Arad et al. (1- 2) have published two reports in this issue of the Journal that examine several issues related to coronary artery calcium (CAC) as a measure of coronary atherosclerosis and as a predictor of atherosclerotic cardiovascular disease (ASCVD) events. One report (1) indicates that CAC scores predict ASCVD events independent of standard risk factors and C-reactive protein (CRP). The second (2) reports that treatment of patients with a cocktail of alpha-tocopherol, vitamin C, and atorvastatin does not modify the progression of CAC over time. The latter study included a small trial of clinical end points. In this trial, the combination of therapies did not significantly reduce ASCVD events, although the small sample size and relatively low event rates appeared insufficient to provide a definite outcome.

To appreciate the significance of these reports it is necessary to examine the numbers in some detail. A striking feature of both studies was the unexpectedly low rates of major coronary events in persons who otherwise were projected to be at a relatively high risk. The first study (1) measured CAC in 4,903 healthy subjects and then followed them for an average of 4.3 years for a combined outcome of coronary event rates (including non-fatal myocardial infarction [MI], coronary bypass surgery, percutaneous coronary angioplasty, and coronary death), non-hemorrhagic stroke, and peripheral vascular surgery. Among all 4,903 subjects, 119 experienced ASCVD events; of the latter, only 34% were listed as “hard” coronary heart disease (CHD) (myocardial infarction + coronary death). In addition, only 40% of total CHD events were hard CHD events. Out of all subjects, 686 had CAC scores of 110 to 399 Agatston units; these scores signify at least moderate coronary atherosclerosis. In these subjects, event rates for total CHD events were only 1.3% per year, and for hard CHD events, only 0.58% per year. Another 450 subjects had CAC scores >400, which should indicate advanced coronary atherosclerosis; in these patients, event rates were 3.26% and 1.1% per year for total CHD and hard CHD, respectively. At the beginning of the study, subjects underwent Framingham risk scoring to determine 10-year risk for hard CHD events as described in the National Cholesterol Education Program’s Adult Treatment Panel III (NCEP ATP III) report (3). Among all subjects, 654 had a projected 10-year risk for hard CHD of <10% (or <1% per year); 506 scored a 10-year risk of 10% to 20% (1% to 2% per year); and 86 projected a >20% 10-year risk (>2% per year). In the intermediate-risk group (i.e., those projected to have 1% to 2% of hard events per year) hard CHD events were in fact only about 0.4% per year. Likewise those predicted to have a >2% per year risk for hard CHD showed an actual rate of only 0.9% per year. In both intermediate- and high-risk groups, only those with the highest CAC scores (upper tertile for projected risk ranges) had actual rates of hard CHD approaching the projected level of risk. Thus, CAC measurement clearly added information to the Framingham projection; furthermore, only in those in the highest tertile of CAC scoring for a given Framingham category did Framingham scoring accord with the observed rate.

In the second study (2), 1,005 healthy subjects with CAC scores above the 80th percentile for age and gender were randomized to the combination of atorvastatin, vitamin C, and vitamin E versus placebo. After 4.3 years, the therapeutic arm showed a non-significant trend toward a reduction of ASCVD events; moreover, progression of CAC was unaltered by drug therapy. The lack of a significant reduction of ASCVD events with therapy that included atorvastatin likely was due in part to the small size of the trial. But also important was the low baseline rate of CHD events in a group that otherwise was projected to be at a relatively high risk. For example, in the placebo group, rates for total CHD events and hard CHD events were only 2.03% and 0.77% per year, respectively.

The findings of this trial suggest that absolute rates of hard CHD events are only about half of those projected from Framingham risk scoring (3). In other recent clinical trials (4- 5), hard CHD events were lower than might have been expected. Until recently, the primary end point of most cholesterol-lowering trials has been MI + coronary death (hard CHD) (6- 8). Unexpectedly low rates of hard CHD events even in so-called high-risk groups raise the question of whether hard CHD is still a meaningful end point for clinical trials. Low rates of ASCVD deaths especially make total mortality a problematic end point. The findings of lower rates of hard CHD in clinical trials than those predicted by older prospective studies may have several causes: healthy volunteer effect, prevention of acute MI through angioplasty or thrombolysis in patients with acute coronary syndromes, institution of preventive therapies in the early stages of coronary disease, improved treatment of established ASCVD to prevent death, and widespread use of preventive measures in the general population (e.g., less smoking, more aspirin use, and improved blood pressure control).

The question therefore must be asked whether projections of total ASCVD events are now preferred over hard CHD as an end point in prevention trials as well as in risk prediction in clinical practice. There is a growing recognition that non-fatal ASCVD events are costly in many ways—financial, social, and personal. In the case of risk prediction, it may be useful to retain the current categories of risk but to change the end points that define these categories to be more inclusive of all ASCVD events (Table 1). Because any ASCVD event will be costly, it is likely that cost-effectiveness of preventive therapies will not be substantially eroded by focusing on all ASCVD events instead of hard CHD events. A disadvantage for clinical trial outcomes of course is that practice standards may vary and thus total ASCVD events will be “softer” than for hard CHD. Nonetheless, the changing presentation of ASCVD events requires a more inclusive outcome to define end points. It seems particularly important to include stroke among outcomes because of recent evidence that preventive measures affect stroke outcomes similarly to coronary outcomes.

Finally, the question can be raised about whether CAC measurement has a role in risk assessment of patients without established ASCVD or diabetes. Framingham risk scoring can still be used as the first step to screen persons for potential risk. As shown in one of the studies (1), in persons found to have a risk of <10% per 10 years, whether hard CHD or total ASCVD is used as the end point, absolute 10-year risk is too low to warrant CAC measurement. For those who have a 20% risk per 10 years, CAC measurement likewise adds little predictive information of value. On the other hand, in the Framingham risk range of 10% to 20% per 10 year, CAC measurements carry the greatest information. Persons who have a CAC score of >100 Agatston units (approximately the upper tertile for this category) can be elevated to a high-risk level. If this approach is taken, the negative predictive power of lower CAC scores must also be used in this intermediate-risk category; thus, a low CAC score should move a person to the lower-risk zone because of the paucity of events in those with lower scores. Presumably there is some CAC range (e.g., 50 to 100 Agatston units) that should keep a person in the intermediate-risk range; unfortunately, the study by Arad et al. (1) did not have the power to definitively identify this middle group. A larger study in the projected intermediate-risk range (10-year risk of 10% to 20%) is needed to more accurately re-assign individuals to risk category using CAC measurement. Moreover, the potential for other risk makers, such as CRP, to adjust risk in this category deserves additional study as well. The treatment arm of this study suggests that CAC measurement is of little utility in monitoring progression of atherosclerosis.

In summary, the studies by Arad et al. (1- 2) provide useful information in two areas. First, they strongly suggest that for a variety of reasons, coronary atherosclerosis is changing in its clinical manifestations. There appear to be fewer non-fatal MIs and coronary deaths than in the past, at least according to projections from Framingham scoring (3). It is likely that many persons with early coronary disease or those admitted to the hospital for acute coronary syndromes are not progressing to myocardial infarction because of earlier and more intensive intervention; in addition, improved management of established CHD results in fewer deaths. Other forms of ASCVD now predominate and probably represent the preferred end point for clinical trials. For this reason, it is suggested that clinical trialists in the cardiovascular field develop a common ASCVD end point through consensus; if this were to be done, it would be possible to more reliably compare the results of different clinical trials. Furthermore, CAC measurements appear to have clinical utility only in persons with intermediate risk in whom testing can more accurately triage individuals into lower-, intermediate-, and higher-risk categories for ASCVD events so that appropriate preventive strategies can be instituted.