortion. Beyond that, there isn’t any cognitive disfunction.

THOMAS C. NEYLAN: Here is a question, the association of long sleep of suicidality might be very tricky to unpack. Have any of these studies accounted for sleep disorders? For example, the 800-pound gorilla, not talked about much today, obstructive sleep apnea? Is that a factor in trying to understand these sleep variables to suicide outcomes? 

STEVEN H. WOODWARD: I can take a shot at that. I would recommend people to take a look at the work of Todd Phillips in that connection. He has demonstrated relationships between obstructive sleep apnea and suicide risk that we suspect that the association between extended sleep and of course, mortality, for example, probably involves sleep apnea and non-restorative sleep. So there is probably something to be investigated more in that domain. 

THOMAS C. NEYLAN: Question for me. What about sedative- hypnotic, acute treatment of sleep problems to prevent PTSD?

That is a great question and there has been some attempts to look at that. I don’t think we have enough data to give you a definitive answer. There was a non-randomized trial published years ago, where they compared people who agreed to take a sedative-hypnotic and those that said no to it. 

Then I know that Tom Mellman did some initial work with benzodiazepines. But we actually don’t really have a great answer to that on a scale for us to say we have a clear answer to that question. There is promising data about Zolpidem, as something that might reduce suicide outcomes. That trial was recently published. I think it is an open question, actually.  

Is this the time to turn it over to David Leitman?  

DAVID LEITMAN: Yes. I am just waiting for my video to go on. Thank you for a wonderful session and to everybody else in the sessions before. I have to admit that I had a bunch of questions that I myself, wanted to ask, but I did get a good light sleep last night so consequently I am not super impulsive and I am able to exercise proper cognitive control.

I guess I would like to close since it is late, and thank you on behalf of myself, NIH, my co-chair, Rebecca Bernert. We are going to reconvene tomorrow 12 noon, EST, for the second and final day of the workshop. After a brief welcome, we will have three sessions, followed by a general discussion across all panelists. 

The sessions for tomorrow include neurocognition, learning and affect regulation. A session on novel therapeutic frameworks and intervention development. A final session on technology innovation and digital medicine in suicide: future directions. 

Please join us then. 

(Whereupon, the workshop was adjourned at 5:10 p.m.)