 Okay, so you heard a reference to the meeting that took place Thursday and Friday that Rex and I co-chaired. We called it the Characterizing and Displaying Genetic Variance for Clinical Action, and this was sponsored by the National Human Genome Research Institute and the Welcome Trust. So we actually had representatives from the U.K., United States, and a few from scattered other places. The goal of the meeting was to consider processes, databases, and other resources that were needed to identify clinically relevant variants, decide whether or not they are actionable and what the action should be, and to provide information for clinical use. The idea being that this would be something that could provide an infrastructure assist to more rapidly disseminate and implement. I'm not going to go through the planning committee, but here are the usual suspects, and many of you are here again today. The rationale for putting this meeting together was that we were getting a lot of data around variants of potential clinical relevance, but that while we were collecting information about the variants and to some degrees about the associations, there really wasn't any sort of systematic effort to collect, synthesize, and then evaluate, most importantly, these findings as to whether or not there was clinical ability to use the information. The organizers thought it would be critical to obtain a consensus on what variants are actionable and what specific actions should be taken given the context of the variant and the clinical scenario, and then to try and develop a way that we can make the information available to clinicians, primarily through the opportunity of making it consumable by electronic health records. Now, as I was thinking about these, I tried to think what are the things that would be most relevant to this group, and at least out of these four backgrounds, I think these last two are the ones that are probably the most relevant for the work that those of us in the room are doing. Now, we had heard from several groups that there was a real need for a centralized resource. There was this genomics and health information technology systems that indicated this, the IOM working group. There was an NHLBI workshop in integration, but most importantly, this was one of the issues that emerged from our genomic medicine meeting in June, that this type of infrastructure was really desperately needed. So the output of that meeting, which is going to be presented here at an extremely 50,000 foot version, was some annotations. I need to thank Erin Ramos from NHGRI who put these slides together, and then Rex and Terry who helped to edit them. So these are some of the organizing questions that came out of the meeting. Do we have adequate and accessible data for making decisions about clinical action ability? And the answer is it depends on the audience. That there are some starting points. ClinVar has been mentioned previously. Ensemble is a European equivalent, and these are good starting points, but they probably don't have all of the information, certainly don't have the annotation and updating functions that would allow us to know that the information about a specific clinical action is supported by good evidence and is up to date. It's clear that there's more data that is needed. This will be reassuring to anybody in the room that's heavily involved in doing large-scale sequencing. There were many calls for more data, especially trying to increase the diversity of populations that are being sequenced. So we have a better sense of the prevalence of variants across diverse populations, and that we need to make sure that the database is going to be able to consume and represent these population-based data. There's a significant need for clinical annotation associated with variants in genes, especially because there are so many of the variants that have unknown significance, and so capturing those, you know, accruing information around them is going to be very important, and I think one of the things that emerged from this particular meeting is that it's highly unlikely that this is going to be able to be done as a centralized function. This is something that probably all of us are going to have to participate in, and so really the point of the resource will be how can the resource consume all of the decisions that all of us are making about whether or not we should be acting on a given variant. There was also the indication that it would be useful for this resource to also capture somatic variation. We need a mechanism to capture one-off associations. Again, this was the question of as we look for the missing heritability, there's at least some reasonable evidence to suggest that some of it is probably in rare, high-impact variants that are perhaps even localized to individual families. We have to be able to capture these and annotate them as well. The bullet comes out primary carry docs, but I think the reality is all docs and other healthcare professionals will need user-friendly clinical decision support tools with integration into electronic health records, and the database needs to be able to carefully model classes of evidence and needs to understand in particular things like prevalence and prior probability to be able to make sense of what's going on. I would say that all of these bullets listed under these two slides are potential opportunities for the pilot projects in our genomic medicine group to take on and explore. Second question, what criteria need to be met to consider a genetic variant or pattern of genetic variants clinically actionable? There was a lot of sentiment that the real focus, at least at the screening level, should be on clinical validity, that it might be relatively easy if we can say this is, we clearly have clinical validity or we don't, to at least use that as a first pass to be able to focus on which variants we think are most likely to have some clinical impact. The idea of taking the low-hanging fruit, using genes that we already know and trying to characterize variants in those genes that we don't have information about and characterizing those in terms of their actionability and utility could be developed now. And again, I think given the projects that we've seen out of this group, this would be something where individuals or consortia of groups in the genomic medicine group could work on how do we actually do this? How do we do this binning process, if you will? And I think that there was also the recognition that this type of characterization of the different variants into is this something that is clinically actionable? Is this something that is promising but needs more information or something where we don't have enough information to use? It's a different process than what a clinical laboratory does in terms of trying to determine is a variant pathogenic or not. And we need to develop processes that will be able to support both. And I think there was a fair amount of concern in the room between the laboratorians that are in the process of having to put out reports when they find something versus the clinicians consuming those reports and then saying, okay, well, they found a variant. What do I do about it? And the point was made that we as clinicians when we're sitting in the room with the patient, we have to make a decision and that decision is a binary decision. Do I do something on this variant or do I not do something on this variant? There's not a, well, we'll have to wait until the data comes. We still have to make clinical decisions based on that information. We should ignore at this point bins with no validity. And the point was made that we should treat variants of unknown significance as innocent until proven guilty as opposed to the converse. And this is the situation, the example of, that was specifically used in a talk was mis-sense variants in a well-characterized gene such as BRCA, where the evidence shows that if you have a truncating mutation in BRCA, you can almost be assured that that's of clinical significance, but that there's very few mis-sense variants that have really been demonstrated convincingly to be pathogenic. But if we treat all of those variants as pathogenic when they're in the unknown category, then clinical decisions can be made that can have significant consequences, particularly when we're looking at interventions such as prophylactic mastectomy, cell ping-a-oo forectomy, these types of things. And so the idea being that if you don't have a family history, you have an individual where you find a variant, the prior probability of that individual being at risk is relatively low, and therefore we should take that into account. So that was an interesting discussion. It's something that I think we need to characterize. And of course, the primum no notre do no harm. What is needed to integrate genomic variants and evidence into EHR and clinical use? We clearly will be needing decision support, and there was a lot of enthusiasm for developing a library of clinical decision support tools or open source clinical support that would be formatted in such a way as to be consumable by standards-based electronic health records. Again, a lot of us around the table here have relatively sophisticated electronic health records, and we're interacting with those teams. And if we are developing decision support rules around some of the activities that we're doing, I think we should also look for opportunities to say, hey, we developed this algorithm. Is this something you would like to take and play with? This is needed to address scalability and access, needs to be able to draw from multiple sources, there needs to be integration, therefore it has to be based on standards, and this would include alignment with HL7. So there was a strong sentiment to begin to include representatives of the Genomic Medicine Working Group of HL7 in some of the work that we're doing. Again, we're not doing a great job even with the genes that we really know we need to be paying attention to, and so in some ways we should use those as our opportunities to learn, and those are the types of genes that we're working on because we recognize that this is where the implementation can have the biggest value, and so I think we really have the opportunity to move this forward. ClinVar, could this be the central repository for variant information? I think there was a sense in the room that this would be a good place, although there are obviously other opportunities as well in terms of ensemble, and so the resource that we began to talk about was something that would sort of sit on top of these other resources and would add in an annotation function related to them. One of the things that I think is most important for this group is the idea of creating what we were characterizing as a dynamic loop. So not only do we want to move actionable variants into clinical practice, but we also want to evaluate outcomes and take those outcomes data and feed them back into the databases to be able to refine variants. In other words, do crowdsourcing in terms of learning about what is the actual impact as we implement programs based on these variants? And so this again would impact curation function. It was recognized that we probably need to generate better interactions between epidemiologists, bioinformaticians, and genomic scientists to facilitate obtaining the needed information on clinical validity and utility. I think that this, some of us are beginning to explore the space in our individual institutions. There was also a call to perhaps establish some training opportunities to cross-train some of these individuals. Also, when we update information about a variant, we need to have versioning of this so that we know when it was done and what information was used to make that. We recognize opportunities that there are large databases of data, such as the MedCo, where we heard a presentation from them, where they have a bunch of data. And if the research question they're trying to answer is amenable to using the data set, we might be able to move very quickly using existing data to develop some at least preliminary answers on some of the questions that we're looking at. So there's an opportunity to try and collaborate in broker relationships. Again, a very important function of this group. We really need to develop approaches to gather long-term follow-up on patients with rare variants to understand the relationship of the variant to the disease and other phenotypes. There is the recognition that there are concerns about privacy, security issues that in some ways have hindered this research. But there's also was the concept that perhaps hasn't been explored as much in this group and something we should think about is what if patients have access to the data and actually control the data. And I know that that's something that partners in particular has been quite interested in and that's something to see is that a model that could potentially be scalable. And I think that our projects as we look at them should incorporate pilots in terms of how to explore the best ways to communicate outcomes results from the clinic back to researchers and define how we can move nimbly but without violating any rules and regulations between the clinical and research spaces. Decision support and physician education, the sort of the signature project that was raised is, you know, I want something, I think Gail brought this up. I want something where you can feed whole genome data into software that produces a concise report regarding relevant genomic variants for a particular patient. I think that's definitely a holy grail type of a situation. But for those of us that are moving into whole genome sequencing, whole lexome sequencing in the clinic, I think as we recognize the amount of intensity that goes into doing the interpretations, if we can find ways to automate this, that would be highly useful. Clinical decision support systems need to be scalable. Clinical decision support to this point has been pretty much institution specific. We all build our own. They don't generally work in other places. I think we're beginning to emerge to call for the need of being able to build something that will then work in other places with minor modifications. We should pursue that more. I mentioned the patient-controlled information. We need to be thinking about what are the sort of educational support that needs to go along with the variant information to providers? And we probably need to do some general consciousness raising about the fact that some of the information that we're using in fact does have utility and can improve patient care. And all of these, I think, represent opportunities for this group. These are the draft recommendations and I'm not going to go through those in great detail. I'm assuming that the slide set will be distributed to the attendees. And this will also be the, there will be a paper that will be coming out about this, hopefully on a relatively rapid time scale. But what I wanted to do is to take a couple of things that I thought would be most important for this group to kind of think about. First of all, I think the recognition that the recommendations from this group were in fact responsive to the recommendations that we had out of the June meeting. That we have a pretty good working relationship, probably because there's a lot of us that sit in both groups that allow liaison. I think it would be critically important for this group to provide input and constructive criticism as the database resource is being developed. As I mentioned through the course of the different questions, there are several of the recommendations that lend themselves to pilot projects that could be added to the implementation programs. And so we should be thinking critically about how that could be done and the role that NHGRI might play in terms of fostering that type of pilot program. I think the Genomic Medicine Group needs to provide input on some of the questions about variant classification that we're raised by the other group. I think, again, the majority of us in this room are in the process of doing it, meaning we've addressed these questions in our individual institutions. What are we going to act on? What are we not going to act on? And that type of information will be critically important as we build this resource. We need to not only understand our successes but also our failures related to variant classification. And I think in part as we look forward to meetings of this group to decide whether or not a portion of or a meeting of this group in the future should be around this particular activity. I'm going to say one other thing before I open it up to questions. And that is that in addition to the ClinAction meeting that we had in this meeting, I wanted to just let you know of one other thing that impact that we will potentially have that Teri briefly mentioned in her remarks. And that is three representatives of this group will be presenting a workshop at the NIH Dissemination and Implementation meeting that's going to be taking place in May of 2012. NHGRI are sponsoring this workshop. And I think this will be a very interesting venue to basically put this work out in front of people whose focus is on dissemination implementation science and really get feedback from the group that is perhaps not quite so inbred as us to say is this the right way to go about things? What suggestions can we have to move things forward? How can we incorporate the tools of dissemination implementation science into our work? So with that I'll stop and entertain any questions that you might have. Or and also ask Rex or Teri if they have any additions or any of the other people that were at that meeting. David. Yeah, so it was a great summary and you alluded to the fact that during the meeting there was some discussion about actionability versus clinical utility. Maybe as one of the people in the anti-actionability subgroup of the meeting I should just reinforce that a small group of people strongly objective to the emphasis on the actionability. Maybe Bob Nussbaum, the best one, said we have these perfectly good terms of clinical validity and clinical utility. So we don't need a new term that reinforces some of the payment biases in medicine that we struggle with today that's procedure related and actionability related. I think it does more harm than good. So I was a little disappointed to see this term clean action group as opposed to something that's more neutral. Right. Well, I'll say two things about that. One is this was frankly an issue of time that in the you know the weekend between that meeting and this meeting we didn't decide that rebranding effort was going to be the best way to use our time. The second thing is I'll say is that I don't think that there was generally consensus on that point. I think that there was debate on that point and some of us were discussing the idea that there are things that clearly fall. One of the problems with the term clinical utility is that for many people particularly on the reimbursement side that implies a very high level of evidence before you really deem that something truly has clinical utility where there are clearly things of variance where there is a defined action that you can do if you know that. And the example that I would use is that based on the currently available evidence I would probably not order CIP2C9 and V-Core C1 variant testing as a standalone test before starting a patient on Warfarin but if the patient walked in with that genomic information I would use it. So I don't think that those Warfarin things for many people would meet the level that we would characterize as clinical utility but they're clearly actionable. And it also is sensitive to the idea that when the NHLBI had a related meeting this was a term that they felt very comfortable landing on. So I think that the takeaway is that we've got a bit more semantic work to do on this as well to make sure that we're understanding what we're talking about. And it looks like David Valley has some comments as well hiding behind them. I would like to weigh in on this exact issue which bothered me a little bit as I listened to your talk. To me the term clinical action ability sounds like biology is black and white and I don't think it is. I think except for a few highly penetrant Mendelian mutations most variation in the human genome interacts with a variety of other variants. So to me the term clinical action ability makes it sound like it's very easy black and white. If you see this do this, don't see this, don't do this. And I like a term like clinical utility which seems to me to be much more nuanced. Okay. And I think that this will be an important role for this group to decide as well. And again in the relatively short time that I had to summarize I apologize for characterizing this is that we came out thinking that this is really a black and white world because clearly that was not the case. I think one of the reasons that there was such a strong feeling that we needed to focus on genes that we currently understand well was around the idea that we have a lot of things that we can do much better and things that we do understand reasonably well as opposed to all of the things that we really understand extremely poorly. And I think the other thing is that it's clearly going to emerge into a situation where it's not just going to be variant by variant but we're also going to be looking at groups of variants along with an associated clinical context that are going to determine how we use things. And so we do need to have that type of a holistic approach to thinking about it. But what we wanted to avoid was trying to work out the details of how this might work with things that we don't understand well, take the relatively small subset of things that we do understand well to work out sort of the general principles of what a resource like this might look like and then we can allow that to scale over time as we gather more knowledge that seems to be ready to move into the clinic. Erwin. Oh, I'm sorry, Catherine first. So I actually think the... Can you just identify yourself? I'm sorry, I'm Catherine Nathan, Symphony Music of Pennsylvania, Kate. I actually think the example of Beersay 1 and 2 is an extremely important example because a lot of inroads have been made in this area and a lot of lessons have been learned. I think that, you know, it's important to realize that in fact large international groups that are working on this area, there's a consortium called Enigma which has representatives from groups all over the world that have sort of done these, answered or tried to answer these questions for Beersay 1 and 2 and how do you address them and what are the levels of evidence and what are the different abilities to do this. And I think it's important in this context that we reach out and we reach out to our groups and in fact even, I think importantly, Larry Brody runs the Breast Cancer Information Corps website for Beersay 1 and 2. We have five levels already within that group and the Steering Committee talks about this every single month about how we classify variants. We've seen variants get to be more deleterious. We've seen variants go back from deleterious to not. And I think one of the issues that you didn't bring up is the issue of how do we get information from proprietary labs. And I think that's a huge issue, not just from myriad which is the obvious example but from something like Athena which does sequencing for lots of many different genes and how do we get that information because that is crucially important in this area. Yeah, both of those things were discussed in question. One of the dangers of summarizing a two-day meeting in 15 minutes is that there's a few things that will be left out. And both of the examples that you brought up were definitely talked about and there was an extraordinarily strong feeling related to the second point that we have to find ways to encourage people not to compete on the basis of proprietary information around variants that this is something that is just critically important for everybody to understand. And in some ways, even if patents disappear, this type of proprietary control about clinical information associated with variants will have a very challenging impact in terms of being able to develop tools that are going to allow clinicians to use this information in an inappropriate way. And we clearly didn't come up with any solutions. But one of the suggestions that did come out of the meeting that wasn't on the bulleted list here was that we needed to do a policy analysis related to potential things that policy changes could address, and that was one of the ones that was teed up very highly on that. Rowan? Yeah, Mark, those are all critically important topics. I was just interested in what the meeting actually brought out with regard to shifting the emphasis a little bit away from what you mentioned. There has been reimbursement and clinical action related to reimbursement towards a more preventive medicine paradigm where perhaps somewhat different standards apply when we actually talk about impact of clinical genomic information on risk factors and assessment of risk factors and reclassification of risk based on genomic information in the context of established traditional risk schemes where I think the considerations are somewhat different. They have to be somewhat different than what I hear has been the bar that's been applied for clinical action ability. Can you comment on that? Yeah, we certainly talked about the idea that some of the results of information could be applied in different ways. There's obviously the pharmacogenomic model where you're using this in either to inform drug dosing, to avoid adverse events, improve efficacy, et cetera. There are clinical situations where we're going to make distinct clinical changes based on information, and then there's clearly information related to can we improve our risk estimates to do a better job of focusing on preventive efforts so that given all of the preventive efforts that we could do on any individual person, how can we target the ones that are most important, and how does that impact the individual's behavior, which is a big question that we haven't been able to answer yet. Will this information in fact help us to address the problem that we've always had, which is this very hard to change behavior? So those things were talked about. I think you're right in the sense that there may be some differences in terms of the bar. When you get into the actual reimbursement side of things, of course the preventive stuff gets even more complicated because our biggest payer in the country, CMS, of course specifically excludes coverage of preventive interventions because in 1965 when the law was passed, prevention just didn't seem to be that important. There are some ways that they can get through this, but it really represents an evidentiary level of the United States preventive services task force, which is probably one of the highest evidentiary bars that we would have to deal with in the country. So it doesn't eliminate the issue of evidence, but I think those are the sorts of things where we need to figure out how can we really get that information much more quickly to say this is really having a positive impact. And I should mention that one of the other things that came out that wasn't represented was the idea that we need to apply some of the new and emerging models of real-world clinical trials, pragmatic clinical trials into the sphere that we should be looking for ways to develop a suite of trial methodologies that are well-validated and could be used in projects such as this. Yeah, Dan. Sorry I wasn't there. It sounds like you obsessed a lot about BRCA 1 and 2 as well as you should. I want to be parochial for a second and make sure that the idea of sudden death susceptibility alleles didn't fall off the radar screen. This is a huge problem in the arrhythmia world because as we start to develop whole genome and whole exome sequencing, we're going to find people who have these variants. And the only advice you can offer those people is you should get a defibrillator. It seems to me irrational. And so I'd echo the idea that a place for NHGRI might be to create these centralized resources so we know what happens to people who have these one-off kind of variants and understand the relationship between genotype and phenotype a little bit better. No one center, probably no one country will be able to do that, but there seems to be some mechanism to look at these really rare variants and try to sort out which ones have been associated previously or are being associated with bad outcomes. That may apply to BRCA 1 and 2. It'll apply to all the sudden death susceptibility alleles, and I'm sure to lots of others. Yeah, we definitely had those come up as well as the discovery of a surprisingly high number of what appear to be deleterious variants in genes that are associated with malignant hyperthermia. And so there are those types of things that are clinically silent up until a very terrible event occurs. And I think in the broader sense, there was a real call for we have to do a better job of understanding phenotypes and gathering that type of information and really understanding, okay, if we find this in one of these, in a sodium channel gene, first of all, how do we enhance family history taking around that particular individual to better understand in a more traditional genetic sense what the prior probability or risk for sudden death would be? And then how do we look at potential interventions? I mean, people were talking about beta blockers and this sort of thing, but I share the same concerns that you have. If we're slapping defibrillators into everybody, which has been approached that has been recommended, we're going to be talking about a huge amount of introduced morbidity and potentially even mortality and cost. And I want to talk about this now. Yeah. So could you introduce yourself? Sorry, Jonas, I'll make that from UAB. Should we be looking at clean VAIR as a decision-neutral data source that we can use to advance research in a position to the clinical decision-making group? And also in relation with that, what will be the relationship between clean VAIR and DBGAP and maybe with the resource like the TCGA part where clinical information is also being added to the resource with no assumptions about decision-making? Yeah, I think that the points that you're making are very good ones. Different resources are going to have different opportunities to collect data and the quality of the data that's being collected, the confidence with which we can attribute the data I think will be important. But we're looking at, we have to be able to aggregate across all of the different efforts. I mean, there's, you know, the Cancer Genome Atlas, the ISCA copy number variant databases, collecting clinical data. There's all sorts of places where people are doing this, but what we don't want to do is to have, you know, 55 different resources. I didn't project it, but there was a slide in our pre-meeting inquiry that we sent out saying where do you go to find information? And there was this huge slide of the different resources that people are looking at. And it's clear that it's a lot of work to go to all these different things and that we're not always aware of everything that's happening. So in my view, the ideal would be that we're somehow capturing all of these under what is, would be perhaps more fairly characterized as a service layer as opposed to a database. We're not, we don't import all the data into this resource, but there's access to all the data along with all the descriptors, the metadata about that data, which is if you, you know, are getting this from DBGAP, then what exactly are the circumstances under which, you know, so yes, we've seen this variant three times, but it turns out it's all members of the same family. So maybe that is something different than if we've seen it three times in 100,000 unrelated individuals, that sort of information. Time for one final question. Hi, thanks, David. Just listening to you, one thing I think we really have to have as an important partner here is the research oversight. I think the importance of as soon as this gets into the clinic that every individual sees themselves as requiring follow-up information, et cetera. So I think that's, you know, a very important piece that oftentimes as soon as it gets into the clinic, you know, what do you mean I have to do follow-up data? Yeah. And I think that that was something that came across loud and clear. The clinical laboratorians are, you know, we're expressing the concerns that they have that when they come across a variant that they've never seen before for a specific genetic test, then they clearly want to go back and do more traditional genetic approaches to say, is this segregating with disease in a family, this type of thing. But it's unclear to them, you know, when do I step over the line from clinical information to research because it's not well-defined. And so one of the outcomes of the meeting was the potential to interact with OHRP to ask them, could you issue a guidance to help us define, you know, where we're doing now? Yeah, well, I understand. You know, we all realize that that's in some ways tilting at windmills. But I'll tell you, we had a similar situation with SACGHS where it was very unclear how family history could be treated with an electronic health record in relation to HIPAA. And they wrote a beautiful guidance about family history that was way more liberal than any of us would have assumed that HIPAA would have allowed it. Basically it says, if the patient gives it to you, you can do everything that you would in that patient record, including name, social security numbers, addresses. They didn't care if the patient provided that to you, that was just fine, which eliminated a lot of the needs for things like shadow charts and all that sort of stuff. So if it happens, it can be extraordinarily useful. But I agree, getting that engagement and actually getting an understandable answer is sometimes challenging, particularly in the context that the whole new common rule, advanced notice of proposed common rule is undergoing accelerated consideration at present. Thank you.