P-R-O-C-E-E-D-I-N-G-S. Drug-Induced Liver Injury Conference XIII. University of Maryland Conference CenteEast. Hyattsville, MD

Transcription

1 P-R-O-C-E-E-D-I-N-G-S Wednesday 0 March 0: Drug-Induced Liver Injury Conference XIII University of Maryland Conference CenteEast Hyattsville, MD DETECTING AND EVALUATING DRUG-INDUCED LIVER INJURY AND DYSFUNCTION: WHAT S NORMAL? WHAT S NOT? WHAT SHOULD WE DO ABOUT IT? Opening Remarks --- WELCOME slides and comments --- MS. PAULS: Good morning, everyone. My name is Lana Pauls and I am the co-chair of this program. Welcome to the th Annual Drug- Induced Hepatotoxicity Meeting. We are very excited that you could be here, but apologize a little bit for the construction. But if you had seen this place three months ago, you would be amazed at what they have been able 1 0 to do in that time. gutted facility. It was literally a 1 This is a new venue. Where we have been for the last several years unfortunately has closed down. We struggled to find a new place and the Marriott came through with full

2 1 0 1 power. As for a couple of logistic points, I wore a red suit today so I am very findable. If you need anything, please find me and I will help you out. My assistant, Heather Sanial, is in the back here in another very findable-colored dress. If you cannot find one of us, feel free to go out to the registration desk, but we can certainly help you. Further to logistics, if you wish to get to the restrooms, go out of this room, turn right, go down a ways, take a left, and there they are, a little bit hidden but there. Now we are going to go ahead, and I tend to be a stickler in trying to keep people on time. We have a very, very packed agenda for today. There is an evening session, which is optional, but we encourage people to come. To encourage people to stay, we have a reception with wine and beer to. provide a little stress test on your liver, appropriate since we are talking about liver all day. Without further ado I welcome Dr. John Senior, co-chair and opening speaker. --- Senior slides and comments ---

3 --- Shi slides and comments Brumfield slides and comments --- DR. SENIOR: Thank you, Martha. Thank you, Tim. I would like now to invite Paul Watkins and Leonard Seeff to come up to moderate the morning sessions. Also Ari Regev and Bob Dufour, who are going to speak. And while they are all coming forward, I shall get started with -- MS. PAULS: I'm sorry to interrupt, John. There are a couple more housekeeping items that I did not mention earlier today. Everyone who registered should have a USB bracelet. All of the slide presentations received as of :00 p.m. last night are loaded there. Any presentations that are provided today that may have a little tweak will be posted on our website and several other websites probably about a month or six weeks post-conference. In addition, this entire program is being transcribed. We have a transcriptionist in the back. If you are going to come up to one of the three mikes to ask a question, please speak your name and

4 you affiliation very clearly. Sorry John. DR. SENIOR: Thank you, Lana. What Lana says is very important. Everyone in this room could be an invited speaker, and many of you have been. But if everybody was going to give an invited speech, we would need to have two weeks, not two days. What I am saying is that your comments and questions are going to be captured, put on the internet for the whole world to see. Your comments will not be lost. It is not an idle thing. When you stand up and give your name and your ideas, everybody is going to be able to know about it SESSION I: WHAT ARE NORMAL TEST RESULTS? Moderators: Paul Watkins and Leonard Seeff SESSION IA: HOW ARE NORMAL RANGES DETERMINED? --- Senior slides and comments Dufour slides and comments Regev slides and comments Seeff slides and comments --- Group discussion of Session IA:

5 DR. WATKINS: Okay, I would like to thank the speakers for staying on time to the second, actually. So we now have at least a half an hour for open questions. There are mikes. For our foreign visitors, it will be nice to speak slowly. But I have one question for Leonard right out of the blocks. DR. SEEFF: I knew it. I knew it. DR. WATKINS: If you used baseline, then what would be your recommendation for the cut-offs for stopping treatment? For instance, times the upper limit of normal could well be times the baseline of an average individual. DR. SEEFF: Well, this talk was not 1 devoted to that part. DR. WATKINS: Okay, fair enough. DR. SEEFF: And that obviously that is something we need to discuss. We have argued 0 about it. As you know, the Europeans, and 1 you know well enough because you are part of that group, say that five times the upper limit of the baseline should be at least a signal. You know, as Arie was saying, I think

6 it doesn't say stop. It says there is a problem. And the problem means you bring the patient back almost immediately, depending upon how high that value was and repeat. And if it continues to go up, then you stop. I think that this was a different issue altogether, but you are right, it will probably be different if you use the individual's baseline value. And if you use that, I suppose five times, let's say 0, is 0. I would be concerned. And I would consider this at least something that we need to look into. DR. WATKINS: Will Lee. DR. LEE: Will Lee, Dallas. Len, how do we settle on ALT over AST or some other more nimble value? So ALT is the one with the longest half-life. Actually, alpha GST is a very nimble marker that goes up more quickly and actually comes down with a halflife of about six hours, instead of. So I am just wondering. Just to confuse it even more, some studies are done with AST, some are done with ALTs. What would be the best?

7 DR. SEEFF: I don't have an answer to that, Will. It is traditional to use ALT as the presumably best surrogate marker of hepatocellular injury, as distinct from the AST -- DR. LEE: the tradition. DR. SEEFF: -- maybe we ought to break Well yes, if you have a better test and if you can show that that is more predictive of a problem, I guess that is fine. But I think that at least for the moment, the ALT has been what we have used and that is what I focused this discussion on. We need a better biomarker and that is what everyone is busy looking for. And we depend on you to do it and then you won't have to ask me that question next time. DR. SENIOR: Will, that is a very good question. And what we are suggesting is not that we know the answer and we are telling people to do something else. We are saying: let's look at the data. Let's get the data. Let's think about it a little differently and see what the data are telling us, maybe we

8 will get an answer what is the best way to do it. We don't know that. The other interesting thing that I alluded to this briefly, the issue of whether we are measuring the whole amount of ALT enzyme that is released or just what is active in the serum. Those studies that looked at ALT mass versus ALT activity found actually in that in people that had acute liver injury that there was very, very little inactive enzyme. The ratio was much higher than you saw, even in healthy individuals. So at least ALT seems to be very sensitive to acute liver injury when you are measuring activity. Bob Dufour already referred to the problem with the alcoholic. In the alcoholic, we don't see ALT rises. AST tends to be higher than ALT. Why? Well we don't really know. But we do know something. As Bob pointed out, pyridoxal phosphate is a necessary cofactor for both transaminases and it is an aldehyde. Now what happens -- what is the first metabolite of ethanol?

9 Acetaldehyde. Is acetaldehyde generated by ethanol metabolism inhibiting, inactivating the ALT? We don't know that clearly yet but it is a strong possibility. Maybe we should find out. DR. LEE: Then there is uremia... DR. WATKINS:... which lowers transaminases, right. The only other comment I just can't help making, I have to say is that there are new biomarkers on the way. There is a big safety initiative. The PSTC is developing translational biomarkers that include SDH actually validating them and showing when they are superior is no minor task. And the hope is that companies will begin archiving samples from clinical trials so that there would be thousands of samples in multiple different patient populations to actually qualify and validate these markers, because that will be no simple task. Bob? DR. FONTANA: Bob Fontana from Michigan. Leonard, I enjoyed that provocative presentation. One question I have, from the

10 1 0 1 existing clinical trial databases, of which there are many that have had hepatotoxic signals, such as bromfenac and troglitazone. If you were to use those criteria of the patient's baseline, rather than whatever the upper limit of normal were in those trials, would that prove to have been more useful for identifying who got into liver toxicity or not? DR. SEEFF: Obviously we don't have that information so I can't answer that. I will say the following, Bob, having spent now the last three years at the FDA reviewing case after case, after case of drugs that are in development to which we have been asked to look and see could this be drug-induced liver injury. It has been very difficult for me, at least, and maybe for John, to really understand what is going on. And it seemed to me that the best thing, and I am absolutely convinced, I know it may be difficult, and I know it may be controversial, but I am absolutely convinced that the preferable thing is to use the

11 individual's baseline value. I am absolutely convinced of that now. I think it has to be tested and I think the point that Paul has raised is that if you start with a lower level and you double that or treble that or quadruple that, does it have the same meaning as if you started with 0? I don't know that but I do know there is variation that I see, it already tells me that this variation is beginning to be less among the various laboratories these days. We still don't know what the upper limit of normal is. We don't. We talk about the upper limit of normal as if we knew it. We just do not. There is no upper limit of normal and it is called the upper reference range anyway. There has to be something that is consistent and takes gender into account, and maybe the people who are doing the trials will tell me that in fact this is being done regularly in the clinical trials, or maybe not. I think that there clearly is a gender difference. And the question is does a doubling in a male, of say 0 to 0, have the

12 same meaning as doubling for a female from 1 to? I don't know. I think this has to be tested. But I feel that this is something that at least should be discussed. And I think we are not the only people. There are publications that have suggested the fact that we use the individual baseline value. DR. FONTANA: So certainly it could be tested retrospectively from the existing databases at the FDA. DR. SEEFF: Right. And maybe we can go back. Can we go back to that? DR. FONTANA: And then prospectively, is it better than the upper limit of normal? Then one other thought, moving this forward, this is obviously true for clinical trial development and drug development. But in clinical practice with sporadic DILI we talk about three times upper limit of normal, five times upper limit of normal and, as you know, most patients don't have baseline LFTs before they start any drug. So a clinician would struggle and have to fall back on their local

13 upper limit of normal. DR. SEEFF: Yes, this is largely for pre-marketing clinical trials. And you are right. We don't have the luxury, but I think we need to think about that as well, to see if there is any other way we could manipulate those kinds of data. But we are stuck with what we have when we see the patients as clinicians. DR. SENIOR: We are going to have a really important opportunity to answer some of these questions because we are now getting ready to handle a flood of new drug applications for patients with underlying chronic liver disease, hepatitis C. And we are going to have a lot of people in those studies with varying levels of abnormality before they get the drug. So we are going to be able to start asking and answering some of those questions. DR. SEEFF: And I'm going to be talking about that this afternoon. DR. SENIOR: We are not ready yet to recommend to practicing physicians. I think

14 the clinical trials are probably going to teach us what should be recommended. DR. REGEV: But Bob, I would just like to play devil's advocate for this recommendation, just adding another complication to the picture. I completely agree with the problem that practicing physicians will not have baselines to compare effects to. So they will have to use upper limit of normal. But when we do use this methodology of using two baseline values before clinical trials, which we do occasionally, there are sometimes some impressive changes between the first one and the second one, that are sometimes far more beyond the 0 percent change that was mentioned. And many times with no pathology just because the patient maybe had McDonald's the day before, which by the way was published. You can actually triple your upper limit -- your ALT, by eating a Big Mac. Using the patient's own baseline I think may get us into a completely new type of problem that it will be very hard to

15 tackle. I am sure there are many issues with this. DR. SEEFF: I respect what you say, since you are involved in this. I don't know whether Dr. Florian is here, but I believe he is giving a talk this afternoon. When I presented this to the antiviral group at the FDA, they actually went back to some database, a huge database to see what kind of differences they may have, if they have two values. And actually, it was not terrible at all, at least in what they looked at. But you know what we are saying is if it is going up and they didn't have McDonald's that day, I would worry. DR. DUFOUR: There are some data that exist into which we have looked when we published the AASLD guidelines on laboratory tests for liver disease. The average day-today variation that has been reported in ALT is about percent. And there are a number of things that do affect ALTs. Very strenuous exercise, if somebody has just started lifting weights. That can affect it.

16 You mentioned a type of food ingestion. There are data on even therapeutic doses of acetaminophen, making aminotransferases go up. So you have to be careful about those issues, as well when comparing baseline values or looking at things on clinical trials. The other issue to think about now is that we are moving towards more electronic health records. I work in the VA system. We have a very good electronic health record. I can look at values for my own institution dating back to the early '0s for most patients. We saw a patient yesterday that we had ALTs back to 1 that we could look at. And we were able to compare values from different institutions in the VA system through our electronic health exchange in real time when we are seeing patients. So I think as we move more towards electronic health records, it will be more feasible to have baseline data that we can ok back on for a patient. DR. WATKINS: And just one other issue,

17 1 0 1 as you will hear. In edish, the FDA now has I think well over 0,000 patients with serial liver chemistries, including comparator arms and placebo, which will be a tremendous resource to ask in individual, in the ACS population in Alzheimer's, et cetera. What kind of variability do you actually see month to month? And then perhaps even have requirements in terms of baseline variance that is patient population specific. DR. KIM: Denny Kim from Takeda in Chicago. This session really resonates with me. Coming from a very practical drug development - drug approval perspective, over the years I have observed directly different central labs and local labs where the upper reference range have varied by greater than twofold, or more. And this has significant implications. We are talking from a very clinical perspective. Especially in this arena of global trials, often now more than 1,000 sites, more than 0 countries, there are significant implications and so where our categories of scrutiny and concern are really

18 based on full upper limits of normal. And so I am very interested in hearing about this from a regulatory perspective and also from an approval perspective. Because if you have a very low upper limit of normal, you are going to, in a controlled setting, of course, a lot of greater than three times and such may -- will probably wash out between the comparator groups. But for categories such as the Hy's categories of the LFT criteria where each one is very important, it really does change the tenor of the data and how you look at the data. And I bring this up because I have certainly have seen where there are individual cases in which just a little bit of change will put it into a different category. DR. WATKINS: Right. And to amplify that, the most severe cases often won't be in the clinical trial database, but in a local hospital and may never return to the study. So how confident are we that for a local lab in India, the actual value is comparable to other settings and the only issue is their

19 reference normal population? DR. SEEFF: By the way, when you were talking about Hy's Law, were you were talking about those ALTs about eight times the upper limit of normal and all the rest of it that was shown? Is that what you were talking about? Because that is not Hy's Law. DR. KIM: No, no. But specifically I mean, it is the cases that meet Hy's LFT criteria. DR. REGEV: I have to make a comment that I was absolutely sure John was going to say, so I wasn't going to say it, but Hy's Law is not just those two numbers. It is very unusual that you will have just those two numbers, ALT more than three times and bilirubin of more than two times and base a Hy's Law diagnosis on that. It has to be drug-related and you have to rule out everything else. So it means it is probably not just a fluctuation that disappeared. It is a very clearly drug-related phenomenon. And in those cases, you will probably see a continuous increase. And the discussion on

20 plus/minus 0 units per liter I don't think will be relevant eventually. DR. SEEFF: Arie, what you have just said is very important, because regularly or not uncommonly at the FDA we get cases sufficient for us to determine if this is a Hy's Law case, based on the fact that the ALT was elevated and the bilirubin. Until you identify the fact that this was probably drug-induced liver injury, and have excluded everything else, it is not a Hy's Law case. Because bilirubin of two times and ALT of three times the upper limt of the reference range (so-called norma ) may be due to alcohol, hepatitis C, or anything else. It 1 0 is only when it is drug-induced that it is a Hy's Law case. DR. KIM: And I recognize that. And again speaking from a very practical on-theground perspective, a lot of times the 1 situation is a lot less clear. And I have seen cases that meet the LFT criteria are blinded, and look probably related. But then we unblind a case, and find it is on the

21 comparator or the placebo arm. don't think it is that clear. And so I Let me give just a final thought, though. I like the baseline approach but then there are regulatory implications. Then what are the categories we are going to use of concern based on that? And just another possible perhaps suggestion is we are talking about data standards but I am not sure if we are going to get away from upper limit of normal. But I am wondering if we could come up with some sort of global data standard for ALT upper limit of normal. I think that would help to standardize quite a bit of the clinical development experience. DR. DUFOUR: Yes, I did have one slide that I didn't show at the end, and that is to get a change from population-based reference intervals to outcomes-based really requires a joint effort of not only people in the laboratory but also clinicians. And that is what happened with cholesterol. And it was only when the National Cholesterol Education Program came together and they said we are

22 1 0 1 going to pick one value and that is going to be where you decide to start somebody on treatment where you consider cholesterol high. That is what laboratories did. Before that when I was a resident, upper reference limits for cholesterol were in the 00s for people my age. We don't do that anymore. It is only if we get a joint effort by interested clinicians and the laboratory community that you are going to make that change to health-based reference limits. DR. WATKINS: Chris Hunt. DR. HUNT: Hi, Chris Hunt, adjunct faculty at Duke. I just wanted to say thanks. Great talks! Lots of questions. And I fully endorse your point, Bob, that we really need to get more outcomes-based information and also if we are looking at full baseline, again, we really need to see the data to best determine how, what actually does play best and pragmatically from a safety standpoint. But also, I just wanted to point out the Kim paper that both Leonard and Bob cited. I don't believe that they

23 actually looked -- they didn't exclude patients with hepatitis B. It is not included in their methods, so that is a significant shortcoming. Understandably, it is a giant population, 0,000 patients. But without hep B data in a population that you know has at least five percent or more of those patients will have hep B, I have always viewed that with some suspicion because you can't say that the difference in outcomes is related -- it is not actually a liver disease-free population, in my view, without that information. DR. SEEFF: Of course, isn't that the point? Isn't that the point that if the upper limit of normal was set at 0 and, indeed, there was a lot of hepatitis B, a subclinical hepatitis B in that group between 0 and 0, that speaks to the issue that the upper limit that was used was not healthy. The upper reference level was not healthy. And as you know, there are other papers, not as well done as this one, and I agree, there are things that are missing from this that we

24 could have used, but there still have been a suggestion from others as well that levels above 0 and 0 and below 0 seem to be associated with more serious outcomes than those below. So I grant you that -- but to me it says the very fact that 0 was taken as the upper reference standard and that there was something underneath it, they didn't show it but it must have been there, the causes for that. I don't know. DR. SENIOR: Well, Chris, you are absolutely right. You should raise this caution. There are two billion people infected with hepatitis B on this planet, worse in Africa and Asia and that is not trivial. A lot of those people don't even know they have it. So you can't ignore that. You are absolutely right. DR. WATKINS: Willis. DR. MADDREY: Willis Maddrey, UT Southwestern. I would like to make two points and get some answers from this learned group. What do you do, since you know from

25 everything you have said starting with Big Mac and then starting with big people, and starting with what happens after bariatric surgery, and talking about chronic disease, when in the course of human events with weight loss would you reevaluate your baseline? If someone comes in say six months after bariatric surgery when they have lost 0 to 0 pounds and their ALT and AST will have almost, in my experience in the bariatric clinic, gone down? float your baseline? Then do you Then I have a second concern that I 1 have been wondering about for the last few years, because I have trouble sleeping and I have to think about something, I have been wondering about the variable we never talk about. We always talk about supply side ALT. We always talk about the release of ALT. But 0 we never talk about degradation of ALT. I 1 can only find three good papers, maybe four, that have looked at this and have suggested where the ALT gets broken down. And the theory in the papers is that the ALT gets

26 broken down in Kupffer cells. I just wondered if anyone could shed a light on that and I will tell you why. More and more of the complex drugs are aimed at such things that could reduce the Kupffer cell numbers. This is a specific example, and a couple of you know about it, but I was mainly interested in those two points. Where does the ALT go and how do you readjust the baseline for loss of weight? DR. DUFOUR: I will try to tackle the second issue, which is a really good point. Most enzymes, and what we are measuring as we have said is enzyme activity, and there are a variety of things that can happen to alter enzyme activity and there are very little data on that, some for a few enzymes like amylase, a very small protein that is cleared by the kidney extensively. But for most other enzymes, there isn't much data on what changes activity and that is what we are actually measuring. So I don't think there is really very good data on that. I think you raise a very

27 1 0 good point. As far as changing baseline over time, you are right, there are certainly some things that will change a patient's condition, you know certainly treatment of hepatitis will change their baseline ALTs. And, as you mentioned, weight loss or change of diet is well known that people with fatty liver disease, when they lose weight, their ALTs do go down. So I think you are right that we would have to consider something about a stable baseline value. And that is a good point. DR. REGEV: Just to continue about the hepatitis C issue, and that was actually one of the things mentioned by Leonard in his talk, with the new drugs, as you know, you may have an ALT response within two weeks. So what is the baseline that you are looking at with this new drug when you are looking 1 for drug-induced liver injury? Is this the baseline at the beginning of the study or is it the baseline after the two weeks, or is it somewhere in the middle? So I think that just

28 adds to the complication of baseline. DR. SEEFF: I am going to be talking about this very issue this afternoon. What you are talking about is an extended period of time. If somebody comes in with -- let's say we use that baseline and there is obesity and/or metabolic syndrome and so on. And you have their baseline value at the beginning of the trial and then six months later, they have lost 0 pounds. In the meantime they will have been treated with this drug and I mean measuring it against that value. When it comes to hepatitis C, as I will talk about, and as Arie has referred to, within a week to two weeks of starting the treatment, down comes the ALT. So there are going to have to be two levels in that particular situation and I will discuss that. And this is one of the controversial parts and I would hope that this really leads to a lot of discussion this afternoon. DR. MADDREY: I'm going to do my best to be awake and alert for that part of the discussion.

29 DR. SEEFF: I hope you will. DR. SENIOR: Don't go away, Willis. Don't go away. You raised a very good point. I don't think we know the answer to a lot of that. Now, Will Lee said AST is more quickly deactivated than ALT. That is true, maybe in general, but how much individual variation is there in the rate of degradation of ALT? We just can't take the average and say that applies to everybody. People differ. Now does weight loss change the rate at which you inactivate? anybody know? I don't know. Does DR. DUFOUR: That is a good question I mean I have some very limited data on this. It is a very bad population use because there are so many things we are looking at it. But we tried to look at whether there is much variation in ALT -- in AST clearance in people with ischemic liver injury. You have got these really high values. You can follow it. And there s very little person-to-person variation in rate of clearance of AST, but a much greater difference in the rate of

30 clearance of ALT person-to-person. I think there are clearly factors that will affect clearance more so for ALT than AST to come back to the point that Jim made. DR. MADDREY: But one of you whispered to me during the break, where is it cleared. DR. WATKINS: There was actually a recent paper from Pfizer that showed it in a couple of different ways if you removed Kupffer cells. DR. MADDREY: talking about. That is the paper I am DR. WATKINS: That's it. But actually the difference in ALT levels, I think, was two to threefold. This was not a dramatic -- this would not account for a dramatic rise but could be a significant factor. DR. MADDREY: Okay, thank you very much. DR. TILLMANN: Hans Tillmann, Duke University. I have a question concerning the change of the ALT. I like that idea very much, but probably the ALT in someone who starts with 00 to go five times is more

31 concerning that someone starting of 0. Have you thought about in biology we always use log. So the log to the ten is probably too drastic. But if you would use a natural log, which I think is -- DR. SEEFF: Again, this afternoon I am going to be talking about what you do in patients who come in with abnormal enzymes. And I am going to suggest something similar to what you have to say, depending upon the level at which you started so the fold increase may decrease, the higher up the ALT goes. That is controversial but I think it is open for discussion. So I guess I don't have to talk this afternoon. It is all being done already. DR. WATKINS: (Laughter.) So I think this is our last question. Lana says we have to break. John Sullivan. DR. SULLIVAN: John Sullivan from Amgen. Very nice talks. I just have a comment about the intra-individual variability in ALT. Those of us who have studied healthy volunteers, usually in Phase

32 I units, have noted that there are often substantial increases in ALT and this has been published. It seems to relate in some changes in nutritional status but it is not at all unusual to get a 0 percent increase. And this has been published with the placebo data. And obviously, looking at the drug, it is very difficult to tease out. And then of course there was that Michael Moore piece where they had this quack doctor saying that the slight increase in ALT was liver failure from the presumably too many Big Macs. DR. SEEFF: Isn't it true that most clinical trials do not permit people in who eat Big Macs? (Laughter.) DR. SULLIVAN: Only Wendy's. DR. WATKINS: All right, well thank you. In minutes we will start again. (Applause.) (Whereupon, the foregoing conference went off the record at :00 a.m. and went back on the record at : a.m.) SESSION IB. HOW SHOULD WE MAKE DECISIONS IN

33 CLINICAL TRIALS? --- Hunt slides and comments Throckmorton slides and comments Pauls slides and comments --- Group discussion of Session IB: DR. WATKINS: Okay, the floor is open. We have an extended discussion period. DR. EL-MOUELHI: Mohammed El-Mouelhi, Novartis. Just a question to Christine regarding the Drug X. What was the drug? I assume it hasn't been taken to clinical development, hopefully. Is that correct? DR. HUNT: Actually, I won't reveal what Drug X is but it was a marketed product. DR. EL-MOUELHI: Based on this, it should have been stopped before going to clinical development, based on the criteria 1 from preclinical studies. sense? Does that make 0 1 DR. HUNT: Yes, I agree with you. I mean, that is a matter of opinion but people here can weigh in. DR. EL-MOUELHI: That is why we need to have this benefit-risk to be reinforced --

34 DR. HUNT: Yes. DR. EL-MOUELHI: -- because we can avoid Hy's cases -- DR. HUNT: Yes. DR. EL-MOUELHI: work up-front. -- if we do diligent DR. HUNT: I am in full agreement. What do other people think? DR. WATKINS: Well you know one of the things that everybody says, and I think it is true, is that acetaminophen would not be approved now, obviously, in terms of total dose four grams a day or even three grams a day is a huge load of molecules, quinone metabolite, covalent binding, oxidative stress, and in healthy volunteers you see an incidence of ALT elevations three times the upper limit of normal in about percent or greater healthy volunteers. So it would never get into man. And if it did in a Phase I study, it would probably be stopped from development, even if it was an effective treatment for multiple sclerosis. So the danger now is that with these

35 criteria for stopping development, we know they are very good drugs that would be safe in their own context that are being banned from development at every stage. But maybe that is the price you have to pay to get truly safe drugs out the other end. But other comments from Hans? DR. TILLMANN: Hans Tillmann, Duke. Along the same line, I think it is very important in the sense of nomenclature to probably discriminate DILI from really life- threatening DILI. And as Paul just pointed out, we don't worry about a little bit of DILI here. We see that with several drugs 1 0 where you then have the transition and they normalize or come to a new normal, we are concerned about the severe cases. And having a biomarker for DILI might help you to identify whom you need to look for but what we really need is the severe DILI biomarker. 1 DR. SEEFF: Chris, could I ask you a question that I really should know the answer to. but I would like to get this confirmed? Is it correct to say that all clinical trials

36 require that there be screening or monitoring for the possibility of drug-induced liver injury in every trial and if so, is there any consistency with regard to the timing and the duration in which screening takes place? Do you screen for a year? Or do you screen for six months and if nothing happens after six months you can stop screening? And is it monthly or every two months? Is there some consistency in there? DR. HUNT: Yes. I think the FDA guidance was probably the best sort of framework that people can use to establish monitoring in clinical trials. And I think all companies continue to do monitoring following within the framework of the FDA guidance. And as people have shared earlier that most liver injury, drug-induced liver 1 injury occurs in the first few months. So 0 that is the richest period. But monitoring 1 is continued, I would guess but I can't speak for all companies, periodically throughout Phases II and III. What do other people say - -- there are lots of folks here --- what are

37 other people's comments on that? DR. AVIGAN: I am Mark Avigan, at the FDA. Just to follow-up on the point about risk assessment, preclinical risk assessment. I think part of the idea of the risk assessment that you were talking about showing those various measures really is risk assessment for the drug development program. It wasn't really risk assessment of patients because the drug never got into patients. So what you were trying to gauge with your roundtable was what is the risk of this drug becoming a problem later. And you were making, based on the preponderance of information, making an informed judgment or whatever, let's called it an informed judgment that this drug might become a problem later. In some circumstances, based on that risk assessment business decision, and it is a business decision to stop development of that drug and go to the next product. The problem with these preclinical markers generally, as a general point, is that there is no individual analysis that has

38 absolute predictive value, including the molecular structure, whether it has the socalled alert structures for electrophilic intermediates, whether there are these BSEP effects on transporters, the toxicology screen in vitro. Each one of these with regards to concordance for how they play out in a clinical population later for severe safety effects falls short. And this is one of the great challenges that I think we are facing now. And part of our creative discussion going forward with regards to biomarker analysis and the collection of biospecimens is to try and bridge that gap. DR. HUNT: Thank you, Mark. To make a further comment, if you look at, for example, the parameters I shared, of compounds that went to market with those parameters, 1 exhibited clinical hepatotoxicity. So you 0 1 could say for something that is less than a required compound -- less than required drug, would you want to continue development? I think the more you look at the data, and the more that data can be extended, it is

39 really nice work, I think the more we can make informed decisions. But what are other people's thoughts? DR. WATKINS: John Sullivan. DR. SULLIVAN: John Sullivan from Amgen. Just talking about, as Mark had mentioned, looking at the preclinical model. So there was a HESI group some years ago that looked at the concordance. There was one 1 0 paper looking backward that Harry Olsen was the first author, there was one looking forward that actually had some problems but I don't think it has ever been published. But the predictive, the false positives and false negatives from animal models were sort of concerning. It was, I think, something around the 0 percent concordance from the animal to human. But getting back to when you get issues in humans and liver toxicity, it is all to do with the risks and benefits and 1 the alternatives. For example, bosentann, the BSEP inhibitor causes Hy's Law cases but is the only drug available for, or at least it appears to have a niche for, pulmonary

40 hypertension. Similarly, just about all of the multikinase inhibitors cause reasonably severe toxicity in a rare number of patients and there are Hy's Law cases there as well. But if you have got Stage renal cancer, then you are not going to be too concerned about one in a thousand, one in ten thousand probability of developing liver failure. It really depends on the context. And it is all to do with the benefits and the risks. DR. HUNT: Yes, I completely agree. Excellent point. DR. EL-MOUELHI: Just to follow up on John's comment regarding concordance between preclinical and clinical, it is true it has been done in the 10s, the first study. There is a current evaluation by the IQ Consortium, which is the PhRMA old organization where they are looking at preclinical to clinical concordance evaluating the safety pattern, including liver injury. That is still in early stages but it should be coming, hopefully soon. DR. HUNT: Great, thanks, Mohammed.

41 DR. WATKINS: I have a question for Dr. Throckmorton, which is: I believe it is true that there has been no drug approved by the FDA in the last five years that has been at least permanently withdrawn from the market. Isn't that right? DR. SENIOR: It s been years. DR. WATKINS: And the FDA rightfully pats themselves on the back for that. But one of the costs has been, in many instances, huge, prolonged clinical trials. So with troglitazone, Rezulin, when it was approved, I think in 1, there were just a little over 1,000 patients that had received the drug for six months total. If you had a first-in-class drug to treat Type II diabetes, I think it would take many thousands of patients and, perhaps, even as long as five years is now required. With rivaroxaban, we have at least one person from the team here for that drug, there was clear efficacy in short-term administration for replacing hip and knee joints. But the FDA withheld approval until very long-term

42 studies were done and I believe over 0,000 patients treated total. So what is happening, at least in some indications, is really getting almost a postmarketing experience in the premarketing setting, at enormous cost, but of course mainly years off-patent. And if you are counting on a billion dollar seller at the end of patent life, every year is a billion dollars. These are huge costs. So I am wondering, you know people talk about an intermediate step for approval, say just approving drugs that are part of the Sentinel Initiative, or even using the Sentinel Initiative as a way to accelerate approval. And I am just wondering, Doug, what thoughts you have about that. How does that stand? DR. THROCKMORTON: Well first off, I will just say we are aware of the impact of time and early relationship between time and money as far as return on the market. So we understand that when we ask for additional trials or when there is a delay in the decision-making, there is real consequence to

43 that. And I also agree with you that there have been indications, I would say diabetes is probably the one place that is most familiar with most people, where we have made a decision that we need to have additional information in advance of marketing than we might have historically required. That is not a huge shock. I mean, that is the nature of drug development. We begin with a certain set of data we require and then as new safety information become available or new concerns are raised, under some circumstances, we have been required to get more information in premarketing than we might have previously. The specific case of the diabetes drugs we took to advisory committees a number of times. We talked quite extensively with the external communities to try to understand that impact. I would also say as far as that particular instance goes, the approach we tried to take was as flexible as we could. And so the guidance offers a mechanism for premarket approval -- for approval in advance

44 of the full safety data package that we believe we would like to have. And that has been a mechanism that we have used under certain circumstances. But the message we got back from the community that we talked to was that for drugs for chronic diseases that was intended to treat a population with high cardiovascular risks, having information, having some level of information about the potential risks of the effects of those products on cardiovascular disease was worth it. That is, it was worth that tradeoff between uncertainty and value proposition, cost to development. We don't do that lightly. I mean there are a lot of examples of places where we have placed those kinds of requirements on recently because we understand its impact. It does fundamentally come down to benefits and risks. Can we put a product on the market saying that it has an acceptable benefit and risk profile, based on what we know at present? And under some circumstances, we have had to ask for additional information.

45 The question you asked about intermediate or staged approval and those sorts of things, lots of interesting ideas that people have brought out. You know from I think my perspective, my own personal view is one important aspect of those relates to what happens if you fail to demonstrate the efficacy that you hope to approve under that circumstance. And so to the extent that there was a mechanism for the removal of that product from the market in the same expedited fashion that there was, that allowed it to go on, then those sorts of things would be interesting to talk about. The challenge would come in, it would be if products got onto the market, failed to demonstrate the efficacy that we were interested in, whether it is under accelerated approval or it is some other mechanism. And then we are challenged to remove them from the market or at least fully recognize their value. I think that is where that conversation is. You know, things like staged approval, fundamentally, are going to require

46 legislative help. So I won't be commenting about those specifically. DR. KIM: Denny Kim, Takeda. Apart from the cost and the time required for achieving more certainty in premarketing, I think that is one issue, there is another issue of sort of risk tolerance. We can achieve a certain level of certainty and of a certain level of risk. And it gets to something that Dr Watkins has said about Tylenol, acetaminophen. And I don't know if Dr. Temple is here from the Cardiac Research Consortium, he says it is a little unfortunate that over the last five years we haven't withdrawn any drugs from the market. In the sense that perhaps our risk tolerance is too low. And so I am not sure how I feel about that. I just have to -- so again, that is concept. I am not attributing this to Dr. Temple. DR. WATKINS: I'll let Bob know. DR. KIM: But I have plenty of witnesses. DR. KIM: (Laughter.) It was just several months

47 ago. But I think it is an important thought, an important concept. Is our risk tolerance too low? And I am wondering about the sort of what the panel thinks about our current level of risk tolerance and what we are willing to accept and what may be or if that is changing, the current environment is changing. It is changing all the time, I think, over the last 0 years or so. DR. THROCKMORTON: Do you mean risk tolerance for some integrated societal sense or do you mean risk tolerance around liver injury? It seems to me that it is likely situational. Isn't it? For oncology, for instance, we did the numbers for products that we have approved despite dramatic adverse effects identified at times of marketing, and they have continued to go up. We have done a really good job, from my perspective, of getting products into the market there. In other places, diabetes or whatever, we have heard from the community when we asked them, that in fact that there was a need for additional information. We

48 1 did need to reduce that uncertainty before putting something onto the market. So maybe in that sense that was a lower risk tolerance in diabetes and a higher or unchanged risk tolerance with regard to the development of oncology products. But it feels situational. I am in the middle of the narcotics overdose, the opioids overdose, issues and things like that. And there is a place where there is an extreme, in some cases, highly personalized, very emotional discussion going on around the nature of the evidence that is available to support the use of these products and the risks of those products when used in bad ways. It is another place where people have very strong opinions about how much tolerance the FDA, society, CMS, whoever, should have for a risk. But it is situational. It is unique to that circumstance. 0 DR. KIM: I agree and AIDS is another 1 good example of a therapeutic area. How about just in terms of risk tolerance for DILI and liver injury? DR. WATKINS: Is there at the FDA an

49 equivalent two people, currently, Leonard and John, for cardiac safety? Is there? I mean are there individuals that what they do is review -- I mean is there special emphasis on liver safety that doesn't apply to the other major area in cardiac? I am asking. I don't know. DR. THROCKMORTON: No, I think in some ways -- John you can characterize this as well. I would say the cardiac approaches to safety have been a little bit broader. That is, we have the Cardiac Safety Research Consortium that we formed five or six years ago now that both Norman Stockbridge and myself and Bob Temple are members of. We have continued to try to build those consortial activities in many ways, like John has been doing in liver. I think we have approached the issues a little bit differently, but both of them have been fairly systematic in terms of trying to build those outside collaborations. DR. WATKINS: And the Cardiac Research Safety Consortium, by all accounts, has been

50 very successful, in your opinion? DR. THROCKMORTON: Oh, I believe it has been very successful, yes. It gives us an opportunity to talk to broad ranges of people with interest in the area of cardiovascular safety, understand their issues earlier. It is not a guidance writing body or anything like that. But it is a mechanism for us to get the non-clinical and the clinical communities together to talk about issues around cardiac safety. DR. WATKINS: I have heard, as a consultant, that people say that you can have somebody in a clinical trial have a hypersensitivity reaction and die but that is given less emphasis than a single possible Hy's Law case. There is an impression with some people, I see Jim Freston nodding back there, that at least in some quarters in the industry, they do feel there is unusual sensitivity to liver injury as opposed to others. But comments? DR. STRAUS: Walter Straus with Merck. I have a comment and a question. And the

51 comment is really around Christine's presentation, and I really applaud the notion of having something that is as simple and intuitive as traffic lights. But as I was listening to you presentation, I was immediately parsing this and thinking well you know I am thinking of cases where it is not green, or yellow, or red. It is somewhere in-between and I think that is often where we land. And I think that really what we try and do in drug development, and I am sure in drug regulation, is balance benefits and risks. And we still, we really lack a very coherent approach that is accessible to most of us around truly balancing benefits and risks. We can quantify risks. We can quantify benefits We can't sort of put them together in a very coherent way and these ultimately have to be contextualized based upon medical need. That is the comment. The question really has to do with a reference that you made, Paul, around the Sentinel Initiative and how much it is going

52 to help. And it seems to me that this initiative, which is bringing together data from many different healthcare organizations around the country is going to provide a mixture of types of data and levels of data. It is not going to provide -- it may be better, frankly, than the spontaneous reports that come in but it is not going to provide what we really would like, which is ongoing true electronic medical record level data on millions of patients. It is going to provide, I think, something that is intermediate between being able to do formal chart reviews and being able to troll these enormous data repositories that are accessible at the FDA which would be useful for signal detection. It is going to be 1 something like enhanced signal detection that is not going to provide a whole lot more 0 beyond. That is my personal opinion. So I 1 don't know what your thoughts are on this. DR. THROCKMORTON: While you are collecting, I think you have the tense wrong. It is not is going to help. It is not