Author + information
- Andrew J. Einstein, MD, PhD⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Andrew J. Einstein, Columbia University Medical Center, 622 West 168th Street, PH 10-203A, New York, New York 10032
Can we improve on coronary computed tomography angiography (CCTA)? Numerous multicenter studies and meta-analyses have found good diagnostic performance for CCTA in the identification of significant coronary stenoses, in comparison with a reference standard defined by percent stenosis on invasive angiography. Like its invasive counterpart, however, computed tomography (CT) angiography does not assess per se the physiologic or functional significance of a particular stenosis. Results from the FAME (Fractional Flow Reserve versus Angiography for Multivessel Evaluation) and FAME II trials have demonstrated the added benefit of physiologic information, specifically fractional flow reserve (FFR), above anatomic information alone, in guiding intervention in patients with multivessel coronary disease undergoing intervention (1) and patients with stable coronary disease (2). Complementing the anatomic information obtained by CCTA with physiologic information, which can be used to better identify patients who would or would not benefit from revascularization, offers the potential to make CCTA a more useful tool for the noninvasive evaluation of coronary disease.
Toward this aim, a variety of approaches have been considered to obtain physiologic information from CT. These methods include CT myocardial perfusion imaging; FFR-CT, which estimates FFR from CT data by means of modeling coronary blow flow using computational fluid dynamics techniques; and several metrics deriving from the gradient in contrast opacification along a coronary artery.
Each of these various approaches is still early in its development, and each has advantages and disadvantages. For example, CT perfusion was noted to increase diagnostic accuracy over CT angiography alone in the not yet published 16-center CORE320 study (The Coronary Artery Evaluation using 320-row Multidetector Computed Tomography Angiography and Myocardial Perfusion Study) (3); however, it generally requires at least 1 additional CT scan, with an attendant radiation and contrast burden (4). The 17-center DeFACTO (Determination of Fractional Flow Reserve by Anatomic Computed Tomographic Angiography) study, although not meeting its prespecified primary outcome goal for the level of per-patient diagnostic accuracy, showed FFR-CT to improve diagnostic accuracy and discrimination versus CCTA alone, with decreased FFR as a reference standard (5). Nevertheless, FFR-CT is tremendously computationally complex, requiring 6 hours of time on an off-site parallel supercomputer to numerically solve the Navier–Stokes equations describing fluid flow, and thus is still of limited utility for patients requiring real-time assessment. Quantification of the transluminal contrast attenuation gradient (TAG), while presently with fewer data to support it than CT perfusion or FFR-CT, has particular appeal insofar as it does not require additional radiation or contrast, or lengthy, off-site computation.
The hypothesis motivating TAG as a metric is that contrast opacification should in theory fall off more rapidly in the presence of a functionally significant stenosis than in the absence of stenosis. TAG is determined using the average contrast opacification (measured in Hounsfield units) in a series of evenly spaced regions of interest, drawn perpendicularly to the centerline of a coronary artery. TAG is defined as the regression coefficient of the line fitting the plot of average contrast opacification versus distance from the coronary ostium, and thus reflects the rate of fall-off of contrast opacification along a vessel.
There have been a few previous single-center studies evaluating the TAG. Initial work by Steigner et al. (6) demonstrated that plots of contrast opacification versus distance are generally linear, and that TAG is of greater magnitude in patients with stenotic lesions. This has been followed by 3 papers from Choi, Toon, and colleagues at Seoul National University Hospital (7–9). In the first, Choi et al. (7) found that TAG increased consistently with maximum stenosis severity, in regard to a reference standard of quantitative invasive angiography, and contributed to improved classification of coronary artery stenosis severity, particularly in severely calcified lesions. Two subsequent papers from this group have compared TAG with FFR, using FFR ≤0.80 as a reference standard. In one (8), Choi et al. demonstrated moderate diagnostic performance of TAG alone, with a sensitivity of 47.5%, a specificity of 91.2%, and a significant improvement in the area under the receiver-operating characteristic curve when TAG was added to semiquantitative classification of the degree of stenosis on CCTA, from 0.73 to 0.81. However, addition of TAG to CCTA did not result in a significant reclassification improvement (8). In the other paper (9), Yoon et al. compared TAG with FFR-CT, finding a greater area under the receiver operating characteristic curve for FFR-CT, although this difference was not statistically significant for vessels with calcified plaques.
One limitation of these studies (7–9), pointed out by Choi et al. (8), is that images were obtained using 64-detector row CT scanners. Because of the limited craniocaudal coverage of 64-detector row scanners, the images they obtain of coronary arteries are composites, comprising approximately 4-cm segments, obtained in successive heartbeats and stitched together. As such, opacification along a vessel does not reflect contrast intensity at a single time, potentially limiting the diagnostic performance of the TAG method.
In this issue of the Journal, Wong et al. (10) assess the diagnostic accuracy of TAG in 78 coronary arteries from 54 consecutive evaluable patients undergoing both CCTA and invasive coronary angiography with FFR. TAG was determined from CCTA images obtained on a 320-detector row scanner, which enables isotemporal, single-heartbeat, whole-heart image acquisition, and FFR ≤0.80 was regarded as the reference standard for a functionally significant stenosis. Thirty of the vessels (38%) were classified as functionally significant. The authors noted a bootstrap-resampled sensitivity of 77% and specificity of 74% for TAG alone. Generalized estimating equation analysis identified TAG and visual assessment of the degree of stenosis on CCTA as being independently predictive of a functionally significant stenosis, whereas integrated discrimination improvement index analysis suggested that TAG significantly improves both sensitivity and specificity over degree of stenosis alone. TAG was highly reproducible and outperformed the contrast opacification difference across a stenosis, a simpler metric also deriving from the gradient in contrast opacification (11).
What implications do this study, and its surrounding literature, have for CCTA practice? The study by Wong et al. (10) further supports the potential complementary roles of anatomic and physiologic information obtained from CCTA scans. Its more optimistic findings, in comparison with those in the recent papers from the Seoul group (7–9), suggest that TAG may perform better using a scanner enabling isotemporal coronary image acquisition.
Nevertheless, several questions remain. The study was performed at a single expert center, with a modest sample size, and surely needs to be more broadly validated. Initial evaluations of diagnostic test performance are characteristically better than those achieved in real-world clinical practice, and so it remains to be seen whether TAG still significantly complements the degree of stenosis under the latter conditions. There are a number of reasons for this discrepancy in test performance. Initial evaluations are generally performed with more narrow and distinct spectrums of diseased and nondiseased patients than are encountered clinically, which may falsely elevate measures of diagnostic performance (12). For example, the present study excluded patients with branch vessel disease and distal vessel disease. Wider adoption of and confidence in a diagnostic test leads to patients with negative test results less frequently undergoing subsequent evaluation, which is often more definitive but more invasive. This change is accompanied by a decrease in observed test specificity, because it is these very subsequent tests that are required to establish the reference standard diagnosis. This phenomenon of preferential selection of positive test responders for verification, caused by early validation of a test, has been referred to as post-test referral bias (13). Were TAG to become more established, patients with negative (high magnitude) TAG values would be unlikely to be referred for invasive angiography with FFR, so true negative test results would go unidentified, and the test specificity estimated in practice (true negatives divided by true negatives plus false positives) would decrease. At the same time, false-negative test results would go unidentified, leading to the sensitivity estimated in practice being falsely elevated (14).
Another question remaining to answer is which functional CT test—TAG, FFR-CT, CT perfusion, or some other—is the best choice for functional assessment, and whether this choice varies between patient populations and clinical scenarios. For practical implementation of TAG, more automated image processing software would need to be developed and validated. It would not be difficult to integrate such software with curved multiplanar reconstruction that is now standard on workstations used to read CCTA scans, and thus to routinely determine TAG, at no additional cost, while reading a CCTA scan. For patients potentially benefiting further from the more sophisticated analysis provided by FFR-CT, such as those with stenoses of questionable significance, the option would remain to transfer image data sets for its performance on an outside supercomputer; and for patients for whom stress myocardial perfusion assessment is likely to provide additional beneficial information, stress testing could be added to the CT protocol. Ultimately, benefit should be assessed, as in the FAME trials evaluating FFR, in terms of improvement in patient-important outcomes. At the present time, however, the value of each of these CT technologies and their comparative effectiveness remain unclear. Thus, perhaps we can improve on CCTA with the addition of physiologic data, but the path forward will require considerably more investigation.
Dr. Einstein was supported in part by a Victoria and Esther Aboodi Assistant Professorship, a Herbert Irving Assistant Professorship, and by the Louis V. Gerstner, Jr. Scholars program; and has received research funding from GE Healthcare and Philips Healthcare.
↵⁎ Editorials published in the Journal of the American College of Cardiology reflect the views of the authors and do not necessarily represent the views of JACC or the American College of Cardiology.
- American College of Cardiology Foundation
- ↵CORE320. http://www.escardio.org/congresses/esc-2012/congress-news/Pages/CORE-320.aspx. Accessed January 23, 2013.
- George R.T.,
- Arbab-Zadeh A.,
- Miller J.M.,
- et al.
- Steigner M.L.,
- Mitsouras D.,
- Whitmore A.G.,
- et al.
- Choi J.H.,
- Koo B.K.,
- Yoon Y.E.,
- et al.
- Yoon Y.E.,
- Choi J.H.,
- Kim J.H.,
- et al.
- Wong D.T.L.,
- Ko B.S.,
- Cameron J.D.,
- et al.
- Chow B.J.,
- Kass M.,
- Gagne O.,
- et al.
- Roger V.L.,
- Pellikka P.A.,
- Bell M.R.,
- Chow C.W.,
- Bailey K.R.,
- Seward J.B.