AI-Assisted Look at Coronary Inflammation on CT Refines CV Risk Assessment

An artificial intelligence (AI)-assisted algorithm that incorporates information about inflammation detected on coronary CT angiography (CCTA) may help identify patients who require more-intensive treatment, especially among the vast majority who have no obstructive disease detected, results of a large, population-based study show.

A high fat attenuation index (FAI) score, indicative of increased coronary inflammation, was associated with significantly heightened risks of cardiac mortality and MACE through several years of follow-up, even after accounting for other cardiovascular risk factors and the presence or extent of CAD, report researchers from the Oxford Risk Factors and Noninvasive Imaging (ORFAN) study.

Moreover, the addition of this metric improved on the prediction of events with a combination of the validated QRISK3 score and the CAD-RADS 2.0 assessment of stenosis severity.

“Detecting and quantifying coronary inflammation reveals an enormous potential for detecting people at risk who are not thought to be at high risk and reclassifying them to receive appropriate advice and therapy,” study author Keith Channon, MBChB, MD (University of Oxford, England), told TCTMD.

Even among the patients who had no CT-detected coronary disease at all, the investigators showed that increased inflammation was predictive of poorer clinical outcomes, meaning everyone who undergoes a coronary CT scan has the potential to benefit from an assessment of inflammation, Channon said.

The 80% of patients with nonobstructive disease “stand to gain enormously from checking and quantifying whether they have coronary inflammation at a high level,” he stressed. “If they do, they’re at very high risk, even if they have no plaque, and they should be stratified and treated as a high-risk population. And that’s up to 40% of patients who are currently not being picked up and—if untreated—we know from this cohort have a very high cardiovascular event rate over the next decade.”

Channon is co-founder and chief medical officer of Caristo Diagnostics, the company that developed the CaRi-Heart technology used in the study to detect coronary inflammation, calculate the FAI score, and incorporate that information with plaque burden and traditional risk factors. CaRi-Heart is approved for use in the UK, Europe, and Australia but has not yet been cleared in the United States.

The findings were published online this week in the Lancet, with lead authors Kenneth Chan, MBBS, and Elizabeth Wahome, PhD, and senior author Charalambos Antoniades, MD, PhD (all from the University of Oxford).

Commenting for TCTMD, Andrew Choi, MD (George Washington University School of Medicine, Washington, DC), said he and many others had been waiting for this paper. There has long been evidence that inflammation plays a role in the development of acute cardiac events, with large prospective trials—such as CANTOS and LoDoCo2—showing that anti-inflammatory therapies can reduce such events, Choi pointed out.

“One of the challenges of inflammation is that the additive value of inflammation is pretty small and that the primary process of development for heart attack comes from atherosclerosis, plaque, plaque burden, and . . . high-risk plaque features,” he explained. “I think the identification of inflammation by CT with [perivascular] FAI is exciting; it has a very large evidence base. I think it’s incremental . . . on top of what we learn from identification of obstructive or nonobstructive CAD, identification of total plaque burden, and plaque characterization, as well as conventional risk factors and age.”

Choi indicated that coronary CT angiography more generally is helpful to guide use of appropriate preventive therapies, including lipid-lowering therapies, anti-inflammatories, and other agents.

“CTA is the most uniquely positioned to act as a gateway to allow for proper allocation and use and potential layering of these medications to allow for really individualized prevention, and to do this in a way that is high-value,” he said. “But also, most importantly, we’re identifying those patients that we are not capturing by using our conventional cardiovascular risk scoring. And this study is really evidence of that.”

The ORFAN Study

Coronary CT angiography is an increasingly important first-line option in the investigation of possible CAD, and the results therefore underly much decision-making around patient management, Channon said. Although historically the focus on the assessment of CAD has been on the extent of obstructive disease, most patients don’t have blockages, he added.

The ORFAN study focused on 40,091 patients (median age 59; 47% women) who underwent clinically indicated CCTA at eight UK hospitals between 2010 and 2021 and were followed for a median of 2.7 years for MACE (MI, new-onset heart failure, or cardiac death).

Overall, 81.1% of patients didn’t have obstructive CAD on the initial CT scan. Even though their overall risk of adverse outcomes was lower compared with patients with obstructive disease, these patients accounted for 66.3% of MACE and 63.7% of cardiac deaths during follow-up.

This “adds enormously to the emerging evidence that stenosis and occlusion of coronary arteries . . .  are not the only important things,” Channon said. “In fact, they may even be [less important] than trying to work out what else is driving the adverse outcomes for the majority of patients who are referred for investigation.”

The researchers then examined how the level of coronary inflammation influences risk of CV events, and how well the AI-assisted algorithm incorporating this information along with plaque burden and risk factors—called AI-Risk—performed as a prognostic tool in a subset of 3,393 patients (median age 62; 44% women) from the two hospitals with the longest follow-up (median 7.7 years).

The FAI score was strongly and independently associated with cardiac mortality and MACE in the overall group, in the patients with obstructive disease, and in the patients with nonobstructive disease. In addition, patients placed in the very-high-risk category using the AI-Risk classification had significantly greater risks of cardiac mortality (HR 6.75; 95% CI 5.17-8.82) and MACE (HR 4.68; 95% CI 3.93-5.57) when compared with those at low or medium risk, with consistent results regardless of the presence of obstructive CAD.

Channon said the addition of an assessment of inflammation is “transformational” because it provides for a better view of risk compared with traditional methods. To highlight the potential impact, he pointed to the results of pilot implementation studies indicating that when given the results of the inflammation assessment, clinicians changed their treatment recommendations in 45% of cases. That included initiation of statin treatment in 24%, an increase in statin dosage in 13%, and addition of other treatments like aspirin, colchicine, and icosapent ethyl (Vascepa; Amarin) in 8%.

Where Inflammation Fits in the Assessment

Choi said this study “does move the needle in application of [perivascular] FAI or this AI-Risk calculator,” adding, “I think the costs are an important part to work out for a large healthcare system like the [National Health Service]. CT is first line because it has shown high value. It improves outcomes as shown in the SCOT-HEART trial, but it also reduces overall healthcare costs, and I think this approach will have to also show the same level of value to NHS for further adoption.”

He pointed out that an assessment of coronary inflammation is not the only approach that is being evaluated to enhance the information that can be obtained from CCTA, with AI-assisted measures of total quantitative plaque burden and assessments of flow and flow surrogates, for example, also being investigated.

“This study is really good evidence of the prognostic value of perivascular inflammation,” but the question is, he said, which additional assessment, or combination of assessments, is needed in the context of a high-value test like CCTA.

As for the AI-Risk measure of coronary inflammation, Channon said “it’s not just reporting more precisely what we already can see on a CT scan,” noting the CaRi-Heart technology is the only one that can detect and quantify coronary inflammation from CT. “So we see this as a real game changer for detecting people at high risk and stratifying them so that the highest-risk people get the most-intensive treatments.”

It may also prove useful, Channon added, for pharmaceutical companies looking for patients with a high level of coronary inflammation to include in trials of novel anti-inflammatory therapies and for monitoring a patient’s response to drugs that target inflammation.

He noted that there is a large trial planned to see whether use of the AI-Risk tool improves patient outcomes, although it hasn’t started yet.