Promise and Perils of Pharmacogenomics
Robert A. Harrington, MD: Hi. I am Bob Harrington from Stanford University, and I am here at the American Heart Association (AHA) Scientific Sessions in Dallas. I am going to take the opportunity to talk to a few colleagues about some of the work we have seen this week on genotype-guided therapy, as well as the promise and perils of relying on pharmacogenomics to make therapeutic decisions. I am joined today by Jessica Mega from Harvard, Magnus Ohman from Duke, and Jennifer Robinson from Iowa. Let's start with you, Jessica. You have done a fair amount of work in the pharmacogenomic space with antiplatelet therapy. Let's start with the foundation. What is the concept of pharmacogenomics?
Jessica L. Mega, MD, MPH: It's the idea that we give many of our patients the same dose of a therapy, and yet we have different enzymes in our bodies, such as the cytochrome P450 (CYP) enzymes. We have genes that encode variants and process these enzymes, and your genetic makeup may be different from mine; in fact, we know that the genetic makeup of many people is different. Let's take clopidogrel, for example. About 10 years ago, we noticed that some of the pharmacodynamic effects of clopidogrel were different between patients. There are many reasons, which we will talk about today, why people may respond differently to drugs, and one of them is genetics. In the case of clopidogrel, there are certain genetic variants in these CYP2C19 genes that metabolize it.
The gene itself doesn't matter so much except that it makes a protein that processes the drug. If you have the variant and I don't, then you process, on average, 30% less of the drug than I do. So a 75-mg dose of a drug in me would be like a 50-mg dose of the drug in you. In certain clinical situations that may not matter, but in other situations it might, particularly after a stent is placed; those are periods of time when you need the highest degree of platelet inhibition. We have seen a 50% increased risk for cardiovascular death and minor stroke for people with this genotype and a 3-fold increased risk for stent thrombosis.[1]
Dr. Harrington: We haven't got the data yet on what happens if you measure my genotype and then you dose me differently?
Dr. Mega: The idea of being able to take a genotype and then titrate therapy has not been shown yet, but I would say that we dose medications according to all sorts of parameters. We take into account clinical factors such as renal creatinine clearance, and we don't have trials dedicated only to patients with certain levels of creatinine clearance. We know that they process drugs in a different way. Their pharmacokinetics are different, and we use those data. Those are the types of data, for example, that the US Food and Drug Administration (FDA) was facing.
COAG and EU-PACT: Genotype-Guided Warfarin Dosing
Dr. Harrington: Magnus, now that we have the foundation established here at AHA this week, we have seen late-breaking clinical trials, one from the United Kingdom and Sweden (EU-PACT),[2] and one from the United States (COAG),[3] that tested this concept of using somebody's genotype to dose a commonly used drug -- in this case, warfarin.
E. Magnus Ohman, MD: Both of these trials are incredibly important for different reasons. We have known for a long time about the genetic profile (it's actually 3 different genes) that describes how warfarin is metabolized and brought to the active compound. In a way, what we have is that when you start a therapy, we are doing a blind assessment essentially, and these studies set out to test whether knowing the genotype would enable us to do a better job of starting the therapy.
Ultimately, because warfarin is a medication that is monitored continuously, we all know that by 3 months the patient should be at the right level, because adjusting the dose will overcome any issues with metabolism. Both of these trials addressed this. The larger US trial[3] looked at it over a longer period of time and identified that this knowledge of genetics and the structured approach for how you got there did not make a difference. That was perhaps a bit surprising, but in some ways it was not all that surprising because physicians are pretty smart sometimes and they adjust for these things. The other study was the UK study,[2] which was much more pragmatic. It said, "We'll give you a lot of information and let you be on your own." In that trial they found that knowing the genetic profile ahead of time improved time in the therapeutic range a little bit but not by a large amount. Because this is only in the early phase, I do not believe that an outcomes trial will be helpful, because in a way the delta is so small.
Beware of Genetic Exceptionalism
Dr. Harrington: The observation in the American trial gets to the question that I want you to start thinking about as an epidemiologist. The American trial provided as much information in terms of the dosing as the clinically based algorithm plus the genetics. From your world of epidemiology and creating risk scores, how do you interpret this?
Jennifer G. Robinson, MD: It's exactly right. If you have a good risk score that already explains 80%-85% of what is going to happen, it is really hard to improve on that with a little bit more information from genetics (unless it's a highly penetrant allele that makes an odds ratio of 9 or 10 for harm). The investigators with the clopidogrel genes are to be congratulated because they truly are doing a biomarker-based test -- ie, does management based on a biomarker make a difference? That's the standard that we need to encourage people to look for moving forward, because it turns out that genetics may not add as much information as we thought it would originally.
I'll give you an example just to generalize it to other fields. In the cholesterol field we have simvastatin. There is a nice study from the SEARCH investigators[4] looking at a particular allele for organic ion transporter that showed that if you were a homozygote, you had about a 30% chance of getting serious myopathy within the first 3 months of starting therapy. If you were a heterozygote, the risk was just a little bit throughout the course of the study. If you were doing this even if the patient were a homozygote, two thirds didn't have any adverse outcomes, and if you chose not to treat them or not to use the correct therapy on that basis, you could potentially be denying them a drug that prevents heart attacks, strokes, and reduces death. The implication is that you have to think through what happens when you make decisions based on whether a biomarker is present.
Dr. Harrington: You are raising a really important point that I want to get your insight on, Jessica: this notion that knowledge of the genome is going to solve everything. Maybe ultimately it will, but certainly not now. It provides us a lot of interesting biology, interesting hypotheses, and ways to take our clinical trials, but for the clinician who is listening to this, they should not be thinking, "Oh my goodness, if I am not genotyping my patients for these specific things in cholesterol treatment, in anticoagulant treatment, antiplatelet treatment, I'm doing a bad job." We don't want them to think that way at all.
Dr. Mega: It's a great point, and the phrase I typically use is "genetic exceptionalism." There's a lot of excitement about genetics.
Dr. Harrington: It's a good phrase.
Dr. Mega: When we were in school we were taught about genetics and about diseases like Marfan syndrome . You have a genetic variant, and most likely you are going to have a physical, phenotypic manifestation of this disease. Complex disease genetics are entirely different. We are looking at odds ratios of 1.15, 1.5 -- incremental probabilities. So, although people (including myself) are very excited about genetics, it can tell us an amazing amount about the underlying pathobiology. The great example is understanding PCSK9. Patients who had this particular variant had lower cholesterol levels. Now we are in phase 3 trials looking at PCSK9 inhibitors.
Dr. Harrington: The first of the genomic targets being identified has been turned into a drug.
Dr. Mega: There are reasons why we may be detecting and doing genetic studies, but we are not talking about risk variants that are telling us the entire picture. I see these genetic variants very much in the same way that we integrate all other clinical data and all other data that we get from the lab currently.
Dr. Harrington: It is just a piece of the total picture.
Dr. Mega: It is a piece of the picture. Let's keep it simple. Let's understand what this really can offer us.
The Murky Genetics of Platelet Function
Dr. Harrington: Magnus, you just did a clinical trial, TRILOGY, that had embedded into it a very interesting substudy dealing with platelet function, and the results perhaps were not as clear as all hoped they would be. Do you want to talk about that because that relates to this story?
Dr. Ohman: I would love to talk more about it. Unfortunately we are right in the middle of working through the genetic component of this. I will deal with the phenotype expression of this. And as you pointed out, we looked at this and we looked at the effect of PRU with 2 drugs, clopidogrel and prasugrel.
Dr. Harrington: Tell people what PRUs are.
Dr. Ohman: PRUs are platelet reactivity units from VerifyNow, which is the company that made the testing device. Basically, high values are bad and indicate little effect of the drug in the case of clopidogrel. A low value is good (ie, the platelets are inhibited). Generally speaking, over time and over many studies, prasugrel has been shown to produce a much lower PRU value than clopidogrel, so one thought was that this would translate into greater clinical benefit. We checked this with the platelet function substudy published by Paul Gurbel in JAMA[5] last year, and we found that clopidogrel and prasugrel behaved just as we suspected. Prasugrel was much lower compared with clopidogrel, but overall when we looked at the outcomes for patients in the trial, there was no significant difference.
Then we went back and asked, "How could that be?" We have a pharmacologic effect that we can measure with the phenotype but this doesn't show up in the outcomes? When we adjusted the outcomes based on the characteristics of the patients and identified the effect that we saw, the difference between PRUs essentially went away, so they are, by and large, driven by other things. Some of that could be genetic (hint hint). We are still working on this, but the reality is that it's very interesting.
Bringing Genetics to the Point of Care
Dr. Ohman: I wanted to say one other thing, however, that was interesting to me. And that is that we have to separate efficacy and safety, because there are some conditions where actually knowing some genetic information is very important. Long QT syndrome is an example of when we would say that having this knowledge would help, because there are some drugs that could potentially be fatal.
Dr. Harrington: Yes, there are some drugs that you should avoid.
Dr. Ohman: Exactly -- that is very important. I see a great opening in the field of electronic medical records (EMRs) that we are all struggling with right now, but the reality is that when that is placed up there in the knowledge base of that patient, that becomes a very powerful tool.
Dr. Harrington: I am going to close, Jennifer, with a practical question for you. What do we want to let the average doctor know is coming? Are we going to be able to have models at the point of care which will incorporate all of this information from the genome?
Dr. Robinson: The EMR is not going to come up with a picture of the patient and all of their risks. It's going to be different. There is this thing called the environment, and you start with the genes. The environment almost entirely determines (with a few rare exceptions) which genes are expressed and how they are expressed. As we understand in a more complex fashion -- "systems biology" is the term we are using these days for how those interactions happen -- at some point we can actually make sophisticated models for at least talking to people about ranking their risk and saying, "You have an above average risk for something, so let's do this."
And we can also start looking and seeing whether certain therapies work better based on this sort of substrate defined by phenotype and genotype and move forward. That is a goal that we are working toward; it is not quite ready for prime time. For prediction of heart attack and stroke, we do an awfully good job with age, sex, smoking, and blood pressure. Knowing more doesn't really change what the prediction is. Hopefully someday we could make it 100% right, because you would like to be able to say to somebody with a 5% new pooled cohort equation whose risk is 100% that "5 out of 100 people like you might have a heart attack or stroke, but actually you are one of those 5." That's the holy grail of what we're after.
Dr. Harrington: I recently reread the Lancet[6] paper on the global priorities set out by the United Nations Summit on the noncommunicable disease burden in the world, and it's interesting that genotyping was not on the list. It was stop smoking, reduce salt consumption, increase physical activity, reduce dietary fats, etc. I'm saying it somewhat tongue-in-cheek because I absolutely agree with you. I'm excited about genetics. I'm excited about proteomics and metabolomics. But let's keep it in context.
Dr. Magnus: It's important also to point out that we have today a very modifiable phenotype of high blood pressure, cholesterol levels, and so forth.
Dr. Harrington: Smoking.
Dr. Magnus: Genotyping might be really important in the areas where we don't have any therapies.
Dr. Robinson: If you don't have a good phenotype to treat, then absolutely.
Dr. Harrington: There is definitely more to come. I want to thank all of you for joining us and having this interesting discussion on genomics and pharmacogenomics and where the world might be going.
© 2014 WebMD, LLC
Cite this: Robert A. Harrington, Jessica L. Mega, E. Magnus Ohman, et. al. Genotype-Guided Drug Therapy in Cardiology - Medscape - Mar 06, 2014.
Comments