 All right. So we're in session two. I'm Jeff Ginsburg from Duke University. This is Resources for Pharmacogenetics Implementation. Our first speaker is Mary Relling, who's going to tell us about CPIC. What we're going to do in this session is leave about 20 minutes for each talk, 15 or so minutes for the talk, and then a few minutes afterwards for some clarifying questions. We have an hour of discussion to follow all the talks in this session, which we really can take a deeper dive into some of the strategic issues. Okay? Mary? And I'm really going to be talking about PharmGKB, which is co-led by Terry Klein and Russ Altman and Stanford, and CPIC, which Terry and I co-lead, and as I just realized this is 18 years worth of work in 18 minutes. So basically PharmGKB is a knowledge base that curates pharmacogenetic knowledge. So it does this via various ways of summarizing the data, including annotating genetic variants that are associated with phenotypic effects, cataloging FDA and other regulatory agency labels for what their effects are, and has been a leader, as we've already heard about, with the Warfarin Group in leading consortia. And just to point out that the CPIC, the Clinical Pharmacogenetics Implementation Consortium, was one of those consortia that PharmGKB took the leadership in facilitating when we started back in 2009. And in these subsequent slides I'm going to show you examples of some of the ways in which PharmGKB curates knowledge about pharmacogenetic variants. So this kind of illustrates the degrees of knowledge that PharmGKB is summarizing, and it does include both automated retrievals of the literature in terms of pharmacogenomic knowledge as well as manual curation that their expert curators work on. They include coming up with very curated things like drug-centered pathways, giving special attention to very important pharmacogenes, and evaluating the evidence that links genetic variation with pharmacologic variation. And finally has some parts of pharmacogenetics that are involved in clinical implementation, which includes guidelines such as the CPIC guidelines. This is from the homepage at PharmGKB, and one point to make is that the PharmGKB webpages are going to have a different look and feel pretty soon. So some of these slides that I'll show you look differently than the slides that are currently up on the live site. One of the things that PharmGKB that is super useful are these annotations, and these include clinical annotations as well as information, as I mentioned, on drug labels, these pathways, and variant annotations. This is an example of a page on PharmGKB that just summarizes drug labels from the FDA, from EMA, and from other international drug regulatory agencies. It includes which drugs have markings in their labels related to pharmacogenetics, and it includes an assessment of the degree of warnings associated with that information, such as genetic testing being required, actionable pharmacogenetics, and this is updated on a regular basis. The PharmGKB pathways are extremely curated and really a unique resource in terms of pharmacology, where pathways that are both pharmacokinetic and pharmacodynamic are summarized. And this is one example for one of our favorite drugs, methotrexate. There's a pathway that indicates how it acts in a cancer cell, and you can see the incredible number of gene products that are known to be involved in methotrexate mechanism of action, as well as the gene products that are involved in the systemic pharmacokinetic variability of methotrexate, and these are freely available on PharmGKB, updated, and usually have a publication associated with them. This is an example of a drug page for phenytoin on PharmGKB, where there's information summarized as to what kinds of warnings, how many warnings are in drug labels, how many clinical annotations there are, how many pathways are on the site, as well as basic information about drugs. So you can see that for those drugs that are annotated in PharmGKB, there's quite a bit of information. The clinical annotations come from this, again, automated and curated review of the literature that PharmGKB curators perform, where here, again, is an example of for phenytoin annotations from the literature that link genomic variation with variation in phenytoin response, and then you can get access to all of that underlying information by clicking on the hyperlinks. This level of evidence that is reflected in the clinical annotations is really important, so they go from levels one to four, with level one evidence being the highest relating genomic variability to variability in response, and, of course, for something that's clinically actionable like the CPIC guidelines, the level of evidence relating genomic variability to phenotypic variability for the drug is very high, usually levels one. They also summarize dosing guidelines on PharmGKB, so you can sort this by drug, you can sort it by gene, and they include a listing of many different professional guidelines, including, of course, the CPIC guidelines. And so, PharmGKB works extremely closely with CPIC on developing these guidelines. And in fact, Terry and I really worked to start as part of the Pharmacogenomics Research Network, CPIC partly based on a survey that we did, a survey of experts in pharmacogenetics asking them, you're an expert in pharmacogenetics, why aren't you implementing pharmacogenetic testing in the clinic at your site? And 95% of respondents said that the process required to translate genetic information into clinical actions was the major impediment to implementing pharmacogenetics with the next two responses being how to interpret the genotype tests and which gene drug pairs should be implemented. So with that in mind, we formed CPIC to create guidelines to help clinicians understand how available genetic test results should be used to optimize drug therapy. And importantly, we realized at this time that all genetic testing was becoming more common. I know that we're in this interim period where it's still not the norm, but genetic testing is getting less and less expensive all the time. And we think that eventually the challenge for clinicians is not going to be whether they should order a genetic test, but how they should use the available genetic test information to guide prescribing. And so that we assume that eventually high throughput preemptive genotyping will become more widespread. Let's not argue about whether the genetic test should be ordered. Let's pretend that we all have the genetic test results, how should we act upon it? And so our aims in CPIC are to create, curate and update freely available peer-reviewed gene drug guidelines. And then as I'll talk about in a minute, to enhance interactions with the external world. The CPIC guidelines that we have are posted on CPICPGX.org, and they highly capitalize on PharmGKB resources. They're freely available with no limits on use. They're peer-reviewed and the journal CPT has the first right of refusal to publish. They follow a standardized format with grading of evidence and recommendations. Importantly, they can be updated on the CPIC website ahead of updates to publication so we can be responsive to new data. And we have strict authorship and conflicts of interest policy and closely follow the IOM practices on clinical guidelines. It's an international consortium, so we have members from all over the world. And this is a screenshot of some of the guideline title pages in CP&T. And every one of those guidelines has its own page on CPICPGX.org where all of the information for that guideline is aggregated. Starting in 2011, we started publishing these specific gene drug pair guidelines. In blue indicates where guidelines have had to be updated. This is a total of 22 guidelines affecting 17 genes where we estimate will be at the end of 2017. Looked at another way, we divide our guidelines into those for which there's strong or moderate prescribing actions, and we call those level A gene drug pairs, those for which there's an optional prescribing action, we call that level B, and those for which we don't think any prescribing actions are possible. We call that level C, and our plan was to try to write guidelines, including all of these gene drug pairs. It turns out that we probably are not going to have the bandwidth to do the CPIC level C guidelines, which is unfortunate because these are often problems for clinicians where patients or lab companies may be pressing to do testing where actually the evidence doesn't really support that testing. The key part of every CPIC guideline is its table two, which is the prescribing recommendations based on the genotype assigned phenotype. So this is for 2C19. We divide patients into these three major phenotypic groups and have therapeutic recommendations for those groups with the classification of the recommendation, and this case is either strong or moderate in each case. And the important part of this is that these prescribing recommendations are backed up by evidence, which is summarized in detail in supplementary tables for every CPIC guideline. What we're trying to do is provide resources so that you can get from genotype to prescribing information, in this case represented by a interruptive CDS alert. And so, again, here's the CPIC guideline page for Voraconazole and Cip2C19. We have hyperlinks to the main guideline, to the supplement, and wherever applicable, we have updated hyperlinks to more complete tables that are present on PharmGKB that we can update that might not be present in the original publication. So we're really trying to get people to go to this CPIC guideline page to look for any updates. And then we have detailed tables that are downloadable for anybody doing clinical implementation to back up the actionability. The first step is translating genotypes to alleles and assigning functions to those alleles. And there's a table that's the allele definition table. Every variant is unambiguously linked using five different systems for coordinating positions on DNA. Every row is a different allele. And the nucleotide position changes relative to the reference is indicated in these tables. And importantly, the assignment of function to every one of these alleles is greatly facilitated by the PharmGKB annotation system. And every one of these functional assignments is backed up with literature that's present in the guideline. Adding alleles to diplotypes is really important for pharmacogenetics. This is illustrated by the first few CPIC-able genes where we looked at what's the difference between prescribing recommendations for someone who's homozygous for a variant versus heterozygous for a variant. And in the vast majority of actionable CPIC genes, there's a huge difference in those two situations. So being able to resolve variants into alleles and diplotypes is critical. At St. Jude, working with Dr. Burkell using the Afimetrix DMET Plus software, it actually is a very nice system for translating every row as a variant here, translating those variants into start alleles that uses good population-based software. So it's easy to go from alleles into diplotypes. But not everyone uses that, and a lot of people are doing sequencing, so we need other systems for aggregating variants into alleles and haplotypes. And that's part of what Farmcat is doing, which you'll hear about from Marilyn Ritchie in a minute. We also have to go from diplotypes to phenotypes and then interpret those phenotypes. And there's tables to back this up, also housed on FarmGKB, where we assign the phenotype based on the diplotype for every individual. And then finally, going from those phenotypes to prescribing actionability. And the CPIC Informatics Group helps with these for every guideline where we again provide some detailed interpretations based on phenotypes and provides sample wording that can be used for interpretation in clinical decision support systems. We also provide sample algorithms that can be used in implementing clinical decision support institutions based on those genetic test results. Updating is really important, as I mentioned. So a few updates that we've posted to CPIC guidelines just in the last few months includes adding the T15 for thiopurines. This is a gene that wasn't even spoken of about two years ago and is gone to actionability already. So we made our users aware of these findings immediately, and they'll be incorporated into the next update. The Blackbox update for Clopidogrel and 2C19 in the last few months, and as we've heard about the GIFT trial for Warfarin. So the other important thing that we do in CPIC is work with other groups. CPIC guidelines are linked to practice guideline filters on PubMed. They're cited in the GTR. They're endorsed by a couple of professional organizations on their websites. And we've really put a lot of emphasis on developing standardized terms for pharmacogenetic results because it's very important for clinical actionability. If we want to fire a warning at somebody based on the absence of a genetic test result, we need to know that that genetic test result is absent. And unless we use standardized language, we cannot find that test result. And if we want to fire an alert at a prescriber based on the presence of a high-risk test result, again, the logic in the EHR needs to find that positive high-risk genetic test result. So in response to this, we led a year-long project, a Delphi process, to come up with standardizing terms for allele function and for phenotypes. And that's been adopted. It's been endorsed by the American Association for Molecular Pathology. It's been incorporated into proficiency testing results for the CAP for pharmacogenetic test results, and adopted by other groups as well. And as you'll hear about in a minute from Heidi, ClinGen is a really important group that PharmGKB and CPIC interact with. PharmGKB is recognized as a three-star submitter, and CPIC is represented as a four-star submitter into the ClinGen system. So PharmGKB and CPIC both have a lot of interactions with ClinGen and ClinVar, and we're trying to work with this group now to assign clinical significance to every pharmacogenetic variant, which you'll hear more about from Heidi, I think, and we've come up with a proposed group of terms for this clinical significance, which I think you'll also hear more about from Heidi. Those standardized terms are really important in working with digitized, and you'll hear about that from Sandy Aronson in a moment. And PharmGKG was the first poster child for this guideline that the IOM's Action Collaborative on Implementation has put together, and HLAB and TPMT were used as the first examples in this. Finally, CPIC lists people who are implementing CPIC guidelines. A lot of times we get asked, who's doing clinical implementation? And that's listed under the Resources tab on the CPIC website, and these are just a few pages of screenshots, and you may recognize your name on these pages of people that are doing implementation, and they are letting other people know that they're doing implementation. Finally, at St. Jude, of course, we put a lot of work into CPIC with our colleagues at Stanford, but we're also doing preemptive implementation of pharmacogenetics locally for children at St. Jude with our long-term goal of doing preemptive testing for all patients for all CPIC guidelines. And I'd like to just acknowledge the team at PharmGKB, as well as our team at CPIC and our steering committee and advisory boards, and thank you for your attention. Thanks, Mary. Can we have time for one or possibly two clarifying questions? It was perfectly... Oh, there's a question. Mary. Oh. Thank you for that. Can you maybe help me understand how much money you think it would take to quit to finish the project from the perspective of what you said, and help me... You may have to help me understand this, because I'm not in your space, but you infer it to me that you have a lack of resource, which is not enabling you to go down to that tier three level of evidence generation. Does that make sense? And if I'm confusing you, shout out. So my understanding was you've got an amount of money, you've taken some of the low hanging fruit, but you need some more money to finish off the tier three of the lower level interactions. Are you talking about to generate the knowledge that's needed to act on these variants? Yeah. Well, so that's much wider than CPIC and... Yeah, exactly. Oh, I'm sorry. Yeah, exactly right. That was much more articulately put. Thank you. Thank you. Yeah, we probably need to have about... I don't know, Terri, what do you think, 60, 75 percent more than we're getting now? We almost double? It depends how far you want to go with that. So yeah, for the MTHFR, people telling people to use different MTX doses based on MTHFR, there's many, many tests that are being pushed to consumers by direct to consumer laboratory testing where the evidence is not there. So that, I can't really control how much that N is going to explode because it's been disappointing to see the number of gene drug pairs for which there's claim clinical actionability balloon in the last few years, whereas the number of gene drug pairs for which there's actual evidence-based actionability is very, very low. So it's hard for me to estimate if we really wanted to address every crazy claim that some companies are pushing. Address the drop 25. Give me a ballpark. I don't know. You need $2 million. You need $20 million. You need $200 million. Yeah, less than $2 million a year, that's for sure. You listening to that, Dr. Minolio? But recall that a huge amount of this is volunteer effort, that it funds infrastructure, but the people that are actually doing the evidence review and writing that, it's all volunteers. And that's one reason that it takes a while. So, Terri, can we get you to come to the microphone, please, because those on the webcast can't hear you. And then with regards to the question about Level C that you had, although there's not as much evidence for them, there is enough evidence, perhaps, that make it worthwhile for there to be expertise that evaluates that evidence and put it out. Now, it may not be at the same level of what we refer to as a CPIC guideline. It may, Mary and I have talked about this. Maybe it's a recommendation or however you want to call it. But there is information that is there that will be of value. So by cutting it, we are really only, we can only do what we can with the resources we have available. And so we are missing out on an opportunity to look at those as well as the fact that as new data becomes available, we may be ignoring it. Thank you. But it's a really good point because we're highlighting the yeses, but we're ignoring the noes. Hey, thank you. That's exactly my point. You're highlighting the yeses, but you're ignoring the noes. Thank you for clarifying what's going on in the back end of CPIC. We're going to move on now to Clint Barr and Clint Jen and Heidi Rehm who's going to speak.