Positive predictive value: adding CACS to current cardiovascular disease risk prediction models

CACS provides incremental value above and beyond current ASCVD risk prediction models, as was concluded in a 2022 meta-analysis of 6 prospective cohort studies comprising almost 18,000 individuals followed for almost 10 years [8]. In addition, McClelland et al. investigated and externally validated a new risk prediction model including CACS [9]. They showed, in two independent validation cohorts, a significant improvement in risk prediction (C-statistics) from 0.7 to around 0.8, when using the new risk equation including CACS. Although the net reclassification improvement (NRI) is not the best way to analyse the incremental value of a marker [10], several studies showed a substantial category-based NRI of 14–32% when adding CACS to different risk prediction models [11,12,13]. A recent prospective study showed the discriminatory power of CACS beyond SCORE2, which is the risk scoring algorithm included in the most recent ESC guideline: C index for cardiovascular events was 0.61 for SCORE2, and 0.75 when CACS was added [14]. Therefore, there is an abundance of evidence that CACS improves the risk prediction of cardiovascular disease in primary prevention among asymptomatic individuals.

Negative predictive value: the role of CACS in ruling out significant CAD (‘the power of zero’)

Continuing evidence shows that the absence of coronary calcification (CACS 0) in asymptomatic individuals is related to a very low risk of ASCVD, even in individuals with a high risk according to the currently used conventional risk prediction models. A systematic review of 13 studies comprising 64,873 asymptomatic individuals, followed for around 4 years, showed that 40% had a CACS 0, and that the 10-year CVD risk of these individuals was only 0.56% [15]. Two other independent cohorts, comprising 16,529 asymptomatic individuals followed for a longer period of 10 [16] or 15 [17] years, both showed that in individuals with CACS 0, the 10-year CVD risk was on average only 3.2%. In all studies the low 10-year ASCVD risk was independent of the presence of traditional risk factors. In fact, in one of these studies, asymptomatic individuals with CACS 0 had an even lower 10-year CVD risk (2.8%) than individuals who were categorised in the low-risk group (meaning absence of traditional risk factors), assessed by current risk models (3.3%) [16]. In addition, in one of the largest cohorts, comprising 44,052 asymptomatic individuals followed for an average of 6 years, individuals with more than three risk factors (e.g. diabetes, dyslipidaemia, hypertension, smoking) and CACS 0 had a 10-year ASCVD mortality risk of only 2% [18]. This indicates that even in asymptomatic individuals classified as (very) high risk because of the presence of multiple risk factors [16, 17, 19], type 2 diabetes [19] or familial hypercholesterolaemia [20], the absence of coronary calcium is associated with an extremely low ASCVD event rate.

Limitations of CACS

There are some limitations to consider when using cardiac CT for cardiovascular risk prediction, such as radiation exposure, healthcare costs and incidental non-cardiac findings. All modern scanners can now acquire a cardiac CT for CACS with radiation doses < 1 mSv [21], a radiation exposure that is much lower than the annual background radiation. The cost of a non-contrast cardiac CT also depends on the time required for calculating the CACS and assessing the scan for incidental findings. Costs can be reduced by automatically calculating CACS with the current software and omitting assessment for incidental findings. On the other hand, one could argue that relevant incidental non-cardiac findings are not a costly burden but a blessing in disguise, such as early lung carcinoma detection. Further studies are needed to adequately assess cost-effectiveness, taking into account incidental non-cardiac findings. Finally, CACS cannot rule out the presence of non-calcified atherosclerotic plaque and possible high-risk plaque features. However, all studies with long-term follow-up show that the negative predictive value of CACS 0 for cardiovascular events is excellent in asymptomatic individuals. Furthermore, Mortensen et al. showed in 23,143 individuals who underwent CACS and CCTA that a CACS of 0 resulted in a very low event rate of 6.9 per 1000 person-years, regardless of whether there was no plaque, purely non-calcified plaque or even non-calcified plaque with > 50% stenosis [22].

Positioning of CACS in the Dutch guideline on CVRM

Conventional risk prediction models based on traditional risk factors are commonly used in a case-finding method at the general practitioners (GPs) office, without systematic screening on a population level. In view of the limited accuracy of risk prediction models and the strong predictive value of CACS for ASCVD, screening by using CACS may improve risk prediction. However, this approach is currently not included in the cardiovascular risk management (CVRM) guideline; the focus is on the potential value of CACS to assist in decision-making regarding preventive medication around treatment thresholds for the risk factor-based risk score (Fig. 1). CACS has clearly shown its added value in risk prediction in asymptomatic individuals, at least those aged between 50 and 70 years (but maybe even a broader range) with intermediate or maybe even high CV risk based on current risk prediction models, but most guidelines do not yet recommend CACS for risk management. This is partly due to the fact that the cost-effectiveness of adding CACS to risk prediction models has not been convincingly established [23]. The Dutch CVRM guideline is currently being updated and will be published in 2024. The position of CACS as a diagnostic tool for reclassification of cardiovascular risk in primary prevention will be revised. The previous guideline (2019) stated that if CACS is available one can consider to take it into account in individuals in whom there is doubt as to whether to start preventive therapy [24]. Indeed, knowing the CACS is high increases medication adherence [25]. The new CVRM guideline advises to consider determining CACS in individuals aged 50–70 years if a discussion about whether to initiate drug treatment does not yield a decision. Importantly, nowadays nearly all Dutch hospitals offer CT for CACS evaluation, and in some regions GPs can already order a CT-CACS, without referral to a medical specialist, making CT more accessible.

Fig. 1

Infographic: Stepped approach with coronary calcium score. CACS coronary artery calcium score; p75 75th percentile; no CACS recommended in low CV-risk group, although 4% have significant CAC [42]

CACS screening for primary prevention

So far, there is no evidence from randomised controlled trials (RCTs) that screening strategies based on CACS improve prognosis. It is important to note that this level of evidence is also lacking for the current risk prediction models, as highlighted in the 2021 ESC Guideline on Prevention [26]. The only two RCTs, the St Francis Heart study [27], and DANCAVAS [28], have not yet shown a clear benefit for treatment according to a population-based screening approach with CACS. However, in a sub-analysis of the St Francis Heart study, individuals with a positive family history of CVD and CACS > 80th percentile had a 45% reduction in events in the cholesterol-lowering arm as compared with placebo [27]. The DANCAVAS study, a population-based screening study comprising 46,611 men, suggested possible benefit for a broad cardiovascular screening approach including CT, but did not reach statistical significance for the primary endpoint of all-cause mortality after a follow-up of 5.6 years (HR 0.95, 95% CI 0.90–1.00, p = 0.06). Importantly, the primary endpoint reached borderline significance, while the study was powered for a 10-year follow-up [28]. On the other hand, in a prespecified subgroup analysis, younger participants (65–69 years) had more benefit from screening than older individuals (> 70 years). Currently, the Dutch ROBINSCA (Risk or benefit in screening for cardiovascular diseases) trial is being conducted [29]. This trial is evaluating men aged 45–74 years and women aged 55–74 years, with at least one risk factor, comparing a CACS-based strategy (intervention arm) with usual GP care (control arm). Management in the two intervention arms is evidence based. Over 43,000 individuals were included and currently 5‑year outcomes are awaited.

Interpretation and implication of CACS

When and how to treat or not to treat based on CACS in asymptomatic individuals for primary prevention of ASCVD is still debated. Despite the lack of evidence from major RCTs, observational data have shown that a significantly elevated CACS, ≥ 100 or ≥ 75th age, sex and ethnicity corrected percentile, causes a sufficiently high risk to consider cholesterol-lowering therapy [16, 17]. Therefore, the combined European and American CVRM guidelines advise to consider cholesterol-lowering therapy if CACS is significantly elevated and to consider withholding cholesterol-lowering therapy if CACS is 0 [26, 30].

It is unknown whether there is a threshold for CACS above which aspirin therapy should be initiated. It is currently not recommended to prescribe aspirin in primary prevention because of the increased risk of bleeding [31]. On the other hand, randomised trials on aspirin use did not take into account CACS, therefore we do not know whether the individuals who benefited from aspirin therapy in these trials might have been the ones with an increased CACS. However, incorporating CACS in an aspirin-prescription decision-making tool, based on the MESA study, suggested that individuals with CACS ≥ 100 would benefit from treatment with aspirin, regardless of their risk status according to current risk prediction models (estimated 5‑year number needed to treat: 173 for low-risk individuals; 92 for high-risk individuals and 442 for a major bleed) [32].

Reclassification to a lower ASCVD risk based on CACS 0, and withholding preventive treatment

ASCVD risk prediction should not be viewed as a static process. Risk of ASCVD evolves over the years, not only because individuals get older, but also because risk factors evolve. Therefore, it is important to realise that when CACS is 0, it needs to be repeated in the coming years. This is in line with the 2018 ACC/AHA cholesterol guidelines, which emphasise the clinical importance of CACS 0, and state that it is reasonable to withhold cholesterol-lowering therapy and, in most individuals, reassess risk with CACS in 5 to 10 years [30]. In certain very high-risk subgroups, such as familial hypercholesterolaemia, guidelines advise to continue cholesterol-lowering therapy in patients with a CACS of 0, considering their very high lifetime risk of ASCVD [33]. On the other hand, two independent prospective cohort studies showed that individuals with familial hypercholesterolaemia and CACS 0, still had a very low 10-year event rate of 0–1.2% [20, 34]. Therefore, CACS could help increase precision in familial hypercholesterolaemia patients and could help to reduce costs related to more intensified cholesterol-lowering regimens. Finally, although multiple studies comprising in total 4248 individuals with CACS 0 followed for 3 to 10 years showed that the warranty period for the development of any CAC is 3 to 7 years [35, 36], another study showed that only 2% of individuals with CACS 0 at baseline progressed to CACS > 50 in 5 years [36]. Furthermore, only individuals with a positive, low CACS level of 1–100 at baseline progressed to a substantial risk (e.g. > 7.5%, in 10 years) for which cholesterol-lowering medication is recommended [16].

Cost-effectiveness of CACS as part of cardiovascular risk prediction

One of the reasons why CACS currently has no basis in risk prediction has to do with the fact that the cost-effectiveness analyses in the primary prevention context have so far shown conflicting results [23, 37]. Critical factors that drove the conclusions in all these studies were the costs associated with cholesterol-lowering medication, the rating of side effects from cholesterol-lowering medications and the general desire to avoid lifelong preventive therapy. In the most recent cost-effectiveness analysis by Hong et al. [23], outcomes were similar for risk prediction based on the addition of CACS or based on the 2013 ACC/AHA guidelines [38] only. They analysed the addition of CACS in the intermediate-risk group in which only 11.4% had a CACS elevated enough to start cholesterol-lowering therapy. Although for the intermediate-risk group it did not seem cost-effective to add CACS to risk prediction models, it did result in less patients receiving preventive therapy, namely 44% of individuals considered or recommended for cholesterol-lowering therapy had CACS of 0. Therefore, as a rule-in strategy CACS does not seem to be cost-effective, but as a rule-out strategy it might be. The authors concluded that cost-effectiveness analyses should not be the only criterion for clinical decision-making, but that a shared decision-making model is similarly important for clinicians, patients and policy-makers.