| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Cardiac Imaging |
1 From the Department of Radiology, College of Medicine (W.S., B.H.T., S.D.H., M.C.B.) and Department of Biostatistics, College of Public Health (T.L.B.), University of Iowa, 200 Hawkins Dr, Iowa City, IA 52242. From the 2002 RSNA scientific assembly. Received July 18, 2002; revision requested September 5; final revision received May 26, 2003; accepted July 15. Supported in part by the Garrick Family Foundation, Atherton, Calif. Address correspondence to W.S. (e-mail: william-stanford@uiowa.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Seventy-eight asymptomatic subjects (37 women, 41 men; age range, 39–78 years; mean age, 54.2 years) underwent multi–detector row CT and electron-beam CT. Volume and Agatston scores were calculated with a workstation. Statistical analyses included assessment of association between calcium scores from two scanners, calculation of percent absolute difference to assess score variability between scanners, equivalence analysis, construction of Bland-Altman plots to assess agreement between scores, and assessment of changes in score grouping and risk criteria based on score differences between scanners.
RESULTS: Electron-beam CT calcium scores were higher than multi–detector row CT scores. Linear association between calcium scores obtained from paired scans was significant (r = 0.96–0.99, P < .001). Mean percent absolute differences were 67.9% and 65.0% for volume and Agatston scores, respectively (48.6% and 46.3% for corresponding natural log–transformed scores). In subjects with a score of 11 or greater, mean percent absolute differences between electron-beam CT and multi–detector row CT scores ranged from 15% to 30% (<10% for natural log–transformed calcium scores). With a 20% equivalence limit, calcium scores from the two scanners were statistically equivalent (P < .05). Score grouping would have been subject to change in 12 (11 increased and one decreased; six with scores of 11 or greater), and possible risk management decisions would have been subject to change in eight (16%) of 51 subjects who underwent electron-beam CT versus multi–detector row CT scanning.
CONCLUSION: Multi–detector row CT appears to be comparable to electron-beam CT for coronary calcification screening, except in subjects with a calcium score less than 11.
© RSNA, 2004
Index terms: Computed tomography (CT), comparative studies • Computed tomography (CT), electron beam, 54.12119 • Computed tomography (CT), helical, 54.12115 • Computed tomography (CT), multi–detector row, 54.12119 • Coronary vessels, calcification, 54.812
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Electron-Beam CT Protocol
After the placement of electrocardiographic leads, subjects were asked to lie on a scanner couch pad that included a calcium analysis phantom (Calcium Analysis, Lexington, Ky). With the subject’s arms extended, a scout image was obtained during full inspiration to determine the level of the left main coronary artery. Scanning was initiated approximately 1 cm above the left main coronary artery to include any tortuosity existing in the left anterior descending coronary artery and was continued distally to encompass the entire heart. The scanning was performed by using a unit (Imatron C-150XP; GE Medical Systems, Milwaukee, Wis) at 135 kV and 630 mA, with 100-msec scanning time. Scanning parameters were triggered at 80% of the R-R interval, with 3-mm contiguous section thicknesses acquired during a single breath hold of approximately 20–30 seconds.
Helical CT Protocol
A four-detector-ring multi–detector row CT scanner (Aquilion; Toshiba, Tustin, Calif) and a sequence that followed the manufacturer’s suggested protocol was used for the comparison. Overall, the scanning protocol was similar to that for the electron-beam CT scanner. With the subject’s arms extended, a scout image was obtained to determine the location of the left main coronary artery. The scanning sequence again began approximately 1 cm above the left main coronary artery. A scan was obtained with 135 kV, 300 mA, and 320-msec scanning time, with triggering of all four images at 80% of the R-R interval. Data were acquired by using conventional transverse scanning, with a 12-mm table advancement per each gantry rotation, which provided contiguous 3-mm sections. The sequence required an approximately 30–40-second breath hold to completely interrogate the heart.
Image Analysis
Scanning data were electronically transmitted to a workstation (NetraMD; ScImage, Los Altos, Calif). The transmission of data occurred as the scans were obtained; there was no masking of the data and no sorting of the scans. A base value of 130-HU peak attenuation and a 3-pixel minimum were used as threshold criteria. The technologist (M.C.B.) who obtained the scans identified each highlighted lesion, verified that the lesion was over a coronary artery, and labeled the lesion as to the coronary artery involved. (The workstation identifies each lesion that meets preset criteria and displays it highlighted in yellow.) Once the lesions were identified, the workstation automatically scored the data. Both the volume and the Agatston scores were determined from images that encompassed the entire coronary arterial tree from base to apex. The volume score was the actual volume of the calcium deposits. The Agatston score consisted of an area score multiplied by a factor for calcium attenuation. If the lesion attenuation was 130–200 HU, the area score was multiplied by one; if it was 201–300, it was multiplied by two; if it was 301–400, it was multiplied by three; and if it was greater than 400, it was multiplied by four (5).
Data and Statistical Analysis
Calcium scores and differences in calcium scores from the two scanners are presented as the mean ± SD, median, and range. Both the raw and natural log–transformed calcium volume and Agatston scores were used for statistical analysis. For the natural log–transformed volume and Agatston scores, a value of one was added prior to the transformation to account for scores that were equal to zero. Linear regression analysis was used to characterize the association between scores obtained from the multi–detector row CT and electron-beam CT scanners. Two multiple linear regression models were fitted, where the multi–detector row CT Agatston and volume scores were the dependent variables and the electron-beam CT Agatston and volume scores were the predictor variables, along with age and an indicator variable for sex (men coded as one; women coded as zero). Pearson (parametric) and Spearman (nonparametric) correlation coefficients were used to characterize the association between scores obtained with multi–detector row CT and electron-beam CT scanners. An equivalence analysis was used to formally test the equivalence of calcium scores obtained with the two scanners. The null hypothesis for an equivalence analysis is that the scores are not equivalent; the alternative hypothesis is that they are equivalent (6).
Two equivalence limits were considered: multi–detector row CT calcium scores within 20% of the electron-beam CT calcium scores and multi–detector row CT calcium scores within 10% of the electron-beam CT calcium scores. A significance level of .05 was used for each analysis. The percent absolute difference (PAD) was calculated as follows: PAD = 100 · [(MDCT - EBCT)]/
(MDCT + EBCT), where MDCT is multi–detector row CT score and EBCT is electron-beam CT score. The PAD was calculated to normalize the difference in calcium scores between the two scanners for the amount of calcium for each subject. The mean of the PADs provides a measure of the variability in calcium scores between the two scanners. In addition, Bland-Altman plots (difference in calcium scores between scanners vs mean of calcium scores) were used to visually assess the degree of agreement between the calcium scores obtained from the two scanners. Because the magnitude of the difference increased with increasing mean calcium score, natural log–transformed scores were used in the construction of the plots. Limits of agreement were computed as the mean of the differences plus twice the SD of the differences (7).
To evaluate noise as a factor for differences in calcium scores, 39 of the subjects were evaluated by placing identical regions of interest over the ascending aorta, approximately 2 cm above the aortic valve. The mean attenuation values and SDs for each of the scanners were calculated for these regions. The subject-specific mean values ± SDs were compared between scanners by using a Wilcoxon signed rank test.
Subject Grouping and Treatment Analysis
Five groups of subjects were identified on the basis of their multi–detector row CT Agatston scores: Group 1 had multi–detector row CT Agatston scores less than 11; group 2, scores between 11 and 50; group 3, scores between 51 and 100; group 4, scores between 101 and 400; and group 5, scores greater than 400 (8).
Treatment recommendations for all subjects with a nonzero calcium score from at least one scanner were evaluated on the basis of a change from one group to another, depending on a change in scores between the two scanners. To assess the potential for a cardiovascular event, risk criteria of less than 25%, 25%–50%, 51%–75%, 76%–90%, and greater than 90%, as per Raggi et al (9), were evaluated.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
|
|
Association between Electron-Beam CT and Multi–Detector Row CT Volume and Agatston Scores
The fitted regression models for multi–detector row CT volume and Agatston scores, along with the Pearson and Spearman correlation coefficients (all P < .001), for the 51 subjects in groups 1–5 are presented in Table 1. Age and sex did not add significantly to either model. However, because the parameter estimates associated with the Agatston calcium score were not meaningfully different from a simple model that included only the Agatston calcium score, age and sex were retained in the models. The slopes associated with the electron-beam CT volume and Agatston scores in the two different models are significantly less than one (P < .001). This indicates that while there is a significant linear association between the calcium scores obtained from the two scanners, there is less than perfect agreement and the difference between the calcium scores is not constant across the range of scores.
|
|
|
|
|
Clinical Implications
Treatment recommendations were assessed on the basis of Agatston scores in the 51 subjects with a nonzero Agatston score on at least one scan. Eleven of 51 subjects increased by one group when the electron-beam CT Agatston score was compared with the multi–detector row CT Agatston score, and one subject decreased by one group. Of the 11 subjects who increased by one group, five were initially in group 1. The one individual who decreased by one group was initially in group 4. For the volume score, 11 subjects increased by one group when the electron-beam CT volume score was compared with the multi–detector row CT volume score, and one decreased by one group. Of the subjects who increased by one group, six were initially in group 1. The subject whose score decreased was initially in group 2. Since a score of less than 11 is thought to represent minimal disease, treatment recommendations in this group would primarily be continued emphasis on optimizing diet and lifestyle.
In assessment of potential cardiovascular events with risk criteria of less than 25%, 25%–50%, 51%–75%, 76%–90%, and greater than 90%, as per Raggi et al (9), we found that treatment recommendations that were based on multi–detector row CT scores would be subject to change in eight (16%) of 51 subjects if the subjects were scanned with the electron-beam CT scanner. However, only one of the eight subjects had an initial multi–detector row CT Agatston score greater than 11. The latter individual had an Agatston score of 73 (69% risk), which increased to 139 (78% risk) with electron-beam CT scanning. In the study of Raggi et al (9), risk of greater than 75% was used as an indicator of significant risk. In the remaining seven subjects—all with a score of less than 11—the risk criteria for four would have increased by one with electron-beam CT scanning (highest risk, 62%), and the risk criteria for the other three would have increased by two (highest risk, 71%), with 50% risk as the average for sex and age.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
The most important observations from this investigation are the very strong linear association between the electron-beam CT and the multi–detector row CT calcium scores and the statistical equivalence of calcium scores (untransformed and natural log–transformed volume and Agatston scores) that were based on a 20% equivalence limit (scores that were based on a 10% limit were not statistically equivalent) for the 51 subjects in groups 1–5. An additional important observation is that the greatest variability (169.2%) as assessed with the PAD was seen in subjects with a score of less than 11, with mean variability ranging from 16% to 30% among subjects in groups 2–5 (natural log–transformed variability in groups 2–5 ranged from 2% to 10%).
Studies by Yoon et al (10), Mao et al (11), and Lu et al (12) about the evaluation of reproducibility of scores obtained with electron-beam CT scanners indicated same-subject variances for two scans obtained at the same sitting to be 13%–43%. Multi–detector row CT variability has also been high; Ohnesorge et al (13) reported prospective triggering variability of 11%–16% for the multi–detector row CT scanner. Thus, the variability of both electron-beam CT and multi–detector row CT scanners has not been optimized, and investigators are not yet in agreement as to how best to minimize that variance.
Investigators are also not in agreement regarding the number of contiguous pixels that should constitute a lesion. While there is general agreement concerning a value of 130 HU for electron-beam CT and a 2–3-pixel minimum threshold, the question of the optimal threshold for helical multi–detector row CT scanners is still being debated. The question of whether the Agatston score or the volume score provides the best accuracy and reproducibility has not been decided, and calcium mass may ultimately prove to be the better determinant (3,4). As initially proposed, the Agatston score provides a method to evaluate both the calcific plaque area and the plaque attenuation. Hence, a threshold attenuation of 130 HU was selected as the threshold attenuation of calcium, and from this datum, the area of each plaque was determined and then multiplied by a factor of one to four, depending on the highest attenuation (in Hounsfield units) of the plaque. More recently, investigators have been less concerned with plaque attenuation and have been emphasizing plaque volume and mass. These latter measures tend to have better reproducibility than does the Agatston score, and thus they appear to be more reliable in following up increases or decreases in calcium scores.
Our experience was that triggering at 80% of the R-R interval of the electrocardiogram for both multi–detector row CT and electron-beam CT appeared to give the best results in variability, and therefore that protocol was used in this study. However, the optimum trigger has yet to be agreed on. It may be that a 40% trigger or retrospective triggering where continuous data are collected that can be analyzed at different triggering intervals, depending on the coronary artery being analyzed, may ultimately prove to be the better protocol. It should be emphasized that one of the drawbacks of retrospective gating is substantially increased radiation exposure; however, recent modulation techniques that ramp down the milliampere-second level, except during data collection, have been introduced. These modifications have greatly reduced the radiation exposure and have made this sequence more comparable to prospective gating (14,15).
Noise is also a factor, and since multi–detector row CT has less noise than electron-beam CT, some authors (3,16,17) have advocated using a 90-HU threshold rather than the traditional 130-HU threshold. In reviewing the noise levels in 39 of our subjects, we found significantly (P < .001) increased noise levels, as well as significantly increased variability, for the electron-beam CT scans (mean, 50 ± 28 [SD]) versus those for the multi–detector row CT scans (mean, 40 ± 11). This issue also remains under debate. In spite of these differences, the scanner used in our study appears comparable to the electron-beam CT scanner in scoring calcium in subjects with scores of 11 or greater.
In evaluation of the management of risk, it is important to recognize that individuals with a score less than 11 are believed to be at minimum risk; if calcium scores alone were used for treatment recommendations, most individuals with these low scores would receive only a recommendation for continued dietary and lifestyle optimization (8). Although institutions vary in regard to the importance of coronary calcification in risk assessment, there appears to be general agreement that a score greater than 400 or greater than the 75th percentile for age and sex can help identify individuals at increased risk for a cardiac event.
The major limitations of this study were the relatively small number of subjects and a nonconsecutive population sample. In addition, although we did not advertise in the media, there was word-of-mouth communication and media advertising in markets outside of the local area, and therefore selection bias was possible.
In summary, calcium scores obtained from the subsecond multi–detector row CT scanner appear to have a high correlation with electron-beam CT calcium scores, except in subjects with a score less than 11. In those individuals in whom scores differed between the two scanners, the multi–detector row CT scores were generally lower than the electron-beam CT scores. Overall risk management decisions might have been altered in eight (16%) subjects by using electron-beam CT rather than multi–detector row CT scores.
|
FOOTNOTES |
|---|
Author contributions: Guarantors of integrity of entire study, W.S., T.L.B.; study concepts and design, W.S., T.L.B., B.H.T.; literature research, W.S., B.H.T.; clinical studies, W.S., B.H.T., S.D.H., M.C.B.; data acquisition, M.C.B.; data analysis/interpretation, T.L.B.; statistical analysis, T.L.B.; manuscript preparation and editing, W.S., T.L.B., B.H.T.; manuscript definition of intellectual content and final version approval, W.S.; manuscript revision/review, W.S., T.L.B., B.H.T.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  R. S. Elkeles, I. F. Godsland, M. D. Feher, M. B. Rubens, M. Roughton, F. Nugara, S. E. Humphries, W. Richmond, M. D. Flather, and for the PREDICT Study Group Coronary calcium measurement improves prediction of cardiovascular events in asymptomatic patients with type 2 diabetes: the PREDICT study Eur. Heart J., September 2, 2008; 29(18): 2244 - 2251. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. L. Assimes, J. W. Knowles, A. Basu, C. Iribarren, A. Southwick, H. Tang, D. Absher, J. Li, J. M. Fair, G. D. Rubin, et al. Susceptibility locus for clinical and subclinical coronary artery disease at chromosome 9p21 in the multi-ethnic ADVANCE study Hum. Mol. Genet., August 1, 2008; 17(15): 2320 - 2328. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Onuma, K. Tanabe, G. Nakazawa, J. Aoki, H. Nakajima, K. Ibukuro, and K. Hara Noncardiac Findings in Cardiac Imaging With Multidetector Computed Tomography J. Am. Coll. Cardiol., July 18, 2006; 48(2): 402 - 406. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. A.S. Cordeiro and J. A.C. Lima Atherosclerotic plaque characterization by multidetector row computed tomography angiography. J. Am. Coll. Cardiol., April 18, 2006; 47(8 Suppl): C40 - C47. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. E. Clouse, J. Chen, H. M. Krumholz, M. E. Clouse, J. Chen, and H. M. Krumholz Noninvasive Screening for Coronary Artery Disease With Computed Tomography Is Useful Circulation, January 3, 2006; 113(1): 125 - 146. [Full Text] [PDF]
|
|
|
|
 |
|
 A. L. Daniell, N. D. Wong, J. D. Friedman, N. Ben-Yosef, R. Miranda-Peats, S. W. Hayes, X. Kang, M. G. Sciammarella, L. de Yang, G. Germano, et al. Concordance of Coronary Artery Calcium Estimates Between MDCT and Electron Beam Tomography Am. J. Roentgenol., December 1, 2005; 185(6): 1542 - 1545. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M.-R. Movahed Failure of Gated Single Photon Emission Computer Tomography Scan to Detect Imminent Acute Plaque Rupture Causing Acute ST-Elevation Myocardial Infarction: Case Report Chest, August 1, 2005; 128(2): 1043 - 1047. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Moselewski, C. J. O'Donnell, S. Achenbach, M. Ferencik, J. Massaro, A. Nguyen, R. C. Cury, S. Abbara, I.-K. Jang, T. J. Brady, et al. Calcium Concentration of Individual Coronary Calcified Plaques as Measured by Multidetector Row Computed Tomography Circulation, June 21, 2005; 111(24): 3236 - 3241. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. W. Strauss, M. Dunphy, and N. Tokita Imaging the Vulnerable Plaque: A Scintillating Light at the End of the Tunnel? J. Nucl. Med., July 1, 2004; 45(7): 1106 - 1107. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
