- Email: [email protected]
Lesion- and vessel-specific coronary artery calcium scores are superior to whole-heart Agatston and volume scores in the diagnosis of obstructive coronary artery disease
Journal of Cardiovascular Computed Tomography (2010) 4, 391–399
Original Research Article
Lesion- and vessel-specific coronary artery calcium scores are superior to whole-heart Agatston and volume scores in the diagnosis of obstructive coronary artery disease Zhen Qian, PhD*, Hunt Anderson, MD, Idean Marvasty, BS, Kamran Akram, MD, Gustavo Vazquez, MD, Sarah Rinehart, MD, FACC, Szilard Voros, MD, FACC, FSCCT Piedmont Heart Institute, 95 Collier Road, Suite 2085, Atlanta, GA 30309, USA KEYWORDS: Computed tomography; Coronary artery calcium; Obstructive coronary artery disease
BACKGROUND: The whole-heart coronary artery calcium (CAC) score has poor predictive value for obstructive coronary artery disease (CAD). We hypothesized that vessel- and lesion-specific CAC scores are more accurate. OBJECTIVES: To evaluate the usefulness of vessel- and lesion-specific CAC in predicting obstructive CAD and to assess the incremental value added by the vessel- and lesion-specific CAC to the conventional whole-heart CAC approach. METHODS: Ninety-one patients with CAC scores and invasive angiography (XRA) data were enrolled. Besides whole-heart CAC, Agatston score (AgSc) and volume score (VolSc) were measured individually for each lesion in the 4 major epicardial coronary arteries. Maximum and average lesionspecific scores in each vessel were also determined. For the primary analysis, obstructive CAD was defined as 50% diameter stenosis by XRA. RESULTS: Whole-heart AgSc and VolSc were not different between patients with and without obstructive CAD (P 5 .23 and P 5 .18), whereas vessel- and lesion-specific scores were (maximum lesion specific AgSc, P , .0001). Maximum lesion-specific AgSc had superior diagnostic performance compared with whole-heart AgSc (area under receiver operating characteristics, 0.71 vs 0.58). Overall sensitivity, specificity, and diagnostic accuracy were improved. When specificity was fixed at 80%, sensitivity of maximum lesion-specific AgSc was superior to whole-heart AgSc (56.6% vs 35.1%). Most importantly, with lesion-specific AgSc, fewer patients were classified as ‘‘indeterminate’’ compared with whole-heart AgSc (17.9% vs 50%). CONCLUSIONS: Vessel- and lesion-specific CAC scores are superior to the whole-heart AgSc and VolSc in predicting obstructive CAD. This simple refinement in CAC scoring may significantly improve the clinical predictive role of CAC imaging. Ó 2010 Society of Cardiovascular Computed Tomography. All rights reserved.
Conflict of interest: The authors report no conflicts of interest. * Corresponding author. E-mail address: [email protected] Submitted May 11, 2010. Accepted for publication September 3, 2010.
Introduction Computed tomographic (CT)–based measurement of coronary artery calcium (CAC) has been introduced as a
1934-5925/$ - see front matter Ó 2010 Society of Cardiovascular Computed Tomography. All rights reserved. doi:10.1016/j.jcct.2010.09.001
392
Journal of Cardiovascular Computed Tomography, Vol 4, No 6, November/December 2010
noninvasive, low-radiation method to assess the overall coronary arterial atherosclerotic burden, by quantifying calcium in the coronary vasculature.1–3 The whole-heart volume score (VolSc)4 is a summation of all voxels along the 3 major epicardial coronary arteries with an attenuation value of R130 Hounsfield units (HU), whereas the wholeheart Agatston score (AgSc)5 also takes into account the overall density of each of the voxels. The AgSc was developed to estimate the overall burden of coronary atherosclerosis and to estimate the cardiovascular risk in asymptomatic subjects without known coronary artery disease (CAD). In this regard, the AgSc has been shown to be independent of and additive to the Framingham risk score6 in predicting cardiovascular outcomes in asymptomatic persons with no prior CAD.7–10 Although not designed for this purpose, the AgSc has also been evaluated in the prediction of obstructive CAD. With the use of myocardial perfusion imaging (MPI) to detect ischemia, it has been shown that more patients had inducible ischemia with increasing levels of CAC,11 but even in patients with an AgSc .400, less than one-half of the patients had inducible ischemia, resulting in poor specificity.12 In addition, several studies have investigated the diagnostic accuracy of whole-heart CAC scores for obstructive CAD with the use of invasive x-ray angiography (XRA) as reference standard.13–16 Similar to MPI, such studies showed that the specificity and positive predictive value (PPV) of the whole-heart CAC score, depending on the cutoff, is poor (specificity 5 23%–69%, and PPV 5 48%–70% at optimal receiver operating characteristics ROC cutoffs16). The poor specificity of whole-heart AgSc for the prediction of obstructive CAD is not surprising, because it is a summation of all calcified lesions along the coronary arteries; therefore, it does not capture the distribution of calcified lesions or the size of each of the individual lesions or clusters. In contrast, the 3-dimensional CAC image data set contains detailed information well beyond the wholeheart AgSc and VolSc, such as the geometric and the structural characteristics of individual calcified lesions, which can be of high diagnostic value. It is intuitive that a significant stenosis may be associated with a larger calcified lesion, and, conversely, a larger lesion would have a higher likelihood to be associated with an obstructive lesion. Surprisingly, a more detailed and localized lesionspecific measurement of calcified lesions has not been previously described, and the diagnostic accuracy of vesseland lesion-specific calcium scores for obstructive CAD remain unknown. We, therefore, hypothesized that vesseland lesion-specific parameters will better predict obstructive CAD compared with the whole-heart AgSc, using invasive XRA as the reference standard.
Methods This was a retrospective, single-center, observational study; the investigational protocol was approved by the
institutional review board of Piedmont Hospital. Consecutive patients (n 5 91) referred to our institution between January 2005 and June 2006 for both CAC scoring and invasive coronary XRA within 12 months were included. This study was designed to investigate the diagnostic accuracy of different CAC measurements for obstructive CAD in patients with positive CAC; those with negative calcium scores were excluded (7 patients).17 Enrollment criteria were met by 84 patients. Demographic and clinical information was collected from medical records. CAC examinations were performed on a 32 ! 2 multidetector computed CT system (Siemens Somatom 64; Siemens, Erlangen, Germany). Noncontrast CAC scans were performed during end-respiratory breath hold with retrospective electrocardiographic gating. CAC images were acquired with 3-mm collimation with a 2-mm interslice gap. Other acquisition parameters included a gantry rotation of 375 milliseconds, pitch of 0.24, tube voltage of 120 kV, and tube current of 250 mAs. Invasive coronary XRA was performed with the use of the Judkins technique, based on institutional protocols, acquiring a minimum of 5 views of the left coronary system and 2 views of the right coronary system. The absence or presence of discrete coronary artery stenoses was examined in 2 orthogonal views, and obstructive disease was confirmed by
Figure 1 Lesion-specific AgSc was quantified in each individual calcified lesion. Vessel-specific score was the sum of the lesionspecific scores in the corresponding vessel. Each vessel’s maximum and mean lesion-specific scores were also recorded and calculated. LAD, left anterior descending artery; RCA, right coronary artery.
Qian et al Table 1
Lesion- and vessel-specific CAC scores to diagnose obstructive CAD
393
Demographic data
All, n (%) Male, n (%) Female, n (%) Age, y, mean 6 SD
Patients
,50% Stenosis
R50% Stenosis
R70% Stenosis
84 52 (62) 32 (38) 66.1 6 8.5
27 (32) 14 (27) 13 (41) 65.3 6 10.0
57 (68) 38 (73) 19 (59) 66.6 6 7.8
33 (39) 22 (42) 11 (34) 65.8 6 8.2
Age distribution was similar in patients with and without obstructive CAD by X-ray angiography with the use of unpaired Welch tests, with P 5 .55 (based on R50% diameter stenosis) and 0.74 (based on R70% diameter stenosis). Sex distribution was similar in patients with and without obstructive CAD by X-ray angiography on the basis of frequency table analysis and chi-square tests, with P 5 .29 (based on R50% diameter stenosis) and 0.62 (based on R70% diameter stenosis).
visual assessment. Two criteria for obstructive CAD, R50% diameter stenosis and R70% diameter stenosis, were used. The coronary vasculature was divided into 4 territories, ie, left main coronary artery (LMCA), left anterior descending artery (LAD), left circumflex artery (LCx), and right coronary artery (RCA). Presence of obstructive CAD was recorded for each corresponding vessel. Total AgSc was calculated by standard methods. The area of calcification was multiplied by an arbitrary weighted density score that was based on the maximum Hounsfield unit value in the identified 2-dimensional (2D) calcified lesion, as follows: 1 5 130–199 HU; 2 5 200–299 HU; 3 5 300–399 HU; 4 5 R400 HU. Calcium scores in all 2D calcified lesions along the major epicardial arteries were summed to derive the whole-heart AgSc. VolSc was calculated as the total volume of the calcification in the major epicardial arteries, where calcification was identified as voxels with attenuation values of R130 HU. As shown in Figure 1, lesion-specific AgSc and VolSc were measured in each single calcified lesion. A single calcified lesion was defined as a complete group of connected calcified voxels that was based on the 6 connectedness criterion in 3-dimension.18 A software system has been developed to allow users to manually annotate calcified lesions followed by automated lesion segmentation. Vessel-specific CAC was determined as the sum of the lesion-specific CAC scores in the corresponding vessel. Lesions that covered R2 vessels were divided according to their locations. We specifically recorded maximum lesion-specific CAC and calculated the mean lesion-specific CAC in each vessel (vesselspecific score divided by the number of individual lesions). Because we excluded patients with negative CAC scores in the study cohort, to achieve comparable sample distribution, we only investigated vessels with positive CAC scores in the vessel- and lesion-specific CAC study. Normality of the data distribution was assessed by the Kolmogorov-Smirnov test. Normally distributed continuous variables were compared by unpaired 2-tailed t test. Nonnormally distributed values were compared by the MannWhitney test. ROC curves and the area under the curve (AUC) were used to evaluate and compare the diagnostic performance of different CAC parameters in diagnosing obstructive CAD. ROC cutoffs were also selected to achieve
80% specificity or sensitivity, and the corresponding PPV, negative predictive value (NPV), and sensitivity/specificity were compared between the whole-heart and vessel- and lesion-specific approaches. Statistical significance was determined by a P value of ,.05.
Results Patient characteristics Mean age (n 5 84) was 66.1 6 8.5 years, and 52 of patients were men (61.2%). Clinical variables were similar between patients with and without obstructive CAD by either definition (R50% or R70% diameter stenosis) (Table 1).
CAD by invasive coronary angiography We evaluated 242 calcium-positive coronary vessels, including 38 LMCA, 79 LAD, 68 LCx, and 57 RCA. On the basis of a criterion of R50% stenosis, 57 of 84 patients
Figure 2 Comparison of CAC scores between patients with and without obstructive disease with the use of the Mann-Whitney test, based on R50% diameter stenosis. WH, whole-heart; VS, vesselspecific; Max LS, maximum lesion-specific. Bar graphs show the median and interquartile range of different CAC quantifications.
394
Journal of Cardiovascular Computed Tomography, Vol 4, No 6, November/December 2010
Figure 3 ROCs of different CAC quantification approaches in predicting R50% (A) and R70% (B) stenosis. The areas under each of the curves (AUCs) are listed in Table 2. Vessel-specific AgSc/VolSc achieved the best AUC values (0.72/0.72). Max LS, maximum lesionspecific score in a vessel; Mean LS, mean lesion-specific score in a vessel.
(67.9%) and 83 of 242 vessels (34.3%) had obstructive CAD (1 LMCA, 47 LAD, 17 LCx, and 18 RCA); based on a criterion of R70% stenosis, 33 of 84 patients (39.3%) and 36 of 242 vessels (14.9%) had obstructive CAD (15 LAD, 11 LCx, and 10 RCA). Table 2
Whole-heart CAC scores, vessel-specific and lesion-specific CAC scores CAC scores were nonnormally distributed in the overall study population (Kolmogorov-Smirnov test for normal
ROC curve analysis AUC
50% Stenosis WH-AgSc WH-AgSc WH-AgSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc 70% Stenosis WH-AgSc WH-AgSc WH-AgSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc
0.58 0.59 0.72 0.72 0.71 0.72 0.68 0.68
0.66 0.66 0.77 0.77 0.75 0.76 0.7 0.7
Cutoff
Sens, %
Spec, %
PPV, %
NPV, %
Accuracy, %
986 100 400 894 149 125 76 132 42 50
31.6 70.5 47.5 29.8 66.3 69.9 73.5 50.6 61.4 62.7
92.6 36.7 60 92.6 74.8 73.6 64.8 88.1 74.8 74.2
90 69.4 70.7 89.5 57.9 58 52.1 68.9 56 55.9
39.1 37.9 36 38.5 81 82.4 82.4 77.3 78.8 79.2
51.2 60.7 53.6 50 71.9 72.3 67.8 75.2 70.2 70.2
790 100 400 484 176 173 182 154 77 41
48.5 77.1 60 57.6 72.2 72.2 61.1 61.1 55.6 72.2
82.4 37.5 64.3 76.5 71.8 75.2 82.5 83.5 80.6 65.5
64 43.5 51.2 61.3 31 33.8 37.9 39.3 33.3 26.8
71.2 72.4 72 73.6 93.7 93.9 92.4 92.5 91.2 93.1
69 51.2 61.9 69 71.9 74.8 79.3 80.2 76.9 66.5
WH, whole-heart; VS, volume-specific; Max LS, maximum lesion-specific; Mean LS, mean of lesion-specific; AUC, area under ROC curve; Sens, sensitivity; Spec, specificity; PPV, positive predictive value; NPV, negative predictive value. Cutoffs were selected to achieve maximum summation of sensitivity and specificity. Conventional cutoffspoints of 100 and 400 of WH-AgSc were also tested.
Qian et al
Lesion- and vessel-specific CAC scores to diagnose obstructive CAD
395
Figure 4 The comparison of test accuracy in predicting R50% and R70% stenosis at the statistically defined optimal cutoffs. The accuracy values can be found in Table 2.
distribution, P 5 .001); mean (6SD) whole heart AgSc was 641 6 709, median was 400. Mean whole-heart VolSc was 536 6 570; median was 338. Whole-heart, vesselspecific and lesion-specific CAC values are shown in Figure 2. With the use of a criterion of R50% stenosis for obstructive CAD by XRA, whole-heart AgSc and VolSc were not significantly different between patients with and without obstructive CAD (median, 436, interquartile range IQR, 95–1177, vs median, 309, IQR, 84–574; P 5 .23; and median, 374, IQR, 108–988, vs median, 265, IQR, 76–480; P 5 .18, respectively). However, vessel-specific and lesion-specific values were significantly higher in patients with obstructive CAD (maximum lesion specific AgSc: 165; IQR, 53–294, vs 47, IQR, 15–110; P , .0001).
Table 3
Diagnostic characteristics of whole-heart, vessel- and lesion-specific CAC scores by ROC analysis ROC curves of whole-heart, vessel- and lesion-specific CAC scores in predicting R50% and R70% obstructive CAD are shown in Figure 3. Corresponding optimal cutoffs with corresponding sensitivity, specificity, PPV, NPV, and test accuracy are shown in Table 2. Test accuracies at statistically defined optimal cutoffs are also shown in Figure 4. For the prediction of 50% stenosis by XRA, all vessel- and lesion-specific scores had higher AUCs than whole-heart values (whole-heart AgSc AUC: 0.58; maximum lesion-specific VolSc AUC: 0.72). Statistically
Comparison of CAC predictive values when specificities are fixed at 80%
50% Stenosis WH-AglSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc 70% Stenosis WH-AglSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc
Cutoff
Sens, %
Spec, %
PPV, %
NPV, %
790 661 194 158 119 97 63 53
35.1 36.8 60.2 61.4 56.6 55.4 48.2 48.2
80 80 80 80 80 80 80 80
80 80.8 61 61.4 59.5 59 55.6 55.6
37.3 37.9 79.9 79.4 77.9 77.4 74.7 74.7
770 661 264 238 155 127 76 67
48.5 48.5 61.1 58.3 61.1 61.1 55.6 52.8
80 80 80 80 80 80 80 80
61.5 61.5 34.9 33.9 34.9 34.9 32.8 31.7
70.7 70.7 92.2 91.7 92.2 92.2 91.2 90.7
WH, whole-heart; VS, volume-specific; Max LS, maximum lesion-specific; Mean LS, mean of lesion-specific; AUC, area under ROC curve; Sens, sensitivity; Spec, specificity; PPV, positive predictive value; NPV, negative predictive value.
396 Table 4
Journal of Cardiovascular Computed Tomography, Vol 4, No 6, November/December 2010 Comparison of CAC predictive values when sensitivities are fixed at 80% Cutoff
50% Stenosis WH-AglSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc 70% Stenosis WH-AglSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc
Sens, %
Spec, %
PPV, %
NPV, %
71 63 34 47 34 31 20 18
80 80 80 80 80 80 80 80
22.2 22.2 35.8 42.1 40.3 42.1 44 42.1
68.7 68.7 39.6 42.1 41.4 42.1 42.9 42.1
35.3 35.3 78.1 80.7 80.0 80.7 81.4 80.7
232 127 114 110 78 71 24 23
80 80 80 80 80 80 80 80
43.1 41.2 63.1 60.7 59.7 62.1 45.1 46.6
48.2 47.4 27.6 26.4 25.9 27.1 20.4 20.9
78.6 77.8 94.9 94.7 94.6 94.8 93.0 93.2
WH, whole-heart; VS, volume-specific; Max LS, maximum lesion-specific; Mean LS, mean of lesion-specific; AUC, area under ROC curve; Sens, sensitivity; Spec, specificity; PPV, positive predictive value; NPV, negative predictive value.
defined, optimal cutoffs for whole-heart AgSc was 986; corresponding sensitivity, specificity, and test accuracy were 32%, 93%, and 51%, respectively. The optimal cutoff for maximum lesion-specific VolSc was 132 with 51% sensitivity, 88% specificity, and 75% accuracy. A similar pattern was also seen for the prediction of R70% stenosis by XRA, although the performance of whole-heart AgSc improved (AUC, 0.66) compared with R50% stenosis. Because the ‘‘optimal cutoff’’ of 986 for whole-heart AgSc is not very clinically meaningful, we also examined the diagnostic performance of each of the parameters with either sensitivity or specificity fixed at 80% (Table 3 and Table 4). This analysis showed that when sensitivity was fixed at 80%, specificity increased from 22.2% (whole-heart AgSc) to 42.1% (maximum lesion-specific VolSc) for the detection of R50% stenosis; for the detection of R70% stenosis, specificity increased from 43.1% (whole-heart AgSc) to 62.1% (maximum lesion-specific VolSc). When specificity was fixed at 80%, sensitivity increased from 35.1% (whole-heart AgSc) to 61.4% (vessel-specific VolSc) for the detection of 50% stenosis; for the detection of R70% stenosis, sensitivity increased from 48.5% (whole-heart AgSc) to 61.1% (maximum lesion-specific AgSc).
with the lesion-specific approach from 50% to only 17.9%, compared with the whole-heart AgSc (Fig. 5).
Discussion The study provides the first analysis of CAC data sets that were based on vessel- and lesion-specific measurements and showed superior diagnostic accuracy for the prediction of obstructive CAD with the use of invasive coronary angiography as reference standard. Indeed, the main finding of this study was that maximum lesion-specific AgSc resulted in a significantly increased diagnostic accuracy compared with the whole-heart AgSc, as evidenced by a higher AUC on ROC analysis (0.72 vs 0.58). Therefore, when specificity was fixed at 80%, as often necessary in clinical practice, sensitivity of vessel- and lesion-specific values increased by approximately 20%, which is a clinically significant improvement.
Classification of patients Another measure of improved diagnostic accuracy is the overall correct classification of patients. Assuming that the 80% sensitivity and specificity cutoffs are clinically meaningful and that patients with values above and below the respective cutoffs are considered positive or negative and those who fall between those values are indeterminate, the proportion of indeterminate patients significantly decreased
Figure 5 Proportion of patients allocated differently in treatment zones with the use of whole-heart and vessel-specific approaches. It is noticeable that the patient number in the indeterminate zone is significantly decreased with the use of the vessel-specific volume score.
Qian et al
Lesion- and vessel-specific CAC scores to diagnose obstructive CAD
397
Figure 6 Two typical cases showing the effectiveness of our vessel- and lesion-specific CAC approach. (A) The patient has a whole-heart AgSc of 790, with a relatively low maximum lesion-specific (LS) AgSc (76). Indeed, invasive angiography showed only minimal luminal disease. However, the whole-heart AgSc is only 347 in patient (B) with a relatively high maximum LS-AgSc (95). Accordingly, invasive angiography showed an 80% stenosis corresponding to that lesion (arrows).
The whole-heart AgSc and VolSc are a simple summation of all calcified lesions in the 3 major epicardial arteries and are therefore inherently limited in the geographic localization of specific stenoses. Accordingly, in our study, whole-heart AgSc and VolSc were not different in patients with and without obstructive CAD, when CAD was defined on the basis of 50% stenosis and diagnostic accuracy for obstructive CAD was poor. However, the 3-dimensional CAC data set contains significantly more information beyond the whole-heart AgSc and VolSc; therefore, it has the potential to improve diagnostic accuracy. Figure 6 shows 2 typical examples that illustrate the better predictive value of our vessel- and lesion-specific CAC approach.
From a clinical perspective, an AUC of R0.70, which was seen with all vessel- and lesion-specific parameters, is comparable to other clinically useful diagnostic tests, such as radionuclide MPI (87% sensitivity and 73% specificity) or stress echocardiography (70% sensitivity and 89% specificity).19 We, therefore, propose that a vessel- and lesion-specific approach is more acceptable for clinical decision making than the traditional whole-heart AgSc. It is interesting to speculate how vessel- and lesionspecific scores may be incorporated in clinical practice. CAC imaging is currently recommended for asymptomatic subjects without known, but with intermediate likelihood of, CAD for the purposes of refined risk stratification.20 The
398
Journal of Cardiovascular Computed Tomography, Vol 4, No 6, November/December 2010
apparent paradox is, however, that revascularization is typically reserved for patients with symptoms resulting from myocardial ischemia; therefore, it has been a clinical challenge to recommend further diagnostic testing on the basis of CAC scanning in asymptomatic subjects. It has been suggested that because many patients may be asymptomatic because of physical inactivity, provocative testing may be performed to uncover potential ischemia in patients with elevated CAC scores (ie, AgSc of R400).21 However, as we pointed out earlier, the specificity and PPVof the whole-heart AgSc, whether using a cutoff of 400 or even 1000, are quite low.12 Therefore, we propose that the improved diagnostic accuracy of vesseland lesion-specific scores may be helpful in refining clinical decision making. As we showed, using cutoffs for 80% sensitivity (to rule out obstructive CAD) and for 80% specificity (to rule in obstructive CAD), a vessel- and lesion-specific approach significantly reduces the number of patients with an indeterminate CAC for the prediction of R50% obstructive CAD (Fig. 5). As a potential clinical approach, patients below the cutoff corresponding to 80% sensitivity (eg, vessel-specific VolSc ,47) clearly do not need further workup for ischemia or obstructive CAD, whereas those above the cutoff corresponding to 80% specificity (eg, vessel-specific VolSc .158) could be referred for MPI or for coronary angiography. Patients in the intermediate zone could be referred for MPI. In this regard, it is important to point out that significantly fewer patients fall into the indeterminate zone with the use of the vessel- and lesion-specific approach, compared with the whole-heart AgSc. Although to our knowledge, no previous investigation has evaluated the diagnostic accuracy of a vessel- and lesion-specific approach, the Calcium Coverage Score (CCS)22 has been previously introduced to improve the predictive value of CAD events beyond the whole-heart AgSc. Brown et al22 showed that the percentage of coronary arteries affected by calcific plaque was significantly associated with coronary heart disease events. However, there was no difference in the prediction of hard cardiac events (myocardial infarction or death) when CCS was compared with the whole-heart AgSc and mass scores.
Limitations Limitations of our study include its retrospective design and relatively small sample size. Because we used XRA as reference standard, this also introduced a referral bias. These limitations can be overcome by a prospective design with more patients, in which downstream testing is driven by vessel- and lesion-specific CAC values. In fact, one such prospective study is currently under way at our center.
Conclusion We showed that vessel- and lesion-specific CAC scores are superior in predicting obstructive CAD compared with
the whole-heart AgSc. Such a simple refinement may refine the selection of patients for subsequent testing on the basis of CAC imaging. Because vessel-specific scores are readily available on most commercial workstations, vessel-specific parameters may be taken into consideration for clinical decision making today.
References 1. Budoff MJ, 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. 2. Guerci AD, Spadaro LA, Popma JJ, Goodman KJ, Brundage BH, Budoff M, Lerner G, Vizza RF: Relation of coronary calcium score by electron beam computed tomography to arteriographic findings in asymptomatic and symptomatic adults. Am J Cardiol. 1997;79: 128–33. 3. Guerci AD, Spadaro LA, Goodman KJ, Lledo-Perez A, Newstein D, Lerner G, Arad Y: Comparison of electron beam computed tomography scanning and conventional risk factor assessment for the prediction of angiographic coronary artery disease. J Am Coll Cardiol. 1998;32:673–9. 4. Callister TQ, Cooil B, Raya SP, Lippolis NJ, Russo DJ, Raggi P: Coronary artery disease: improved reproducibility of calcium scoring with an electron-beam CT volumetric method. Radiology. 1998;208:807–14. 5. Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Viamonte M Jr., Detrano R: Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol. 1990;15:827–32. 6. Wilson PW: Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97:1837–47. 7. Arad Y, Spadaro LA, Goodman K, Newstein D, Guerci AD: Prediction of coronary events with electron beam computed tomography. J Am Coll Cardiol. 2000;36:1253–60. 8. Becker A, Leber A, White CW, Becker C, Reiser MF, Knez A: Multislice computed tomography for determination of coronary artery disease in a symptomatic patient population. Int J Cardiovasc Imaging. 2007;23:361–7. 9. Fallavollita JA, Brody AS, Bunnell IL, Kumar K, Canty JM Jr.: Fast computed tomography detection of coronary calcification in the diagnosis of coronary artery disease. Comparison with angiography in patients ,50 years old. Circulation. 1994;89:285–90. 10. Greenland P, LaBree L, Azen SP, Doherty TM, Detrano RC: Coronary artery calcium score combined with Framingham score for risk prediction in asymptomatic individuals. JAMA. 2004;291:210–5. 11. Berman DS, Wong ND, Gransar H, Miranda-Peats R, Dahlbeck J, Hayes SW, Friedman JD, Kang X, Polk D, Hachamovitch R, Shaw L, Rozanski A: Relationship between stress-induced myocardial ischemia and atherosclerosis measured by coronary calcium tomography. J Am Coll Cardiol. 2004;44:923–30. 12. He ZX, Hedrick TD, Pratt CM, Verani MS, Aquino V, Roberts R, Mahmarian JJ: Severity of coronary artery calcification by electron beam computed tomography predicts silent myocardial ischemia. Circulation. 2000;101:244–51. 13. Akram K, Voros S: Absolute coronary artery calcium scores are superior to MESA percentile rank in predicting obstructive coronary artery disease. Int J Cardiovasc Imaging. 2008;24:743–9. 14. Bielak LF, Rumberger JA, Sheedy PF, Schwartz RS, Peyser PA: Probabilistic model for prediction of angiographically defined obstructive coronary artery disease using electron beam computed tomography calcium score strata. Circulation. 2000;102:380–5. 15. Detrano R, Hsiai T, Wang S, Puentes G, Fallavollita J, Shields P, Stanford W, Wolfkiel C, Georgiou D, Budoff M, Reed J: Prognostic
Qian et al
16.
17.
18.
19.
Lesion- and vessel-specific CAC scores to diagnose obstructive CAD
value of coronary calcification and angiographic stenoses in patients undergoing coronary angiography. J Am Coll Cardiol. 1996;27: 285–90. Haberl R, Becker A, Leber A, Knez A, Becker C, Lang C, Bruning R, Reiser M, Steinbeck G: Correlation of coronary calcification and angiographically documented stenoses in patients with suspected coronary artery disease: results of 1,764 patients. J Am Coll Cardiol. 2001;37:451–7. Akram K, O’Donnell RE, King S, Superko HR, Agatston A, Voros S: Influence of symptomatic status on the prevalence of obstructive coronary artery disease in patients with zero calcium score. Atherosclerosis. 2009;203:533–7. Qian Z, Marvasty I, Anderson H, Rinehart S, Voros S: Lesion-specific coronary artery calcium quantification better predicts cardiac events. Sixth IEEE International Symposium on Biomedical Imaging. 2009; 237–40. Marcassa C, Bax JJ, Bengel F, Hesse B, Petersen CL, Reyes E, Underwood R: Clinical value, cost-effectiveness, and safety of myocardial perfusion scintigraphy: a position statement. Eur Heart J. 2008;29:557–63.
399
20. Greenland P, Bonow RO, Brundage BH, Budoff MJ, Eisenberg MJ, Grundy SM, Lauer MS, Post WS, Raggi P, Redberg RF, Rodgers GP, Shaw LJ, Taylor AJ, Weintraub WS, Harrington RA, Abrams J, Anderson JL, Bates ER, Grines CL, Hlatky MA, Lichtenberg RC, Lindner JR, Pohost GM, Schofield RS, Shubrooks SJ Jr., Stein JH, Tracy CM, Vogel RA, Wesley DJ: ACCF/AHA 2007 clinical expert consensus document on coronary artery calcium scoring by computed tomography in global cardiovascular risk assessment and in evaluation of patients with chest pain: a report of the American College of Cardiology Foundation Clinical Expert Consensus Task Force (ACCF/AHA Writing Committee to Update the 2000 Expert Consensus Document on Electron Beam Computed Tomography). Circulation. 2007;115: 402–26. 21. Naghavi M, Falk E, Hecht HS, Shah PK: The first SHAPE (Screening for Heart Attack Prevention and Education) guideline. Crit Pathw Cardiol. 2006;5:187–90. 22. Brown ER, Kronmal RA, Bluemke DA, Guerci AD, Carr JJ, Goldin J, Detrano R: Coronary calcium coverage score: determination, correlates, and predictive accuracy in the Multi-Ethnic Study of Atherosclerosis. Radiology. 2008;247:669–75.
Original Research Article
Lesion- and vessel-specific coronary artery calcium scores are superior to whole-heart Agatston and volume scores in the diagnosis of obstructive coronary artery disease Zhen Qian, PhD*, Hunt Anderson, MD, Idean Marvasty, BS, Kamran Akram, MD, Gustavo Vazquez, MD, Sarah Rinehart, MD, FACC, Szilard Voros, MD, FACC, FSCCT Piedmont Heart Institute, 95 Collier Road, Suite 2085, Atlanta, GA 30309, USA KEYWORDS: Computed tomography; Coronary artery calcium; Obstructive coronary artery disease
BACKGROUND: The whole-heart coronary artery calcium (CAC) score has poor predictive value for obstructive coronary artery disease (CAD). We hypothesized that vessel- and lesion-specific CAC scores are more accurate. OBJECTIVES: To evaluate the usefulness of vessel- and lesion-specific CAC in predicting obstructive CAD and to assess the incremental value added by the vessel- and lesion-specific CAC to the conventional whole-heart CAC approach. METHODS: Ninety-one patients with CAC scores and invasive angiography (XRA) data were enrolled. Besides whole-heart CAC, Agatston score (AgSc) and volume score (VolSc) were measured individually for each lesion in the 4 major epicardial coronary arteries. Maximum and average lesionspecific scores in each vessel were also determined. For the primary analysis, obstructive CAD was defined as 50% diameter stenosis by XRA. RESULTS: Whole-heart AgSc and VolSc were not different between patients with and without obstructive CAD (P 5 .23 and P 5 .18), whereas vessel- and lesion-specific scores were (maximum lesion specific AgSc, P , .0001). Maximum lesion-specific AgSc had superior diagnostic performance compared with whole-heart AgSc (area under receiver operating characteristics, 0.71 vs 0.58). Overall sensitivity, specificity, and diagnostic accuracy were improved. When specificity was fixed at 80%, sensitivity of maximum lesion-specific AgSc was superior to whole-heart AgSc (56.6% vs 35.1%). Most importantly, with lesion-specific AgSc, fewer patients were classified as ‘‘indeterminate’’ compared with whole-heart AgSc (17.9% vs 50%). CONCLUSIONS: Vessel- and lesion-specific CAC scores are superior to the whole-heart AgSc and VolSc in predicting obstructive CAD. This simple refinement in CAC scoring may significantly improve the clinical predictive role of CAC imaging. Ó 2010 Society of Cardiovascular Computed Tomography. All rights reserved.
Conflict of interest: The authors report no conflicts of interest. * Corresponding author. E-mail address: [email protected] Submitted May 11, 2010. Accepted for publication September 3, 2010.
Introduction Computed tomographic (CT)–based measurement of coronary artery calcium (CAC) has been introduced as a
1934-5925/$ - see front matter Ó 2010 Society of Cardiovascular Computed Tomography. All rights reserved. doi:10.1016/j.jcct.2010.09.001
392
Journal of Cardiovascular Computed Tomography, Vol 4, No 6, November/December 2010
noninvasive, low-radiation method to assess the overall coronary arterial atherosclerotic burden, by quantifying calcium in the coronary vasculature.1–3 The whole-heart volume score (VolSc)4 is a summation of all voxels along the 3 major epicardial coronary arteries with an attenuation value of R130 Hounsfield units (HU), whereas the wholeheart Agatston score (AgSc)5 also takes into account the overall density of each of the voxels. The AgSc was developed to estimate the overall burden of coronary atherosclerosis and to estimate the cardiovascular risk in asymptomatic subjects without known coronary artery disease (CAD). In this regard, the AgSc has been shown to be independent of and additive to the Framingham risk score6 in predicting cardiovascular outcomes in asymptomatic persons with no prior CAD.7–10 Although not designed for this purpose, the AgSc has also been evaluated in the prediction of obstructive CAD. With the use of myocardial perfusion imaging (MPI) to detect ischemia, it has been shown that more patients had inducible ischemia with increasing levels of CAC,11 but even in patients with an AgSc .400, less than one-half of the patients had inducible ischemia, resulting in poor specificity.12 In addition, several studies have investigated the diagnostic accuracy of whole-heart CAC scores for obstructive CAD with the use of invasive x-ray angiography (XRA) as reference standard.13–16 Similar to MPI, such studies showed that the specificity and positive predictive value (PPV) of the whole-heart CAC score, depending on the cutoff, is poor (specificity 5 23%–69%, and PPV 5 48%–70% at optimal receiver operating characteristics ROC cutoffs16). The poor specificity of whole-heart AgSc for the prediction of obstructive CAD is not surprising, because it is a summation of all calcified lesions along the coronary arteries; therefore, it does not capture the distribution of calcified lesions or the size of each of the individual lesions or clusters. In contrast, the 3-dimensional CAC image data set contains detailed information well beyond the wholeheart AgSc and VolSc, such as the geometric and the structural characteristics of individual calcified lesions, which can be of high diagnostic value. It is intuitive that a significant stenosis may be associated with a larger calcified lesion, and, conversely, a larger lesion would have a higher likelihood to be associated with an obstructive lesion. Surprisingly, a more detailed and localized lesionspecific measurement of calcified lesions has not been previously described, and the diagnostic accuracy of vesseland lesion-specific calcium scores for obstructive CAD remain unknown. We, therefore, hypothesized that vesseland lesion-specific parameters will better predict obstructive CAD compared with the whole-heart AgSc, using invasive XRA as the reference standard.
Methods This was a retrospective, single-center, observational study; the investigational protocol was approved by the
institutional review board of Piedmont Hospital. Consecutive patients (n 5 91) referred to our institution between January 2005 and June 2006 for both CAC scoring and invasive coronary XRA within 12 months were included. This study was designed to investigate the diagnostic accuracy of different CAC measurements for obstructive CAD in patients with positive CAC; those with negative calcium scores were excluded (7 patients).17 Enrollment criteria were met by 84 patients. Demographic and clinical information was collected from medical records. CAC examinations were performed on a 32 ! 2 multidetector computed CT system (Siemens Somatom 64; Siemens, Erlangen, Germany). Noncontrast CAC scans were performed during end-respiratory breath hold with retrospective electrocardiographic gating. CAC images were acquired with 3-mm collimation with a 2-mm interslice gap. Other acquisition parameters included a gantry rotation of 375 milliseconds, pitch of 0.24, tube voltage of 120 kV, and tube current of 250 mAs. Invasive coronary XRA was performed with the use of the Judkins technique, based on institutional protocols, acquiring a minimum of 5 views of the left coronary system and 2 views of the right coronary system. The absence or presence of discrete coronary artery stenoses was examined in 2 orthogonal views, and obstructive disease was confirmed by
Figure 1 Lesion-specific AgSc was quantified in each individual calcified lesion. Vessel-specific score was the sum of the lesionspecific scores in the corresponding vessel. Each vessel’s maximum and mean lesion-specific scores were also recorded and calculated. LAD, left anterior descending artery; RCA, right coronary artery.
Qian et al Table 1
Lesion- and vessel-specific CAC scores to diagnose obstructive CAD
393
Demographic data
All, n (%) Male, n (%) Female, n (%) Age, y, mean 6 SD
Patients
,50% Stenosis
R50% Stenosis
R70% Stenosis
84 52 (62) 32 (38) 66.1 6 8.5
27 (32) 14 (27) 13 (41) 65.3 6 10.0
57 (68) 38 (73) 19 (59) 66.6 6 7.8
33 (39) 22 (42) 11 (34) 65.8 6 8.2
Age distribution was similar in patients with and without obstructive CAD by X-ray angiography with the use of unpaired Welch tests, with P 5 .55 (based on R50% diameter stenosis) and 0.74 (based on R70% diameter stenosis). Sex distribution was similar in patients with and without obstructive CAD by X-ray angiography on the basis of frequency table analysis and chi-square tests, with P 5 .29 (based on R50% diameter stenosis) and 0.62 (based on R70% diameter stenosis).
visual assessment. Two criteria for obstructive CAD, R50% diameter stenosis and R70% diameter stenosis, were used. The coronary vasculature was divided into 4 territories, ie, left main coronary artery (LMCA), left anterior descending artery (LAD), left circumflex artery (LCx), and right coronary artery (RCA). Presence of obstructive CAD was recorded for each corresponding vessel. Total AgSc was calculated by standard methods. The area of calcification was multiplied by an arbitrary weighted density score that was based on the maximum Hounsfield unit value in the identified 2-dimensional (2D) calcified lesion, as follows: 1 5 130–199 HU; 2 5 200–299 HU; 3 5 300–399 HU; 4 5 R400 HU. Calcium scores in all 2D calcified lesions along the major epicardial arteries were summed to derive the whole-heart AgSc. VolSc was calculated as the total volume of the calcification in the major epicardial arteries, where calcification was identified as voxels with attenuation values of R130 HU. As shown in Figure 1, lesion-specific AgSc and VolSc were measured in each single calcified lesion. A single calcified lesion was defined as a complete group of connected calcified voxels that was based on the 6 connectedness criterion in 3-dimension.18 A software system has been developed to allow users to manually annotate calcified lesions followed by automated lesion segmentation. Vessel-specific CAC was determined as the sum of the lesion-specific CAC scores in the corresponding vessel. Lesions that covered R2 vessels were divided according to their locations. We specifically recorded maximum lesion-specific CAC and calculated the mean lesion-specific CAC in each vessel (vesselspecific score divided by the number of individual lesions). Because we excluded patients with negative CAC scores in the study cohort, to achieve comparable sample distribution, we only investigated vessels with positive CAC scores in the vessel- and lesion-specific CAC study. Normality of the data distribution was assessed by the Kolmogorov-Smirnov test. Normally distributed continuous variables were compared by unpaired 2-tailed t test. Nonnormally distributed values were compared by the MannWhitney test. ROC curves and the area under the curve (AUC) were used to evaluate and compare the diagnostic performance of different CAC parameters in diagnosing obstructive CAD. ROC cutoffs were also selected to achieve
80% specificity or sensitivity, and the corresponding PPV, negative predictive value (NPV), and sensitivity/specificity were compared between the whole-heart and vessel- and lesion-specific approaches. Statistical significance was determined by a P value of ,.05.
Results Patient characteristics Mean age (n 5 84) was 66.1 6 8.5 years, and 52 of patients were men (61.2%). Clinical variables were similar between patients with and without obstructive CAD by either definition (R50% or R70% diameter stenosis) (Table 1).
CAD by invasive coronary angiography We evaluated 242 calcium-positive coronary vessels, including 38 LMCA, 79 LAD, 68 LCx, and 57 RCA. On the basis of a criterion of R50% stenosis, 57 of 84 patients
Figure 2 Comparison of CAC scores between patients with and without obstructive disease with the use of the Mann-Whitney test, based on R50% diameter stenosis. WH, whole-heart; VS, vesselspecific; Max LS, maximum lesion-specific. Bar graphs show the median and interquartile range of different CAC quantifications.
394
Journal of Cardiovascular Computed Tomography, Vol 4, No 6, November/December 2010
Figure 3 ROCs of different CAC quantification approaches in predicting R50% (A) and R70% (B) stenosis. The areas under each of the curves (AUCs) are listed in Table 2. Vessel-specific AgSc/VolSc achieved the best AUC values (0.72/0.72). Max LS, maximum lesionspecific score in a vessel; Mean LS, mean lesion-specific score in a vessel.
(67.9%) and 83 of 242 vessels (34.3%) had obstructive CAD (1 LMCA, 47 LAD, 17 LCx, and 18 RCA); based on a criterion of R70% stenosis, 33 of 84 patients (39.3%) and 36 of 242 vessels (14.9%) had obstructive CAD (15 LAD, 11 LCx, and 10 RCA). Table 2
Whole-heart CAC scores, vessel-specific and lesion-specific CAC scores CAC scores were nonnormally distributed in the overall study population (Kolmogorov-Smirnov test for normal
ROC curve analysis AUC
50% Stenosis WH-AgSc WH-AgSc WH-AgSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc 70% Stenosis WH-AgSc WH-AgSc WH-AgSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc
0.58 0.59 0.72 0.72 0.71 0.72 0.68 0.68
0.66 0.66 0.77 0.77 0.75 0.76 0.7 0.7
Cutoff
Sens, %
Spec, %
PPV, %
NPV, %
Accuracy, %
986 100 400 894 149 125 76 132 42 50
31.6 70.5 47.5 29.8 66.3 69.9 73.5 50.6 61.4 62.7
92.6 36.7 60 92.6 74.8 73.6 64.8 88.1 74.8 74.2
90 69.4 70.7 89.5 57.9 58 52.1 68.9 56 55.9
39.1 37.9 36 38.5 81 82.4 82.4 77.3 78.8 79.2
51.2 60.7 53.6 50 71.9 72.3 67.8 75.2 70.2 70.2
790 100 400 484 176 173 182 154 77 41
48.5 77.1 60 57.6 72.2 72.2 61.1 61.1 55.6 72.2
82.4 37.5 64.3 76.5 71.8 75.2 82.5 83.5 80.6 65.5
64 43.5 51.2 61.3 31 33.8 37.9 39.3 33.3 26.8
71.2 72.4 72 73.6 93.7 93.9 92.4 92.5 91.2 93.1
69 51.2 61.9 69 71.9 74.8 79.3 80.2 76.9 66.5
WH, whole-heart; VS, volume-specific; Max LS, maximum lesion-specific; Mean LS, mean of lesion-specific; AUC, area under ROC curve; Sens, sensitivity; Spec, specificity; PPV, positive predictive value; NPV, negative predictive value. Cutoffs were selected to achieve maximum summation of sensitivity and specificity. Conventional cutoffspoints of 100 and 400 of WH-AgSc were also tested.
Qian et al
Lesion- and vessel-specific CAC scores to diagnose obstructive CAD
395
Figure 4 The comparison of test accuracy in predicting R50% and R70% stenosis at the statistically defined optimal cutoffs. The accuracy values can be found in Table 2.
distribution, P 5 .001); mean (6SD) whole heart AgSc was 641 6 709, median was 400. Mean whole-heart VolSc was 536 6 570; median was 338. Whole-heart, vesselspecific and lesion-specific CAC values are shown in Figure 2. With the use of a criterion of R50% stenosis for obstructive CAD by XRA, whole-heart AgSc and VolSc were not significantly different between patients with and without obstructive CAD (median, 436, interquartile range IQR, 95–1177, vs median, 309, IQR, 84–574; P 5 .23; and median, 374, IQR, 108–988, vs median, 265, IQR, 76–480; P 5 .18, respectively). However, vessel-specific and lesion-specific values were significantly higher in patients with obstructive CAD (maximum lesion specific AgSc: 165; IQR, 53–294, vs 47, IQR, 15–110; P , .0001).
Table 3
Diagnostic characteristics of whole-heart, vessel- and lesion-specific CAC scores by ROC analysis ROC curves of whole-heart, vessel- and lesion-specific CAC scores in predicting R50% and R70% obstructive CAD are shown in Figure 3. Corresponding optimal cutoffs with corresponding sensitivity, specificity, PPV, NPV, and test accuracy are shown in Table 2. Test accuracies at statistically defined optimal cutoffs are also shown in Figure 4. For the prediction of 50% stenosis by XRA, all vessel- and lesion-specific scores had higher AUCs than whole-heart values (whole-heart AgSc AUC: 0.58; maximum lesion-specific VolSc AUC: 0.72). Statistically
Comparison of CAC predictive values when specificities are fixed at 80%
50% Stenosis WH-AglSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc 70% Stenosis WH-AglSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc
Cutoff
Sens, %
Spec, %
PPV, %
NPV, %
790 661 194 158 119 97 63 53
35.1 36.8 60.2 61.4 56.6 55.4 48.2 48.2
80 80 80 80 80 80 80 80
80 80.8 61 61.4 59.5 59 55.6 55.6
37.3 37.9 79.9 79.4 77.9 77.4 74.7 74.7
770 661 264 238 155 127 76 67
48.5 48.5 61.1 58.3 61.1 61.1 55.6 52.8
80 80 80 80 80 80 80 80
61.5 61.5 34.9 33.9 34.9 34.9 32.8 31.7
70.7 70.7 92.2 91.7 92.2 92.2 91.2 90.7
WH, whole-heart; VS, volume-specific; Max LS, maximum lesion-specific; Mean LS, mean of lesion-specific; AUC, area under ROC curve; Sens, sensitivity; Spec, specificity; PPV, positive predictive value; NPV, negative predictive value.
396 Table 4
Journal of Cardiovascular Computed Tomography, Vol 4, No 6, November/December 2010 Comparison of CAC predictive values when sensitivities are fixed at 80% Cutoff
50% Stenosis WH-AglSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc 70% Stenosis WH-AglSc WH-VolSc VS-AgSc VS-VolSc Max LS-AgSc Max LS-VolSc Mean LS-AgSc Mean LS-VolSc
Sens, %
Spec, %
PPV, %
NPV, %
71 63 34 47 34 31 20 18
80 80 80 80 80 80 80 80
22.2 22.2 35.8 42.1 40.3 42.1 44 42.1
68.7 68.7 39.6 42.1 41.4 42.1 42.9 42.1
35.3 35.3 78.1 80.7 80.0 80.7 81.4 80.7
232 127 114 110 78 71 24 23
80 80 80 80 80 80 80 80
43.1 41.2 63.1 60.7 59.7 62.1 45.1 46.6
48.2 47.4 27.6 26.4 25.9 27.1 20.4 20.9
78.6 77.8 94.9 94.7 94.6 94.8 93.0 93.2
WH, whole-heart; VS, volume-specific; Max LS, maximum lesion-specific; Mean LS, mean of lesion-specific; AUC, area under ROC curve; Sens, sensitivity; Spec, specificity; PPV, positive predictive value; NPV, negative predictive value.
defined, optimal cutoffs for whole-heart AgSc was 986; corresponding sensitivity, specificity, and test accuracy were 32%, 93%, and 51%, respectively. The optimal cutoff for maximum lesion-specific VolSc was 132 with 51% sensitivity, 88% specificity, and 75% accuracy. A similar pattern was also seen for the prediction of R70% stenosis by XRA, although the performance of whole-heart AgSc improved (AUC, 0.66) compared with R50% stenosis. Because the ‘‘optimal cutoff’’ of 986 for whole-heart AgSc is not very clinically meaningful, we also examined the diagnostic performance of each of the parameters with either sensitivity or specificity fixed at 80% (Table 3 and Table 4). This analysis showed that when sensitivity was fixed at 80%, specificity increased from 22.2% (whole-heart AgSc) to 42.1% (maximum lesion-specific VolSc) for the detection of R50% stenosis; for the detection of R70% stenosis, specificity increased from 43.1% (whole-heart AgSc) to 62.1% (maximum lesion-specific VolSc). When specificity was fixed at 80%, sensitivity increased from 35.1% (whole-heart AgSc) to 61.4% (vessel-specific VolSc) for the detection of 50% stenosis; for the detection of R70% stenosis, sensitivity increased from 48.5% (whole-heart AgSc) to 61.1% (maximum lesion-specific AgSc).
with the lesion-specific approach from 50% to only 17.9%, compared with the whole-heart AgSc (Fig. 5).
Discussion The study provides the first analysis of CAC data sets that were based on vessel- and lesion-specific measurements and showed superior diagnostic accuracy for the prediction of obstructive CAD with the use of invasive coronary angiography as reference standard. Indeed, the main finding of this study was that maximum lesion-specific AgSc resulted in a significantly increased diagnostic accuracy compared with the whole-heart AgSc, as evidenced by a higher AUC on ROC analysis (0.72 vs 0.58). Therefore, when specificity was fixed at 80%, as often necessary in clinical practice, sensitivity of vessel- and lesion-specific values increased by approximately 20%, which is a clinically significant improvement.
Classification of patients Another measure of improved diagnostic accuracy is the overall correct classification of patients. Assuming that the 80% sensitivity and specificity cutoffs are clinically meaningful and that patients with values above and below the respective cutoffs are considered positive or negative and those who fall between those values are indeterminate, the proportion of indeterminate patients significantly decreased
Figure 5 Proportion of patients allocated differently in treatment zones with the use of whole-heart and vessel-specific approaches. It is noticeable that the patient number in the indeterminate zone is significantly decreased with the use of the vessel-specific volume score.
Qian et al
Lesion- and vessel-specific CAC scores to diagnose obstructive CAD
397
Figure 6 Two typical cases showing the effectiveness of our vessel- and lesion-specific CAC approach. (A) The patient has a whole-heart AgSc of 790, with a relatively low maximum lesion-specific (LS) AgSc (76). Indeed, invasive angiography showed only minimal luminal disease. However, the whole-heart AgSc is only 347 in patient (B) with a relatively high maximum LS-AgSc (95). Accordingly, invasive angiography showed an 80% stenosis corresponding to that lesion (arrows).
The whole-heart AgSc and VolSc are a simple summation of all calcified lesions in the 3 major epicardial arteries and are therefore inherently limited in the geographic localization of specific stenoses. Accordingly, in our study, whole-heart AgSc and VolSc were not different in patients with and without obstructive CAD, when CAD was defined on the basis of 50% stenosis and diagnostic accuracy for obstructive CAD was poor. However, the 3-dimensional CAC data set contains significantly more information beyond the whole-heart AgSc and VolSc; therefore, it has the potential to improve diagnostic accuracy. Figure 6 shows 2 typical examples that illustrate the better predictive value of our vessel- and lesion-specific CAC approach.
From a clinical perspective, an AUC of R0.70, which was seen with all vessel- and lesion-specific parameters, is comparable to other clinically useful diagnostic tests, such as radionuclide MPI (87% sensitivity and 73% specificity) or stress echocardiography (70% sensitivity and 89% specificity).19 We, therefore, propose that a vessel- and lesion-specific approach is more acceptable for clinical decision making than the traditional whole-heart AgSc. It is interesting to speculate how vessel- and lesionspecific scores may be incorporated in clinical practice. CAC imaging is currently recommended for asymptomatic subjects without known, but with intermediate likelihood of, CAD for the purposes of refined risk stratification.20 The
398
Journal of Cardiovascular Computed Tomography, Vol 4, No 6, November/December 2010
apparent paradox is, however, that revascularization is typically reserved for patients with symptoms resulting from myocardial ischemia; therefore, it has been a clinical challenge to recommend further diagnostic testing on the basis of CAC scanning in asymptomatic subjects. It has been suggested that because many patients may be asymptomatic because of physical inactivity, provocative testing may be performed to uncover potential ischemia in patients with elevated CAC scores (ie, AgSc of R400).21 However, as we pointed out earlier, the specificity and PPVof the whole-heart AgSc, whether using a cutoff of 400 or even 1000, are quite low.12 Therefore, we propose that the improved diagnostic accuracy of vesseland lesion-specific scores may be helpful in refining clinical decision making. As we showed, using cutoffs for 80% sensitivity (to rule out obstructive CAD) and for 80% specificity (to rule in obstructive CAD), a vessel- and lesion-specific approach significantly reduces the number of patients with an indeterminate CAC for the prediction of R50% obstructive CAD (Fig. 5). As a potential clinical approach, patients below the cutoff corresponding to 80% sensitivity (eg, vessel-specific VolSc ,47) clearly do not need further workup for ischemia or obstructive CAD, whereas those above the cutoff corresponding to 80% specificity (eg, vessel-specific VolSc .158) could be referred for MPI or for coronary angiography. Patients in the intermediate zone could be referred for MPI. In this regard, it is important to point out that significantly fewer patients fall into the indeterminate zone with the use of the vessel- and lesion-specific approach, compared with the whole-heart AgSc. Although to our knowledge, no previous investigation has evaluated the diagnostic accuracy of a vessel- and lesion-specific approach, the Calcium Coverage Score (CCS)22 has been previously introduced to improve the predictive value of CAD events beyond the whole-heart AgSc. Brown et al22 showed that the percentage of coronary arteries affected by calcific plaque was significantly associated with coronary heart disease events. However, there was no difference in the prediction of hard cardiac events (myocardial infarction or death) when CCS was compared with the whole-heart AgSc and mass scores.
Limitations Limitations of our study include its retrospective design and relatively small sample size. Because we used XRA as reference standard, this also introduced a referral bias. These limitations can be overcome by a prospective design with more patients, in which downstream testing is driven by vessel- and lesion-specific CAC values. In fact, one such prospective study is currently under way at our center.
Conclusion We showed that vessel- and lesion-specific CAC scores are superior in predicting obstructive CAD compared with
the whole-heart AgSc. Such a simple refinement may refine the selection of patients for subsequent testing on the basis of CAC imaging. Because vessel-specific scores are readily available on most commercial workstations, vessel-specific parameters may be taken into consideration for clinical decision making today.
References 1. Budoff MJ, 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. 2. Guerci AD, Spadaro LA, Popma JJ, Goodman KJ, Brundage BH, Budoff M, Lerner G, Vizza RF: Relation of coronary calcium score by electron beam computed tomography to arteriographic findings in asymptomatic and symptomatic adults. Am J Cardiol. 1997;79: 128–33. 3. Guerci AD, Spadaro LA, Goodman KJ, Lledo-Perez A, Newstein D, Lerner G, Arad Y: Comparison of electron beam computed tomography scanning and conventional risk factor assessment for the prediction of angiographic coronary artery disease. J Am Coll Cardiol. 1998;32:673–9. 4. Callister TQ, Cooil B, Raya SP, Lippolis NJ, Russo DJ, Raggi P: Coronary artery disease: improved reproducibility of calcium scoring with an electron-beam CT volumetric method. Radiology. 1998;208:807–14. 5. Agatston AS, Janowitz WR, Hildner FJ, Zusmer NR, Viamonte M Jr., Detrano R: Quantification of coronary artery calcium using ultrafast computed tomography. J Am Coll Cardiol. 1990;15:827–32. 6. Wilson PW: Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97:1837–47. 7. Arad Y, Spadaro LA, Goodman K, Newstein D, Guerci AD: Prediction of coronary events with electron beam computed tomography. J Am Coll Cardiol. 2000;36:1253–60. 8. Becker A, Leber A, White CW, Becker C, Reiser MF, Knez A: Multislice computed tomography for determination of coronary artery disease in a symptomatic patient population. Int J Cardiovasc Imaging. 2007;23:361–7. 9. Fallavollita JA, Brody AS, Bunnell IL, Kumar K, Canty JM Jr.: Fast computed tomography detection of coronary calcification in the diagnosis of coronary artery disease. Comparison with angiography in patients ,50 years old. Circulation. 1994;89:285–90. 10. Greenland P, LaBree L, Azen SP, Doherty TM, Detrano RC: Coronary artery calcium score combined with Framingham score for risk prediction in asymptomatic individuals. JAMA. 2004;291:210–5. 11. Berman DS, Wong ND, Gransar H, Miranda-Peats R, Dahlbeck J, Hayes SW, Friedman JD, Kang X, Polk D, Hachamovitch R, Shaw L, Rozanski A: Relationship between stress-induced myocardial ischemia and atherosclerosis measured by coronary calcium tomography. J Am Coll Cardiol. 2004;44:923–30. 12. He ZX, Hedrick TD, Pratt CM, Verani MS, Aquino V, Roberts R, Mahmarian JJ: Severity of coronary artery calcification by electron beam computed tomography predicts silent myocardial ischemia. Circulation. 2000;101:244–51. 13. Akram K, Voros S: Absolute coronary artery calcium scores are superior to MESA percentile rank in predicting obstructive coronary artery disease. Int J Cardiovasc Imaging. 2008;24:743–9. 14. Bielak LF, Rumberger JA, Sheedy PF, Schwartz RS, Peyser PA: Probabilistic model for prediction of angiographically defined obstructive coronary artery disease using electron beam computed tomography calcium score strata. Circulation. 2000;102:380–5. 15. Detrano R, Hsiai T, Wang S, Puentes G, Fallavollita J, Shields P, Stanford W, Wolfkiel C, Georgiou D, Budoff M, Reed J: Prognostic
Qian et al
16.
17.
18.
19.
Lesion- and vessel-specific CAC scores to diagnose obstructive CAD
value of coronary calcification and angiographic stenoses in patients undergoing coronary angiography. J Am Coll Cardiol. 1996;27: 285–90. Haberl R, Becker A, Leber A, Knez A, Becker C, Lang C, Bruning R, Reiser M, Steinbeck G: Correlation of coronary calcification and angiographically documented stenoses in patients with suspected coronary artery disease: results of 1,764 patients. J Am Coll Cardiol. 2001;37:451–7. Akram K, O’Donnell RE, King S, Superko HR, Agatston A, Voros S: Influence of symptomatic status on the prevalence of obstructive coronary artery disease in patients with zero calcium score. Atherosclerosis. 2009;203:533–7. Qian Z, Marvasty I, Anderson H, Rinehart S, Voros S: Lesion-specific coronary artery calcium quantification better predicts cardiac events. Sixth IEEE International Symposium on Biomedical Imaging. 2009; 237–40. Marcassa C, Bax JJ, Bengel F, Hesse B, Petersen CL, Reyes E, Underwood R: Clinical value, cost-effectiveness, and safety of myocardial perfusion scintigraphy: a position statement. Eur Heart J. 2008;29:557–63.
399
20. Greenland P, Bonow RO, Brundage BH, Budoff MJ, Eisenberg MJ, Grundy SM, Lauer MS, Post WS, Raggi P, Redberg RF, Rodgers GP, Shaw LJ, Taylor AJ, Weintraub WS, Harrington RA, Abrams J, Anderson JL, Bates ER, Grines CL, Hlatky MA, Lichtenberg RC, Lindner JR, Pohost GM, Schofield RS, Shubrooks SJ Jr., Stein JH, Tracy CM, Vogel RA, Wesley DJ: ACCF/AHA 2007 clinical expert consensus document on coronary artery calcium scoring by computed tomography in global cardiovascular risk assessment and in evaluation of patients with chest pain: a report of the American College of Cardiology Foundation Clinical Expert Consensus Task Force (ACCF/AHA Writing Committee to Update the 2000 Expert Consensus Document on Electron Beam Computed Tomography). Circulation. 2007;115: 402–26. 21. Naghavi M, Falk E, Hecht HS, Shah PK: The first SHAPE (Screening for Heart Attack Prevention and Education) guideline. Crit Pathw Cardiol. 2006;5:187–90. 22. Brown ER, Kronmal RA, Bluemke DA, Guerci AD, Carr JJ, Goldin J, Detrano R: Coronary calcium coverage score: determination, correlates, and predictive accuracy in the Multi-Ethnic Study of Atherosclerosis. Radiology. 2008;247:669–75.