Author + information
- Received March 31, 2008
- Revision received April 15, 2008
- Accepted April 21, 2008
- Published online July 29, 2008.
- Philip Greenland, MD⁎ and
- Donald Lloyd-Jones, MD, ScM
- ↵⁎Reprint requests and correspondence:
Dr. Philip Greenland, Northwestern University, Feinberg School of Medicine, 750 North Lake Shore Drive, 9th Floor, Chicago, Illinois 60611.
Screening for coronary artery disease or for other forms of cardiovascular disease (CVD), is a superficially attractive idea that has been proposed by many cardiovascular experts. However, guidelines panels that employ medical outcomes and evaluation of a full range of benefits, risks, and harms of screening have repeatedly advised against the adoption of screening programs for CVDs. In this commentary, we discuss the limitations of the predictive capacity of selected cardiovascular screening tools. Possible harms and risks of screening tools are also discussed, and the conclusion is reached that available evidence does not provide support for adoption of a strategy of routine screening of asymptomatic people for CVD with available testing tools or devices. More evidence is needed to justify a policy of screening for CVD.
Screening for presence of atherosclerotic coronary artery disease (CAD) or cardiovascular disease (CVD) in general, as proposed by Cohn and Duprez in this issue of the Journal (1), is superficially attractive. The widely accepted World Health Organization-defined criteria for screening (2) for disease in medicine are shown in Table 1, and CVD satisfies many of these criteria. However, as we discuss in the following text, screening for CVD does not meet several of the criteria considered essential for screening tests (Table 1), leading us to different conclusions than Cohn and Duprez (1).
Despite the immediate appeal of cardiovascular screening with such common tests as coronary calcium measurement (3), vascular ultrasound for intimal-medial thickness (4), or a host of other tests as described by Cohn and Duprez (1), routine screening for subclinical CAD/CVD has met with nearly universal negative recommendations by expert panels (5–7). If medical screening of this type is feasible and well-accepted for certain forms of cancer and patients are willing to purchase screening tests even without a personal physician's referral (8), what is the problem?
Although Cohn and Duprez (1) criticize traditional cardiovascular risk factors as being poorly predictive of cardiovascular events, they do not consider the substantial limitations of prediction with the screening tests that they advocate. A critical assessment of one of the most promising of the methods for early detection of CAD, coronary artery calcium (CAC) measured by rapid computed tomography, provides enlightening insights into the negatives of such screenings. On the one hand, recent reports from MESA (Multi-Ethnic Study of Atherosclerosis) demonstrated convincingly that CAC testing is predictive of coronary events in whites, blacks, Hispanics, and Chinese-Americans, and MESA found that increased CAC scores are predictive over and above traditional risk factors in both men (9) and women (9,10). Despite showing unusually strong hazard ratios for major coronary events ranging as high as approximately 7 compared with a reference group with CAC = 0, MESA results also showed that CAC has remarkably poor positive predictive value for near-term events regardless of the cutoff chosen for a positive test (9). Depending on the cutoff chosen for a positive test, sensitivity is also relatively poor. As shown in Table 2, for a cutoff value of CAC >0 as the definition of a positive test in MESA, sensitivity for near-term events was 91% but specificity was only 51% in this population of relatively low-risk people (major coronary event rate during 3.9 years of median follow-up was slightly above 1%, with estimated 10-year risk extrapolated to approximately 5%) (9). With a cutoff of CAC >0 for a positive test, positive predictive value was only 2%. Furthermore, the positive predictive value would be substantially lower in younger populations than in the MESA cohort, which had a mean age of 62 years.
Screening, by definition, involves the routine evaluation of asymptomatic people, and detection, not exclusion, of disease is the primary goal. This certainly seems to be the expected outcome of the approach proposed by Cohn and Duprez (1). In MESA (9), when a more stringent cutoff of CAC >100 is used, sensitivity for near-term major coronary events drops markedly to only 63%, and positive predictive value remains low at 3.5%. Thus, although hazard ratios are impressive for CAC and far exceed those for carotid intima-media thickness (4,11), C-reactive protein (9,12), or any of the traditional risk factors for CHD (9), positive predictive values for CAC measurement are poor. Indeed, even with these impressive hazard ratios, the C-statistic, which is a measure of discrimination, improves only from 0.79 in MESA for risk factors alone as a predictor of CHD events to 0.83 for risk factors plus CAC score. Although this increase in the C-statistic is consistent with improved discrimination, the value of 0.83 indicates that CAC scoring on top of risk factor measurement still has important limitations in prediction of events—another fact ignored by Cohn and Duprez (1).
An important distinction between cancer screening tests and those available for cardiovascular risk assessment is availability (or lack for CVD tests) of cost-effectiveness or outcomes data. We do not have a full understanding of the risks of CVD screening; nor do we have a full accounting of benefits and costs. These issues have been the major cause for lack of endorsement of such tests as coronary calcium scoring for the general population by groups that include the U.S. Preventive Services Task Force (USPSTF) (5) and the U.K. National Health Service Research and Development Health Technology Assessment Program (6). Not only do we lack evidence that these sorts of screening confer a benefit, we know that there is also risk of harm. Harms of screening include false reassurance due to a test with low sensitivity, anxiety due to a false positive test, and in the case of certain tests such as mammography and coronary computed tomography, radiation exposures as well as the additional anxiety and possible risks due to discovery and evaluation of incidental findings that result in considerable added medical expense and invasive procedures (13). The added burden of follow-up invasive angiography (that might or might not be necessary) observed after coronary calcium scoring is an object lesson that screening for vascular disease is not without obvious although often neglected hazards.
There should be no double-standard in comparing cardiovascular screening tests and cancer screening tests. As noted by Cohn and Duprez (1), mammography is recommended by the USPSTF, with a quality rating of B (i.e., this level recommends that clinicians provide this service to eligible patients on the basis of at least fair evidence from clinical trials that the service improves important health outcomes and concludes that benefits outweigh harms). Evidence is strongest for women ages 50 to 69 years, the age group generally included in screening trials (14). With the identical review standards, the USPSTF (5) recommended against routine screening with resting electrocardiography, exercise treadmill test, or rapid computerized tomography scanning for coronary calcium for either the presence of severe coronary artery stenosis or the prediction of coronary heart disease events in adults at low risk for CHD events (i.e., screening of the general asymptomatic population). The USPSTF concluded that there was at least fair evidence that these tests were ineffective or that harms outweigh benefits (5).
The preceding discussion points regarding coronary calcium measurement pertain to what is arguably the best and best-studied screening test currently available for detecting subclinical CAD. Our concerns are magnified and far more meaningful when applied to measures with large potential sources of error (relative to signal), such as carotid intima-media thickness and arterial stiffness measures. There are substantial difficulties associated with broad application of these newer technologies in clinical practice, including: technical issues, especially technician-dependence of scanning; reader error and bias; and lack of agreement on the most appropriate parameters to be measured and the best instruments to use. It is also largely unclear how often (if ever) any of the screening tests proposed by Cohn and Duprez (1) should be repeated or what a meaningful change in these values is.
In our judgment, additional studies of screening strategies, including patient outcomes, are needed before recommending any form of routine CVD screening for the asymptomatic patient. As an example of the complexity of this issue, a study of coronary calcium screening among Army personnel (15), which attempted to use knowledge of subclinical CAD to motivate patients to make evidence-based changes in risk factors, indicated that such knowledge was not associated with improvement in modifiable cardiovascular risk at 1 year. In that study (15), case management was superior to usual care in the treatment and control of risk factors. The “obvious” and expected benefits of screening did not occur. Although not definitive throughout the entire spectrum of possible patients and potential levels of cardiovascular risk, these findings demonstrate that we cannot assume clinical benefit simply because the tests we employ are feasible and have anatomical or physiological face validity.
In conclusion, we agree with the USPSTF (16,17) that the best approach at this time for CVD assessment and prevention lies in routine testing for traditional coronary risk factors. As the USPSTF has noted (16,17), this is a proven approach to identifying a group of patients who can benefit from preventive treatments shown in clinical trials of blood pressure lowering and cholesterol lowering. In contrast to the more controversial forms of screening previously discussed, the USPSTF gave blood pressure and cholesterol testing and screening its highest grade of quality—an “A” grade. Accordingly, the USPSTF strongly recommends that clinicians routinely provide these risk-factor measurements to eligible patients on the basis of solid evidence that these services improve important health outcomes and the benefits of such tests substantially outweigh the harms (16,17). On the basis of the considerations we have discussed, we believe this is the best and most rational current approach to “screening” for CAD risk.
- Abbreviations and Acronyms
- coronary artery calcium
- coronary artery disease
- coronary heart disease
- cardiovascular disease
- Received March 31, 2008.
- Revision received April 15, 2008.
- Accepted April 21, 2008.
- American College of Cardiology Foundation
- Cohn J.N.,
- Duprez D.A.
- Greenland P.,
- Bonow R.O.,
- Brundage B.H.,
- et al.
- ↵(2004) Screening for Coronary Heart Disease, Topic Page (U.S. Preventive Services Task Force. Agency for Healthcare Research and Quality, Rockville, MD) February, http://www.ahrq.gov/clinic/uspstf/uspsacad.htm. Accessed March 30, 2008.
- Waugh N.,
- Black C.,
- Walker S.,
- McIntyre L.,
- Cummins E.,
- Hillis G.
- Mosca L.,
- Banka C.L.,
- Benjamin E.J.,
- et al.,
- Expert Panel/Writing Group
- Kassirer J.P.
- Folsom A.R.,
- Kronmal R.A.,
- Detrano R.C.,
- et al.
- Wald N.J.
- ↵(2003) Screening for Breast Cancer, Topic Page (U.S. Preventive Services Task Force. Agency for Healthcare Research and Quality, Rockville, MD) November, http://www.ahrq.gov/clinic/uspstf/uspsbrca.htm. Accessed March 30, 2008.
- ↵(2001) Screening for Lipid Disorders in Adults, Topic Page (U.S. Preventive Services Task Force. Agency for Healthcare Research and Quality, Rockville, MD) March, http://www.ahrq.gov/clinic/uspstf/uspschol.htm. Accessed March 30, 2008.
- (2007) Screening for High Blood Pressure, Topic Page (U.S. Preventive Services Task Force. Agency for Healthcare Research and Quality, Rockville, MD) December, http://www.ahrq.gov/clinic/uspstf/uspshype.htm. Accessed March 30, 2008.