Thank you, Ashwani and Craig, and a privilege to share the podium with Philip and Lorenzo. So I've got to thank one of my colleagues in oncology for some of the slides that I'm going to be using. In this talk, I will go over stratification of AH, endpoints for trials in phase two, and spend most of my time on virtual trials in AH and adaptive designs. So traditionally, we've been treating only patients with severe AH, but those who do not meet those criteria also have mortality. So we have to be thinking of studies where we treat less severe patients. And so for the purpose of stratification for future studies by Mel-Skord, it's been decided that 10 or less is mild AH, moderate is 11 to 20, and severe is 21 or more or greater than 20. And so I hope we see more trials in patients with moderate AH. The second is endpoints in phase two trials. Phase two trials are for efficacy signals and safety assessment, and usually require 100 or more patients take up to two years or more for completion. So there's a lot of work involved in these trials, and I think it's important to try and get as much information as possible. So I'm trying to make a case here that in these trials, at least in a majority of patients, we should try and get a liver biopsy and molecular pathology, because I think this is going to inform who responds to treatment, and I think it's also informative for phase three studies. And for phase two trials, a decrease in Mel-Skord of five or more at 90 days, corresponding to 30% relative reduction in mortality, may be an endpoint. And how do we come up with this number? So Mel-Skord of 20 has a mortality of 20%, 15 has a mortality of 13%, 28 has a mortality of 54%, and 23 has a mortality of 35%. So moderate AH, which is 11 to 20, a Mel decrease of five corresponds to a mortality decrease of 7%, which is about a 30% relative reduction in mortality. Severe AH, a decrease from 28 to 23, that's five, corresponds to a decrease in mortality of 19%, which is again a 30% relative reduction in mortality. I do understand that the mortality rates in more recent studies have been less than this, but the calculation is still the same. Five percent decrease in Mel corresponds to a relative risk in decrease in mortality of 30%, which is a good signal of the drug being effective. So in future trials, Mel decrease by five or more may be used as an endpoint for phase two studies. I think the big focus in future should be on virtual trials in AH, and there are difficulties in recruitment and follow-up in this population. So the question is, are virtual trials a solution? Since the 60 seconds which came on from the last slide, 3.8 million Googled, one million went on Facebook, and 18.1 million texts were sent, and I'm sure there are also some texts which were sent from this room during that past one minute. So the question is, can we use this digital media to facilitate recruitment? So in general, we know that 37% of investigator sites fail to meet recruitment goals, and 10% fail to recruit even a single patient. Facebook, 900 million people log on 14 times a day for 20 minutes per day. And recently, one of the sponsors in a trial in obesity using Facebook on the first day identified 5,000 potential subjects, and at the end of the day, knew which 300 patients they wanted for the study. So once they knew where the patients were, then they identified the potential sites. So now you can use this to pick sites around the patients. Traditionally, we pick the sites first, and then we struggle to get the patients. So digital media might help us recruit patients. The question, of course, that is still unknown is whether patients with AAH are an optimal group for virtual trials or not. So this is an AI-based system for recruitment into breast cancer studies, which one of my colleagues is involved in at Mayo. So patient data is entered into an electronic record data, and there are the clinical trial protocols. So that gets into an AI system, and a screening team of research coordinators and data abstractors look at that information. They let the physician know before the patient has actually come in there. So at the point of care, a decision can be made between physician and patient, whether the patient enters into a clinical trial protocol or a standard of care protocol. By the end of this year, the patient will also get information digitally, what the potential protocol is and what the standard of care protocol is, and again, that's going to make it a little easier for that discussion to take place. And as a next step, there are also going to be standard of care protocols to be entered. This is a little more complicated because each patient has had a different treatment before, but this design can be used to recruit treatment, and this has really accelerated the recruitment of patients for the breast cancer trials. So initially, using this system, they had about 45% accuracy, and now it's up to about 75%. And why is this not 100% accurate? You should think it should be 100% accurate. The problem is the patient record system does not account for everything. These tend to be inaccurate. These do not tend to record all possible treatments that the patient has, but certainly this is something that we should be thinking about in the future. So using digital technology for clinical trials is called a virtual trial. It can facilitate recruitment. You can collect information, data, and adverse events real time. So remote monitoring, we're doing that to prevent readmission. We can monitor the patients remotely, and that way it increases and improves convenience of the study. If the patient does not show up, that's a failure and loss to follow up. But if you follow the patient at home, that's much more convenient for the patient. We are more likely to complete the trial, and we can also increase confidentiality in the trials by following up these patients at home. We can also use all the information that we have here for outcome modeling. So I think in the future, we should strongly consider virtual trials. And in fact, this is a goal of the National Academy of Sciences, Engineering, and Medicine, and I would encourage all of you to look at this publication, which has come out just a few months ago. And finally, adaptive designs. Adaptive designs plan for failure. So if a study started and it fails, then no one knows what to do. Adaptive designs help you plan well in advance if one arm is failing. So if we look at alcohol-related hepatitis, in patients with severe hepatitis, severity at baseline and response to treatment at day seven will determine who's alive at six months. And Alex and Philippe have very clearly shown that if you're alive at six months and have no alcohol relapse, and you had an initial response to treatment, you have a minimal risk of death. If there's no alcohol relapse and initial non-response to treatment, such patients have a substantial risk of death, and they are candidates for liver transplant. But if you have an alcohol relapse, and especially if you're not responding to treatment, there's a higher risk of death and to consider palliative care. So therapeutic targets in the short term should target liver injury, and in the longer term should target alcohol behavior. So the question is, can you have a study design which seamlessly moves from targeting liver injury to targeting alcohol behavior? So let's look at what are the advantages of an adaptive design. Larger proportion of participants assigned to the treatment arms performing well. If one arm is performing well, move patients to that arm. Reduce participants in treatment groups which are performing poorly. Move them away from that arm. Not only that, you can investigate higher doses and longer durations rather than the traditional fixed dose trials. So you can change things in there. Those are the adaptive designs. And they will also allow seamless transition of phase two trials to phase three trials. So what is the big advantage of adaptive design goals? You're trying to reduce mortality in all possible arms. So let's give you two examples. So this is a two-stage population enrichment design where we've stratified patients. So let's say subgroup S, alcoholic hepatitis with infections, okay? That's one-third of the patients. 50% get treatment, 50% get control. And we have a second subgroup which has two-thirds of the patients, let's say this is alcoholic hepatitis without infections. 50% treatment, 50% control. So at the interim analysis, we have N events and N patients in this subgroup and similarly in this subgroup. So there are three possible outcomes here. We find that no response in both groups. So I think this study has to be stopped because of futility. Or we see that both groups are responding. So we have to continue the study, continuing stratification into two groups. But there's the third possibility. Only patients in this arm are responding to the treatment. Patients in this arm are not responding to the treatment. But if you continue the study with only one-third of the patients, you won't have the required number of events to have the adequate power. So then what we do is no longer enter patients into this subgroup, move all patients to subgroup S. And so that that part of the study is adequately powered. So arms performing well, move patients into that arm. And then you finally have a subgroup analysis. What if everyone fails? You have to plan for failure. And that's the advantage of Bayesian adaptive design, say, for a three-arm study. For the purposes of this discussion, futility is a less than 40% probability that the drug reduces mortality. And success is a greater than 90% probability that the drug reduces mortality. Sorry, this is reduces, yeah, greater than 90%. So let's take three arms. For instance, anechyndra, GCSF, and steroids. We do an interim analysis at 20%. Bayesian means with every event you do an analysis. But actually the first analysis, you wait a little. So 20% of events. And we find that treatment A is not working, 40% less chance that it's reducing mortality. So when you continue the study, no more recruitment into A. You're going to recruit only into B and C. Say at 30%, you do another interim analysis. And we find that C is looking much better than B. At that stage, change the randomization and roll three-to-one into treatment C and look at the outcomes. You're getting moving patients into arms which are performing well. So at the end of this phase, let's say we know treatment C works well. Let's say this is steroids for four weeks works well. But the question is, what if you give steroids for 12 weeks? Is that better than four weeks? So anechyndra for 30 days rather than 16 days. Is that better? So this is incorporated in these designs. So then we seamlessly move into treatment two. What is the study duration? What about a longer duration versus a standard duration? At the end of this, now we know which drug and which duration is the best. But this is still not the end of the study because we don't know about alcohol use in these patients. We're going to keep them alive only for six months with this treatment, but long-term we know it's only with treating alcohol behavior. So then you target alcohol use in here, treatment A, B, and C, and you can continue. What's the advantage? If all this failed, that there's the end of the study, you've got no information. But if this worked, then you want to say, what if we give it longer? Typically it takes at least two years between two phases of studies. So there's a two-year share and say a two-year share. If you went in between studies, you'd be spending five years more. But if you had a Bayesian adaptive design, you save a lot of time. So to summarize, adaptive design trials improve efficiency and reduce costs. They maximize success and the data obtained. Power is calculated on the study performance. They can accelerate the discovery process, but we have to understand this. It's a complex trial design and it requires acceptance of complex statistical methods, not only both by investigators, but by reviewers too. And an adaptive design will not save a poorly planned trial or an inefficient treatment. So the take-home messages on clinical trial design in AH for the future are trials in AH should be aimed at targeting liver injury and alcohol use behavior. We have to test the use of digital media for recruitment as well as for conduct of studies. And a Bayesian adaptive design might be ideal for AH studies which target alcohol injury and alcohol use behavior. Thank you. Thank you. I'm gonna make a couple final comments while people are coming up here. Number one, we've got great drugs and great targets for alcoholic hepatitis. So if you're a drug company out there why now it's a good time to invest. One of our problems though is relapse after six months is really the thing that's driving mortality. So again we need better drugs for alcohol use disorder and I think we've heard that from Lorenzo. In transplantation we need more clarity on who we're transplanting. It's obviously different from the United States and Europe. And a real question is how bad is relapse? So in the alcohol use disorder there's a thing called harm reduction. So I have a drink every day. I don't think that's a problem or at least I hope not. We've heard about unique trials from Patrick and I think that's a very important thing. We should all want him entering patients in our studies. And lastly I think we want to be doing more trials in a broad range of people with liver disease and not just alcoholic hepatitis. So with that let's have some questions. Thank you. Brian Lee from San Francisco. So excellent presentations, very interesting. But my specific question was for the selection process for early transplant and ALKEP. You know it's such a unique sub-population and especially they come in with such high acuity and you have to evaluate them so quickly you're really losing time with that initial evaluation. So my specific question was you know in Accelerate the majority of patients had overt encephalopathy during the evaluation. So what is the European experience with that? Like how can we learn? How do you assess a patient with encephalopathy when you're relying on surrogates for accurate information? How can you assess their commitment to sobriety? So we build up two algorithms. One algorithm is for patients without encephalopathy and as you mentioned 30% of patients have clinical. We never talk about over-encephalopathy. As you know it's difficult to assess but at least on clinical encephalopathy. So there is an algorithm who is in patients with clinical encephalopathy and this most of the response to this algorithm item are fulfilled with a family. In fact in order to have a kind of evaluation of the history of alcohol intake according to the family members. So most of it is also patient are selected on this with this algorithm. Thank you. Hi David Wong from Toronto. It's great to look at MELD score response in terms of disease severity but why don't we look at inflammation markers to help drive how long we treat therapy? For example do we actually need longer therapies or are shorter therapies with steroids more beneficial to have lower risk of infection? So the question is why MELD scores and why not inflammatory markers? Because currently we don't have an inflammatory marker level which is associated with a particular risk of mortality. You know we need a number to say this decrease accounts but I think we should actually incorporate. I think there are several inflammatory markers which are prognostic. Ramon has looked at CRP and things like that lipopolysaccharide. I think we should incorporate all that. It's just that this is the easiest available and the best studied. Thank you. Ramon Bataller from Pittsburgh. Thank you for the sake for this symposium. I'm glad to see more and more people in the growing field of alcohol. One question to Lorenzo about the treatment of alcohol disorder. We focus always in alcoholic hepatitis that we know they're sick. We give an alcoholic something that we know they have a sick disease. What about the early liver disease? I have seen four people dying because of dysulfuron-induced fulminant hepatitis and all four patients have silent cirrhosis. None of them were checked for liver fibrosis before dysulfuron was given. In my opinion this is another reason why we should do fibroscan in all addiction centers before giving dysulfuron to patients without checking the degree of liver disease. What do you think about that? Yeah thanks Ramon. I agree. I think what you say and the examples you bring up I think are consistent with the bigger picture problem that maybe the addiction doctor is look at the addiction piece and is not look at the more general medicine. I know checking something about the liver you know could be cardiovascular consequence. So you're right even if we don't go to the extreme of alcoholic hepatitis people with already excessive alcohol use they have a variety of medical problem that they should really be treated in sync. Silent cirrhosis is a great example. Bowels I would say cardiovascular problems etc. Yeah absolutely I couldn't agree more. Thank you. Hi Gene M from Mount Sinai, New York. This question is for Philippe. In patients who, setting aside transplant for a second, patients who present to alcohol rehab programs their rate of relapse is 60-80 percent. So it's quite remarkable that these patients who undergo early liver transplant are able to maintain abstinence or rare drinking and with excellent outcomes. Certainly the transplant played a role in that. There's good insight and selection process. But I wonder if we're probably going to get to a point where we can get a general sense of a decent candidate. You know we're not going to necessarily get to certain specific Milan criteria for example for these patients for inclusion. But these are patients where we could do after transplant something to actually induce a good outcome with things like pharmacotherapy or some sort of addiction therapy. So in your group in Lille, have you made any plans to do some post transplant interventions where you can actually make a borderline patient into a good patient? So we don't have that but I will tell you that I just finished with my color of you on how we're looking at alcohol relapse. We need to go to the addiction guide now and to have to use the same endpoints and them as an example what is a failure. You know because selection patients means that we know who failed, who succeeded. So yes or no question for alcohol relapse is no more relevant I think so. So we need to probably get there and then as mentioned just before by Craig, we know so to look at a patient. Let's say a patient with a sustained alcohol use which 50 gram per day after the transplant. It go down to 20 gram. Limited risk on the liver graph. Is this a failure or a success? It's a success until there is no life society effect on the society. So therefore in order to decide what will be the good candidate we need to at least to agree on what is a failure, what is a success in term of alcohol after transplantation. If when we will be there I think the field will be more clear. So this is why up to now we are still sticking with the question yes and no and this is why we cannot completely address your question. But I think we need now to define a nomenclature of relapse after alcohol and I'm encouraging you guys to take Lorenzo and over to help us out to define it clearly. Because percents of every day now I understand everything in their topic. Percentage of every drinking days which is very important. Alcohol intake relapse according to the double HO level two weeks now. I already learned everything. I think we need to introduce a liver transplantation now. At that time will be an accurate way of looking at it. But I don't understand your question as usual. I'm sorry about it because I don't have the answer. But I try to say that at least we can find out answer we agree on that. But up to now we are facing the same problems as you. Sorry about the non-response. We're going to need short questions and short answers. They're going to kick us out of here. Hi Juan Pablo from Chile. So in the same line of gene but for Dr. Leggio. So after transplanting, the post transplant, what is your experience of pharmacotherapy for alcohol use disorder considering drug interaction and all those things? Do you use this whole theorem or not? When did you start it? Yeah I mean that's a very good question. The problem we don't have as you know very well. You know we don't have a design study that have a look at this question in a systematic controlled way. In general I will not use that. So for me in part also for the reasons that Ramon was talking about. And I think the question is not really at this stage about which specific drug to use. But I think it's more about the bigger picture problem which is A, as Philip pointed out, how we define a relapse or not. And the importance of really providing some treatments. It doesn't need to be a medication. It could be a behavioral treatment. So I will keep it short. But that's the bottom line to make sure that the addiction is integrated in the post transplant. Elena Cortes Pint from Lisbon, Portugal. My question is for Dr. Leggio regarding abstinence. The issue is to which extent we should try for total abstinence or just a reduction in consumption. I think there has been some controversial regarding this issue and so I'd like to know. And the second question is to which extent do you think that the gastroenterologist or an epithologist should try to use the drugs for abstinence or always refer to psychiatry treatment? Yeah, these are excellent questions. I will try to be very short. So let me start with the second question. I mean this is critical. The problem is of course that people need the training. And that's something I also often talk like with Cathy Young who directs the Division of Metabolism at AAA. We need to provide them more training. But once people have done the residency fellowship it's harder to convince people to still stay in a fellowship position forever, right? But the point is we don't need everybody to be able to do CBT or to prescribe nartrexone or whatever. But we need the virtually everybody to identify that there is a problem. Maybe some hepatologists will be able to even feel comfortable to prescribe the medication just like some PCP are comfortable to prescribe an SSRI. And other people will not, which is fine, but they will need to refer to the addiction specialist. But identifying that there is a problem is really critical. And so we need the training education to at least be able that every clinician can identify there is an alcohol problem. For the first question, so the challenge in the liver field is that of course the safer or the safest position would be to say total abstinence because of course we don't have any data suggesting that the moderated drinking could still be safe in people with liver disease. And I think it goes back to what Philippe was saying, you know, this could be a little ideal. Like are we asking a patient with diabetes to always keep their glucose levels below 100? No, we are not. But if we take out the liver disease for one second, in the AUD field in general, I think now we clearly understand that the total abstinence is not what we should really pursue. It's too kind of extreme. And very briefly, FDA does already accept, thanks to work done by Rayleight and NAAA, the percent of subjects without heavy drinking days as an outcome, which Philippe was mentioning. And now NAAA, again Rayleight and then Falk and also people in the academia such as Katie Wickwith, their work on the WHO criteria, the WHO criteria, to basically look at a reduction in harmful drinking as acceptable criteria. Again, the point is, if the glucose is below 126 or below 140, that's already an achievement, right? So we should go into that mindset for AUD as well, which also will help to really understand that we are talking about a medical problem. We are not talking about a bad behavior to be kind of a classified dichotomous, yes or no. Very short. So NAAA defines problem drinking as seven drinks a week in women and 14 drinks a week in men. And they define a typical, a standard drink as a 12 ounce can of beer with 5% alcohol. Now increasingly across microbrews across the US, I've seen a can of beer, 12 ounce can of beer with 12 to 15% alcohol. Do you think that could have contributed to the increase in the increased liver disease and the 25 to 35 year old demographic that has been like, that was a subjective Washington Post, you know, article, et cetera. Has someone looked at what these 15% in 12 ounces? So I think that's highly unlikely. So if you look at what people are drinking who come into a alcoholic hepatitis study or to his alcohol use disorder program, it's 15 drinks a day. And so whether you increase the alcohol a little bit in your drink or not, I think is probably not real relevant. At least that's my answer and I'm ending the question. So NAAA has a great website. So go to that if you want to find out more about drinking. Thank you everybody for staying and thank the speakers.

alcoholic hepatitis

Mel's score

phase two trials

liver biopsy

adaptive trial designs

post-transplant pharmacotherapy