 I know it's been a very intense day, particularly yesterday was a long day and I know many of you continue to work at dinner and other places, some did not, but many did, so yeah, that's good. Thank you. Great. Thanks. And so what we often try to do is kind of summarize some lessons learned. Because our time is short, I'm tempted to kind of run through these quickly, but then come back to them if we have time for discussion. We will put all of these up on the website from what I understand from Wukul, they'll be up really soon. Is that right, Wukul? Yeah. Like in the next couple of days or so, really? For the slides. So the videos, they have to do some fair amount actually of processing that they do in an incredibly short time. So we'll have these available. I should also mention, and this is kind of in our next steps section, that we do typically try to produce a paper for publication from these conferences. Usually the participants in those are the people who are the moderators and the presenters of the various discussions during the day. Of course, that presupposes that they do participate in the development of the manuscript. And so you do have to be active in your back, that's great. You do have to be sort of at least respond that you received it and that you're okay with it. But we'll go into that in a little more detail in a moment. So lessons learned, I think our number one lesson and our closing lesson was that behind every statistic is a patient like Angela. When we were talking about cases that could be reported, I think Josh Denney and Dan Rodin have that outstanding case on the Clopidogrel, the patient who had nine stents and restenoses in that before somebody actually did genotyping on her and found that she was a homozygote unable to metabolize Clopidogrel. Implementing pharmacogenomics seems like we'll likely only benefit a small proportion of the population for each drug. Recognizing that, 95% of us have at least one pharmacogenetic variant. Probably all of us have at least one. And there's a number of folks who have many, many. And they are probably the ones who end up with a lot more adverse drug reactions. And I give the TPMT, one of the TPMT examples here. And the question that I believe Marilyn asked, how do we get clinicians to focus on outliers? We do practice medicine based on averages. On the other hand, we all like to think as clinicians that we do individualize and personalize what we do for patients. But the outliers are the ones that get us into trouble as well as the patients. The lesson from the Hong Kong experience of not adequately educating clinicians what to do when you advise them to do genotype testing prior to using a particular drug. So an appropriate education effort is definitely needed. And we heard from Heidi that quality can be vastly improved of the quality of the genotyping and interpretation by a requirement to submit to ClinBAR and undergo peer review. Because as now, as we understand it, being required by at least one of the major insurers. So before I go on to this, is there anything on here that gives anybody terrible heartburn that we should change? Yes? Not heartburn. But as I think about the idea of who will benefit from pharmacogenomics, I wonder if there isn't some value to knowing that you're a CYP2D6 star 1, star 1, for example. As we start to, you know, I don't know how many years down the road and we start to integrate drug-drug interactions with pharmacogenetics, you know, a drug-drug inclusion of addition of a CYP2D6 inhibitor like fluoxetine, for example, is not going to impact the kinetics in a poor metabolizer because there's nothing to inhibit. The impact will be greatest on those individuals who have the highest activity because they've got the furthest to fall. So you know, I don't know how you measure that in terms of its value. But I don't think we should be too cavalier to imply that there's no benefit to somebody who's not in the tails. We just may not recognize what that is just yet. The immediate benefit, for sure, is people in the outliers. But as we become, as we have access to more of the factors and information that make us unique as individuals, it's got to be of some value. No, that's an excellent point. You can all watch me try to struggle with a keyboard I'm not familiar with, but I'll get it here in a second. And as Josh knows, every time he gives his talk, I always stand up to the microphone and say, 100 percent of us have pharmacogenomic variants. Because as Steve pointed out, if you're wild type, that still is informing dose for specific medications. And so I think that's the message we should be telling. The impact is more nuanced. No, and I think you've made the point previously, Mark, that because we dose to the average and the average includes herziagots and homoziagots for loss of function alleles, that we may be underdosing the people who do not have the risk variant. Yeah, the Warford study has indicated that if you're wild type for all of it, that probably six milligrams is the appropriate starting dose rather than five. So I do have to just, I'll call on Jeff in just a second, but I do have to take you to task because I love taking you to task, and many of us do the same thing. The term wild type is a little bit hard for patients to kind of understand. They envisioned Howard last night having his party and that sort of thing. And look at each other and go, are you the wild type, or am I? Definitely a badge of honor, whatever it is. Yeah, exactly. And so perhaps we need to shift more to terminology like common allele or in the scientific literature ancestral allele, that's not always the ancestral allele, reference allele, something like that. So just a little bugaboo I thought I would put in your mind. Jeff Strick. As to the second, just the way the second bullet is phrased, it would seem like it could be phrased in a slightly more sort of positive, like, although for any given gene drug pair might affect only a small proportion of the population, kind of getting to Mark's point. Yeah. Well, and I mean, and it depends what your definition of small is, CYP2C19, which is one of the most common genes that we have actionable variants. It's 30% of the population. That's not small, probably. Okay. So how about a subgroup of the population? I think I would be inclined to phrase it as the greatest benefit will occur in a relative, a subgroup of the population. Fair enough. Remember, he's a pediatrician and a Canadian, so he's going to be a glass half full guy over here. Three quarters full. I am now officially a U.S. citizen, and this last election was my first one that I got to learn. Yeah, way to go. Good job. How'd that work out for you? But it is worth noting that, like, I remember when there was a story in one of the globe articles saying that a patient got 23 ME testing and had the warfarin gene, so he made sure his mother wasn't dosed with that drug, right? So the potential for complete misunderstanding and actually avoiding drugs in people who don't have the risk, alleles, just because of total misunderstanding I think is something that we also have to think about in, you know, implementation of genomics to say you don't have that allele, you're good to go can benefit the rest of the population, too, just something to think about. Yeah, excellent. Yes. While we're kind of nitpicking on words, can we pick on the word population? Because since that's my livelihood, we're talking mainly right now about a European population and we need to really think about the impact and the health disparity and inequity that face minority populations in even the ability to implement. So we wrote this one for you mentally, but also for all of us because we clearly are not addressing that, you know, people outside the European ancestry population on whom almost all the pharmacogenetic and pharmacodynamic studies have been done. So you're absolutely right. So since I see that we are drifting a bit into wordsmithing, what I would suggest is maybe we go past the lessons learned, we can come back to them if we have time because we really want to, you know, capture your energy on research opportunities. That's, you know, one of the major objectives of this meeting is what directions should we go in. So we heard the suggestion to analyze early on from REX to analyze genotype data from past trials. I might sort of pin Dan down a little and just ask, the CAST trial, the cardiac arrhythmia suppression trial was, you know, one of the major trials showing harm and withdrawal of drugs and the question has always been asked, is that because of some genetic variant and could that ever be tested in CAST? So it can't be tested because the DNA samples, trials we might are not around. A more interesting question is, could the result of CAST have been predicted? And I hate to waste half an hour for people's time explaining why that might or might not be true. But the answer is no, you couldn't do that in CAST right now. And the extent to which the genetics play a role in mediating that adverse drug effect is, I think there probably is a genetic story, it's going to be impossible to prove. Again, that's another half hour talk. But REX's point I think was, you know, is it possible for us to get access to some, REX I won't make your point for you, you can make it, in terms of ongoing trials or completed trials. This is something we might want to explore. We also heard about the value of patient-driven contribution of data and samples. So trying to encourage patients to contribute such data, again how one would do that in the infrastructure for receiving it, and all of that is challenging at best. But it did come out as something from this group, I think, that was something we should explore. We've heard about harnessing existing quality assurance or quality improvement projects for a couple of years now at this meeting. We haven't thought of a good way of doing it. And would welcome those of you who are doing QI projects as to whether there's a way to put a registry on top of that or some other kind of approach, and Dick, you know, maybe in your project or Steve, in yours. Is there any way for us to try and capture some of that experience that can be generalized to other centers? I mean, obviously you'll be publishing your results. Do you want to comment, Dick or Steve? Maybe not. Push your microphone. You can use your mic. I think what might come out of this, Terry, is that the more we know about what other people are doing, the more we can collaborate. I've seen that happen in the corridors here, and I assume that's part of what you're thinking about. Yes. So we've been approached already by several people who said, are there ways we can work together? And I would hope that there'll be more of that. Fabulous. So, Terry, just to be clear, for UF for the Clopidogrel project, that was a QI project. We did not have research informed consent, so it was all done under a QI mechanism. So I think that's a great example of where you can do a clinical implementation, under clinical consent, and with appropriate approvals then go in and follow up. Yes, and it's fabulous that those are happening. I think what we're asking is, is there a way to bring those together, to harness them? Is the word that we use. And if we can think of some clever ways of trying to convene those. There are quality for Dan, again, who has sort of pointed me towards some quality people at your place, and it hasn't kind of gotten anywhere. But anyway, so that was one that I think came up as something that we saw as a research opportunity. Looking at diverse approaches to it. Can I just ask, follow up on that. So if things are done under quality improvement, have you gone forward and published any of those? And how do you deal with that issue of publication and dissemination beyond your own institution? So we have an IRB approval eventually to do the quality improvement project and then publish. And so, for example, we're given the approval to pull data in a certain date range and then publish off of that. And so we had to keep doing revision, so we did the first analysis that Laurie presented and then we had to extend our date range. But I think, I mean, I think one of the issues maybe Terry to additionally capture is, I think a lot of people don't appreciate that mechanism. They don't sort of understand the how to do that. And it was new ground for us, you know, but it's really quite feasible if you sort of just sit down with IRB. And it's a very, I mean, in our center, it's a simple sort of IRB application. Yeah, and I think there's confusion amongst some that says, well, unless you've sort of set this up on the front end that you can never use data. But the reality is, is the quality improvement projects are not set out with the intent of generalizable knowledge. But in the course of conducting it, you sometimes come across very important generalizable knowledge. And so, both at Intermountain and at Geisinger, there's a mechanism by which we can do sort of an exemption of an IRB just for publication of data that is pretty straightforward. But I do know that there are other institutions that do not have that type of a mechanism available. And if it wasn't conceived as a research study, they don't allow publication. But you can also, in several cases, request a non-human subjects declaration from the IRB because if the data are de-identified and they're already in existence at the time of the study. So I think there are mechanisms. Pat, did you have a? Yeah, I just wanted to go back to your second bullet and say, you know, I think we really appreciated hearing about Angela Anderson's story. And we don't have a strong patient presence here right now. But I think if we want to think of patients as the pool for pharmacogenomics, not just us pushing the information out, but as a pool, then I think we don't want to just think of their contribution in terms of data and samples. But I think they can have a much more engaged role in thinking about both the design of research, but also the design up to demonstrate clinical utility, but also the design of implementation research projects. And I know we talked last night about maybe having an entire meeting focused on sort of the patient's role in genomic medicine. But I just would maybe think of that patient bullet less in a passive way, but more in a sort of a PCORI way of sort of patient engaged research and thinking about the design of the research questions and the measures and the implementation of the study, the interpretation, and the dissemination of the results sort of across that continuum. Because I think that they could make a really valuable contribution. Well, and particularly in pharmacogenomics, I mean, patients get this. Howard, you've told us in the past that you have patients who will pay for it themselves because they think it's so important. So somebody, Steve? Yeah, I mean, for us, the patient and family engagement is really, really important, especially if you're looking at studies that are going to follow the effective treatment over a period of time. You don't want to limit your data collection points to visits that may be a month or three months apart or something like that. And so taking advantage of apps on cell phones is a way of, for example, with an ADHD study. It's a way of collecting information from the teacher, the parent, and the child themselves about how they felt they performed at school. And, you know, we've got the nutrition people. We'll have people take a picture of their meal from which calorie counts can be calculated. There are different ways to engage patients and families to get rich phenotype data in a real time and like data rich manner. But often, we're not the people who know what those are. It's engaging whoever it is, nutritionist nurses, whoever it is that sees the patients on a daily basis. Great. Okay. We also heard we should be studying diverse approaches to implementation, including not only hospital or specialized center-based pharmacists and clinicians, but also community pharmacists. One of the things we discussed briefly was, you know, should there be some meeting of community pharmacists? Mary, you have a colleague who's considering galvanizing that community, which is a very powerful one. Should we look at using a pharmacogenetic card like a credit card versus a QR code on your phone versus some other kind of approach? We heard from Steve the importance of children versus not children. Versus adults, but differences in children and adults, and obviously from Enoli and Melissa about the differences in non-European ancestry populations. So all of those would be. And probably it's not only the ancestry, but cultural differences, challenges in underserved populations. We heard from Lynn Dressler about the barriers to even up, you know, what some of us might consider to be a low-cost test at $200, $250. It depends on your point of view, decidedly. We also heard about requiring or at least encouraging use of standardized outcomes across multiple studies so that data can be pooled. We are trying to do this to some degree in NHGRI's genomic medicine programs, at least come up with some standard outcomes. I know there are patient-driven outcomes, a set from PCORI and others. So that seemed to be. Yes. For the diverse approaches, would you include using different types of technologies like mobile health technologies? And also, now I just lost it, but it's not just engaging those particular groups, but how do you engage those groups? Right, yes. Yeah, yeah, it's more the how than the what. Good. And then, repeatedly, we heard we need other measures of benefit than cost. Cost is one driver, but it's obviously not the only driver. Anything on here? Yes. The only thing for the community pharmacist that I would add is I think that it's really a community-based practitioner, whether that's a pharmacist in a community pharmacy. It's an outpatient clinic. It's a physician who is engaging. I think there's a number of diverse groups. There are really more community-based than in a community pharmacy, specifically. I know we see a lot of the clinic-based pharmacists growing in this area to have a clinic-type service, consult-type service, and outpatient setting, too. Great. I want to leave pharmacists on there, because I think it's the only place where we mention you. And you're an important part of this puzzle. And I don't know if it's on other slides, but I think on their last bullet, I would think that we want to expand to patient-reported outcome measures as one of the measures. So on the standardized outcomes, are the measures a benefit? Well, I guess both. OK. And patient. Well. Yeah, I would just say require use of standardized outcomes, including patient-reported outcomes, parenthetically. Thank you. I had writer's block in front of 30 people, including patient-reported. Now I have typer's block. OK. Great. We do have a few more of these, so we'll go on. Developing methods for studying outliers. I think we heard from that side of the table about how do we get not only people to pay attention to the outliers, but how does one go about studying them, identifying them, finding them, treating them appropriately. Creating a system for pulling together rare adverse drug reactions, patients who have suffered rare adverse drug reactions nationwide. We did hear some discussion of the FDA's system for doing that, which the FDA will tell you they recognize is imperfect. Are there better ways for trying to identify these, particularly in these days of fancy electronic medical records and other things? Jeff was encouraging us to consider creating registries of pharmacogenomic patients. We hear about exome patients and genomic patients, so PGX patients. And Jeff, I was going to, sorry I should have warned you in advance, but I know you're up to the task. Could you elaborate a little bit on what data would be collected and what uses those could be put to? So I think this would be, first of all, in the usual care environment that sites like the ones around the table, but also a number of the affiliated sites that Julie and her team have put together for the pharmacogenetics clinical research network are already deploying pharmacogenetic testing as part of their care systems, either as research programs or as clinical care projects. So the idea would be to capture that vast number of individuals and their clinical outcomes that occur in the context of usual care, often through the electronic medical record. And I think this is really taking something from the playbook of particularly the cardiology, large-scale registries that have been both used to develop quality metrics by which performance is based, as well as to provide evidence on which reimbursement, as well as FDA policy have eventually come out. And but this is a really long-term sort of play over the next 10 years, not something that I think we're going to reap the benefits of in the next year or two, but probably over the next five to 10 years. Mark? I think the other opportunity, although it's a little difficult to imagine how it could potentially work, is to build on a large-scale program like Sentinel, whose intent is to try and determine if there are adverse events occurring with medications in the clinical practice. And if you could somehow tie pharmacogenomic information into something that looked like Sentinel, you could then have a very large group of patients using a standardized methodology that would need to be invented from scratch. Conceptually, it's fairly attractive, but the operationalization of that would be a nightmare, I'm sure. One potential might be the CDRN network of PCORI. Good idea. Yes, Sandy? Yeah, just as a quick note on that, there is an effort to tie Sentinel to the PMI, oh, there is an effort to tie Sentinel to the PMI, which may start to provide some of that. I mean, wouldn't that include everything that's been done in eMERGE? Sorry, everybody who's had pharmacogenomic testing. Yeah, but isn't that tens of thousands of patients? No, it's 9,000, and they're representative of... For PGRNC. For PGRNC. But to us, has it been done on how many thousands? On a smaller panel. Yeah, and the ability, I think the strength of the registry, is Jeff, correct me if I'm wrong, is getting updated information on what happens to them long-term. Right, they're continually updated, and that is the strength of it, is it's the long-term longitudinal data on outcomes. Great, okay, cool. There was a suggestion that we shouldn't hold pharmacogenomics to different standards from other kinds of clinical care that has been implemented without clinical trials, and perhaps trying to compare on a more systematic way, pharmacogenetic testing to implementations to other routinely accepted, now routinely accepted testing, or new testing that gets added into clinical care. So why do those get accepted, and this has more of a problem? I might comment, just parenthetically, we have seen that there is exceptionalism in genetics, and we all decry it, but actually it can be strangely motivating. The director of the National Institutes of Health tells the story of knowing that he needed to lose weight, and knowing that he needed to lose weight, and not actually doing it until he looked at his genotyping scan, and found that he had several risk alleles for type two diabetes, and that stimulated him to lose 20 to 25 pounds. So, and there are other examples, I think, of genotype-driven trials that have had much stronger results among patients, because somehow this information to them is more compelling. So, providing users with data on which gene drug pairs aren't actionable, so this was the nose, as well as the yeses, really important topic, and Mary and Terry, if you're still here, you do have this level C in CPIC, right, but they don't get written up or published, right? We do, but actually, you know, the way that we prioritize which gene drug pairs we write guidelines for is based on many things, including user feedback, and so based on what we've heard at this meeting, we'll probably plan to survey CPIC members and see if there's interest in moving up some widely used drugs for widely tested genes, which might include some in the ADHD or psychiatric area, and maybe we will move that up and write one of those guidelines. It's something that Terry and I in the CPIC group need to discuss. It'll move something else out of the queue, but maybe there's bigger priority on doing a couple of nos. The Action Ability Working Group of ClinGen has the same problem, which is, you know, we've adjudicated, I don't know, about 100 genes, which leaves us with 19,900 to go, but one of the things that we've done in terms of the evidence review is that the review of the evidence is the real-time consuming part of it, and so what we developed was a quick screen where we developed a few sort of screening questions which is to say, is there any evidence at all, you know, on a quick review? Is this something that's actionable? Are there recommendations? And so we can go through genes and we're using it primarily for prioritization, but when we have nos on one or more of those categories, we know it's not something we move forward in the queue, and then we can publish to say, we looked at this gene, there's insufficient evidence at this time to move it forward, which is kind of what C-level is, but there's at least been a cursory, standardized review process that's been used, and so it might be a compromise given that you don't have the resources to do a full-blown review. Yeah, I mean, that is what I tell people when people come to me and say, should I act on this gene that this company charged me for for this drug? I say, here's our CPIC listing, here's how we categorize gene drug pairs, your gene and drug are a category C, that means there's not enough evidence to act on, and this information that they gave you to suggest that they act on it is not backed up by evidence, but it seems like we're asking for, we're hearing a request for a more detailed review for a few of these, maybe. Yes. Yeah, yeah, it sounds like that would be a useful thing, so. Okay, great. There was also a suggestion to link national data on drug use, I mentioned the IMS, IMS Health, how many are familiar with IMS Health? Just out of curiosity. Yeah, so it's a large nationwide systematic survey of clinicians and hospitals, I believe, John, you might know it better than I, but at any rate, very, very useful in terms, that's where you hear that a meprosal is the eighth or Xth most commonly prescribed drug in the U.S., that sort of thing. But anyway, taking national databases on drug use or prescribing and linking that with allele frequency to make some estimates of numbers of people at risk would sort of avoid our having to do this in studies like the EmergePGX or in various EMR systems or healthcare systems. But then having made that prediction, then those healthcare systems can go in and look and see, well, do they really have adverse responses or not? Just a note about IMS, they bought or merged with quintiles a year or two ago, so it's not just a data warehouse, but they're actually capturing data from clinical trials, some of which have genomic information. Oh wow, didn't realize that, thanks. Okay, so anything here that people have angst over? And again, you get another shot at these, but just while we have you all in one room, it's nice to get some input. I also wanted to mention that IMS is also starting to collect CPT coding information. So when 2D6 or 2C19 or things like that that have tier one codes are being billed for, they're starting to collect that information as well. Great, thanks. Okay, so, oh, sorry. Before we move on to that, can we maybe talk real quickly about a conversation that John really started in our panel and continued over dinner and then drinks after dinner? Wild type. So, and it's not exactly a research opportunity, but I think as we talked more to John, I think we all got a better understanding of what he was talking about. And so I think he's really pleading for two things and I'll sort of lay it out in my words and then you can fill in the gaps. And that is that we have CPT guidelines that tell you what to do if you have the genotype. He's requesting and it doesn't have to be CPT, but that somebody says you should test these drugs. And so within ASCPT and CPT, there's an effort to develop guidelines for best practices. And so I think there's an opportunity for people in this group or a larger group to really sit down and develop the set of drugs that we would say you must test and whether it would be two or three tiers. And I think Michelin's an editor in CPT and sort of feels like this is a great opportunity. The second, again, where the payers are having challenges and I think the thing as we talked more to John that was really refreshing is we've had payers at multiple, I think there's been a pair meeting here. Ignite has had me payers at multiple meetings. And I think we've just never gotten the clarity of input from them that we've constantly been seeking. So the second thing is clarity on sort of minimum testing within the pharmacogenetics. So he was talking about quality and we're like habit that we know that the genotyping is of high fidelity and that's not really what he's talking about. And so I think as we talked through, for example, 2D6 versus 2C19, he's like, we have no idea that those two are different in complexity. So what is the minimum that should be tested? And so I'll just sort of throw out, if you're doing CIP 2D6 and you're only testing star four, we would say in a best practice that doesn't meet acceptable minimal standards. So if you are not testing copy number variation at CIP 2D6, that doesn't meet a minimum standard. So if we can convene a group that would do those two things and then write a paper, what we're hearing at least for one major payer, that would make a huge difference in ability to understand what's reasonable versus not reasonable. So I'll let you take it from there. And just one other note is that within that paper, we also use that IMS data and that we actually say these are the drugs that these genotypes would affect and this is the amount of people that are on those particular drugs. So again, I think this is a topic my colleagues at UnitedHealthcare, the payer, would be, I think it's a very interesting point of discussion. I mean, to me, if we can figure that piece out, it hits that second point which I was making yesterday around, how do we improve the quality in this space? And if we don't know what's getting paid for at a level of precision, then by definition, how do we know quality? So this entire thesis of how do we improve the quality? How do we understand what was actually done? I think needs to be a comment, my humble opinion needs to be a common thread. And I'd love to get some folks together to talk about how that comes to life. Great, thanks. With respect to the quality and the conversations we were having last night to be very clear, it's not only quality in terms of the coverage of 4, 2D6, it's a variant and a copy number variant. It's also things about the quality of either the genotype calls or the sequencing that was done. So as we were talking, I said, well, so depth is important. Well, we don't have depth in a genotyping array. You have depth in a sequencing assay. Well, is a depth of 20 enough? I don't know, what's your question? Like, what are you trying to find? And so as we started talking, I said, well, with this gene, you really need this technology. And with this gene, this technology's fine. And with this gene, you really need that technology. And he said, payers don't understand or know any of that. So I think if we could come in this paper, put together what you do need in terms of not only the coverage of which variants you need, but the technologies to get you there, perhaps without mentioning specific companies, because that's not the goal, but you need sequencing with reads that are at least this long and depth of at least this. If that can go into this paper, I think it will go a long way for the payers to see it and say, okay, this is what the standards are. Right now we don't have those standards and we need them if we want things to get paid. It may be better to say, because the technologies may change. So it may be better to say this particular technology at this point in time is adequate for this and this one has these drawbacks. And actually the conversation went to where, if we could come up with what that list looks like and put that in front of some of the technology companies, if they knew that all they would need to do is create a machine that does X, Y and Z and that would put them ahead in the market, they might develop such a machine. Until the standards change six months later. Just to remind everyone, we put a lot of effort into this in the pharmacogen nomenclature group and there was no way. It's just, it's an incredibly complex undertaking. So good luck, you're right, that's not CPIC's job. You can definitely refer to CPIC guidelines to see which variants are functional and what their allele frequency is at any given time. But the data always change. Yeah, well and that's. But I think the suggestion was what's the minimum? Right, what's the minimum for, and I think we could go around the room and pick our favorite gene and probably say, if you don't at least have this, you shouldn't bother. And those could be updated just like other things are updated. Yeah, sorry. I think that would be great. Great, okay. So you'll see, we actually separated out things that were not necessarily research opportunities, but more consensus building exercises. We at NHGRI love consensus building exercises because we get lots of free effort and people come in and do things for us that doesn't necessarily need to cost a huge amount but can have a tremendous impact on the field. So something to think about. We'll get to those in a moment. So I moved that from the research opportunities into the consensus building part. But then we had this, oh sorry, are you pointing at somebody, H.E.? Oh, okay, great. So we had as a fourth objective and now I'm a little embarrassed that we left it on. We had debated within the genomic medicine working group which remember is all the people at that end of the table so you can pick on them as well as anyone else you'd like to for some of these suggestions. But we thought, well gee, maybe we should bring a design of a study or a couple of designs to this group and put it in front of you and have you bat it around. Decided against that because we wanted to have a little more free flowing dialogue. But given that we've heard a lot of discussion about strategies for large scale evaluation and implementation. And we had quite a debate on it earlier this morning so we don't wanna go back through all of that. But I think we'd like to try to capture some of the key points that were made. Among them had to do with the ethical questions of randomizing people to know genotyping. And then the point made that well, given that it's only the 1% that's actually getting genotyping at all it may be more acceptable in places that wouldn't have any access to genotyping unless they were part of a clinical trial or some kind of a study. Yes, Howard. I was trying to wait till you were done for that. I think it also, by leaving it in there it would also give some framework for a potential RFA or something that would allow some of the issues that were raised would be issues that the applicants would have to have a solution for. Including what sites they use and can they randomize and all that. And so using the Trident, well I don't know Trident True but the existing peer review process that we already have at NIH some of these things would be filtered in that context. And so I'm glad that you've kept it in as objective four because then it'll help capture that part as we go forward. Because I think at some point in time it would be good to have that kind of a system. This process versus that process type of study as opposed to this gene versus that gene. But that could be argued and has been. And will be. Okay. We also need to understand better when trials are needed. So when is the observational data something that one can go forward with and when is it something that should be tested. And as I said we debated the strategy it was suggested yesterday that one design strategy would be a pragmatic trial to randomize people to genotype guided treatment versus a standard clinical care. And yet we heard this morning that G for many of the gene drug peers that are widely accepted that is of questionable ethics at least to those in this room. So one thing that could be done would be to include a secondary outcome since there is that outliers problem. Could you have as a secondary outcome of your trial what happens to the people who are risk allele carriers. And I think there was general consensus that identifying people who you know are not going to be able to metabolize a drug and then going at and randomizing them anyway is probably difficult to justify. So would largely agree with what's up here. I know y'all are starting to fade. So all right then moving on to opportunities for building and disseminating consensus. I think we have heard repeatedly we need some standardized terminology for genetic results and for phenotype designations and Mary's group and others have gone a long way in this but she corrected my it's solved to it's partly solved and we probably need some more efforts in this area. Heidi had suggested that this question of sort of ever changing panels that with people add another variant or another gene to a panel and suddenly it's a brand new panel and needs a brand new cost and a new CPT code could be addressed by the approach that's been done in some of the monogenic diseases where you basically have a minimum set of genes and variants that need to be tested for and if a lab wants to add on others that's fine as long as the cost doesn't change. Carol. Yeah, I just want to go back to the first point which is there's a there's a difference between standards development and actual adoption. So maybe that's the partly solved part. So there are a lot of there are a lot of cases where there are well defined, well accepted standards but they simply aren't adopted especially as it relates to nomenclature. So when it's partly solved was the was it that the standards still needed development or that the implementation and adoption of the standards still needed development or both those right. But as opposed to a lot of standards development where they occur in a vacuum we're engaged with the organizations that have the ability to actually promulgate the standards into clinical care. And so as we solve the nomenclature problem we have a direct pathway into getting the appropriate codes and loyne can snow med and that that will ultimately are generally used. And so I think we have a pathway that's laid out and so it's more just completing the work as opposed to having to develop that. Just in the general idea one of the ways to compel people to do something is to get the USPSTF United States Preventative Services Task Force on board. And that's an uphill battle. But when we're talking about things that are prevention related including adverse events you may think about that. They have lay out their criteria ahead of time so you can even develop your studies toward those criteria. They have a survey period every year which I think is either just closed or just about to close this year where you can say we think there's enough evidence in this area that you need to review this area. Anybody can participate in that survey and then they will probably need some extra expertise in this area. So even if you're not on the push end or you're the developing end if you can help them with expertise that's one way to do it. And as you may know USPSTF has some teeth in terms of the current situation with the Affordable Care Act requiring grades A and B USPSTF Preventative Services to be covered. Excellent point, thanks. Okay, next we heard about inability to code or inability to bill with the codes we have, 200 CPT codes for 65,000 odd tests being a major barrier and the radical proposal from John that perhaps some new kind of coding approaches is needed. Was this a suggestion? John, you wanna comment on it? Yes, let me just push on that a little bit more. To me it's not just around coding for billing. It's, I mean that's part of it. It's also coding for quality. So the payer needs to understand what have they just paid for. So to the comment, Julie and I had five minutes ago, if there are certain specific alleles that need to be tested, it would be useful to know that. And my point is a very simple one. I don't think the CPT structure handles this use case of genomics particularly well. A, given the speed of change. B, given the level of specificity which you'd really like to get. And that specificity is important not only to ensure what has been paid for is the right thing but also what has been paid for is the right thing of quality. So we use this kind of data not just for the purposes of reimbursement. And we would use, instead of the term billing, we would use the term reimbursement, but not just for the purposes of reimbursement but also for the purposes of quality. Does that make sense? Yeah, and I think I would add there that my takeaway is that we went beyond developing to actually, I heard a certain commitment to convene the group to do it. So that's number one and number two is we may in fact have a solution at hand depending on the genetic testing registry and whether they have unique codes for all the tests that are in fact registered in the GTR. So we may not have to develop this from a tabular raza. Yeah, that's a good point. Yes, definitely on your first point, I'm very keen to move on this thread. But if there's an existing NIH or if there's an existing open source asset that can be leveraged or extended to solve the problem, I think that's entirely in my mind within scope. Yeah, and the genetic testing registry is something that we like to think is pretty comprehensive in terms of available genetic tests but it's a voluntary registry. And so one thing to think about is whether payers might want to incentivize testers to put their tests into the GTR in order to be reimbursed or that they get reimbursed at a higher level if they are in the GTR. And Terry, can I also say that that would help pull the information out of the EMR? So if you're doing retrospective data and you wanna search, that would really help figure out who's getting those types of tests and drill it down. And that would help efforts like Sandy's with the digitize as well. I mean, I don't wanna connect with Sandy on this but I would also suggest that the use cases around decision support that you folks have been talking about in EMRs are very similar use cases to either a payer or a provider or an ACO are trying to handle. So I think that the discussion point around this third bullet shouldn't just be payer based in my mind, it's around decision support irrespective whether you're on the payer side or on the provider side. Yeah, I 100% agree. The decision support has to know what test was run or hasn't been run and it needs to, and it needs to know that particularly if the test was not done in the same institution as the care is happening. I just wanted to make a quick point that I was recently made aware of a new addition to CPT codes by the AMA called proprietary laboratory analysis codes or PLA codes. I know next to nothing about them but just looking at the website, they're an alphanumeric CPT code with a corresponding descriptor for labs or manufacturers that want to more specifically identify their test. So that might be another mechanism that's already in place. My understanding is an optional pathway that laboratories can take but could be leveraged alongside GTR. Great, okay, cool. We did earlier allude to our lengthy debate on randomized clinical trials but we also noted that there weren't many NACIRs in the room and should we consider reconvening such a debate and have the people that need to be convinced. So how do you find the people that need to be convinced? I don't know. Oh, I can get you a list of those. Is that right? Okay. But clinicians as well as I think because they're the ones that have to actually use it. Yes. I was just gonna say that I think the ones that need to be convinced include the authors of the guidelines like AHA, ACC or whatever gene drug payer. So in that case it would be, you know, in addition to the payers, it'd be easy to identify those key people who are thought leaders in cardiology or whatever. One, it might be fun to have somebody, one of you, not me, right, to professional societies and say, we think this needs to be implemented now without a randomized clinical trial and then see what kind of a response you get. So, okay, all right. And then other opportunities, the building and assuring quality we talked about a little bit earlier, so I won't go over that again. We also talked about some needs in clinical informatics and I should point out that there have been kind of needs identified throughout these two days. Dan gave us a very nice list of them at his kickoff talk, but at any rate, this is trying to capture those, specifically in the clinical informatics realm, improving standardization and updating of clinical decision support implementations for a whole variety of guidelines and other resources. We recognize that EHR data need to be updateable with new knowledge, so there has to be a way to go in, which is what Teji, I think, one of the things you were alluding to is if you know that they've had this test, then you can go back and say, oh, you know, if you didn't even test the star four allele and now we have the star six allele that should be tested, somebody should know that perhaps. Plug-ins for drug-drug interaction are available, we heard, we heard they weren't perfect, but at least there are some, so should we encourage people to develop some plug-ins for drug gene interactions, perhaps through our small business mechanism, which some of you may be familiar with, has been active in the informatics space, engaging clinical IT personnel more in grants, providing funding for them in grants, so it was a great idea, also inviting them to our conferences and programs. Some of you are here, but we could probably use more, Teji. I think one of the challenges for IT personnel is they're usually people that are staff of hospital administrations, and so they can't technically engage in a grant mechanism if you're getting to the nuts and bolts of it, so thinking about how you would do that, bringing in people who would be contractors, or bringing them on as thought leaders if they're at the level of CMIO or something like that. So Rex or Mark, you might wanna comment on that because you've done this different. Yeah, and we actually have a group that we've had some engagement with, the American Medical Informatics Association has a lot of interest in this space, and a lot of us are members of that, and so I think that, and a lot of the folks that are there have operational roles, but also many times have clinical research interests as well, and so I think we could look to partner with a group like EMEA to get representatives from the groups who are interested in to whatever the activity is that we want to. Yes, Sandy. And I would just say I'm actually staff as opposed to faculty and have participated in a number of different grants, and I really do think in the community right now, there is increased interest. I think that there's a desire to figure out how to become more involved, so I think there's opportunity. Wonderful. So maybe we should think about, at the next EMEA meeting, having us proposing a seminar or something in this area or symposium on it where we get some thought leaders to actually address this. I think part of Teji's comment was that at her institution, they're not eligible to be a investigator. Not that they couldn't do it, but just structurally, some places are easier than others to acknowledge that staff can be major contributors. Industry leaders. Absolutely, okay. Well, usually though, that doesn't limit them from participating, it just usually limits them from being the PI. And we heard very emphatically that trying to curate manually, haplotype, phenotype assignments is a fool's errand. And yet it was what was going on in many sites until Marilyn, you and your group began the farmcat effort, which I think we're all grateful for and hope that it will continue. But trying to do those kinds of things and perhaps identifying other areas. Remember, ClinGen was born out of basically a conference like this where we found out that multiple groups were trying to identify actionable variants and they were all sitting around tables like this one and going over the same evidence and largely, but not entirely coming to the same conclusions. So it seemed like trying to harness those efforts made a lot of sense and then disseminate them effectively. And then there was the note that we have also heard repeatedly, the patient is the one constant in all of the medical system interactions and if there's a way for the data to follow the patient since the genomic data largely do not change, that would be awfully nice. But coming up with a system, Sandy you can do that as your next small project. You have a son who does this, right? It'll probably be several lifetimes before that happens. Other things related to data quality, we need some kind of an infrastructure, again, if the data are gonna follow the patient, infrastructure for storage and accessibility of these data and you need that in your institution if you're gonna use these data right now and still in many institutions, these are PDFs, we need to get past that. Dan, were you just scratching or did you have a comment? Oh, okay, all right. Yeah, sorry. I'm sorry, I meant Dan, did you have a question? For those who are viewing, you're scratching his head. So, moving on, it is getting late. So also, should we, in regard to the data resources that we heard about, is there a way that we can aggregate them all or point to them all? The big pharmacogenomics tools and sort of one mega-site, the one site, as Jeff said, to rule them all if there's a way of doing that or at least to link them together. Were there other clinical informatics opportunities that pop into people's heads that aren't listed here? Just go back to the longer side. Okay, all right, we're almost there, guys. Education and workforce development, we had a session on that just before this one and so these may be in a little bit wetter clay than perhaps some of the earlier sessions. But we did hear repeatedly throughout these two days the key role of the pharmacist in this process, the need to sort of grow that group into the pharmacogenomics space and many of them are quite interested in it, at least that's what we're hearing and I think, Julie, you're seeing that and Kristen, in your courses and that. And Dick told us how clinicians, those are the people that clinicians contact and we wanna be sure that there are people available to consult. So perhaps as Howard described, an extra year and Julie, I think you have a program like this too. So doing an extra year in pharmacogenomics, people are beginning to see this as a viable career path including nurses and patients in education efforts was mentioned repeatedly. Engaging community pharmacists is a little bit redundant, sorry. The value of multidisciplinary training as Howard described what used to be known as molecular tumor boards, but now is a clinical genomic action committee, which is a great name for it. But really the point being that even having trainees being involved in these processes where they actually learn the value of the various disciplines that need to be brought to bear in this area. The plus one year is another approach for a completed clinical pharmacologist or molecular pathology trainees and even having sort of various levels of potential training. So a one month possibility, a short boot camp, a full year fellowship. We heard Genomics England, actually it's Health Education England that is doing this in multiple tiers so that you can have a certificate, you can just take one course, you can have a certificate, you can get a master's degree. There's kind of a series of possibilities. Sorry, there's a duplication there. A lot of discussion toward the end of the day, which is a tough time to have discussion, but you all did, on including compelling case reports that can really grab clinicians. The case that Josh and Dan presented on The Woman with the Nine Stents and Restent Stenosis is something that I heard Josh present probably two and a half years ago at a, I can't even remember what meeting it was, but I remember that slide and I emailed him and I said, Josh, I really need that slide, it's critical. And by the way, what happened to her? And he said she hasn't had, it's been two years I think, and she hasn't had another event, which is pretty amazing, so. Publishing lessons learned in implementation. Again, something that doesn't get into top tier journals, but is really important and needs to be done, considering webcast cases. Genomic medicine obviously is near and dear to my heart, but for the purposes of this conference, cases in pharmacogenomic medicine is probably a useful thing to consider. Again, who would do this and how, and that is something that we need to explore. Providing a sustained online forum, following courses like the course that you lead, but some way for people to continue to interact, similar to the approach taken by the City of Hope, include clinicians in the design of education programs, always makes sense to include the learners in what you're trying to get them to learn. And trying to sneak education, Howard, I thought that was a stealth approach to putting it into things that are already happening and avoiding focusing on, much as I love the term, leaving the genome sometimes out of our advertisement and publicity material so we don't scare people away, because I think we sometimes have had that effect on folks. Anybody have any problems with what's up here? Yes, Mark? No problems, but there is an issue that straddles education and clinical informatics, and we touched on it, but didn't state it explicitly, and that's point of care just in time education. So when a physician has to make a clinical decision now, how can we get the information that they need to make that decision into their hands? And so that's, it was part of what I was asking about, but it wasn't very explicit about it, but I think that needs to be articulated as well. So it's sort of point of care, education? Point of care, quote unquote, just in time education. So giving them the education that they need right at the time that they need to make a decision. Actually, I'm gonna save here before something awful happens. Mary, was there somebody else asking a question? Maybe not. You know, just one additional comment on that. Doing a project like that that involves the EHR would also have the effect of really driving standards into use. I think that type of thing can really help with standards adoption. The point of care education would? If the point of care education is delivered contextually through the EHR, then you need to get, in order for that to work, you need to receive the data in an appropriately standardized way. And so now all of a sudden, everyone has an incentive to send it to you that way. Makes sense. Yes. And I think one of the things we've learned that I don't think I articulated well in focusing on implementation is really focusing on what is it that our target audience needs to know. So you have the incorporating clinicians within the educational design process, but I think it's also thinking about what it is we're actually asking them to do. So we want them to make whatever a medical change based on genomic information and making sure that our educational programs are focused on that end versus giving them everything in the world they need to know. Well, and I remember hearing one geneticist bemoaning the status of the state of medical students and junior residents who didn't even know what an intron was, how could they possibly practice, and it's like, why do you need to know that? So let's try and focus on what's important. Because you don't want to sequence that first. And they wouldn't be doing that. That's the thing. Other points on education? One thing that Bob alluded to is bringing in other types of educators into designing the education. So doing it out of the box, I hate that. But finding a different way to educate people. Yeah, we don't have geneticists doing surgery. Right. So we need educators to be part, I mean professional PhD educators to be part of the teams that are designing the effective methods. So clinicians would tell you they are educators, doctor means teacher, but some of us are better at it than others. Yeah, I think it's a part. It's about the techniques and what's working currently and what the evidence is behind from an educational standpoint. And then also you mentioned adult learners. So how do you teach an adult stuff in different ways? Because everyone learns differently. So sometimes we have to step out. And I think ISCC is very good in facilitating that. The meeting that I mentioned that really focused on short snippets of kind of educational best practices and genomics from various associations, there was so much within that content that was really just good educational design and adult learning. And so I think CME does a really good job from an association standpoint of really considering that in the design. And there were some great, great techniques. And I'm not saying that physicians can't learn how to do that. But if you're gonna look at every single specialty and every single specialty society and say, do the physicians who rotate through their CME committees all know how to do that well, I would say, probably not. Absolutely, okay. Great, I wanted to leave a little bit of time for a last question that was posed to us by John. And I have said to him personally and we'll say more broadly, it has been incredibly valuable having your voice at the table. And if you guys know of other, because John can't come from Minneapolis for every one of our conferences, but if you all know of people like him and many thanks to Dick for suggesting we include him, we would very much like to have this voice at the table because I think we've seen how valuable it is. So he asked us the question. So let's see, what is the sort of one, three and five year, John, you should ask your question. But this is how I captured it. I don't think you captured it well. What is the one, three and five year projection for PG action implementation? Also reframe the smartest people in America around this table right now. What do you folks see happening one year out, three years out, five years out? How do you see the technology and the science evolving? So in other words, if you're in software development, you say, paint me a roadmap, help me understand. Now that's a little bit of gazing into a ball. But it's a useful exercise because it starts to think about how this could meaningfully impact the system. And subsequently, folks, whether they're on the pair side or the provider side or in life sciences didn't get surprised by the technologies. They can start talking about them at a much earlier stage in their thinking as they do their own internal business plans. Mark. So I'd break that out into a couple of different things that you'd wanna project over at that timeline. One is availability of the data. And I think we could anticipate that there's going to be some type of an acceleration of the availability of the data over that time. The second question is then is the use of the data in evidence-based ways, which will presumably lag behind to some significant degree. And then the thing that probably won't be accomplished, at least other than perhaps to build the systems would be capturing the outcomes from the implementation. So I would see those things as being sort of the foundational elements and those things are going to be different. So for my organization, I'm going to have available data in a very short period of time. So under one year timeframe, I'm going to have PGX data on 90,000 patients that assuming we can figure out some of the issues of how to do the validation we'll be able to use. So then the question is, is how do we actually implement that? Which is probably going to be more in the three-year timeframe. And my hope would be at the same time that we're looking at the implementation, we're also defining the outcomes and building the mechanism to capture those outcomes. So at five years, I would have one to two years of outcomes data related to the implementation. So that's how I would look at this for my specific organization. So I wonder, that is one way of looking at it. I wonder if another is to say, what do we think is going to be happening in clinical care five years from now? And we can all have our views of that. John originally asked us for what we think will happen in 10 years. And I've been saying as have several others that in 10 years everybody will have a whole genome sequence in their medical record. We've been saying that for about five years now, but even so. But yeah, five years from now, do we think that there's going to be, pharmacogenetic sequencing is going to be, sorry, pharmacogenetic testing at least for major gene drug pairs, et cetera, is going to be standard of care? I kind of think it probably will be in five years. But I could be just a cockeyed optimist, but. Well, I think one trend that we has been alluded to that we need to be aware of is patients getting their own testing done outside. And what is the workflow when a patient comes to their primary care provider with a list of 20 variants, maybe 10 of which are actionable? What do they do with that information? So just to push on that. So one thesis here is to break it down, like Mark said, look at it from a technology lens. I think that's important. I think you also want to break it down from a consumer lens. Is there going to be a new PGX company doing consumer-based testing? I don't know. The other lens would be to look at it by the disease state. And again, you folks will know this much better than I do. I'm certainly not technically deep in this domain, but is it going to be more impactful in certain clinical conditions than in other clinical conditions? That's a useful thing to know. And I think visibility to that maybe outside of this community is limited. So if you can help in that journey, in painting this picture, I think it would be really useful. You had three lenses, John, and I only got two of them. What do you remember? So I had a technology one, which to me is Mark's piece. Mark's talking more than technology, so maybe I shouldn't call it technology, but technology and data, I think there's a consumer piece to this. I think there's a disease-specific lens. Probably some thought needs to be done to think about how you would break this down into meaningful chunks. But off the top of my head, those look fair. How do people feel about that? Howard? I agree with that. I think there's a caveat with a disease-specific is that there's certain areas, a solid organ transplant, probably cardiovascular disease, psychiatry, where there's enough data emerging that it'll become useful, if nothing else to weed out some options as you try to choose from the buffet of possible treatments. But then there's some areas, very big areas, like diabetes, that I would argue, and someone may hit me for this, but I would argue that there is not even enough research going on that it's possible that we'll have enough data in five years. And who knows about 10, but if there were studies going on today that I thought would answer that question, then maybe in five years we'd have it. But the studies that I'm aware of today are not going to answer that question. They're still trying to find a gene that might be associated with something. And so looking out for those diseases, I don't think, and then that's a wonderful place where there was seven different types of treatment. It'd be 13, no, not 13. But it'd be great to be able to choose from amongst those various options, but yet the data is not even being generated to start that in my personal view. So my personal view is to disagree with Howard. There are, one percent of the diabetes population has a genetic cause and that should drive different therapies for them. And there are 13 different kinds of that. That we know of. That we know of. That we know of. And that doesn't sound like a big number, but that's a half a million people in the United States because there are 50 million diabetics and one percent of 50 million is 500,000. And so that's a pharmacogenomic challenge in a different way because it's disease-based pharmacogenomics, but it's certainly something that John wants to hear about. And I think that it's an interesting example because it does represent the sort of intersection between rare, not so rare disease and targeted drug therapies that had nothing to do with the pharmacogenetics, sorry, the pharmacokinetics piece that we spent a lot of time talking about, but a rational nevertheless. And I think would be easier for an adoption pathway than clopidogrel, for example. You make an excellent point, Dan, and we do have an ignite, as always. I should have made it earlier. And we do have an ignite project that is trying to identify those monogenic forms of diabetes. And it occurs to me, if as part of the workup of a new diabetic, you not only got the anti-gaba, whatever it is, antibodies, GAT antibodies, which everybody tests, but at least looked at one of these genes. I mean, Dick, you and I treat hypertensives. When you diagnose a new hypertensive, you're supposed to get a renal ultrasound to make sure you're not missing renal hypertension. How often do you see that? It happens, but it's rare. You're supposed to test urinary metanephrines. Sorry? And no, it hasn't. Urinary metanephrines, every hypertensive is supposed to get this. But we don't do testing for sort of basic genes that lead to these diseases. In all honesty, I think we're surely, not because we thought about it this way at the beginning. In pharmacogenomics, because of the PK side, which is the place we started because we had a phenotype we could measure easily and it's quantitative in his blood drug levels, we have a spectrum of genes that are totally impervious to the way we organize medicine. That is, 2D6 is found in psychiatry, tomoxifen and breast cancer. It's all over the place because these genes are drug metabolizing, genes encoding drug metabolizing enzymes and transporters. So this is one area of genomics that basically covers every discipline in medicine in one way or another to varying degrees of our current understanding. Now Dan's right, we're gonna move increasingly to the pharmacodynamic side and we're gonna find, I remember when someone explained to me in words of one syllable, that receptors aren't going to show variation because unlike drug metabolizing enzymes, God made them, they're really important as opposed to what I was doing. So, but they do. So blocking those receptors will be fatal. I understand Dan, but nevertheless, we're gonna move increasingly to PD side. Frankly though, what's gonna happen, I think with, and would I have guessed that during my career I could say this, is that every patient in academic centers you're gonna have a panel of sequence based data on the genes we currently have data on and that's gonna happen during a fairly brief time period, three to five years. I have absolutely no doubt that that's going to happen. And then it's really the infrastructure, building the infrastructure within the institution. The technology of how you do 2D6, which is complex, they're not gonna really be, and you guys will know who does that well. It is really hard as the people around this table know better than anybody to build the infrastructure with if it's not going to be alerts, it won't be alerts, it'll be some way that'll be invisible to the doctor. That it'll be in the pharmacy some place and it'll be a hard stop on a prescription if you're going to, if it's TPMT, star 3A, star 3A, and then go back and ask for an alternative drug. So I think you're gonna see this very quickly, which is why I made those provocative comments about clinical genomics for everyone everywhere because I really think that you're asking a good question and I can't say how happy I am to have somebody who understands the mind of the payer and what they're concerned about and seated at this table. So John's done us all a big favor and I wanna thank him too. But I think the timeline is going to be very rapid and that's been driven by a lot of things out of the NIH. I wasn't just gratuitously saying that yesterday without eMERGE, without the PGR and none of us would be having this. But the NIH doesn't have the resources to do the implementation, not really. It takes in a major investment. You went through that at St. Jude, we've gone through it at Mayo, but we're gonna see that academic medical centers that want the U.S. News and World Report to rate them even though the CEOs will say we never look at it right, they're going to all be doing this. So I think you're gonna see this very quickly and we need your help to do it right. Great. Thanks, Dick. One final point, thanks, Dick, for that. I appreciate the comments. One final point, I wouldn't just say useful for payers. I actually think this is useful for everybody. Yeah, you're absolutely right. I would argue this is useful for the healthcare system. For the patients. Yeah, exactly. I genuinely believe this, yeah, exactly. It's useful for... Sorry, this is an inside joke because the PMI has changed its name to the All of Us Research Program. And so whenever you talk about All of Us, it's like what are we talking about? But anyway, the model future needs not only for reimbursement but for infrastructure as we heard about for education. No, I'm just playing. I'm playing. I had one other intriguing question maybe and I don't know if it directly affects payers or indirectly affects payers, but I could see it coming down the line as we talk about UPGX. As Dick talked about carrying a card around, there's going to be a portability piece. Consumers, health systems are going to want to port this data. Genomic data that as we all keep saying only has to be done once, especially once we get to sync with the data. And I don't know where that plays into it, but I think it's coming down the line. Let me actually, I would, sorry, I was just going to say I would keep the capital A and U in that as well because I think if we're successful we will actually enable all of us to implement. Good point. Okay. Steve, and I should just say we are aiming to end at three as promised, so go ahead. So I was just going to extend what Dick was saying to even further to the realm of totally off the wall. And, you know, kind of part of this discussion is we talk about patients as everybody else, but the people in this room. And it got me to thinking my institution is self-insured and Lynn said that mission was self-insured. And I wonder how many of our home institutions who are investing heavily in pharmacogenomics are prepared to do a pharmacogenome or a panel on each of their employees and see if it makes a difference in their costs over a five-year period. I'm going to actually go back to my institution. We are, I think, 7,500 employees. I think we could do our targeted next-gen panel for maybe 50 bucks a pop. I wonder how that stacks up against the annual physicals that's provided free, the vaccinations, and some of these other costs. Might we get some data for the study that NIH won't fund from our own institutions? Just a horrible question. So we've actually looked at that. We have a self-insurance program that's not UF-wide and it's called GatorCare. And we pulled data on the drug use. GatorCare? It's called GatorCare. The best in veterinary medicine. We have an excellent vet school. But we pulled the data on drug use in the GatorParek population, which is physicians, Shands Hospital employees, Athletic Association. They are healthier than the people that are in the Vanderbilt database and in the Emerge database. And so we have, because that's exactly what we were going to do. And then we decided that it might, that might not be our best test case because their drug use is actually much lower than sort of the average person because they're well-educated and probably healthier. So you might want to look at that before you jump on that. There's also issues of safe harbor in the sense that you'll be hit by Gina twice as an insurer and as an employer if you don't do it correctly. Well, and Marilyn stepped out of the room but the Penn State experience on encouraging their faculty and staff to be genotyped. I think there was a financial incentive to that if I recall correctly. That they would get a lower insurance rate if they had genetic testing. And they got in all kinds of trouble. You don't know, okay. Maybe I'm hallucinating, it often happens. There was another comment. Oh yes, I'm sorry. So the chances of pharmacogenetic implementation to a community type of health providers would increase within five years if the research community could provide and forgive me what I'm going to say in my next statement. Because I think the list of gene drug target is not extremely impressive. So if the research community could generate more such a target. More exciting drugs and genes? Yes, so. I want to see what we could do about that. I'm suggesting the body like this maybe we could develop the net to catch new targets. And my suggestion would be for example anti-hypertension pharmacogenomics which I think it's a very fertile area. Because like in my institution, it's a huge institution, 16th hospital, 160. But if I would show them the list for the administration, I might say they would be hard to impress them and mobilize. You should have them talk to Mark and Steve and a couple of others. I bet you'd have good arguments to make for, or Dan for implementing even the unimpressive list we have so far, you can avoid some pretty important adverse reactions. Yeah, and I don't know that it's for dent of looking. I think that the reason that you see the drugs on the list is because they're the ones for which we have the largest signal of effect size. And so I don't think it's because we haven't looked at hypertension or the SSRIs or that sort of thing, but we're just not finding as much that moves the needle as significantly as we're seeing with clopidogrel and morphine and that. So I think that's one of the pragmatic reasons that we've sort of initiated focus on that even though the point that you make is well taken which is we're treating way more people with hypertension and diabetes. And so even if we find effect sizes that are somewhat more modest in the pharmacogenomic realm, they might in fact have a larger cumulative effect if we're able to implement. Okay, I think at this point, we've probably tapped everyone out. We've worked pretty darn hard in the past two days and we very much appreciate it. I'm gonna leave Mary with the last word but before we get to the last word just a reminder in terms of next steps. Our colleagues here, Melpy and Ellen and Collette who was here yesterday, will be writing up a summary of the meeting. We'll send it around to all the participants and ask for a rather quick turnaround on it. It's a summary that basically goes on our website and says we talked about, you can see some of them from our previous meetings. We usually do an executive summary that's a one to two page. Here's the key stuff and then maybe 10 pages of other stuff. And then following that, we would very much like to put together a manuscript. My colleague, Simone of Olpe has volunteered or been volunteered to lead that and with Mary and me, we hope to get a draft out to you in the next six weeks or so. So that would be our target and having said it publicly, it gives it a little bit more force but recognizing that all of us are busy, we'll do the best we can with that. We would ask for your responses back within about a two to three week window. We usually have two or three, about one or two back and forth on these things and then we submit them. Target Journal, probably clinical pharmacology and therapeutics, which is the group that publishes the CPIC guidelines if they'll have us. And just a reminder that you do need to respond in order to be included. So I think with that, Mary, you have the last word. Well, I guess, Terry, thank you and all of the staff. You don't have the last word. Okay. Thanking the staff. Yes, so I'm so sorry. So very important and Jeff, you can tell me who I've forgotten but particularly Kim Davis who is outside the room but who did a fabulous job in keeping us all organized and Teji particularly who kept us all excruciatingly organized. Alvaro and Mukul who are doing the video taping and webcasting for today and anyone else. Oh, I already made, well, and of course, Melpy and Colletin and Ellen who are taking our notes. So many thanks to you all. Now you can. And also to thank the Genomic Medicine Working Group who talked about this meeting for many, many months at our regular conference calls. And to all of you for participating, I mean, when I look back over our objectives, I think we really did get a survey of the landscape of what's going on in pharmacogenomics in a way that has been really helpful to me bringing this many people together in this room who are really doing some of the cutting edge research. We talked a lot about limitations. If you remember, my very first slide was on the hype. And I do think that that last question helps to illustrate, you know, there is a tremendous hype about pharmacogenomics. There are horrible adverse drug reactions and there's an unacceptably high percentage of patients who don't respond to medications. The way that we would like them to. And we do have to keep in mind that pharmacogenomics is probably not going to solve the majority of those problems. But it is going to solve a significant minority of the problems that we have. And the point is that we have the knowledge, we increasingly have the genotyping, we just need to put in place the processes to do the implementing. And I commend all of you for doing what you are doing because many of you are making tremendous progress in implementing genetic testing to improve patient care to the extent that we can. And obviously sharing your information with everybody else, which is really great. And to thank NIH and especially NHGRI for supporting the work because they've definitely planted the seed for us to do what we've done. So thank you. Sure, thank you. All right, safe travel soon. Safe call.