false
ar,zh-CN,zh-TW,en,fr,de,hi,it,ja,es,ur
Catalog
CONFIRM-2: Artificial Intelligence Enabled Quantit ...
Clinical Conversation
Clinical Conversation
Back to course
[Please upgrade your browser to play this video content]
Video Transcription
Well, hello, everybody. Welcome to a Sky Conversations in Interventional Cardiology event. I'm really pleased to have all of you join us. My name is Jader Sandoval. I'm an interventional cardiologist with the Minneapolis Heart Institute and the co-chair for the Center for Coronary Artery Disease at the Minneapolis Heart Institute Foundation. I'm honored to represent Sky for this conversation on interventional cardiology. And we're here today to discuss the important late-breaking science that was presented just recently at TCT of the CONFIRM-2 study, which is Artificial Intelligence-Enabled Quantitative CT Assessment of Atherosclerosis and Major Adverse Events of Multicenter International Registry. I'm joined by esteemed panel of leaders and experts involved in this field, multidisciplinary and a key member from Sky. We have Dr. Alexander Van Rosendael from the Department of Cardiology at Leiden University. And at Netherlands, we have Dr. Mervet Alesnak, the Director of the Cardiac Activation Laboratory at King's Faher Armed Forces in Jeddah. We have Dr. Omar Khaliq, the Director of the Division of Cardiovascular Imaging at St. Francis Hospital of Catholic Health in Long Island in New York. And we also have Dr. Evan Schlafly, the Director of Intravascular Imaging at St. Francis Hospital at the Heart Center in Rosslyn, New York as well. So it's again, really a unique opportunity for us to discuss this trial. I'm going to turn it over to Dr. Van Rosendael to start the conversation and discuss this important study. Thank you, everybody. Thank you all for having me here. And I'm presenting the results of the CONFIRM-2 in name of all the CONFIRM-2 investigators for which many of you in this discussion are like great contributors and co-authors. Thank you for that. So in patients with chest pain, today's workup is usually based on ischemia, the assessment of ischemia, although many recent trials have shown that this actually is not a very effective approach. It doesn't really improve prognosis. Meanwhile, we know from studies from Intravascular Ultrasound, NCT, that especially the plug burden and the plug morphology like low attenuation plug or like high-risk plug features are especially prognostic for future major adverse cardiovascular events. And a tool that is especially well suited to assess the plug burden from the entire coronary tree is coronary CT, a non-invasive tool to assess the coronary arteries. And from large prognostic studies, for instance, the CONFIRM-1, we've seen that the CT provides very strong prognostic value for future heart attacks. So in CONFIRM-2, we performed artificial intelligence-aided quantitative CT of the whole heart of every coronary segment from the coronary tree. And we know that this AIQCT, as it's called, provides reproducible and quantitative automated whole heart plug quantification with high degree of correlation with invasive cold standards such as IVUS. So in the CONFIRM-2, we included patients without known coronary artery disease and who underwent a clinically indicated coronary CT, and we followed them for about five years for major adverse cardiovascular events. So we included three and a half thousand patients, as you can see on the left, after excluding asymptomatic patients, patients with prior coronary artery disease, and two patients with missing clinical information. Patients were included from 18 sites all over the world, and we had 5% of event rates of our primary endpoint. As you can see, the primary endpoint was consisted of death, myocardial infarction, cerebrovascular accidents, hospitalization for heart failure and unstable angina, and late revascularization. We also had a secondary endpoint, which was only consisting of myocardial infarction and death. So this is the findings we had from the CT scan. On the left, you see the conventional CT characteristics that everyone is familiar with. So most patients had non-obstructive coronary artery disease, 60%, 15% had obstructive disease, and 25% had none or minimal coronary artery disease. One interesting thing that's different than other prior CT registries is that we combined none and a little bit of plug together, because what we see with the AIQCT actually that it picks up very, very small amounts of plug. Only like 7% of our patients had a normal CT scan, and we are now in the process also of investigating what is this prognostic value of this minimum amount of plug. On the right, you see the findings we had from the quantitative CT. So we provided patient-level values, so the total plug burden from every coronary segment, the total was an average of 150 cubic millimeters of plug, and most of this plug was non-calcified. As you can see, about 90 of the 150 was non-calcified plug, and 60 was calcified plug, and we also defined how many patients had high-risk plugs. That is defined as a remodeling index of more than 1.1, and together that the plug has a large burden of low-attenuation plug, that's the necrotic core. About six patients had high-risk plugs. So one of the aims of our confirmed study was to identify the most strong predictors for a primary endpoint, and after we had 24 univariable CT predictors, only two of them were significant and most strongly related to their primary endpoint, and that was the burden of non-calcified plug, so not calcified, but the non-calcified plug burden, together with the stenosis severity, and just looking at those two variables already gave an area under the curve of 0.76. Here you see a comparison with the Diamond-Foerster pre-test likelihood ratio. So for instance, if you look at the AI-QCT model, so you base your risk stratification on the CT, when you are in the lowest risk, you only have 1% events, and when you compare it with the highest risk group, you have 10% events, so 10 times more risk for events. And for the Diamond-Foerster groups, low versus high group, it only gives you a two to three factor increase in risk. So that means that you have better risk stratification with our AI-QCT. And here you see a stepwise model of how much extra information does the AI-QCT bring you. If you go to the left, the area under the curve of the Diamond-Foerster is 0.58. If you add in the clinical risk factors, it goes up significantly to 0.65, but adding in the AI-QCT model, as I said, only consisting of like the burden of non-calcified plug and stenosis severity, it significantly increases to 0.77. And here you see the results for a secondary endpoint myocardial infarction and death, and basically you see a similar picture of having much increase in discrimination over secondary events when you add in the AI-QCT model. So in conclusion, CONFRAME-2 is the first multi-center, multinational registry with AI-based quantification of coronary artery disease, especially the lumen diameter stenosis and the non-calcified plug burden from the quantitative CT were most predictive for events, and they were much more strongly predictive than clinical risk scores. And to bring it in a clinical perspective, the atriosclerotic profile, as we can assess with AI-QCT, it may guide your intensity of anti-atriosclerotic therapies you will provide to your patient, and perhaps it will guide in performing coronary intravascular procedures all aiming to reduce cardiovascular events in the future. So these were the findings I wanted to show you. Alexander, wonderful presentation, really congratulations to you and CONFRAME-2 investigators for this very exciting work and the presentation at TCT and now sharing this data with us. Maybe to get started, and we'll go and discuss with our expert panel here, but just to get started the discussion, can you speak a little bit more about, I understand these were patients that were having a clinical CT being performed. Are these patients mostly with stable coronary disease outpatients, or did you have a proportion of patients that were coming to the hospital acutely, let's say with suspected ACS or even myocardial infarction? Can you speak to that a little bit? So these are actually all outpatient clinics, outpatient clinic patients that were referred for a clinically indicated CT, and they were all symptomatic. So these are stable outpatients with suspected stable coronary artery disease, and usually as we see with CT is that it's a lower risk type of patient, because that's according to the guidelines as we have now the predominant indication for a CT scan, and that's also reflected by the relatively lower number of events, but these are stable outpatients with suspected coronary artery disease and without known coronary artery disease. Wonderful. And maybe let me shift here to Dr. Alasnag, I know she was involved in this important study as well, and she's an interventional cardiologist, and this is of course a conversation for SCI, an interventional community, so I'm just very curious about your input on, I mean this is something that I know the imagers, Dr. Kalik and others, talk about this all the time, but how should we in interventional cardiology understand what the opportunities are here? You know, I think we're moving into exciting times where the demarcation between interventional cardiology and imaging is starting to blur. We're starting to see imagers understand coronary angiograms well, and interventional cardiologists resorting to CTs and imaging a lot more. This particular trial was largely reassuring for me as an interventional cardiologist, so these are low-risk patients, outpatients, 25% of them were normal, but actually over 60%, you know, had non-obstructive coronary artery disease, so that's reassuring to a certain extent that these are patients that are outpatients walking and so on, and these are the findings that we have. It is still interesting that even with AI, one of the most important predictors is percent stenosis, so diameter, in addition to non-calcified plaque, and again, so that just tells us these are the areas that we need to be investigating a little bit more, but the data is fairly consistent when you look at the complete revascularization study and you look at the complete OCT, for example, maybe Evan can talk to that because I know you do a lot of intercoronary imaging. Again, it showed that there was good correlation between stenosis, obstructive coronary disease, but perhaps thin cap atheromas, and so we're going to just have to start putting all these together to understand a little bit more what is, you know, a higher-risk plaque that we need to be dealing with. And let me shift here to Dr. Slaffman. Sevan here, I mean, you're an expert in intravascular imaging. I've associated and have looked at this many times, you know, invasively for a long time. I also know you're interested in secondary prevention and primary prevention. How do we see this? I mean, I should be proactive about medical therapy. What are the actions that particularly us in the cath lab should be, you know, embracing here when you see this? Definitely. I agree with all the comments and, you know, really such an interesting presentation. For exactly what Mirva said that, you know, two-thirds of patients had non-obstructive coronary disease. And, you know, traditionally what would happen, we'd see these patients in the cath lab, you do high-resolution intravascular imaging, and you see there's TIKVA, and it's non-obstructive. You're not stenting them, but for me and my practice, it changes my management, how aggressive I am with intensifying PCSK9 inhibitors to thicken the fibrous cap and stabilize that plaque and really just being aggressive with that. But what's nice is we don't have to wait until a patient ends up on a cath lab table here. You see this enhances the information we get from a CAT scan. And in that, you know, two-thirds of patients with non-obstructive disease, they may never get to a cath lab in clinical practice. But if they have high-risk features, which pretty common, one in 20 patients in this study, you're intensifying their pharmacotherapy. And then also you're probably gonna have a lower threshold to refer that patient to the cath lab if over time that patient has progressive symptoms. So I think this is really impactful for clinical practice. Dr. Kaleik-Omar, you lead one of the leading imaging programs, super high volume. And so, you know, this is very interesting data from all these AIQCT features, you know, total plaque volume, non-causative volume, causative plaque volume and whatnot. Can you speak, you know, we've used some of this technology, but I think maybe I speak more for myself, but I think many of us have yet to understand what these numbers mean and how are they reported? How do you see this moving forward? Are there something that you guys are ready or foresee that this should be in the clinical report and how people will act on that? What is your take on how to, you know, how's the field moving forward in this aspect? Yeah, so I think this is gonna be a big move forward in the field. I think that currently if you ask most clinical cardiologists and I have asked a lot of referrings, what would they do with a plaque burden of one millimeter squared versus 100 millimeters squared, would it change their management and then the composition? And most of them will say that if the patient was naive, they would start them on a statin and none of the other factors would really change anything. But I think we're now entering the area with the various studies that are gonna be done and more pharmacotherapies where we can start to investigate and develop a tiered approach where we can say, give ranges, right? Plaque burden of, you know, I'm just making up the numbers, but one to 100 with calcified plaque, you get agent A. With mixed plaque, you get agent B. If it's a higher plaque burden, you add agent C. So I think that's really where we need to be moving into a tiered therapy based on not just obstructive and non-obstructive, but what are the plaque characteristics and what is the risk? And I think from the imaging standpoint, the AI is gonna help us because I think this is a key point. Why not just read all this yourself? Well, if a patient has 10 lesions and you're reading 30 CTs a day, it's really hard to do all of that manually. So just from the sheer workload and not missing any information, it's important to have AI that makes this much faster and much more automated. Thank you, Omar. And Alexander, you know, and going back to you, the, I understand, you know, is this like an initial data set of confirmed? So I understand the study, it's thousands of thousands of patients. So where are you guys going with this? What are the next steps for confirmed two? Yeah, so this is the first report from the confirmed two, the first study that has been performed with three and a half thousand patients. We are aiming to include much more patients as now we are about to close our second data set, which aims to have 10 to 12,000 patients. And we aim to keep enrolling patients and to also look at like different cohorts of asymptomatic ones, perhaps patients with primary coronary artery disease and other subgroups. And yeah, our aim is to learn much more from the quantitative CT, especially like what do these values mean? As you said before, like, is this like, like this 150 cube millimeters of plaque, is that much? Is it not much? How should I treat it? Like, that's all things we need to learn. Like how important are these high risk plaque features above burdens of overall plaque burden as we have now. Like that's all things we need to learn. And that's where we need more patients for. For some subgroups, we may need like 20,000 patients. So we are currently enrolling much more and hope to do upcoming years and try to get many more interesting publications from it. Well, that's wonderful. And I know that, I mean, this for, there was of course the confirmed, the initial confirmed study, right? Back I was in 2011 or so. So I know that that really provided unique features as sex-specific differences and non-obstructive disease. So I guess now you guys will have the ability to look at this, but now with AI, QCT and this sort of like very modern sort of way of FASL to do this. Do I understand you guys will also be looking at all these subgroups, right? Sex-specific groups and whatnot. Yeah. Yeah, absolutely. Yeah. One of the, like confirm one is like for 15 years out now, like the main difference is that it's now confirm two is with AI quantitative CT. And I think that's a very large like improvement because as we've seen non-calcified plaque burden was most predictive. That's in my personal opinion, that's not something you can just like do very well with your eyes to say like on this scan, I see a large or a small burden of non-calcified plaque. It's really difficult to do that. Like calcified plaque, we have the calcium scoring, that's more easy. But now we know that the non-calcified plaque is most important. I really think that we need AI tools for that. And also to do it reproducible and to do it accurate, to do it fast, we just need that. And that's where the big advantage of AI is. And that's the difference with confirm one, confirm two, everything automated with quantitative CT. In Miravet, you know, for an international audience, I mean, like I'm sure there's some differences in practice between all of us and our centers and whatnot, but in whom and which patients should we be thinking about obtaining this? Can you maybe speak to your practice and which patients are, of course, if somebody is symptomatic, you're ordering CT to exclude obstructive coronary disease, but I guess now we're shifting our thought to this issue of plaque burden and all these issues. When are you, how are you using this? So just really quickly before I actually answer that question, the one fantastic thing about confirm two is that it is actually involving a multinational database population. And so it'd be really fantastic not just to look at, you know, sex differences, but to look at population differences. So an Arab population versus a Caucasian population and so on. And I think with the number of patients enrolled, this is gonna be the first of its kind really that's gonna permit us to do that and see there are actually ethnic differences in our patients. In terms of our own center, you know, honestly, we're still looking at patients who are stable coronary disease or chronic coronary syndrome type of patients. I think the patients who have prior CABG, these are the kind of patients that are fast-tracked into CT. We're still very reluctant to take patients who have had prior stents. And I think part of the reason is not that the sophisticated CTs can't look too much, but sometimes they actually look and you don't know what to do with the information. And it's a bit difficult to quantify incidentary stenosis by CT and to know what it means and what to do with it. I mean, sometimes I'm at loss, even when I'm doing a cath and I wanted to go and do intracoronary imaging to decide what is the mechanism, is it an under expanded stent and so on. So I think we're going to find our way over the next decade, but I think for now, the bulk of the cases are going to be chronic coronary syndrome and perhaps very low risk ACS cases. Wonderful. And Evan, going back to you, I'm just interested in how you see this integrating with the other tools. Is this something that you see as a complimentary to some of the information that we, seeing the cath lab when we're putting imaging tools seeing the cath lab when we're putting imaging catheters, how do you see this? Yeah, it's very complimentary. This enhances your ability to, for the patients who have obstructive disease and are going on to PCI, it really helps with procedural planning. We have a lot of intravascular imaging enthusiasts on this call, but the reality is globally intravascular imaging rates are extraordinarily low. So there's a lot of information for those who are still practicing angioguided PCI. You may see the obstructive stenosis in one location, but you see that there's this non-calcified plaque adjacent to it. That's helpful in procedure planning so that we could avoid landing our stent in that lipidic plaque that's high risk and you make sure you're covering the entire lesion. So I think ideally would be universal intravascular imaging. We're certainly not there. And this is going to help patients today get improved results who end up going to a cath lab and really have an enhanced CT-guided PCI procedure. Thanks, Evan. And Omar, I'm just interested to hear from you. I mean, of course, CONFIRM is ongoing and Alexander and the CONFIRM group are doing all this work, but how do you see the field moving forward? I mean, we're now reporting all this very valuable and informative prognostic AI, QCT data, black volume, all these things, but can you speak to the future directions? I mean, how does this get into the guideline or to practice? Do we need a trial that by doing X intervention, we help patients or I know you're very involved in new societies and how do we get this into practice? What are the next steps? So I think we talked a little bit about it previously at McCall, but there are investigations starting which are going to study different tiered approaches based on these various characteristics. Those are ongoing and the PREVENT trial started some of the kind of preventative therapies that could be beneficial based on not only stenosis, but the type of plaque and the plaque burden, right? That's an important area, not just the percent stenosis. And then I think the other areas, now with the AI and the newer technology scanners, we're getting information that's much closer to what you get in the cath lab with intravascular imaging, right? Closer to the spatial resolution and differentiation of plaque. We're not there yet, but we're getting closer and I think we'll get even closer in the future. And then it's the type of collaborations we're having on this call, which are actually pretty old in the structural heart community where we're speaking the same language between the imager and the interventionist for coronary cases and having a closer collaboration getting rid of some of the old stuff, like CAD-RADS scores, which were useful in a different era maybe, but are largely irrelevant to the rest of you on this call, the interventionists. So I think we need to really tighten things up there to really have people uptake coronary CT, which I think has proven to be very valuable and to really move us forward in the field. Well, thank you, Omar. As we wrap this up, maybe I wanna circle back with Alexander. Again, congratulations to your group and this exciting data. Any other major message, maybe a key finding that you wanna share with the audience that maybe we haven't discussed or any closing thoughts from your side? Yeah, well, what I personally like most from the CONFIRM-2 is that it really brings, the AI quantitate, like it really brings the plaque burden measurement, the quantitative CTC into clinical practice. It's already existing for many, many years, but it was always too much work to do. And now it can really be like get it in a few minutes and we're starting to learn what do these values mean? And also like it's the way to have the measure of plaque burden, which I personally, in my patients, I really use that as the main driver of how intense am I gonna treat my patients? And I'm really excited that we have this tool now to really get this information accurately. And now we need to learn how to optimally use it into clinical practice. That's it. Well, wonderful, Alexander. Well, I think, we're pretty much ready to close this session, but I do want to really highlight, and I think Dr. Ahlersnack made that point early on, that for us, I think it's really a wonderful step to see our interventional society, really see the opportunities here for multidisciplinary collaboration. We're all interested in improving outcomes for patients with coronary artery disease. And I think it's increasingly, I think Evan and Omar also, all of us alluded to this. Increasingly, we're trying to see these opportunities irrespective of the image modality and really learning from each other from different specialties, which I think is a really great thing for us to move the field forward. So again, thank you for all of us for joining for Sky Conversations Interventional Cardiology. This will be available online. Congratulations, Alexander and your group. Thank you all. Thank you.
Video Summary
The Sky Conversations in Interventional Cardiology event, led by Jader Sandoval from the Minneapolis Heart Institute, focused on the CONFIRM-2 study. This multicenter research leverages AI in quantitative CT imaging to assess atherosclerosis and predict major adverse cardiovascular events. The study involved 3,500 patients from 18 international sites, focusing on non-calcified plaque burden and stenosis severity as key predictors. Dr. Alexander Van Rosendael highlighted the significance of AI in accurately measuring these factors, which contribute to better risk stratification and treatment planning. Discussion among experts like Dr. Mervet Alesnak, Dr. Omar Khaliq, and Dr. Evan Schlafly emphasized the merging of interventional cardiology and advanced imaging techniques, underscoring the potential to enhance preventative strategies and informed clinical decisions. The ongoing CONFIRM-2 aims to expand data for better understanding and application of these findings in clinical practice, ultimately improving patient outcomes.
Keywords
Interventional Cardiology
CONFIRM-2 Study
AI in CT Imaging
Atherosclerosis Assessment
Risk Stratification
Preventative Strategies
×