Beyond coronary artery calcium score: can lipoprotein(a) and artificial intelligence-coronary computed tomography angiography unmask ‘invisible’ residual cardiovascular risk?

This editorial refers to ‘Lipoprotein(a) selectively associates with vulnerable coronary plaque phenotypes in comparison with other established risk markers’, by R. Fisher et al., https://doi.org/10.1093/ehjci/jeag024. For more than a decade, lipoprotein(a) (Lp(a)) has been unequivocally recognized as an independent, genetically determined risk factor for both atherosclerotic cardiovascular disease (ASCVD) and aortic stenosis.1 On this basis, the 2022 European Atherosclerosis Society (EAS) consensus statement recommended measuring Lp(a) at least once in adulthood to identify individuals at increased cardiovascular risk.2 The report by Fisher et al.3 builds on this foundation, providing further evidence for the strong association between elevated Lp(a) and high-risk coronary plaque morphology, and underscoring the potential of non-invasive coronary imaging to refine risk stratification in primary prevention of coronary artery disease (CAD) beyond traditional approaches. The traditional cornerstone of non-invasive coronary risk stratification has been the coronary artery calcium score (CACS).4 Although CACS robustly predicts total plaque burden, it has an important limitation: it cannot detect low-density, non-calcified ‘vulnerable’ plaques, which are the primary triggers of acute coronary syndrome. The Mount Sinai analysis of a large cohort of asymptomatic patients with a mean age of 56 years undergoing coronary computed tomography angiography (CCTA) for risk stratification—using artificial intelligence (AI)-driven plaque quantification and characterization—highlights a critical divergence. While CACS correlates strongly with overall plaque volume, it fails to predict the presence of low-density, non-calcified ‘vulnerable’ plaques. In contrast, Lp(a) emerges as a significant independent predictor of these high-risk plaque features, outperforming established markers such as LDL-cholesterol (LDL-C) particle concentration and high-sensitivity C-reactive protein.3 A divergent effect of Lp(a) and CACS on ASCVD risk was also observed in the analysis of long-term outcomes from the combined multi-ethnic study of atherosclerosis and Dallas Heart Study cohorts.5 Elevated Lp(a) levels and CACS were independently associated with increased ASCVD risk, even after adjusting for traditional risk factors. However, Lp(a) levels did not correlate with CACS, and elevated Lp(a) levels were associated with a higher risk among individuals with a CACS ≥100. Elevated Lp(a) increases cardiovascular risk through its unique structure, which promotes atherosclerosis, inflammation, and thrombosis.6 Oxidized phospholipids contribute to plaque formation and instability, while the apolipoprotein A component impairs fibrinolysis. These biological effects suggest that Lp(a) may be a critical driver of coronary atherosclerosis progression and cardiovascular risk, potentially exceeding the impact of LDL-C. An observational, single-centre study of 357 stable CAD patients using quantitative CCTA found that high Lp(a) levels (>60 mg/dL) were associated with increased total plaque volume, greater low-attenuation plaque, and abnormal pericoronary fat attenuation. These patients also exhibited higher computed tomography (CT) fractional flow reserve pullback pressure gradients, reflecting more diffuse disease. Furthermore, high Lp(a) was linked to worse cardiovascular outcomes; ∼40% of this prognostic effect was mediated by these high-risk features, and including Lp(a) and CCTA attributes in risk models significantly improved cardiovascular risk stratification.7 Further evidence from a 12-month serial CCTA study in patients with advanced multivessel stable CAD showed that high Lp(a) levels (≥70 mg/dL) were associated with faster progression of low-attenuation plaque. In contrast, no significant differences were seen in total, calcific, or non-calcific plaque progression between high- and low-Lp(a) groups.8 These results suggest that elevated Lp(a) increases residual cardiovascular risk by driving the growth of vulnerable plaques, even in patients receiving guideline-directed preventative therapies. This is reinforced by a long-term, 10-year prospective serial CCTA study of 299 patients with suspected chronic coronary syndrome. Over this decade, higher Lp(a) levels were associated with greater progression of coronary plaque burden, a higher presence of low-density non-calcified plaque, and pericoronary adipose tissue inflammation.9 Collectively, these studies suggest that the increased risk of coronary events associated with elevated Lp(a) is primarily driven by the formation of rupture-prone, low-density non-calcified plaques. These findings highlight a key mechanistic distinction between traditional and emerging risk markers. While CACS primarily reflects the ‘scar’ of atherosclerosis—stabilized, calcified lesions accumulated over time—Lp(a) is a superior predictor of vulnerable plaque in middle-aged people and, consequently, of the imminent risk of acute atherosclerotic complications. This distinction underscores their complementary value: CACS captures historical disease burden, whereas Lp(a) identifies the active, rupture-prone processes that drive clinical events. Until recently, only limited evidence from small studies supported adding CCTA data to risk stratification based on classical factors and CACS.10–12 However, a landmark analysis of the SCAPIS study—which followed 24 791 cardiovascular disease-free individuals aged 50–64 for a median of 7.8 years—demonstrated that CCTA data modestly, but significantly, improve coronary event prediction.13 By accounting for the extent of non-calcified and obstructive atherosclerosis, CCTA enabled a 14.2% upward reclassification of individuals who experienced events, with the most notable impact observed in the low-risk group. The future of risk assessment may lie in integrating high-resolution imaging with molecular markers. AI-enhanced plaque analysis now allows for more precise quantification of disease from CCTA. Furthermore, the advent of photon-counting detector CT promises even greater accuracy due to its superior spatial resolution and tissue differentiation.14 By precisely delineating plaque subtypes—particularly low-density, non-calcified components—this technology may refine risk stratification beyond traditional metrics. When combined with Lp(a) measurements, this approach could enable more accurate cardiovascular risk prediction, supporting targeted interventions such as aggressive lipid management or emerging Lp(a)-specific therapies. Prospective studies are required to confirm the incremental prognostic value of this integrated approach and to establish its place in routine clinical workflows. The forthcoming results of the SCOT-HEART 2 trial (NCT03920176) and the innovative TRANSFORM trial (NCT06112418)—which incorporates AI-driven analysis and the formal staging of coronary plaques—will provide critical data on CCTA-guided prevention; subsequent post hoc analyses of these cohorts may provide further answers regarding the role of Lp(a) as a driver of active coronary disease. Ultimately, incorporating Lp(a) testing—aligned with the 2022 EAS recommendations—alongside advanced imaging could facilitate truly personalized preventive strategies, mitigating the substantial residual risk in high-Lp(a) individuals and reducing the global burden of ASCVD. Christiaan J. M. Vrints (Writing—original draft [lead])), and Emeline M. Van Craenenbroeck (Writing—review & editing [equal])) This is an editorial, and there is no new data to share.