To the Editor:
Secci and colleagues are to be commended for the integrity required to question the accuracy of electron-beam CT (EBCT) scanning as a screening test for coronary disease,1 particularly when at least one member of the group had anticipated that the method would prove successful. However, their conclusion that the amount of coronary calcification fails to predict nonfatal myocardial infarction (MI) and coronary death is at odds with their own and other data and is probably the result of a critical flaw in the design of the South Bay Heart Watch.
In the larger cohort of 1461 South Bay Heart Watch subjects, the risk of nonfatal myocardial infarction and coronary death has been proportional to the number of calcified coronary arteries as determined by fluoroscopy.2 The fact that there is a threshold for the detection of coronary calcification by any method (and, in the case of a method as insensitive as fluoroscopy, a rather high threshold3 ) implies first, that the greater the number of calcified vessels, the higher the coronary calcium score and second, that the sample of 326 high-risk subjects was too small to confirm what was already demonstrated in the larger study.
Two other studies4 5 of EBCT scanning in less-selected populations have reported unprecedented accuracy of the EBCT-derived coronary calcium score in the prediction of nonfatal MI and coronary death. When pooled with data from the South Bay Heart Watch, these studies yield odds ratios of 4.7 to 9.5 for nonfatal MI and coronary death for subjects with calcium scores in the upper third versus the lower two thirds of the population (P<0.001). These findings are consistent with autopsy evidence of calcification of the coronary arteries in victims of sudden cardiac death and fatal MI,6 as well as with results of a large study7 that reported 2 to 5 times as much calcium in the coronary arteries of victims of coronary heart disease (n=214) as persons dying of other diseases (n=1028).
Although Secci et al have reported MIs in several persons with coronary calcium scores of zero, this is undoubtedly a rare event. We have >1500 patient-years of documented follow-up in middle-aged men and women with coronary calcium scores of zero, with no coronary deaths and just 1 MI. That patient, a 58-year-old woman, had angiographically normal coronary arteries. Our favorable experience in asymptomatic persons with low coronary calcium scores is also consistent with the autopsy literature and correlations between EBCT and coronary angiographic findings. The majority of middle-aged victims of fatal MI have extensive coronary atherosclerosis, and although the culprit lesion is not necessarily calcified, calcified plaques are generally present elsewhere in the coronary arteries.8
Secci et al have concluded that a strategy of screening for coronary disease based on EBCT is flawed because “noncalcific plaques may be more likely to rupture than calcific plaques.” Available evidence certainly supports the latter statement but, as cited above, also indicates just as clearly that coronary calcification is a marker for the presence of noncalcified plaque. A more likely explanation for the failure of the EBCT-derived coronary calcium score to predict nonfatal MI and coronary death in the South Bay Heart Watch is the nature of the population studied. In order to qualify, potential study participants were required to have a ≥10% risk of coronary heart disease events over the ensuing 8 years, which placed them at or above the 75th percentile in the population distribution.2 Bayes’ theorem predicts that the likelihood that a negative test result is a true negative is inversely proportional to the pretest probability of disease. Because the relationship between coronary calcification and the severity of underlying coronary atherosclerosis is imperfect, Bayes’ theorem applies to EBCT scanning. Thus, with respect to the phenomenon of coronary calcification, a high-risk population like that of the South Bay Heart Watch contains an unusually high prevalence of false-negative examinations. Similar results have been obtained in a cohort of symptomatic patients undergoing coronary arteriography, in which obstructive coronary disease was present in the majority of patients with angina pectoris and 4 or 5 risk factors for coronary artery disease regardless of the coronary calcium score.9
Stated differently, the design of the South Bay Heart Watch makes it difficult if not impossible to prove that the coronary calcium score predicts coronary events. Imagine trying to determine the relationship between cigarette smoking and coronary events by confining a study to smokers, or between serum total cholesterol and coronary events by restricting the study to persons with serum total cholesterol >5.9 mmol/L (roughly 230 mg/dL). One might observe a significant relationship, but it would be much weaker than that observed in a nonselected population. Indeed, the high-risk profile of subjects in the South Bay Heart Watch probably explains the failure of diabetes, smoking, hypertension, family history, HDL cholesterol, or LDL cholesterol to predict coronary events (Table 9 in the article by Secci et al). Only left ventricular hypertrophy, which is evidence of end-organ damage, predicted nonfatal MI and coronary death. At the same time, left ventricular hypertrophy failed to predict need for coronary revascularization, raising the question of whether some of these deaths were the result of a primary (ie, hypertrophic) or secondary (hypertensive) cardiomyopathy rather than coronary atherosclerosis.
Although the balance of evidence demonstrates that EBCT predicts coronary death and nonfatal MI as well as coronary revascularization, there are still many questions about the pathogenesis and predictive accuracy of coronary calcification. Unfortunately, it appears that the South Bay Heart Watch will provide few answers.
- Copyright © 1998 by American Heart Association
Secci A, Wong N, Tang W, Wang S, Doherty T, Detrano R. Electron beam computed tomographic coronary calcium as a predictor of coronary events: comparison of two protocols. Circulation. 1997;96:1122–1129.
Arad Y, Spadaro LA, Goodman K, Lledo-Perez A, Sherman S, Lerner G, Guerci AD. Predictive value of electron beam computed tomography of the coronary arteries: 19-month follow-up of 1173 asymptomatic subjects. Circulation. 1996;93:1951–1953.
Agatston AS, Janowitz WR, Kaplan GS, Lee D, Prashad R, Viamonte M Jr, Lamas GA. Electron beam CT coronary calcium predicts future coronary events. Circulation. 1996;94(suppl I):I-360. Abstract.
Kragel AH, Reddy SG, Wittes JT, Roberts WC. Morphometric analysis of the composition of atherosclerotic plaques in the four major epicardial coronary arteries in acute myocardial infarction and in sudden coronary death. Circulation. 1989;80:1747–1756.
Eggen DA, Strong JP, McGill HC. Coronary calcification: relationship to clinically significant coronary lesions and race, sex, and topographic distribution. Circulation. 1965;32:948–955.
Arad Y, Spadaro LA, Goodman K, Lledo-Perez A, Sherman S, Guerci AD. 3.6 years follow-up of 1136 asymptomatic adults undergoing electron beam CT (EBCT) of the coronary arteries. J Am Coll Cardiol. 1998;31:210A. Abstract.
Guerci AD, Spadaro LA, Goodman KJ, Lledo-Perez A, Lerner G, Arad Y. Comparison of electron beam tomography and conventional risk factors in the prediction of coronary artery disease. J Am Coll Cardiol. In press.
Drs Guerci, Arad, and Agatston state (1) that the results of our preliminary report on the accuracy of EBCT in the South Bay Heart Watch are at odds with the results of our other reports regarding the accuracy of coronary calcium in predicting coronary events, (2) that our results are also at odds with the “unprecedented accuracy” of EBCT for predicting events reported by themselves, and (3) that differences between their results and our own are due to high pretest probability in the South Bay Heart Watch cohort.
1. Our 5-year follow-up of 1461 asymptomatic South Bay Heart Watch participants who had undergone digital fluoroscopyR1 and our report of follow-up of 1196 of these subjects who underwent EBCT scanningR2 both clearly demonstrate that there is a weak relationship between the presence and quantity of coronary calcium and subsequent coronary death and infarction. The preliminary report to which Guerci, Arad, and Agatston refer treats a subset of 326 of these subjects who had undergone both 3-mm and the more reproducible 6-mm scanning to determine if 1 of these scanning protocols was more accurate in predicting events. Since the primary purpose of this preliminary analysis was to compare protocols, it was not powered to detect weak associations between coronary calcium and future events. Therefore, there is no discrepancy between the conclusions of these 3 reports. Coronary calcium quantitated with EBCT predicts events about as well as standard risk factors.
2. There are indeed large differences between the “unprecedented accuracy” of the reports from Drs Guerci, Arad, and Agatston and the weak association found between calcium and coronary events in the South Bay Heart Watch participants. These may be due to methodological differences in cohort recruitment and analysis of scan results. The South Bay Heart Watch cohort was drawn from a population-based sample living in the Los Angeles, Calif, area. The Heart Watch had funding from government and private foundations and thus did not request payment from the subject volunteers, all of whom signed consent forms and understood that they were participating in a study of an unproven technology. The subjects participating in the studies of Drs Guerci, Arad, and Agatston were largely paying customers who were recruited through advertising campaigns directed at convincing them that the scan was a useful way of protecting their future health. A South Bay Heart Watch committee of adjudicators blindly assessed medical records to determine outcomes, whereas the method of assessment in the report of Arad et alR3 is unclear. Thus, we agree with Drs Guerci, Arad, and Agatston that differences in study design might have affected results of either or both studies, though perhaps not in the way that they suggest.
3. Guerci, Arad, and Agatston attribute differences between their results and our own to a higher pretest probability of coronary events in our South Bay Heart Watch subjects. They criticize this aspect of the design of the South Bay Heart Watch study and believe that the weak relation between test results and coronary events is due to the relationship between pretest and posttest probabilities resulting from application of Bayes theorem.
Indeed, we chose our subjects to be at higher risk for 3 reasons: (1) higher risk would increase the number of coronary events and thus the power of the study; (2) higher-risk subjects would be subjects for whom risk factor reduction interventions would be appropriate and therefore for whom the results of an accurate screening test would be useful in making clinical decisions (eg, to treat with lipid-lowering medications); and (3) higher pretest probability (closer to 50% probability), according to Bayes theorem, increases rather than decreases information gained from diagnostic testsR4 R5 and would therefore increase, not decrease, the utility of coronary calcium screening. Indeed, Bayes theorem and most physicians would agree that coronary screening with any method applied to 25-year-old women without risk factors would not be useful.
Despite the fact that the South Bay Heart Watch cohort consisted of subjects who should be most appropriate for coronary calcium screening, this test was only as accurate for predicting outcomes as are the more easily verified and modifiable coronary risk factors. These research results should mandate an immediate moratorium on commercial coronary calcium screening of asymptomatic adults.
Detrano R, Wong ND, Doherty TM, Shavelle R. Prognostic significance of coronary calcific deposits in asymptomatic high risk subjects. Am J Med.. 1997;102:344–349.
Detrano R, Wong ND, Tang W, Doherty TM. Determining coronary event risk in asymptomatic high-risk subjects: a risk factor versus an anatomic approach. Circulation. 1997;96(suppl I):I-404. Abstract.
Arad Y, Spadaro LA, Goodman K, Lledo-Perez A, Sherman S, Lerner G, Guerci AD. Predictive value of electron beam computed tomography of the coronary arteries: 19-month follow-up of 1173 asymptomatic subjects. Circulation.. 1996;93:1951–1953.
Diamond GA, Hirsch M, Forrester J, Staniloff HM, Vas R, Halpern SW, Swan HJ. Application of information theory to clinical diagnostic testing: the electrocardiographic stress test. Circulation.. 1981;63:915–921.