AI-enhanced detection of coronary inflammation predicts 10-year risk for cardiac mortality and MACE in individuals with no or minimal coronary atherosclerosis on routine CCTA

Abstract Background Coronary CT angiogram (CCTA) is one of the first line investigations for patients with chest pain suspected of coronary artery disease (CAD). Current CCTA interpretations mainly focus on luminal obstruction, which is identified in the minority of patients, whereas the majority who have no or minimal atherosclerotic plaques visible on CCTA are often provided with reassurance. However, residual risk from inflammation that drives atherosclerosis and plaque vulnerability are not captured with luminal stenoses alone. Fat attenuation index (FAI) Score quantifies coronary inflammation on routine CCTA, and integrate with clinical risk factors/plaque burden in an artificial intelligence enabled risk algorithm (AI-Risk) that could help to risk stratify this patient group. Purpose To evaluate the prognostic value or FAI Score and AI-Risk in the risk stratification of patients with no or minimal atherosclerosis. Methods In a nested cohort within the Oxford Risk Factors and Non-invasive imaging (ORFAN) study, consecutive patients (n=3,393) who underwent routine clinical CCTA were followed up over a median(IQR) 7.7(6.4-9.1) years for cardiac mortality and major adverse cardiac events (MACE, including myocardial infarction, new onset heart failure, cardiac mortality). CADRADS 2.0 classification 0 or 1 were used to define no or minimal luminal stenoses. FAI Score was calculated for each of the 3 epicardial coronary arteries. Results A total of 1,678 patients had no or minimal atherosclerosis on CCTA [median (IQR) age 53(45-63), 49.8% male]. Cardiac mortality occurred in 3.1% and MACE occurred in 8.8% of the patients during the follow-up period. Patients at the lowest quartile of FAI Score or low/medium AI-Risk classification showed very low event rate during follow up. Patients with the highest quartile of LAD FAI Score showed 9 times higher risk of cardiac mortality (Panel A) and 5 times higher risk of MACE (Panel B) compared to those at lowest quartile. Similar findings were observed for LCX and RCA. Finally, integrating coronary inflammation with clinical risk factors and plaque burden, patients at very high AI-Risk classification showed 5 times higher risk of cardiac mortality (Panel C), and 4 times higher risk of MACE (Panel D) compared to those with low/medium risk classification. Conclusion Coronary artery inflammation measured by FAI Score could stratify the risk of patients with no or minimal coronary atherosclerosis on CCTA, and could help to guide early preventative treatments.