Session II Discussion: Afternoon, 8 April 2009CO-CHAIR PAULS: Before we get started, I have a couple of announcementsfor those of you who were not here earlier today, we are arranging shuttle services to the threeairports. Please make sure you sign up at the registration desk before4:00 today. In addition to that, a couple of people have asked me about dinner tonight. The waywe arranged the food is that the people that are staying here have already paid for the food. Sodinner is on your own tonight, if you are actually staying here. At the same point in time, there isa pretty hefty chunk of appetizers with happy hour this evening so we can all test our livers.Also in addition to that, we do have the slides. The slides and the transcriptswill be posted, as soon as possible. Lots of people are asking for copies of the slides. Anythingthat is shown here today will be posted on the FDA website and there will be a link to that to theAASLD website.And lastly, I have three people that if you could please report to the registrationdesk at your convenience, it would be appreciated. Ken Borow, Antonio Paredes, and CosetteSarbjit Singh. So if all of you could report to the registration desk at your convenience, I wouldappreciate it. That being said, I am going to turn it over to our moderator this afternoon, PaulWatkins.DR. WATKINS: Thanks, Lana. Before I get started, I just, you know, have tosay how excited I am having come to this meeting now for seven or eight years, and what aterrific program John put together this morning and I think for the rest of the meeting -- could wejust have a round of applause? (Applause.)DR. WATKINS: And I think for senior management and companies who think ofthese rare idiosyncratic events as a bolt of lightening, an act of God, there is no way you canpredict it. It is just the luck or bad luck of drug development. I wish they were here to see, we areon the verge. I mean, you can see in the next few years, I think, we are really going to be able tocrack this thing. But anyway with that editorial comment, we are now going to talk about geneticfindings in two of the major drug-induced liver injury gene banks in the world. The third, the DILInetwork I will mention briefly, but we are really just now getting underway and starting a millionSNP analysis of our gene bank. So the first speaker is Matt Nelson from GSK, who is going toget us up to date on what is going on with the Severe Adverse Event Consortium.
Link to Notes: Nelson SlidesDR. WATKINS: Lana tells me we must stay on time so there won't be anyquestions right now. But we will have Matt back for questions after our next talk, which is goingto be Raul Andrade from the University of Malaga telling us about the Spanish DILI network.Link to Notes: Andrade SlidesDR. WATKINS: All right. So, we have the panel next. So, if Matt can come onup, too. And we are going to hear a little more about genetics from both Roger Ulrich and DavidGoldstein. So, if the questions travel over into something you are going to cover in great detail,please put up your hand we will stop. I can't argue with John Senior. I think we want Ann Dalyback up here and Arthur Holden as well. So Arthur, why don't you come up? And Ann. Do wehave enough chairs? Here we go. So I guess we have a few mikes there.Let me just start by asking about the very striking home run in terms of wholegenome association studies of flucloxacillin with the particular HLA and the exciting possibility isthat is the missing link for all DILI. That in fact, there is something that stresses the hepatocytebut the real issue is does the person have the right genetic makeup from acquired immune sideof things to attack the liver cell, once it is stressed in a peculiar way. From the work that has beendone so far in the SAE Consortium, is that a tenable general hypothesis or are flucloxacillin andAugmentin and ticlopidine going to be the exceptions to the rule or don't you know?DR. NELSON: I can speak to some of the other drugs that we have looked atand I don't think those will be a general -- just I don't think that we have seen -- I think we haveseen enough counter examples because not everything looks like flucloxacillin certainly. And Idon't think even some of them would look like what you would expect for co-amoxiclav. So, Ithink it will be very specific. I think even by internal as well.DR. WATKINS: David, in the corner?DR. GOLDSTEIN: Can you just clarify what you were asking? From theexamples of 5701, I was just wondering exactly what you are asking because I guess the ideawould be the example of abacavir and also the fact that 5701 is particularly good at the analogyimmune attacks for HIV-type cells, we know that. The idea would be that it is actually just animmune attack that is taking place with 5701. So that would just suggest that for whateverreason 5701 mounts an attack on liver cells and so that would make it very drug specific.
DR. WATKINS: So the question was how general would this think pan out forother uses? I don't think the sample sizes are large enough to find it. Or is there evidence thatsuch an association would exist?DR. GOLDSTEIN: I think the absence of association can not be taken asinformative at all because 5701 is at a frequency of 6 to 7 percent in the European population,which means it is exactly on the border of genome-wide association studies exactly. So the factthat we don't see things for other drugs means nothing.DR. WATKINS: So questions from the audience in this area? Roger.DR. ULRICH: So my question is, it is still on the same topic of HLA, is it is allplausible. It all makes sense in the detection. I absolutely agree because you don't see itdoesn't mean it isn't associated. But is it actually causal or is it reactive? In other words, I mean,to your question, Paul, it is what stresses the liver. I mean, is that what varies from one individualto the next and is that component drug specific? Or I mean, we all agree that mitochondrialpoisons are a bad thing. Well maybe there are different degrees of a bad thing and it is really theexpression in the liver that makes the individual susceptible but it just happens to be that thatindividual also has to have the HLA as an additional risk factor. But it is reactive and not causal.DR. WATKINS: Well if you say it is drug-induced liver injury and it is a multistep,multi-hit process, then each hit is important. But as you go along, each one is not sufficient.They are essential but not sufficient. Ann?DR. DALY: I would like to comment. I kind of agree with David Goldstein thatflucloxacillin is an extreme case. I think it is at one end of the spectrum. Possibly something likeacetaminophen may be at the other end of the spectrum and everything else is somewhere inbetween.The co-amoxiclav data, for example, is a lot less clear-cut than the flucloxacillindata and also I was going to ask Raul to talk in a little bit more detail about why perhaps the coamoxiclavdata from Spain in relation to HLA Class II seems to be panning out to be a little bitdifferently from that for other studies, the Northern European, in particular.DR. ANDRADE: I think the main reason may be that the implementation inSpain is quite different. We find a lot of more patients expressing the hepatocellular damage.
I want to also know the difference in control population accounting for data tofind more strong association with HLA frequencies.DR. DALY: I think not. It is only not about controlled population frequencies,too.DR. NELSON: But the following question is for Matt.DR. DALY: It is about the control population frequencies and the power, Ithink.DR. NELSON: Oh, in terms of power, yes. So, certainly I didn't catch thenumber of controls from the slides that you had shown. But certainly the affect of coamoxiclavand I think some others were four or five and so with the loss of controls, it would be difficult todetect.DR. WATKINS: Questions?PARTICIPANT: So I want to pursue with this HLA theme. I understand, sittinghere this morning you assume that a large number of cholestatic drugs have had the HLA comeup. Dr. Andrade, I was kind of interested in your cartoon there which said that the cholangiocytesmight be antigen-presenting cells.DR. WATKINS: We have to have more than one mike here.DR. ANDRADE: In some examples they have been shown that they are ableto express that.DR. WATKINS: If I could make one comment. With cholestatic injuries, classicexample is Augmentin, is often a delayed reaction at least becomes clinically evident days oreven weeks after the drug has been stopped and it can sometimes progress to a vanishing bileduct syndrome. So at least in my mind, it always seemed immunologic; whereas what we arenow seeing are HLA associations with hepatocellular injury. So from a clinical hepatologistperspective, I always thought the HLA associations would only be with cholestatic injuries but infact they seem to be more general.Comments from anybody here? All right, so I'll ask another question.Are you going to make the genetic data from the SAEc public? And thenindividual researchers are going to be able to access the terabytes of data through the net andactually it will be in some sort of form that I can test hypotheses, , for instance, checkassociations with my own favorite polymorphism. Can you expand on that a little bit?DR. NELSON: Yes, so the bulk of the download will probably make this a bitdifficult for an individual researcher who doesn't do this type of research to access and usebecause it is a huge file and there are specialized formats that are probably foreign if you are not
computationally inclined. But the interface that I showed you that they are working on, and I don'tknow exactly what the status is of that, but they are using the CAB framework which should allowyou to pick specific general regions and define the hypothesis you want to test and actually carrythem out. I don't think that that functionality is there yet but it is coming.
DR. WATKINS: Arthur?DR. HOLDEN: The Consortium updated data at the end of December, earlyJanuary. There was limited functionality in terms of what we will call varied capability. And foreach of the major clinical and demographic and genetic elements it could be a percent of a queryresult. You would specify that element, go down into the data warehouse, extract the related dataand review it and review the profile of it. And then if you want to amend that query based onwhat you see, you will be able to build on it.And again, this is the CAB architecture. What we are trying to do here is,hopefully not only with the continued effort of the Consortium, to build these open sciencedatabases. And I would hope again, I would encourage everyone as you go over your datasets atthe appropriate time. We will figure this out much better together than anyone who goes it alone.The concept is to get as much of these data out to the public domain. But I would say practicallythat by the middle of the summer, the query capability will be there completely but we will havethe serious skin rash data and the initial release of the DILI data. But this will grow over time. So,it should be a very convenient way for you to extract the data that you want if you are a, youknow, blood researcher, there will be a great possibility to define it if you are developing the drugand you want to look at specific types of drugs and a specific phenotype, you will be able to drilldown and extract that data.DR. WATKINS: David?DR. GOLDSTEIN: Just a very specific comment to your question. There isexisting software available, publicly available software that allows researchers to ask exactly thequestion they propose: which is this given polymorphism, what is its degree of association in thestudy; or in this given gene, what is the past association for any polymorphism in the gene just byentering that polymorphism and the gene. And that is certainly how these data should bepopulated so the researchers could do exactly that. So there are existing routines for doing thatand it should help you exactly as you describe.DR. NELSON: All right. So the data that are up there now are actually just thetheater. We didn't publish any of the association studies that we have run yet. But I agree, oncewe feel like we have run the definitive analyses, we should make those available as well in thesequeries.
DR. HOLDEN: One other comment. And for those of you who are academics inthe crowd, this may seem crazy, but we are actually pulling data out before we finished ourpapers. Think about how exciting that would be for moving the science forward.But we simply put in a moratorium. If you access that data, then the resourcesthat are developing, analyzing, putting everything together, we will place a moratorium on it sothey have a chance to completely finish their work. So for example, the serious skin rash data,there is a moratorium at the end of December this year before anyone can extract and use thatdata in a publication.So again, what we are trying to do here is to get more data out because of allthe questions that we have been discussing here and many more. It will still allow the publicationand academic process to go on with its integrity.DR. WATKINS: Well can I just finish up on this? Because the follow-upquestion to that is, if I -- and so with some of the drugs, the number of individual cases is verysmall. So that if I knew my uncle had hepatotoxicity from one of those drugs, I could find hisgenotype profile in the database. The question is what safeguards are there for geneticinformation and privacy in this?DR. HOLDEN: Right. So this is not -- this is an important issue, a veryimportant issue. It is also not a novel issue. And it is an evolving context. If you take a look at thegame program, I don't know if there is anyone here from NIH, get involved in that, that is theevolving mechanism for providing access. And they continue to look at the ethical and social andlegal issues associated with it.As Matt alluded to in his presentation, what we are doing is that we ask thatsomeone have a legitimate academic or industrial organizational affiliation and that the individualand the organization sign a data use agreement. And that within that data use agreement, arevery specific conditions about how the information can be used and that they are legally boundby that. So it is that legal binding and it is a put through agreement. We cache store it and thenwith any incident like that, we have got a legal record. We would have legal recourse. It wouldfall to the legal system. That is really the only practical way that we can handle it at this point.But that being said, it is an evolving context. Wellcome Trust has beenintimately involved in this and so we are really taking their guidance.DR. WATKINS: Dr. Regev?
DR. REGEV: So just to make sure I understood correctly, so you described inpatients with bile acid secretion, the secreting channel, as a possible explanation for the fact thatsome people get more cholestasis. Maybe. We don't know. But the question is, do we have anyevidence when we look at Augmentin patients that had cholestasis, this is those few that had thepositive reaction. Do we see differences in these sensations?DR. ANDRADE: No, we didn't find any of this. They couldn't find anydifference in any of the patients. We couldn't find any difference.DR. DALY: We wouldn't have sufficient power to look at hepatocellular casesthough because there are very few.DR. WATKINS: We didn't hear that response.DR. DALY: Okay, I will repeat that. I think it is a good question. We didn't see itin hepatocellular injury cases but I would say that the power we had in hepatocellular caseswouldn't be sufficient to determine whether there was a role for these SNPs.PARTICIPANT: Again from another perspective, as a practitioner's tool for datamining, is there any way that a clinician faced with a new drug can look in some source and say,yes, there have been 15 cases of cholestatic hepatitis worldwide with such and such a drug, tohelp them know when to look for an adverse event?DR. HOLDEN: I think what you just described is the ideal we would love to getto. But no, there isn't, is the short answer.And I mean from an epidemiological point of view is that the work in this area isso weak and general and we know that we dramatically under-report these events. The FDA hasbeen working on a project called the Sentinel Project, which is trying to, in the United States, useelectronic medical record data to get a better sense of what the incidence of these events there.But there are so many problems in the underlying data, in the medical recordswith compliance, it is, we are a long ways away from that. But one of the things that theConsortium is getting involved with that. And in the second phase, we are putting a verysignificant emphasis is looking at a compile of studies we have used in different electronicmedical record formats and standardize definition from a phenotype point of view of a variety ofserious adverse events and hopefully build a system where we can identify these in EMR andIMS and then have a separate basis to enroll and to build a crosswalk. So that will certainly helpus do a good job of looking at all of the ethnic drug differences in this area but I think it is going tobe a good long time before we would be able to get the type of system you described.
Now there is an activity called GTAB, which is sponsored by the NIH which hasa data site which is really focused on kind of pharmacogenetic affects and they are building data.But it would be -- that is not what you are talking about either but that is another element thatover time that could mature to that. So, sorry, no.DR. WATKINS: As you say, the NIDDK has an effort which is well underway tocreate a website, a hepatotox website that will have sort of living or real time text book that willgive you information on cases. We will have the up-to-date experience of the DILIN network interms of how many cases have been found with various drugs, including the identified narrativesand liver biopsy information. I think that will be something that fills the gap.But do either Leonard or Jose have anything to add?PARTICIPANT: One of the reasons that he said these two are working on acomprehensive and useful website to both professional conditions and public. And we haveabout now a hundred or so records are going to be uploaded with cases. And we are alsodeveloping a case form, a comprehensive case form that will allow those who encounter theseindividuals that otherwise are not publishable. To meet those cases do as we all build thisdatabase, we all benefit. We are developing a prototype and we expect to have this evaluated bythe end of this year.DR. WATKINS: Other questions.PARTICIPANT: Just one follow-up on that.DR. WATKINS: Sure.PARTICIPANT: Just one follow-up on the availability on the DILI data on anongoing basis. In addition to the NIH National Library of Medicine Initiative. This includes onlyU.S. drugs, by the way. There is a group being developed including many others in the room,including yourself, that integrates achieved cases from the Spanish Registry, the SwedishRegistry and DILLIN and does so for acute liver failure as well. And the end is to publish thatsoon in drug safety and possibly update it as drugs come in. Dr. Suzuki, our operator is overthere, she is a lead author on that but I think that will be a contribution in making all these caseshave already been adjudicated in a first class action by the way in making information available.DR. WATKINS: Other questions? John.
DR. SENIOR: In all of this work that you are doing in surveying these largepopulations around the world and identifying individual genetic haplotypes and stuff, is there anyclue to explaining the latency between exposure to a drug and the first appearance of anyevidence of injury? Even if they aren't taking the same drug, is there anything in the geneticmakeup that explains the latency?DR. WATKINS: You mean other than an acquired immune response?DR. SENIOR: Well, of course. Not previous immunization.DR. NELSON: We haven't looked at that yet but it is a fascinating question. Ithink there is probably limited data within the cohort we have to be able to ask that.DR. WATKINS: Well that's right. So you would be dividing within a drug earlyonset versus medium versus late in trying to sort it out.DR. HOLDEN: John, I just had one thing. The pilot that we are doing with bothCerner and the VA, we are also structuring a similar one with the HMO Research Network andthe first case we are using is DILI. And the first medical records of the cases we identify shouldhave that information or at least being explored to see if there is that sort of information, whichover time could easily address a question like that but we are not there now.DR. WATKINS: So the SAE Consortium doesn't have time on drug, forinstance?DR. HOLDEN: Most of it.DR. WATKINS: You do? Okay, so you have some of them.DR. DALY: In the flucloxacillin cohort, there are a small number of patientswho developed disease after a relatively short time after they have been on a drug. We have notlooked at this issue detail yet. I think we may have sufficient power to do so. I mean, it issomething we can do. But we have not always had any kind of chance to follow upasymptomatic patients who have had flucloxacillin treatment. And it would be interesting look atB5701 and individuals who were prescribed this drug for any reason to see if there were anyerrant changes in liver enzymes.PARTICIPANT: When you say short, what do you mean?DR. WATKINS: The question is, what do you mean by short.DR. DALY: Well, I think the what I showed you was about 25 days forflucloxacillin. So I am probably talking about something in the region of 10 or 12 days.DR. WATKINS: John.
DR. SENIOR: I was talking to Jim Freston at lunch about Ketek, the first of theketolides as distinct from the macrolide antibiotics. Will Lee and Leonard Seeff, I think looked at50 cases of hepatotoxicity and they found a lot of them showed toxicity very shorlyt afterwexposure. Within days. Four days, six days, three days. Very short and going almost too fast toexplain the development of the immune response in such a few days. Except for the fact, if therewas a hint that some of those patients may have been given a macrolide in the past. So theremight have been some sort of a cross-immunization between macrolide drugss, such aserythromycin or clarithromycin, and the ketolide, which is a different drug. So I don't knowwhether it would be a genetic effect on that or if it is truly immunologic.DR. WATKINS: One comment I might make -- oh, David, sure.DR. GOLDSTEIN: My question is kind of following up about this questionabout Class II presentation mechanism. Suppose you could consider two quite differentpossibilities. One, we know that 5701 presents different epitopes that lead to different T cellbased effects on cells and we also know that it is a ligand which could lead to changes in thepopulations of NK cells that are changing because the ligand is bringing present changes,changes all kinds of NK cells you get.So I guess then to you, Ann, do you have a thought on which of these are likelydirections or how one might distinguish those possibilities?DR. DALY: I am by no means a neogeneticist so it is difficult for me tocomment on that. But I think we do need to try and get a handle more on re-challenge in theflucloxacillin cases. And I think if we could understand that a bit better, that would help us answersome of these questions. Because, as we suggested, my suspicion would be that some of theseindividuals who develop flucloxacillin toxicity may have seen the drug before.DR. KAPLOWITZ: The question that was just asked is something that I wasgoing to touch on later. But it is, you know, it is clear that the interactions between MHC I and IIclass molecules and killer immunoglobulin receptor and you know, NK and other innate let's sayreceptors, can be antigen independent. So you know, I probably wouldn't argue that a strikingassociation like that is likely to be antigen dependent but certainly there is room to consider thepossibility that the so-called adaptive immune response regulates the innate immune response.And so this may be a very important contributor to injury. Not everything that exists with respectto HLA indicates an adaptive antigen for hapten-definitive process.
PARTICIPANT: I will follow up on that observation. I was intrigued this morningby the results of Paul Watkins' acetaminophen cases. We have had some cases that were reallyclose to acetaminophen and did not have that dramatic response. I wonder if the genetics ofthose patients stayed the same or is that an example of a disturbance in say adaptive immunityversus immune immunity?DR. WATKINS: Yes, so what Jim is saying is true. If the people who arevolunteers who develop the characteristic features on day five on acetaminophen come down.And you bring them in, the patient say, there were just three people we brought in two weekslater that had the same phenotype. But if you bring them in three months later or further down,they have already blended responses or almost no response. And if you take the nonrespondersand bring them back, they are still are non-responders. And that was also seen withximelagatran, confirmed that they had intentional and inadvertent challenges that people whohad the classic ALT elevations delayed more than a month clearly due to drugs and had blendedor no response on re-challenge. And I have seen that with several drugs.So, and at least with ximelagatran, there is a strong HLA association. So thereis probably, just as the acquired immune system can cause a problem, the mechanisms with theT regs, et cetera, can make it go away. So, this is a complication obviously of genetic associationstudies. Re-exposure might make it more likely for someone to have a reaction. In the sametoken, it may change the phenotype completely. And what actually causes that toxicity, is it justdrug exposure, nobody knows.PARTICIPANT: Is it drug specific?DR. WATKINS: Is it drug specific? I don't know that. In other words, if you getsomebody who adapts and doesn’t now respond to acetaminophen, if we brought them in withanother drug, would it be generalized or drug-specific? I suspect it is drug-specific.The other issue --and so any comments about that? It actually was observedalmost ten years ago who was looking at gassing rodents with a liver and kidney toxin. Thosethat survived the initial toxicity recovered essentially completely, despite continued exposure tothe toxin. And what I am told is that when they stopped exposure and some of the animals thathad recovered were re-exposed, they got much less toxicity then. So there may be animalmodels for this. John?DR. SENIOR: The other comment I would make is that it may not just begenetics because there are other things that change over time, just as the data suggested.
There are a number of dietary factors that change. It may be that different times of year andexercise, or exposure to sunlight. There are so many other things that happens. These aren'tcontrolled populations in the same way as you --DR. WATKINS: I believe there were also negative rechallenges withximelagatran even with the first episode was clearly due to the drug.DR. SENIOR: And there is data much better than I did but as I remember, therewas a fantastic adaptive response in some patients. So that is kind of -DR. WATKINS: With continued treatment.DR. SENIOR: Continued treatment, yes.DR. WATKINS: That would make sense. But this was stopping and rechallenging.Who would know to confirm or refute it? Do you know?DR. SENIOR: And a single dose was a re-charge as I understand it.DR. CEDERBRANT: Yes, there was a group of people who were rechallengedwith one dose and with those we couldn't say. But this was a very small group of oddpeople.DR. SENIOR: But then the original studies up to in that one study, you didn'tsee the effect of the of the male to it anyway --DR. WATKINS: Okay, so I am wrong. But I actually have been told that peoplewere restarted for a prolonged period.DR. SENIOR: Well, I surmised that.DR. WATKINS: INH is a very good example but -CO-CHAIR SENIOR: They stopped one and they restarted?DR. WATKINS: Right. And frequently in India and China, very commonly, ifthey see ALT elevations, they stop and then reintroduce later and they don't see any ALTelevations at all. There is only a weak HLA association.DR. SENIOR: You can't just expect genetics and an adaptive immuneresponse to explain all that. There is so much more that changes, its time, its diet, it gets older,you stop smoking, you walk home, whatever it is. There are so many things which can confoundus.DR. WATKINS: Right, which is the next question which I think I know theanswer to. You can't say AstraZeneca, you have shown a metabolomic correlation which was alow pyruvate in the serum was predictive and you also have a genetic association. Whathappens when you put them together? And I know the answer is you can't say because youdon't have serum and genomic DNA in the same patients or at least there was blinded in such a
way you can't make that correlation. Which is a plea I think from this morning's talk was veryconvincing if you are collecting genomic DNA delivers kind of a liver safety signal. But I think themissing link is also collecting and banking serum to be able to start putting phenotypic datatogether with genotypic data because that may well be what the real goal is.
Other questions? Rick.PARTICIPANT: Yes, I just pulled up one study and you were dosing the twoweek study and the re-challenge.DR. WATKINS: yes.PARTICIPANT: And so we had actually three responders who came back, rechallenged,and they responded as well.DR. WATKINS: Are you talking about the animals or humans?PARTICIPANT: No, the humans.DR. WATKINS: They all were less, though, in that study. But some had someresponse, yes.PARTICIPANT: I have got the figures, if you want to see them.DR. REGEV: I would just like to suggest an explanation that is not just ahistory that may quickly be unrelated to immunological mechanism. And this story withconclusion of the liver transplant in animals and humans where after transplanted, there is acertain amount of damage to the liver; we have to understand ischemia and so on and retrofusion.But if you primarily go with hypoxia in all the transplants, then after transplant, you won'tactually see any patients where this will not occur. It has been well done with --DR. WATKINS: Yes.DR. REGEV: But this is probably unrelated to immunology it is just some typeof causing damage and having the liver in some way protect itself from the next insult. So wedon't really know what it is.DR. WATKINS: Yes, well in terms of coming back. But the idea is you cancondition the liver to withstand additional insults drug or otherwise is well established. And thereis quite a bit of data on mechanisms behind that.One other comment that maybe didn't come out today but with the HLA*5701association with flucloxacillin it may not be good enough to screen people before they go on thedrug. I guess that is debatable. But it is the single, I believe, diagnostic test if you have a patientthat develops liver injury and they took flucloxacillin. We had only one piece of information. Notage. Not when did they take it, when did they stop. There is nothing that would come close toknowing their genetic status in that regard. That would be the single most important thing. Sothere is a lot of value in these things, even if it is not in terms of screening the populations andmaking drugs generally safer.
DR. NELSON: You bring up a good point about clinical utility and you know, ifyou crunch the numbers and think about well is it worthwhile to pre-screen for 5701 beforeprescribing the drug, it is probably a pretty hard argument to make. But if we reach a point wherethere is a common panel of thing that we tend to test for because they are involved in drugresponse and people are tested for these at some point early in life or at some milestone wherethey might be receiving these kind of drugs, then that kind of information can come into aphysician's decision-making about which drug to treat a patient with. If they knew someone was5701 positive and they had an alterative to flucloxacillin, then it might be a good choice to treatthem with something else.DR. WATKINS: Right. And the other thing that came out of that study was thatthe actual causality assessment that came along with the patient was almost irrelevant. Am Ioverstating this? That the genotype for that drug would become a better gold standard in thesetting of a liver injury than what some filling out a RUCAM or expert, whatever the assessmentwas.DR. DALY: Okay. I don't know that we have enough data really to make thatconclusion definitively. All I can say is I believe there were three or four cases who had a lowRUCAM score. And as far as I can recall, the majority of those were B*5701 positive. Thepercentage that were B*5701 positive wasn't any different between the three categories that wedefined on the basis of RUCAM score.DR. WATKINS: What is a pre-game ideal dead-on data but we will see howthis works out.DR. HOLDEN: Paul, I just want say what that said is where we had theRUCAM score, giving all the pluses and minuses of the patients on that, we still are going to tryto take a look at that in aggregate and see what kind of effect that has.DR. DALY: One thing that we did find when we did the scoring was that a druglike flucloxacillin is probably relatively easy to score for a number of reasons because patientstend to be on it for short courses and it is relatively easy to pick up a number of the factors thatare needed.I think it is more difficult than other drugs like the NSAIDs to do the scoringproperly. So possible the story we did was easier than say some of the work at the DILLINnetwork have published recently where a different group of drugs have been studied.
DR. WATKINS: Any final comments before we take a break? All right, if not,let's have a round of applause for our speakers.
(Applause.)DR. WATKINS: And if everyone could be back at a quarter to 4:00.BREAKDR. WATKINS: All right, I guess we can get started again. Our next speakeris Roger Ulrich, of Abbott to Rosetta to Rosetta Merck and now one of the founders of acompany called Calistoga. He is going to talk to us on genotype-phenotype relationships.Roger.Link to Notes: Ulrich SlidesDR. WATKINS: I think we will have questions for the panel later.All right. So, the next speaker will be David Goldstein. I won't make anyDuke/UNC basketball comments because it would be cruel. He will be talking about implicationsof genetic susceptibility to DILI.Link to Notes: Goldstein SlidesDR. WATKINS: Okay, so David has got to go so questions have to be askedright now. Mark?DR. AVIGAN: The idea of a rare variant causing the problem suggests anothernon-phenotype, classic DILI phenotype. So what we heard this morning from one of thecompanies is about the drug lumiracoxib, where in fact with particular abnormality, particularlythe HLA marker, there was in fact a gradation of abnormalities, depending on the level ofenrichment making it sound like that HLA allele was a necessary but not sufficient effect to themost extreme phenotype. So again, a little bit against what you were saying which was that youwould expect that that enrichment for them, as you march up severity, that enrichment would goaway or not occur. In fact, the showing factor was greater representation of the more severephenotype.
DR. GOLDSTEIN: Yes, the basic framework emphatically does not depend onMendelian-like causation. It just depends on strong causation. So the framework is one of highpenetrant but certainly not Mendelian causation. What you need is strong enrichment infrequency and you need this recognizability. It most certainly does not have to be the onlycause. It most certainly can be modified by things that it is interacting with exactly like 5701.DR. RUSYN: David, you gave an excellent talk. I wanted to ask you aboutwhat Matt was showing as earlier as population stratification or these signs of populationstratification. You know, if you collate your samples from various regions, you know subpopulations,have you looked at that with your data and if there is a challenge, is that going toaffect your framework and do you have ways to deal with it?DR. GOLDSTEIN: Actually that is an absolutely excellent question. There havebeen, as I am sure you know, there have been methods developed that essentially perfectlycontrol for this problem of population stratification when you are looking at common variants. Soit became essentially not a concern. There are no such methods available for working with rarevariants. And exactly how that is going to be handled right now, nobody really knows. And so,what I would say is that that is an issue. One could certainly imagine that there are rare variantsthat are -- well, there certainly are many rare variants that are highly population specific. Youcan imagine that certain phenotypes are also differentially distributed amongst groups and soyou could get a set of variants that are associated for exactly that reason.And there are no statistical methods available now that are appropriate fordealing with that. So, I would say it is a problem. The only point I would make is that the variantsthat we are going to be able to identify in this way, are going to be the obviously functional ones.And what we are describing for stratification will apply to anything. And so if it is only the obviousfunctional ones that jump out, then you can be more comfortable. But in general terms, it is avery good point.PARTICIPANT: I think earlier in your talk you mentioned copy number variantis going to be a major factor but the whole genome sequencing will not pick that up.DR. GOLDSTEIN: No. Whole genome sequencing will pick that up. Certainly itis the best thing out there for copy number variants but it is not very good. So copy numbervariants are not culled, depending on the copy number variant, with high reliability right now fromnext generation sequencing data. And the methods for culling them are in a state of flux.
So, right now I would say copy number variant culling is not very good but it isthe best that we have. As far as how important they will be, we will see. But I feel to a certaindegree there is more, maybe a little bit more excitement and focus on them than is warrantedwhen there seems to be a little bit of a view that is going to be mostly copy number variantcausation for a lot of traits.I think what we have to keep in mind is the ascertainment problem. For thecommon disease associations for copy number variants, those will all come from the G chipsused for genome-wide association studies. Those are not designed for copy number variance, ofcourse. Those are designed for SNPs. And the only thing to catch are these whopping big copynumber variants. So those are the only things we can see that we have made a connection to.Those are the only rare things that we can see and we have made a connection to commondisease. So I think it is extremely likely that other kinds of rare variants that you can't see onthese genome-wide association platforms are also important. They may be more important. Wejust can't see them right now.I don't feel like we need to orient all of our discovery efforts toward those copynumber variants. It may be short indels are important, certainly single site rare variants will betoo, and so on.PARTICIPANT: David, I guess we have seen the genetic community to swingfrom one pendulum to the other. You know, first, they are candidate genes, the whole genomescans. I will not disagree with that because we have had this debate a long time and we havealways that whatever is important to DILI. But my question is, apart from that, what do you thinkabout epigenetic changes? Do you have a sense of that? I would like to hear you commentabout those kind of changes.DR. GOLDSTEIN: Yes, so I am kind of a traditionalist. So I am sureepigenetics is important but my basic take is that we do have what some people call a darkmatter problem. We have high heritabilities for most common diseases, very high heritabilities.And we now know, and I think this is now unequivocal, that for many of them, common variantsdo very little. So then you ask what is the rest of it and that is the dark matter problem. And somepeople suggested that it is epigenetic processes that end up looking like heritability and I do notbelieve that. I think that the 85 percent heritability for height is because genetic differenceinfluence height. It is just that we haven't found them.So I am not prepared to endorse the idea that it is all epigenetic because of afailure of genome-wide association studies. And that is especially because with hindsight, thelimited return from genome-wide association studies maybe shouldn't have been so surprising.
Some people say they weren't surprised by it. I, myself, am quite happy to say I was surprised.Okay? But maybe it shouldn't have been so surprising because after all, we are talking aboutvariants that selection doesn't do anything to. For them to get to a frequency of ten percent, 15percent, selection doesn't do much to them unless they have been selected for it in a veryspecific set. And therefore, probably that is the reason they don't do much.
DR. REGEV: David, just a comment. That story was great. A great talk. Butyou did say something that was something like we must choose patients with extremephenotypes and DILI is an extreme phenotype. And I want to make sure we understand that weare talking about the same thing. Here DILI is an extreme phenotype. We can't do these studieswith patients with more than three times the upper level of normal. We have a mess of data thatwe won't be able to understand. It has to be severely - it was from the assessment standpoint,was perfectly related to the drug. Otherwise, we will never get out of this study data.DR. GOLDSTEIN: I think that that is exactly right. So, for this framework towork, you really have to have a trait that is rare in the population. And it also has to be geneticbut you can't know that going in but it has to be rare in the population. So you really need to lookat the extreme end of the distribution. And that is why I say you have got to be really carefulabout the selection when you push into this. But having done that, I really think that it isreasonable to be optimistic that you can discover in this framework.DR. WATKINS: Can I just make one comment? Because I feel like I have tochallenge a little bit what you just said. Maybe I am wrong but the idea that common variantsprobably don't do very much because they are common.DR. GOLDSTEIN: That is exactly what I said.DR. WATKINS: But how many people take grams of an anti-seizure and thedrug has really only been around for a hundred years ago and some drugs only 20 years. Soisn't that a whole new environmental challenge?DR. GOLDSTEIN: To be sure. So in fact that is exactly. That is why I said Ipart company slightly with Matt and myself from several years ago. I have made that argumentmyself exactly in print. And I still believe it to a degree, which is that, exposure to a drugrepresents a new challenge that in general evolution may not have spoken to. And so one cancertainly postulate that it could be more common, that common variants have a major effect inthat setting than for something like blood pressure where if you have a variant that doessomething dramatic for blood pressure, the individual is going to be in some difficulty. I think thatis true and I think that some of the examples that we know about are probably for reasons likethat, like warfarin, for example.
I guess the other point there is that if much of why we differ from one another isdue to rare variants, and I think that is the lesson of looking at other complex traits, part of thatwill have to feed through to drug response, too, even if there is more tolerance for big affects. Sothat is the point I was making.PARTICIPANT: I have a question about your recommendation of doing thegenome-wide testing in severe individuals. What do you suggest for the controls? Are yousaying that we should do this in a setting where we are already getting genetic testing of sort of amore basic nature in the study or are there, can we use the database or whatever for ourcontrol? In clinical trials, a lot of times what happens is these things happen in like Phase II andmaybe they didn't really plan to do the genetic testing or you are in a country where they make itreally difficult. Whereas, if you have an individual patient who really got into trouble, I think youhave about 100 percent acceptance of genetic testing that is going to possibly figure out whathappened to them. But doing it on a broad basis, we are doing the best we can but sometimeswe just can't there.DR. GOLDSTEIN: So are you talking about the discovery phase or once youhave a set of variants, genotyping them in advance to decide about what drug or both?PARTICIPANT: I'm saying that we have done a clinical trial and we get our firstcase.We can do a genome-wide testing in that case and then maybe as we plan the programand figure out what that case has, then we can do more.DR. GOLDSTEIN: Yes, that would work. So for this framework it is in fact theexact situation as Matt talked about for the genome-wide association phase. You do not have tohave exposed controls for carrying out this kind of study. You just have to have controls. Youhave to have population samples. So in talking about DILI at the extreme, so if you take a set ofDILI cases at the extreme and do whole genome sequencing on them, you can compare whatyou get to random population samples but have also that whole genome sequence to look forfrequency differences that will work just fine.PARTICIPANT: That would be very, very doable.DR. GOLDSTEIN: That would work just fine. If you pursued an alternativething like modifiers of 5701, you would want exposure. But in the general discovery framework,you would not need it.
PARTICIPANT: I had a question about the 10,000 critical sort of variant sitesand how confident you are that that is the window that you need to look at and no other place inthe room. Because again, of course stop codons and changes in open reading, frames andillusions, those are no-brainers. But there are a lot of important sort of functions of DNA,transcription, regulons, et cetera, which may play an important role if you are outside of that10,000 window. So that is important.The other question is let's say you do discover the cause of DILI in patient Xwith the extreme variant. What is the lesson learned about patient Y?DR. GOLDSTEIN: Yes, so those are actually, those are both very good points.So on the first one, 10,000 sites in the category being likely to be functional, no I do not think thatwill catch everything. The point that I am making is that I am sure that it will catch some of it andit is doable now. To catch the rest of it that is not so obviously functional, it a gene that hasn'tbeen annotated yet or it some distant regulatory sequence that has a variant in it, that will haveto depend on only association evidence, which means we need more samples. So things can bepicked up only when we can do whole genome sequencing in the kind of large numbers that westarted pursuing for genome-wide association. That will come later. But what we can do now isget the obvious functional ones.To buttress the view that some of them will be like that, I just refer to Mendelianthings. So we are talking about high penetrant things, not necessarily Mendelian but highpenetrant. If you take the example of Medelian variants, those are very often recognizable in astrictly genome context. So I think that that example is fair. As far as what the cause in patient Xtells you about patient Y, that is the big downside to the direction that genetics is going. Geneticsis going in this direction. I think that is pretty fair to everyone. But what that means is that patientsare broken up into more subgroups than we were imagining before and that makes things harder.There is no other way to put it.PARTICIPANT: Quick question for you and for the representatives of DILINand the Spanish Network. Are these networks gearing up to jump right in to this framework withyour repository materials?DR. WATKINS: Unfortunately, David is on our genetics committee in the DILInetwork and we listen to everything he says and he still says do the GWAS, as common as it istrivial, I just want to get it out of the way. But it is definitely doing some sort of sequencing,whether it is exon sequencing or whole genome sequencing, it does seem like the direction togo. And I think you may want to speak, I think Arthur and the SAE Consortium is going the sameway.
And our thought is, you know, rare variants may just find out the cause in oneperson but presumably, it is going to then give you a clue to pathways and things like Roger wastalking about, as we get smarter. And that is the reason why there is this new initiative now withFDA and my center to actually come up with computer models to store all this information so youcan begin to say, well, if this rare variant causes catastrophe, then we should look at variation inthese other genes, maybe common combinations of SNPs.PARTICIPANT: Are you going to start with your extreme cases? You havesome cases of acute liver failure very carefully phenotyped.DR. WATKINS: Yes, I think the -- oh, so the question is were you going to startwith people with extreme phenotypes like needing a liver transplant or came close to liver failure.I think there are two considerations. That would be if that it was very clear thatit was the drug. But also patients that we have a lot of numbers in would be ideal. So the DILINnetwork, something like isoniazid would be something close to perfect pristine cases. And wemaintain the identity links and can actually call these people on the phone and do genotypephenotypestudies in them as well. And so that would probably be the strategy and we will listento Dave and I am quite sure as we go forward on this thing.Jose over there has to keep coming up with the money to fund all of this.PARTICIPANT: He is going to wait until the end of the year when it is down to$5,000.DR. RUSYN: Again, I would like to commend you for proposing this frameworkbecause there should be some sense of direction. I also want to warrn you a little bit about astrategy for managing expectations because I think that we really have to be very careful aboutwhat we are promising with this, you know, next big thing and where and how we can reallydeliver. And you know, I am putting myself in the strange position of being arguing for cautionand more research being a younger person and trying to solve everything today if not tomorrow.But I think you know, a combination of approaches that Roger was talkingabout. You know, taking a more holistic approach and thinking about pathways. As Paul wassaying, in addition to all of this beautiful discovery, we probably need to be open to the idea thatwe will not find a sample or a gene pool in each DILI case and we actually have to look deeper,rather than charge too fast. And that is just a word of caution and I am sure you have plan forthat.
DR. GOLDSTEIN: Well I think it is a point well worth making. And I think thatyou are absolutely right that we shouldn't over promise what discoveries will be made. And thepoint that I want to push the hardest is that this discovery framework is implementable. Becausethere are skeptics out there that take the view that we can't do this right now. The bioinformaticsis too hard, we can't sequence genomes right now. So what I want to push really hard is that wecan do this.Now, we don't know about the genetic architecture. And what that means isthat we still don't know what will come out of it. That is absolutely right. And I don't think for aminute that we are going to do this sequencing and we are going to have all of the individualswith DILI with identified causes. I certainly don't believe that. I have never seen anything looklike that and that is not how it is going to go. I do think we will find some things this way. And Ithink it is a doable. So that really is the thing that I want to push hard to do. It is a very goodpoint to make.DR. WATKINS: Okay. Let's give a round of applause for David.(Applause.)DR. WATKINS: All right. Well now to bring it all together, synthesize everythingwe have heard and explain everything, is none other than Neil Kaplowitz.Link to Notes: Kaplowitz SlidesDR. WATKINS: So, Roger why don't you come on up here. And we have about20 minutes until the wine arrives. Questions from the group after a long but terrific day?PARTICIPANT: You are talking a lot about oxidative stress. It is an area, Ithink a lot of us are interested in. I think there has been some data in the literature suggestingmaybe people with NASH maybe would be more susceptible to drugs. Well, we don't really havea good clinical test for oxidative stress. And if we were going to look to develop one, what wouldyou recommend and how should we approach it? Because we really have no idea if a patient isentering clinical trials, whether they are under any oxidative stress at all.DR. KAPLOWITZ: That is a very important question, a very tough one.Because obviously, what you would like to have is a measure of oxidative stress in the liver.
And certainly there are, you know, isoprostanes and circulating products of oxidative stress thatcan be measured in patients but I am very dubious as to what, if anything, they mean. And so inan non-invasive fashion, you know, I think it is very challenging. I have often thought aboutlooking at GSH redox status but it is a technically, it is certainly feasible to do it but it istechnically demanding. So if you capture, you know, a plasma sample very rapidly and collect insome derivatizing agent, you can accurately measure GSH and GSSG in plasma. And so, it ispossible that GSSG, an alteration in GSH, GSSG can be measured.The problem in liver is that the selective route of excretion of GSSG is in the bile. Butmaybe under some circumstances, that may not be the case. But other than that, I really don'thave any great insights into that. And I will let Roger tell you how to do it.DR. ULRICH: I love this.DR. KAPLOWITZ: With mice, it's easy.DR. ULRICH: I love to speculate. The problem is, I mean, reactive oxygen hasgot such a, in reactive oxygen species, has such an incredibly short half-life that to take a sampleand actually do the measurement, I mean, I don't know how you would do it. I mean, even if youcould do it under a protocol where you could biopsy, I think the kinetics are just impossible. Butso I am going to speculate.You know, years ago we found that reactive oxygen, and this is a study we didwith diquat, which is profoundly specific at redox cycling and generating reactive oxygen, upregulatesglucose-6-phosphate dehydrogenase, a part of the pathway that deals with reduction.And so I am wondering if there isn't now at least the potential of finding a marker. I don't know ifyou could do this in blood. I wouldn't do it by biopsy.PARTICIPANT: Would there be a change in pyruvate?DR. ULRICH: I don't think you would see a change. I don't see a change inpyruvate until you really knock down glycogen and just depleted the heck out of glucose stores.PARTICIPANT: We had no glycogen and no fat --DR. KAPLOWITZ: So I mean, that doesn't necessarily have anything to do withoxidative stress.PARTICIPANT: No, no, no.DR. KAPLOWITZ: Yes.PARTICIPANT: But I just want to know what to measure. And I think that maybe useful --
DR. KAPLOWITZ: I think you know, certainly in animals it is not. There aredirect measures. I mean, we can isolate mitochondria and measure hydrogen peroxide and atleast for mitochondria measure antioxidant status of mitochondria. So that is much -- and alsogene expression. So what you see in Nrf2 activation for example, you know, in the gene, thatprofile repertoire of genes that generally reflects oxidative stress response.PARTICIPANT: Well maybe then pre-clinically --DR. KAPLOWITZ: That's more reasonable.DR. WATKINS: Arie?DR. REGEV: I have a question to Neil. So during those 36 hours from thewhen the patient took a bottle of acetaminophen until ALT starts going up, which part of themechanism goes into play? And could you measure something that will tell us that there is thebeginning of liver injury at that time?DR. KAPLOWITZ: Yes, that is a very good question. And obviously, it is on apathophysiological level, much better understood in a mouse model than it is in humans. In amouse model, of course, the time course is very contracted. So you know, the glutathione isdown. By an hour it is down to practically zero after a large dose. You know, the activation ofJNK is beginning to occur. There are changes in mitochondrial function that are beginning tooccur. And then after a few more hours, one begins to see some cell necrosis followed by someactivation of the innate immune system.And, you know, how this plays out over the course of 36 hours in a human isnot very well understood. So I mean, it is very hard for me now to explain that.I was very impressed with a paper that Chris sent me recently from NIH fromFrank Gonzalez' group, looking at the time course of changes that occur after acetaminophenchallenge in mouse models. It was very interesting. What he showed was that one of the earliestchanges was the accumulation, and this is in press I think somewhere, of acylcarnitines,suggesting that one of the earliest things that you see is an impairment in mitochondrial function.And so one potential biomarker that even in serum was the appearance within the first hour ortwo before any changes in ALT had occurred was an increase in a marker that might reflectmitochondrial dysfunction.DR. WATKINS: Other questions? Yes.PARTICIPANT: I was wondering if you would be willing to comment on howbiologic products protect which cell would fit into your hypothesis. And I am presuming youwould think it would alter the innate immunity. One particular example would be to sample --
DR. KAPLOWITZ: Well I think that would be the presumption. But there is alsothe issue of activation of underlying autoimmune hepatitis. And this is seen, I think with the TNFantagonistic approaches especially. I don't know if it has been seen with this or not. But in thesecircumstances, the individuals would have quiescent autoimmune hepatitis or activated. I mean,it has been seen with interferons as well. So the beta interferons in MS treatment have beenshown, have been associated with severe liver disease, although ALT abnormalities occur in afairly high percentage of severe overt liver diseases is rather rare. And a number of those casesappear to have the phenotype of autoimmune hepatitis. So that is another potential mechanism.PARTICIPANT: So just about any therapeutic which alters, which modulatesthe immune system potentially could -DR. KAPLOWITZ: In theory, yes.PARTICIPANT: Yes, so we have some data suggesting a transcript on thiscanon provides some insight into that early oxidative stress, mitochondrial stress. And I wouldsuspect Roger does, too. Roger, do you want to comment on that?DR. ULRICH: The answer is yes. But I was trying to think how you couldequate this to a human clinical trial situation. I mean, I am assuming you are probably showingwhere you can actually do blood profiles I think is worth looking into further but I think --PARTICIPANT: We will see some of that tomorrow.DR. ULRICH: Good. Okay so I won't say anything more about that. I thinkthere are still might be so compound-specific that you have to start that work on that specificcompound, before you enter the clinic with it. And so that comes into the reluctance factor of whywould we do anything that isn't required when the most it could do is throw fuel at a program.And I am not defending it. That is just sort of how the industry tends to think. But to get beyondthat, yes. I mean the signals are there. I mean, those signatures I think are really convincing.DR. WATKINS: All the way in the back.PARTICIPANT: We have saw more studies talking about immunogenetics andwe noticed that, for example, in the first talk, that you look at many genes that has changedexpression. And if you take the same people six months later, it is a different profile. And but ifyou do sequencing of genes, you are going to be almost the same as some new small mutations.
If you look at the proteome, like we heard at the last talk, you look at all thedifferent changes in the genetics, they will be underexpressed. And so between the two there is abig variability. And that goes for today, six months from now, year from now, you have thedifferences. So how could you put all this together with that JNK, for example? It is the sameexpression, JNK is toxic but no activational junk is very high to the liver for liver operation. Sohow do you see this genotypically, if you need the proteome to tell you what is going on exactly?DR. ULRICH: Paul mentioned this as well. I think that the important thing is notthat specific protein, or that specific gene, or that specific transcript. I mean, that is where westarted because that is really exciting. When you see something that looks familiar, yourecognize it, you can plug it into a pathway. But I think where science has evolved now, and thisis really where the integrative genomics comes in, which can include of course, proteinexpression and protein function, not just expression, is understanding the network. And so whereyou see a defect, whether it is a genetic or functional defect, or transcriptional defect today, youhave to think not about that specific gene or protein, but about the environment around it andask, you know, a month later, if I see a change in a protein, a different protein, is it still on thesame pathway, is it still in the same network? That is telling you something about the phenotype.DR. KAPLOWITZ: I think, Mohammed, that is a really important issue. And itreally gets to the fundamental question about the interplay of genetics and environment andallele. You know, it is almost like if the planets are lined up in the right way today, that is in yourenvironment, and you have the right genetic predisposition, then something will happen; the nexttime maybe the planets are not lined up that way and having that genetics is not enough.That issort of the simple-minded way to look at it, but it is the best that one could do at the present.DR. WATKINS: Just one comment there. It has been my impression and theimpression of others that with drugs that do have a potential to cause acute hepatocellular injurythat is severe and potentially life threatening, that event occurs within the same time window asthe ALT elevations observed, more frequent ALT elevations. So it does appear that there is aninitial something that goes on in a subgroup and then a subset of those that go on. And webelieve it is pretty rare to see DILI after a year on a drug. Now in part that is an artifact becauseif it occurs and you have been on a drug more than a year, we assume it is not the drug. So Ithink you can only take sort of the changing environment so far. There does seem to be a solidelement of sort of preprogrammed, a person is going to get it or they are not in this, at least in theinitial -- do you agree?
DR. KAPLOWITZ: Yes, I do. You know, there is always nitrofurantoin and afew oddballs that go out there every couple years. But the problem is even in those cases, wedon't really know what has been going on in the patient. It is not as if the patients have beenmonitored. So they just may have a low grade liver disease or some interplay with autoimmunehepatitis that we are not quite clear about. So yes, I do agree with that. It is just that you know,that it is easy to see how sensitization would initiate an allergic reaction in, you know, a fixedwindow of a month or so. What is harder to understand is the variability of the response. Sosome people just have ALT three times the upper limit of normal and some are killed by thedrug? I mean, that seems odd to me. At the same time, you are seeing this trickling of casesdistributed out to six months or sometimes longer with troglitazone. So, did it take that long forthese people to develop an adaptive immune response? So it is a very challenging problem.PARTICIPANT: You just answered my question and that is some patients dodevelop DILI with autoimmune features who were rock solid normal going into the event. That is,had normal ALTs. What is going on there, in your opinion?DR. KAPLOWITZ: Well, I mean, there are two scenarios that one can envision,or three. One is nothing but I think that is unlikely, if you have baseline data. There are somedrugs, I think the statins being an example, where there has been some suggestion thatindividuals who take statin drugs have an increased propensity to exhibit underlying autoimmunediseases of a variety of types, including autoimmune hepatitis, where the statin itself is notviewed as being toxic but perhaps is doing something to alter or to activate an underlyingautoimmune disease, through its biological action, you know, through the complicated things thatmethalonic acid metabolism controls.The other is the kind of situation that one sees well documented with acuteHepatitis A. So, I have seen it and I know the literature has a number of cases where a patienthas no evidence of liver disease. They get acute hepatitis A, well documented and then theirautoimmune hepatitis becomes activated and just remains. So then there are patients whoprobably develop drug-induced liver injury which can activate autoimmune hepatitis. And then ofcourse further complicating the situation are individuals who develop an autoimmune hepatitislike DILI. So those are individuals whose liver disease disappears when you stop the drug. So itis a drug-dependent autoimmune reaction which has the clinical features that areindistinguishable from autoimmune hepatitis, meaning a kind of plasma cell or alymphoplasmacytic kind of infiltrate in the liver and your antibody hypergammaglobulinemia, arather unique set of features or phenotype.
And in this case, the phenotype is drug-dependent. This was actually firstdescribed by one of my former colleagues, the late Telford Reynolds who described this. And itwas really one of the hallmarks in the history of drug-induced liver disease, oxyfenasitin, whichwas a laxative. And being a great clinician back in the '70s, I think, he identified the number ofhis patients with what looked like autoimmune hepatitis progressing to cirrhosis when taking thesame laxative. It just seemed odd to him. He just, you know, he just made note of that. And sohe stopped the laxative in some of them and the liver disease went away and re-challengedsome and it came back. And it even progressed to cirrhosis in those who continued to take thedrug.DR. WATKINS: All right, well I am getting the high sign from Lana, whichmeans the wine is here. But let's give a round of applause for a terrific panel. (Applause.)morning at 8:00.CO-CHAIR PAULS: Thank you everybody. We will reconvene tomorrow