By Turna Ray
Vanderbilt University earlier this month launched the Vanderbilt Electronic Systems for Pharmacogenomic Assessment, or VESPA, project concurrent with the opening of its DNA biobank BioVu.
VESPA holds the distinction of being the PGx research project to receive the most funding from the National Institutes of Health under the American Recovery and Reinvestment Act of 2009, according to analysis by Pharmacogenomics Reporter [see PGx Reporter 11-11-2009]. With more than $5 million in federal funds, VESPA aims to analyze DNA samples from the BioVu database and electronic medical records to investigate the genetic underpinning for disease and drug response.
BioVu currently holds 75,000 DNA samples, but researchers are hoping to grow the biobank to 250,000 samples in the next three years. VU's database of electronic medical records currently includes information on 1.9 million people.
According to Dan Roden, assistant vice chancellor for Personalized Medicine at VU, the aim of the VESPA project is to identify clinically validated gene variants linked to drug response and generate a list of such important variants that will aid healthcare providers recognize pharmacogenomics opportunities in the care of their patients.
"The only way you're going to implement something" like VESPA in a routine clinical setting is "in an electronic healthcare environment," Roden told Pharmacogenomics Reporter this week, while discussing the project.
Below is an edited transcript of the interview.
Can you provide some background on how Vanderbilt University came to launch the VESPA project?
For the last five years we've been building a DNA biobank [called BioVu] that links DNA obtained from patients who come to our medical center with de-identified medical records. The goal is to create a resource that will do two things. One, it will allow investigators to do discovery of new relationships between genetic variations and important human conditions, such as susceptibility to disease or variability and response to drugs. We've been collecting samples since the spring of 2007. As of [last] week we have samples from more than 75,000 patients. Arguably, this is the biggest resource in the country.
The second thing we want is for BioVu to act like a giant clinical laboratory. The idea is that we all see the increasing robust science of genetic or genomic variation as ultimately coming to the bedside ... It's a pretty commonly held vision that at some point in the future a doctor will write a prescription and the electronic health system will say, 'That's the wrong drug,' or 'That's the wrong dose of the drug for that patient,' or 'This particular patient doesn't have the disease that you think they have because of genetic variations of some type.' While people talk about that kind of vision, actually executing it presents a lot of practical problems, such as which genetic variants would you actually want to act on? What would be the strength of evidence? How expensive is it to do this? What information technology challenges are there? How do you store huge amounts of genetic data on huge numbers of patients, and access it rapidly?
So, BioVu is designed both as a tool for scientific discovery, as well as a tool to help [answer] those kinds of questions. I say it's a tool for discovery, and the discovery is not just in new genetic variations, but in actually how to execute that kind of vision. As we were building BioVu, we conducted a demonstration project, in which we tried to, before investigators started to use it, prove to ourselves that the things that are there, that ought to be discovered, or ought to be present, are detectable. For instance, there are many studies that implicate variations in certain genes as susceptibility variations for type 2 diabetes. So, the demonstration project that we executed asked the questions, 'Can we in our database identify patients with type 2 diabetes? Can we identify patients who do not have type 2 diabetes? And genotype all those patients and demonstrate that in fact, the ones with type 2 diabetes do have a higher incidence based on susceptibility alleles than do controls, just like the literature says we should?'
In the demonstration project we looked at around two dozen genetic variants that have been associated over and over with susceptibility to five different diseases. Each one of those replicated what we thought it should replicate. We are very comfortable with the idea that BioVu can do what it is supposed to do ... So then VESPA can be viewed as our effort to extend that demonstration project logic to the area of drug responses ... All the information we use is mined from electronic medical records, so to develop automated tools that are good enough to find those cases and controls is a challenge. And it's an important challenge to meet in order to execute the long-term vision of getting this information into the medical records, so physicians can begin to use this information to make better decisions about drugs.
The first phase of VESPA will be similar to the demonstration project, in trying to find things in the data bank that we expect to find. The second phase will try to look at other kinds of abnormal drug responses and discover new genetic variants that determine variability in drug responses. The long-term vision is to find those [gene variants] that are robust, that are clinically important, so eventually we will generate a list of genetic variants that are important for physicians to know about or electronic health delivery systems to know about, so that drugs can be chosen more rationally.
[ pagebreak ]
For how long will this project run, and when are your most immediate deliverables due?
VESPA is a project funded through the stimulus grant project. The project started in the fall and it runs for two years. The first phase really involves dense genotyping in a large number of subjects chosen because they have been exposed to many, many drugs. So, this will create a database, which will then be mined in the second year. The way the stimulus grants were written was that you had to say what you wanted to do, and once you got the money you actually had to come up with a plan of doing that. We're in the process now of selecting the platform that we will use for genotyping and the specific patients in which the genotyping will be done.
It sounds like we're starting from scratch. Of course, we're not. We're in the middle of negotiating ... with Illumina and Affymetrix ... as to which platform is best for us and will allows us to genotype the most number of patients. That process is just about done.
In the first three months of the project, which are ending just about now, we have to decide which of our 75,000 samples we want to genotype. We aren't going to genotype them all because the money won't allow that. The funds will probably allow us to get GWAS-level data somewhere between 10,000 and 20,000 samples. We may get supplementary funds from our own institution.
The next step is to move forward with the genotyping and in parallel the development of the processing algorithms. By the end of the first year we hope to have all the genotyping completed. We might overflow a little bit into the second year. But we want to build the genomic database in a big hurry. And once that database is built, we're not going to just sit back and relax. Once we have dense genetic data on lots of patients that is a resource that will last for many years. The mining doesn't have to be completed in the second year. But for the ARRA project we do want to have some deliverables in terms of publishing manuscripts by the end of the second year.
At the end of it, is the main aim of this project to try to give a picture of the type of resources that would be necessary for this type of genetic assessment to be done on a routine clinical basis?
Yes. One way to look at it is we'll develop a list of genetic variants that are [linked to various diseases and drug response]. For instance, there is a well-recognized story around genetic variability in response to [the anti-platelet agent] Plavix, which is a drug many people are aware of since it's advertised on TV every night. There are one or two genetic variants that play an important role in determining whether you're going to be a Plavix responder or a non-responder. The state of the art right now is to say, 'Here are the patients I'm starting on Plavix. Let me do a genetic test and then decide whether this is the right drug for them or they are getting the right dose.' With the cost of genotyping plummeting and our ability to get very dense genetic information for really the same amount of money it takes to get a single genetic test right now, the vision has to be that you'll deposit your genetic information into your record before you need Plavix. And then when the Plavix prescription is written, you don't have to have a separate test; you don't have to wait around a couple of days to see what the results of the test are. If you're a physician you don't have to contact the patient [again] after three days to tell them that their genetic test shows they should be on a different dose. The idea is that the electronic health system will look up what you have and say, 'This is the right dose for this person.'
If we're going to get to that vision, and that's the one we really want to get to, we have to start somewhere. We're not going to start with the entire genome. So, we're going to start with a list of important genetic variants. I don't think I have to do VESPA to do that list today. That [being] said, how are you going to decide the next genetic variant that comes along? The decision will be based on the strength of the data, how important the adverse effect [is], and how reproducible the effect [is that] you think you're looking for. So what VESPA will provide to us is the list that we are sure in our dataset we can replicate and provide a framework for how to look for anything else that might come down the pipe … It's important to create that initial list ... using criteria that we can apply rationally and prospectively to anything else, so VESPA will allow us to calibrate how big a signal do you get, how important the genetic signal is, and those kinds of things.
So you have a plan for the gene variants, but do you have a priority list for the drugs you will look at?
The first two drugs we're going to look at, so we can confirm we can find what we intend to, are Plavix in stent thrombosis and the anticoagulant warfarin. There is pretty clear data that says there is genetically determined variability in steady-state warfarin dosage. There are a lot of other issues around warfarin, but it is clear if you look at patients who are for long-term therapy, and look at their steady-state dosage, their genetic variants will determine that [dose]. We're in the process of identifying those patients right now. We're not waiting around for the Illumina or Affy platform decision. The algorithms to identify those patients and their dosages and their responses are in development right now.
Then there are a list of 20 or so reported relationships between drug responses and genetic variants that we're going to look at. We're going to look at as many as we can in the allotted time. Then there is another list of 20 or so variable drug responses where there is variability that is not well understood and may well have a genetic [component]. So, that's why the GWAS set will be particularly important for that.
Where do you see [a system like] this being most immediately implemented? Would pharmacy services providers, such as Medco and CVS Caremark, be able to implement a system like this right now in providing PGx testing to customers within their systems?
The only way you're going to implement something like this in an electronic healthcare environment. We're never going to get to a vision of personalized medicine that includes genomic variation without electronic health records, because the alphabet soup of genetic variants is way too big for a human being to handle.
Additionally, both Medco and CVS Caremark are actively engaged in looking at genetic variation, and I think what they're doing may be the way this field really gets prodded forward. The reason I say that is that there are two examples I can think of, both relevant to what we're doing in VESPA, one around Plavix and one around warfarin, where there are new drugs that are about to supplant those drugs. Plavix is about to come off patent and warfarin has been off patent for a long time. So, they are both cheap drugs that work in most patients. But in patients who have genetic variations, there is a risk [of adverse events.] So, one thought that many people have had is to do a genetic test on someone who needs, say, Plavix, to see if he has a genetic variation that makes him an aberrant responder. If he does, then he gets the new, expensive drug. If he is programmed to be a normal responder, let's give him the inexpensive, old drug. That's a simple way of looking at it, because there are other differences between new and old drugs.
The same is true for warfarin. There are new drugs coming along that look to be, depending on how you look at it, a little bit worse or a little bit better than warfarin. But if you're genetically predestined to have a bad response to warfarin, then the new drug would be the right thing to do. If you're someone who will have a pretty ordinary and okay response to warfarin, maybe the right thing to do, if you're a pharmacy benefits manager, is to take the old drug.
That's one of the reasons [the pharmacy services firms] are very interested in looking at genetics. We're of course interested for exactly the same kinds of reasons. It's about making sure people get the right drug, and that is ultimately related to cost. If you get the wrong drug, and have a bleed into your head, that's a pretty expensive event in terms of dollars and lives.
So will you do cost effectiveness analysis as part of VESPA?
We're going to try to work cost into that, although that's very difficult. Cost [analysis] is not part of the VESPA grant itself. We thought about that, and we felt it was way too much to bite off. That said, VESPA is not the only thing we do in personalized medicine at Vanderbilt. We have a big program and the institution has a big commitment to moving forward an agenda in personalized medicine. So, the agenda ultimately includes a look at economics and outcomes, because if you develop systems to deliver better electronically driven healthcare, you ought to be able to look in those systems and say, 'Aha! We've saved lives!' or 'We've saved money.' So, that's part of the bigger program, but it can't be part of VESPA, because VESPA is a pretty big project as it is and there are limits to what the NIH will fund for us.