Highlights
- The risk factor-weighted clinical likelihood (RF-CL) model represents a major refinement over older age-sex-symptom pre-test probability tools for estimating the probability of obstructive coronary artery disease (CAD) in patients with suspected chronic coronary syndromes (CCS).
- By incorporating five clinical risk factors, and coronary calcium information when available, the model improves calibration in contemporary populations and can safely identify many patients at very low likelihood in whom further testing may be deferred.
- Its strengths are simplicity, external validation, and clinical utility for front-end triage; its limitations include subjective symptom classification, imperfect handling of women and non-obstructive ischemia, and a semi-quantitative upper range that is less transparent for very high likelihood estimates.
- The RF-CL model should be viewed as an initial gatekeeper for obstructive CAD testing, not as a comprehensive predictor of ischemia, plaque vulnerability, microvascular dysfunction, or total cardiovascular risk.
Background
Estimating the pre-test probability of obstructive CAD is a foundational step in evaluating patients with chest pain or other symptoms suggestive of CCS. Historically, clinicians relied on age, sex, and symptom typicality. Those classic models were attractive because they were simple, but they progressively became poorly calibrated in modern practice. Several factors explain this drift: better risk factor control, broader use of preventive therapy, referral pattern changes, declining prevalence of flow-limiting epicardial stenosis among stable outpatients, and increased recognition of ischemia with non-obstructive coronary arteries (INOCA), especially in women.
Overestimation of disease prevalence has practical consequences. It can drive excessive downstream testing, increase false positives, expose patients to radiation or contrast, and overload imaging and catheterization services. Underestimation, by contrast, risks delayed diagnosis in patients with genuine obstructive disease. Thus, contemporary guideline-directed care requires a model that is sufficiently simple for routine clinical use, yet better aligned with present-day populations.
The 2024 European Society of Cardiology (ESC) guidelines for chronic coronary syndromes endorse a structured estimate of disease probability after history taking. The RF-CL approach combines age, sex, symptom characteristics, and five clinical risk factors, with coronary artery calcium data used as an optional enhancer where available. In their 2026 European Heart Journal “Great Debate,” Andreotti and colleagues argue that the new RF-CL model is useful, while also acknowledging important practical criticisms. That debate is clinically relevant because the model now influences which patients are reassured, which undergo coronary CT angiography (CCTA) or functional testing, and which may proceed directly to invasive angiography.
Key Content
From traditional pre-test probability to RF-CL: why change was necessary
Older pre-test probability frameworks were largely built from historical angiography cohorts enriched for obstructive disease. In modern ambulatory populations, especially those evaluated with CCTA, the prevalence of obstructive CAD is substantially lower. This mismatch led to systematic overestimation when legacy tools were applied unchanged.
The central conceptual advance of RF-CL is that symptom type alone is not enough. Age, sex, and pain phenotype remain important, but the probability of epicardial obstructive CAD also depends materially on the burden of conventional risk factors. By incorporating these variables, the RF-CL model moves from a coarse phenotype-based tool toward a more clinically realistic estimate.
This recalibration is especially relevant in three groups: first, younger patients with chest pain but little risk factor burden, in whom older models often overstated disease likelihood; second, older patients with apparently atypical symptoms but multiple risk factors, in whom disease may still be substantial; and third, women, among whom symptom descriptors and anatomic disease burden frequently diverge from traditional male-derived assumptions.
What the RF-CL model includes and how it is used
As summarized by Andreotti et al., the model uses age, sex, symptom characteristics, and five clinical risk factors, with coronary calcification data added if available. The output is a numerical estimate of the initial likelihood of obstructive CAD. Up to roughly 45%, the estimate is quantitative; above that, the framework becomes more semi-quantitative or qualitative. At the very high end, an estimated probability above 85% supports invasive coronary angiography in guideline pathways, although one of the critiques raised in the debate is that reaching this threshold in routine practice may not always be operationally transparent.
Clinically, the model functions as an entry-point triage tool. Patients at very low likelihood can be considered for deferral of further diagnostic testing, provided the clinical context is stable and there are no high-risk features from history, examination, or ECG. Intermediate-likelihood patients are candidates for noninvasive anatomic or functional testing, often CCTA in contemporary practice. High-likelihood patients may warrant more direct escalation depending on symptom severity, treatment response, and suspected disease burden.
Why the model is useful: evidence synthesis from derivation, validation, and guideline adoption
The case for RF-CL rests on three linked strengths.
First, improved calibration in contemporary populations. Compared with earlier age-sex-symptom models, RF-CL better reflects the current prevalence of obstructive CAD among patients presenting with suspected CCS. This matters more than small differences in discrimination alone. In triage models, calibration drives whether too many or too few patients are sent for testing.
Second, broad external validation. As highlighted in the debate article, the model has been externally validated across European, North American, and Asian cohorts. This geographic breadth is important because symptom reporting, referral thresholds, preventive therapy exposure, and disease prevalence all vary by healthcare system. A model that performs reasonably across regions is inherently more credible for guideline use.
Third, practical de-risking of symptomatic patients. One of the model’s strongest clinical contributions is not simply finding more disease, but identifying those unlikely to have obstructive CAD. That front-end reassurance can reduce unnecessary testing and reallocate resources toward patients in whom imaging or angiography is more likely to change management.
In other words, the RF-CL model offers a pragmatic balance: it captures more clinically relevant information than historical tools, but remains rapid enough for outpatient use. That trade-off likely explains its guideline success.
Where the debate is justified: the main limitations in routine practice
Despite these gains, the criticisms summarized by Andreotti et al. are substantial and deserve close attention.
1. Symptom classification remains subjective. The distinction between typical angina, atypical angina, and non-anginal chest pain is less robust than most probability models imply. Many patients struggle to describe their symptoms precisely; clinicians vary in interpretation; and women, older adults, and patients with diabetes often present with less canonical symptom patterns. A model that relies on symptom categories inherits that variability.
2. Risk factor definition and weighting can be clinically messy. The presence of hypertension, diabetes, dyslipidemia, smoking, or family history is easy to list, but less easy to standardize. Is controlled hypertension weighted the same as longstanding resistant hypertension? Does former smoking count the same as active smoking? Are treated dyslipidemia and untreated severe hypercholesterolemia equivalent? Practical models usually compress such distinctions, which preserves usability but sacrifices granularity.
3. The model predicts obstructive anatomy, not ischemia. This is arguably the most important conceptual limitation. Patients may have angina because of microvascular dysfunction, vasospasm, endothelial dysfunction, or diffuse non-obstructive atherosclerosis. Such patients may have a low RF-CL estimate for obstructive CAD yet still have clinically important ischemic syndromes. Therefore, a low obstructive CAD likelihood cannot be interpreted as “symptoms are benign” without broader clinical judgment.
4. The upper probability range is less transparent. The model becomes semi-quantitative above about 45%, and the pathway recommending invasive angiography for >85% likelihood has been criticized because, in many routine scenarios, it is not obvious how the score reaches that threshold. This can reduce clinician confidence precisely where consequences are greatest.
5. Enrichment factors remain partly qualitative. Guidelines often acknowledge that additional findings may “upgrade” likelihood: abnormal resting ECG, left ventricular dysfunction, known extracardiac atherosclerosis, severe risk factor burden, or abnormal prior testing. These are clinically sensible, but if the enrichment process is insufficiently standardized, the seeming precision of the base score may be undermined by discretionary adjustment.
Women, atypical symptoms, and INOCA: where RF-CL must be interpreted carefully
The debate around symptom specificity is especially important in women. A 2026 review on premature CAD in women emphasized that diagnostic delay is common because women more often present with fatigue, nausea, dyspnea, or mixed symptom clusters and may have non-obstructive coronary abnormalities or spontaneous coronary artery dissection rather than classic focal obstructive plaque. Similarly, sex-specific literature on lipoprotein(a) and early CAD suggests that biological and phenotypic heterogeneity in women is not fully captured by conventional symptom-risk models.
These observations do not invalidate RF-CL; rather, they define its scope. RF-CL is useful for estimating obstructive epicardial CAD. It is not a sex-specific ischemia model, and it should not be used to dismiss persistent or high-impact symptoms in women whose disease substrate may differ from classic obstructive stenosis.
Residual risk and the limits of a parsimonious model
Another line of evidence from recent PubMed-indexed studies supports the idea that parsimonious clinical scores cannot capture the full biology of CAD. For example, metabolomic work in patients with CAD but no traditional risk factors identified elevated unsaturated fatty acids as a potential residual-risk signature. Other studies have reported associations between CAD and circulating MALAT1, ApoB/ApoA-I ratio, or lipoprotein(a), sometimes with sex-specific effects. These studies are not ready for routine frontline triage, but they illustrate an important principle: the biology of CAD extends beyond age, sex, symptoms, and a handful of conventional risk factors.
From a translational standpoint, RF-CL should therefore be viewed as a population-level triage model, not as a precision-medicine tool. It helps decide who likely needs imaging today; it does not fully characterize plaque biology, inflammatory activation, microvascular disease, or future event risk.
The role of coronary artery calcium: simple enhancer, powerful reclassification
A particularly important feature of the RF-CL construct is the optional incorporation of coronary calcium information. Coronary artery calcium scoring is attractive because it directly reflects cumulative coronary atherosclerotic burden and can substantially reclassify likelihood estimates when symptoms and risk factors alone are ambiguous.
In patients with low-to-intermediate baseline RF-CL, a calcium score of zero may support conservative management in the appropriate clinical setting, whereas substantial calcification can shift the pre-test estimate upward and strengthen the indication for CCTA or further evaluation. Mechanistically, calcium provides an anatomic anchor that pure clinical models lack. Practically, it also helps reduce dependence on symptom semantics.
That said, calcium is not perfect. Younger patients, especially smokers or those with soft non-calcified plaque, may still have clinically significant disease despite low calcium burden. Conversely, heavy calcification may overstate the probability of flow-limiting lesions. Thus, calcium improves RF-CL, but does not replace downstream imaging when indicated.
Clinical workflow implications
In day-to-day practice, the model is most helpful when used to answer a narrow question: how likely is obstructive CAD before selecting a diagnostic test?
A reasonable workflow is as follows:
- Take a structured history and identify symptom pattern, timing, and exertional relationship.
- Estimate RF-CL using age, sex, and conventional risk factors.
- Integrate calcium data if already available or easy to obtain within the local pathway.
- If likelihood is very low and there are no red flags, consider deferring further CAD testing and addressing alternative diagnoses.
- If likelihood is intermediate, favor noninvasive testing, often CCTA.
- If likelihood is high, especially with treatment-refractory symptoms or high-risk clinical features, escalate to more definitive testing and consider invasive assessment.
The model is less useful when clinicians attempt to extract from it judgments it was not designed to provide: whether a patient has microvascular angina, whether symptoms are ischemic despite no obstructive disease, or whether future myocardial infarction risk is high.
Expert Commentary
The 2026 “Great Debate” is valuable because it avoids two common errors: uncritical enthusiasm and blanket dismissal. The pro-RF-CL position is persuasive. Compared with older models, RF-CL is clearly a methodological improvement. It incorporates readily available data, aligns better with modern disease prevalence, and supports a crucial stewardship goal in cardiology: reducing unnecessary testing without missing important obstructive disease.
Yet the critics are also right. Probability models only appear objective when their inputs are objective. Symptom characterization is not. The definitions and weights of conventional risk factors are only partially standardized. The enrichment factors that modify baseline likelihood can reintroduce subjectivity. And the model’s shift from quantitative to semi-quantitative estimates at higher probabilities is conceptually awkward, especially if invasive angiography decisions hinge on those upper categories.
Perhaps the most important expert takeaway is that RF-CL solves a narrower problem than many clinicians or health systems may want it to solve. It estimates the likelihood of obstructive CAD, not the likelihood of clinically meaningful myocardial ischemia in general. This distinction is not academic. A large proportion of symptomatic patients, particularly women, patients with diabetes, and those with endothelial or microvascular dysfunction, may have disabling ischemic symptoms without obstructive epicardial stenosis. If RF-CL is used too rigidly, these patients risk being under-investigated or mislabeled.
From a health-system perspective, however, a simple model has major advantages. More complex tools may yield marginally better statistical performance, but often fail in real clinics because they are too cumbersome, poorly remembered, or inconsistently applied. RF-CL appears to have found a pragmatic middle ground. It is good enough to improve care, even if not comprehensive enough to settle every diagnostic ambiguity.
Future refinement should probably move in three directions. First, better standardization of symptom descriptors, potentially with digital questionnaires, could improve reproducibility. Second, sex-aware and phenotype-aware refinements may help address women, younger patients, and INOCA populations. Third, low-burden biological enrichers—such as calcium, selected biomarkers, or imaging-derived plaque features—may eventually enhance prediction without making the tool unwieldy.
Conclusion
The new RF-CL model is useful, and the 2024 ESC recommendation to adopt a structured probability estimate for suspected CCS is well justified. The model improves upon older pre-test probability tools by incorporating risk factor burden, calibrating better to contemporary obstructive CAD prevalence, and helping identify many patients who can safely avoid unnecessary testing.
Its limitations are equally important. RF-CL is not a universal chest pain algorithm, not a predictor of ischemia without obstruction, and not a substitute for clinician judgment. Symptom ambiguity, semi-quantitative high-end categories, and incomplete capture of sex-specific and non-obstructive phenotypes mean the model should guide, not dictate, decision-making.
In practical terms, RF-CL is best understood as an efficient first-step gatekeeper for suspected obstructive CAD. Used in that role, it can de-risk low-likelihood patients, rationalize test selection, and improve resource stewardship. Used beyond that role, it risks over-precision and under-recognition of important syndromes outside obstructive epicardial disease. The real achievement of RF-CL is not that it is perfect, but that it is more clinically honest and more operationally useful than the models it replaces.
References
- Andreotti F, Thiele H, Vrints C, Jobs A, Winther S, Gröninger J, Capodanno D, Desch S, Nedios S, Gorog DA. Great debate: the new risk factor-weighted clinical likelihood model is useful to estimate the initial pre-test probability of obstructive coronary artery disease in individuals with suspected chronic coronary syndromes. Eur Heart J. 2026;47(22):2777-2792. PMID: 41823519.
- 2024 ESC Guidelines for the management of chronic coronary syndromes. European Society of Cardiology, 2024. PubMed-indexed guideline article.
- Mechanisms linking hypertension to cardiovascular and cerebrovascular diseases and their clinical implications: A comprehensive review. Clin Exp Hypertens. 2026. PMID: 41721701.
- Elevated circulating unsaturated fatty acids: A novel metabolic feature of coronary artery disease without traditional risk factors. Clin Exp Hypertens. 2026. PMID: 41952487.
- The ApoB/ApoA-I ratio supersedes conventional lipids in predicting coronary artery disease and clinical phenotypes requiring revascularization. Clin Exp Hypertens. 2025/2026 issue. PMID: 41430776.
- Elevated MALAT1 expression predicts coronary artery disease severity as a potential biomarker for risk stratification: a cross-sectional study. Ann Med. 2026. PMID: 42059375.
- Premature coronary artery disease in women: sex-specific risk factors, pathogenetic mechanisms and clinical implications. Ann Med. 2026. PMID: 41614676.
- Associations of lipoprotein(a) concentrations with cardiovascular disease in men and women with primary hypercholesterolemia. Atheroscler Plus. 2026. PMID: 42099700.
