Introduction

The 2013 American College of Cardiology/American Heart Association (ACC/AHA) guidelines for cardiovascular risk assessment introduced the pooled cohort equations (PCE) for estimating the 10-year risk of developing a first atherosclerotic cardiovascular disease (ASCVD) event (non-fatal myocardial infarction [MI], coronary heart disease [CHD] death, or fatal or non-fatal stroke) among asymptomatic individuals [1]. The work group formulating these guidelines reviewed the contribution of non-traditional risk markers to variables used in the PCE [1]. Current guidelines recommend that if a risk-based treatment decision is uncertain, four non-traditional markers may be considered for revising risk assessment upwards: coronary artery calcium (CAC) score ≥ 300 Agatston units or ≥ 75th percentile for age, sex, and ethnicity; high sensitivity C-reactive protein level (hsCRP) ≥ 2 mg/L; and family history of premature CHD or low ankle-brachial index (ABI < 0.9). Performances of these tests were all given a IIb ACC/AHA class of recommendation [1].

Simultaneous with the risk assessment guidelines, the ACC/AHA published guidelines on treatment of blood cholesterol to reduce ASCVD risk in adults [2]. High-intensity statin use was recommended for adults with clinical ASCVD, low density lipoprotein cholesterol (LDL-C) ≥ 190 mg/dL, and those 40 to 75 years of age with diabetes and LDL-C between 70 and 189 mg/dL. Among the primary prevention group of 40- to 75-year-old patients without diabetes along with LDL-C between 70 and 189 mg/dL, the guidelines recommended shared decision making among clinicians and patients regarding statin use in those with ≥ 5% 10-year ASCVD risk [2]. Pencina et al. applied the ACC/AHA cholesterol guidelines to the National Health and Nutrition Examination Surveys and found that 56 million US adults (nearly 50%) would be classified under the four statin benefit groups [3]. This increased the statin eligible population by 12.8 million compared to prior guidelines, with the majority (10.4 million) belonging to the primary prevention group [3].

The current guideline-based approach results in a higher sensitivity (statins recommended for adults who would develop ASCVD in the future) at the expense of lower specificity (statins recommended for adults who would not develop ASCVD in the future) compared to prior guidelines [3]. Interestingly, a systematic examination of the PCE revealed that any hypothetical man or woman, aged 40 to 79 years, can reach clinically relevant risk thresholds depending on risk factor burden, but age is the dominant determinant of 10-year ASCVD risk [4].

In contrast to the ACC/AHA approach, the recent United States Preventive Services Task Force (USPSTF) statement on statin use for primary prevention is much more selective and recommends initiating low- or moderate-dose statins in adults aged 40 to 75 years, who have one or more risk factors (dyslipidemia, diabetes, hypertension, or smoking) and a 10-year ASCVD risk ≥ 10% [5]. Additionally, low- or moderate-dose statins can be selectively offered to those with a 10-year ASCVD risk of 7.5 to 10% [5]. A different approach was recently recommended by the American Association of Clinical Endocrinologists (AACE) and American College of Endocrinology (ACE) [6]. These guidelines recommend statin use to target a LDL-C level of < 55, < 70, < 100, and <130 mg/dL for patients at extreme risk, very high risk, high or moderate risk, and low risk of ASCVD, respectively [6]. Lastly, the European Society of Cardiology (ESC) and European Atherosclerosis Society (EAS) recommend using the Systematic COronary Risk Evaluation (SCORE) tool for assessing 10-year risk of CVD mortality (< 1%, 1 to < 5%, 5 to 10%, ≥ 10%) along with target LDL-C levels (< 70, 70 to < 100, 100 to < 155, 155 to < 190, ≥ 190 mg/dL) to guide risk-based intervention strategies among asymptomatic adults [7]. Similar to the ACC/AHA guidelines, the ESC/EAS recommend up-classifying risk among asymptomatic individuals at intermediate 10-year CVD mortality risk and a CAC score > 400 [7]. However, unlike the ACC/AHA guidelines, the population-wide implications of these guidelines have not yet been determined.

The increased eligibility for statin use per the ACC/AHA guidelines highlights an unmet need to refine cardiovascular risk assessment techniques, particularly to identify patients who are at lower risk for developing ASCVD events but are recommended statins. Non-traditional risk markers might be a solution to this problem as they can improve cardiovascular risk assessment and can guide preventive treatment decision making. Indeed, multiple studies have evaluated the improvement afforded by these markers in addition to traditional risk factors (TRFs).

Among non-traditional markers, CAC has consistently improved risk discrimination and correctly re-classified individuals to appropriate risk categories as reviewed below. CAC is measured by non-contrast, electrocardiographically gated computed tomography of the chest and is a marker of total subclinical coronary atherosclerosis burden, reflecting the cumulative exposure to both measured and unmeasured risk factors [8]. The aim of this review is to summarize recent literature focusing on improvement in cardiovascular risk assessment with CAC and how it compares with other non-traditional risk markers.

Improvement in Cardiovascular Risk Assessment Using CAC

CAC is a well-documented independent predictor of CHD events, strokes, and all-cause mortality among asymptomatic individuals [9,10,11,12]. In the landmark Multi-Ethnic Study of Atherosclerosis (MESA) cohort-based study, Detrano et al. described the independent association of CAC with incident coronary events [9]. CAC is independently associated with incident strokes among participants of MESA and Heinz-Nixdorf Recall (HNR) studies, although the strength of association was not as strong as observed with CHD [10,11]. In a large, single-center study of 9715 asymptomatic patients followed for nearly 15 years, CAC accurately predicted long-term all-cause mortality [12].

Improvement in Risk Discrimination and Reclassification

Perhaps the most significant advantages of CAC assessment are an improvement in discrimination of individuals that will develop an event in the future from those that will not and reclassification of individuals to appropriate risk categories. This consistent finding stems from six unique, prospective, population-based cohorts of adults without overt cardiovascular disease (Table 1) [13,14,15,16,17,18]. MESA is the largest and most diverse among these, and Rotterdam has the oldest participants while the Dallas Heart Study (DHS) has the youngest. Of note, almost all participants in the HNR, Rotterdam, and Framingham Heart Study (FHS) were Caucasians, though HNR and Rotterdam were European populations. While the outcome of interest differed slightly, each study included MI and CHD death in their composite outcome (Table 1). The addition of CAC to baseline models containing TRFs from each study led to significant and consistent increases in C-statistics and significant net reclassification improvements (NRI).

Table 1 Improvement in cardiovascular risk assessment by using CAC score in addition to traditional risk factors
Full size table

Heterogeneity of CAC Across Risk Factor Subgroups

Among individuals with a high TRF burden, absence of CAC marks a low risk of future events and conversely, despite normal values of traditional risk factors, a high CAC score identifies those at elevated risk. In a study by Martin et al., MESA participants were stratified based on the burden of lipid abnormalities (high LDL-C, low HDL-C, and high triglycerides) and their CAC scores [19]. The authors observed that those with no lipid abnormalities and CAC > 100 had much higher event rates (22.7 CVD events/1000 person-years) than those with 3 lipid abnormalities and CAC = 0 (5.9/1000 person-years) [19]. In another MESA study, Silverman et al. stratified the population based on presence of modifiable TRFs (smoking, high LDL-C, low HDL-C, hypertension, and diabetes) and CAC scores [20]. Participants with no TRFs and CAC > 300 had an incident CHD rate 3.5 times higher than individuals with ≥ 3 TRFs and CAC = 0 (10.9/1000 versus 3.1/1000 person-years) [20]. These studies, among several, highlight the heterogeneity between risk factors, subclinical atherosclerosis, and event rates [21].

The potential utility of CAC in two low-risk populations has been explored recently [22•,23]. In a meta-analysis of low-risk women with 10-year ASCVD risk < 7.5% from MESA, HNR, Rotterdam, DHS, and FHS, Kavousi et al. reported that CAC presence (36% with score > 0) was associated with an increased incidence of ASCVD (4.33 per 1000 person-years) as compared with CAC absence (1.41 per 1000 person-years) [22•]. After adjustment for confounders, low-risk women with any CAC were twice as likely to have an ASCVD event as compared to women with CAC = 0 [22•]. Similar to previous studies, addition of CAC to TRFs significantly improved C-statistic from 0.73 to 0.77 and yielded a continuous NRI of 20% [22•].

Among participants of the Coronary Artery Risk Development in Young Adults (CARDIA) study, Carr et al. observed that young individuals (aged 32 to 46 years) with any CAC were five times more likely to experience a CHD event as compared to those without CAC [23]. The CARDIA and Kavousi studies highlight how CAC presence identifies those at high CVD risk among patients thought to be at low risk of events due to low TRF burden.

The MESA CHD Risk Score

Capitalizing on the principle of improving risk assessment with CAC, McClelland et al. created a novel 10-year CHD risk score from a MESA analysis by integrating CAC scores with TRFs [24••]. The score was validated in two other population-based cohorts (HNR study and DHS) and showed very good discrimination (C-statistic 0.779 in HNR and 0.816 in DHS) and calibration (calibration slope 0.90 in HNR and 1.19 in DHS, calibration-in-the-large − 0.50% in HNR and − 0.46% in DHS) during external validation [24••]. The MESA CHD Risk Score is available online and can be a very useful tool for communicating 10-year CHD risk to patients that have undergone CAC assessment (https://www.mesa-nhlbi.org/MESACHDRisk/MesaRiskScore/RiskScore.aspx) [25].

The Power of CAC = 0

A unique aspect of CAC scoring is the strong prognostic implication of its absence (i.e., CAC = 0). It is well-established that individuals with CAC = 0 have a low short-term risk of adverse cardiovascular events [9,13,14,15,16,17,22•,23]. The long-term implications of CAC = 0 were less clear until recently. While the above studies highlighted the short-term event rates or long-term mortality rates, there was a need to prospectively measure ASCVD event rates over a longer term among those with CAC = 0. To address this knowledge gap, we evaluated the 10-year ASCVD events among MESA participants with a CAC score of 0–10 [10, 26]. The 10-year ASCVD event rate was 2.9 per 1000 person-years for those with CAC = 0 vs 5.5 per 1000 person-years for those with 1–10 score [26]. Notably, in participants with a 10-year ASCVD risk between 1 and 15% and CAC = 0, the ASCVD event rate was 4.4 per 1000 person-years, which is well under the ACC/AHA recommended threshold for starting statin therapy [26]. Additionally, Valenti et al. have reported a less than 1% annual all-cause mortality rate among asymptomatic individuals with CAC = 0 that were followed for nearly 15 years [27].

Utility of CAC for Guiding Primary Prevention Strategies

Several recent studies evaluated the impact of CAC assessment on primary prevention strategies [28,29,30,31,32]. Yeboah et al. examined the role of non-traditional risk markers for up-classifying risk among MESA participants with a 10-year ASCVD risk < 7.5% and not taking statins [28]. After excluding individuals with diabetes, high CAC (≥ 300 or ≥ 75th age/sex/ethnicity percentile) up-classified risk in 6.8% of the population. These individuals had significantly higher ASCVD event rates (13.3%) over 10.2 years of follow-up [28]. Abnormal CAC was noted to be the most effective non-traditional risk marker for re-classifying risk [28].

Given the increased number of individuals eligible for statin therapy with the new ACC/AHA guidelines, a larger body of recent literature has tried to explore the role of CAC = 0 to ‘de-risk’ individuals that are identified as statin eligible. Specifically, these studies have assessed if the ASCVD event rate in individuals recommended for statins, but with CAC = 0, was below the risk threshold for statin initiation.

Pursnani et al. analyzed statin-naïve participants of the Framingham multi-detector computed tomography (MDCT) study that were followed for ASCVD events over 9.4 years [29]. They observed that the current ACC/AHA guidelines had greater accuracy and efficiency for identifying individuals with increased risk of incident CVD and subclinical atherosclerosis, as compared to the previous guidelines [29]. However, among individuals eligible for statins as per the ACC/AHA guidelines with CAC = 0, the ASCVD event rate was only 1.6%, which is much lower than the ACC/AHA statin-initiation threshold [29]. Interestingly, the ASCVD event rate among statin ineligible participants was 1%, and if CAC absence was used to make more individuals ineligible for statins, the event rate in the new statin ineligible group did not significantly increase (1.1%) [29].

Nasir et al. systematically evaluated the ACC/AHA cholesterol guidelines in the context of CAC scoring in MESA [30••]. Nearly 50% of the study participants would be recommended statins based on either LDL-C ≥ 190 mg/dL (4%), diabetes (19%), or 10-year ASCVD risk ≥ 7.5% (77%) [30••]. Among the remaining participants, 12% would fall under the statin consideration category (10-year ASCVD risk 5 to < 7.5%) and 38% would not be recommended statins (10-year ASCVD risk < 5%) [30••]. Notably, the ASCVD event rate for all participants with CAC = 0 in the statin recommended group was only 5.2 per 1000 person-years [30••]. Non-diabetic participants with CAC = 0 and 10-year ASCVD risk of 7.5 to 20% experienced 4.6 ASCVD events per 1000 person-years, which was below the threshold for statin initiation [30••].

Mortensen et al. evaluated the applicability of CAC = 0 for down-classifying risk among participants of the BioImage study [31]. They reported that among participants without ASCVD at baseline, a majority (86%) qualified for statin therapy per the current ACC/AHA guidelines [31]. CAC = 0 was observed in 32% of the cohort and yielded a binary NRI of 20% for CHD events and 14% for CVD events [31]. Lastly, Mahabadi et al. analyzed the utility of CAC = 0 for making primary prevention decisions among participants of the HNR study [32]. Among participants that met the ACC/AHA statin eligibility criteria, those with CAC = 0 had an ASCVD event rate of 2.7 per 1000 person-years follow up as compared with 9.1 events per 1000 person-years among those with CAC score ≥ 100 [32]. Again, the event rate among participants with CAC = 0 was much lower than the recommended threshold for starting statin therapy.

These studies have helped solidify the consensus that CAC = 0 can ‘de-risk’ guideline-based, statin-eligible individuals. In such individuals with a low baseline event rate over 10 years, statin therapy should be viewed as a measure to reduce longer term, or lifelong, risk and the clinician-patient discussion over risks and benefits should proceed accordingly.

Comparing CAC with Other Non-Traditional Risk Markers

The 2013 ACC/AHA risk assessment guidelines weighed evidence for several non-traditional risk markers other than CAC - hsCRP, family history, ABI, carotid intima-media thickness (CIMT), apolipoprotein B (ApoB), glomerular filtration rate (GFR), microalbuminuria, and cardiorespiratory fitness [1]. As described earlier, CAC, hsCRP, family history, and ABI may be assessed when risk-based treatment decisions are unclear [1]. These guidelines do not recommend CIMT testing and acknowledge that the contribution of ApoB, CKD, albuminuria, and cardiorespiratory fitness to risk assessment is not certain at present [1]. It is prudent that the performance of these markers be compared against each other so that physicians can decide which marker would be most useful for guiding risk assessment and treatment decisions.

Peters et al. systematically reviewed the comparative efficacy of three modalities of subclinical atherosclerosis imaging (flow mediated dilation [FMD], CAC, and CIMT) for improving risk stratification in 2012 [33]. They observed heterogeneous results, but the range of improvement in C-statistics and NRI was greatest with CAC among the three measures [33]. However, a major limitation was the lack of comparison of these non-traditional risk markers within a single-study cohort.

This limitation was subsequently addressed in the Early Identification of Subclinical Atherosclerosis by Noninvasive Imaging Research (EISNER) trial along with MESA and Rotterdam studies. [34, 35•, 36]. Rana et al. evaluated the utility of CAC and five biomarkers (hsCRP, interleukin-6, myeloperoxidase, B-type natriuretic peptide, and plasminogen activator-1) for predicting incident CVD events among EISNER participants over 4.1 years [34]. Among these, only CAC resulted in an improvement in C-statistic (from 0.75 to 0.84) and offered a significant NRI (35%) [34]. Yeboah et al. compared the efficacy of six non-traditional risk markers for improving risk assessment among non-diabetic participants of the MESA study at intermediate risk (Table 2) [35•]. Over a follow-up period of 7.6 years, CAC resulted in a much higher change in C-statistic (0.161) and NRI (65.9%) as compared to other markers for predicting CHD risk [35•]. Kavousi et al. compared the utility of 12 non-traditional risk markers when added to Framingham risk factors among participants of the Rotterdam study (Table 2) [36]. Again, CAC scores resulted in the maximum change in C-statistics (0.05) and the highest NRI (19.3% for overall and 39.3% for intermediate risk) as compared to other markers [36].

Table 2 Comparison of CAC with other non-traditional risk markers for improving cardiovascular risk assessment
Full size table

With the inclusion of strokes as an outcome of interest in the new guidelines, the incremental contributions of non-traditional risk markers were recently evaluated in the MESA and HNR studies [37, 38]. Yeboah et al. re-evaluated of utility of CAC, hsCRP, family history, and ABI in MESA (Table 2) [37]. They found that CAC was the only non-traditional marker that resulted in a significant C-statistic improvement and NRI for predicting incident ASCVD [37]. Comparable results were obtained when the outcome of interest was changed to incident CHD events (Table 2) [37]. Geisel et al. compared CAC, CIMT, and ABI in the HNR study and similarly observed that only CAC that resulted in a significant improvement is C-statistic (Table 2) [38]. All three measures resulted in a significant category-free NRI, but the magnitude afforded by CAC was the highest (Table 2) [38].

As discussed previously, the new guidelines increase the number of statin-eligible individuals. It is also well established that patients can be averse to starting statins as a lifelong therapy in primary prevention [39]. In this paradigm, the importance of a test that can accurately ‘de-risk’ patients appropriately and further personalize risk assessment cannot be understated. In a seminal study from MESA, Blaha et al. studied the role of negative markers, including CAC = 0 and 12 others for ‘de-risking’ participants [40••]. The authors used multivariable-adjusted diagnostic likelihood ratios (DLRs) to assess the change in risk after knowing the presence of a negative risk marker (posttest risk) as compared to not knowing the result (pretest risk) [40••]. They also evaluated NRIs offered by each negative marker for down-classifying risk among MESA participants with a pretest 10-year CVD risk ≥ 7.5% [40••]. Among all negative markers, a zero CAC score was the strongest modifier of posttest risk (mean DLR 0.54 for CVD events and 0.41 for CHD events) and resulted in the most significant down-classification of risk (NRI 13.8%) [40••]. This suggests that on average, the posttest risk for ASCVD is approximately 50% of the pretest risk when CAC = 0 [40••]. The average DLR was observed to be lower in older and higher pre-test participants, the population that would benefit the most from appropriate ‘de-risking’ [40••]. CIMT < 25th percentile (DLR 0.75 for CVD and 0.65 for CHD, NRI 4.9%) and no family history of CHD (DLR 0.81 for CVD and 0.76 for CHD, NRI 3.6%) were also important negative markers [40••].

Our CAC-Based Approach to Primary Prevention

Based on the robust evidence described above, among the non-traditional risk markers discussed in the ACC/AHA guidelines, we recommend CAC assessment to improve shared decision making and guide statin initiation among primary prevention patients, particularly in those who are reluctant to start a lifelong therapy (Fig. 1). We recognize that patients with familial hypercholesterolemia (or those with LDL-C ≥ 190 mg/dL) are at particularly elevated risk and would likely benefit from statins, regardless of their CAC score. Therefore, we strongly encourage initiating statins in this group. While there is significant heterogeneity of ASCVD risk among diabetic patients, we typically encourage statin therapy in this group as well. Similarly, for those with ASCVD risk > 15%, we find a CAC = 0 unlikely to significantly reclassify their 10-year risk below 5–7.5% and so we typically recommend statin initiation without recommending CAC scoring.

Fig. 1
figure 1

Algorithm for CAC assessment to improve cardiovascular risk assessment in primary prevention patients. LDL-C low density lipoprotein cholesterol, ASCVD atherosclerotic cardiovascular disease, CAC coronary artery calcification

Full size image

For all other primary prevention patients, we focus our discussion with patients on 10-year ASCVD risk as well as the risks and benefits of initiating potentially lifelong statin therapy. We find CAC scoring to be useful in low-risk patients (10-year ASCVD < 5%) with a family history of ASCVD and in intermediate risk patients with 10-year ASCVD risk between 5 and 15% [41].

If a patient would like to take a statin, we do not encourage CAC scoring to dissuade them (i.e., if CAC = 0). Importantly, for any patient (regardless of baseline risk) that is hesitant about initiating statins, CAC assessment should be performed only if the patient plans to change treatment decisions based on the CAC results. In these patients, if CAC = 0, we reinforce lifestyle modifications and recommend withholding statins with repeat CAC scanning in ~ 5 years. If there is any CAC, we suggest initiating statin therapy (typically with a moderate to high intensity statin), low dose aspirin if no contraindications, and strong risk factor control; at this time we do not recommend repeat CAC scanning once a decision to start statins has been made. Higher scoring thresholds including CAC ≥ 100, CAC ≥ 300, or > 75th percentile for age/race/gender (2013 ACC/AHA guidelines) could also be substituted. We prefer dichotomizing at CAC = 0 vs > 0 because any CAC represents atherosclerosis and a significantly higher ASCVD risk compared with CAC = 0.

Future Directions in Risk Assessment

Risk assessment is continually evolving with the advent of novel risk markers and approaches to risk assessment. Extra-coronary calcification at various sites (aortic valve, aortic root, mitral valve, thoracic aorta, or abdominal aorta) has been shown to have an independent association with CHD events and may be a useful marker that can be used in conjunction with CAC [17,42]. Additionally, a recent MESA study reported an incremental benefit of measuring CAC distribution (an indicator of diffuse, multi-vessel subclinical atherosclerosis) in addition to the traditional Agatston score [43].

The utility of combinations of non-traditional risk markers to improve risk prediction is also being investigated. In DHS, the combination of family history with CAC suggested additive risk [44]. Those with prevalent CAC who also had a family history of CHD were at the highest short-term risk for events as compared with those with prevalent CAC alone [44]. As novel markers of risk are discovered, these should also be tested for their incremental contribution to the existing risk prediction paradigm. One recent example stems from cholesterol efflux capacity measured in DHS [45]. Higher than median cholesterol efflux capacity identified individuals at lower risk among those with prevalent CAC, prevalent FH, or elevated hsCRP levels [45]. Cholesterol efflux capacity additionally improved discrimination indices and net reclassification index when added to a model containing TRFs and either prevalent CAC, FH, or elevated hsCRP [45].

Conclusion

Coronary artery calcium is a well-validated marker of cardiovascular risk and a high-yield aid for shared decision making. The absence of CAC has emerged as a powerful tool for guiding primary prevention strategies in the wake of recent ACC/AHA guidelines recommending statins in a larger proportion of adults. Finally, in the studies comparing CAC with other non-traditional risk markers, CAC uniformly outperforms other markers for cardiovascular risk assessment and guiding treatment strategies.