From stenosis to atheroma: repositioning lipoprotein(a) in coronary artery disease risk assessment

This editorial refers to ‘Incorporation of Lipoprotein(a) levels improves calibration of pre-test likelihood estimates of obstructive coronary artery disease’, by G.S. Brix et al., https://doi.org/10.1093/ehjci/jeag021. Angina pectoris was first described in 1772 by William Heberden and later remarked by Dr. John Warren, the Latin angina, ‘infection of the throat’; from the Greek ἀγχόνη, ‘strangling’; and from the Latin pectus, ‘chest’, at a time when understanding of disease pathology was predominantly grounded in clinical observation and anatomical dissection.1 Advances in medical technology through animal studies, electrocardiography, echocardiography, and other imaging modalities such as invasive coronary angiography (ICA), intravascular imaging, and coronary computed tomography angiography (CCTA) complemented by clinical observations and post-mortem anatomical findings have since provided pivotal understanding of the pathophysiology and natural history of coronary artery disease (CAD) beyond the symptomatology itself. Given the acuity and catastrophic consequence of cardiovascular events, risk stratification is critical for prevention and diagnosis. The modern cardiovascular risk stratification approach recommended by clinical guidelines began with the Framingham Heart Study in 1948. Cardiologists, biostatisticians, and epidemiologists collaborated in a joint effort which allowed them to demonstrate that elevations in blood pressure and cholesterol levels were associated with increased incidence of ischaemic heart disease and acute myocardial infarction, often for women occurring later in life than for men.2 The identification of these risk factors unveiled opportunities for risk stratification and preventive interventions for a gradual reduction in cardiovascular mortality and morbidity in the developed countries. Several years later, the advent of ICA that confirms the pathological connotation of coronary obstruction and angina/cardiac ischaemia and later the availability of revascularization techniques either surgically through coronary artery bypass grafting or percutaneous coronary intervention shifted the focus from atherosclerotic disease to ischaemia-causing coronary obstruction. To facilitate adequate resource allocation of ICA, Diamond and Forrester developed the first model for calculating the pre-test-probability of finding obstructive CAD on ICA.3 The model was imperfect yet rapidly adopted and improved by implementing conventional clinical risk factors. The current risk factor-weighted clinical likelihood model (RF-CL), recommended by the ESC guidelines for chronic coronary syndromes,4 aids in the management of patients with de novo anginal symptoms by predicting [area under curve (AUC) 74.9] the risk of obstructive CAD based on symptoms, age, sex, and classical risk factors (family history, smoking, dyslipidaemia, hypertension, and diabetes). Hereby, the RF-CL model functions as a gatekeeper tool to more costly non-invasive testing and ICA. However, emerging evidence on novel risk factors require updates and recalibrations of these models to contemporary cohorts. Brix et al.5 evaluate whether incorporating lipoprotein(a) into the RF-CL model improves estimations of obstructive CAD. Two cohorts encompassing patients with de novo chest pain, Dan-NICAD as a derivation cohort and PROMISE for external validation, were used to assess the impact of Lp(a) on risk for obstructive disease. Participants from the Dan-NICAD cohort underwent CCTA and if significant stenosis were observed (>50% in ≥2 mm diameter vessels) subsequent fractional flow reserve guided ICA. In contrast, in the PROMISE cohort a selection was made of patients that underwent either CCTA or ICA. The authors develop several models, Lp(a) alone, RF-Lp(a) model with Lp(a) embedded as a continuous predictor, and a simple model with a multiplicative rule (RF-CLLp(a), where the predicted risk is multiplied by 1.5 based on the observed relative risk when Lp(a) ≥ 125 nmol/L), and compare them against the original RF-CL in terms of discrimination (AUC), calibration, and reclassification. The methodology is rigorous, the cohorts are large with minimal missing data, and the practical simplicity of the proposed adjustment is appealing. The authors demonstrate that RF-CL systematically underestimates obstructive CAD risk in patients with elevated Lp(a) ≥ 125 nmol/L, and that a simple multiplicative correction largely resolves this miscalibration. Although the adjustment for elevated Lp(a) appropriately up-classified patients that had obstructive CAD, the statistical significance was only met in the derivation cohort were re-classification affected fewer than 7% of patients. However, the incremental discriminatory value is modest: absolute differences in AUC between models were ≤0.02 in both cohorts, with overlapping confidence intervals. This is conceptually surprising, as Lp(a) has long been recognized as a causal independent risk factor for cardiovascular events. Nevertheless, as Brix et al. have shown, lipoprotein(a) does not perform well to predict the presence of coronary stenosis. This observation warrants careful interpretation, not as a failure of the biomarker, but as a signal that in terms of CAD risk the community might be asking the wrong question. The RF-CL model, while clinically useful, does not adequately address the fundamental course of atherosclerotic CAD (ACAD): ACAD is a lifelong, progressive condition that begins well before symptoms arise. For decades, CAD has been understood through the paradigm of epicardial obstruction causing myocardial ischaemia resulting in angina or anginal equivalents that move patients to consult physicians. These consultations often result in diagnostic testing which lead to interventions—medically or interventionally. Identification and management predominantly focus on coronary ischaemia; yet, the dominant mechanisms for myocardial infarction are atherosclerotic plaque rupture or erosion, events mostly emerging from non-obstructive lesions and thus at earlier stages of the disease that would not trigger the RF-CL diagnostic cascade. A recent analysis of nearly 5 million US patients with a first myocardial infarction revealed sobering evidence of this disconnect: half had no documented antecedent symptoms, and approximately one in five had no previously identified standard modifiable risk factors.6 Moreover, the majority who visited a physician prior to their infarction were not receiving any preventive medical therapy. Even those with identified risk factors and symptoms. These data expose the inadequacy of current symptom-triggered, stenosis-focused approach and indicate room for improvement as by the time obstructive CAD is detected, the window for optimal prevention has passed. The distinction between treating symptoms and preventing events deserves emphasis. Revascularization primarily alleviates angina as an acute phenotype of the disease; it does not prevent myocardial infarction or reduce mortality in patients with stable disease. The therapies altering the natural history of ACAD like antihypertensive drugs, statins, and other emerging lipid-lowering agents work by modifying factors associated with the underlying atherosclerotic process, not by addressing stenosis. It is therefore unsurprising that Lp(a), a biomarker that is associated with risk modification, demonstrates only modest discriminatory improvement when evaluated against anatomical obstruction in a low-risk population. According to the European Atherosclerosis Society consensus, Lp(a) functions primarily as a causal risk factor and the risk is amplified in the presence of elevated LDL cholesterol, inflammatory markers such as high-sensitivity C-reactive protein and hyperglycaemia.7 Evaluating Lp(a) against the endpoint of obstructive CAD therefore represents a conceptual mismatch: the biomarker is being asked to predict anatomy when its strength reasonably lies in predicting atheroma burden (and maybe subsequent events). To capture the natural history of ACAD, the path forward requires a paradigm shift from symptom-guided to screening-guided prevention.8 Current risk models, including RF-CL, are derived from and applicable to symptomatic individuals and do not identify the asymptomatic patient with progressive ACAD. Bridging this gap will require innovative strategies where genetic predisposition, including polygenic risk scores, biomarkers not in isolation but as comprehensive panels (CRP, LDL cholesterol, Lp(a), and HbA1c) and early non-invasive imaging play more prominent roles capture the multidimensional nature of cardiovascular risk (Figure 1). The SCOT-HEART trial demonstrated that CCTA-guided care improved cardiovascular outcomes over standard care, with benefits persisting beyond 10 years, largely through earlier initiation of preventive therapies.9 Finally, imagomics extracts imaging features imperceptible to the human eye, potentially enabling earlier and more precise cardiovascular risk stratification.10 History reminds us that the Framingham Heart Study transformed prevention through collaboration. A similar collaborative effort across disciplines is pivotal to shift from stenosis to atherosclerosis, with funding that matches the global burden of disease. Required paradigm shift in natural history of coronary artery disease for optimal prevention strategies. A paradigm shift from symptom-triggered intervention to early atherosclerosis detection. Atherosclerotic coronary artery disease progresses along a continuum from early risk factor exposure through subclinical plaque formation to eventual obstruction and clinical events. Current approaches using the risk factor-weighted clinical likelihood model (RF-CL) operate at the symptomatic stage, missing opportunities for early intervention. Comprehensive assessment using biomarker panels, genetic risk scores, and early imaging could identify at-risk individuals decades earlier. Data from Nurmohamed et al.6 demonstrate that half of patients with first myocardial infarction had no documented antecedent symptoms, highlighting the inadequacy of symptom-triggered approaches. CCTA, coronary computed tomography angiography; CRP, C-reactive protein; HbA1c, glycated haemoglobin; LDL-C, low-density lipoprotein cholesterol; Lp(a), lipoprotein(a); MI, myocardial infarction. Brix et al. have provided valuable data demonstrating that Lp(a) is independently associated with obstructive CAD and that a simple multiplicative adjustment improves calibration in patients with elevated levels. These findings have practical utility for clinicians by demonstrating individuals with high Lp(a) levels face a higher risk for obstructive ACAD. Yet the study results also serve as an illustration of why the field must evolve beyond coronary obstruction and ischaemic endpoints. The question is not whether Lp(a) helps predict stenosis, it is whether including it in a comprehensive risk assessment can enhance identification of individuals with greater atheroma burden subjected to potential destabilization that in the end lead to cardiovascular events while there is still time to alter their risk trajectory. As the Lancet Commission on CAD recently articulated, the future of cardiovascular risk assessment and subsequent prevention lies not in refining our detection of obstruction, but in halting, and ultimately eradicating the atherosclerosis that underlies it.11 Mick Renkens (MD (Conceptualization [lead]; Investigation [lead]; Visualization [lead]; Writing—original draft [lead]; Writing—review & editing [lead])), John Tsai (MD, PhD (Conceptualization [equal]; Writing—original draft [equal]; Writing—review & editing [equal])), and Patrick W. Serruys (MD, PhD (Conceptualization [equal]; Supervision [lead]; Visualization [equal]; Writing—original draft [equal]; Writing—review & editing [equal]))