 Okay, I think most people know me. I'm Rex Chisholm from Northwestern, and it's my pleasure to moderate this final panel discussion. Although we've had an enormous amount of discussion already today, I'm sure the panel is uniquely qualified to take us in some new directions. Just to let you know what we have planned for you, each of the panelists is going to make some comments for about three to four minutes just to set their perspective on this question of what are the gaps in terms of research implementation, and I guess maybe we can even broaden that a little bit, although I know they've all thought about this in detail. Not just research implementation, but I think also thinking about quality improvement implementation, because we've heard a lot of discussion today about whether this should be done as research implementation or quality improvement implementation, and that might even be something that we should focus a little bit on. And then I'll ask each of them just to reintroduce themselves to you as they speak. And then after that, each of them has made their comments. We'll start to have a little bit of discussion driven by their comments and very quickly open the floor to participation, not just for questions, but also for comments from people in the rest of the audience. So, Dick, I'll ask you to go first. First of all, let me say that I found the whole day really invigorating, because all of us are dealing with the same issues and we're approaching it in slightly different ways, and I think that's all to the good. Before I talk about where things might be going, I might mention the sort of setting within which what I'm going to say comes from. Number one, and somebody should say this very clearly, anything that we've done at the Mayo Clinic has been entirely because of the foundation laid by NIH funding through the Emerge grant, which we've had for years, and the PGRN grant. So the NIH has laid the foundation for what I'm going to say, and I feel obliged to say that very clearly. What we're now moving to do is to build on that foundation and really the use of the PGRN Seek version 1 for 1,000 patients from our Biobank to sequence that information and put it preemptively in the electronic health record was a pilot study for what we are now doing because that taught us a lot and we're now moving forward with 10,000 patients from our Biobank in collaboration with Baylor for $300, that's the cost with the regulatory cost of CLIA, about half of it's regulatory, about half is sequencing, for 77 pharmacogenes sequence-based because as we look to the future, number one, I heard the discussion about panels versus not panels. All we are putting in the electronic health record are those variants that are part of the 19 drug gene pair alerts we currently have, but we've got all the other genes that are put aside and they were done in a CLIA environment. They are all the transporters, all the drug metabolizing enzymes, phase one, phase two, all the things you would think about. So against that background then, how are we proceeding to use that information? Because our electronic health record goes back 20 years, we're immediately able to do retrospective studies and we're doing that. Number two, we're doing prospective studies and we're going to do that and the approach that we have taken is to ask each of the clinical departments and divisions involved and all you have to do is take $300 times 10,000 and you know what the institution paid in order to develop the data. So we are asking people to write protocols for data access and we're putting the focus on the clinical departments and divisions. Now why are we doing that? Because what we see really happening is that it's a young faculty member with a fellow or two within each department who knows their drugs and their practice and they've taken ownership. Now immediately we have a champion in that division of department. The other thing that I haven't heard here which surprised me but which we're now learning, they do know their practice better than anyone and the one thing, Terry I haven't heard here, is that what I'm surprised to find is the procedures within the hospital in the academic medical center can be changed by the kind of information they're developing and I heard Vanderbilt made a whole lot of money on those multiple stents because that was clinical income. I heard that comment made. So what we're talking about now is actually making our processes that are drug related within the hospital. I can give you a concrete example if someone in the question answer wants to ask about it. And I was surprised to see that kind of efficiency coming forward out of having these data. So what you're hearing is one model in which number one at the top of our list are the economic analyses and that's why Dr. Wilson is here because clearly we went to him for advice and guidance to think about how to do that. Clearly we need to be thinking about what are we doing because I think this is something I hadn't thought through. The doctors didn't order this test. Up until now all of our drug gene pair alerts were when a doctor wrote a prescription. We're now going to take 10,000 local patients who get all of their care at the Mayo Clinic. The doctor didn't order the test and now that information is going to be there. These are mainly in our primary care clinics so there'll be hundreds of these patients for any one doctor, for some doctors. The implications of having information suddenly appear that has clinical implications we already are thinking through that from an ethical and a legal point of view and are making certain that we don't find ourselves with a patient suddenly having an alert firing the alerts only fire when you write the prescription having information that says what your doctor ordered for you is not good for you. Think through doing that with 10,000 patients and we now have 4,000 of the samples are already being sequenced by the early fall all 10,000 of those data will be back. There are some significant legal and ethical issues that we're going to find ourselves facing and clearly what we haven't even talked about cancer and psychiatry which the people at NHGRI know that I have some interest in where we have no biological markers. Those are all things that I'd hope to talk about before we're done and hopefully I didn't make Rex too mad and did I go over my... Thanks, Dick. Steven. So I too will place my comments into context. I lead a pediatric clinical pharmacology group in a pediatric hospital and our institution largely supports our activities beyond what we can bring in in grants. And the reason it does so is that there's not all that information on genotype guided dosing in children because there's not very much information on dosing of medications in children that has been generated in children so therefore it's difficult to know how much to reduce a dose. So part of our role is to generate the information that's going to be needed before we can do any implementation and the other part of our responsibility is to educate the pediatric sub-specialists who are going to take ownership in implementing this in their particular areas. So I've got three pediatric-related knowledge gaps and sort of three general ones. In terms of pediatric, again, the first thing is we have to generate the knowledge base related to genotype guided drug dosing information for kids and preferably in the population that's going to be treated with the medications. One of the things that we have learned is that there is no substitute to generating this information in kids. There are real limitations to trying to extrapolate adult experience in kids and I'll give an example of this with Symbostatin tomorrow. And another important thing that we need to think about is engaging the families, the patients and their parents because this is a two or three-way process. The idea is to implement precision therapeutics into the patient care and you really can't do that if it's just a one-way street. And so we are developing the tools that help to explain to patients and their families what we're talking about when we're talking about pharmacogenetics is an important knowledge gap that needs to be addressed. In general, I think it's important for us to remember that designing a study to look at CYP2D6 pharmacogenetics and the response to an antidepressant misses the point that the proximal phenotype for CYP2D6 is not drug response, it's actually drug exposure. And so we need to start thinking about how we're actually going to integrate pharmacogenomics at the level of the drug target into some of these studies and into clinical care. Dick kind of addressed my next point and that is, is genotyping for one or two common variants really good enough. And there's lots of work coming out now. Mary and St. Jude's methotrexate and the burden of rare variants in SLC1B1, for example. I think we've got to look beyond common variants. And then finally, this is just something that we face on a daily basis, is the genotyping results that are coming back to the direct to consumer. Genotyping companies in the lack of data, prospective validation of some of those genetic markers, particularly with respect to drug response. And if you do a 2x2 table from GWAS data, because there will be some people who respond that have the wild type and the reference to the variant genotype, there will be some non-responders that do. When you do a 2x2 table, sometimes the specificity and the sensitivity don't add up to a clinically useful test. So that's a prospective validation is the last thing we need to do. Last point I'm going to make. Hi, I'm Micheline Picat and I'm from University of Toronto, which is in Canada. Just in case someone might not know. That's why I talk funny, exactly. And there isn't as many implementation studies in Canada as there are in the United States, partially because of the funding. It's a much smaller country, a lot less money for funding. But one of the things is that I see, my vision being a number of the people at the faculty see the pharmacogenetics being implemented in a real life situation. Everyone has their pharmacogenetic test results done. And so in that case, we wanted to see, there's very, the research gap is that there's not really that many research studies that have actually tested the pharmacogenetic feasibility services in a real life situation. So one of our questions was that which health care provider is best positioned to provide pharmacogenetic services in primary care. And this is both logistically as well as economically. And so in our study, we did, we were looking at positioning pharmacists, pharmacists as the front line person for providing pharmacogenetic services. They see all the prescriptions. They often have relationships with their patients. And many of them in the small community pharmacies have very good relationships with their physicians as well too. So the pharmacists, if we educate them, they can educate the community. So we see the pharmacists providing the education to the patients and providing the education as well to the physicians. So we developed some training programs in order to try and address that. And also, I was working on a continuing education program that's online that's available to all pharmacists in Ontario free of charge. So they can do their pharmacogenetics. I also teach in third year pharmacy students. I teach them a course on personalized medicine. And I tell them you guys have to get ready to do this because you will be doing this in your lifetime. But a couple of things that when trying to determine get the payers to agree to cover the costs of pharmacogenetic services is that they want to know the cost effectiveness. So we've talked about this a lot, and especially what's lacking is the cost effectiveness of preemptive testing across a panel of common gene variants. And so you see the single drug, single gene, that's been, there's cost effectiveness studies there. But when we ask for a panel, we can't really, we haven't got the data behind us to tell them how this will help them. And especially in Canada, we have a separate, we have a universal healthcare system. The government pays for our hospitalizations and for our doctor's visits. But all of us have private insurance policies for our drug. So drugs and hospitalizations and physician visits are separated. And so the other, the last thing I think is also, in case of this, is that what's the best outcomes to measure, particularly in community settings where patients may be relatively healthy. So we can't, you know, deaths and heart attacks may be too pronounced. We may have more subtle differences in improved patient care. And so do we measure the number of prescriptions, number of physician visits? Those are the types of things we are tackling with and how to design proper implementation studies. Thank you. Good afternoon again. I'm John Wilson for the folks that don't know me. Once again, I'm the Chief Medical Information Officer at a company called Optum. And to be clear, Optum is one half of United Health Group. The other half of United Health Group is United Health Care. So if you think about United Health Group as two platforms, Optum is one platform. United Health Care is the other platform. Most folks here, when you talk about payers, you're really referring to United Health Care. To be clear, I don't work for United Health Care. I work for Optum. And we do a lot of work with our payer siblings, if you will. Let me paint you a picture about Optum because it will help contextualize some of my comments and questions in this space. So Optum is a health care services company with an application for size and scale. We employ about 140,000 people. And we will... Last year our revenues were out of United Health Group. United Health Group's total revenues were about $185 billion. About $100 billion was from United Health Care. And approximately... Again, these numbers are approximate. $85 billion, let's say, was from Optum. Optum has a number of divisions. We have an analytics division which provides products and services to other payers, to providers, to life sciences companies. We have a... a PBM, a Pharmacy Benefit Manager. We are the third largest PBM in the United States. Directly or indirectly, we'll touch 63 million people through that group, processing close to a billion scripts. So when we think about pharmacogenomics, that's obviously an interesting area. And then finally, we provide services directly to consumers. We are... We have about 20,000 affiliated physicians that work with us and for us. So why do I share that with you? Because when we think about this domain, we think about it from multiple different levels. We don't just think about it from the payer perspective. We think about it from the provider's perspective as well as the consumer perspective. Let me just hit on... I'm getting flagged for speeding up. Let me hit on some of the key issues. And you've heard me talk about these earlier through the course of the day. And I cannot stress this to you enough. We have to get the coding right. It's unfortunate with my accident when you say coding. I don't mean coding the drug. I mean coding... Just call it war for it and it'll be easier for us all. I mean the codes. And let me be clear what I mean by that. And I'll share some perspective. One perspective is that CPT codes in this domain will never be sufficient to meet the needs of it. So let me tell you what I mean by that to push on it one step further. There needs to be a net new coding system. CPT codes may not cut this because the rate of change of pace in this space is far, far, far faster than it takes to get a CPT code. And so part of that, part of this community is if we can work together with you folks to explore some different domains... Maybe I'm getting an alarm. If we can work together with you folks to find new ways to code for this that will help in two fundamental important areas. One is around reimbursement. I cannot stress that enough. The second is around quality. So a lot of people have interpreted my comments saying well if we can get the codes right we can figure out how to pay for them. That's definitely the truth to that. The other part I'd stress to you is that if we can get accurate data by understanding exactly what test has been done we can link that up to all of our other data assets. Let me just digress for how long have I got 30 seconds and what I mean by that. We sit on about 180 million lives worth of claims data. We pull close to a million charts a year. We have access and we extract from about 80 million EMRs. We have a lot of data assets. We're a PBM. We know a lot about what gets prescribed, where we can the ability to link those data sets together with an understanding and appreciation about what genomic test has been done is clearly valuable to this space. I think it could help a lot of the questions that are being answered. However, in order to do that we need to make sure we exactly know what test was ordered. Right now in my opinion the CPT codes don't support that plus to, I think it's Mark, is it Mark? Mark's comments earlier I may disagree with you about some of your comments. We can have a conversation about that later. I think that's healthy. But one area where I do agree with you on is it is important whoever's making a decision about CPT codes actually be equipped to make that decision. And I'm curious I don't know enough about the mechanics of the AMA about how many geneticists they've actually gone on staff when they make these decisions. So there is a point around coding and I cannot again I'll stress it enough. If there's five things I could ask you to do the first three would be let's help figure out the coding, the second one would be let's help figure out the coding, and the third one would be let's help figure out the coding. That will reduce a lot of the tension and friction in this space. I'm probably over time. I'm going to hit you with two other comments. Quality. I cannot again stress that to you enough. It's a real worry right now when you see the variety of quality in this space like any new technology there's variance and that's understandable but the ability to measure that variance, understand that variance and try and drive to an improvement in that space I think it's going to be fundamental. We do worry about this variation. Not everybody can be the best interpreter. We need to have better methods to understand that and again we would be actively seeking folks to engage with us to help figure out what does quality look like. So if any of you are interested in that please grab me later. And the final point is obviously the data in the health economics. There is a sparsity of data in this space. I think that's a common consensus. There needs to be better quality of health economic studies here to help any pair regardless of whether they're an individual insurer or a large pair like CMS or one of the delegated providers or a large employer group understanding exactly what the benefits are going to be. So to summarize coding, quality health economics. So I'm Julie Johnson and I think there are lots of gaps but I'm going to pick one and my argument is that our most important gap remains an evidence gap and we talk a lot in pharmacogenetics about frustrations about genetic exceptionalism, pharmacogenetic exceptionalism look at renal function drug interactions as examples and the reality is those things are often inferred now through simulations not through actual clinical studies but the reality is 10, 20, 30, 40 years ago there were lots of clinical studies not necessarily randomized controlled trials but clinical studies that documented in human studies drug interactions and documented the impact of renal function on drug dosing and things like that. So I think that we are not quite there for most of our drug gene examples in terms of the evidence. We have a few where we are. I think we've been there for a long time with diapurines I think with the data Larry will show tomorrow with clopidogrel. I mean so I think they're coming but we have to continue to build the evidence hopefully eventually we will get to the point at least for the pharmacokinetic related pharmacogenetic examples that how many times do our predictions have to be right. So if we're right 20 out of 20 and pretty much all of our examples so far they line up with what the clinical pharmacologist would predict and hopefully eventually we will become like drug interactions and renal function where you don't have to do a lot of studies. You just have to have that background evidence of the relationship between genotype and in this case pharmacogenetics. I think drug targets are going to be harder. So I think that in the absence of evidence we're going to have a hard time convincing payers, we're going to have a hard time convincing clinicians and I think it sort of underpins many of the challenges. There are many but that's to me the most important one. You showed a lot of restraint. So it's been great to hear that everybody loves CPIC guidelines for facilitating implementation so I guess the first gap I would just like to point out is that we are not done with CPIC guidelines. There's still some that need to be written and what we've learned is that they take constant care and feeding so even if we reach some steady state of 20 to 30 guidelines we don't see this as something that's just going to end. At least in the foreseeable future is there's more and more genetic variance being discovered. I think that we're going to be in a period of 5, 10, 15 years where there's a lot of evidence just relating to genomic variation and phenotypic variation that we're going to need to keep tracking and locally in our own implementation program we're behind in implementing even those CPIC genes that already have guidelines because of the work that's involved in actually putting this together so I still think that building infrastructure tables that can be imported into sophisticated EHRs clinical decision support that gives real prescribing information based on genetic variation I still think we're in a period of a few years where we have to work this out in dedicated tertiary care settings I'm all for moving this into the community and community pharmacies as soon as we can but I think it's going to be very difficult to do that well until we've proven that we can do that well in tertiary care settings. One thing I heard from today is that there are probably opportunities to aggregate outliers with very rare types such as we heard about this morning the Stevens-Johnson syndrome type reactions that are just incredibly rare now with the web there's ways that we should be able to share information about that and maybe this community should do better at creating a single system for putting together these rare outlying patients rather than having a hundred different databases to share information about them and generate evidence about what might be behind their unusual outlying response reactions. I also think that we should come up with metrics of benefit that can be accepted that don't necessarily involve cost because we've all got plenty of examples where we think that patients really benefit and as we heard about from Dan's talk it's going to be very difficult to demonstrate cost effectiveness when it's by definition a small percentage of the population that's going to benefit for most of these things so research on establishing other metrics of benefit would be good I still think we need to come up with standardized terms for test results to be used in the EHR so we can share information and that that will contribute to making these tests that are lifelong tests that can be used by patients long term so I also think we should look into methods for providing genetic information on a per patient basis where they have it in some kind of a chip or a card or a smart phone symbol or something so that over the next 10, 20, 30 years patients can move from pharmacy to pharmacy and from provider to provider with an assessment of their genetics that could be read for the foreseeable future so I think there's people already doing that that we can copy like our friends in Europe and apparently their cards are made here so I guess those are a few of the gaps that I see that we could address okay thanks that gives us a wide range of things to talk about so maybe to structure what I'd like to do is put it into a few categories and then maybe we can open the floor for discussion in those categories so the first category that several of you referred to that I think it would be good to tackle is the category of evidence gaps so I think we've heard about a few examples of evidence gaps we've heard about maybe problems in study design so maybe we need to think about study design I think we heard about what the endpoints are are they outcomes are they a lowered serum measurement are there and then sort of a third kind of gap that a couple of you alluded to but we haven't talked explicitly about which I think would be very interesting to talk about is how many CPIC guidelines do we need ultimately and how many other genes are there out there that we don't know about that we need to be talking about so let's just open the floor up for discussion about how do we make sure we get to some resolution on evidence gaps in terms of whether people actually believe the data is significant and then what's the scale of the problem how many things are there out there that we don't know about I'm going to sound like Dick Cheney pretty soon so we'd like to start with that evidence gaps how do we fix them I think that we have to pick the low hanging fruit we've sort of done that I think the work that has been done at St. Jude with the Cyperines starting a couple decades ago is one example of that low hanging fruit I think Clopidigral was low hanging fruit Warfarin was low hanging fruit it just had some challenges I think because of study design so I think we have to think about do we continue to sort of focus on this one drug at a time approach or one class of drugs at a time approach or think about a model like is being done in Europe which has sort of the US would probably be the CPIC level drugs and sort of tackle all of those because people aren't one drug in their lifetime kind of people and then if you do that though then the important outcome may differ so how do you define what the outcome is what is the important outcome if you're looking at a PPI the important outcome is different than if you're looking at Clopidigral or Warfarin and so figuring out how you would sort of collect all of those together and the group in Europe has done that I think those have looked closely at that may not think about it in exactly the same way but I think they have given us a roadmap of where you might start and I mean I think my sense is there's sort of universal agreement that a gene or one drug and the genes that go with it at a time isn't the logical approach so I think the challenge is how do we sort of cover the evidence gap that builds evidence for a lot of different drugs but does it with a preemptive panel and sort of does a collective approach Mary I guess I'm distinguishing between discovery, research and implementation and I think that one drug at a time and however many genes it takes is the way to go when we're trying to generate evidence when we take real life populations that are having all their messiness and all their concurrent drugs and their altering disease status that generates more noise and the increased sample size that you get will allow you to make discoveries even for thiopurines in the setting of childhood ALL one can't find an association between TPMT genotypes and myelosuppression depending on the therapy so the more extra drugs you put in the more real life confounders that you add in studies like the UPGX study the less likely you are to discover gene phenotype associations when we come to implementation yeah I'm all for test every possible gene and implement for every possible drug but if you're still not convinced that there's an association between CYP2D6 status and tricyclic pharmacokinetics you're not going to discover it by studying 8,000 patients with all their messiness I think we're probably talking about two different kinds of evidence so the evidence gap that I'm really talking about not to dismiss discovery in the context of this meeting I'm really talking about the evidence that genotype guided approach is beneficial clinically it won't work when there's that much noise one of the things is it if it's pharmacokinetic based drug gene interaction that would be just blood levels that we need and you need very very small sample size to determine whether or not the blood levels are altered and that's what the case that happened in renal disease is that they use a very small number of patients with renal failure and they measure the blood levels and they determine yes the blood levels are higher so therefore they need a lower dose same thing with drug-drug interactions they just monitor the blood levels very small sample size as opposed to outcomes right but a lot of clinicians don't gravitate to drug levels I would propose a way forward that reconciles those two issues because they're both important Mary is absolutely correct that in terms of trying to develop the evidence it needs to be developed in a relatively constrained way but the other point that's relevant to evidence is that not all elephant elephant evidence gaps are elephant in size no elephant evidence gaps are the same and that's where I think the role of modeling is really important we've talked a little bit about economics here and one of the tools of economics is to do modeling and the advantage of modeling first of all is that you don't need any data so that's useful but the point is that by using assumptions around different decision points in the model you can decide we can vary this node a tremendous amount and it makes no difference at the end of the day so we shouldn't be expending resources to try and get that piece of evidence whereas this piece of evidence off by 10% there the model goes from being cost effective and so we really need to focus there so focusing limited resources and closing the correct evidence gaps is really important and then the other approach from modeling that I think can reconcile the two issues of single versus the multiple is that you can use a threshold approach by which you can take situations where we have good evidence of the cost effectiveness like the clopidical example or the back of your HLA B and a few others and you can say well if we add these up and given a test cost for a given panel at which point do we cross the threshold where essentially every other test we do is not going to you know somehow flip us back into not being cost effective we'll make it more cost effective except in the fact that as Howard pointed out we still have to deal with the issue of you know the unintended consequences generated in cost downstream but I agree with him that I think that that's relatively minimal but then you get to the point of saying hey you know if we have these three genes and these variants that cover these CPIC variants at that point for any patient we essentially achieve a threshold an acceptable threshold of cost effectiveness therefore other things we add on there we have to go back to the same argument every single time Other thoughts about evidence gap Howard I think one another elephant that was sorry one thing that John highlighted earlier was around Category C or whatever they're called we haven't put as much emphasis on the drugs who don't need or unlikely to benefit from the pharmacogenomics the things where the data is weak that was brought up I think shaping that would really be useful because one it would highlight some of the drugs that have never been tested if you take the 200 most prescribed drugs not the money on the money list but the most prescribed volume wise there's still almost half of them that you can't find a pharmacogenetic study even a candidate SNP study much less than what you would have seen or GWAS and so it would highlight the areas where there is a drug that has variability toxicity or whatever that needs some work it also would put some to rest I think having some nose allows people to take the yeses more seriously and right now we've been for very practical and right reasons focused on the yeses I don't think it's going to be hard to get a lot of people excited about being on the no committee for CPIC going and buzzing through that and getting some of those done will allow us to then have those types of discussions a little more clearly because I think there are gaps on the drugs that are used where it was no one's pet drug it's a generic drug maybe there's no there's a term that you used to be able to use that I'm told you I can't use anymore there's no sugar daddy for those for those drugs I know it has a different meaning now in the old use of the term there's no one favoring it trying to help it go through so there's an opportunity there to really define that for the field that I think we can make I like to know it really does help people think that you're not just a true believer if you can actually say notice some things so I think that's actually an important point Jeff so one of the lessons from the Ignite that we've discussed is that it's probably insufficient for a single site to be able to carry out a pharmacogenetic study with sufficient numbers to generate the evidence required and I think at least we've discussed hypothetically if we had it to do over again it might be a series of a network that focuses on maybe one study or a series of studies but not each of the five or six sites doing something on their own which is what I was trying to get at before the collective expertise of all the different pharmacogenetic implementers in the room and others that are not in the room could be harnessed to generate the kinds of evidence that are necessary that's what you know that's what is done with drugs right you know they're large multi-center multinational clinical trials because you need the numbers you need the numbers to generate sufficient evidence so that's one thing that should be seriously considered I think and the second is again barring from the playbook of the pharmaceutical industry registries you know registries are really huge repositories of clinical outcome data as well as economic data that are often used to convince payers to to make the right coverage decisions so you know we have again we have the basis for large pharmacogenetic registries given the centers in the room how do we pull somebody said it earlier you know how do you pull all this data together in a common repository that makes it usable and longitudinal and just to build on what Jeff is saying I think to make either of those models work a multi-center clinical trial building on the Ignite Network for example or to think of a registry capitalizing the idea of common outcome measures and deciding what that would be would be really critically important and I keep hearing these two uses of the term evidence when I think of implementation research it's really to promote adoption of an evidence-based intervention so if we as a community would have to decide is what we're trying to test evidence-based yet and there's a lot of variation some of the things we're still talking about discovery for certain drug gene pairs and other things we have CIPIC guidelines for so clearly there's a whole continuum in terms of the degree of evidence so if we're on the end where there's a CIPIC guideline and I think that's really the focus of today's meeting I've heard the most for is really this idea of preemptive testing for a pharmacogenomics panel and so that seems to be where there's the most energy and agreement that that's where the proof of concept needs to go for this type of trial and so if you would marry that with common outcome measures it seems to me like that would be sort of taking it to the next level that's sort of what I've heard so I think the question of evidence gap is an interesting one and in some ways it seems to me that in some regards the problem is the enemy is us that often is the case some of you know I've told the story before at Northwestern when we tried to implement the emerge PGX project initially my Pollyanna approach was oh we'll implement all the CIPIC guidelines and for everyone that we because we were sequencing all the genes for everyone that there was we'll provide that evidence we'll provide that information back to the participants we were stymied in that because we couldn't persuade the physicians that we were working with and so the idea of lack of clinician buy-in I think fits into this whole question of the evidence gap they just weren't persuaded that the data is there so we need to think about designs that actually give us robust evidence that will be persuasive we're going to get 100% people to agree because that's just the way the world is but are there things that we need to be doing to deal with this problem of lack of clinician buy-in we'll just better studies with better data do that or do we need to be doing something fundamentally different I mean so I think that's I mean what you just said maybe captured my concern about or my point about the evidence gap I mean so within the CIPIC guidelines even in the current guidelines there's really two groups I mean there are groups where there's very strong data that you know few people would argue that there are different outcomes I mean I think there's others where there's good evidence there's good associations between a phenotype and a genotype and yet you know a high percentage of clinicians are not going to be comfortable based on the evidence available I'll take voriconazole for example we just published a paper on voriconazole genotype and drug concentration relationship that showed that in patients with one or two copies of the star 17 allele they're significantly more likely to have a subtherapeutic trough 50% of the population that we studied in this sort of prospective but not an active genotype intervention study died and yet our clinicians are you know they want to do under a research protocol and implementation and so I think that you just you know you have to sort of meet clinicians where they are at some level now hopefully you know if we're 10 years down the road and we have 20 examples then it's enough that you have the drug concentration and the genotype relationship but we're not there and so I think we have to your experiences I think exactly the challenge that despite the fact that there's a group of people who are comfortable with the CPIC recommendations which I think are great it's just that there's not you know the vast majority of clinicians aren't comfortable that the level of evidence is such that they want to actively do something different based on genotype unless it's under a research protocol so it's sort of true clinical intervention they're just not ready for so I'm not surprised you had that experience because I think for a lot of the things like the CPIC let alone some of the things that are being recommended with commercial firms there's this sense of an evidence gap and we just need a little just give us a little more to push us over the edge I think is the general sentiment that we see with clinicians so Julie that's what I was trying to say at the beginning first of all the way we got our 19 gene pair drug gene pair rules implemented because that was already constituted to have subspecialty expertise which could be consulted with regard to what drugs go in the formulary and that made it possible for John Black who was as you know runs our pharmacogenomics lab but he's also a board certified psychiatrist to deal with those groups and he said the major skill set that made it possible to get those drug gene pair rules implemented was his background in psychiatry I don't know what he was trying to say but the point I was making earlier now that we have across the board a good deal of information about thousands of our patients what we've done is gone back and ask the physicians themselves what they want to do to make themselves feel comfortable some of them are approaching this as if it was a you know randomized trial but others have taken somewhat different approaches and we empower them now did I think at the beginning that this would be as useful as it has been frankly not but as it turns out it's extremely useful because with all due respect if I go to the chair of our division of cardiovascular disease which is bigger than many departments of medicine and I'm just a general internist and I suggest something it will I'll get a polite response but that's all if it's a young faculty member in his division and fellows from his division then he begins to take ownership and all I'm saying is that this sounds so simple and straightforward as a matter of fact it's turned out to be really important in terms of getting their enthusiasm going and the anesthesiologists for example who have phenotypes that we in medicine frankly would salivate for because they keep track of everything and they're now seeing oh my goodness I can use this so that I think all of us in our own micro environments every environment is different you've seen one academic medical center you've seen one academic medical center I think we need to be thinking about how can we move forward to learn from our colleagues and I frankly have found that it's been much more useful than I would have thought initially and we all have the same problem when you were just talking about this morning about design and talked about having to do designs where we were really doing a genotype based design is there a sense that if we went to more studies like that that we would get better buy in from the clinician colleagues I mean within our Ignite projects I think that the design that we've had a fair bit of success with is sort of a pragmatic design where patients either to patient level and clinic level we've done both are randomized to a genotype guided versus not and we give them the genotype at the end and so I think it accomplishes several things one it allows the physician to develop a comfort level with having genotype and making clinical decisions and we haven't wrapped up actually we're closing our pain study hopefully in the next couple weeks so we'll see how they turn out but in terms of sort of getting physician buy in and what we have found and I think I completely agree with Dick one of our biggest lessons learned within Ignite is find the enthusiastic clinicians and so for example we at the very beginning of our program went to psychiatry and asked if they would be interested in doing CYP-2C19 2D6 guided antidepressant therapy they're like nah that's okay don't think so so we said okay we're plenty busy well they have come back to us we've now I think enrolled our last patient in a pragmatic trial in pediatric psych and so there is this element of just sort of momentum and focusing on those where you have a local advocate and you know if you will embedded in that clinical group because it's so much more effective so I mean I don't know what the right answer I think these you know this attempt at the perfect randomized controlled you know tightly regimented design is not the answer that's what co-ag was and I don't know that it helped us understand anything that's true to real life so thinking about you know can you do pragmatic designs that still allow you to do data collection and answer real questions but it also feels more natural and you can flow then easily into a clinical implementation Mark? I wanted to come back to a point that I think it was Julie brought up earlier because I think this is a critical piece if we think about the evidence needed to implement something an evidence based guideline say I think it's significantly different if we look at something that's a patient safety issue and when we look at some of the significance of some of these adverse events or the implications of giving clopidogrel to somebody that's a poor metabolizer and knowing what the data shows you know could patient safety approaches for those that really meet that threshold get us away from this evidence you know there's never enough evidence to implement something and I'd be interested if there are folks that have taken that approach and whether that you know changes the tenor of the conversation within the organization because I think we frankly have a pretty good case in many areas to make in the safety realm Howard? So Mark I think you might already know this but the name of our intervention panel is the therapy risk mitigation panel doesn't mention the word genome or anything to do with DNA doesn't have analytics doesn't have big data but it has it's therapy, risk, mitigation and happens to be a panel so that way people can focus on what we're trying to achieve as opposed to getting lost in a double helix and I think it's really important because patient safety moves the needle other stuff is for tomorrow and I can actually say that it was framing the personalized medicine program at mission in the context of patient safety and patient centered care that got the board of directors to approve having a program to begin with because you know although they asked you know about cost and savings it was the safety issue so although that is the context within which we've developed this program still within that context there are various responses to quality we even have our quality officer be our physician champion but that's not enough for our P&T committee there has to be several clinicians from each of the service lines to then give testimonial on why this is clinically useful so you know yes so it works but not all the time in my experience I mean I agree it has a short term panache but you know it depends on how serious the disease that you're treating is and what the alternative therapy is available for making that decision based on pharmacogenetics so at the beginning we got away with decreasing thiopurine doses in ALL based on TPMT status to prevent life threatening toxicity but very quickly people said well how do you know you're not compromising the antilukemic effect and without those data we never would have got by and I just had a conversation at the back of the room this morning with somebody about UGT-1A1 and Irenote can which clearly predisposes to toxicity from Irenote can but it's not been widely adopted from the cancer community partly because those patients have a chance of cure from that drug and everyone's afraid that reducing the drug to reduce toxicity will be a problem even if it's based on genotype and let's not forget that the FDA has just done the best way to avoid toxicity is to not use the drug and so that's been the response of the FDA for codeine for all children less than 12 years of age so I don't think we can only base our studies on preventing toxicity Steven? Yeah and I certainly wasn't meaning that this is a panacea but I think for certain select cases and again I think the advantage that most of you and most of us that have done this is that once you get a successor to under your belt where you can really at your institution implement it and show that it really does make a difference and I think and also do it in a way that doesn't totally disrupt the physician's workflow a very important consideration in fact maybe even makes your life better if we're really intelligent about it then I think the energy barrier to overcome for subsequent ones where you can use a different approach may be slightly lower and particularly when you begin to potentially you know publish outcomes at the practitioner level which many of us do across many different quality indicators and notice that those people that use it have different outcomes of those that don't I think that is also something is that peer comparison that also can make a difference Steve? Well I think my comments now are maybe not so quite so relevant to where the discussion is gone but because we're finding that we need to generate some of the knowledge if we're going to implement this into into pediatrics we've taken a little bit of a we're taking a little bit of a different approach to it and that is rather than starting with the drug or the gene we're trying to get our folks to think about what is the outcome that is expected or desired from the therapeutic intervention and then working back and saying well what is the exposure or the concentration of the active form of what's being administered that needs to be present to have a high probability of achieving that response and then what dose needs to be administered to the individual patient to achieve that response and if you work through that process you start to find out what genes may or may not be important and then those are the ones that should be paired with the drug that's going to be administered to the outcome of interest the challenge with that is that if you're going to work backwards like that you have to have a means of giving everybody the right exposure or the same exposure you have to for the drugs that are subject to a polymorphically expressed clearance pathway there's going to be a lot of variability and exposure and so I think this is one of the barriers that we're going to have to deal with if we can get people to buy into that approach the underlying question now I'm taking away from my presentation tomorrow but the underlying question is a failure to respond to the medication of an inadequate exposure or is it a function of something that's inherently different at the level of the drug target because we don't go after the drug target very often it's hard to get at some of these things if your drug target is in the CNS for example so the depression, the autism, the ADHD the things that affect a lot of the patients that our clinicians that have drank the Kool-Aid at least the flavor of the Kool-Aid that we're serving up those are the things that they're interested in so one of the buckets that we've sort of gotten into a few times was the whole bucket of payer and reimbursement issues and obviously some of that's going to be based on evidence gaps but John I'm going to look to you so one of the things that we've talked a lot about today is the relative value of economics versus outcomes especially outcomes that you know may save life how do you at Optum and how do payers in general putting you on a spot as a representative for them how do you think about that so thank you for that I mean the first thing they do is obviously look at the evidence base they look at the clinical guidelines they look at what the societies have produced they run the literature reviews and that will go through a assessment committee of some sorts and they make a decision and there's a variety of different criteria that are put into that mix I think one of the things that's fascinating about this space and I think actually warrants further discussion with members of this group that's different about this space is the persistency of the data and that means I think we I think there is time I think it would be time worth spent thinking about how is this field different to a regular lab test how do we think about this differently to a cholesterol test I'm not saying I've got all the answers but what I would say is I think there's an appetite to listen so if folks are interested around how should we think about the economics of this differently I think that would be a discussion that would be an appetite to have so put simply I think we want to listen to you folks so I think we've it's been mentioned a couple of times there's a lot of concern I think amongst people in the field that we all are subject to some level of genetic exceptionalism because it is maybe no it is no different from a cholesterol level especially if you have a cholesterol levels in a family right so you actually so I think it is different my my purview is that there is I mean Mark said it earlier and I concur with them there is value this data is persistent you all know that much better than I do that has implications for how you think about the space so I think the lifetime value of some of these tests I think that's a really interesting construct which I don't think often gets described so I'm interested in pushing that space forward can you just help how does that jive with the fact that we have so much impermanence in our insurance for any one individual switches from insurer to insurer to insurer so what's in it for the primary insurer the current insurer in paying for a lifelong test because they won't have that patient under their insurance policy the 20 or 34 I mean what do insurance people say about that use your microphone please I'm going to answer a different question that is not an attempt to avoid the question but it does an attempt to give a different perspective on I think where we're trying today what fascinates me about the space is the cost of some of these testings falling dramatically I think everyone would agree the cost of some of these testings falling dramatically so these notions of what if it's a $250 test I mean that's more, that's cheaper than an MRI isn't it quite a bit so I think what's interesting is is the lifelong value discussion is that a little bit of a red herring given the costs of this thing have fallen so much so there is persistency there is value in that sense because you don't want to get the patient retested but there is value in the fact that the costs alone have come down substantively I think that is an interesting so is that not well appreciated in your sector which bit the low price I think that again it comes back down and I hate to beat a dead horse for folks that you probably hear me say this again it comes down to the coding we see such variation in what test is being done for it is difficult to ascertain precisely for this test this cost has dropped because often the codes get the tests get bundled together in CPT code and is difficult to tease that out so again to beat the dead horse a little bit more it comes back down to the coding if we can get the accurate coding then we can answer some of these questions more appropriately so when you talk about persistence understanding how you're thinking about persistence so most of us believe that at least for germline related issues cancer, somatic cancer issues are a different story but for germline issues by persistence do you mean the fact that once you've done the test once you know what the answer is whereas for cholesterol maybe you need to do it multiple times so that should reduce the threshold for you to accept evidence then so again I'll say two things this this whole notion right now of persistence this almost if you think about it as an asset you capitalize the asset and depreciate it that's the way you think about it I think there is an entire thread of discussion around how do you actually measure and manage the health economics in this space different to other domains do you need to look at this space with net new eyes I think that is the discussion that folks inside our organization will be willing to have I'm not saying they're going to agree with you all but I will say that we're interested in having dialogue so to be humble the audience around this table around this room is probably the experts in the United States from our perspective we are dealing with multiple different issues but I think what we'd like to do is engage dialogue and start thinking through how would some of this actually work so I'm not saying we've got all the answers what I am saying is we're happy to listen so I think the persistence piece of it just to dial in on that a little bit more suggests that actually you know this idea that some of us have actually championed that do a genome sequence at birth and have that sequence do it once and if it costs pick a number a thousand five hundred two thousand dollars and you've prevented some percentage of people from getting a three thousand dollar BRCA test obviously the economics of that would change over time but you know so that's an argument for that persistence actually go straight to the juggler let me push you on that one the other kind of function of this and again I'll beat the horse a little bit more coding is one facet of it quality is the other facet of it how do you know how do we know how does an insurer know a payout whether it's united or another payer that what they're paying for is actually good quality right I'll say it again how do you know that what you're paying for is good quality and I would urge you all to in the evidence generation spaces you think about this to think about what is the evidence what is the quality evidence here right Mark so I think there's a piece that's not been articulated here that's really important when it comes to the insurer which is it's a different issue if we do a test for an indication at a low cost because if you say well this person's coming in for a stent so we're going to do we need to have the SIP-2C19 status to see if we use Kilputigral or not we could do that on a panel at a cost of $250 which for most payers would not even trigger their review because it's under the under a threshold and so unless there's a you know some sort of an exclusion genetic testing in general you might not even you know need to go through power authorization or anything of that nature you then have the information from all of the rest of that that you essentially can use for free from the payer perspective because you've got the information you don't have to ask them are would you be willing to pay for this this this would be a different issue than what we're proposing which is to do preemptive testing absent an indication and then expecting a payer to pay for that because that is not the payer model payers pay on medical necessity and we don't have a medical necessity argument for preemptive testing which is why you know many of us are looking to other mechanisms to generate data that we can then use in that setting until unless we can prove it's cheaper well but no actually not you would be for Medicare for example it doesn't matter how cheap it is because the legislation says we only pay for medically necessary tests and almost all policies say we only pay for medically necessary tests and there's never a situation where a non indication based test would be medically necessary it's anathema so that's a fundamental issue that it's not a matter of the economics that answers the question about why would a payer pay for a test without an indication what about a medical checkup they cover a lot of medical so there are preventive services which Medicare doesn't cover by the way that's an exclusion any preventive services covered by Medicare was either legislatively approved or through the ACA has a meets the USPSDF task force recommendations in the HMO model where you have a wellness benefit but it's defined in the benefit so there could be a situation you could imagine a payer saying we are convinced enough that we are going to include preemptive pharmacogenomic testing as part of our benefit package that you pay for and so that would be a potential model for coverage but approaching them you know outside of that type of a discussion it would not really be fruitful so Mark could we push that just a little bit further because the fact of the matter is as a result of work by a lot of people in this room what 30 years ago before half the people in this room were born some of us TPMT looked really exciting and it was totally new and it wasn't anything that we had much precedent for now we have a whole panel where for individual drugs the evidence is pretty strong so what we're really talking about I find myself and it won't surprise those of you know me saying that really pharmacogenomics is clinical genomics for every patient everywhere and I sincerely believe that it wasn't 30 years ago it is today what's the difference between this since all of us I've heard rumors are going to get older with the exception all of us are going to get older and as we get older we're going to be exposed to some of these drugs not all of them but to some of them and that's why I say it's for everybody everywhere this is beginning to approach the situation that we see with vaccination I heard what you said about prevention but as a matter of fact this is a little unusual even for genomic medicine and it may be a societal good to have this information your electronic health record is one at mine I am beginning to approach those ages so you know just because Medicare doesn't cover prevention societies decided prevention is pretty important and this is a little different than the way we usually talk about this and I see John over here fainting get away but as a matter of fact I think that there is some comparability with vaccination now it wasn't 30 years ago but it is now now that's the strangest thing anyone said in this room today but that's alright that's why you have me that's why I invite you again I think you can so I think I mean the point that you're making is a perfectly reasonable one but the the case that you need to make for society to say everybody's in like we've done for vaccination I'm going to say I'm going to do a panel because I have an indication to do a pharmacone I know that I just made I just said what I thought right right so but we're seeing in the payer world particularly in cancer genetics that they're comfortable moving from the model of BRCA 1 and 2 testing to say you know what we know that the phenotype is not as clear cut as it is and I work with cancer too and I've seen that that much so that's a model that actually if you want to talk about something that could be implemented in the short term is probably very pragmatic go ahead I just when we're talking about the life-long the persistence of these things that we make sure that we're careful not to promise that once we test do the 2D6 genotyping test we will never need to do that again because with the exception of if we were 100% coverage on a whole genome sequence which to my knowledge there aren't anybody that can do that yet that the variants we're testing now you know if 2 years from now there's another one that comes that's a 20kb away that's the middle of some enhancer thing that's really the thing that really could contribute as much to some of these the genetic variation as any of the other ones except the knockout ones we just got to be careful that we're not promising that we will never do this again as the technology is advancing even you know pgrnc that doesn't cover the main gene but there's still other places and maybe the algorithms for aligning it are a little bit different or something so that would be the only okay well then maybe that we can move to another one of the sort of buckets that I think we've heard about today and I'll lump sort of two buckets together the one bucket was the idea of data infrastructure do we have the right data infrastructure to support robust pharmacogenomics going forward and I think there are a couple of issues one we heard a lot of discussion about nomenclature we don't even have a good nomenclature for pharmacogenomics and for phenotypes and how do we go about fixing that and then a related sort of data infrastructure issue which Heidi was passionate about is how do we make what can we do and I think also relates to the quality issue that you raised can't we just agree that everybody has to sort of deposit their data somewhere as a consequence of ability to get paid and use that as a way to measure quality and begin to assess quality I think there's nothing magical about you know most of these genomic tests there are quality measures that are built into them that one can provide so there's the issue of data infrastructure I want to hear people's thoughts about that but obviously related to data infrastructure is how do you store and present that data you know and Sandy talked about this a little bit earlier in terms of the IT infrastructure and again I think the IT infrastructure is important in terms of a how do you provide downstream to non-experts the ability to use that data in an effective way and then there's a second issue of if we do a lot of this preemptively and make sure the data is stored and broadly available to everybody this is Rodin's idea of you carried around on a chip in your wallet maybe that's not the best way maybe some place in the cloud is better but so comments about sort of those two issues the idea of IT and obviously your idea of persistence also relates to the value of data infrastructure and IT infrastructure so John that term data infrastructure I think you've crystallized that really nicely and it's kind of fascinating to me to see this space mature and I would again I'll beat the horse a little bit more I would strongly suggest that this is one of the biggest limiting factors to this approach getting adopted more widespread because if we can't understand what we are talking about then we will never make decisions around reimbursement or quality or appropriateness or choose your favorite use case so I would strongly urge the powers to be to focus on this area because I think it will be an accelerant in an appropriate way to ensure appropriateness of utilization I'm working your feedback as well I think it's important to think through this space at different levels I think there's a a piece around this are we all talking the same thing I don't know if we are I think there's confusion it's a bit like British English and American English I think we're talking past each other sometimes so I think there's an amygdala piece and then I think there's a real practicalities piece how do we move the data around what do we need to do to analyze the data is the bioinformatics pipeline stable do we know what version control looks like there's two threads of dialogue that would be my immediate reaction to that but I would say in closing I would be very happy to work with folks as part of that journey we would be highly engaged in that process if other people around this table would be so interested Mark so the nomenclature piece I think is close to being a resolved issue and I think the pieces that are needed to bring that to closure are in place and so I don't see that as being a gap that needs additional effort to fill I think the work that CPIC is doing in terms of trying to bring phenotypic terms together and that we're pretty close there are a few edge cases that need to be solved but I see the end of that pretty quickly the others are more problematic but I think one of the rate limiting steps for representation of the data in a structured form in the electronic health record environment is that barrier is being lowered because there are more and more end user demands to say we need to have this data in there and there's more and more experience at the vendor level where their customers are saying we need to have the ability to have access to this data in a structured format and we need to have sufficient support to support these types of things so I see that as moving along fairly substantially although there's still much more to do I think the biggest gap that I would see is the portability of the data with the patient and that I think is a fascinating area that would be very amenable to some research Sandy If I could make two points and the first one is a little out there so I'm going to put that first but I want to be more concrete so within this meeting we've talked about clinical decision support of the form where you have a defined rule that you implement it executes it generates an alert it doesn't generate an alert type of clinical decision support there has been a dream that's been out there for at least 10 years that to best of my knowledge has not been realized yet but with deep learning, machine learning techniques there's increasing hope of it that be able to generate another kind of clinical decision support that would work by generating indexes of similarity to from your patient to other patients that have been previously seen in the electronic health record select out patients that are similar to the patient that you're seeing and then use that to generate statistics about what you are likely to see in this patient which if that was done there's a whole bunch of challenges by the way in pharmacogenomics one of the interesting things that it makes easy relative to that is one of the biggest challenges there is lining up past patients in terms of time for determining what point in their life is equivalent to the point in your patient's life now whereas in pharmacogenomics you're going to prescribe a drug so that piece gets easier but it's challenging but if this was done then that starts to address I think some of the evidence things that we're looking at etc. something that I wanted to throw out there by the way Amazon does that every day so in general within health IT the fraction of what we do in the health space relative to what everybody else does everywhere is incredible the so something just a little more specific so on the coding issue one of the things that I do wonder one of the things that I think still as Mark points out we're making good progress on this but something that's still out there is multiple things that are still out there but tests and really defining the tests and what I'm wondering I just don't know the answer to this in terms of what insurance companies provide what you're saying you need from an insurance perspective is pretty much the exact same thing that these rule-based clinical decision support algorithms need they need a very well defined you know what test was run as well as for the clinical decision support what the answer was do you get Loint codes or SNOMED codes when you receive claims that's just hold on a second there so inside Optum when we pull the data out of the EMR by definition we will get the Loint code and the SNOMED code where we're doing the chart extract now I'm wearing the Optum app if I look at a an 837 or a claims form if you will the criteria the metrics that are typically on those are CPT codes and IC9 IC10 codes these days there are additional fields that you can leverage inside those claims forms for additional information and I think that's what's interesting about this domain we may not I think SNOMED is limited in its own right I think there's too many different reference labs doing too many different tests that make it amenable to a SNOMED solution but your challenge the challenge that you're facing in clinical decision support in the EMR is a very very similar use case to the challenge a payer is making about whether to pay this claim or not so the base predicates of what's the data element how do I build a declarative logic off the back of it to drive some kind of decision you can take that path and you can apply it to a payer hence my belief is we've got to get some kind of coding going or it's a reference piece that facilitates this both on the payer side of the house and also on the provider side of the house and again my lens on this isn't just UHC being the payer it's also optimum with our own physician groups that make sense yeah and it feels like the line codes in the CPT codes that they should be aligned well I mean let me be clear my personal view and I really would welcome further dialogue in this is that CPT codes are not sufficient and do not meet the use case need in this space for testing I personally believe that it's not sufficient because the rate of change in this field is so fast and the time to get a CPT code so so long there's always going to be a mismatch there needs to be a much I think you folks call it GTGTR is that what you all the 5000 you have a database 5000 order will test we know the 65,000 tests in the market so instantly there's a mismatch there right so this issue is in my opinion is something that's inhibiting decision support in the AMR what pay is painful how do we measure quality to write down some comments earlier around consumers and how do we know what's the right test for a consumer it's got manifestations everywhere again I welcome discussion on that Marilyn did you my comment was actually on your first point the way out there one so I have come across two pieces of commercial software that will do that theoretically they advertise they're not advertised one is advertised for healthcare the other one is advertised just for big data the one is a half a million dollars per user ID it's a cloud based system the other one is a half a million dollars for two user IDs on a cloud based system so I think that realization is coming I know of about five or six research labs in the US that are trying to develop those algorithms so I have a Pennsylvania State Department of Health grant where one of the aims is to work on that trying to come up with a strategy that's not half a million dollars you know a lot of health systems can't afford such a thing certainly for one person to run it so I think we are moving towards that realization just not as fast as I think we would all like okay so I think we're running the end of our time I hate to say this but Mark you might have the last word oh god and I'm going to dive into code so this is really horrific so everybody's going to be definitely ready for a drink after this I just wanted to make sure that we weren't going in a direction because what I heard in your comments Sandy was that there is one solution one code to rule them all and I really don't think that that's going to be the case I think that the LOINC code while very specific if you can imagine I have to fire a decision support rule for CYP2C19 I could get it from a panel I could get it from an exome but the LOINC code that I would need to run the CDS is specific but it doesn't help the payer to know where that was derived from and so there's not a one-to-one match between those things what we're really talking about here is a system that as yet doesn't exist and the problem that you reference between GTR and what you know is that GTR is completely volunteer and unless you have something that ties to the payment then it becomes very problematic and as you well know that the additional fields were used for a period of time when CPT went down the really bizarre direction let's just do it on procedures which was extremely helpful for everybody was that they said well you could enter I think it was a G code or something that would give a name of a test like BRCA so you at least knew what was happening but again it was voluntary and it wasn't used and so the question is how this is a really diverse set of stakeholders and how to get them in a room together and to agree that this is the problem and that we need a solution that actually would work in all the different scenarios I think is an extremely challenging one but is a very high priority so we'd be sorry I'd be very happy to host that meeting so if we can get the right people in the room I'd be happy to host that we've got to figure that one out because it's slowing down this field Yeah I mean earlier you talked about the powers that be the problem is the powers that be this is so far from what they think about on a daily basis that I think we're suffering from that so I think this was I was a little worried about how robust the discussion was going to be because we had already had a lot of discussion today but this was fantastic thank you everybody for participating it was just sort of a quick little survey I'm curious about so I want to show off hands how many people today here think we're ready for clinical implementation of pharmacogenomics today wow almost everybody how many people don't alright well on that happy note I'd like to thank everybody on the panel and everybody in the audience for all of their participation so