Electron-Beam Computed Tomography for Coronary Calcium: Title and subTitle BreakA Useful Test to Screen for Coronary Heart Disease?

Electron-beam computed tomography (EBCT) uses x-rays generated by electron-beam irradiation of a tungsten target to identify and quantify coronary artery calcium (CAC).¹ Initial studies focused on the use of EBCT for risk stratification or diagnosis among individuals with clinical symptoms of coronary heart disease (CHD).^{1 - 3} However, more recent articles,^{4 - 14} as well as marketing by for-profit companies,^{15 - 16} have focused on using EBCT for screening; that is, for detection of CHD in asymptomatic individuals.

When considering the potential value of EBCT for screening, it is critical to keep in mind generally accepted principles for what constitutes an effective screening test^{17 - 18} :

Coronary heart disease is the leading cause of death among men and women in most nations. In the United States, CHD accounts for 2 million hospitalizations, 850 000 myocardial infarctions, 660 000 deaths, and $133 billion in costs annually.¹⁹ In autopsy studies of war casualties, atherosclerotic lesions were found in men in their 20s and 30s,²⁰ indicating an extended subclinical stage for detection of asymptomatic disease. Initial symptoms can be catastrophic. More than 50% of CHD deaths occur outside the hospital, many as the first manifestation of CHD.¹⁹ Myocardial infarction survivors are also at increased risk for subsequent morbidity and mortality.¹⁹ Thus, waiting for symptoms to appear before providing treatment is too late for many patients. These features make CHD an attractive target for screening.

To evaluate the efficacy of EBCT for screening (rather than diagnosis), individuals without known or clinically suspected CHD must be tested and then followed up for clinical events. Alternative end points, such as findings on coronary angiography or perfusion imaging, do not differentiate between relevant subclinical cases (the cases who will progress to clinical disease) and “pseudodisease” (subclinical cases who would never progress to clinical disease or would die from another cause).

Six cohort studies have investigated the ability of EBCT to predict risk of future clinical events in asymptomatic individuals, totaling 24 622 participants with 754 incident events during 103 473 person-years follow-up.^{4 - 9} Studies varied in clinical outcome examined (combined cardiovascular events to total mortality), mean follow-up (1.6-6.3 years), and highest CAC cut point evaluated (≥107 to >1000). In each study, individuals with an abnormal test result (3%-26% of the population at the highest CAC cut point, depending on the study) had several-fold higher risk than the overall population, with between 4.6% and 15.4% experiencing a clinical event compared with 1.5% to 8.2% overall. Nevertheless, the majority of individuals with an abnormal test result (approximately 85%-95%) did not experience a clinical event during follow-up. A normal or negative screening test at the highest CAC cut point was less helpful, with clinical risk only slightly lower than the overall population risk (ie, the risk predicted by chance alone).^{4 - 9}

A positive EBCT appeared least discriminatory among women (risk with a positive test, 4.6%; overall population risk, 2.2%),⁶ although in higher-risk subgroups (women smokers or women with hypertension), a positive EBCT predicted a cumulative incidence of events more similar to that observed in men (10%-12%).¹² This heterogeneity is indicative of spectrum bias or effect,²¹ variation in test performance depending on the population examined. This may reflect true biological heterogeneity or a Bayesian phenomenon (different underlying risks of the screened populations).

Determinations of clinical risk depended on duration of follow-up, and shorter or longer follow-up periods could modify results. Individuals undergoing screening (and their physicians) were not blinded to EBCT results, so determinations of test performance could be biased if the screening test result altered subsequent treatment (work-up bias). For example, individuals with higher CAC scores may have been more likely to seek or receive care leading to elective coronary revascularization, causing overestimation of clinical risk in studies including revascularizations in the end point.

Neither observational studies nor randomized trials have reported the effect of EBCT screening on clinical outcomes. Without an appropriate comparison group of individuals not receiving EBCT, potential effects of EBCT screening on therapy cannot be directly evaluated. In the absence of such evidence, plausible clinical scenarios can be considered to assess potential consequences of screening. These suggest that most preventive treatment goals would not be altered (Figure).

Could EBCT screening affect treatment goals for serum cholesterol? Current recommendations are based on CHD risk factors and Framingham risk, with persons having a 10-year risk of less than 10% receiving less intensive therapy; persons having at least 2 risk factors and a 10-year risk of less than 20% receiving intermediate therapy; and persons having clinical CHD, diabetes mellitus, or a 10-year risk of more than 20% receiving maximal therapy.²² Stratified by Framingham risk, for individuals with a positive EBCT-screening test (CAC score above the highest cut point), the cumulative incidence of events during 4- to 6-year follow-up was 4% to 5% for a Framingham risk of less than 10%, 10% to 14% for a Framingham risk of 10% to 20%, and 16% to 20% for a Framingham risk of more than 20%.^{5 ,7} Although observed event rates would be higher with longer follow-up, these results suggest that within each Framingham risk category, a positive EBCT may not substantially alter cholesterol treatment guidelines.

Conversely, a negative EBCT might alter cholesterol treatment goals. For individuals with a CAC score below the lowest cut point, the cumulative incidence of events was 0% for a Framingham risk of less than 10%, 1% to 4% for a Framingham risk of 10% to 20%, and 9% for a Framingham risk of more than 20%.^{5 ,7} This suggests that a low CAC score could plausibly reduce an individual's risk categorization and allow less intensive preventive treatment than currently recommended. Although this would not reduce CHD morbidity or mortality in the screened population, the usual goal of screening programs, it might result in more cost-effective treatment of the population. Additional studies with longer follow-up are needed to confirm the event rates in strata of Framingham risk predicted by negative or positive EBCT-screening tests.

Other potential EBCT-screening scenarios could be considered. Highly selected populations could be screened (eg, nondiabetic individuals with a Framingham risk of 10%-20%). However, as with general screening programs, targeted screening would not change goals for most preventive therapies, such as management of smoking, hypertension, obesity, diabetes mellitus, physical inactivity, or poor dietary habits, and data are unavailable to determine whether targeted screening would change cholesterol treatment goals. Screening programs could use multiple CAC cut points, allowing prediction of low clinical risk (<2%) with very low scores and modest risk (12%-15%) with very high scores. However, it is still unclear how a very low or very high score would modify most preventive treatments. In addition, many individuals' scores would be in the large intermediate category. The usefulness of other EBCT screening scenarios, such as serial testing over time, is similarly unknown.

Thus, plausible clinical scenarios do not indicate how EBCT screening would alter most preventive treatments received in the absence of screening. It could be argued that EBCT screening per se would bring individuals, and their untreated risk factors, to medical attention. However, this is an unconvincing argument when less expensive and simpler programs could be used for such a goal. It could also be argued that a positive EBCT would increase patient motivation, so that clinical benefit would occur due to improved adherence even with similar treatment recommendations. On the other hand, a negative EBCT could decrease patient motivation and adherence, causing potential harm in such individuals. Potential effects of EBCT screening on patient behavior, both positive and negative, require investigation.

The same attention and resources that have been given to EBCT screening^{4 - 14} should now be devoted to investigating the effectiveness of such screening. Large-scale randomized trials are needed to determine whether a screening strategy based on EBCT would result in fewer myocardial infarctions or premature deaths, after careful consideration of screening scenarios and populations in which EBCT might best alter management and improve outcomes. Although the design and implementation of such trials would not be trivial, randomized trials of screening strategies for breast cancer,²³ colon cancer,²⁴ and aortic aneurysms²⁵ have demonstrated that such studies are feasible and informative. Promotion of EBCT for screening cannot be justified until clinical benefit is demonstrated.

Fast imaging times and electrocardiographic-gating allow EBCT to be performed in 1 to 2 breath holds, and intravenous access and contrast administration are not required. These considerations minimize risks, and the large numbers of individuals voluntarily obtaining EBCT^{4 - 9} indicate that the procedure is acceptable to potential screening populations. The radiation exposure raises concerns.²⁶ Although little evidence exists for substantial harm from a single scan in a nonpregnant adult, even small increases in risk may be noteworthy in a population screening program, particularly among individuals with negative test results (who generally do not receive benefit from screening).

Costs for EBCT vary but typical charges are approximately $500.²⁷ Such amounts may be cost-effective for CHD diagnosis in patients with clinical symptoms.²⁷ However, cost-effectiveness for screening of asymptomatic individuals is not established, in part because the ability of EBCT screening to reduce morbidity or mortality is unknown. In 1 risk model,²⁸ the addition of EBCT screening to the Framingham risk score resulted in a marginal cost of $86 752 per quality-adjusted life-year saved, assuming a 30% improvement in life expectancy associated with EBCT screening, which may be optimistic given that EBCT screening may not alter most preventive treatments received in the absence of screening. Assuming a 25% reduction in mortality from early intervention due to EBCT screening, the cost was $1 700 000 per quality-adjusted life-year saved.²⁸ The practicality of EBCT screening might also be limited by the availability of scanners, particularly in rural areas or outside the United States.

In addition to demonstrating risk reduction among individuals with true-positive test results, the consequences of screening in the entire target population must be considered. False-positive screening tests may result in further unnecessary testing or treatment. Based on evidence to date,^{4 - 9} approximately 85% to 90% of screened individuals with a positive EBCT would not experience a clinical event over the ensuing 4 to 6 years. Follow-up testing, which may include invasive procedures, and adverse effects from use of unnecessary medications create real possibilities of harms. Insurability or insurance rates could also be affected by CAC results. Thus, possible benefits of early diagnosis and treatment in the minority of persons who would have sustained a clinical event must be weighed against harms of unnecessary testing and treatment in the majority of individuals who would not have sustained a clinical event. Negative screening results may also have consequences, reducing attention to or compliance with other preventive therapies, lifestyle habits, or follow-up. Thus, the sum of effects in the entire screened population must be evaluated.

Given public health consequences and extended subclinical progression, CHD is an appropriate target for screening. Although EBCT augments prediction of CHD risk in asymptomatic individuals, the relatively low underlying population risk results in only a small proportion (<15%) of individuals with a positive test experiencing a clinical event in the ensuing 4 to 6 years. Given the often severe initial clinical presentation of CHD, such modest predictive value may be acceptable for screening. However, the ability of early detection by EBCT to reduce morbidity or mortality—a fundamental goal of screening—has not been convincingly demonstrated. Furthermore, the cost-effectiveness and consequences of negative tests and false-positive tests are unknown. Given uncertain clinical benefit and the possibility of some individuals experiencing harm, large-scale randomized trials of EBCT screening are needed. Based on the principle of primum non nocere, use of this promising modality for detection of subclinical CHD in asymptomatic individuals cannot be advocated until additional evidence demonstrates that benefits outweigh harms.