false
Catalog
Unbreak My Heart: Progress Towards Personalized Ar ...
Can Polygenic Risk Scores Help Personalize Antiarr ...
Can Polygenic Risk Scores Help Personalize Antiarrhythmic Therapy? (Presenter: Christopher H. Newton-Cheh, MD, MPH)
Back to course
[Please upgrade your browser to play this video content]
Video Transcription
Okay, so the next talk is going to be performed by Dr. Christopher Kuhn, and it is going to be about risk scores and polygenic risk scores to personalize antiarrhythmic therapy. Great. Thanks. And I do consults for a few pharmaceutical companies, but I am going to be talking today about ways in which their drugs can hurt patients through proarrhythmia. So I have very little conflict in that regard. To date, genome-wide association studies of common genetic variants of sudden cardiac death as a crude and broad phenotype have identified no common variants meeting genome-wide significance. However, the study of using genome-wide association to test common genetic variation of intermediate traits or risk factors for sudden cardiac death have been far more successful. As an example, QT prolongation is a recognized risk factor both in the general population and in subsets of the population who have, for example, coronary artery disease. It's thought that at least part of this risk of sudden cardiac death is mediated by the risk for Torsade-Dupont, a quintessential arrhythmia arising from disordered repolarization, although some of the QT prolongation risk for sudden death could be an epiphenomenon. It's also recognized that over 100 currently marketed therapies are associated with proarrhythmia. This has led to the withdrawal from the market of medications like terfenidine or cisipride used to treat non-life-threatening conditions such as seasonal allergies or gastroesophageal reflux for which tolerance of even a one-in-a-million risk of sudden cardiac death is unacceptable. But there are many currently marketed therapies for which, for example, I showed just a few examples of antimicrobials or psychiatric medications for which these are off-target effects. And, in fact, the alteration and repolarization is the on-target intended effect of medications that have QT prolongation, but limited utility of these medications due to proarrhythmia has already been mentioned by prior speakers, and, in fact, that has motivated some of the efforts to develop the cellular models that have been discussed earlier. For purposes of reference, I'll just say that QT prolongation, and any time I say QT interval, I mean heart rate-corrected QT interval. BASETs is often used clinically, although we'll typically use regression-based adjustment for heart rate effects, but I don't mean to ignore the heart rate dependence. On the order of 8 to 10 millisecond QT prolongation for novel therapeutic agents certainly increases the concern about a repolarization signal. Many of the associations with sudden death are only discovered after clinical trials have led to successful registration for medications because, in many cases, the event rates are low enough that they can't be detected in clinical trials. There are well-recognized Mendelian syndromes of predominantly long QT syndrome, but even short QT syndrome, in which strong familial aggregation of either extreme long or short QT interval is associated, in the case of long QT, with sudden death from TORSAD. These are relatively highly penetrant, certainly compared to the common genetic variants I'm going to describe, and can be induced by stereotypical stimuli such as exercise or awakening from sleep with an alarm, or the use of QT-prolonging drugs. And the list of genes and proteins that are involved are steadily growing, although the top three genes underlying LQT1, 2, and 3 explain the majority of explainable congenital long QT syndrome. I'm going to highlight two common genetic variants before I talk about polygenic scores, just to try to give you a little granularity. There's a variant that was detected in sequencing studies of KCNE1 in the 1990s. It was initially unclear whether it could be a causal variant for long QT syndrome, but when its frequency of 1.5 percent was recognized, and considering the frequency in the general population of about 1 in 2,000, depending on the estimate of long QT syndrome, it was felt that this would be unlikely to be sufficient to cause long QT syndrome. But I'm going to talk about this in a general sense, and I'm going to pretend that this had not already been recognized in the 90s, but I'm going to start from a common variant focus, to say that in an unselected general population, such as 6,000 people from Finland, the minor allele of this D85N single nucleotide polymorphism was associated with about a 10 millisecond longer QT interval. And in the Framingham Heart Study, a study of mostly people of Western and Northern European descent in the Boston area recruited in Framingham was associated with an 8 millisecond longer QT interval. Those were subsequent to the work that I'm now going to describe, in which this variant was in fact found to be enriched in cases of drug-induced arrhythmia, work that Al George and Dan Roden worked on in the 90s. A small study of drug-induced arrhythmia in Europe demonstrated a slight excess of the D85N allele. And as well, in those who were negative for mutations in LQT1, 2, or 3, it was also found to be enriched on the order of 8 to 10 percent compared to controls that had roughly 2 percent. This has also been supported in an East Asian population from Japan. In Japan, in controls, the frequency of this variant is about 1 percent, and the prevalence of those with mutation-negative long QT syndrome was about 7.5 percent. In a subsequent study in which some of the prior TORSAD cases were examined that included up to 176 patients with TORSAD and 207 controls, controls who were selected both to have one set of controls selected because they had relatively reduced QTC prolongation in response to exposure to a class 3 agent or also comparison to ancestry-comparable population controls. In this study, there were 1,400 single nucleotide polymorphisms in 18 candidate genes that were examined, and the top result was KCNE1, D85N. With an odds ratio that was higher in this initial sample, a smaller attempted replication study showed an odds ratio that was on the same side of 1, but possibly due to smaller sample size, was not supported. And so in total, I think we could say that there is a common polymorphism in a gene that's well known to have rare variants that cause congenital long QT syndrome and that is probably a modifier of long QT syndrome risk, although the underlying basis of that risk is not always known, both in a drug-exposed TORSAD ascertainment as well as those ascertained on congenital long QT syndrome. Genome-wide association studies, so that variant was well recognized because that gene had been amply sequenced. Common variant genome-wide association studies really only had the potential to be used in the mid-2000s when large-scale common variant genotyping platforms were developed. In one of the earliest studies to successfully use that design, a common non-coding variant in the region of nitric oxide synthase one adapter protein was found to be associated with QT interval. This is, in fact, a genome-wide association study that simply sampled 200 women from or 100 from the low end of the QT distribution and one from the 100 from the high end of the distribution with successive ways of replication that confirmed the association. And in fact, this common variant ascertained here originally through the association with QT variability in the general population was also found to be enriched and was associated with a roughly 30 percent increased odds of sudden death in studies from the ERIC, CHS, and Rotterdam studies collectively. Thirty percent may seem like a lot. I'll say that for Mendelian variants, the odds ratios are on the order of 10 to 100 fold increased odds. And if one considers that the general population risk in middle-aged adults is on the order of 0.1 percent per year, an increment in risk of 30 percent means it's 0.13 percent as opposed to 0.10 percent. So you can appreciate the absolute increment in risk from such a variant is not clinically useful. In fact, these variants at NOS1AP have also been found to be related to Torsade de Point in individuals ascertained in a study published by Yalda Jamshidi. And the NOS1AP variants have also been associated both with QTC among those who have Long QT syndrome, including South African Long QT syndrome 1 founder population that has been very well studied, as well as outbred collections from across Europe in which association with cardiac events being syncope or cardiac arrest. And so, in total, I would say these are two variants. One that has a frequency of about 1 percent, the second variant, the NOS1AP variant, has a frequency of about 20 percent in the general population, and these are associated with respectively a 10 millisecond and a 3 millisecond longer QT interval per allele. So these are weak effects, typical congenital Long QT syndrome variants. If you look within families at mutation carriers versus non-carriers, associated with about a 20 to 50 millisecond difference when they're bona fide Long QT syndrome variants. So we've done now larger and larger scale genome-wide association studies. In total, the published literature includes over 60 independent variants at 35 loci that explain collectively about 10 percent of variability in QT interval. About 40 percent of variability in QT interval appears to be due to inherited genetic factors. This is about a quarter of the heritability that's explained by these variants so far. Most of the studies that have been well-powered have included people of European descent, although larger and larger collections of those with of African descent have been examined. And while we can't implicate specifically each of those 68 variants that were proven in people of European descent, we certainly see a trend of effect and a correlation with a correlation coefficient of 0.6 for the European allele effect in those of African descent. We did look to see if any of the genes at some of our novel loci could explain genetically elusive Long QT syndrome by looking at about 300 probands ascertained from collaborators in Amsterdam, London, Mayo Clinic, Nantes, Pavia, and Toronto. And to cut to the chase, we could not specifically implicate coding variants or loss-of-function variants because we couldn't demonstrate segregation and we were not able to prove that any of these genes is implicated. We have built what are called polygenic risk scores. An example of such a score using 14 SNPs from one of our earlier genome-wide association studies was tested in the Health 2000 finish collection, and demonstrated a 15-millisecond difference in the top quintile compared to the bottom quintile when aggregating the per-allele effects across all 14 of these variants. That study used an early GWAS that explained about 8.5 percent of variants. As mentioned, we now are up to about 10 percent variants explained. And these have a continuous and graded effect on QT lengthening, here shown on the millisecond scale on the y-axis, and the red bar is showing the odds ratio for having a prolonged QT, C being greater than 450 milliseconds in men and 470 in women, achieving a threefold increased odds relative to the bottom quintile for those in the top 20th percentile for a QT score. And as I already mentioned, the slightly greater estimate of the between quintile difference in another study. So if variants detected in QT genome-wide association study are collectively associated with resting QT interval, and with clinically defined QT prolongation in general population, what about drug-induced QT response, or drug-induced TORSAD, or sudden cardiac death? Well, so the polygenic risk score hypothesis is that if long QT syndrome mutations are strong risk factors for drug-induced TORSAD, I think this is well accepted, and two SNPs with relatively large effects, not as large as LQTS mutations, increase the risk for TORSAD, then common variant genotypes that prolong QT interval might collectively lead to exaggerated QT response to medications. And in a collaborative study with David Strauss, who's in the research arm at the Food and Drug Administration, they've had a big interest in trying to identify either individual or drug factors that contribute to risk of sudden death, 17 European-descent individuals, four African-descent individuals, and one Asian-descent individual were studied. This is not a study that was designed to look at genetic questions. And because of power, I'll focus your attention on the 17 European-derived individuals. These people were admitted to a CRO, and with a seven-day washout in between, were administered a single dose of 500 micrograms of tefetalide, or quinidine, or ranolazine, or placebo. So they had seven days of washout in between, and then they had, over the span of 24 hours in which they were resident in the CRO, they had 15 electrocardiograms and simultaneous plasma samples to measure achieved drug levels. And we used as a phenotype the QT interval adjusted using the Fridericia correction, adjusted for the baseline within individual QT, and it adjusted for the placebo effect to account for diurnal variation, and for the achieved plasma drug concentration. So it's about as refined, if you consider these genome-wide association studies, which take all comers with a variety of conditions, this is about as clean a design as you can get. And we tested the score. We were able to get 61 SNPs genotyped from among 68 based on our prior work, and I'm going to not mention further the African-derived sample, which was underpowered. This shows basically what the phenotype is, again, as a reminder. These were monitored over 24 hours. Here they received defedolide. This is their defedolide plasma concentration. Here is shown over those 24 hours the change in the QTC compared to baseline, and then the phenotype that we measured was the change in QTC compared to placebo compared to baseline relative to their achieved defedolide concentration. And here is just an example of a high responder and a low responder. These are, again, randomly selected individuals, not specifically selected because of any recognized risk. And if you look at the QT genetic score, when tested against the baseline QT interval, not surprisingly there was a weak association. This is 17 people. Our prior genome-wide association study included 70,000 people, so not surprising that you would see some confirmation. But I think importantly to us, the genetic QT score was, again, modest in terms of statistical significance with just 17 individuals, predictor of defedolide QTC response, as well as quinidine QTC response, as well as ranolazine. And this last panel is just to show that, in fact, the defedolide and quinidine responses within individuals were correlated, which would not be surprising if there's an inherent repolarization phenotype. So genetic score predicted the QT response in whites. In African-Americans, four individuals is not enough to draw any conclusions. And it explained roughly about a quarter of the variability in QT response. QT response to different drugs is correlated, as mentioned. And then we went on to dip into a TORSAD case control collection that is now a superset of that earlier study. And we found, again, that the higher QT score predicts higher TORSAD risk, appearing to explain about 12 percent of variants in TORSAD risk. This is a score that includes many variants, some at the same locus. If we restricted to just one variant per locus, it was still a significant predictor. And skip that. So we're now doing tests in healthy volunteers who are administered moxifloxacin. We have a multi-center consortium, including collaborators at Mayo Clinic, Cleveland Clinic, University of Colorado, and Mass General, approaching patients who are started on defetalide or Sotilol and who are monitored for routine ECG monitoring. So far, we've recruited 550 patients, planning to get 1,000 in whom we'll test our genetic score. There's an ongoing updated GWAS in which we have up to 240,000 individuals. It's multi-ethnic, imputed to lower allele frequencies. Will Young and Najeem, not Smith, Najeem LaRouche, are the analysts, and Patricia Monroe, Nana Sodadene, and myself are leading that study. So our goal is to try to identify individuals at risk based on their genetic profile, with the hope that maybe novel drugs could come to market more safely. And just to conclude, I would say intermediate myocardial phenotypes offer well-powered quantitative readouts of fundamental biological processes, and may expose liability to pathologic responses to the environment medications, and I'm happy to answer questions. Thank you. So we have time for questions. Chris, that was very nice. Where are we now with thinking about weighting the individual SNPs in a polygenic risk score, and is that something that should be done, or is there a statistical approach that you depend on to guide that? Yeah. In fact, we take the weights for each of the variants from the population level impact on the QT. So, for example, the D85N allele, it wasn't present in the 17 individuals in the FDA study, but that gets a weight of 9 milliseconds. The NOS1AP variant gets a weight of 3.5 milliseconds for each copy of the QT lengthening allele. So we do take into account the strength of the effect on QT interval from population data. Yes. Chris, very nice. So I might have missed this, but do you take into account drug concentration, plasma concentration when you're using these risk scores? Because it seems to me that depending upon the drug and the metabolic pathway, variants not only in that defined repolarization reserve, but also drug absorption, drug metabolism are going to be critical to try and better understand who's at risk here. Yeah. You're absolutely right. We are doing an ADME panel in our sotolol-defetolide study. Now, the relationship of genetic variation to drug effect is not so well known. In that FDA study, achieved plasma concentration was measured. But in our clinical study, while we are getting plasma samples, we're not timing them adequately to be able to get robust measures of achieved blood levels. So I certainly would acknowledge that there are many different factors that contribute to the repolarization effect of a drug in an individual patient. Because drug repolarization effects exist across a variety of medications that have very different targets and have different metabolism, it'll really need to be drug-specific accounting for the ADME effects. And so here we're looking for on-target side effects that presumably a lot of it is due to the effects on ion channels that we already know about. But as you point out, there are likely other patient factors that contribute to absorption and metabolism and availability in the heart. Yes. Nice presentation. At best, only 10 percent of the variability in QT interval is explained by genetics. What's your explanation for why we've not been able to find the remaining variability? Well, I would say that anybody — I'm guessing many in the room put people on a treadmill and try to measure their QT response to exercise or can recognize the fact that PVCs and other things make it difficult to recognize those effects. I don't think — 10 percent is not terrible. Forty percent of the heritability — or 40 percent of variability in QTC is heritable. Clearly sex, LVH, use of diuretics, achieved potassium concentration, all of those factors contribute to variability. So I would say I think we'll probably get to about 20 percent, 20, 25 percent with common variants. I'm guessing that rarer and rarer variants that may be individually numerous across seven billion individuals but within a finite set will also be contributors. Yes. Thank you, Chris. We have devoted eight years to three of your SNPs in ubiquitin ligases protein, which is a lot of fun. So in our rabbit models, we think that there are two mechanisms of arrhythmia. One is the cellular scale of EADs induced in cells, but a very important emerging model for arrhythmia is dispersion-induced VPCs or dispersion-induced triggers. So when you're looking at QT in humans, it doesn't tell you much about dispersion. Yes, I am aware of the clinical. So I'm wondering, in terms of drugs, some drugs induce more dispersion than others, and it looks like in your studies that there is a straight correlation between the different drugs that you are using. So how would you account for mechanistic insight into the drug-induced arrhythmia? Well, I would say this polygenic score approach is a completely Luddite approach. It ignores all of the individual mechanisms. In fact, it also ignores that you don't have to have the smoking gun variant. It could just be correlated with the functional variant. So I think you're right that not all QT prolongation is the same. We have used QT interval in the general population because it's routinely measured and it's available in hundreds of thousands of individuals. But I don't mean to suggest that the QT interval by itself is a sufficiently potent way to capture the risk of repolarization. So dispersion is another approach to try to examine this. Unfortunately, there are just not scores of thousands of individuals who've had dispersion measured for genome-wide association. To that end, if you test amiodarone, which does prolong the QT interval, how would you skew it? Would it skew your results? Well, I'm assuming that we are testing a mixture of genetic variants that influence the QT interval and have no role in increasing risk of arrhythmia, and variants that increase the QT interval and increase the risk of arrhythmia. Now, it will take very large sample sizes to be able to establish which of these individual variants contributes to arrhythmia risk. I think, you know, ranolazine compared to sodalol, if you were to give equipotent doses to get the same QT effect, we would recognize that there would be a different influence on risk. So I think the point is well made. We're currently not in a good position to be able to distinguish which of these QT lengthening effects and which of their specific mechanisms are influenced.
Video Summary
Dr. Christopher Kuhn discusses the use of risk scores and polygenic risk scores to personalize antiarrhythmic therapy. He explains that genome-wide association studies have been successful in identifying common genetic variants related to intermediate traits or risk factors for sudden cardiac death, such as QT prolongation. Dr. Kuhn highlights two common genetic variants that have been studied: a variant in KCNE1 and a variant in the region of nitric oxide synthase one adapter protein (NOS1AP). These variants have been associated with a longer QT interval, and in some cases, an increased risk of drug-induced arrhythmia or Torsade de Pointes. Dr. Kuhn then discusses the development of polygenic risk scores, which aggregate the effects of multiple genetic variants to predict an individual's risk of drug-induced Torsade de Pointes. He presents findings from a study that demonstrated the predictive ability of a polygenic risk score using 61 genetic variants and suggests that these scores could be used to identify individuals at risk and guide the development of safer medications.
Meta Tag
Lecture ID
15811
Location
Room 213
Presenter
Christopher H. Newton-Cheh, MD, MPH
Role
Invited Speaker
Session Date and Time
May 10, 2019 10:30 AM - 12:00 PM
Session Number
S-053
Keywords
risk scores
polygenic risk scores
personalize antiarrhythmic therapy
genome-wide association studies
genetic variants
Torsade de Pointes
Heart Rhythm Society
1325 G Street NW, Suite 500
Washington, DC 20005
P: 202-464-3400 F: 202-464-3401
E: questions@heartrhythm365.org
© Heart Rhythm Society
Privacy Policy
|
Cookie Declaration
|
Linking Policy
|
Patient Education Disclaimer
|
State Nonprofit Disclosures
|
FAQ
×
Please select your language
1
English