Predictive Value of Electron Beam Computed Tomography of the Coronary Arteries
19-Month Follow-up of 1173 Asymptomatic Subjects
Background Coronary electron beam computed tomography (EBCT) detects atherosclerotic coronary artery disease by measuring calcium deposition in the walls of coronary arteries. EBCT-derived coronary artery calcium (CAC) scores correlate with the severity of underlying coronary artery disease.
Methods and Results We followed 1173 asymptomatic patients who underwent EBCT between September 1993 and March 1994. During average follow-up of 19 months, 18 subjects had 26 cardiovascular events: 1 death, 7 myocardial infarctions, 8 coronary artery bypass graft procedures, 9 coronary angioplasties, and 1 nonhemorrhagic stroke. For CAC score thresholds of 100, 160, and 680, EBCT had sensitivities of 89%, 89%, and 50% and specificities of 77%, 82%, and 95%, respectively. Odds ratios ranged from 20.0 to 35.4 (P<.00001 for all).
Conclusions Coronary EBCT predicts future atherosclerotic cardiovascular disease events in asymptomatic subjects.
Atherosclerotic CVD is the leading cause of death in developed countries. In up to 50% of patients, atherosclerotic CVD presents initially as either an acute MI or sudden death.1 Approaches that use cholesterol,2 assessment of nonlipid risk factors, or stress testing3 lack precision for the detection of preclinical CAD in asymptomatic subjects.
CAC content correlated closely with the severity of underlying CAD at autopsy.4 5 6 7 Coronary EBCT can accurately measure CAC content,8 9 and EBCT-derived CAC scores correlate with angiographically documented CAD,10 11 12 13 suggesting that EBCT might allow for more accurate screening for CAD.
To assess the potential predictive value of coronary EBCT-derived CAC content for future CAD events, we followed 1173 asymptomatic men and women for a mean duration of 19 months.
All 1238 EBCT patients scanned between September 1, 1993, and March 11, 1994, were asked to complete initial and follow-up questionnaires, including a medical history, occurrence of cardiovascular events, and the modified Rose Angina Questionnaire.14 Results were analyzed for 1173 of 1175 consenting, eligible subjects (99.8%) without documented or symptomatic atherosclerotic CVD. Mean follow-up was 19 months (range, 14 to 23 months). A positive response to the modified Rose Angina Questionnaire required the presence of exertional chest discomfort relieved by rest in ≤10 minutes. Events were verified by telephone inquiries and reviews of medical records. Patients were self-referred or referred by physicians in response to information in newspapers, local and national networks, and active advertisements. In all cases, coronary EBCT was described as potentially useful for screening purposes only.
The diagnosis of MI required two of the following three criteria: ischemic myocardial pain lasting at least 30 minutes, CK elevation to more than twice the upper limit of normal with CK-MB ≥5%, and development of new Q waves ≥40 ms on the ECG. The diagnosis of a stroke required the development of a new, persistent neurological deficit confirmed with head CT.
Electron Beam CT
EBCT was performed with a Siemens Evolution scanner as previously reported.15 Forty contiguous slices 3 mm thick were obtained during a single breath-hold, beginning at the lower edge of the carina. Scan time was 100 ms per slice, with synchronized ECG triggering at 80% of the RR interval. CAC scores were calculated according to Agatston et al.15
The relationship between CAC scores and events was analyzed with Student’s t test. An ROC was generated to determine the predictive power of EBCT-derived CAC scores for hard cardiovascular events over the average 19-month follow-up period. Increased area under the ROC curve indicates increased predictive value of a diagnostic test. We used contingency tables and the χ2 function to examine the relationship between risk factors and events. Only one event was counted per patient.
Mean age (±SD) was 53±11 years. Seventy-one percent of the subjects were male. During the follow-up period, there were 1 cardiac death, 7 nonfatal MIs, 1 spontaneous thromboembolic stroke, 8 coronary artery bypass procedures, and 9 coronary angioplasties in 18 patients, for an event rate of 1.53% (counting only 1 event per patient). Indications for revascularization included MI in 7 patients. Revascularization procedures not preceded by an MI were performed because of new unstable angina (2 subjects), positive early stress tests (3 subjects), extensive reversible defects on thallium scan (3 subjects), or severe dyspnea on stress testing out of proportion to effort and age (1 subject). No subject went directly from EBCT to coronary angiography, and all stress tests were performed at the direction of the patient’s primary physician after the development of symptoms consistent with myocardial ischemia (chest pain or exertional dyspnea). Forty-four patients developed angina pectoris.
CAC scores were 935±1070 for patients with events versus 144±446 for patients without events (P<.0001). The distributions of CAC scores for both groups are shown in Fig 1⇓. CAC scores were 363±746 for all patients who developed angina versus 147±455 for those who did not (P=.063). Men with angina had CAC scores of 555±843 versus men without angina, 165±492 (P=.053). Women with angina had CAC scores of 203±627 versus women without angina, 98.9±337 (P=.427).
For CAC scores >100 (previously shown to correspond to a worst stenosis in any major coronary arterial segment of 20%, see Reference 1313 ), sensitivity, specificity, and odds ratio for predicting future hard cardiovascular events in this population were 89%, 77%, and 25.8, respectively (CI, 5.9 to 113). For CAC scores >160 (the value that maximized the sum of sensitivity and specificity), sensitivity, specificity, and odds ratio for hard cardiovascular events were 89%, 82%, and 35.4, respectively (CI, 8.1 to 155). For CAC scores >680 (previously shown to correspond to a worst stenosis of 50%, Reference 1313 ), sensitivity, specificity, and odds ratio were 50%, 95%, and 20.0, respectively (CI, 7.6 to 52) (Table⇓). Negative predictive values were all >99%, and the positive predictive value was 14% for patients with CAC scores >680. The area under the ROC was 0.91 (Fig 2⇓) and was unaltered after correction for the effect of age on CAC scores (by ANCOVA).
Of the reported incidence of coronary risk factors (age, high cholesterol, low HDL cholesterol, hypertension, smoking, diabetes, and family history of premature atherosclerotic disease), only hypertension correlated with CAC scores (χ2=14.3, P=.0012) at baseline, and only age showed a positive correlation with subsequent atherosclerotic cardiovascular events (χ2=6.2, P=.035).
Autopsy7 8 9 and angiographic10 11 12 13 data consistently indicate a correlation (r=.75 to .88) between CAC content and the severity of CAD. EBCT can quantify coronary calcifications, thus identifying subjects with preclinical CAD.
This is the first large, written report with a high degree of completeness of follow-up (99.8%) documenting the prospective short-term predictive value of EBCT of the coronary arteries in asymptomatic patients. The strong positive correlation of CAC scores with clinical events is consistent with a study that used cinefluoroscopy16 and with three preliminary follow-up reports of subjects who underwent EBCT.17 18 19
We do not believe that subjects with high CAC scores underwent bypass surgery or angioplasty preferentially, because EBCT has not been accepted by the local medical community, no subject went directly from EBCT to coronary angiography, all stress tests were done for clinical indications, and all revascularization procedures not preceded by an MI were performed for clinical indications.
Compared with an area under the ROC of 0.74 for NCEP II guidelines found in a recent 12-year analysis of the Lipid Research Clinic Prevention Follow-up Studies,20 the area under the ROC of 0.91 (Fig 2⇑) suggests that EBCT is a better predictor of CVD events in a much shorter time period. This is supported by the high negative predictive values and a positive predictive value of 14% for CVD events, in only 19 months, in initially asymptomatic subjects with CAC scores >680.
Finally, the lack of correlation between most traditional CAD risk factors and baseline CAC scores or subsequent events must be interpreted with caution because of the use of reported rather than measured risk factors. However, measured values are subject, in turn, to the problem of prior behavior modification and medical intervention. In this mostly self-referred, middle-class population, most middle-aged adults are likely to be aware of the presence or absence of risk factors.
We conclude that EBCT-based screening for CAD shows great applicability to the development of CVD events in a relatively short time period (average, ≈1.5 years) in an initially asymptomatic middle-aged population (mostly men). These data support the routine use of coronary EBCT to screen for occult CAD in asymptomatic middle-aged men and women.
Selected Abbreviations and Acronyms
|CAC||=||coronary artery calcium|
|CAD||=||coronary artery disease|
|EBCT||=||electron beam CT|
|ROC||=||receiver operating characteristic curve|
We thank Sean Callahan, Lania Bianco, Joan Scordo, Denise Brennan, Laura Cook, Stephen Boyd, Lawrence Balletti, Frederick Bernstein, Michael Kortbus, Lorraine Lindsay, Puja Mahindra, David Sacrestano, Usman Saleem, and Sameer Sayeed for their tireless efforts in collecting the follow-up data and Bill Schoepfer for his technical expertise.
- Received February 14, 1996.
- Revision received March 11, 1996.
- Accepted March 13, 1996.
- Copyright © 1996 by American Heart Association
American Heart Association. 1989 Heart Facts. Dallas, Tex: American Heart Association; 1988.
Wald NJ. Cholesterol and coronary heart disease: to screen or not to screen. In: Marmot M, Elliott P, eds. Coronary Heart Disease Epidemiology. Oxford, UK: Oxford University Press; 1992.
Blankenhorn DH, Stern D. Calcification of the coronary arteries. Am J Roentgenol. 1959;81:772-777.
Beadenkopf WG, Daoud AS, Love BM. Calcification of the coronary arteries and its relation to arteriosclerosis and myocardial infarction. Am J Roentgenol. 1964;92:865-871.
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.
Rumberger JA, Simons DB, Fitzpatrick LA, Sheedy PF, Schwartz RS. Coronary calcium area by electron beam CT and coronary atherosclerosis plaque area. Circulation. 1995;92:2157-2162.
Budoff M, Georgiou D, Brody A, Agatston AS, Kennedy J, Wolfkiel C, Stanford W, Shields P, Lewis RJ, Janowitz WR, Rich S, Brundage BH. Ultrafast computed tomography as a diagnostic modality in the detection of coronary artery disease: a multicenter study. Circulation. 1996;93:898-904.
Kajinami K, Takekoshi N. Noninvasive prediction of coronary atherosclerosis by quantification of coronary artery calcification using electron beam computed tomography: comparison with electrocardiographic and thallium exercise stress test results. J Am Coll Cardiol. 1995;26:1209-1221.
Guerci AD, Spadaro LA, Popma JJ, Goodman KJ, Brundage BH, Lerner G, Vizza RF. Electron beam tomography of the coronary arteries: relationship of coronary calcium score to arteriographic findings in asymptomatic adults. Am J Card Imaging. 1995;9:5. Abstract.
Wong N, Vu A, Abrahamson D, Eisenberg H, Tobis JM. Prediction of coronary events from noninvasive calcium screening by ultrafast CT. Circulation. 1993;88(suppl I):I-15. Abstract.
Puentes G, Detrano R, Tang W, Wong N, French W, Narahara K, Brundage B, Baksheshi H. Estimation of coronary calcium mass using electron beam computed tomography: a promising approach for predicting coronary events? Circulation. 1995;92(suppl I):I-313. Abstract.
Balogh T, Hoff J, Rich S, Wolfkiel J. Development of coronary artery disease in asymptomatic subjects undergoing coronary artery calcification screening by electron beam tomograph. Circulation. 1995;92(suppl I):I-650. Abstract.