Video

AI in Quantitative CT Assessments: CONFIRM2 with Dr Alexander Van Rosendael

Published: 06 Nov 2024

  • Views:

    Views Icon 212
  • Likes:

    Heart Icon 0
Average (ratings)
No ratings
Your rating
View Transcript Download Transcript

In this episode of the "Innovations in Cardiology" series, host Dr Dipti Itchhaporia (Hoag Heart and Vascular Institute, Newport Beach, CA, US) is joined by Dr Alexander Van Rosendael (Leiden University Medical Center, Leiden, NL) to discuss the findings from the CONFIRM2 trial. Conducted from up to 50 international clinical CCTA sites with 35,000 patients, CONFIRM2 aims to improve comprehensive quantification of coronary computed tomography angiography findings using an artificial intelligence algorithm that will enable an accurate quantification of coronary artery disease, including total coronary plaque volume and low attenuation plaque volume.

Findings showed that AI-guided quantative computational tomography quantification of coronary CTA lumen diameter stenosis and the noncalcified plaque volume were most predictive of MACE and added discriminative accuracy to the Diamond and Forrester predictive model. These atherosclerotic profiles as assessed by AI-QCT could guide the tailoring of anti atherosclerotic therapies or coronary interventional procedures to reduce the instance of cardiac events.

If you have any questions or suggestions for topics to cover on the Radcliffe Podcast, please email managingeditor@ecrjournal.com.

Transcript

Dr Dipti Itchhaporia:

Hello, I'm Dipti Itchhaporia, and welcome to another series, Radcliffe Innovation series. I'm here at TCT 24 with Dr. Rosendael from the Netherlands, and we're here to talk about the CONFIRM2 trial, which is looking at quantitative, artificial intelligence, quantitative CT assessment of atherosclerosis. Dr. Rosendael, welcome.


Dr Alexander Van Rosendael:

Thank you. Thank you for having me.


Dr Dipti Itchhaporia:

So tell me a little bit about the CONFIRM2 trial.


Dr Alexander Van Rosendael:

Yeah. So in patients with suspected coronary artery disease, a diagnostic approach based on ischemia has not resulted in improved clinical outcomes, what we know from the latest randomized controlled trials. But what we know is that CT is a very good tool to assess the total burden of atherosclerosis from an entire coronary tree. And previous IVUS and CT studies have shown that this burden and composition is a very strong predictor for events in the future.

So, usually, the most CT studies before have evaluated the plaque burden by looking at the plaques and scoring them based on a visual scoring system. But with CONFIRM2, every CT scan got an artificial intelligence quantitative CT analysis of the whole heart of every coronary segment. So what that did was that quantified every plaque in vascular morphology and gave reproducible and reliable numbers, for instance, on plaque volume, plaque burden. And these values have been validated with invasive measures such as IVUS before. So that's what we did with CONFIRM2.


Dr Dipti Itchhaporia:

So go back and tell me about the patients that you enrolled. And tell me how many centres were there, how many number of patients, what types of patients?


Dr Alexander Van Rosendael:

Yeah, so for this type of study, we need large populations, large patient bases because the event rates are usually low. So we took symptomatic patients without prior coronary artery disease, and we took them from all over the world, from 16 sites in 13 countries through Asia, Europe, and the U.S. And we followed them for almost five years for major cardiovascular events.


Dr Dipti Itchhaporia:

How many total patients?


Dr Alexander Van Rosendael:

3,551 patients.


Dr Dipti Itchhaporia:

Okay, so now tell me a little bit more about what you did exactly.


Dr Alexander Van Rosendael:

Yeah. So the aim of the CONFIRM2 was to see which are the most predictive CT atherosclerotic features predictive for the primary event because it gives you many. It gives you like 30, 40 variables, but not all of them are important, and not all of them are clinically as interpretable as possible. So we wanted to see what are the most important ones for our primary and secondary endpoint.

And our primary endpoint was a composite of all-cause mortality, stroke, myocardial infarction, late revascularization, unstable angina, and heart failure. And our secondary endpoint was myocardial infarction and death.


Dr Dipti Itchhaporia:

Okay, so you did this and you correlated all of this data with IVUS data, is that what you said?


Dr Alexander Van Rosendael:

No, it's not correlated with IVUS, but we know from previous studies that quantitative CT correlates very well with IVUS. So that means that if you have a plaque volume of, let's say, 100 cubic millimeters of plaque, that means that that's very similar to 100 cubic millimeters of plaque on IVUS.


Dr Dipti Itchhaporia:

Okay, and did you, how did you decide about the risk score on these patients prior to doing the CT scan?


Dr Alexander Van Rosendael:

Yeah. So all of these patients, they had a clinical suspicion of coronary artery disease. So the referring doctor thought they had coronary artery disease and they wanted a CT scan in these patients. And for this study, we wanted to compare the prognostic value of the CT features with the clinical traditional scores we usually use, such as a ASCVD risk score or the Diamond-Forrester score, together with risk factors. Because the Diamond-Forrester score we know it's not very well predictive. So we compared the CT features for prognostic value, so the ability to risk stratify, with the clinical risk scores we usually use.


Dr Dipti Itchhaporia:

And so tell me now about the results.


Dr Alexander Van Rosendael:

Sure. So out of 24 quantitative CT features, we saw that two were the most strongly predictive for the primary endpoint and that was the burden of noncalcified plaque. So every noncalcified plaque volume from the tree together on a patient level. And the second variable was the diameter stenosis. So the highest stenosis degree in any of the segments in that patient. And those two variables gave a very accurate risk prediction for the primary endpoint.

So let's say if you were in the lowest, let's say if you were in the lowest tertile of AI QCT risk, you had 10 times as low risk as those in the highest, highest risk group based on AI QCT. And if you compare that with, let's say, the ASCVD risk score, the lowest versus the highest only gives you a factor two or three. So that means that if you have a low ASCVD risk score, your risk is not that low, but it becomes much lower if you have very little plaque. And if you have a lot of plaque, your risk is very much higher than if you have a very high ASCVD risk score.


Dr Dipti Itchhaporia:

It's interesting because what you said, and I want to highlight this for our listeners, which is that it's the total volume of the noncalcified plaque. Correct?


Dr Alexander Van Rosendael:

Yeah.


Dr Dipti Itchhaporia:

And so tell me a little bit more about that because we all are used to listening about calcium scores. So did you correlate this with calcium scores? And obviously that's different than what you're talking about now: plaque volume.


Dr Alexander Van Rosendael:

Yeah, thanks for this question. This is a very, very good question we all want to know. So there is of course a certain amount of correlation between plaque volume, noncalcified plaque volume and calcified plaque volume, but it's definitely not the same. So what came, like, calcium score or calcium plaque volume was one of the variables that could be the strongest predictor, but it wasn't. In fact, when you compared it, noncalcified plaque volume was much more predictive than calcified plaque volume.

So that means if you have a very high calcium score, your noncalcified plaque burden is low, your risk may not be that high as we have thought about on the calcium score. And the other way around also. If you have a younger patient with a lower calcium score or a low burden of calcium, but it has a lot of noncalcified plaque, that patient is actually still high risk. And CT angiography is the only way we can see this noncalcified plaque because a calcium score is based on a non-contrast scan, and you can only see the calcium, but you miss the noncalcified plaque, which is actually more important than calcium.


Dr Dipti Itchhaporia:

Okay, so tell me, translate this for me. So what does this mean? So if you find a patient with a stenosis and high plaque volume, noncalcified plaque volume, those are the highest risk patients, that's what you identified?


Dr Alexander Van Rosendael:

Yeah, yeah.


Dr Dipti Itchhaporia:

So tell me about what, translate this into clinically. What do I do with this information? I got a CT. You know, I believe in CT and now I'm getting this. What do I do? Tell me about the spectrum of patients.


Dr Alexander Van Rosendael:

Yeah, so the patient you just mentioned needs like two interventions. First, like, needs risk reduction and we do that with our anti-atherosclerotic therapies. So this patient with the highest burden of noncalcified plaque is at highest risk for events. And we just need to give them like the most intensive anti-statin, like anti-lipid therapies, like statin species, K9 inhibitors, like anti-platelet therapies or, and all the other therapies, we have anti-inflammatory therapies that are coming up. So that's what we need to give. And if the patient is symptomatic, which were all in CONFIRM2 too, and have a high degree of stenosis, that's also likely that patient that's going to benefit from the revascularization.

On the other hand, if you have a patient with symptoms suspected of coronary artery disease, but it has almost no plaque and only very little degree of stenosis, this patient is actually very low risk. Even, despite his ASCVD risk score may be intermediate, but if there is very little plaque, very little stenosis, the event rates are really, really low, like below 1% at three years of follow-up. So that patient likely doesn't need a lot of medication. Of course, like lifestyle intervention is always good, but it's not that patient that needs the most aggressive therapies.


Dr Dipti Itchhaporia:

I think this is really exciting because now you have something more than just calcium score. You've already told us that traditional risk scoring does not really give you as much information as you need. So maybe this newer emerging technology with AI quantitative is going to be the way to go. So tell me, where do we go from here? What are some of the unanswered questions? Where do you see the next phase of clinical trial with this?


Dr Alexander Van Rosendael:

Yeah, this is a very good question. Thank you. First of all, we need to know what patients do need CT. Is it all the asymptomatic patient ones? Do we need to do mass screening, or do we need to do screening with patients with only a little bit elevated risk factors, or do we need to do only the symptomatic patients? So that's first the question we need to select the patient that needs a CT scan.

Second is we need to improve our understanding of the findings of the plaque volumes. Let's say there is already staging systems out, but we need better understanding of like what is a lot of noncalcified plaque. And also if you have this amount of noncalcified plaque, what medications do you need to give? Does this patient need aspirin therapy based when the plaque volume is this high or low? Does it need like a PCSK9 inhibitor when it's this high or low? So that's what we definitely need better understanding on.


Dr Dipti Itchhaporia:

Yeah, and it sounds like we need more of an idea of what the cutoffs are for that plaque volume. And then also if you give them these therapies, does that really make a difference?
So serial scans to see what happens with regards to that.


Dr Alexander Van Rosendael:

Yeah, and randomized trials are also upcoming on this new approach of treating your patient based on the plaque burden instead of on the risk factor. This new approach, trials are coming up on this.


Dr Dipti Itchhaporia:

I think this is an exciting topic. I want to thank you, Dr. Rosendael. This was fantastic, and I wish you luck for more of these trials. I'm looking forward to talking to you in the future to see where we are.

Comments

You must be to comment. If you are not registered, you can register here.