OK, well, I'd like to welcome everyone to our symposium, which will be focused on advances in precision medicine approaches, precision diagnostics, and precision therapeutics for alcohol use disorder, and for some of the common comorbidities of AUD, including PTSD. I'm Charlie Marmer. I'm chairing this panel. Our first speaker will be Dr. Nancy Diaz-Granados, who is deputy clinical director of the National Institute for Alcohol Abuse and Alcoholism. And Dr. Diaz-Granados will present on neurocognitive markers to guide diagnostics and precision treatment for alcohol use disorder. I will be presenting second. And I am the chairman of the Department of Psychiatry at NYU, where I direct an NIAAA-funded Center for Precision Medicine in Alcohol Use Disorder and PTSD. And our third distinguished presenter will be Dr. Nancy Sugarman, who is an assistant professor at Harvard Medical School and is a research psychologist and scientist at McLean Hospital, and will be presenting on some advances in experimental precision therapeutics using digital therapeutics for alcohol use disorder. So welcome, everyone. We also have a number of people who are joining us virtually online. This is being simultaneously broadcast online. So I want to welcome whoever is joining us virtually as well. And the first presentation will be by Dr. Diaz-Granados, Addiction, Neuroclinical Assessment, and Precision Medicine for AUD. Dr. Diaz-Granados. Thank you, guys. I might speak fast. So if you don't understand me, please let me know. I have no conflict of interest to disclose. And I don't think I need to make the case to you that in psychiatry in general, we really mix pears and apples when we diagnose someone. And it is a problem because giving a diagnosis doesn't necessarily tell us a lot about treatment and doesn't tell us a lot about prognosis. And that's something that we need to improve as physicians. I'm presenting kind of briefly two cases to illustrate these. So Mary and Carol are both 46-year-old females that are drinking a pint of vodka a day. When we look at their cases, they have very different courses. Mary started drinking very early. First drink was at age six. She remembers that drink vividly. She loved that drink. It was the first time in her life that she actually felt relaxed. She felt tension melt. She's a 46-year-old woman who remembers drinking at age six and who considers that something that kind of changed her life. She drank all her life, had a diagnosable disease by age 16. And by the time she comes to us, she has a 30-year history of an alcohol use disorder. On the other side, Carol is someone that started drinking also as a teenager but drank socially most of her life, never really had a problem with alcohol until three years prior to her presentation when she had a big loss. She lost a child. And at that point, her coping mechanism was drinking. So she started drinking daily to be able to sleep and to numb her feelings. And three years later, when she comes to us, as I said, developed an alcohol use disorder. She's also drinking a pint of vodka a day. And her life is in crisis. Her work is on the line. Her personal relationship is on the line. And she's realized that she has an alcohol use disorder and is seeking treatment. When we look at these two cases, the first patient, Mary, with a 30-year history of an alcohol use disorder, only meets 10 criteria of the 11 criteria for a diagnosis, while Carol meets all the criteria. And Mary only meets 10 because she is very clear that she never drinks more than intended. She intends to drink a pint of vodka a day. And that's exactly what she drinks. So she doesn't meet the first criteria of the DSM. While Carol meets every criteria, she's been trying to drink, she's been trying, I'm sorry, she's been trying to cut down, she's been trying to do better. But she can never drink less than a pint because she gets very anxious and will have a hard time if she does. She has significant withdrawal when she does. So if we looked at the DSM, in theory, Carol will have a more severe illness, even if it's mildly more severe illness. She'll have 11 criteria instead of 10. But the truth is we know as clinicians that Carol has a better prognosis. We know that treating someone with a three-year history that is for the first time coming to treatment will be a little easier than treating someone with a 30-year history that has no significant sobriety in 30 years. So again, our diagnostic system at this point is really letting us as clinicians down and it's letting our patients down. We need a system that helps us identify treatments and that helps us give a prognosis to the patients, a more accurate prognosis than what we can give at this time. And that's why I'm advocating for precision medicine in psychiatry. And I think that's why we're all here. If these two patients, Carol and Mary, showed up at a doctor's office with a lump in their breast, no one will tell them they have breast cancer. There will be cellular, genetic, molecular, and imaging techniques used before they get a diagnosis. You would need biopsies, you would need imaging, you will need a lot of assessments before you can tell a patient that they have breast cancer and that they need one treatment or another. While we are only with a clinical history, diagnosing someone and giving them, you know, if we diagnose someone with an alcohol use disorder, we're giving them a poor prognosis. We're saying they have a chronic relapsing illness and that they are gonna have a very hard time managing their illness for the rest of their life. So I think we need to do better for our patients and I think we need to classify their diagnosis or their alcohol use disorder better. One of the things that we're doing at NIAAA, and I think we have a small audience and I'm sure you're all interested in addictions, so you've probably read this paper, but it's a paper that I always recommend. It's the Kubian-Bolkow model for addiction from the New England in 2016. It shows, it's kind of the neurobiology of addiction paper and it kind of proposes this function in three domains. Associated with three different brain circuits and these are executive functioning, negative emotionality and incentive salience. And at NIAAA in the intramural side, we have been collecting for years deep phenotyping data on patients with alcohol use disorder and patients in the full spectrum of alcohol use. So I have a natural history protocol where we invite patients that never drank and patients that cannot stop drinking on their own and we do deep phenotyping. We collect imaging, genetics, neurocognitive testing, biomarkers on all of these patients and then we try to assess how different levels of drinking or not drinking might affect their quality of their life or their functioning in different ways. And a few years ago, two colleagues, Dr. Laura Kwako and Dr. Mel Schwant, did some preliminary work with our deep phenotyping database. They just ran a factor analysis on about 500 individuals and this was in the American Journal of Psychiatry in 2019, looking to see if with a factor analysis model, they could find groupings that would understand and that would explain and help diagnose an alcohol use disorder patient better. And of course, we had a hypothesis with these three domains but the factor analysis pretty much showed the three domains that we were hoping to see. So this was the preliminary work to what I'm presenting today. As a psychiatrist, for me, it's important to note that incentive salience, and I'll talk about, these are psychology concepts. So I think for us psychiatrists, they're not as clear sometimes. But incentive salience is cravings and depression actually showed up under cravings. I was of course expecting it or we were all expecting it under negative emotionality because that's depression. But other than that, things really fell into place where we expected them. And this is, next slide, is important because just dysfunction in these three domains is essentially diagnostic of an alcohol use disorder. So if you're looking at dysfunction on incentive salience, you could not ask for better numbers. So just dysfunction in this area will give you 0.96 correct diagnosis on sensitivity and specificity for these patients. The other two areas, I would consider them very good screening tools. So for executive function and negative emotionality, it's 0.85 and 0.84 sensitivity and specificity. So they're still really good but not as good as incentive salience. It's important to note that there's a lot of correlationship within this dysfunction. So when one is off on the alcohol use disorder, they are all off. So, and I'll show you more about this but this at the end of the day means we don't know which happens first or it's hard to predict exactly how this happens but when a patient has a dysfunction in one domain, they usually have dysfunction in the others. As I said, I'm gonna let you like share what incentive salience, negative emotionality and executive function mean, since at least for me as a psychiatrist, it wasn't necessarily clear to start but this is similar to the research domains from the NIMH group where they wanted to group patients not by diagnosis but kind of neuro functional areas and incentive salience is essentially craving. It's anything that makes you kind of desire a substance or it was discussed earlier today, game, sex, sugar, there's so many other things that can be something that you crave but it makes it very hard once you are in front of the substance to not engage it. So that's what incentive salience is. Negative emotionality, I think the word we, our director George Koob has been talking about hyperkatifeia for a couple of years now and it's a word that I really like because I think it really embraces what it is. It's all negative affect, all negative emotions, everything that has a negative balance for the patient. So it's depression, anxiety, anger, irritability, loneliness, if you treat someone in the middle of withdrawal that's negative emotionality. It's that state of irritability and almost despair that patients have when they are in withdrawal but that might also, withdrawal or not, lead someone to drink so in one of the cases I presented with Carol, that negative emotionality of losing a child, it's part of this area. We all heard patients that say that they drink as a self-medication for depression and anxiety, that's negative emotionality. And executive function, as we all know, that's the one I think we understand better, is planning, understanding the consequences of your behavior, impulsivity, irritability, anything that helps the patients really understand the consequences of their behavior and be able to control their behavior accordingly. So these are the three domains that we decided to study and Dr. Cuaco decided to create this addiction neuroclinical assessment because we don't want to necessarily take the clinical history as the only tool to assess our patients. Dr. Cuaco, for the people that know her, moved to the extramural side of NIH, NIAAA and so Dr. Gunawan and Dr. Aram Chandani have really helped me move this project forward for the last few years after Dr. Cuaco moved out. And what I'm gonna present here is essentially Dr. Gunawan who's sitting in the back and will be happy to answer questions. It's the work we've been doing for the last, I want to say about four years. So the ANA battery, and this is, I know, less than ideal at this time, but we're working on improving it. It's about four hours of testing that patients do. We have tasks, we have self-report measures and we have other ancillary measures. Some are measures that are clinician rated and some are measures that are also patient rated but not necessarily self-report. And we are testing these three domains. Again, executive functioning, negative emotionality and incentive salience with different, what I would call cognitive biomarkers. And what we're trying to do is understanding with all the deep phenotyping data we have, we also have in these patients genetic information and imaging information. We have biomarkers from their just general assessment when they come in for treatment. So we want to see if we can group patients better. So as I said, we usually have this kind of salad of patients where there's a lot of things mixed and we don't really know how to treat each independent group. At this point, we don't have data on this. We just know that someone has an alcohol use disorder and we have a couple of agents and a couple of therapies that are recommended for everyone. But we don't have a lot of information on what therapy or which agent is better indicated for an individual in front of us. So we're hoping with this neuroclinical assessment, we will improve our diagnostic capacity. This is the sample, this is the demographics and I'm gonna skip through these, but I just want to point that we have, this was essentially collected through the COVID years. So we have many more cases. We could only bring cases into the hospital that controls, but we have non-AUD patients in the sample and we have AUD patients. And we've been again, testing them with the same battery of tests. And again, we have really good numbers with incentive salience. We're pretty much able to diagnose with incentive salience dysfunction. We're pretty much able to diagnose patients as having an alcohol use disorder. I want to point that alcohol motivation, which is the drive to consume alcohol, obviously, it's probably the best tool we have at this point. But alcohol insensitivity is also really important and I think genetics might play a big role on this. With negative emotionality, there's kind of three subdomains of negative emotionality. Internalizing, which is driving negative thoughts inward. Externalizing, which is driving them outward. And resilience, which is a protective factor. And it's important to note that resilience really is protective in these patients. And it's one of the areas, if we move down the road into prevention and early detection, it's something that we need to work as, you know, as healthcare providers. It's something that we need to really understand better and work more on this. But internalizing can really drive some of the responses on negative emotionality. So these are patients that are driving inward some of their emotions. They're, again, anxiety and depression and this is harming them. Finally, with executive function, we also have really good numbers on how diagnostic or not this could be. There's many domains in executive function. The one that didn't show up much was working memory. But impulsivity actually was one of the domains where we had some of the most interesting results in executive function. Of course, with impulsivity, we all know of the kind of common comorbidity of ADHD or bipolar or any of the diagnosis in psychiatry where impulsivity is a problem. And this graphic, I really like this. This is, I forgot the name, spaghetti plot. This is a spaghetti plot of the patients and I want you to see how diverse the sample is. So this patient, this group of patients have very different profiles, if you want. And again, with these different profiles, we can anyway group them and diagnose them just by dysfunction in some of these domains. These are the dysfunctions or the domains that I mentioned. I want, on incentive salience, to point out how dramatic is the difference on alcohol motivation between AUD and non-AUD patients. They are, almost everything is statistically significant. Let me see, so psychological strength is the other one that I really like. So again, this is resilience. It's the other one where there's also kind of an inversion of the response between AUD and non-AUD patients. But I think they're essentially all very good tools to help identify patients. Finally, I want to go back to the cases and discuss what is their dysfunction and how can we use this information that we have essentially from kind of testing different domains on them. Should we, once we understand that different domains are affected differently among patients, do we need to give everyone naltrexone and ask them to have an ME therapist or do we need to target our treatments more specifically to these patients? On the biomarker side and the precision medicine, something else that I always like to advocate is that we are, in many cases, I am the only medical provider my patients have seen in years. And it is really important for me to understand their health not just as a psychiatrist and as a psychiatric diagnosis of AUD, but they are very complex patients. And with Mary's case, she had significant liver dysfunction and those are biomarkers that are more predictive of her outcome than the diagnosis of an AUD for this patient. So really managing those biomarkers, understanding her liver function and her overall functioning, give her a much adequate prognosis than just telling her that she has a severe alcohol use disorder. Being part of her treatment team, instead of just referring her to a liver doctor, is her liver disease a comorbidity or is it part of her alcohol use disorder? And that's something that I keep advocating is we cannot really split the patients. Once they have liver disease, there's someone else's problem. And down the road, what we want to do down the road is what treatments work better for someone that has executive dysfunction? What treatments work better for someone that has negative emotionality or a dysfunction on incentive salience? I think on incentive salience, some of the medications we have work well. And I have the graphic, but I don't have time to show it. But once, as I said, one of these areas is in dysfunction, and especially on a case like Mary, everything falls apart. All the other areas have some degree of dysfunction. And it's really important to understand if improving one of these areas helps improve all of the others, or if we really need to target each area individually to help these patients really recover. And if we know these two patients have kids, do we need to test these kids to figure out before they ever start drinking, if they already have a dysfunction in one of these areas? As I said, ADHD could be one of those dysfunctions for someone that has already impulsivity and executive dysfunction. So do we need to test their children? Do we need to do early prevention or interventions on these kids? And could we, with genetic markers and imaging markers, also improve the accuracy and not only treatment, but also prognosis that we're giving our patients? I'm gonna do a small pitch for my study. If you have patients with an alcohol use disorder that need treatment, I am happy to recruit them where the federal government, there's no cost for the patients. And we'll even fly some of them to Bethesda for treatment. So 18 and over, thank you for asking. So I was asked what ages. So 18 and over. I am hoping at some point to recruit kids for biomarkers and everything I just said on early detection and prevention, but at this point, I'm only treating 18 and over. But I'm happy to take all patients, all comers. And just quickly, I have a huge team of people helping me do this work, so I want to acknowledge my team as well and our patients and their families because we couldn't have any of this work without it. Where I got him. Let me close this. Okay, wonderful. So, I'm going to present briefly on a combination of biomarkers and precision therapeutics. Let's think I'll just use this to sketch through. I do have a few conflicts to disclose. I used to, earlier in my career, I used to say, my only conflict to disclose was unresolved oedipal conflicts, but at this point in time, they seem to be fading, but they've been replaced by massive industrial conflicts. Those are all of the sources of my funding. I don't think they'll influence what I have to tell you today. So, precision medicine, why precision medicine? We're all presenting on this today. Psychiatry is lagging far behind oncology and cardiology in the use of precision medicine, and our challenges, we have challenges, and I'm going to propose that precision medicine offers some solutions. First, our diagnostics are mostly based on the patient's subjective complaints. Psychiatric disorders are highly heterogeneous, so it's difficult to find a one-size-fits-all treatment, which is true for all medical disorders. There are variable risk and outcomes within psychiatric disorders, and there's limited knowledge about which patients are likely to respond to which treatment. Some proposed solution, as we've heard, objective biomarkers, neurocognitive, molecular, and circuit markers may help us. Personalizing treatment based on clinical, cognitive, molecular, and circuit features will help us move closer to the way we treat breast and prostate cancer. Subtyping based on stratifying risk, course treatment, and biomarkers will help us. Obviously, there's not one AUD. There's not one PTSD, and there's not one... We know, for example, in depression, that we don't treat unipolar depression the way we treat bipolar depression. We don't treat typical depression the way we treat atypical depression, and we don't treat psychotic depression the way we treat non-psychotic depression, but for most psychiatric disorders, we do not have those clear subtypes, and then we also need to be able to use advanced computational modeling. Hundreds of clinical trials, maybe thousands, have been done in AUD, PTSD, and related disorders. Many of them show minimal or no separation of drug and placebo, but in point of fact, there are subgroups of patients that respond or don't respond to all these treatments, so I'll be offering you some ideas about how to make those distinctions computationally. So, I'm going to present today, in my 20 minutes or so, it's a little bit of a data blitz, I will present very briefly five studies. Blood molecular markers for screening for PTSD in veterans, voice markers as an objective biomarker for screening for stress disorders, and I'm going to present you quickly the results of reanalyzing three major previously published studies for the treatment of alcohol use disorder. Gabapentin, topiramate, and catiopine, these are all previously published, the references are there. Dan Falk, Hank Krensler, and Ray Litton published these studies, and we were able to access the data with their permission and reanalyze them to try to answer the question, which patients with which characteristics are more likely to respond to those treatments? So, the general idea is to leverage circuit, molecular, and computational psychiatry to advance precision diagnostics. My first study will be a DOD-funded study for blood biomarkers for high-throughput screening for PTSD. The second study will be voice markers. So, let's start with the two diagnostic precision meta-studies. We published this article as multi-omic, that is to say genetic, genomic, proteomic, and metabolomic blood biomarkers for identifying people with, veterans with and without PTSD. And the broader goal is a 16-year, nearly $100 million Department of Defense-funded collaboration among seven universities and the Army's elite molecular biology laboratory, which has as its goal to do something that has not been done successfully much in psychiatry, beginning to in neurology with blood tests for dementia. But in psychiatry, for most of our disorders, we don't have high-throughput screening tests that can be used in primary care, and that's the goal of this study, to identify and validate them. We started with male veterans. We started with a discovery set. We've now gone on to larger validation sets, and we had some independent validation, and we used a systems biology approach to look at genes, gene expression, epigenetic changes, proteins, and metabolites. So, what did we find? We started with one million features in the blood of each of our subjects and using a process called Wisdom of the Crowds, we were able to identify 50 candidate biomarker panels consisting of 343 unique molecular features in blood and also one physiological feature, heart rate. And we then used machine learning to identify 77 of the most important of those features. This is an output from a random forest plot to look at the most important features from the classifier, and we looked at the top features, the top 30% of the feature, 28 markers. This was heart rate, some clinical labs, three metabolites, some epigenetic marks, and proteins. And with that, using those features, we were able to accurately classify 81% of people who had been deployed to war and have PTSD, had been deployed to war, and do not have PTSD. So, that was on the initial testing and external validation sample. Second study is a, these are published studies. If you're interested in the details, I'm happy to share the publications and all the details in the analysis are in the supplements. Second study was to see, to answer a simple question, can the human voice help us to understand who does and does not meet diagnostic criteria for post-traumatic stress disorder? This is published as well. Human voice is very interesting. It evolved phylogenetically, evolved from a survival point of view before the internet and before other rapid forms of communication. Humans are herd animals, and they survive by rapidly signaling danger and safety to each other in the group through voice and some other sensory modalities. And speech is the primary form of human communication. And as they say, evolutionary developed to communicate emotional states, especially danger and safety and attachment. So, why look for speech voice markers? It's simple, it's non-invasive, it's low cost, it's ubiquitous. It can be used in a high throughput screening. And in the age of telepsychiatry, our algorithms can be built into clinician programs for treating patients to give real-time readouts of voice quality to signal emotional states. But for the purpose of today, in the interest of time, I'm just gonna say that we studied voice quality in veterans with and without PTSD. We looked at 35 minutes of speech. We collaborated with the best speech engineers in the world. By the way, this is not the content of speech, this is the biophysics of speech. We didn't look at natural language processing, we're doing that now. So this is simply the physics of speech. We have our colleagues at SRI develop, when you speak to your iPhone, when you ask Siri for directions to this conference, it was the engineers in this project that developed Siri for Apple. And they developed the coding for Dragon Naturally Speaking. And we developed a classifier with them, we looked at 40,000 unique biophysical elements in the human wave spectrum, used Fourier analysis, and we were able to find a set of 18 biophysical features in the speech that classified PTSD cases from controls with nearly 90% accuracy. Interesting, the features, which I thought would have been high anxiety features, were features of atonal, flat, muffled speech, probably consistent with emotional numbing. This work is ongoing. We have a large DARPA-funded study to validate these markers in 1,000 stressed healthcare workers. And that study is ongoing now. So now quickly, let me talk about heterogeneity of treatment, this is PTSD, but I'm gonna actually talk about AUD heterogeneity. But the idea is a very simple idea. The number needed to treat, with a drug and many psychiatric disorders, to have an effective treatment is often 10, 20, or 30 patients. But if you knew with precision who would respond to which treatment, which patients with AUD responded specifically to which treatment, the number needed to treat can fall in precision medicine to two. That's what's happened with, for example, the use of Herceptin in estrogen negative, progesterone negative, and HER2 positive breast cancer. The number needed to treat is very small because you have a precise knowledge of which monoclonal antibody for which type of cancer. And that's what we're trying to do here. The mathematics of this are complicated. All I can tell you in the interest of today's time very briefly is we've developed a mathematical function, which is the probability that any given patient with a set of clinical or biological features is likely or not likely to respond to a treatment. And we use those features to get in the placebo group, a similar group that has the same probability of responding to the treatment. This allows us to make causal comparisons between the drug and the placebo group and to reanalyze data. So, for example, let me move forward in this and show you some quick results here. Why is this not moving forward? Sorry, one second. So, the first study was a study of gabapentin. This was published by Dan Falk and Ray Litton. It was conducted by the intramural group at NIAAA. And our team, headed up by our mathematician, Gene Laska, reanalyzed this data using precision medicine. We'll also show you results from a topiramate and a catiopine study. So, for the likely responders to gabapentin, you can see from this, while in the original trial that Dan and Ray published, the drug did not separate from placebo. But the point that's very important to make here is, while drug did not separate from placebo, using this mathematical approach, we were able to find subgroups of patients for which placebo was dramatically better than drug. Placebo and drug were equal in the middle, and drug way outperformed placebo in the quintile to the far right of the figure. So, the question then becomes, among which subgroup of patients responded to gabapentin? And the answer is, those that before treatment had higher levels of heavier drinking days, they were heavier drinkers, they had lower levels of internalizing symptoms, anxiety and depression, the negative affect symptoms you referred to from your schema, and they had higher levels of externalizing symptoms, behavioral discontrol and cognitive impulsivity associated with executive dysfunction. So, the domains you presented are highly predictive of HUDL. So, those who are heavy drinkers with higher levels of impulsivity and lower levels of anxiety and depression did very well to gabapentin, which is an anticonvulsant and might be seen to be reasonable for a discontrol. We looked at topiramate, another anticonvulsant. In this study published by Hank Kranzler, overall there was a main effect of drug over placebo, but when we did the precision analysis, we looked at polygenic risk scores for alcohol use and asked whether polygenic risk would determine who responds better or not to topiramate, and here's what we found. We found that polygenic risk scores for time until alcohol relapse, time until heavy drinking, and problematic alcohol use were predictive of abstinence. We found that polygenic risk scores for time until heavy use and time until relapse were important for how many drinks per day the patients drank at the end of the trial, and we found polygenic risk scores, again, for problematic alcohol use and time until relapse predicted percent of heavy drinking. So genetic risk factors, and what's very nice about this study, the polygenic risk factors were developed on a sample of a million people and were applied to this small sample of a few hundred people in the clinical trial. In the past, we've been looking for genetic predictors of treatment inside our treatment samples, but inside clinical trials, there's no power to detect polygenic risk factors. So the polygenic risk factor scores are developed on large external populations, and then the subjects in the trial are genotyped, and then we can get very precise prediction. Hold your question just one second, and we'll come right back to it. And the final study is one of catiopine fumarate. This was published by Ray Litton. This trial showed no difference between drug and placebo. Let me read you what Dr. Litton wrote in the paper. In this study, we hypothesized and recruited an alcohol-dependent population of very heavy drinkers. Even though the primary analysis were negative, drug did not separate from placebo. There was no overall benefit of the atypical catiopine over placebo. Dr. Litton wrote we should explore for various subgroups. So this is a very sophisticated subgroup analysis approach, and you can see, again, we found subgroups in which placebo was much better than drug, in which drug and placebo in the middle were equal, and to the far right, in which catiopine highly outperformed placebo. So the reason that drug and placebo are equal is that there's an interaction, and a third of the sample, placebo is better than drug. A third of the sample, drug and placebo are equally effective. And for a third of the sample, actually catiopine is more effective than placebo. So as clinicians, what we want to know is for whom is catiopine more effective than placebo? Here's what we found with these sophisticated models. We found the patients did better in this catiopine trial if before treatment they had higher blood pressure, which is a sign of heavier alcohol use. They had different occupational status. I don't know which direction that is yet. We're still analyzing the data. They had greater trouble sleeping, which is common in people with heavier alcohol use and anxiety and depression, and they had lower levels of total plasma proteins, probably a marker of heavy alcohol use. So in essence what we found, catiopine outperforms placebo for heavy drinkers who are more heavily affected by AUD severity. So that's just a beginning, this is the early fledgling footsteps to move our field from a one-size-fits-all treatment to a precision medicine model where we can begin to select which drug for which person with which clinical and biological features. And as was the case for Nancy's presentation, it takes a village to raise a project like this. So this is a collaboration of eight universities and the Army's molecular biology lab. Yes, quick question, come to the mic please for the quick question. For the pyramid study, did you control for age of onset and if so how did that change the results? We have age of onset, we have duration of alcohol use, we have many variables in the models, they did not emerge in the final model. But that that can be because in machine learning model one variable can stand as a proxy for another. Okay, thank you. I'll stop at that point and it's my pleasure to welcome Dr. Shigerman who's going to present next. Let me just get my slides off. Okay, so I'm going to switch gears a little bit and talk about some more preliminary work around using technology to address alcohol and co-occurring disorders in women and some of the work we've done in developing a digital intervention in this space. So I have no conflicts of interest, I receive grant funding from NIDA. So just to start with some background, we know overall that the prevalence of alcohol use disorders is greater in men than women. However, what we're seeing is that this gender gap has been narrowing both in the US and internationally. So when we look at the epidemiological data, you can see the the top lines are past month alcohol use and the bottom ones are past year alcohol use disorder. And you can see the trend over the years where males are slowly decreasing and females are increasing so that those lines are starting to converge. And when we pull out the younger age group, so this this graph is just 12th graders, you can see longitudinally for drinking in the past month and getting drunk in the past month that those lines by 2018 are nearly equivalent. Recent data from the youth risk behavior survey also shows this, but in this case girls surpassed boys in prevalence of alcohol use and binge drinking as well as prescription opioid misuse. So there are several contributing factors to these trends. For one, we know that in countries where there's more gender equality, there tends to be more equivalence in alcohol use. But there's also been over the years really increased marketing by the alcohol industry towards towards women. You can see that there's wines, Mommy's Time Out, there's Skinny Girl cocktails, there's beers that are targeted toward women, and this is it's been very successful. We also see this in social media. There's particularly this really ramped up during COVID in that these these messages around women's alcohol use and using alcohol as a tool to cope. So to cope with stress, to cope with anxiety, and in particular the demographic of women with young children were getting bombarded with these messages. And then just the idea that alcohol is a necessary part of being a mom. It's, you know, mother's sacrifice isn't giving birth, it's nine months without wine. And just to sort of drive home the point that this has really pervaded the culture now. You know, in the past, if somebody wanted to avoid drinking and going out to bars with friends, we might recommend that they do an alternative activity like join a book club, you know, do some athletics, something like that. But that has all now become really paired with alcohol as well. Paint and sip classes, knit and sip, podcasts around wine moms, and athletic events where there's wine or alcohol at the end. So this has really become, particularly for women, paired in our culture as part, as normalized. And why this is concerning is that we know that women show what's called the telescoping course of illness. So by the time, when they first use alcohol to when they develop an alcohol use disorder and enter treatment, it's a much shorter time period than for men. So they've had fewer years of use, but they have more medical, psychiatric, and adverse social consequences. And this is due to both sex and gender differences. So biological sex differences between males and females, and the effects of alcohol. We know that women have higher blood alcohol concentrations than men when they drink the same amount due to differences in metabolism and body water. There are also gender differences based on the psychosocial cultural experiences. So certain co-occurring mental illnesses are more common in women. Women are more likely to have child abuse, sexual abuse, and intimate partner violence as well that all contribute to this. And we're seeing a lot of these medical consequences as well. Alcohol and cancer have been linked, of course, but for women there's the concern around breast cancer. And we know that for each alcohol drink consumed per day, the risk of breast cancer increases by about 7%. And when they have two to three alcohol drinks per day, that goes up to 20% risk compared to women who don't drink at all. And we're seeing recent trends around increases in alcohol-related disease in women as well as mortality. So although alcohol-related deaths are increasing among both sexes, we see a significantly higher rate of increase for females compared to males. And while these things are happening, there are significant barriers to treatment for women that they face compared to men. So pregnancy can be a barrier to treatment, but we have fewer programs that treat pregnant women. And there's also concerns for pregnant women about entering treatment and what legal implications that might have for them. Childcare is a barrier, financial costs, and as I said, trauma and co-occurring disorders can also be a significant issue for women, and that's not always addressed in treatment. And although there's stigma for individuals with alcohol and substance use disorders in general, women, it's been shown, really feel that stigma a lot more than men, and they feel that it's less acceptable for them, and it makes it harder for them to enter treatment. So given all these factors, gender-specific treatments were developed to address these needs that women have, and what was found was that these treatments are associated with better treatment outcomes for women. The problem is that it can be really difficult to access these treatments. So these data show a national survey of substance use treatment facilities and the specially tailored programs that they offer, and you can see that less than half offer programming for women, just over half co-occurring disorders, and then when you go down the line of the things that disproportionately affect women with alcohol use and other substance, you can see that very few facilities offered programming in those. And even when programs want to implement something that's women-specific, it can be a logistical challenge because our treatment programs tend to be predominantly male. So when you look from 2010 to 2020, you can see the lines are basically completely flat. It's just under 70% male-to-female admissions in substance use treatment facilities. It can make it really difficult to implement anything woman-specific when there's few women in a program at a time. And that's really where technology and digital interventions can kind of bridge this gap. We can reach more women, we can get at some of those barriers, they can be more cost-effective, particularly in mixed gender programs that are predominantly male. We can make them individualized, and for women who have trouble accessing traditional services, they could engage with these programs in an anonymous way. So where we started was taking an evidence-based, gender-specific treatment, the Women's Recovery Group, that was developed by my colleague Shelly Greenfield. This was designed as a 12-session, in-person group treatment. It's a relapse prevention group with structured sessions and women-focused content. It was studied in a stage 1 and stage 2 trial and found to be effective for women with alcohol and other drug use. So we started in our pre-pilot study with just taking three of the topic modules, really the core ones, the effect of drugs and alcohol on women, managing the co-occurring disorders, and then women and their partners. We adapted that to a web-based format, and then we collected data from participants and ended up adding two more topic modules after that, based on their feedback. So I'll present both of the studies at the same time, since they're very similar. So these are the components of the intervention. There's gender-specific psychoeducation, some introduction to coping skills. We tried to make it interactive, to have interactive questions and did these knowledge check questions, again, to keep people engaged. There's links to resources, and at the end of each topic is a take-home message. So as I said, the two studies were similar in their inclusion-exclusion criteria. Women had to be 18 or older, and really the exclusion was if they had anything psychiatric or medically that would impair their ability to engage with the intervention. For the pre-pilot study, where we had the three topics, we recruited 30 women who were receiving mixed-gender inpatient treatment, and for the pilot study with the five topics, we recruited 60 women across our levels of care, inpatient, partial, and outpatient treatment. And it's a single-session intervention. So the samples were very similar for both, just under 40 years average age, predominantly white, about half had children, and then a good portion in each had a partner who uses substances, which we know can make it difficult for women. So for the pilot study, we collected information on their mental health problems, and you can see, as expected, there was a high co-occurrence of anxiety, depression, PTSD. These are things that we know that women with substance use come in with. The sample predominantly had problems with alcohol, so over 50% said alcohol had caused them the most problems of the past year, and then another 11% said the combination of alcohol and drugs. So for results, we got high satisfaction with the intervention. People thought it was easy to use, they thought it was visually appealing, they really liked the gender-specific information. The three topics took about 26 minutes to complete, and it's about over 41 minutes to get the five topics complete, and we didn't see any difference in satisfaction by the level of care, and we didn't see any difference by the number of times that women had been in treatment previously. These are the elements that they rated as most relevant to their recovery. You can see they were similar across the both studies. Self-care came up as well as the co-occurring disorders, particularly depression, anxiety, and the link between that and their alcohol and other drug use. We collected qualitative feedback. One thing we had wondered is that some of these women had been in and out of treatment many times, and when they say this is all information I've heard before, and what we found was overwhelmingly that was not the case. As this woman said, just learning all this is crucial for women. I was unaware of a lot of it. And then this quote really speaks to the technology piece, that this person felt it was really helpful to see the information this way, that although groups are great to discuss these things, it's easy to miss something, so having it all laid out on the iPad was helpful to remember and see it visually. So for that piece, we were recruiting women who were already in treatment for alcohol and other drug use disorders, but at our hospital we have women coming into several different programs for a primary psychiatric disorder who also have co-occurring alcohol and other substance use, and oftentimes that piece isn't addressed. So we wanted to see how could we adapt this for those women to provide this to them while they're in the hospital. So to start, we did a needs assessment to figure out what changes we needed to make to the intervention. We interviewed 15 women for this project. We were focused on young adult women, 18 to 25 age range, and they were women in inpatient and residential treatment who had problematic substance use. So the majority of participants mentioned that their substance use was not adequately addressed in their current treatment. As this person said, I think I've always had a problem with alcohol and smoking too much and probably drinking too much. I'm clearly addicted and I think it's a huge part of PTSD that's not really addressed in any of the classes or anything. And then another theme that came up was the lack of integrated treatment, that it's difficult to treat the other issues, the psychiatric issues, when there's substance use going on as well and would be helpful to understand the connection between those two. So we took all that information, we made adaptations to the program, particularly we expanded the information on the connections between the alcohol and drug use and co-occurring disorders. We increased the interactivity to add more coping skills practice. For this young adult age range, peer relationships came up a lot, that there was a lot of concern about navigating social situations with their peers who might be using alcohol and other substances. And then another topic that came up was women noted that their menstrual cycle affected their alcohol use or their craving for alcohol as well. So that was an important topic. So once we made the adaptations, we brought this back to, we recruited 44 women from these programs that, again, had a primary psychiatric disorder. They didn't have to have an alcohol or other drug use disorder, they just had to have noted problematic alcohol or substance use, either in their chart or by their clinician. So in fact, about 14% of the women that we recruited, they themselves did not feel that their alcohol or substance use was problematic. So you can see the the Audit C scores just above the threshold and the DAST-10 were in the moderate range. Again, alcohol was the predominant substance for people and mood disorders was the primary diagnosis for most of the participants. So we, again, we got good satisfaction with the digital intervention. For this study, we found that 93% of the participants reported an increase in knowledge post-intervention and that from pre to post-intervention, we measured their interest in making changes to their substance use and their willingness to make changes and found that both of those increased, which was encouraging, as I said, given that there was a subset of women who, before they did the intervention, did not feel that their alcohol or drug use was problematic. We collected qualitative feedback so that we can further iterate the intervention. Participants liked that this was informative, they felt they could relate to the information, they liked the interactivity components, and we, for this, we had put in these real-life scenarios to sort of get at the peer, navigating the peer situations, where we gave them a situation, asked them how they would respond, and provided some feedback. And we also asked them to give us some feedback, is this scenario realistic or not, and if not, what would you change? So again, we can further iterate the program. The suggested changes were they wanted even more interactivity, there was some content wording changes they suggested, and although they liked the scenarios, some of them they felt were a little too black-and-white, so they wanted them a little more nuanced and offered some suggestions for that. And although we do have information about LGBTQ resources, they wanted more of that. So the next steps are to incorporate the feedback from the participants to further modify the intervention. We, you know, for this study we didn't have a control group, and they were in the hospital at the time that we collected the information, so it's important to collect follow-up data on their alcohol and substance use after they leave. But I think really the promise of this for the topic today is that, you know, we hope to, in the next iteration, to really leverage the technology to personalize these interventions so that it can be adaptive to women's needs. As others have said today, there's, you know, a lot of heterogeneity in alcohol and other substance use, as well as the co-occurring disorders and profiles that people come in, and technology can be a really good platform to individualize these interventions and tailor them to what people need. And so, you know, the conclusion is that, as I said at the beginning, this is a preliminary data, but I think that it speaks to the idea that technology and these tools may be effective to address the complexity of co-occurring alcohol use and mental health problems in a way that's gender responsive, and that will really benefit women. And like my colleagues, I want to end with my acknowledgments to my mentors, collaborators, and staff, as well as acknowledge the funding as well. Thank you. Terrific. We have a hardcore end of the meeting through faithful groups, so we're in it with you as long as you're in it. Questions? Please come to the mic, because we're online also. So where do you see, I spent some time working with Ben Coley Johnson's group way back when, and we were looking at age of onset and polygenic, a whole range of stuff. So where do you see genetics, epigenetics, fitting in with the model that you were presenting? Right. So up until now, one of the limitations of the data sets we've had to do these sophisticated analysis is the data available on the subjects prior to randomization is primarily clinical data. There's a little bit of neurocognitive data and very minimal biological data. Where we're going now, including in my own lab, is to really collect very deep data. So in my current PO1, which is also focused on a topiramate treatment for co-occurring AUD and PTSD, we have DNA, RNA, plasma, and other blood products on every subject before and after treatment. We have neuroimaging before and after treatment. We have neurocognitive assessments before and after treatment. And we're going to, as well as all the clinical features on onset of alcohol use, duration of alcohol use, kind of alcohol use, kinds of trauma, et cetera. And where we see it going in the future is to be able to model molecular circuit neurocognitive and clinical features and to try to understand how these can all be integrated in these models to better understand who's a good candidate and who's not a good candidate. So the early trials are very well-conducted, including, obviously, Ben-Coley Johnson's trials and others. We're collaborating with him. We're collaborating with Hank Kranzler. We're collaborating with Ray Litton and Dan Falk at NIAAA and others. And we would, to put in for a pitch, anyone who has a very good clinical trial that they've conducted that was somewhat disappointing in terms of drugs separating from placebo who would like to work with us on our computational models, we would be thrilled to work with you to try to see if we can understand for whom the drug or other treatments indicate. So it's very interesting. One of the problems in our field, let's say compared to some areas of medicine, is we tend to get larger placebo effects. Because a placebo in the context, to say the obvious, in the context of a psychiatric relationship, patients who come, particularly underserved patients who have difficulty accessing mental health services, when they come into one of our clinical trials, they get massive amounts of attention from very kind and caring faculty and research coordinators. So there's a huge nonspecific effect of being cared for in the context of a clinical trial, which means an enormous amount to our patients. So for a drug to separate from placebo, it's separating from that, you might say, therapeutic alliance effect, right? So it takes some sophistication to find that. But we're very encouraged, and the great thing is these advanced computational models can handle large numbers of variables in relatively small samples and get reliable signals. So we're very excited, and we're going to go back and look at many of the trials. I mean, I've published in PTSD, I published a large sertraline trial, which was negative in veterans. I published a large adrenergic antagonist trial that was negative. There was a multi-site prazosin trial for PTSD that Murray Raskin published that was negative and published in the New England Journal. But the fact is, we know for a fact, in the case of prazosin, which is an alpha-1 antagonist, that there are definitely a subset of patients with chronic PTSD who definitely benefit from prazosin, and we know those who have higher systolic and diastolic blood pressure before treatment tend to benefit more. So that's the direction we're going, and if any of you or your colleagues have well-conducted randomized control trials, for which you invested a lot of money and had disappointing results, we'd like to work with you to find the meaningful differences. So food for thought. One of those I would think about automatically would be Petanadi's study of alcohol, where the early onset treated for depression got worse, as opposed to the later onset. So you could really tease it out a bit more. It would be very interesting to use these new analytic models to try to understand that better. If I can add to that answer, I think in psychiatry and in substance abuse, we are treating essentially what I would consider almost the terminal cases. In the case I presented, I'm talking about a woman who's had an alcohol use disorder for 30 years. This would be equivalent of having a woman that comes in with stage 4 breast cancer. We're not doing early detection. We're not doing prevention. I think polygenic risk scores and some of these biomarkers would move us into early detection and prevention, where we can really help patients, not just to prevent an alcohol use disorder, but if they already have it, to help understand who is the patient that will develop cirrhosis. I had patients that drank all their life and have normal livers. My hepatic colleagues do not understand how it is possible, but I see them routinely. I have patients that drink very little and need a transplant. What is it that will get you into a transplant, and what is it that will protect your liver somehow of massive doses of alcohol? I think we need biomarkers like polygenic risk scores to really understand and move psychiatry from treating cases that are very late for treatment to doing treatment that is in prevention or in early detection. I just want to comment also, because I found your results with the executive functioning markers very important. For example, this is from my work in PTSD, but I completed a collaboration with a group at Stanford headed up by Amit Ek and his colleagues. We did a very simple experiment. Cognitive behavior therapy is considered to be a standard of practice for many patients with PTSD, alone or in combination with medication. CBT, emphasizing exposure and cognitive reprocessing, is a first-line treatment for PTSD. We asked a very simple question. Does it matter whether PTSD patients have intact or impaired executive function when they're treated with CBT? We found about 40% of men veterans with PTSD, male veterans, and about the same percent of women sexual trauma survivors had significant executive dysfunction. That is to say they were significantly impaired with respect to attention, concentration, working memory, and other measures of executive function. They did not respond at all to CBT, zero. There's an example where precision medicine can be advanced. Assessing attention, concentration, and working memory is not difficult. It can be done in everyday clinical practice. It doesn't require the whole genome to figure that out. That's the direction we're moving in. Don, did you want to come up to the conversation as well? Yes, thank you. Hi, I actually had a question for Dr. Sugarman. Yeah, I had a question for Dr. Sugarman. On the technology aspect for women with alcohol use disorder, I'm just wanting to know about the access in terms of women who may be underserved and maybe not have access to some of these technologies, and also the generalization. How can we generalize results based off of the demographic profile given that most of the women were white and educated? Yeah, that's a really good point. When we iterate and do the next trials, we are hoping to get a more diverse sample. One of the reasons we created it as a web-based intervention was that it would be more accessible. We do find that women in treatment, even if they have a phone, they don't always have that phone a few weeks or a few months later. Sometimes apps can be more problematic. We did design it as a web-based program for that reason, that it could be accessed from anywhere. The model that we've done for these studies is to have the technology embedded in the clinic and that people can engage with it while they're in treatment, particularly when they're in treatment for primary psychiatric disorders where the substance use and the alcohol isn't being addressed, which could also have implications for doing in primary care clinics or things like that. We are experimenting with ... Also, we have a study now where we're having people engage after they leave treatment with these programs and just to see the feasibility. As you said, how many people can continue to engage? Do they have the technology? For studies, we could provide phones to people, but it doesn't always last. If they have two phones, it's problematic. It's certainly a challenge that we're trying to address, and point well taken. Dr. Botoraco, hardcore aficionado. Dr. Botoraco is our vice chair at Bellevue and is responsible for the care of more patients with AUD than perhaps any psychiatrist in America. Thank you all. Thanks for the talk. My main question is a clinical one, which is I very much like your discussion of the different domains and the McLean protocol and then looking for biologic markers and speech patterns. How far do you think we are from being able to personalize interventions with patients who people are seeing now? How far do you think? I don't think we're far. I think we just need to use some of the tools we have better. Again, I know I'm not in the right place to complain about the DSM-5, but it's not a good diagnostic system. If we want to diagnose someone with an alcohol use disorder, we actually need blood chemistries. We need liver function. It should be part of the diagnosis like it is with any other medical condition. I think if we use some of the tools we already have, we can go really far. Of course, I'm really looking forward to polygenic risk scores and to many other tools that we're developing, but I believe it's a few years away. I don't think it is really that far away. I think we just need to use some of the data we have better and stop the division between psychiatry and a more holistic approach to the patient where we actually care about liver function or pancreatic enzymes or anything else that this patient has as part of our diagnostic system. Marianne, I would say with regard to the pharmacotherapy of AUD, of which we have antabuse, acamprosate, naltrexone, topiramate, gabapentin, and then some more experimental drugs, I think we're fairly close to being able to offer some specific guidelines of which one to start with, with which person, with which co-occurring disorders, because there are a sufficient number of clinical trials that can be mined using this method. The advantage is that it takes five years or more and millions of dollars to run a single trial. It takes us about one month and probably 15 or 20 thousand dollars to do the work to answer the precision medicine question once we have access to the data. So I think we can really super advance the practical guidelines. In other words, I'm hoping three or four years from now, when we talk about clinical practice guidelines for the drug treatment of AUD, we can talk about it in these terms rather than we have this, start, try this, if that doesn't work, try that, and if that doesn't work, try that, and if that doesn't work, add CBT. That's not the way we want to treat these problems. We really want to begin to understand what we do. And I think also, to me, it creates a great optimism for our field. There's a lot of, there's a sense in psychiatry often of discouragement that we haven't advanced experimental therapeutics far enough. I don't think that's actually true. It's I think what's more true is we don't fully understand what we've done so we can extract the signals that are actually in there in the enormous amount of careful work that's been done. So I'm very optimistic about it. And I don't think that we, as the APA and as practicing psychologists and psychiatrists, are going to use clinical practice guidelines as a way to approach the patient in front of us in our office. I think I think that's that was better than not having guidelines. But it's not the same as knowing what is in the best interest of this patient that I'm consulting on today. Yeah, well, that makes sense. Even something, though, as simple as what you described about working, testing, working memory before prescribing CBT, we don't do that. That we could start doing. It's easy to do. Ask a person to remember five objects for five minutes. You don't you don't have to send them for ten thousand dollars worth of neurocognitive testing. I have a little secret. My bedside neurocognitive testing accounts for 94 percent of the variance of the five thousand dollar neurocognitive assessment. Sorry, neuropsychologists. Thank you. Thanks. I'll just add that, you know, in the space of technology, I think that there's tremendous potential and excitement that, you know, we have these these technologies and just getting the research to catch up. I mean, I think with there's a lot of good work with just in time, adaptive interventions to figure out when to give the right intervention at what time. But I think we still need to know a lot more about how to keep people engaged with these technologies and how much human interaction needs to be involved as well and how to do that in a way that's not burdensome to clinicians and overwhelming them with with data. That's something we hear a lot from. Providers, when we ask them about integrating technology, is that how do I do this with everything else I need to do? And it's great that we have all this data, but it's it's almost too much unless it's distilled in a certain way. But I think there is tremendous potential there. Thanks. So research and clinical practice have to be advanced in a thoughtful and harmonious way. So, for example, at the leading edge of research, we look at a million molecular features in every subject and we image everyone with structural resting state and task based fMRI. That's a very little practical value to what we do when I move over from my research lab and move into my faculty practice, which is in the same office. So I'm going from one to the other. I'm studying a million molecular features in a PTSD patient in my research lab and somebody you refer me a patient with PTSD and I'm assessing them in my office. So how can we make that leap? Well, there's some simple ideas. Cognitive testing can be done in a very simplified form. Molecular testing may turn out that routine clinical blood tests, which we get on every patient, contain much of the information of complex genetic testing. It may turn out that simple EEGs that can be hooked up on an outpatient office contain much of the information of an fMRI. Let me give you an example. One of my colleagues, Toma Handler at Tel Aviv University, had the following idea. Can we develop a simple, practical neurofeedback treatment for ADHD or PTSD, which could be done in a clinician's office? Well, we know we can do a simplified EEG in the office. So what she did is she simultaneously acquired EEG inside the magnet while getting fMRI. So using the fMRI signal, she acquired amygdala activity and using the simultaneously acquired EEG signal and machine learning, she found a derivative of the fMRI amygdala signal in a simplified EEG form, which we're now using in the office. So I think one of the things we're going to do in the future and combining this with technology would be to use biomarkers like simple memory tests, like voice quality, like facial expression, like simple EEGs and routine blood tests to try to give us a window into these more complex processes and relate one to the to the other through advanced analytics. So I'm very bullish on two things that we'll crack the code on precision psychopharmacology through the reanalysis of existing trials. And the second thing I'm bullish on is we'll be able to implement biomarkers in clinical practice that you and I would think are reasonable to do in our office. So, of course, I failed all medications and psychoanalysis to cure my eternal optimism. And you know, you know me very well. So so you have to take it all with that grain of salt. No, it's hopeful. Good. Thank you. Thank you all. Any more hardcore? This is an incredible group. This is the Wednesday 530 group. All right. Thank you so much for everyone.

precision medicine

alcohol use disorder

comorbidities

neurocognitive markers

DSM-5

genetic techniques

computational models

gabapentin

topiramate

digital interventions

gender-specific treatment

personalized treatment