Author + information
- Donald M. Lloyd-Jones, MD, ScM∗ ()
- Department of Preventive Medicine and Division of Cardiology, Department of Medicine, Northwestern University Feinberg School of Medicine, Chicago, Illinois
- ↵∗Reprint requests and correspondence:
Dr. Donald M. Lloyd-Jones, Department of Preventive Medicine, Division of Cardiology, Feinberg School of Medicine, Northwestern University, 680 North Lake Shore Drive, Suite 1400, Chicago, Illinois 60611.
For the past 2 decades, the consensus paradigm for prevention of atherosclerotic cardiovascular disease (ASCVD) has been based on the concept that the intensity of prevention efforts should match the absolute risk of the individual patient. Application of this concept in international guidelines has improved identification of those most likely to benefit from preventive therapies. The risk scores upon which recommendations have been based to date use data on traditional risk factors and risk markers to provide probabilities of disease and guidance for instituting preventive therapies when net clinical benefit can be expected from doing so. This risk-based paradigm has been adopted broadly for cholesterol guidelines (1,2) and recommendations regarding use of aspirin (3).
During the same time period, advances in imaging technology have allowed for accurate detection of subclinical atherosclerotic disease and important surrogates, which add independent information and improve accuracy for prediction of risk for future events. These modalities, such as coronary artery calcium (CAC) screening, can identify individuals with subclinical atherosclerosis who are at higher risk than expected and those with minimal or no evidence of atherosclerosis, who tend to be at very low risk even in the presence of risk factors. In fact, CAC screening does a very good job of discriminating those most likely to have events among low- and intermediate-risk subgroups and identifying those who are not likely to have events among the predicted high-risk groups. This is not surprising, because CAC represents the actual disease of interest and not merely the presence of risk factors or risk markers for disease. But, to date, no national evidence-based guidelines have adopted a “disease-screening” approach; the current paradigm remains fixed on “risk-based” screening.
In this issue of the Journal, McClelland et al. (4) use data from MESA (Multi-Ethnic Study of Atherosclerosis) to develop novel risk equations for 10-year prediction of coronary heart disease (CHD) that include traditional risk factors and markers as well as the CAC score as covariates. As expected, they describe significant improvements to the discrimination and calibration of the risk equations with the addition of the CAC score. In addition, they assess the utility of the new equations in 2 well-described and similar external cohorts, the German HNR (Heinz-Nixdorf Recall) Study and the U.S.-based DHS (Dallas Heart Study), and demonstrate good to excellent discrimination and calibration in these cohorts. The methods used by these MESA investigators are sophisticated and highly appropriate, and their overall conclusions fit well within the context of the participants of these studies. It is understandable, but unfortunate, that the authors chose only to predict CHD events, because this ignores substantial amounts of preventable risk from atherosclerotic stroke events, particularly in women and African Americans. The authors suggest that these new risk equations could be used in clinical practice, if a CAC score has been obtained, to guide decision-making.
However, before we can conclude that now “we are there” with CAC screening, we must consider some methodological issues and underlying assumptions and understand their potential consequences. Most importantly, we must remember how we use absolute risk prediction equations in clinical practice. They are typically designed to predict the natural history of disease in the absence of intervention. MESA, although contemporary, is not a natural history cohort. By design, MESA participants underwent CAC scoring at the baseline examination and at least once more during follow-up. For all the right reasons, the MESA investigators shared this CAC score information with participants and encouraged them to discuss it with their own physicians. As might be expected in the current environment, knowledge of this information has been associated with increased usage of evidence-based preventive therapies in MESA participants (5). The increased use of statins, aspirin, and antihypertensive medications has been proportional to the reported CAC scores. Participants with higher CAC scores, who are therefore at higher risk and would have experienced more events, have been significantly more likely to initiate preventive therapies after the baseline examination. It was recently reported that >80% of MESA participants have received some sort of preventive therapy during follow-up (6). It is certain that participation in MESA, with its attendant CAC screening, has led to the prevention of many events that would otherwise have occurred in these individuals, but it is difficult to quantify the magnitude of effect. True natural history prediction has thus been lost. Similar issues likely apply to the HNR and DHS cohorts, whose members were also informed of their CAC results. This does not discount the immense value of the cohorts, but it does mean we must understand the observed event levels in context.
That the CAC score remains an independent predictor of CHD events is, in part, a testament to the potential effect of disease screening rather than risk screening. One suspects that CAC would be an even stronger predictor in a true natural history cohort. However, because the new MESA risk equations would be used to predict absolute risk for CHD in the natural history clinical environment, it is difficult to know how to translate these results to the point of care for more representative populations of patients. How can a patient and clinician use these new equations in the context of a “risk discussion” of natural history, when some of the MESA participants received downstream therapy that prevented events? Which ones were they? Would that apply to this patient? Can we accurately classify risk without knowing what therapies this patient will take in the future? Although still valuable, the information provided by these equations is somewhat garbled.
Furthermore, incorporation of the CAC score into clinical risk prediction equations implies universal screening (as happened in MESA and the other cohorts). The lack of a randomized screening trial demonstrating the efficacy, utility (including potential adverse events), cost-effectiveness, and net clinical benefit of CAC screening, particularly in intermediate-risk patients when there is uncertainty regarding decisions, is a substantial barrier to widespread adoption. Without these data, current clinical practice guidelines cannot provide strong evidence-based recommendations to guide practice.
Until such data are available, or a consensus is reached regarding switching to disease-based screening for ASCVD, there may be a better way to think about the use of CAC screening. Nasir et al. (7), also in this issue of the Journal, provide data from MESA to inform this approach (although the same caveats about natural history apply). These investigators observed that one-half of MESA participants were in the net clinical benefit groups identified by the 2013 American College of Cardiology/American Heart Association cholesterol guidelines (most due to a 10-year predicted risk ≥7.5%) who therefore might receive statin therapy after a patient-clinician discussion. In this group, the 41% with a CAC score of 0 had a lower, but nonzero, event rate (∼0.5%/year) compared with the group with any CAC (1.2%/year).
The optimal approach at present appears to be a sequential screening approach, with quantitative risk assessment followed by selective CAC screening, rather than universal screening. The value of classifying individuals into very-low and very-high predicted risk groups on the basis of risk factors remains, and there appears to be little debate about instituting drug therapy in the highest-risk groups (except in those with high competing risk for mortality) and avoiding it in almost all of the low-risk group (except perhaps in those with a strong family history or extreme risk factor levels). It is also clear that the utility of screening will vary on the basis of predicted risk—the yield of finding a high CAC score is low in those at low predicted risk, and a CAC score of 0 is increasingly uncommon with higher predicted risks, as previously shown by Okwuosa et al. (8) and confirmed by Nasir et al (7). In the broad middle, perhaps from 5% to 20% 10-year ASCVD risk, there is room for the patient-clinician discussion espoused by recent guidelines (1,2), which could well be informed by judicious use of CAC screening. Starting with a quantitative risk-based assessment, the patient and clinician first calculate the 10-year risk. If, after discussion, they are uncertain whether the individual patient is likely to benefit from initiating a statin, obtaining a CAC score would be reasonable. Finding a CAC score of 0 in someone otherwise thought to be in a net benefit group is a powerful reason to consider withholding statin therapy. Likewise, the presence of a high CAC score in an individual at only moderate predicted risk should be a powerful motivator to initiate and adhere to statin therapy. A screening trial to test this approach is needed. In the meantime, sequential screening with judicious use of CAC scoring seems appropriate when the patient and clinician want more information to guide their decision.
↵∗ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
Dr. Lloyd-Jones is a MESA investigator; and he has reported that he has no relationships relevant to the contents of this paper to disclose.
- American College of Cardiology Foundation
- JBS3 Board
- Stone N.J.,
- Robinson J.G.,
- Lichtenstein A.H.,
- et al.
- McClelland R.L.,
- Jorgensen N.W.,
- Budoff M.,
- et al.
- Nasir K.,
- McClelland R.L.,
- Blumenthal R.S.,
- et al.
- Nasir K.,
- Bittencourt M.S.,
- Blaha M.J.,
- et al.
- Okwuosa T.M.,
- Greenland P.,
- Ning H.,
- et al.