false
Catalog
Hypertrophic Cardiomyopathy: EP Considerations
Using Risk Stratification Algorithms to Predict SC ...
Using Risk Stratification Algorithms to Predict SCD in HCM (Presenter: Perry Elliott,)
Back to course
[Please upgrade your browser to play this video content]
Video Transcription
So, Dr. Maron, dear colleagues, thank you very much, and thank you to HRS for this invitation to come and talk today. Okay, so here we are. Here's the beast, hypertrophic cardiomyopathy, a disease with a well-deserved reputation for being a killer of young people, not difficult to understand when you see severe hypertrophy disarray, small vessel disease. And yet, despite this very powerful substrate, we now know that the incidence of ventricular arrhythmia is actually very low. What you see here on this slide is the decline in annual sudden death rates from the 60s, 70s, 80s, to the modern era, and some of this almost certainly reflects the impact of therapy, but it also probably is influenced by the fact that we're seeing a different demographic. So you go back to some of these earlier studies, and they were largely young, highly symptomatic patients, often with very severe obstruction, but in the modern era, as a consequence of screening and modern imaging technologies, we're starting to see milder disease, and we're also seeing cohorts which are much larger and perhaps more representative of the true spectrum of the disease. As a consequence, sudden death rates now, or I should say ventricular arrhythmia rates, are less than 1%, 0.8% or less. So the first take-home message is that sudden death is a rare event in this disease, but of course, that is exactly why we need methods for identifying that small group of patients who are at risk. If you have a condition where you have an event rate of 90%, it's not very hard. When you're dealing with something where you have a very low event rate, what you're trying to do is to identify that high-risk cohort. So this situation is pretty easy, isn't it? So someone has declared themselves to be at high risk, and if they survive this, they will have a much increased risk of having further events, but that's not really the group we're after. We're after those who've not had such an event. Now if you summarise the world literature going back decades, this is a very short list, but a whole variety of aspects of this disease have been associated with an increased risk of dying suddenly. For almost two decades or more now, we've taken from this list a small number of characteristics of the disease that are relatively easy to assess in the clinic, and which we believe are most strongly associated with that risk of dying suddenly. I've listed them there as family history, unexplained syncope, abnormal blood pressure response, non-sustained VT on a Holter monitor, and severe left ventricular hypertrophy. And the analysis of those variables has formed the basis of our approach to risk for almost two decades. The idea took a little while to get some traction. Here you see two studies separated by ten years, showing this here, the decline in survival or increase in rates here, with increasing numbers of so-called major risk factors. And this notion, this summation of risk factors, one, two, three, four risk factors, form the basis of the last ACC ESE guidelines, the last joint document. The more of these things you have, the greater is your risk. Now that reflected the knowledge and practice at the time, but things have moved on, and we're slowly, as a community, pushing up the bar for rigor in terms of the studies and the statements that we make. So if you take this one, two, three, four risk factor model, does it actually predict risk? Does it work? So this is, if you like, an ROC curve, looking at the ability of that one, two, three, four risk model to separate the high from the low risk patient. A really good model would be a right angle curve here, a straight line is random. We can see that it's better than random, but could be significantly improved. So how do we do that? Well, in the last ACC AHA guidelines, the approach taken was to consider risk factors as a hierarchy, so not all risk factors are equal. And here you can see that family history, or indeed severe hypertrophy, were considered to be more predictive than non-sustainability or an abnormal blood pressure response. The notion of what has been called arbiters was also introduced. Other features of the disease, which in the gray case can be used to help you in decision making. Entirely reasonable approach. One of the problems with this approach, though, is that we have very little evidence that any one risk factor is more predictive than another. Here you see from our own experience and bits of experience of others, looking at outcomes for individual major risk factors, and there is no difference. If there is any difference, non-sustainability always just edges slightly ahead of the others against all expectation. Another approach is to look for other aspects of the disease which we can use to assess risk. Now, that may be genetic status, it may be scar burden, and all of these things make sense, but collectively in medicine, we're moving away from this notion that there is any single biomarker which is sufficiently predictive to tell you what's going to happen to a patient, because you're dealing with complex biology here, so you're going to need more than one parameter to separate out those who are more likely to develop one particular outcome than another. This forms the whole basis now of predictive tools or decision tools undergoing a whole variety of different names. I'm sure you're very familiar with the Framingham study, and if you go to the Framingham site, you'll see here tools for estimating outcomes in atrial fibrillation, congestive heart failure, coronary disease, diabetes, hypertension, based upon large-scale population data. And this notion of using tools is rapidly, has rapidly gained traction in cardiovascular medicine as in other areas of medicine. This is from a recently established database at Tufts, showing here you can see the number of publications in 2012 reporting new risk tools or decision tools, more than 500 such examples, and this is probably a conservative estimate. Notably absent from that until relatively recently was a similar method for estimating risk in hypertrophic cardiomyopathy. That was until the publication of this study, which was a multi-center study of European centers with about 3,600 patients. Nothing particularly revolutionary in what went into this study, but we took those variables from the literature that had been shown in at least one study to be an independent predictor of risk. And they're shown on here, the usual suspects with a few additional, perhaps the most important is age. Because we know that age is a significant modulator of your risk of dying suddenly with this disease. Each of the things listed on this slide were, as they had been shown previously, to be significantly associated with risk. There was no difference between them, again, if you want to look for differences, non-sustained VT, again, edging out slightly in front of the others. But essentially all of these things were associated with an increased risk. And we then used that to create a model. Now I just put this slide up more as a language tool more than anything else. So we use tools all the time, and there is a vocabulary that goes with the use of predictive tools. That if you want to understand the validity of what you're using in your clinic, you probably have to become familiar with. How has the model been validated? Internally with its own data or with an external cohort? Is the model predicting one, five, ten years? Is it powered to do that? And a variety of other terms which are used to determine how well the model performs. If we look at the Hocum risk model, this is basically does what the model says is going to happen, actually happen. So here the black bars are what actually happens. The gray bars is what the model predicts will happen. And in the original source paper, there was a slight overestimation of risk in the low risk categories and a slight underestimation of risk in the higher risk categories. But if you compare that to what we had before, which was the one, two, three, four risk factor model, there was a significant improvement in performance. With an area under the curve very similar to that of CHADS-VASC. Now because this was a validated tool, because it appeared to provide some incremental value over the previous model, this was incorporated into the 2014 ESC guidelines as an online decision making tool in which you can calculate individualized five year estimates for sudden cardiac death. Now inevitably, because this was new for this particular area of cardiology, although not for the rest of cardiology, we've gone on to do further validation of this. This is now in a second global cohort of a similar number of patients, three and a half to 4,000 patients from North America, from Europe, from the Middle East, and from the Far East. And if you look at the performance of the model here on the left, the gray dots are the predictors, the black dots are actually what happens. And what happens in this more ethnically diverse and geographically diverse cohort is that there is some disparity at the higher end, but the model is over predicting. It's pretty good at telling you what's going to happen at low risk, but it over predicts, not under predicts, it over predicts, which means you're going to put in more ICDs than perhaps you necessarily need to. If you look at the survival curves for the populations, those who were estimated at low risk, those who were estimated at high risk, intermediate risk category, most of those individuals didn't have events. We've subsequently gone on to do a meta-analysis of all of the studies which have been published since that risk calculator was published, and what you see here in a pooled data set of around 8,000 patients is those who are estimated at low, medium, and high risk, and the gray boxes here are those who go on to have an event, and an event may be a death or it may be an ICD discharge, and you can see that the majority of the events occur in those individuals who are medium to high risk, and the overwhelming majority of those who are estimated to be at low risk have no events. Okay, so what next? One of the major omissions from this work to date has been that it is exclusive to add adult patients with the disease, but there are now efforts afoot to correct this deficiency, and this is work being led by one of my colleagues in London, Juan Caskey, with a pan-global pediatric cohort, which is setting out to develop a specific pediatric tool. But I think more fundamental to this, and this is, if you like, part of the evolution in our thinking about how you use these tools, is what do you do with the number that you get? Now this is a decision tree that's in the ESC guidelines, and there we've created three risk categories, high, intermediate, and low risk, with some recommendations, and this comes, I think, from the existing philosophy of guidelines, which is, you know, people want to be told what to do. Of course, these thresholds are arbitrary. I was there at the time. I know how they were defined. It was a group of 20 guys and one woman, sitting around a table, saying, what do you think? Because there is no such thing as an acceptable risk. No one can define that. Risk depends on your philosophy as a physician, on the patient's philosophy, and also what the pay is prepared to pay, and this is something which I think is a debate now, which is taking place amongst organizations charged with developing guidelines, and people are moving to what is called shared decision-making, and one of the problems with that, it trips off the tongue very easily, that shared decision-making, but often what people are referring to is what they do in the clinic every day, which is to talk to their patients, but there is a science to shared decision-making. Is the patient actually a participant in that decision? Are you presenting the patient with sufficient data and information that allows them to be an active participant? And when you start to assess that in real world, you do get some challenging responses to that from a clinical community, and I've listed some of them on there. Nevertheless, I say this is a science, and with training, and with additional tools that allow you as a physician, first of all, to understand what risk means to you, but more importantly to help you discuss that risk with your patient, we can use tools like this, visual representations of risk. If you have an intervention which reduces risk, showing it in this way is very helpful, and it's really important to talk about absolute changes in risk rather than relative changes in risk. If you go for something which affects 2% of a population to 1% of a population, that's fantastic relative reduction, but absolute reduction, a whole different matter, and you can demonstrate that with these kind of representations. If you do that for hypertrophic cardiomyopathy using the Hocan risk tool, that the number of ICDs you have to put in at different thresholds of risk to save a life, this is not quite the same statistically, but this is if you like your NNT, your number needed to treat, you know, at a risk of 4% or more, you've got to put 22 devices in, 16 one device, 6% one device to save 13 lives. So it is possible. It allows you to individualize, and it allows you to engage the patient in a discussion about intervention based on real data. This slide summarizes some of the people who've helped with the generation of this work, and I'd like to thank them, and thank you for listening. Thank you. Okay. We've sort of almost obliterated the question period, but let me just ask you, as you know, some of us have been critical of the risk score, or at least concerned about its perceived lack of sensitivity, okay, perceived lack of sensitivity, and you haven't talked about sensitivity in the context of applying the score to an independent population and measuring against individual patient outcome, such as defibrillator implants and shocks. So I thought maybe you could address the issue of sensitivity in context of score. I'm not quite sure what you mean by sensitivity in this context. Well, how many patients are left, how many high-risk patients that would have life-saving interventions are left vulnerable when applying the score, or is the score always perfect? No, but nothing in life is perfect, but if you look at that figure I've just shown here, this is pretty darn good for any risk tool that we use in clinical medicine. So if you look here at the low-risk group, 5,000 patients, 58 of those 5,000 go on to have another event. So it's not absolutely perfect, although who these people are is something that we're just starting to learn. Some of them are older, some of them have died from non-sudden death mechanisms, in terms of ventricular arrhythmia at least. So I can only go on the basis of the data in 8,000 patients, and to me what that's showing is that if you define risk in this way, and there's no magic to it, because the risk calculator is just taking what we were doing before and improving on it, you know, left ventricular warfariness, non-sustained VT, syncope, all of these, these are all the things that without any quantification we were doing before, here we are quantifying. Okay, I still don't see sensitivity, but we have a question, and we have to keep it very quick, and hopefully the answer very concise. Thank you for your talk. Both septal thickness and outflow tract gradient are used in the risk score. What's your approach to risk stratifying a patient who has undergone septal reduction therapy and thus modified one or both of those parameters? Yeah, that's a good question. So the bottom line is this tool was not designed to look at post-intervention outcomes. We have modeled that in the validation cohort, and it doesn't seem to make a substantial difference to the predictive power of the model, because those factors individually, whilst they're statistically associated with risk, they're not overwhelmingly associated with this greater than any other risk factor. So if you've got the substrate, you've blacked out, you've got non-sustained VT and a family history, having a gradient there or not doesn't make a huge difference to your estimated risk. Does not make a difference applying their pre-septal reduction therapy values or their posts? I mean, you can put whatever you like into the calculator, but it wasn't validated for that purpose. So retrospectively, in the validation cohort, we didn't see an influence of septal reduction therapy on the performance of the model. Thank you. OK. Again, quick and quick. Informed decision making is the key of all aspects of medicine, but in brief, one question. What do you tell a patient who is deemed at low risk, but he says, Doc, I do not want to be at any higher risk than a person who does not have HCM? And you've classed him as being at low risk, but he states, he categorically states, that I do not want to be subjected to any higher risk of dying suddenly than a person without HCM, even after you've deemed him at low risk. Absolutely. What do you tell? I mean, do you implant a defibrillator or not? So as your question, what do I say to the very low risk patient about defibrillators? I use the tool. I have the tool open on the screen, and I explain that based on the current knowledge, this is what we would estimate your five-year risk to be. Depending on their age, it's often close to, sometimes lower than their actuarial risk, and then we talk through the pros and cons of a defibrillator and what that means in terms of long-term morbidity. Most patients, once they see the data, choose not to have a defibrillator. Okay, this is going to have to be the last one, sorry. Very quick and quick to answer. There's obviously convincing data out there of the effect of scar burden and the dose effect in terms of increasing scar burden and its impact on sudden cardiac death. Could you comment on why or did it drop out as the calculator was made, and in patients that... I missed the beginning of your question. What? I missed the beginning of your question. Oh, I'm sorry. Exactly. There's convincing data of scar burden and its influence on sudden cardiac death risk and a dose effect of increasing scar by late gadolinium enhancement on cardiac MRI and its impact on overall mortality. Did this drop out as you were making your calculator? No. Why was it not incorporated in the model, and how are you incorporating it in your decision? We haven't put that into the calculator for two very good reasons. In order to get sufficient event rates, you have to use a cohort which goes back three or four decades. So the data just didn't exist to put into the model. It may be that in time, and as you're aware, HCMR is ongoing at the moment, we'll have data that shows that it's an independent risk factor. This model can be reconfigured in any way you wish. What I would say is that the interim data from HCMR are showing that the event rates are low, and actually what it's predicting at the moment is atrial fibrillation rather than sudden death risk. Okay. Well, the fact of the matter is the MRI is not in the score.
Video Summary
Dr. Maron discusses the evolution of risk assessment in hypertrophic cardiomyopathy (HCM) and the development of an online risk calculator called the HCM Risk-SCD. The risk calculator uses various clinical variables, such as family history, syncope, abnormal blood pressure response, non-sustained ventricular tachycardia (VT), and severe left ventricular hypertrophy, to estimate an individual's risk of sudden cardiac death (SCD) over a 5-year period. The calculator was validated in a global cohort of HCM patients and showed good predictive performance for SCD risk. However, there are still some limitations and areas for improvement, such as the need to include pediatric patients and the challenge of incorporating scar burden and genetic status into the risk assessment. Dr. Maron emphasizes the importance of shared decision-making between physicians and patients in determining the appropriate interventions, such as implantable cardioverter-defibrillators (ICDs), based on the estimated SCD risk.
Meta Tag
Lecture ID
4998
Location
Room 152
Presenter
Perry Elliott,
Role
Invited Speaker
Session Date and Time
May 09, 2019 10:30 AM - 12:00 PM
Session Number
S-012
Keywords
risk assessment
hypertrophic cardiomyopathy
HCM Risk-SCD
online risk calculator
sudden cardiac death
clinical variables
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