 We need my co-chair up here, so Neil, if you would come to the front of the room, maybe not. But we're going to start anyway as in the true Neil Shear tradition, whether you're seated or not, here we go. So Mary, if you could take your seats, Mark, Josh, let's see, other people who I know who are talking. Now, all right. Great. And the moderator for this session is Mike Lee. Hi. Welcome back. So this session is a special topic. So we have four speakers talking about issues from regulatory to money. The first talk is by Mike Pekanowski. He is the Associate Director for Genomics and Targetist Therapy at the Office of Clinical Pharmacology Center for Drug Evaluation and Research at the US FDA. All right. Can folks hear me okay? So thanks for the introduction. This has been a very enlightening meeting for me. I think this is really an important conversation for, from a drug safety standpoint and precision medicine standpoint, and hopefully the outcomes of this meeting will translate something that does improve public health and patient safety. So for the next 15 minutes, I'm going to talk a bit about some of the regulatory considerations that have been involved in labeling discussions related to incorporation of pharmacogenomic biomarkers. So what I'll do is spend a couple minutes on just some general considerations for safety pharmacogenomics labeling in general, then walk through the Steven Johnson attends case study for carbamazepine. I know it's a story we're all very familiar with, but I think it tells a story and illuminates a thought process from a regulatory perspective. And then we'll talk about some other additional pathways to inform regulatory decisions from the FDA-US perspective. So just generally speaking, from a labeling standpoint, we have regulations that state that we must revise labeling when there is a clinically significant hazard that is identified and there's some evidence for causal association between a drug and the clinical outcome. And in some cases, when events are particularly severe, i.e. the event leads to death, there's serious irreversible harm, we're required to present those findings in a boxed warning. So it has a high level of prominence. I understand, based on past comments here and from other meetings, nobody reads labeling, but I think it is important to recognize that all of our tertiary resources sometimes depend on labeling and the global community often looks to FDA labeling as well, and it's very important for the sponsors and what they can communicate with physicians about. So labeling is important as a communication tool from an FDA standpoint. We have a number of other pathways that we can use to communicate risks about medications, most noble of which is drug safety communications, which are posted on the FDA website. Now the trick in the devil in the detail is really the evaluation of these emerging risks. Here in this is spelled out in guidance that we have for industry and the community about our communication about risk, there's a number of things that FDA looks at. So we look at the reliability of the data, obviously the magnitude of the risk, how serious of an event it is, whether it's plausible, and a number of other factors. So there's really a lot to consider when a safety signal is identified and with respect to this discussion, a genetic predictor is identified that may prevent that safety signal. So there's a lot of hand wringing on a variety of issues that goes on in terms of how to communicate and what to say. And additionally, we have wherewithal to consult with our Drug Safety Oversight Board as well as advisory committees on such matters. Generally speaking, if we look at the whole landscape of biomarkers and genetic factors in labeling, these numbers aren't totally accurate as of today, but there's upwards of well over 150 drug gene pairs that have been described in labeling covering well over 100 drugs in almost 50 biomarkers. Now most of this relates to metabolism and transport, about a third relate to the drug target and its pathway, and then there's about a quarter of the labels that actually relate to some immunologic reaction or some other safety signal, and this would be things like potentially Factor V Leiden for thrombosis. But it's important to emphasize that not all of these are necessarily actionable drug gene pairs that are included in labeling. Sometimes it may be descriptive of how a clinical study was conducted, how a clinical trial was run if it enriched for a certain patient population. But about 76 of these are actionable, so we do provide some management recommendation for patients in labeling, whether it be dose adjustment or patient selection. So in looking at the history of what we've included in labeling over the years, we've clearly had a lot of lessons learned. I think the salient points that are common themes for each of these labeling revisions has been that most of the data emerge in the post-marketing setting, obviously, because you need large numbers of patient exposures to be able to even detect these signals. But critically important is that a lot of this is external to a sponsor's development program, so it's not necessarily something that FDA will immediately have first-hand access to. We receive reports from other countries and a variety of different sources. The clinical events that we focus on for pharmacogenetics labeling are usually severe. The findings are very highly replicated. Often it extends known pharmacology of a drug. So for instance, if we know a drug is QT pro-longer and you have high concentrations of a drug in patients who can't metabolize the drug, that's amenable to warnings. And part of this extension of known pharmacology is important because we have to look at the totality of evidence. We often don't have perfect data in hand to be able to inform recommendations or communications. And we're infrequently going to have cases like a baccalaureate where we have some prospective validation study to inform a clinical management recommendation. So we rely on the totality of evidence. And we also have to pay close attention to the treatment context. So as I'll mention in a couple of slides where we seldom say that every patient must be tested for a certain genetic marker. And that is often the case because we don't necessarily know what the consequences of going to a different therapy are and the risks and benefits of those alternative therapies. So there's a lot of different considerations. But to the extent that you have alternative treatments that are safe and effective, that you could have some enhanced clinical vigilance in managing a patient or monitoring for toxicities or that you can adjust doses, those are cases that have been more amenable to making pharmacogenetic recommendations. So I mentioned before that we're often silent on actually making explicit testing recommendations. And this was mentioned before that we have, this is really done in a way to acknowledge clinical judgment and individual decision-making for a patient. So you'll see in our labeling often that we have reference to things like known status or considered testing, which is soft. And even the GAUC guidelines were relatively soft in how they recommended HLA testing for allopurinol, something that should be considered but isn't necessarily compulsory in clinical practice. When we do recommend testing, we've taken a variety of different approaches. In some cases, we've said you should test everyone, as was the case for oligolistatin abacavir. For some cases, we recommend testing targeted subsets. So for carbamazepine, that's an ethnically defined population. For valproic acid, that depends on, for pole G mutations, that depends on the clinical picture and the age of the patient and how they're presenting. For drug metabolism issues, we've also recommended testing above a certain dose threshold. That's for Pimazide and tetrabenazine, where once you achieve a certain dose, then you start looking at how the drug is genetically broken down. Some other considerations, we have reference to specific alleles in most cases when testing is recommended. In some cases, the genetics are so complex that we don't always reference them. Sub-2D6 is not necessarily the easiest gene to genotype, and it's done not necessarily uniformly across labs. Additionally, it's important to recognize differences in population prevalence of certain genetic factors, and we don't necessarily always acknowledge variability in population prevalence across different global groups, but to the extent that you can focus testing on certain subsets of patients, for example, CYP-2C19, where the prevalence is high, we do try to spell that out to the extent possible. So I'll talk a little bit about Stevens-Johnson, carbamazepine, we're all familiar with the story with that, so I won't belabor the point, but just, you know, I presume a lot of people haven't actually read the labeling for FDA, so we'll walk through some of the elements of it that I think are salient. So the box warning is listed up here on the slide. It's obviously small print, you may not be able to read it, but I'd just like to point out that we basically recommend testing for patients whose ancestry is of a population where the prevalence of these alleles is expected to be high. And this is in part because, you know, the Southeast Asian populations have a higher prevalence of this, but also the incidence of Stevens-Johnson syndrome tends to be higher in those populations as well. So in the United States, testing is focused to this specific population. We have a mix of ethnicities obviously in this country. The rest of the warnings are quite lengthy and I think really serve to elaborate on many of the considerations that have to go into making a decision to test. So at the top, we talk a bit about who, what population actually generated the signal, what the characteristics of the study that led to this labeling were. So this was, you know, specifically at the time that this was written, patients of Chinese ancestry in whom the studies were conducted. We do go on to talk a bit about allele distributions in various populations that are known to express this particular allele. We provide management recommendations basically saying that if a patient tests positive for HLA-1502, do not treat them with the drug, absent acceptable alternatives. There's some considerations for screening, specifically with respect to the timing of the adverse event. With more recent studies, we've generated some information on lesser forms of severe cutaneous adverse reactions. So there is some acknowledgement for MPE address in the labeling. And then there's two paragraphs that really spell out a lot of the areas of uncertainty so that we don't necessarily know what the impact of HLA-1502 is on other drugs and also that none of this really should replace clinical management and clinical decision making at the patient level. Now one thing that's been somewhat of a challenge is, you know, we have these emerging data that surface in the literature every couple of years. There's new studies and we have to make some decision as to how to approach it. I think 3101 reflects a fairly complex issue because clearly the risks that were identified in that study published in the journal were not necessarily as high as those observed for HLA-1502 in Southeast Asian populations. So what do you say about that? Do you recommend testing at the same level that you would for HLA-1502 or do you simply warn that this is a potential risk factor in European populations and also since this extends across other different forms of cutaneous reactions there has to be some consideration for the fact that many of these reactions may not necessarily be treatment limiting and we don't know what the risk of this is with other drugs as well. So there's a lot of considerations that go into some of these newer studies that have been reported. Now we also acknowledge in labeling that we don't know what the outcome is with alternative therapies, alternative anti-epileptic drugs. So I thought I'd put up here just for the sake of illustration some of those alternative drugs that went into that decision-making process. So we have oxcarbazepine and phenitone which both have established associations with HLA-1502, S-carbazepine while it's structurally related to oxcarbazepine didn't have very well validated reports of Siemens Johnson in the development program and certainly no studies at this point that describe HLA-1502 it's only been on the market for a short time. So you know there was some very elegant studies that have been published showing that there's common structural features of all of these molecules that mediate HLA binding and could potentially translate that risk across this class of molecules so that's information but we don't have clinical information but you know can we rely on experimental evidence to support screening or labeling recommendations. Given some of this uncertainty we actually required of the sponsor that they conduct that they collect specimens for cases of Steven Johnson in the post-marketing setting so the language of that requirement and the sponsor is required to do this is spelled out here but basically we're having them whenever they identify a case collect a specimen and this is for a range of cutaneous reactions and then we're having them do high throughput genotyping approaches to be able to identify risk factors or at least understand if 1502 is a shared risk factor in clinical specimens. So in the last couple minutes talking about various pathways that can inform regulatory decisions so I've mentioned before we have a number of considerations that go into the decision-making process about incorporating genetic factors into labeling depends on the risk benefit of alternative therapies, how efficient screening would be in the context of bad decisions in the clinic in addition to the event rate and what we're trying to to prevent. We have to consider the diversity of the populations and the spectrum of adverse events and how much we can generalize findings of a single investigation. We also weigh very heavily the validity of the signal and the quality of the research methods as well as experimental evidence that supports the biological plausibility. So when you have a case control study that has 10 cases and eight of them are positive for some genetic factor you know you're investing a lot of belief in biology at some level and it's helpful to have these other streams of evidence to support the decision-making. So we receive these signals of potential safety concerns from a variety of sources. There's some North American efforts spelled out here you know we've looked at serious adverse events consortium studies but we also have some new developing infrastructure in place I think that'll help us characterize these risks and make more robust recommendations as time goes on. Things like emerge the large biobanks being developed at Kaiser and various other safety networks. We also focus on DNA collection and pre-market reviews so we do have guidance to industry that they should collect DNA to the extent possible from all subjects in a clinical development program. If we know that there are factors that could be a potential risk there to we expect that they test these things prospectively but when you have signals such as a safety issue or something of that sort you collect DNA from at least those specific patients who are experiencing adverse events to enable these types of studies to be conducted at later time points. Now this has been a challenge because every investigational program that's evaluated genetic factors on case four cases of some type of skin reaction all of which may be different and of different adjudication it's very difficult to interpret those results but nonetheless you're building a database for future use. We've also in the post marketing setting negotiated studies with sponsors to conduct a variety of different studies so we've looked at some to validate known signals for genetic factors such as 2C19 for clopidogrel, bilinostat for safety and PK by UGT genotype but also on the discovery side. So we have things like tilapivir where we negotiated exploratory GWAS to help identify factors that predicted skin reactions to that product and a number of others where we have required or negotiated agreements to collect DNA specimens from cases as part of their post marketing experience. We have a couple of additional efforts basically we've tried to look at these administrative claims databases to figure out if we can collect DNA samples from those as well as existing biobanks to look at more common adverse events and we've looked at FAIRs to determine what the extent of reporting is there that's proven somewhat disappointing because there's obviously reporting in that in and of itself is a challenge but looking at genetics is even more complicated and then actively surveilling for safety issues in our current infrastructure through Sentinel and whatnot. It's proven quite difficult because the ICD-9 codes that are used to capture testing are very insensitive and there's no way for us to really monitor whether a recommendation translates to clinical care. So I'll close with this slide basically to provide some thoughts on research directions. So I think infrastructure establishment is something that's I think in the U.S. going to be part and parcel to advancing regulatory decisions, patient management and I think we really do need to build a framework for capturing cases and specimens and having some interoperability in our biospecimen repositories, implementation infrastructure as a consequence of creating interoperable repositories, you know, could feedback into this research and clinical application type of model but we do need dynamic clinical decision support. You know, one of the things we often belabor over is, you know, if we make a recommendation to test every patient that could disrupt clinical practice to a very significant extent, you know, if tests were available we could, it would ease those concerns quite a bit. We also need methods for case adjudication, electronic health record studies. This is in the context of real world evidence being very in vogue at the moment. Strengthening experimental models and one thing that we haven't talked too much about is developing efficacy biomarkers because this, while we can identify safety biomarkers, efficacy biomarkers would also serve to further shift the risk benefit as well for these products. And with that, I'll close. So we have time for one question. Can I ask a question? Can I ask a question? Yeah. Thanks very much, Mike. This is really important. I guess my, I hope my question is not too simplistic. Who actually authors the label? Is it written by the company and then there's discussion with the FDA and there's back and forth and ultimately who signs off and says it's done? Does the company, the manufacturer have to agree? Because I think it's important who's in charge of this, who's driving it in a way, not just who's navigating. The FDA and the sponsor negotiate labeling in most cases. When it's a safety issue, there is safety related labeling changes that we can more or less impose. So there is a negotiation. So the sponsor authors the label, there's negotiation and you have more or less influence, which is sort of not, you have final approval as you. Yes. So if you're not happy with it, you just keep sending it back until they make you happy. Yes. Okay. So they author it and they have to make you happy. And in some cases, FDA will author it as well. Okay. For a product that's okay. It's like my wife and I. All right. Thank you. Thank you. The next talk is by Mary Cullington, identifying cultural variation in studies of disease association with MSC genes. And Mary is senior investigator laboratory of experimental immunology at the cancer information program cancer for center research National Cancer Institute. Thank you.