Detection of coronary artery calcium by ultrafast computed tomography and its relation to clinical evidence of coronary artery disease.

Ultrafast computed tomography (CT) was used to evaluate the relation of coronary artery calcium, a marker of atherosclerosis, with a reported history of coronary artery disease (CAD) in 928 men and 290 women (mean age 53 +/- 10 years; 11% with previous CAD). Total calcium score was calculated as the sum of each lesion-specific score, calculated as the product of pixel area and density > 130 Hounsfeld units. Total score was 3 to 6 times greater (p < 0.01) and the probability of coronary artery calcium 30 to 40% greater (p < 0.01) in patients with a reported history of myocardial infarction, positive angiography, bypass surgery or angioplasty. From score cutoffs ranging from 1 to 500 for defining calcium, a negative test was accurate 93 to 98% of the time in ruling out CAD, whereas specificity increased from 43 to 93%; however, sensitivity decreased from 92 to 42%. A score cutoff of 50 showed modest sensitivity (78%) and specificity (71%); however, the predictive value for CAD from a positive test remained low (< or = 40%), regardless of score cutoff. From multiple logistic regression, total score was also an independent indicator of CAD after considering any effects due to age, sex and other CAD risk factors. Further study is needed to document the long-term prognostic use of coronary calcium screening, including criteria that best project future risk of CAD.

ultrafast computed te (CT) was used to evaluate the relation of coronmy artery calcium, a marker of atimosdrosls, wlth a reported history of wronary artery disease (CAD) in 928 men and 290 women (mean age 53 + 10 years; ll% with previous CAD). Total cakium score was cakulated as the sum of each lesiorrspeeifsc score, cakulat-edastheproductofpixelareaanddensityrl30 Hounsfeld units. Total score was 3 to 0 times gNate# (p <O.Ol) and the probabilii of coronary arterycalcium3Oto4O%greatef (p4.01)inpa tientswitharepoddhistoryofmyocardialinfa~ tion,poaitivea~~phy,bypasssu~~orarr gioplasty.Fromscomwtoffsrangingfrom1t0 SO0 for defining calcium, a negative test was accurate 93 to 98% of the time in ruling out CAD, whereasspecificityincreased Wnll43to93%; however, sensltlvlty dmma!mdflvms2to42%. Ascorecutoffof5Oshowedmodestsensitivity (78%)mdspecificity(71%);however,the predie tive value for CAD from a positive test remained low @IO%), regardless of score cutoff. From muC tiple logistic m, total score was also an independent indicator of CAD after considering any effects due to age, sex and other CAD risk facbrs.Gtherstudyisneededtodocumentthe long-term progno&ic use of coronas calcium screening, including criteria that best project future risk of CAD.
(Am J Cardiol l-73223-227) A close correspondence between coronary artery calcium deposits and atherosclerotic plaque has been shown by pathologic studies.1-3 Recently, ultrafast computed tomography (CT) was used to determine the arterial site and relative burden of coronary artery calcium, as a marker of atherosclerosis.4,5 Studies in subjects who had both coronary angiography and ultrafast CT have shown a direct relation between the extent of coronary artery calcium present and the number or severity of diseased vessels on angiography.4-6 Those studies have documented a high sensitivity and modest predictive value of ultrafast CT coronary artery scanning in detecting significant coronary artery disease (CAD). The present report evaluates the sensitivity and predictive value of ultrafast CT coronary artery scanning in relation to reported history of CAD, including the impact of age and preexisting risk factors on the ability of this test to identify CAD. METHODS Study population and interview: Data obtained in 1,218 men and women, primarily self-referred or referred by their physician for ultrafast CT screening between May 1991 and December 1992 at a private radiology clinic, are included in this report. Reasons for referral included family history of premature CAD, known cardiac risk factors or personal history of documented CAD. The study evaluated data obtained as part of a baseline questionnaire regarding the medical and risk factor history of patients, specifically including history of hypertension, diabetes, hypercholesterolemia, tobacco use, premature family history of myocardial infarction (age 1.55 years in parent or sibling), chest pain suggesting angina pectoris (by Rose questionnaire), reported previous myocardial infarction, positive angiography, angioplasty and bypass surgery. Use of medication to control hypercholesterolemia, hypertension or diabetes was also included in the category of a positive history of these conditions, and the number of cigarettes consumed or time since quitting was recorded also.
-forU-conputed~ . All studies were performed by a standard protocol previously described, using an Imatron C-100 Ultrafast CT scanner (South San Francisco).4 Brielly, the protocol involved subjects being positioned supine, head tirst, into the scanning aperture with no couch angulation, and instructed to take 3 deep breaths and hold their breath (at end-expiration) while a preview scan was obtained. Patient positioning was checked, and if necessary, adjusted, so that scanning began from near the lower evaluated sequentially to determine the presence and margin of the bifurcation of the main pulmonary artery. quantity of coronary artery calcium. From the es-Coronary visualization was achieved without contrast by tablished protocol,4 the threshold for a calcified lesion using the high-resolution volume mode of the ultrafast was set at a CT density of 130 Hounsfeld units in 21 CT scanner in conjunction with a 100 ms scan time, 3 pixel (area 20.51 mm2). At each level, all pixels with a mm slice thickness, electrocardiographic triggering (to CT density Y30 Hounsfeld units were displayed. A re-80% of RR interval) and breath holding for approxi-gion of interest was manually encircled around each vismately 45 seconds. Twenty to 30 contiguous slices were ible lesion within each coronary artery and computerobtained with no interslice gaps. Each of the levels acquired measurements of lesion area in mm2 and max-(which encompassed the proximal portions of the coroimal Hounsfield number of each region of interest were nary arteries where nearly all calcium is present) was recorded. A density score was determined based on maximal Hounsfield number, as follows: 1 = 130 to 199; 2 = 200 to 299, 3 = 300 to 399, and 4 = >400 Hounsfield units. A score for each region of interest was calculated by multiplying density score by area. A total calcium score was determined by adding each score for all lesions across all slices. Statistical analysis: The prevalence of coronary artery calcium (defined as the prevalence of subjects with a total calcium score >O, 50 or 250), and the mean total calcium score were calculated and compared (by the chisquare test of proportions to compare prevalence, or Student's t test comparing the mean natural logarithm of [total score plus I] to normalized data) between men and women with and without history of myocardial infarction, bypass surgery, angioplasty and positive angiography. Total CAD was defined as a reported history of ti of the following: myocardial infarction, bypass surgery, angioplasty and positive. angiography. Reported chest pain was not included in the aforementioned definition, because of its nonspecificity as an indicator of CAD. Analyses were performed also by age group (~50, 50 to 59, and 260 years) and sex. Sensitivity, specificity, and positive and negative predictive values of ultrafast CT-detected coronary artery calcium (using cutoff points of total score of 1, 50, 100, 250 and 500) in relation to CAD were plotted for the overall sample, and analyzed by age/gender group for a cutoff score of 500 (arbitrarily chosen on the basis of maximized overall sensitivity and specificity). Finally, the independent value of the natural log of [calcium score plus l] in predicting prevalent CAD was examined by multiple logistic regression after allowing for adjustment of known reported risk factors of CAD.
Overall, 6.1% (n = 56) of men and 5.5% (n = 16) of women had a reported history of myocardial infarction, and including bypass surgery, angioplasty, and positive angiography, CAD was reported in 11% (n = 102) and 9% (n = 26), respectively. Table I shows comparisons by prevalence using different score cutoffs and mean calcium scores, comparing patients with and without each coronary end point. The prevalence of coronary artery calcium and mean total calcium scores were significantly greater in patients with than in those without each coronary end point (p ~0.01). Although 90.3% of patients with history of myocardial infarction had coronary artery calcium defined by a score >O, 58% of those without this history were positive for coronary artery calcium. In patients (n = 128) with any manifestation of CAD, 92.2% were positive for coronary artery calcium by a score criterion >O and 57% by a score >250, which are substantially greater than in those without previous CAD (56.5 and From the negative predictive value plotted by calcium score cutoff (Figure l), regardless of cutoff, a negative calcium scan was found to be associated with the absence of CAD 93 to 98% of the time. Sensitivity, however, decreased from 92% at a score cutoff of 1 to below 70% after score cutoffs exceeding 100 for defining a positive test. Specificity (absence of calcium in the absence of CAD) increased to nearly 90% using a score cutoff of 250. A score cutoff of 50 appeared to yield modest levels of both sensitivity and specificity in the study population as a whole. The predictive value of a positive test, however, remained low (<40%), regardless of score cutoff used. Table II compares the prevalence and mean calcium scores of patients with and without each CAD end point by age/gender group. In men aged <50 years and women aged ~60, good discrimination in coronary artery calcium prevalence is shown in patients with compared with those without each end point; however, after age 60, calcium prevalence is high regardless of history of event. Mean coronary artery calcium score, however, remains much higher in patients with than in those without each end point, irrespective of age.
When comparisons of sensitivity, specificity, and positive and negative predictive values were evaluated by age and gender group (with a total score cutoff of 500 arbitrarily chosen based on Figure l), both specificity and the predictive value of a negative test to conlirm absence of CAD were best in younger subjects. Sensitivity \?ras best in older subjects (aged >60 years) who had adequate data for analysis.
Multiple logistic regression showed that the natural log of [calcium score plus l] remained an important indicator (p <O.Ol), discriminating between patients with and without CAD (Table III). Only score and age were associated independently with CAD status, with no other risk factors entering in the model by stepwise selection. DISCUSSION This report confirms a prevalence of coronary artery calcium that is generally in the 90 to 95% range for subjects with a reported history of CAD, which is substantially higher than in those without reported CAD. Mean calcium score is also considerably higher in patients with a positive versus negative history of CAD. The data also suggest that if total calcium score is used as a criterion for defining CAD, a score cutoff of 50 units appears to provide modest (>70%) specificity and sensitivity for discriminating CAD end points, although this criterion may be too strict for women or younger subjects who have a lower prevalence of coronary calcium. The use of age-and sex-matched isopercentile score cutoffs to risk stratify patients was examined recently by other investigators.' However, longitudinal investigation in relation to future clincial events in a larger population will be required to co&m the use of such an approach.
Although the predictive value of a negative test for excluding clinical CAD remains high regardless of the score cutoff used, the predictive value of a positive coronary artery scan for documenting a positive history of CAD remains low (~50%). This is not surprising, because coronary calcium or atherosclerosis does not necessitate clinically significant CAD, and many subjects with low or moderate levels of coronary calcium may never develop symptomatic CAD. However, the large number of asymptomatic middle-aged subjects in this population who have not developed clinical CAD may underestimate the positive predictive value and overestimate the negative predictive value that would eventually be realized.
The findings from this large study population corroborate those from a previous report of a decreasing sensitivity and increasing specificity with respect to calcium score.4 Furthermore, in angiographic subsets, a perfect or nearly perfect sensitivity documenting positive calcium in subjects with significant angiographic obstruction is observed, as well as a high negative predictive value to rule out significant CAD.4q6 However, if the presence of any calcium is used to deline a positive test, it has been shown that almost half of patients negative for obstructive CAD will remain positive for coronary artery calcium (low specificity).6 In these cases, CAD sufficiently significant to result in infarction cannot be ruled out, because these events frequently occur in the presence of mild angiographic CAD.
A possible explanation for lower levels of sensitivity in women and younger subjects, who in general have less calcium, may relate to partial volume effects on identifying calcified pixels, or patient movement that may result in missing smaller lesions, especially if only 1 scan is performed, as in this study. Furthermore, the tinding of an increased sensitivity, but reduced positive predictive value with age for higher amounts of calcium suggests that older, highly calcified vessels may not necessarily reflect the severity of stenosis or threat of acute events, but may largely be a reflection of increased calcium prevalence with age. The lower sensitivity in women suggests that calcium could be less important in the pathogenesis of CAD in women, although overreporting of myocardial infarction by women could be an explanation also. However, underreporting of CAD status would tend to result in underestimated specificity.
A role of coronary artery calcium as a marker for coronary risk is also implicated from studies showing its association with coronary risk factor&to and from earlier fluoroscopic studies showing that coronary artery calcium predicts subsequent prognosis,"J2 although not all reports show an association.t3 The present data show important differences between subgroups of patients with and without a reported his-226 THE AMERICAN JOURNAL OF CARDIOLOGY VOLUME 73 FEBRUARY 1, 1994 tory of CAD in the distribution and quantity of coronary artery calcium. Further refinement of scoring criteria, and improved scanning methodology to minimize technical error are needed to best define the role of coronary artery calcium screening in the diagnosis of CAD. Furthermore, investigations involving longitudinal follow-up of asymptomatic subjects with and without coronary artery calcium, and evaluation of threshold levels and patterning of coronary calcium in relation to clinical events are warranted.