AI-enabled technology independently predicts cardiac events based on coronary inflammation
Click Here to Manage Email Alerts
Key takeaways:
- An AI-based technology can detect coronary inflammation invisible to the human eye.
- Incorporating inflammation data from CT scans into a risk score predicts risk for cardiac death and major heart events.
PHILADELPHIA — An artificial intelligence-based technology that evaluates coronary inflammation predicted risk for cardiac events in patients undergoing coronary CTA, according to new study data.
In the ORFAN study presented at the American Heart Association Scientific Sessions, risk based on coronary inflammation was a strong predictor of cardiac events even in patients with no obstructive CAD showing no plaque or zero calcium score. The AI-based model including a risk score derived from inflammation assessed by the technology (CaRi-Heart, Caristo Diagnostics) enabled reclassification of certain patients to higher or lower levels of risk, resulting in changes of management in approximately half of the patients. Most changes were due to a clinician decision to target previously undetected coronary inflammation, according to the researchers. On May 29, 2024, the findings were published in The Lancet.
Image: Adobe Stock
“This is a technology driven by artificial intelligence, which takes the information that is part of routine cardiac CT scans done to investigate the possibility of underlying coronary artery disease, and it looks for a signal that’s not visible to the human eye that indicates the presence of inflammation in the coronary arteries,” Keith Channon, FMedSci, FRCP, professor of cardiovascular medicine and head of the Radcliffe Department of Medicine at Oxford University and co-founder and chief medical officer of Caristo Diagnostics, told Healio. “This is important because, currently, cardiac CT scans are analyzed carefully to look for narrowings or blockages in the coronary arteries. And if those are found, that’s important, and there are treatments that those patients are recommended to have. However, the majority of patients who undergo coronary CT scans don’t have significant narrowings or blockages, and right now, those people don’t have major recommendations made. But, in fact, a substantial proportion of those patients have inflammation in the arteries, which is the precursor and the driver of the development of narrowings, and ultimately heart attacks and death. This technology ... analyzes the information from the CT scan in a new way using an AI-driven algorithm developed by Caristo Diagnostics called CaRi-Heart Technology, which identifies the invisible inflammation even in patients who have no or minimal narrowings or blockages on their scan. The impact of this is that it identifies a group of people who would otherwise be unaware that they are at risk, and it means that doctors can use the information provided by this new technology to give them advice and treatment and reduce their risk of future heart attack and death.”
The technology enabled researchers to derive a Fat Attenuation Index Score (FAI-Score) to predict inflammation-related risk and to use the FAI-Score plus plaque characteristics and conventional risk factors to reclassify patients in terms of CV risk, Charalambos Antoniades, MD, PhD, FRCP, FESC, professor of cardiovascular medicine and consultant cardiologist at the University of Oxford and co-founder and chief scientific officer of Caristo, said during a presentation. Antoniades and Channon were among the Oxford researchers who discovered the technology behind CaRi-Heart and formed Caristo Diagnostics to develop it.
The present analysis of ORFAN included 40,091 patients (mean age, 59 years; 53% men; 78% white) who underwent coronary CTA, of whom 18% had obstructive CAD.
The cohort with no obstructive CAD was much larger and the group had about twice as many cardiac deaths (1,118 vs. 636) and major adverse cardiac events, defined as cardiac death, nonfatal MI and new HF (2,857 vs. 1,450), compared with the obstructive CAD group, Antoniades said during the presentation.
In a median 7.7 years of follow-up of 3,393 patients who had an FAI-Score calculated, compared with those with no obstructive CAD, those with obstructive CAD had elevated risk for cardiac death (adjusted HR = 1.41; 95% CI, 1.28-1.56; P < .001) and major adverse cardiac events (aHR = 1.57; 95% CI, 1.47-1.68; P < .001), according to the researchers.
When patients were stratified into quartiles by FAI-Score in the left anterior descending (LAD) artery, the three highest quartiles had elevated risk for cardiac death and major adverse cardiac events compared with the lowest quartile, with the risk ascending by quartile, Antoniades said, noting that compared with the lowest quartile, the highest quartile had 20-fold elevated risk for cardiac death (HR = 20.2; 95% CI, 11.49-35.53; P < .001) and sixfold elevated risk for major adverse cardiac events (HR = 6.76; 95% CI, 5.21-8.78; P < .001).
The effect was consistent regardless of whether patients had obstructive CAD, he said. Among those with obstructive CAD, those in the highest quartile of FAI-Score in the LAD had elevated risk for cardiac death (HR = 5.15; 95% CI, 3.26-8.13; P < .001) and major adverse cardiac events (HR = 3.15; 95% CI, 2.3-4.31; P < .001), and the same was true to an even greater degree for those without obstructive CAD (HR for cardiac death = 10.49; 95% CI, 5.25-20.95; P < .001; HR for major adverse cardiac events = 4.77; 95% CI, 3.4-6.69; P < .001).
The risk prediction for FAI-Score was similar in other coronary vessels and independent from risk factors and Coronary Artery Disease Reporting and Data System (CAD-RADS) 2.0 score, he said.
FAI-Score in the LAD also predicted risk even in patients with no plaque or calcium (32% of the cohort), Antoniades said (HR for cardiac mortality for highest vs. lowest quartile = 11.6; 95% CI, 3.51-38.21; P < .001; HR for major adverse cardiac events for highest vs. lowest quartile = 5.3; 95% CI, 3.06-9.18; P < .001).
In an independent validation process, predicted and observed events based on FAI-Score matched closely, he said, noting that those classified as very high risk or high risk had greater risk for events than those classified as low risk or moderate risk, regardless of whether they had obstructive CAD, no obstructive CAD or no plaque/calcium.
The AI risk model reclassified approximately 30% of patients to a higher-risk category and approximately 10% to a lower-risk category than they would have been in without a FAI-Score, according to the data presented.
“What we have learned from this study, at unprecedented scale ... is that this discovery which we made a few years ago ... is extraordinarily powerful,” Channon told Healio. “The signal that is derived from measuring coronary inflammation predicts the risk of heart attack and death over a 10-year period, above and beyond narrowing and blockages in the coronary arteries, [which] cardiologists currently use to make treatment recommendations for patients who have had a heart scan.”
U.S., U.K. and European guidelines currently recommend CT scans as a first-line test for patients with suspected CAD, but the advent of this technology means that “having a CT scan and purely looking at it visually ... will be an incomplete investigation,” Channon told Healio. “Because a very high percentage of those scans won’t have severe narrowings or blockages, and yet we now know that there is invisible inflammation going on in the arteries that can be detected using this new technology. Our expectation is that an integral part of the interpretation and the clinical use of cardiac CT scans will incorporate this new technology.”
CaRi-Heart Technology is being used in the U.K. and the European Union and regulatory work for entry into the U.S. and other global markets is currently underway, Frank Cheng, CEO of Caristo, told Healio.
Reference:
Perspective
Back to Top
Coronary CT angiography was originally validated around 2008 for identifying stenoses within the coronary arteries. However, further research has raised the important observation that most instances of ACS occur from nonobstructive plaque. Subsequent research identified plaque morphologies and quantities of noncalcified plaque as potential risk markers for future ACS; however, prediction of events remains difficult. Some combination of plaque-based risk, combined with patient-based risk typically manifesting as inflammation, which destabilizes existing plaque, was required to cause an event. However, the ability to detect coronary-specific inflammation was limited.
In 2017, Dr. Antoniades and his team published a seminal paper (Antonopoulos AS, et al. Sci Transl Med. 2017;doi:10.1126/scitranslmed.aal2658) showing that there was a bidirectional relation between the coronary arteries and the surrounding fatty tissue of inflammation, whereby the fat around the arteries appears different when there is inflammation in the artery. This technology could be applied to traditional coronary CT angiography, and for the first time, raised the possibility of adding inflammation-based information to plaque-based information.
It was subsequently trialed within a large cohort, published in 2018 (Oikonomou EK, et al. Lancet. 2018;doi:10.1016/S1040-6736(18)31114-0), showing that this measurement of fat attenuation improves prediction of cardiac mortality over other markers. Around this same time, there has been increasing success with drugs that specifically target inflammation to reduce cardiovascular events, including an FDA approval of low-dose colchicine in 2023.
However, the ability to measure both quantitative plaque and peri-coronary fat attenuation has been difficult due to it being a relatively manually intensive process, which has historically made it less feasible for routine clinical use. In this abstract, the researchers demonstrate that fat attenuation has significant independent predictive abilities for cardiac mortality, and also validate a combined score that integrates plaque, fat attenuation and risk factors, within a large prospective U.K. population with coronary CT angiography, demonstrating that this significantly helps improve risk prediction for heart attacks, and may reclassify patient risk. The take-home point is that we’re nearing an inflection point for coronary CT angiography, where we are able to leverage more information, including information about inflammation to more directly target the underlying mechanisms of cardiovascular events.
While high-sensitivity C-reactive protein does also measure inflammation in the body, the fat attenuation information is more localized and specific to the coronary arteries. I think an apt analogy would be an LDL measurement vs. plaque presence, in that while the blood marker is able to suggest and is associated with the disease process, what we typically care most about is the specific disease. That being said, we don’t yet know the degree of incremental benefit for treating based on coronary fat attenuation over more nonspecific markers, but I would predict that the direct visualization of coronary disease would be superior.
The integration of this extra information will turn the coronary CT results from a more binary result (yes/no obstructive disease) to something more subtle but more important to managing patient risk. In this way, it will benefit even those who do not have obstructive disease by identifying, classifying and quantifying risks that may lead to a heart attack independent of obstructive disease, and allow medical intervention that may be more effective at preventing future events.
As long as the workflow to obtain this additional information from CT scans becomes accessible from both a cost and feasibility standpoint, this will allow patients to receive more information from their coronary CT angiography, and potentially provide markers which we could directly target and acutely modify. This may improve our ability to prevent progression of CAD.
This research represents another significant step forward in more effectively utilizing diagnostic imaging through post-processing to improve patient care, and Dr. Antonaides’ group in particular has played a central role in identifying and developing this fat attenuation measurement. The effects of this, and particularly with growing ability to perform automated analysis of quantitative CCTA measurements of plaque and inflammation, will further drive coronary CT angiography as a touchstone within management and prevention of CAD. While this occurs, it is critical that we improve access to this technology (both physical and financial), both the ability to receive a CCTA, as well as the ability to receive the advanced post-processing which is just starting to emerge, in order to avoid exacerbating health care disparities. Given the major role of CAD within the health of American adults, CCTA is a technology that may end up being beneficial to a great many people at some time in their lives.
Collapse
Chan K, et al. LBS.08. Artificial intelligence at the bedside. Presented at: American Heart Association Scientific Sessions; Nov. 11-13, 2023; Philadelphia.
Click Here to Manage Email Alerts