Task force #4—how do we select patients for atherosclerosis imaging?

Improved diagnosis

The goal of cardiovascular disease (CVD) screening is to accurately determine risk early in the natural history of disease. Adding subclinical disease markers to traditional CVD risk-factor screening has the potential to facilitate more appropriate, targeted interventions that will further reduce CVD morbidity and mortality in clinical and population-based settings. Various studies have determined that subclinical disease markers of atherosclerosis improved the ability to identify the subset of individuals who are at increased risk for CVD outcomes. Examples of specific markers that have been shown to provide additional information beyond traditional CVD risk factors include ankle brachial index (9,10)and carotid intima-media thickness (IMT) (11–13). For example, based on these data, it is logical to anticipate that the addition of noninvasive markers of atherosclerosis may enhance our ability to diagnose the amount and potential severity of early/asymptomatic CVD. Other atherosclerosis markers (magnetic resonance imaging [MRI], coronary artery calcium (CAC), and brachial artery vasoreactivity) appear to have potential but do not yet have the depth of scientific evidence documenting their validity, reproducibility, and value in predicting CVD events beyond risk factors (14–16).

Incremental management impact

A major potential benefit of screening for atherosclerosis is to enhance CVD prevention strategies. The ability to select higher risk asymptomatic subsets from the population that would benefit from either an earlier or more aggressive risk factor intervention strategy is a key advantage of subclinical disease screening. Theoretically, if these additional markers are used, preventive measures (lifestyle interventions and/or pharmacologic interventions) can be implemented earlier in the course of disease, with the potential not only to reduce the burden of clinical outcomes but also to reduce subsequent subclinical or atherosclerotic disease progression. Observational data are key to improving management of vascular disease, but diagnostic imaging utility should also be tested with randomized clinical trials.

Published studies that used noninvasive CHD risk assessment in this situation have generally not been restricted to prespecified initial probabilities, and some have been limited by selection, observer, and publication bias. Undertaking an experimental design, including blinding the involved patients and their physicians would allow rigorous testing of the utility of the new procedures. Appropriate exclusion criteria within such an experimental design would be necessary to address concerns over withholding information for persons with very “abnormal” test results. Alternatively, rigorous analysis of testing strategies in this situation might be undertaken by randomizing patients to testing or no testing, then prospectively assessing outcomes.

It is important to frame both the testing schema and the hypothesis that would be tested. The null hypothesis would be that the newer noninvasive testing provides no additional benefit beyond the traditional risk-factor assessment. For example, information from the new diagnostic procedure would be put into a Cox prediction model that included a CHD risk estimate score and results for the new diagnostic test. A statistically significant relation between the new variable and the outcome in the statistical model would provide evidence of the incremental utility of the new diagnostic procedure. The noninvasive test score could be considered in various ways to test the hypothesis—the data could be as a continuous variable, as a “positive” test, or as a “negative” test. It is also possible that the utility of a “negative” test that significantly decreased the posterior probability of disease would be helpful in terms of clinical care, as aggressive therapy for persons with abnormal risk factors but little risk of disease would be useful information. Important interpretive considerations include both the presence and the clinical relevance of the observed results.

Young versus old patients

Assigning the same Framingham Risk Score (FRS) points to all individuals of the same chronological age does not take into account the great variation in plaque burden at a given age. More accurate determinations of risk through measurement of subclinical atherosclerosis may also be useful in older people as a way to determine one’s biological age rather than simply one’s chronological age. The Adult Treatment Panel (ATP) III pointed out that measurement of coronary calcium may be useful for older persons in whom traditional risk factors lose some of their predictive power. A high CAC score may “tip the balance in favor of a decision to introduce low-density lipoprotein (LDL)-lowering drugs for primary prevention in older patients” (17).

No studies have directly compared the accuracy of multiple imaging modalities for cardiovascular prognosis across a broad age range of patients. Furthermore, the practicing physician would optimally seek to couple accurate information on cardiovascular risk to a change in management that appreciably alters that risk. Whereas brachial artery reactivity testing and MRI are potentially more suited for atherosclerosis assessment in younger individuals (Task Force 3) in whom absolute cardiovascular risk is expected to be relatively low, a shift in management to more vigorous recommendations for lifestyle interventions would be more likely than an alteration in the use of pharmacologic therapies. In comparison, CAC detection and carotid ultrasonography may be best matched to middle-aged and older individuals where the data related to cardiovascular prognosis are most robust. Finally, the utility of ankle-brachial index (ABI) testing may be limited to older patients in whom even asymptomatic abnormalities could alter the approach to cardiovascular risk reduction.

Men versus women

No data indicate a clear role for gender in the selection of atherosclerosis imaging for cardiovascular risk detection. However, the recognition of gender differences in the prevalence and severity of abnormalities found with individual modalities has importance in rendering accurate risk prediction. For example, CAC scores and IMT values are generally lower in women than in men, although the relative risk attached to an individual test value may exceed that seen in men. Once women are postmenopausal, atherosclerosis imaging in men and women appears to perform comparably, as shown in a recent sample from the Framingham Offspring Study. In a stratified sample of 318 men and women with a mean age of 60 years studied with electrocardiogram (ECG)-gated magnetic resonance scanning, evidence of aortic atherosclerosis was present in 38% of the women and 41% of the men. In both genders the presence of atherosclerotic plaque was correlated with the Framingham risk score (18). In middle-aged women, because false positive exercise stress tests are common, atherosclerosis imaging may be more cost-effective than traditional noninvasive testing (19).

Ethnic differences in subclinical disease

Carotid ultrasonography is predictive of cardiovascular outcomes in both black and white individuals, although differences in the extent and location of carotid atherosclerosis varies somewhat by race. In the Cardiovascular Health Study (CHS), including a limited sample of 244 black adults at least 65 years of age, common carotid walls were thicker and ABI ratios were lower in blacks of both genders, whereas internal carotid walls were thinner in black women, after adjusting for traditional CHD risk factors (20). The relationship between race and carotid atherosclerosis varies depending on the site of analysis. For example, in both the Insulin Resistance Atherosclerosis Study (IRAS) and the ARIC study, blacks had the same or less atherosclerosis in the proximal internal carotid artery, yet greater atherosclerosis at other carotid sites (21). Among Hispanics, atherosclerosis in the common carotid artery was less severe than that of whites, after risk-factor adjustment (22). It is unclear whether the relatively minor quantitative differences in these measured carotid atherosclerosis values would cause a shift in the clinical cardiovascular risk assessment and subsequent cardiovascular management.

The relationships between CAC and race are similarly complex. Several groups have found that blacks have less CAC than whites at middle age and older (23–25). In one study, despite the finding that the prevalence of CAC was 36% in blacks and 60% in whites, black participants sustained more CVD events than did whites during 70 months of follow-up (23). Few data are available for CAC assessments in other ethnic groups. These data suggest caution in applying CAC assessments to ethnic minorities until ethnic-specific outcome studies have been completed.

Diabetes mellitus and renal disease

Although diabetes mellitus is classified as a CHD risk equivalent and, thus, the diagnosis of subclinical CHD might not be expected to shift the management strategy, recent data from Kuller et al. have challenged this notion (23a). Examining a population of diabetic subjects in CHS with carotid ultrasonography, the investigators detected a significant gradient of cardiovascular risk in diabetics associated with the presence of subclinical atherosclerosis. In that study, the presence of subclinical atherosclerosis increased the risk for incident CHD by 100%. Similar data are not yet available for other imaging modalities. In a study of asymptomatic diabetics, no significant age differences were seen in CAC scores between women and men (26). This suggests that the premenopausal protection afforded women in the development of CAC is lost, and potentially extends the relevance of coronary calcium scanning to women diabetics of younger age. Thus, testing for subclinical atherosclerosis, even in a clinical high-risk group, appears to modulate the coronary risk assessment. Although such a finding is unlikely to broadly alter the management of these patients, such data could lead to increased vigilance on the part of patients and providers for the warning signs of CHD, particularly in a setting of limited financial or personnel resources. Similar arguments potentially apply to patients with end-stage renal disease (ESRD), another high-risk group for CVD. An elevated level of coronary calcification is seen in ESRD patients at a much younger age than in the general population. Even young adults on dialysis may have rapidly progressive CAC. Whether the detection of subclinical atherosclerosis in such high-risk populations can meaningfully direct therapies to achieve enhanced patient outcomes should be the subject of clinical trials.

Individuals with a family history of premature CVD

The FRS does not take into account family history as family history analysis did not demonstrate sufficient incremental risk for a family history of premature CHD to be included in the risk assessment equations. Nevertheless, a large body of case-control and cohort studies report that a family history of premature CHD independently predicts CHD events. These discrepant findings may be due to the way in which family history was assessed in the various studies. It appears that the risk for CHD is higher the younger the age of onset in the affected family member and the greater the number of affected first-degree relatives (3).

Recently, Valdes et al. (27)reported that CAC was more prevalent in asymptomatic adults with a positive family history for premature CHD (male first degree relative less than 5 years and female less than 65 years). Traditional risk factors accounted for only 20% to 30% of the variance in calcium score. This study included only whites; subjects with diabetes, poorly controlled hypertension, current smoker, or cholesterol greater than 300 mg/dl were excluded. A measure of subclinical atherosclerosis such as coronary calcification determination may be very helpful in persons with a family history of premature coronary disease, because this risk factor is not accounted for in the FRS.

Screening for atherosclerosis in “real world” settings

It is important to note that the vast majority of data that documents the importance of subclinical disease markers to predict CVD outcome has been collected in highly controlled research settings (28,29). Thus, excellent quality control measures, very detailed protocols, and highly trained personnel were involved in all phases of the imaging and reading components. Translating the results of clinical studies to real-world settings will require similar attention to quality control and accuracy. Without such controls, the potential exists for misclassification of subclinical disease, resulting in errors in the cardiovascular risk assessment.

False positives

Definitions of a positive test procedure are necessarily problematic for a test that is used prognostically without immediate clinical and pathologic correlation. Large-scale observational projects, such as the National Institute of Health (NIH)-sponsored investigation entitled “Multi-Ethnic Study of Atherosclerosis” (MESA) that is underway, will address this issue. It is inevitable that any screening program would have false positives owing to variability in measurement of subclinical atherosclerosis (e.g., improper imaging of the carotid artery via ultrasound, errors in reading, transposition of data). Although the proportion of false positives would be expected to be relatively small, the aggregate impact on the number of misclassified individuals would be increased should a screening program be implemented on a large scale. The adverse impact of a false positive test is that individuals will be unduly alarmed and perhaps would be subjected to a more aggressive treatment course than would be warranted based on their “true” risk of CVD.

False negatives

Just as with any screening test, the potential exists for false negatives. A variety of reasons exist for not detecting “true” subclinical atherosclerosis. Similar to the case for false positives, variability in measurement or reading techniques could also result in classifying an individual with atherosclerosis as being disease-free. In addition, other examples may include being unable to identify a noncalcified atherosclerotic lesion using a computed tomography (CT) scan or when carotid ultrasound focuses on a specific area of the vascular bed and misses an adjacent area with a significant plaque. As reported by Detrano et al. (15), some coronary disease events occur in persons with CAC scores less than 75 Agatston units, and both physicians and the public should be aware that CAC evaluations help to define prognosis but are not definitive. Similarly, IMT scores in the top quintile were predictive of later CHD in the CHS cohort, but the overall vascular disease risk in this cohort was high, and individuals with “negative” tests also experienced events relatively commonly during follow-up (13). Individuals should not be given a false sense of security when a test is “negative,” thereby missing an opportunity to reduce the burden of atherosclerotic disease by treating a known risk factor.

Incidental findings

In the process of conducting an assessment for subclinical atherosclerotic disease measures, there is the potential for the identification of other incidental findings (either nonatherosclerotic or non-CVD findings). For example, when noncoronary pathology in the field of view is assessed in studies using CT scanning to screen for CAC, approximately 20% of participants have other findings (ranging from benign calcified nodules to undiagnosed lung cancer). Although identification of asymptomatic disease may be of benefit to some individuals, a substantial burden is placed on these participants and their health care providers to determine if additional diagnostic tests or treatment are required. Thus, these incidental findings may result in increased health care costs to rule out other disease processes and may cause undue anxiety on an individual basis. Detection of incidental findings is much less frequent for some other subclinical atherosclerosis modalities (i.e., ABI, carotid IMT, brachial artery endothelial function) in which imaging is limited to a specific vascular bed location.

Effects on insurability

Clinical events certainly impact on both an individual’s long-term prognosis and their cost of obtaining insurance. Subclinical disease is highly related to the potentialfor the development of CVD events and should be considered a modifiable factor. It remains unclear how data collected in a subclinical atherosclerosis screening program would be used by actuaries in underwriting life insurance and individual health insurance policies. Normative data for IMT, MRI, ABI, and CAC have not been scrupulously developed with the same degree of accuracy and precision as some other diagnostic testing, such as cholesterol and blood pressure measurement. Knowing more about individual CVD risk can be beneficial to an individual’s health, but the question remains as to whether collection of these data on a high-risk individual will increase the person’s cost of obtaining insurance.

Other considerations

When considering the risk benefit of subclinical atherosclerosis screening, it is important to state that CAC assessment using CT involves exposure to ionizing radiation. Although discovery of subclinical atherosclerosis in the coronary bed may change the CVD treatment strategy and be a valuable addition to an individual’s care, the radiation exposure from this test should be considered as we determine the appropriateness of using this test in low-risk individuals or as part of a nationwide screening program. Finally, cost is a major consideration of screening, and that topic is discussed in the Task Force 5 report.

Benefit and harm of self-referral for noninvasive testing for atherosclerosis

Access to noninvasive testing varies greatly and depends on the type of test and the extent to which the test is available and commercialized. The concern of self-referral access to atherosclerosis imaging is for a patient to be either falsely reassured if the result is negative, despite significant risk factors, or needlessly alarmed with a result that could be very common and may not pose immediate risk.

Despite convincing evidence from population-based studies showing increased ABI to predict a wide range of cardiovascular end points, few physicians currently use ABI in clinical screening, and the financial incentives have not been established. This may be partly due to the absence of advertising and commercialization of this test, which limits its current use to primarily a research tool and not to widespread use as a self-referred test by the public. The relatively low cost of equipment and performance of the required measures for determination of ABI, and its ability to detect subclinical peripheral arterial disease, suggest there may be benefit for self-referral in certain populations at potential risk, such as persons aged 50 years and over or those with multiple risk factors (1).

Magnetic resonance imaging of atherosclerotic plaque has great promise to noninvasively image the high-risk vulnerable plaque and allow serial evaluation of the progression/regression of atherosclerosis (30). At this time, the procedure is limited in availability and is used almost exclusively as a research tool. The expense and complexity of acquisition and interpretation limit this technology to a few research sites, suggesting self-referral is not appropriate at this time.

Carotid ultrasonography of IMT has been advertised in the form of “stroke screening,” with testing often being done in the form of mobile test teams. Although carotid IMT has clearly been shown to be associated with risk of cardiovascular events and stroke in large-scale population-based studies, guidelines do not exist to recommend specific follow-up above certain age- and gender-based cut points for IMT, nor how these recommendations may be modified according to an individual’s cardiovascular risk factor profile. In addition, the reproducibility of the measurement may be in question unless done at a highly skilled facility.

The success of numerous CT scanning centers has depended on self-referred asymptomatic patients, often with one or fewer risk factors, as a result of mass media advertising campaigns, despite the insistence by many physicians, including those overseeing such centers, that physician referral of persons with multiple risk factors is the most appropriate way to access the technology. In addition, although some centers have taken the initiative to provide one-on-one physician consultations that attempt to explain clearly the meaning of the results, this is not the typical practice. Some evidence suggests that patients with a positive scan may worry more and seek consultation with their physician, but may also try to lose weight, start a low-fat diet, or possibly comply better with cholesterol-lowering or blood pressure-lowering medicine (31). Although most of these could be construed as benefits, a potential disadvantage results from patients who have a negative scan, believing perhaps less determined to undertake healthful lifestyle changes or to comply with physician orders. The additive management impact of these tests has not yet been completely defined by rigorous clinical trials.

Does an abnormal atherosclerosis imaging test shift management to a secondary prevention strategy?

Primary prevention efforts for individuals with multiple risk factors may be considered insufficient when their cumulative risk is high enough that it is similar to patients with existing vascular disease. The Third Adult Treatment Panel of the National Cholesterol Education Program (ATP III, 2001) extended previous guidelines in recommending that those with two or more risk factors whose calculated 10-year risk of CHD exceeded 20%, or if diabetes, peripheral arterial, or symptomatic carotid disease were present, be treated as a CHD risk equivalent. Following this paradigm, the presence of a sufficient burden of subclinicalvascular disease could be construed as CHD risk equivalent when the additive risk of conventional risk factors (including patient age) and atherosclerosis burden exceeds the 2% annual threshold. In general, these threshold values require better definition for all modalities of subclinical atherosclerosis testing. However, among available modalities, sufficient evidence exists to recommend persons with peripheral vascular disease diagnosed by an ABI below 0.90 to be candidates for secondary prevention management. In such individuals, relative risks are similar to those seen in secondary prevention, considered a justification for moving a patient with apparent “intermediate risk” based on office risk assessment to high-risk status (1).

In the case of carotid ultrasound assessment of carotid IMT, epidemiologic data show significantly increased cardiovascular event risk among those with IMT of 1 mm or greater, or for persons in the highest quintile of IMT (32). Relative risks similar to that seen in secondary prevention have also justified that such individuals who would otherwise be considered intermediate risk should be elevated to a “coronary risk equivalent” (1). Moreover, such persons may be at greater risk of stroke than many of those whose 10-year CHD risk is estimated to be 20% or more, but who may not have increased IMT.

Currently, MRI can quantitate plaque burden in peripheral arterial beds (i.e., aorta and carotid), and it has the unique, but unproven, potential to morphologically characterize the vulnerability of atherosclerosis (30). However, absent large population-based data on MRI of atherosclerosis, including how such findings may relate to clinical events, no recommendations have been made as to whether persons with identified plaque (and to what extent) should be candidates for secondary prevention, although as such data accumulate in the future, experts may make such recommendations.

Efforts to use results from noninvasive testing for the purposes of risk stratification have been perhaps most active with coronary calcium imaging by CT. Rumberger et al. (33)first published guidelines recommending more aggressive risk factor modification efforts for persons with coronary calcium scores exceeding 400. Others have suggested that anyone with calcium scores at or above the 75th percentile, associated with substantially increased relative risks, to be candidates for treatment according to secondary prevention guidelines. Despite the lack of consistent recommendations, the current practice by numerous physicians is to consider a significant calcium score to warrant atherosclerosis that must be treated aggressively, as in the case of a person with known coronary artery disease (CAD). However, it will be several years before the results of the NIH-sponsored MESA are published. This study investigates the incremental value of CT coronary calcium scores for prediction of cardiovascular events over both standard and novel coronary risk factors. Data from other cohorts (34,35)as well as a recent meta-analysis from earlier studies (36), and other reports documenting significant calcium scores to signify clinically significant atherosclerosis, suggest the use of high calcium scores (400 or higher, or at or above the 75th percentile for age and gender) may be reasonable, among intermediate risk individuals (e.g., those with a premature family history of CHD or risk factors achieving at least a 10% risk of CHD over 10 years), to warrant aggressive treatment as a CHD risk equivalent.

The role of serial testing

Serial testing for evidence of subclinical atherosclerosis using coronary CT scanning (30,37)and other more experimental techniques (30)has been identified as an opportunity to study and track arterial changes in patients on medical therapies. Serial evaluation of coronary calcium by CT has been limited to studies from self-referred or clinical cohorts, where annual progression rates of 22% to 52% have been reported, with a wide range of interscan reproducibility (37). The highly variable estimates of CAC within individual patients, particularly if calcium scores are low, raises questions regarding the utility of serial scanning to track atherosclerotic disease. Ongoing observational and randomized clinical trials will help establish the validity of serial coronary calcium scanning as a surrogate measure of atherosclerosis risk and to test whether changes in CAC severity translate into an altered risk of coronary disease risk assessment.

The use of serial carotid ultrasonography for tracking of IMT has perhaps the strongest evidence base. Numerous clinical trials have documented the effect of treating dyslipidemia, blood pressure, and other risk factors to slow progression of IMT. Moreover, studies have also shown risk factor levels associated with progression of carotid IMT (38)to be greater in persons with, versus without, coronary artery disease (39), and in those with new coronary events (40). Despite these data, serial evaluation of carotid IMT is not widely used clinically, nor are widespread recommendations regarding the appropriateness and time frame for repeated assessments in either asymptomatic or symptomatic individuals. Moreover, should repeated carotid ultrasonography be performed, it is essential that such repeat scans be read by research-quality laboratories to ensure standardization. Such laboratories are not widely available in the U.S.

Cardiovascular MRI has great future potential as a means to track the progression of overall atherosclerotic plaque burden. One recently published clinical trial (41)showed significant reductions in atheroma plaque cross-sectional area resulting from simvastatin therapy over a treatment period of only 12 months.

For serial testing of atherosclerotic imaging modalities to be practical, such evaluations: 1) should be standardized to ensure accurate determination of change/progression, assessed by research-quality laboratories; 2) should be sufficiently reproducible—e.g., change deemed to be clinically significant should be substantially greater than intertest measurement error; and 3) there should be agreed-upon guidelines for more aggressive clinical management based on a known degree of progression. Although standardization of measurements can be acceptable and sufficiently reproducible for several of the imaging technologies, there is wide variation and great dependency on which laboratory is used, as well as in reading or evaluating images. Although more aggressive treatment might be recommended for those demonstrating progression of atherosclerosis, specific guidelines do not exist, in part because there are no currently agreed-upon criteria used to define clinically significant progression of disease for any of the imaging modalities reviewed in this report. Until then, however, routine serial testing of any imaging modality in patients receiving assessments of noninvasive testing of atherosclerosis is not recommended.

When is further testing (e.g., stress testing, invasive testing) required after atherosclerosis imaging?

In the above-noted guidelines first published by Rumberger et al. (33), potential further testing was suggested for persons with coronary calcium scores exceeding 400. More recently, Berman et al. (42)have recommended the use of coronary calcium screening in persons with a low-to-intermediate (0.15 to 0.50) pretest likelihood of CAD, and when scores are in the range of 100 to 400, recommending treatment according to AHA secondary prevention guidelines; for those with a score of 400 or greater, they suggest direct referral to a stress nuclear test. One preliminary report showed nearly half of those with a score of 400 or greater to demonstrate a positive nuclear scan test, although these persons who were tested both with a myocardial perfusion single-photon emission computed tomography (SPECT) test and electron-beam CT scanning for coronary calcium had other indications for nuclear testing, resulting in their referral (43). No guidelines exist for direct referral to coronary angiography or other invasive testing given a particular calcium score. In addition, no clear cut points exist for other atherosclerosis imaging modalities for referral to further diagnostic testing. It is clear that many variables determine whether a patient should be referred for further noninvasive or invasive diagnostic testing, such as medical history, presence, and extent of any current symptoms, as well as existence of other risk factors. Physicians should carefully evaluate these criteria in combination with the results from any atherosclerosis imaging tests in making a prudent decision as to the need and type of additional diagnostic testing.

Integrating clinical and atherosclerosis screening

Information obtained from noninvasive imaging of atherosclerosis can be valuable in refining risk-stratification efforts, particularly for intermediate-risk patients, which could comprise as much as 40% of the U.S. adult population (1). It is of interest to note that the 1999 AHA Prevention V Conference considered persons aged 50 or older or those at intermediate or higher risk of CHD to be possible candidates for ABI assessment or carotid B-mode ultrasonography. Specialized screening could possibly provide incremental value over standard risk factors in asymptomatic persons, justifying such use in an intermediate risk group. Algorithms have been proposed that use a Framingham risk score and arterial calcification and then the “age points” in the Framingham risk algorithm, based on the extent of coronary calcium (subtracting points if coronary calcium score is below the 25th percentile for age and gender and adding points if the score exceeds the 75th percentile) (44). Although this approach is reasonable, its validity has not yet been demonstrated. Rather, an individualized approach with respect to enhancing risk level in the presence of significant atherosclerosis detected from imaging techniques seems prudent at this time. A decision for atherosclerosis imaging should be based on physician recommendation and referral, but only after a careful consideration of known medical history and evaluation of major standard cardiovascular risk factors by office-based techniques.