Coronary artery calcium: from risk prediction to treatment allocation and clinical trials

CAC and traditional risk factors

Atherosclerosis and CVD are complex multifactorial processes, which are rarely caused by any single risk factor. Commonly used CVD risk prediction models only use the measurement of CVD risk factors at a single point in time and do not take into account an individual’s dietary and exercise habits. Furthermore, while some genetic alleles and genetic diseases have been identified, these only provide a small sample of an individual’s overall genetic makeup, and the clinical utility of polygenic risk scores remains uncertain. Therefore, while traditional CVD risk scores work reasonably well on a population level, they work less well to estimate an individual’s CVD risk. In contrast, CAC scoring effectively integrates an individual’s lifetime positive and negative risk factor exposures along with their genetic susceptibility and/or resilience to provide an individualised quantification of CVD risk.4

Accordingly, CAC is a superior predictor of CVD events and individual CVD risk compared with standard clinical measurement of traditional risk factors and novel risk factors such as high-sensitivity C reactive protein, carotid intimal medial thickness and ankle-brachial index.5 It also outperforms American College of Cardiology (ACC)/American Heart Association (AHA)-identified risk-enhancing factors, even among individuals with ≥3 risk-enhancing factors.6 Indeed, even among patients with familial hypercholesterolaemia (FH), severe hypercholesterolaemia or an elevated lipoprotein(a) (Lp(a)), CAC provides better risk stratification.7 8 Among middle-aged persons with FH and elevated Lp(a), the prevalence of CAC=0 is approximately 40% and it is robustly associated with a low CVD risk over at least an intermediate follow-up time period.9

CAC to guide primary prevention therapies

CAC also provides a significant improvement in CVD risk stratification beyond an individual’s estimated 10-year atherosclerotic CVD (ASCVD) risk. Approximately 40% of persons with an intermediate or high 10-year estimated ASCVD risk have CAC=0 and a low 10-year observed ASCVD event rate.10 11 Conversely, among patients with a low 10-year ASCVD risk estimate who have CAC >0, the risk of ASCVD is significantly higher and identifies persons who may benefit from primary prevention therapies.

Based on these data, the 2018 AHA/ACC Cholesterol Treatment Guideline and 2019 ACC/AHA Primary Prevention Guideline provide a IIa recommendation to perform CAC scoring to improve ASCVD risk stratification among borderline to intermediate-risk patients (5%–<20% 10-year risk) to aid in the shared decision on whether to withhold, postpone or initiate statin therapy. These guidelines recommend that for patients with CAC 1–99, at least moderate-intensity statin therapy should be considered, while patients with CAC ≥100, moderate to high-intensity statin therapy should be started (figure 2). The 2022 ACC Expert Consensus Decision Pathway (ECDP) Guideline further expands upon the role of CAC scoring and recommends consideration of ezetimibe therapy for patients with CAC ≥100 whose low-density lipoprotein cholesterol (LDL-C) is ≥70 mg/dL.12

A number of other international guidelines including those from Canada, Europe and the UK have similar recommendations to consider CAC scoring among persons ≥40 years old who are asymptomatic and at intermediate risk. However, in the 2021 European Society of Cardiology CVD Prevention Guideline, CAC scoring has a IIb recommendation compared with the 2019 ACC/AHA Prevention Guideline in which CAC scoring has a IIa recommendation. These guidelines generally agree on the use of CAC=0 to identify low-risk patients in whom statin therapy may be withheld and then generally agree that when CAC ≥100, initiation of statin therapy should be considered. There is less consensus among these international guidelines about whether CAC scoring should be used to allocate aspirin therapy or antihypertensive medications. Conversely, while guidelines from China and Japan recognise CAC scoring for risk assessment, these guidelines emphasise traditional risk factors for allocation of prevention therapies.

It is also now recognised that primary prevention patients with very high CAC scores have a similar CVD risk compared with patients with a prior CVD event. In an analysis of patients without known CVD from the Multi-Ethnic Study of Atherosclerosis (MESA), a CAC score of approximately 900 had a major adverse cardiovascular event (MACE) rate that corresponded to the MACE rate for patients in the Further Cardiovascular Outcomes Research with PCSK9 Inhibition in Subjects with Elevated Risk (FOURIER) trial.13 Based on this evidence, the 2022 ACC ECDP recommends consideration of PCSK9 inhibitors among patients with CAC ≥1000 whose LDL-C remains ≥70 mg/dL on statins and/or ezetimibe.

Antiplatelet therapy with low-dose aspirin is no longer routinely recommended among primary prevention patients, as a number of trials have shown that the risk of bleeding outweighs the potential CVD risk reduction. However, persons with a CAC score ≥100 have an increased CVD risk commensurate with benefit from low-dose aspirin therapy and among patients with a CAC score of ≥100, low-dose aspirin therapy may be considered.11

While patients with CAC=0 have a low long-term risk of CVD and approximately 40% have healthy arterial ageing with persistent CAC=0 10 years later, detecting CAC early when there is a low burden of atherosclerosis represents the optimal time to start primary prevention therapies.14 15 Accordingly, determining the best methods to identify patients most likely to convert to CAC >0 and the optimal time for repeat CAC scoring has significant clinical implications for detection of atherosclerosis at an early stage.16 Dzaye et al demonstrated that a higher estimated 10-year CVD risk was associated with increased likelihood of conversion from CAC=0 to CAC >0, but even over 10-year follow-up, the majority of patients with incident CAC >0 converted to a CAC score between 1 and 99. Using these data in order to have a 25% yield of conversion to CAC >0 (ie, one out of four patients scanned), approximate rescan intervals of 5–7 years for low-risk patients, 3–5 years for intermediate risk and 3 years for high-risk patients with diabetes can be considered.17

There is an increasing awareness that CAC scoring should be considered among appropriately selected persons <40 years old, especially among men who typically develop CAC about 10 years earlier than women, as the presence of any amount of CAC in a young person is premature and associated with a significantly increased lifetime risk of CVD. This was highlighted in a recent pooled analysis using data from participants 30–45 years old from the MESA, Coronary Artery Risk Development in Young Adults (CARDIA) and the Walter Reed Study in which there was a prevalence of CAC >0 of 26% among white men, 16% among black men, 10% among white women and 7% among black women (figure 3).18 Accordingly, it is especially important to calculate age, sex and race/ethnicity CAC percentiles among younger persons, which helps to quantify their long-term degree of increased risk, which has the potential to be underestimated if only the absolute Agatston score is used. If ≥75% percentile, the ACC/AHA Cholesterol Treatment and Primary Prevention Guidelines, along with the 2022 ACC ECDP Guideline, recommend initiation of statin therapy.12 Patients most likely to have CAC >0 and in whom CAC scoring may be considered before the age of 40 years old are those with an increased risk factor burden or a strong family history of premature CVD.14 16 In general, among persons <40 years old with any individual traditional risk factor (eg, hypertension, hyperlipidaemia, smoking), there is an approximately 5%–10% probability of CAC >0 for men and <5% probability for women (figure 4).

Access to CAC scoring

A primary advantage of CAC scoring is the standardised protocol, which is simple to perform and does not require the use of specialised scanners, intravenous contrast or imaging technicians/readers with highly specialised training in performing cardiac CT. Accordingly, CAC scoring can be performed at nearly any centre that offers CT scanning, which greatly increases access to performing CAC scoring. While CAC scoring is not routinely covered by insurance companies, more are beginning to provide coverage and even without insurance coverage, the price is generally $75–$100. Additionally, CAC scoring is strongly associated with increased CVD risk across race/ethnicity, which addresses a major limitation of the Pooled Cohort Equations and can help to more accurately allocate prevention therapies.

CAC on non-cardiac CT

There are many non-cardiac indications to perform a CT chest scan, of which approximately 80 million are performed annually in the USA compared with approximately 2 million cardiac CT scans. Important information on a patient’s atherosclerotic burden can be obtained from these studies, and non-cardiac CTs of the chest have minimal difference in quantitative or visual categorisation based on the presence/absence of CAC or as mild, moderate or severe CAC.19 These CAC scores from non-cardiac CT also display similar associations with CVD risk. Therefore, reporting a patient’s CAC burden from non-cardiac CT chest is recommended by the Society of Cardiovascular CT (SCCT).20 The SCCT recommends using the CAC Diagnostic Reporting System (CAC-DRS), which is a score that combines the absolute CAC burden with the number of coronary arteries that have CAC. For the CAC-DRS, the burden of CAC can be measured by either formal Agatston CAC scoring (A) or visual categorisation (V, for example, mild, moderate, severe) assessment from non-cardiac CT along with the number of coronary arteries with CAC. CAC-DRS is simple to score and an increasing CAC-DRS has a graded relationship with higher CVD risk.21 In particular, there is an especially high utility for measuring CAC among patients undergoing CT of the chest for lung cancer screening in whom there was a 45% prevalence of CAC ≥100 and the SCCT recommends reporting CAC for non-cardiac CT of the chest for patients undergoing cancer staging or surveillance.22 23

Artificial intelligence (AI) algorithms can be especially helpful for identifying and reporting CAC on non-cardiac CT scans and are rapidly being adopted in clinical practice. The use of AI algorithms can increase the speed, efficiency and accuracy of CAC scoring in order to reduce barriers to measurement and reporting of CAC on non-cardiac CT scans.24 Currently, a number of companies are pursuing Food and Drug Administration approval for reporting CAC based on AI algorithms, and the NOTIFY-1 Study which used an AI algorithm to identify and score CAC from non-cardiac CT demonstrated the clinical promise of this approach. In the intervention arm, primary care clinicians were notified of patients with prevalence CAC and provided with the 2018 AHA/ACC Cholesterol Treatment Guideline IIa recommendation to start statin therapy. At the 6-month follow-up, 51% of patients with CAC had been prescribed statin therapy vs 7% in the standard of care group.

CAC versus coronary CT angiography

The ACC/AHA guidelines primarily recommend CAC scoring for asymptomatic, borderline to intermediate-risk primary prevention patients. The focus on asymptomatic patients is in part because luminal stenosis and non-calcified plaque cannot be measured by non-contrast CT imaging. However, the 2021 AHA/ACC Chest Pain Guideline is the first national US guideline to recommend consideration of CAC scoring in symptomatic patients, which is based on the very low ASCVD event rate for patients with CAC=0. It provides a IIa recommendation for the use of CAC scoring among low-risk patients with stable chest pain and no known CAD to exclude calcified plaque and identify patients with a low likelihood of obstructive coronary artery disease (CAD). Adding support to the utility of CAC scoring among symptomatic patients, a meta-analysis of 19 studies in which 79 903 participants with chest pain had CAC scoring and coronary CT angiography found that 58% of patients with acute chest pain had CAC=0. Among those with CAC=0 and acute chest pain, only 13% had non-calcified plaque among whom 9% had non-obstructive plaque and 4% had obstructive plaque. In this group of symptomatic patients with acute chest pain, there was a negative predictive value of 98% for patients with CAC=0 (table 1). Furthermore, the risk of CVD among persons with CAC=0 was similar across age groups with a 1.4% risk over 4-year follow-up for younger patients compared with 1.8% for older patients.

While luminal stenosis cannot be measured from a CAC scan, only a small percentage of asymptomatic individuals with CAC >400 have obstructive CAD and it is now recognised that measurement of obstructive luminal stenosis (eg, ≥70%) is best used to determine if patient symptoms are due to angina, while coronary plaque burden is the optimal measurement for CVD risk prediction. There are two main reasons for this observation. First, vulnerable or high-risk plaques, while associated with an increased risk of plaque rupture, are uncommon compared with non-high-risk plaques and therefore only provide at most a minimal increase in CVD risk.25 Additionally, the total number of plaques or burden of atherosclerosis is more informative, because the greater the number of plaques, the higher the risk that any one of these plaques may rupture leading to myocardial infarction. Accordingly, while the Agatston score does not directly report the number of plaques, a higher score is strongly correlated with an increased number of plaques and increased atherosclerotic burden. This was highlighted in an analysis of 23 759 individuals with symptomatic chest pain in which CAC scoring and coronary CT angiography were performed. When the presence or absence of obstructive CAD was examined by CAC group, those with obstructive CAD did not have a higher CVD risk compared with persons in the same CAC category with non-obstructive CAD.26 Second, a number of studies have shown that other measures of coronary plaque phenotype have no clinically significant improvement in CVD risk prediction beyond CAC scoring. This absence of added value for coronary CT angiography beyond CAC scoring was demonstrated in the Coronary CT Angiography Evaluation for Clinical Outcomes an International Multicenter (CONFIRM) registry where among 1226 asymptomatic individuals, the incorporation of coronary CT angiography showed no added prognostic benefit beyond a model that included CAC scoring.

Improving upon the Agatston score

The Agatston score is simple to calculate and strongly associated with CVD risk. However, an abundance of information beyond an individual’s CAC burden, such as a higher number of coronary arteries with CAC and greater number of plaques, is associated with increased risk and can be extracted from these scans to further refine CVD risk. Because the CAC score is based on a peak density weighting factor of each plaque, plaques with a higher density contribute a larger amount to the score than a similarly sized plaque with a lower peak density. However, increased density is indicative of a more stable plaque and studies have demonstrated an inverse relationship between plaque density and CVD risk prediction.27 Similarly, a primary pathophysiological benefit of statin therapy is plaque stabilisation that occurs via a reduction in volume of vulnerable low-attenuation plaques and increased progression of high-density calcified plaques.

Beyond calcified plaque, extracoronary calcification can also aid in CVD risk prediction. Measurement of ascending and descending thoracic aorta calcification has a graded relationship of CVD within CAC groups.28 Cross-sectionally, higher aortic valve calcium (AVC) scores are strongly correlated with increased aortic valve gradients and the American Society of Echocardiography/European Association of Cardiovascular Imaging recommend AVC scoring as a tiebreaker in the diagnosis of severe aortic stenosis.29 Recent work from our group has also demonstrated that AVC has an extremely strong association with the long-term risk for severe aortic stenosis.

Future directions

It is now increasingly recognised that while traditional risk factors work reasonably well for CVD risk prediction on a population level, they work less well on the individual level and that direct measurement of atherosclerosis by CAC significantly improves identification of high and low-risk individuals. Accordingly, persons with very high CAC scores have CVD event rates similar to patients with a prior CVD event. Therefore, use of CAC scoring to enrol high-risk individuals most likely to benefit from novel treatment strategies or therapies is beginning to emerge in clinical trials.

To date, there has not been a randomised clinical trial examining the efficacy of statin therapy allocated based on CAC score. However, the Danish Cardiovascular Screening trial (DANCAVAS) invited 46 611 community-dwelling individuals to a comprehensive CVD screening protocol that included CAC scoring.30 In the screening group, 10 471 men (63%) underwent CVD screening and those with an above median CAC score based on their age were recommended to start low-dose aspirin and moderate-intensity statin therapy. The overall primary outcome of CVD mortality did not reach statistical significance between screening groups, but individuals <70 years old had an 11% reduction in total mortality (HR 0.89 (95% CI 0.83 to 0.96)) and a 7% reduction in the risk for MACE (HR 0.93 (95% CI 0.89 to 0.97)). While multiple screening tests were performed as part of this trial, the observed CVD risk reduction was most strongly attributable to aspirin and statin therapy, which was guided by the results of CAC testing. A 10-year follow-up is planned for this trial and whether the additional 5 years of follow-up impacts the overall and CVD results is an eagerly anticipated finding.

In order to more directly address the utility of CAC scoring for allocation of primary prevention strategies, the Coronary Calcium (CorCal) Study is a pragmatic open-label randomised trial currently under enrolment and conducted by Intermountain Health (Utah, USA) (table 2). In this study, half of the anticipated 9000 participants will receive a CAC scan and be randomised to statin therapy based on ACC/AHA guidelines, while the control group will receive standard risk assessment based on the Pooled Cohort Equations. In addition, the Pragmatic Evaluation of Events and Benefits of Lipid-lowering in Older Adults (PREVENTABLE) trial is currently underway to test the efficacy of statin therapy among individuals ≥75 years old for CVD risk reduction. As a part of this trial, approximately half of the participants will have blinded baseline CAC scoring performed. While the intervention will not be randomised based on CAC scoring, results from this trial will provide important information on the utility of statin allocation by CAC score for CVD risk reduction in older persons. The largest of these trials is the Dutch Risk or Benefit in Screening for Cardiovascular disease (ROBINSCA) trial, which has enrolled 43 447 middle-aged men and women who have been randomised for screening based on traditional risk factors and CAC. Enrolment has been completed for this trial in 2021 and the primary outcome results for CAD morbidity and mortality at 5 years follow-up are expected to be the first available of these ongoing CAC trials.