Review. Fractional Flow Reserve and Coronary Computed Tomographic Angiography. A Review and Critical Analysis

Review This Review is in a thematic series on Cardiovascular Imaging, which includes the following articles: T1 Mapping in Characterizing Myocardial Disease: A Comprehensive Review Fractional Flow Reserve and Coronary Computed Tomographic Angiography: A Review and Critical Analysis Prognostic Determinants of Coronary Atherosclerosis in Stable Ischemic Heart Disease: Anatomy, Physiology, or Morphology? Noninvasive Molecular Imaging of Disease Activity in Atherosclerosis Transcathether Valve Replacement and Valve Repair: Review of Procedures and Intraprocedural Echocardiographic Imaging Advances in Echocardiographic Imaging in Heart Failure With Reduced and Preserved Ejection Fraction Viability: Is it Still Attractive? Guest Editors: Jagat Narula and Y. Chandrashekhar Fractional Flow Reserve and Coronary Computed Tomographic Angiography A Review and Critical Analysis Harvey S. Hecht, Jagat Narula, William F. Fearon Abstract: Invasive fractional flow reserve (FFR) is now the gold standard for intervention. Noninvasive functional imaging analyses derived from coronary computed tomographic angiography (CTA) offer alternatives for evaluating lesion-specific ischemia. CT-FFR, CT myocardial perfusion imaging, and transluminal attenuation gradient/corrected contrast opacification have been studied using invasive FFR as the gold standard. CT-FFR has demonstrated significant improvement in specificity and positive predictive value compared with CTA alone for predicting FFR of 0.80, as well as decreasing the frequency of nonobstructive invasive coronary angiography. High-risk plaque characteristics have also been strongly implicated in abnormal FFR. Myocardial computed tomographic perfusion is an alternative method with promising results; it involves more radiation and contrast. Transluminal attenuation gradient/corrected contrast opacification is more controversial and may be more related to vessel diameter than stenosis. Important considerations remain: (1) improvement of CTA quality to decrease unevaluable studies, (2) is the diagnostic accuracy of CT-FFR sufficient? (3) can CT-FFR guide intervention without invasive FFR confirmation? (4) what are the long-term outcomes of CT-FFR guided treatment and how do they compare with other functional imaging-guided paradigms? (5) what degree of stenosis on CTA warrants CT-FFR? (6) how should high-risk plaque be incorporated into treatment decisions? (7) how will CT-FFR influence other functional imaging test utilization, and what will be the effect on the practice of cardiology? (8) will a workstationbased CT-FFR be mandatory? Rapid progress to date suggests that CTA-based lesion-specific ischemia will be the gatekeeper to the cardiac catheterization laboratory and will transform the world of intervention. (Circ Res. 2016;119:300-316. DOI: 10.1161/CIRCRESAHA.116.307914.) Key Words: atherosclerotic plaque cardiac catheterization coronary angiography myocardial perfusion imaging sensitivity and specificity Background: Invasive Fractional Flow Reserve Obstructive coronary artery disease (CAD) produces symptoms and increases the risk of adverse cardiac outcomes. However, determining whether a particular coronary stenosis is responsible for causing myocardial ischemia can be challenging, even with invasive coronary angiography (ICA). To help guide decisions in the cardiac catheterization laboratory, Pijls et al 1 and De Bruyne et al 2 introduced the concept of myocardial fractional flow reserve (FFR) more than 20 years ago. Original received February 24, 2016; revision received May 18, 2016; accepted May 25, 2016. From the Department of Cardiology, Icahn School of Medicine at Mount Sinai, New York (H.S.H., J.N.); Department of Cardiology, Stanford University School of Medicine, CA (W.F.F.). Correspondence to Harvey S. Hecht, MD, Mount Sinai Saint Luke s Medical Center, 1111 Amsterdam Ave, New York, NY 10025. E-mail harvey.hecht@ mountsinai.org 2016 American Heart Association, Inc. Circulation Research is available at http://circres.ahajournals.org DOI: 10.1161/CIRCRESAHA.116.307914 300

Hecht et al Fractional Flow Reserve 301 Nonstandard Abbreviations and Acronyms CTA CTP FAME FFR FFRCT MPI MRI NPV PCI PPV coronary computed tomographic angiography computed tomographic stress myocardial Fractional Flow Reserve Versus Angiography for Multivessel Evaluation fractional flow reserve fractional flow reserve form computed tomography myocardial perfusion imaging magnetic resonance imaging negative predictive value percutaneous coronary intervention positive predictive value FFR is defined as the ratio between the maximum myocardial blood flow in the presence of an epicardial coronary stenosis and the maximum myocardial blood flow in the theoretical absence of the stenosis. 1 Briefly, when both epicardial and myocardial resistances are minimized by administering vasodilators and assuming microvascular resistance is similar in the presence and absence of an epicardial stenosis, coronary flow becomes proportional to coronary pressure. Because there is little pressure loss along a normal epicardial coronary artery, the proximal coronary pressure of a diseased artery is a reflection of what the distal pressure would be in the absence of disease. Hence, FFR can be defined as the distal coronary pressure divided by the proximal coronary pressure during maximal hyperemia. Traditionally, FFR is determined invasively in the cardiac catheterization laboratory with a coronary pressure wire in the vessel measuring distal pressure and a guiding catheter measuring the proximal pressure during hyperemia induced with adenosine. FFR has several unique attributes that separate it from other indices, such as coronary flow reserve. 3 First, it has an unequivocal normal value of 1.0 in every patient and every vessel. Second, it has a well-defined ischemic cutpoint of 0.80 with a gray zone from 0.75 to 0.80. Third, because FFR is measured during maximal hyperemia, it is not affected by changes in heart rate and blood pressure and is extremely reproducible. Fourth, FFR incorporates the contribution of collateral blood supply because distal pressure takes into account both antegrade flow and collateral flow. Noninvasive assessments of FFR do not account for collateral flow in the same manner. Finally, FFR specifically interrogates the epicardial vessel independent of the microvasculature, assuming the microvascular dysfunction if present is fixed and chronic. For example, in the setting of an old myocardial infarction, a residual stenosis, despite its angiographic appearance may have little pressure gradient and a nonischemic FFR during maximal hyperemia because it supplies nonviable myocardium. In this setting, FFR remains accurate and informs the clinician that there is no benefit in relieving the stenosis because it is not responsible for inadequate myocardial blood flow. However, in the setting of an acute myocardial infarction, in which case transient and reversible degrees of microvascular damage occur, FFR may not be reliable. FFR was first validated by measuring it in 45 patients with intermediate single-vessel coronary disease and comparing the FFR value to the result from 3 different noninvasive stress tests. 4 Because there is no true gold standard for identifying myocardial ischemia, if any one of the noninvasive tests was abnormal, then the lesion was felt to be capable of producing myocardial ischemia. Because each test has an accuracy of roughly 80%, applying sequential Bayesian analysis, the combination of all 3 tests provides an accuracy of >95%. In this study, the investigators found at a cutoff value of 0.75, FFR had 100% specificity, 88% sensitivity, and an accuracy of 93%. To improve the sensitivity of FFR, the more recent clinical trials have used an FFR cutoff value of 0.80. Several multicenter, randomized clinical trials have provided further validation of FFR as an index, which can predict clinical outcome. The DEFER trial established the safety of deferring percutaneous coronary intervention (PCI) on coronary stenoses, which were not functionally significant based on FFR. 5 Patients with intermediate lesions and FFR value of 0.75 were randomized to performance of PCI anyway or to deferral of intervention and continued medical therapy. The study demonstrated the safety of medical therapy for functionally nonsignificant lesions because the event-free survival rate at 2 years was similar between the 2 groups (89% versus 83% for the deferral versus performance groups, respectively; P=0.27), as was the degree of angina. This cohort has now been followed for 15 years with no significant increase in revascularization rate in the deferral patients and a significantly lower rate of myocardial infarction (2.2 versus 10%, P=0.03). 6 Other observational registries have confirmed the safety of deferring revascularization of coronary disease, which is not significant based on FFR, even if involving the proximal left anterior descending artery 7 or the left main coronary. 8 The Fractional Flow Reserve Versus Angiography for Multivessel Evaluation (FAME) trial was a large, multicenter, randomized study that further confirmed the safety of deferring PCI of lesions that are not hemodynamically significant, but more importantly demonstrated the safety, cost-effectiveness, and clinical benefit of routinely measuring FFR in patients with multivessel CAD undergoing PCI. 9 The FAME study included 1005 patients with 2- or 3-vessel CAD eligible for PCI and randomized them to either angiography-guided PCI, in which case the operator performed PCI based on the angiographic appearance of the CAD, or FFR-guided PCI, in which case only those lesions with an FFR of 0.80 underwent PCI. The FFRguided patients received significantly fewer stents; yet, they had similar relief of angina and a significantly lower rate of the primary end point: death, myocardial infarction, or repeat revascularization at 1 year (13.2 versus 18.3, P=0.02). The rate of death or myocardial infarction was also significantly lower at both 1 and 2 years. 10 Most of this benefit was because of better assessment of stenoses between 50% and 90% narrowed because the vast majority of those stenoses >90% narrowed had an ischemic FFR value. A cost-effectiveness study showed that the FFR-guided strategy was a cost-saving approach, leading to a lower event rate and utilizing fewer resources. 11 By 5 years, there was a durable benefit from FFR-guided PCI. 12 The FAME trial showed that functional angioplasty, performing PCI on lesions responsible for ischemia, while treating lesions medically that are not hemodynamically significant results in

302 Circulation Research July 8, 2016 excellent clinical outcomes by maximizing the benefit of PCI while minimizing its risks. One of the criticisms of the FAME trial was that it did not include a medical therapy arm. Some argued that medical therapy alone would have performed as equally well as FFRguided PCI. The goal of the FAME 2 trial was to compare outcomes in patients who had stable CAD, which was functionally significant based on FFR, and who were randomized to either guideline-directed medical therapy alone or to PCI and best medical therapy. 13 In this manner, FAME 2 was the mirror image of the DEFER trial. The key difference between the FAME 2 trial and previous studies comparing medical therapy to PCI in stable patients was that FFR was measured first, so only those patients with significant myocardial ischemia were randomized. Patients with CAD but no functionally significant lesions received guideline-directed medical therapy alone and formed a reference group. Enrollment in the FAME 2 trial was stopped early by the data safety monitoring board after a mean follow-up of 7 months because of a highly significant increase in the primary end point of death, myocardial infarction, or the need for hospitalization and urgent revascularization in the patients assigned to medical therapy alone (12.7% versus 4.3%, P<0.001). This was driven primarily by an increase in the need for urgent revascularization in the medical therapy patients. The death and myocardial infarction rate was numerically higher, but not statistically significant in the best medical therapy alone patients (3.9% versus 3.4%, P=0.22). The complete 2-year follow-up showed similar findings with a significantly higher rate of the primary end point in the medically treated patients (19.5% versus 8.1%, P<0.001). 14 A landmark analysis evaluating the rate of death or myocardial infarction occurring 7 days after enrollment (eliminating the periprocedural myocardial infarctions occurring in the PCI group) demonstrated a significantly higher rate of death or myocardial infarction in the medically treated patients (8.0% versus 4.6%, P=0.04). Importantly, in the patients in the reference group who had stable CAD that was not functionally significant based on FFR and who received medical treatment alone, the event rate was low and similar to the PCI group of patients. A cost-effectiveness analysis from FAME 2 found that FFR-guided PCI was attractive at $36 000 per quality-adjusted life years. 15 Historically, FFR use in the United States has been low; however, as a result of the above data and an increased emphasis on appropriate application of PCI, the percentage of cases involving FFR has been increasing on a yearly basis and is now 20%. 16 Because of this increased usage of FFR, there is interest in streamlining the measurement. One proposed method is to eliminate the need for hyperemia by measuring only resting gradients, such as the resting ratio of distal pressure to proximal pressure or the instantaneous wave-free ratio that is the ratio of resting distal pressure to proximal pressure during the so-called diastolic wave free period. Unfortunately, multiple large, multicenter studies have shown that these indices are only 80% accurate when compared with FFR. 17,18 The FAME 2 trial highlighted an important concept that the greater the burden of ischemia-producing CAD, the higher the event rate with medical therapy alone. Other studies have also found a relationship between the FFR value across the lesion and the risk of adverse events with medical therapy alone, which is attenuated by revascularization, 19 even when focusing only on the gray zone of FFR values. 20 These data emphasize the need to focus our noninvasive evaluation of CAD on both anatomy and ischemia. One of the limitations of the traditional method for measuring FFR is that it requires an invasive procedure. Recently, there has been great interest in developing a noninvasive method for assessing FFR by using the anatomic data derived from coronary computed tomographic angiography (CTA). Noninvasive FFR FFR derived from CTA (CT-FFR) has the potential to be a game changer in the noninvasive assessment of the significance of coronary artery stenoses by providing the proverbial one stop shop of anatomy and function in a single low-dose scan. 21 The notion of noninvasive measurement of CT-FFR seems, on the surface, to be ludicrous. How could FFR be derived noninvasively from a resting CTA, without the administration of a vasodilator? The application of computational fluid dynamics to the CTA data set has allayed these fears and initiated a series of studies, which may fundamentally transform the cardiac catheterization laboratory into a truly interventional rather than diagnostic facility. Why We Need CT-FFR The National Cardiovascular Data Registry highlighted the need for a more reliable gatekeeper to the cardiac catheterization laboratory by evaluating the percentage of 661 063 patients undergoing elective catheterization who had >50% stenosis. 22 Preprocedural noninvasive testing was performed in 64% of patients; of those, 51.9% were abnormal. The percentages of patients correctly identified as having >50% stenosis ranged from 44% to 45% for exercise treadmill testing, stress echocardiography, stress myocardial perfusion imaging (MPI), and stress magnetic resonance imaging (MRI); for resting CTA, the percentage was 70%. The inadequacy of the commonly used functional tests may be explained by their validation paradox: exercise treadmill testing, single-photon emission computed tomographic MPI, and stress echocardiography have used >50% stenosis on ICA as the gold standard, on which their sensitivities and specificities were validated. Unfortunately, >50% stenosis is a severely flawed gold standard, correlating well neither with intravascular ultrasound nor with invasive FFR, which is validated by the gold standard of outcomes. Functional imaging is even more problematic because it uses the test to judge the results of the gold standard by which it is validated. For instance, a normal stress echocardiogram in a patient with an 80% proximal left anterior descending stenosis is by definition a false-negative finding. However, the overwhelmingly more likely interpretation in the noninvasive community would be a nonischemic stenosis, on the erroneous assumption that the stress test is the gold standard. The National Cardiovascular Data Registry suggests that anatomy trumps function for the identification of patients with >50% stenosis, a finding confirmed by numerous nonrandomized studies, and 3 prospective randomized controlled trials. In the Prospective Multicenter Imaging Study for Evaluation of

Hecht et al Fractional Flow Reserve 303 Chest Pain (PROMISE) trial, 23 10 000 symptomatic patients were randomized to CTA or functional testing (MPI, 67.5%; stress echocardiography, 22.4%; and treadmill testing, 10.2%). The percentage of patients undergoing catheterization with <50% stenosis was 3.4% for CTA versus 4.3% for functional testing (P=0.02). The Coronary CT Angiography in Patients With Recent Acute-Onset Chest Pain (CATCH) trial 24 randomized 576 patients with chest pain to CTA versus usual care (treadmill testing, 76% and MPI, 22%). In those undergoing catheterization, 29% of the CTA arm and 64% of the functional arm had <50% stenosis (P=0.002). In the Cardiac CT for the Assessment of Chest Pain and Plaque (CAPP) trial 25 of 488 patients with chest pain randomized to CTA or treadmill testing, 30% of the CTA arm and 65% of the functional arm that had catheterization had <50% stenosis. In the Scottish Computed Tomography of the HEART (SCOT-HEART) trial 26 of 4146 subjects randomized to CTA versus standard care, similar numbers of patients underwent invasive angiography, but the percentage with normal coronary arteries was 4.8% for CTA versus 14.0% for usual care (P<0.001), and the percentage with >50% stenosis was 69.2% for CTA versus 57.3% for usual care (P=0.005). From the prognostic perspective, in the Clinical Outcomes Utilizing Revascularization and Aggressive Drug Evaluation (COURAGE) trial of patients randomized to intervention plus optimal medical therapy versus optimal medical therapy alone, 621 patients had MPI. 27 The odds ratio for prediction of acute events was 1.01 (95% confidence interval, 0.98 1.03; P=0.54) for ischemic burden versus 1.05 (95% confidence interval, 1.02 1.08; P=0.002) for atherosclerotic burden by quantitative coronary angiography. In addition to demonstrating the superiority of CTA over functional testing in identifying >50% diameter stenosis (DS), the randomized controlled trials are suggestive of the beneficial effect on prognostic outcomes of CTA versus functional testing guided management. The PROMISE trial 23 revealed a better outcome at 12 months for CTA versus functional-guided treatment (hazard ratio, 0.66; P=0.049) for death plus MI, although there was no difference at 36 months. In the SCOT-HEART trial, 26 there was a 50% reduction (hazard ratio, 0.50; 95% confidence interval, 0.28 0.88; P=0.02) in fatal and nonfatal myocardial infarction in the CTA-guided group. The Long-Term Clinical Impact of CATCH trial 24 also noted a 38% reduction in the primary end point of cardiac death, myocardial infarction, unstable angina, late revascularization, and chest pain readmissions (11% versus 16%, P=0.04). Thus, the anatomic superiority of CTA to functional testing in identifying obstructive disease and producing better outcomes in relatively short-term randomized trials, plus the prognostic superiority of anatomy to ischemia in the Courage trial, provides support for the increasing use of CTA. However, the ascendancy of functional invasive FFR as the gold standard, with outcome-based superiority of FFR-guided intervention to anatomy-guided intervention, would seem discordant with the above discussion. In reality, there may be no contradiction because FFR is not simply another functional test that infers ischemia by stress-induced perfusion defects or wall motion abnormalities. Rather, as discussed in detail above, it is a direct measure of flow inadequacy. A recent meta-analysis (Table 1) highlighted the inferior accuracy of MPI and stress echocardiography to CTA alone for identifying invasive FFR of 0.80 and may explain the better outcomes discussed above for CTA-guided treatment. 28 In addition, in a study of 67 patients with multivessel disease, MPI was particularly suboptimal, with 61% sensitivity, 69% specificity, 47% positive predictive value (PPV), and 80% negative predictive value (NPV). 29 Interestingly, stress MRI and positron emission tomography (PET) performed at least as well as CTA alone, but these technologies are expensive and are not widely used. Whether the ability of CTA to improve outcomes by identifying abnormal invasive FFR will be further enhanced by adding FFR capabilities, paralleling the improved outcomes after invasive FFR-guided treatment, is yet to be determined and is the focus of this article. Computed Tomography Fractional Flow Reserve Accuracy Fractional flow reserve form computed tomography (FFRCT; HeartFlow, Inc, Redwood City, CA) is the first noninvasive functional test to use an outcome-based gold standard, that is, invasive FFR, rather than one fundamentally based on coronary stenosis measurement. Computational fluid dynamics are used for the calculation of FFR at any point in the vascular tree, based on projected adenosine-induced vasodilation, from a CTA of at least moderate quality acquired at rest without adenosine infusion; the step by step methodology and clinical examples are presented in Figure 1. 30 Table 1. Diagnostic Performance of the Noninvasive Imaging Modalities Compared With Invasive Fractional Flow Reserve Gold Standard Modality n Sensitivity Specificity PLR NLR AUC Pt/V Pt, % V, % Pt, % V Pt V Pt V P V SPECT 533/924 74 81 79 0.84 3.13 3.76 0.39 0.47 0.82 0.83 Echo 177/NA 69 NA 84 NA 3.68 NA 0.42 NA 0.83 NA MRI 798/1830 89 83 84 89% 6.29 8.27 0.14 0.16 0.94 0.95 PET 224/870 84 83 87 89% 6.53 7.43 0.14 0.15 0.93 0.95 CTA 316/1074 88 78 80 86% 3.79 5.74 0.12 0.22 0.93 0.91 AUC indicates area under receiver operator characteristic curve; CTA, coronary computed tomographic angiography; Echo, stress echocardiography; MRI, stress magnetic resonance imaging; NA, not available; NLR, negative likelihood ratio; PET, positron emission tomography; PLR, positive likelihood ratio; Pt, per-patient; SPECT, single-photon emission computed tomographic myocardial perfusion imaging; and V, per-vessel.

304 Circulation Research July 8, 2016 Figure 1. Top, Step by step method for calculation of fractional flow reserve form computed tomography (FFRCT). A, Acquisition of coronary computed tomographic angiography (CTA); (B) coronary artery segmentation to second- and third-order vessels; and (C) application of subvoxel resolution techniques. In this example, a cross-section of a coronary artery shown with image intensity data (left) and image-gradient data (right) illustrates typical coronary CTA reconstruction with increasingly improved image resolution (middle and bottom) demonstrating subvoxel resolution techniques. D, Discretization of mesh elements for calculation of computational fluid dynamics at millions of points in the coronary vascular bed. Note that the tetrahedral vertices are reconstructed in 3 dimensions (3D) and are continuous even at the branch points to accurately calculate coronary computed tomographic angiography derived FFR (FFRCT) at these areas commonly affected by plaque. Reduced order methods that do not use 3D analyses are less accurate at these points. E, Relationship of the location and size of coronary arteries to the left ventricular mass they subtend; (F) relationship of coronary vessel caliber and flow and resistance; (G) demonstration of reduced coronary resistance index at an adenosine dose of 140 mg/kg per min (McKavanagh et al 25 ); (H) Navier Stokes equations that govern the fluid dynamics of blood (nonlinear partial differential equations related to mass conservation and momentum balance are solved); and (I) example of a patient-specific FFRCT. Bottom, Clinical examples of FFRCT. J, Coronary CTA demonstrating severe calcifications of the left main, left anterior descending (LAD), and left circumflex arteries (left). FFRCT model demonstrating coronary ischemia of the LAD with a value of 0.74 (middle). Invasive coronary angiography (ICA) demonstrating intermediate stenosis severity that causes ischemia according to FFR with a value of 0.71 (right). K, Coronary CTA demonstrating diffuse coronary artery disease (CAD) from multiple tandem lesions in the LAD (left). FFRCT demonstrating coronary ischemia of the LAD with a value of 0.70 (middle). ICA demonstrating diffuse luminal irregularities that result in ischemia according to FFR with a value of 0.74 (right). L, Coronary CTA demonstrating an intermediate stenosis of the LAD (left). FFRCT demonstrating coronary ischemia of the LAD with a value of 0.78 (middle). ICA demonstrating intermediate stenosis severity that causes ischemia according to FFR with a value of 0.78 (right). M, Coronary CTA demonstrating a small left main and LAD with mild diffuse calcific plaque (left). Despite the absence of high-grade stenoses, ischemia is noted by using FFRCT with a value of 0.75 (middle). ICA demonstrating similar concordance with a value of 0.81 (right). N, Example of virtual stenting. In the first pane, an FFRCT model demonstrates coronary ischemia of the obtuse marginal branch of the left circumflex artery with a value of 0.61. In the second pane, ICA and invasive FFR confirm a high-grade stenosis and ischemia with an FFR value of 0.52. In the third pane, a virtual stent is placed with predicted resolution of the ischemia by FFRCT with a value of 0.83. In the fourth pane, a postpercutaneous coronary intervention FFR demonstrates resolution of ischemia with a value of 0.88. QCA indicates quantitative coronary angiography. Reprinted from Min et al 30 with permission of the publisher. Copyright 2015, Elsevier. CCTA indicates cardiac computed tomographic angiography; and MBF, myocardial blood flow.

Hecht et al Fractional Flow Reserve 305 Figure 1 Continued.

306 Circulation Research July 8, 2016 The brief history of CT-FFR trials is summarized in Tables 2 and 3 and reveals somewhat discordant results for the 2 multicenter FFR CT trials, attributed to technical issues as will be discussed. Because of these differences, pooled values are not presented. The initial report (DISCOVER-FLOW [Diagnosis of Ischemia-Causing Stenoses Obtained Via Noninvasive Fractional Flow Reserve]) of 159 vessels in 103 patients from multiple centers undergoing FFRCT and ICA with FFR yielded remarkable improvements in specificity (25% 82%) and PPV (58% 85%) in both patients and vessels compared with CTA alone, with invasive FFR as the gold standard, without differences in sensitivity and NPV. 31 The study met its primary end point to detect a relative improvement in diagnostic accuracy of 25% for FFRCT when compared with CTA stenosis. A larger multicenter study (DeFACTO [Determination of Fractional Flow Reserve by Anatomic Computed Tomographic Angiography]) of 615 vessels in 252 patients yielded different results. 32 The sensitivity and NPV were similar for CTA and FFRCT, as in DISCOVER-FLOW, but there were no significant improvements in specificity (42% 54%) and PPV (61% 67%). A subgroup analysis restricted to 150 vessels in 82 patients with intermediate stenoses of 30% to 69% revealed improved sensitivity (34% 72%) and NPV (78% 90%) compared with CTA. 33 There was no improvement in specificity (72% 67%) and the PPV, while improved, (27% 41%) was unacceptably low. The study did not achieve its prespecified primary outcome goal for the level of perpatient diagnostic accuracy of >70% of the lower bound of the 95% confidence interval. Pursuant to the disappointing DeFACTO results, the investigators addressed the perceived technical and methodologic shortcomings. Better lumen boundary identification was achieved by improved automated image processing methods. Physiological models of microcirculatory resistance that yielded better diagnostic performance than the previous studies were used. In particular, more aggressive rate control Table 2. Diagnostic Performance of CT-FFR and CTA Compared With Invasive FFR Gold Standard and universal nitroglycerin administration were mandated to strictly adhere to the best practices for image acquisition. The results of the third multicenter study (NXT [Analysis of Coronary Blood Flow Using CT Angiography: Next Steps]) results reflect the remedies. 34 In 484 vessels in 254 patients, FFRCT significantly improved the per-patient specificity and PPV (32% 84% and 40% 65%, respectively) and the per-vessel specificity and PPV (60% 86% and 33% 67%, respectively) compared with CTA. There were no significant changes in sensitivity and NPV. Similarly, the area under the receiver-operating characteristic curve increased from 0.81 to 0.90 (P=0.0008) for the per-patient analysis (Table 2). In the 235 patients with intermediate stenoses in the 30% to 70% range, the per-patient specificity and PPV improved from 32% to 79% and 37% to 63%, respectively, compared with CTA. The study met its primary end point per-patient diagnostic performance as assessed by the area under the receiver-operating characteristic curve of FFRCT ( 0.80) versus CTA (stenosis>50%) for the diagnosis of hemodynamically significant stenosis (FFR 0.80) in patients with CTA stenosis of 30% to 90%. The sensitivity, specificity, and accuracy of FFRCT persisted across the quartiles of coronary artery calcium score, as did its superior specificity and accuracy compared with CTA. 35 Inclusion of the 12% of studies deemed nonanalyzable because of calcium blooming, and other artifacts might have decreased the accuracy. The ability of FFRCT to predict a nonischemic invasive FFR response to stenting was evaluated by recalculating the FFRCT after projected improvement of the stenosis by percutaneous intervention. In 44 lesions in 44 patients, they reported 100% sensitivity, 96% specificity, 50% PPV, and 100% NPV. 33 In addition to the prospective, multicenter FFRCT reports, there are several studies that use local workstation-based computational fluid dynamics algorithms. In a retrospective single-center study by Renker et al, 37 with <60 minutes analysis time and invasive FFR of <0.80 as the gold standard, 67 Study/y n n Modality Sensitivity Specificity PPV NPV Design Screened Pt/V Pt, % V, % Pt, % V, % Pt, % V, % Pt, % V, % DISCOVER-FLOW 2011 Prospective Single center 31 NA 103/159 DeFACTO 2012 Prospective Multicenter 32 285 252/615 NXT 2014 Prospective Multicenter 34 357 254/484 Renker et al 37 Retrospective Single center Kruk et al 38 Prospective Single center 61 53/67 CTA 94 91 25 40 58 47 80 89 FFRCT 93 88 82 82 85 74 91 92 CTA 84 NA 42 NA 61 NA 72 NA FFRCT 90 83 54 78 67 NA 84 NA CTA 94 83 34 60 40 33 92 92 FFRCT 86 84 79 86 65 67 93 95 CTA 94 90 32 34 38 37 92 89 CT-based FFR 94 85 84 85 71 71 97 93 CTA 100 100 2 2 46 43 100 100 NA 90/96 CT-based FFR 76 76 71 72 69 67 78 80 CTA indicates coronary computed tomographic angiography; FFRCT, fractional flow reserve from computed tomography; NA, not available; NPV, negative predictive value; PPV, positive predictive value; Pt, per-patient; and V, per-vessel.

Hecht et al Fractional Flow Reserve 307 lesions were retrospectively evaluated in 53 patients. They reported 85% sensitivity and specificity, PPV of 71%, and NPV of 93% on a per-lesion basis, with the area under the receiver-operating characteristic curve of 0.92 compared with 0.72 (P=0.0049) for CTA alone. On a per-patient basis, the area under the receiver-operating characteristic curve was of borderline superiority to CTA alone (0.91 versus 0.78, P=0.078), with 94% sensitivity, 84% specificity, 71% PPV, and 97% NPV. In a prospective single-center workstation-based study with a 20-minute processing time, Kruk et al 38 analyzed 96 lesions in 90 patients with invasive FFR of <0.80 as the gold standard. They reported 76% sensitivity, 72% specificity, 67% PPV, and 80% NPV compared with 100% sensitivity, 2% specificity, 43% PPV, and 100% NPV for CTA alone. Important CT-FFR Considerations CTA Quality The importance of strict adherence to quality guidelines cannot be overemphasized because the ability to calculate CT-FFR is dependent on the quality of the study. In the 2 multicenter trials, 11% and 13% of the CTA studies were unsuitable for technical reasons. 31 34 Because scanner temporal and spatial resolution continue to improve and strict adherence to quality guidelines increases, motion and calcium blooming artifacts should be less problematic and the percentage of analyzable studies should increase, although this remains to be demonstrated. Unfortunately, there will always be nondiagnostic studies, as there are for all imaging modalities. Test Accuracy Are the CT-FFR sensitivity and specificity, which characterize the accuracy of the test compared with the invasive FFR gold standard, sufficiently robust? The NXT sensitivity and specificity were 86% and 79% for CT-FFR versus 94% and 34% for CTA alone. 34 The benefits clearly reside in the reduced false positives compared with CTA alone, with the concomitant reduction in unnecessary ICA. In this context, evaluation of FFRCT as a gatekeeper to ICA will be examined by Coronary Computed Tomographic Angiography for Selective Cardiac Catheterization (CONSERVE). 30 However, the 79% specificity is still suboptimal, and the lower sensitivity is of concern. The PPV and NPV, which are dependent on the prevalence of ischemic lesions in the specific patient population and cannot be easily extrapolated to the community at large, were 65% and 93% for CT-FFR versus 40% and 92% for CTA alone. 34 In the NXT population, therefore, there was no benefit compared with CTA alone for identifying those patients whose invasive FFR will warrant intervention, and 35% of the abnormal results were false positives. In addition, the discrepancies between the specificities of CTA for invasive FFR of 0.80 in the FFRCT trials (patients 25% 42% and vessels 40% 60; Table 3) 31 34 and those reported in a meta-analysis (patients 80% and vessels 86%; Table 2) 28 are striking and require explanation. Although head to head comparisons in the same group of patients are more compelling than meta-analyses, and even assuming that there may have been a disproportionate number of patients with extensive calcification and, therefore, more false positives in the FFRCT trials, the discrepancies need to be further elucidated. Table 3. Areas Under the Receiver Operator Characteristic Curve for CTA and CT-FFR per-patient Analysis Compared With Invasive FFR Study AUC CTA FFRCT P Value DISCOVER-FLOW 31 0.75 0.90 0.001 DeFACTO 32 0.68 0.81 <0.001 NXT 34 0.81 0.90 <0.0008 CT-based FFR Renker et al 37 0.72 0.92 0.0049 Kruk et al 38 * 0.66 0.84 <0.01 AUC indicates area under receiver operator characteristic curve; CT-FFR, fractional flow reserve form computed tomography; CTA, coronary computed tomographic angiography; DeFACTO, Determination of Fractional Flow Reserve by Anatomic Computed Tomographic Angiography; DISCOVER-FLOW, Diagnosis of Ischemia-Causing Stenoses Obtained Via Noninvasive Fractional Flow Reserve; and NXT, Analysis of Coronary Blood Flow Using CT Angiography. *Per lesion analysis; all other are per-patient. Moreover, the CTA specificities are considerably inferior to the meta-analysis reported specificities for single-photon emission computed tomographic MPI and stress echocardiography, 28 a surprising result. Nonetheless, CT-FFR implementation will be warranted if it proves to be superior to other functional tests in head to head comparisons in studies evaluating the triage of patients to ICA. Future trials to address this critical issue are underway or are being planned. Computed Tomographic Evaluation of Atherosclerotic Determinants of Myocardial Ischemia (CREDENCE) 30 will be a direct comparison of CTA plus FFRCT against MPI and positron emission tomography. PERFECTION will pit FFRCT against single-energy CT rest/ stress perfusion imaging 30 and Dual Energy CT for Ischemia Determination Compared to Gold Standard Non-Invasive and Invasive Technique (DECIDE-Gold) will compare FFRCT and dual-energy CT rest/stress perfusion imaging. 30 However, these are trials of diagnostic performance, rather than patient outcomes. CT-FFR, Decision Making, and Outcomes Application of population-based accuracy to the individual is far from straightforward, and decision making is always easiest at the extremes. For instance, it is a simple choice to proceed to ICA with a CT-FFR of 0.65 and to defer ICA with a value of 0.90. Intermediate values present more difficult choices and closer to 0.80, the more likely is a false negative or false positive compared with invasive FFR. There is no obvious remedy for the uncertainty of the middle ground other than to temper expectations of the agreement with invasive FFR in this range. However, there are more complicated issues than simple triage to ICA. 1. Below what CT-FFR value will invasive FFR confirmation be unnecessary before proceeding to intervention, thereby resulting in cost savings incremental to those associated with avoiding unnecessary ICA? Current practice requires confirmation of CT-FFR by invasive FFR because CT-FFR is not yet guideline recommended, and

308 Circulation Research July 8, 2016 there are no ongoing or proposed trials to evaluate the outcomes of proceeding to intervention based on CT-FFR without invasive FFR confirmation. In the meantime, FFRCT has been approved by the FDA and is being used for triage to the catheterization laboratory without longterm trials demonstrating improved outcomes with this approach. Interestingly, intervention is already deemed appropriate after abnormal MPI and stress echocardiography without requiring confirmation by invasive FFR, with which their correlation is suboptimal. Thus, logic would dictate that abnormal CT-FFR, which correlates well with invasive FFR as discussed above, should also suffice for intervention. However, as noted, it is not yet the standard of care, and should be discouraged until the requisite outcome data are available. b. Although CTA has been shown to increase the revascularization rate compared with usual care, 24,25 the incremental role of CT-FFR on the number of revascularizations is unknown, and there are no outcome studies to determine the beneficial effect of CT-FFR guided revascularization. The prognostic significance of FFRCT will be evaluated by Assessing Diagnostic Value of Non-Invasive FFRCT in Coronary Care (ADVANCE) 30 in a prospective longitudinal registry. 2. There are short-term outcome, 39 quality of life, and resource utilization data. 40 In the PLATFORM study of 380 patients for whom ICA was planned but were instead randomized to usual care versus FFRCT-guided care, the major adverse coronary event rate at 90 days was 0% versus 1%, a nonsignificant difference. 39 There were also no differences in event rate at 90 days for the 204 patients for whom a noninvasive testing strategy was planned but were randomized to usual versus FFRCTguided care; there were no events in either group. Significant differences were noted in the percentage of patients with <50% stenosis on ICA (the primary end point), in the planned invasive arm for the FFRCT (12%) and usual care (73%) groups, and ICA was cancelled in 61% of the FFRCT-guided care group. In the same patient cohort of 584 patients, costs and quality of life issues were addressed. 40 With projected charges for the FFRCT ranging from none to 7 the cost of the CTA (there is currently no insurance reimbursement), 90- day costs were 32% lower in the FFRCT patients in the planned invasive group than in the usual care patients ($7343 versus $10 734 P<0.0001) at no charge for the FFRCT, with no differences in the planned noninvasive group. The lower cost persisted with FFRCT charges 7 the CTA ($8619 versus $ 10 734; P<0.0001). Three quality of life scores improved more in the FFRCT arm of the planned invasive strategy without differences in the noninvasive cohort. CT-FFR Lesion Candidates Which CTA stenoses should be evaluated with CT-FFR? CTA uses the traditional >50% DS cutoff for a significant lesion, on which its sensitivity and specificity using ICA as the gold standard to ICA are determined. Patients with stenoses <50% are traditionally not referred for further evaluation. Unfortunately, the percent DS does not correlate well with FFR. In FAME, 9,41 only one third of 50% to 70% lesions demonstrated FFR of 0.80 and one fifth of 71% to 90% stenoses were associated with FFR of >0.80. In another report, 42 FFR of >0.80 was noted in more than half of >50% stenoses and 1 in 7 to lesions <50% Figure 2. Influence of high-risk plaque on decision making. Clinical scenario: A 65-year-old men with typical angina on dual-antianginal therapy underwent computed tomographic angiography (CTA) that revealed an intermediate (50% 70%) left anterior descending stenosis. The plaque revealed positive remodeling, low attenuation plaque (LAP), and spotty calcification. Invasive angiography confirmed an intermediate stenosis and the fractional flow reserve (FFR) was 0.81. Images: Multiplanar reconstruction of the left anterior descending coronary artery (A) demonstrates an intermediate (50% 70%) proximal stenosis (arrow). Cross-sectional analysis (C) and schematic (D) of the straightened vessel (B) reveal LAP adjacent to the lumen ( 8 to 38 HU) and spotty calcification (745 HU); PR compared with the normal segment (remodeling index, 1.18) was also noted. Reprinted from Hecht et al 53 with permission of the publisher. Copyright 2015, Elsevier.

Hecht et al Fractional Flow Reserve 309 lesions had FFR of 0.80. Therefore, it would be reasonable to perform CT-FFR on all stenoses in the 30% to 90% range to minimize both unnecessary ICA and the failure to detect potentially ischemic stenoses in the lower range. The subsets of bifurcation and complex lesions have not been addressed in any study other than to be excluded from the study by Renker et al. 37 CTA Plaque Characteristics and FFR It has become evident that high-risk CTA plaque characteristics influence outcomes and ischemia. Patients with low-attenuation plaque and positive remodeling (2 feature positive plaques) accounted for 50% of acute coronary syndromes in both shortand long-term studies 43 45 with no differences in DS between those plaques with and without ischemic events. Several studies have confirmed the relationship between CTA plaque characteristics and FFR. CTA and invasive FFR were obtained in 58 patients with intermediate (30% 69%) stenosis. 46 The area under the receiver operator characteristics curve for ischemic FFR was highest (P=0.001) for total atherosclerotic plaque volume (0.85), followed by minimum luminal area (0.78), minimum luminal diameter (0.75), DS (0.68), and area stenosis (0.66). In the much larger and more detailed second study, 47 a multicenter evaluation of 407 lesions in 252 patients undergoing CTA and invasive FFR, there was a 50% increased risk of FFR of 0.80 per additional 5% atherosclerotic plaque volume compared with a 30% increased risk per 5% lumen area stenosis. In multivariate analysis of adverse plaque characteristics, positive remodeling, low attenuation plaque, and spotty calcification were associated with odds ratios for ischemic FFR of 5.4 (P<0.001), 2.2 (P=0.028), and 1.6 (0.177), respectively. The presence of 1 and 2 characteristics was associated with odds ratios of 4.5 (P<0.001) and 13.2 (P<0.001), respectively. Finally, in the equally large and detailed third study, 48 484 vessels in 254 patients were evaluated with FFR and FFRCT using optimal plaque thresholds derived from area under the receiver operator characteristics curves. Lowdensity noncalcified plaque, total noncalcified and total plaque volume, positive remodeling, and plaque length were all significantly related to FFR of 0.80 in univariate analysis; only low-density noncalcified plaque volume was significant (relative risk, 4.3; P<0.001) in multivariate analysis, independent of the degree of stenosis. The area under the receiver operator characteristics curve for detection of FFR of <0.80 was 0.71 for >50% stenosis and increased to 0.79 with the addition of low-density noncalcified plaque volume (P<0.001), and it was further increased by FFRCT of <0.80 (P<0.001). Endothelial dysfunction secondary to the oxidative stress imposed by the necrotic lipid core has been thought to be the pathogenesis of the ischemic response. 49 52 Thus, stenoses in or below the lower range of presumed significance should be evaluated with CT-FFR in the setting of high-risk plaque characteristics. In addition, it has been proposed that borderline normal invasive FFR values, for example, 0.81, be regarded in a less benign light when accompanied by 2 feature positive plaques in a suggestive clinical setting (Figure 2). 53 Impact of CT-FFR on Noninvasive Imaging There are already multiple outcome studies demonstrating the superiority of CTA alone to functional testing. 23,24,26 The publication of additional positive CT-FFR data, particularly outcome and cost effectiveness studies compared with other functional imaging technologies, will likely significantly decrease the use of MPI, stress echocardiography, and stress MRI. These will be reserved for those patients with renal dysfunction and contrast anaphylaxis for whom CTA Figure 3. Myocardial perfusion imaging (MPI) by computed tomography (CT). Stress (dipyridamole) perfusion CT (A, middiastole 4- and 2-chamber multiplanar reconstruction [MPR] using a smooth filter are shown) was performed in view of the high pretest coronary artery disease likelihood using conventional static CT-MPI protocol with a 256-detector scanner and demonstrated an extensive reversible perfusion defect at the anterior wall (*) and a mild subendocardial reversible perfusion defect at the inferior wall (arrow). Single-photon emission CT confirmed the findings (B). The patient was referred to invasive coronary angiography (C), which showed 3-vessel disease with totally occluded right coronary artery, critical lesion at the left anterior descending artery, and severe lesion at the distal circumflex artery. Reprinted from Gonçalves et al 54 with permission of the publisher. Copyright 2015, Elsevier.

310 Circulation Research July 8, 2016 is contraindicated and for institutions without CT-FFR capability. The additional time, contrast load, and radiation of computed tomographic stress MPI (CTP) may also dampen enthusiasm. On a cardiology community level, as noninvasive functional imaging provides a substantial proportion of the private cardiologists income, its further decline will likely accelerate the rapidly ongoing hospital acquisition of private practices with all the implications thereof. This is particularly salient if insurers mandate CTA and CT-FFR as the first test in the evaluation of chest pain. Practical Issues Although immediate access to CT-FFR results is not imperative in stable angina patients, convenience and cost favor an on-site workstation solution, provided that the FFR values are as accurate as the FFRCT of Heartflow, which must be sent out for derivation. Ultimately, all vendors will likely develop on-site CT-FFR, with the additional cost absorbed by insurers at a level not likely to exceed the cost of the CTA itself. Whether it will be a purchasable software upgrade or a per use fee is yet to be determined. Reliance on Invasive FFR The remarkably rapid ascendancy of invasive FFR to the gold standard for intervention as a class 1 (level of evidence A) recommendation is predicated on the results of the FAME and DEFER trials. 6,9,10,12 14 In the unlikely event that this role is downgraded, CT-FFR will suffer accordingly. Of note, the clinical use of CT-FFR is not yet guideline recommended. Other CTA Functional Testing Although not measuring FFR itself, there are other CTA modalities that have been compared with invasive FFR: CTP and transluminal attenuation gradient/corrected contrast opacification (TAG/CCO). CT Myocardial Perfusion Imaging Similar to single-photon emission computed tomographic MPI, static CTP requires both rest and stress acquisitions, with the order determined by the pretest likelihood of significant obstructive disease. Adenosine, regadenoson, and dipyridamole have been used as the stress agent, with semiquantitative comparison of rest and stress myocardial perfusion to determine the extent of ischemia (Figure 3). 54 There are extensive CTP data favorably comparing its accuracy to ICA, MPI, and stress MRI for the detection of >50% stenosis by ICA. 55 58 The ability to reduce false-positive findings by discounting perfusion abnormalities in the distribution of normal vessels offers an advantage over the other functional tests, all of which lack an anatomic component. There are several studies comparing CTP with invasive FFR (Table 4). In 195 vessels in 65 patients undergoing invasive FFR, CTP yielded 95% sensitivity, 74% specificity, 49% PPV, and 98% NPV compared with 98% sensitivity, 54% specificity, 37% PPV, and 99% NPV for CTA alone. 59 Ko et al, 60 in 86 vessels in 42 patients, reported 76% sensitivity, 84% specificity, 82% PPV, and 79% NPV for CTP. The values for CTA alone were 93%, 60%, 68%, and 90%, respectively. The combination of computed tomographic myocardial perfusion (CTMP) defect and CTA >50% stenosis improved the specificity (98%) and PPV (97%) but decreased the sensitivity to 68%. CTP was compared with FFR in 101 patients, only 27 of whom had directly measured FFR with the remainder imputed from total occlusions of severely stenotic lesions. CTP alone was associated with 68% sensitivity, 93% specificity, 88% PPV, and 79% NPV. With the addition of CTP to CTA, they reported sensitivity, specificity, PPV, and NPV of 89%, 83%, 80%, and 90%, respectively. 61 Wong et al 62 reported 89% sensitivity, 65% specificity, 57% PPV, and 92% NPV in 127 vessels in 75 patients undergoing FFR. The addition of CTP to the CTA resulted in sensitivity, specificity, PPV, and NPV of 76%, 89%, 78%, and 88%, respectively. 62 Finally, Choo et al 63 noted that the addition of CTP to CTA improved specificity from 83% to 94% and PPV from 75% to 90% without change in sensitivity and NPV. The variable results of CTMP compared with CT-FFR, as well as the additional radiation, contrast, and scan time inherent in 2 separate acquisitions, are current disadvantages of this modality. Dynamic CTMP, and dual-energy CT that reduces beam-hardening artifacts, may improve the accuracy of CTMP. 64,65 Transluminal Attenuation Gradient/Corrected Contrast Opacification Resting flow impairment measured by differences in resting CTA contrast densities proximal and distal to a stenosis, with or without CCO to correct for the downstream decreases in attenuation inherent in studies not acquired with a single beat, has been evaluated in multiple studies with initially favorable results, despite an expected reduction in resting flow only with stenosis >90% (Figure 4). 54 Subsequent data have not been supportive of additive value. TAG is defined as the linear regression coefficient between luminal contrast attenuation (Hounsfield units [HU]) and length from the ostium (cm). The diagnostic performance of TAG and CCO is summarized in Table 5. Although the sensitivity, specificity, PPV, and NPV of TAG were inferior to CTA alone (77%, 74%, 67%, and 86% versus 94%, 66%, 64%, and 94%) in 54 patients and 78 vessels with invasive FFR as the gold standard, the addition of TAG to CTA provided incremental value. 66 In a second study of 97 vessels in 63 patients with invasive FFR, the sensitivity, specificity, PPV, and NPV of TAG were 48%, 91%, 79%, and 71%, respectively; TAG was additive to CTA alone (c-statistic increased from 0.726 0.809, P=0.025). 67 For CCO, the corresponding values were 65%, 61%, 54%, and 71%, and it did not significantly increase the c-statistic of CTA (0.726 versus 0.784, P=0.09). Comparison of TAG with CTA and FFRCT was performed in 53 patients and 82 vessels from the DISCOVER-FLOW trial; TAG sensitivity was 38% versus 72% for CTA >70% DS and 81% for FFRCT. 68 Evaluation of well-developed collateral flow in the setting of chronic total occlusions by TAG yielded 65% sensitivity, 73% specificity, 52% PPV, and 82% NPV, as well as accurately differentiating between retrograde and antegrade flow. 69 A subsequent comparison with invasive FFR revealed no incremental value for TAG (69% sensitivity, 44% specificity, 27% PPV, and 83% NPV) over CTA alone (95% sensitivity, 75% specificity, 54% PPV, and 98% NPV) in 253 vessels in 85 patients; the addition of CCO and exclusion of calcified

Hecht et al Fractional Flow Reserve 311 Table 4. Diagnostic Performance of CTP, CTA, and CTA+CTP Compared With Invasive FFR Gold Standard Study Per-vessel segments only slightly improved the results. 70 Finally, in a 3 part study of phantoms, dogs and 152 vessels in 62 patients undergoing ICA, no significant TAG differences were noted between stenotic and normal vessels. Proximal to distal decreases in attenuation were a function of vessel diameter rather than stenosis. 71 The variable TAG results may reflect the use of 64 detector scanners that do not provide the isotemporal, single-beat acquisition best suited for TAG, as emphasized by Lardo et al 72 in their pilot animal and human study in which they calculated coronary blood flow using CT transluminal attenuation flow encoding. They found a strong correlation with microsphere myocardial blood flow (R 2 =0.90, P<0.001) in the animal model and appropriate values in patients. Conclusions The profound impact of invasive FFR on intervention has had major repercussions on the world of noninvasive imaging. It is no longer acceptable to gauge the efficacy of an imaging modality by its ability to identify a >50% DS. Rather, its diagnostic accuracy for the detection of an invasive FFR of 0.80 Pt/V Greif et al 59 65/195 Ko et al 60 42/86 Bettencourt et al 61 101/303 Wong et al 62 75/127 Choo et al 63 37/81 Per-patient* Greif et al 59 65 Bettencourt et al 61 101 Modalities Sensitivity Specificity PPV NPV V, % V, % V, % V, % CTA 98 54 37 99 CTP 95 74 49 98 CTA+CTP 95 75 50 98 CTA 93 60 68 90 CTP 76 84 82 79 CTA+CTP 68 98 97 77 CTA 95 67 48 97 CTP 55 95 78 87 CTA+CTP 71 90 68 91 CTA 89 66 57 92 CTP NA NA NA NA CTA+CTP 76 89 78 88 CTA 93 83 75 96 CTP 93 90 84 96 CTA+CTP 93 94 90 96 CTA 100 44 67 100 CTP 97 66 74 96 CTA+CTP 97 69 76 96 CTA 100 61 67 100 CTP 68 93 88 79 CTA+CTP 89 83 80 90 CTA indicates coronary computed tomographic angiography; CTP, computed tomographic stress myocardial perfusion imaging; FFR, fractional flow reserve; NA, not available; NPV, negative predictive value; PPV, positive predictive value; Pt, patient; and V, vessel. *Per-patient data not available for the studies by Ko et al, 60 Wong et al, 62 and Choo et al. 63 is paramount and will remain so for as long as invasive FFR remains the gold standard. In this regard, the most commonly used modalities of MPI and stress echocardiography are clearly inferior to CTA alone, which, in turn seems to be inferior to CT-FFR in specificity and PPV. In addition, the ability of CTA to analyze plaque characteristics offers a clear advantage over nonanatomic functional imaging. Considerable work remains to be done: 1. Longer-term outcome studies of major cardiac events and cost effectiveness of CT-FFR guided treatment compared with the other imaging modalities, including MRI and positron emission tomography, which seem to have diagnostic performance rivaling CTA alone and CT-FFR. 2. Outcome studies comparing CT-FFR leading directly to intervention without confirmation by invasive FFR versus requiring confirmation by invasive FFR. 3. Development of the most accurate and cost-effective CT- FFR workstation-based technology.

312 Circulation Research July 8, 2016 Figure 4. Representative examples of transluminal attenuation gradient (TAG) measurement. A1, Calcified lesion in midleft anterior descending artery that was indeterminate by coronary computed tomographic angiography, but diameter stenosis (DS) was 28.7% by quantitative invasive coronary angiography. Red arrows correspond to stenotic sites. A2, The intraluminal attenuation in distal left anterior descending artery does not decrease, demonstrating no significant obstruction. A3, Cross-sectional views with gray border and sloped legend in italics represent excluded intervals. B, Severe stenosis is shown in both coronary computed tomographic angiography and invasive coronary angiography. It is confirmed in cross-sectional views. Gray dots in A2 and B2 represent intervals that were excluded because of significant calcification or stenosis. HU indicates Hounsfield units; MLD, minimum lumen diameter; NG, nitroglycerin; QCA, quantitative coronary angiography; and TAG, transluminal attenuation gradient. Reprinted from Gonçalves et al 54 with permission of the publisher. Copyright 2015, Elsevier.