Author + information
- Received April 20, 2011
- Revision received June 21, 2011
- Accepted June 27, 2011
- Published online November 1, 2011.
- Bon-Kwon Koo, MD, PhD⁎,
- Andrejs Erglis, MD, PhD†,
- Joon-Hyung Doh, MD, PhD‡,
- David V. Daniels, MD§,
- Sanda Jegere, MD∥,
- Hyo-Soo Kim, MD, PhD⁎,
- Allison Dunning, MD¶,
- Tony DeFrance, MD#,
- Alexandra Lansky, MD⁎⁎,
- Jonathan Leipsic, BSc, MD†† and
- James K. Min, MD‡‡,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. James K. Min, Division of Cardiology, Department of Medicine, Cedars-Sinai Medical Center, 8700 Beverly Boulevard, South Taper Building 1258, Los Angeles, California 90048
Objectives The aim of this study was to determine the diagnostic performance of a new method for quantifying fractional flow reserve (FFR) with computational fluid dynamics (CFD) applied to coronary computed tomography angiography (CCTA) data in patients with suspected or known coronary artery disease (CAD).
Background Measurement of FFR during invasive coronary angiography is the gold standard for identifying coronary artery lesions that cause ischemia and improves clinical decision-making for revascularization. Computation of FFR from CCTA data (FFRCT) provides a noninvasive method for identifying ischemia-causing stenosis; however, the diagnostic performance of this new method is unknown.
Methods Computation of FFR from CCTA data was performed on 159 vessels in 103 patients undergoing CCTA, invasive coronary angiography, and FFR. Independent core laboratories determined FFRCT and CAD stenosis severity by CCTA. Ischemia was defined by an FFRCT and FFR ≤0.80, and anatomically obstructive CAD was defined as a CCTA with stenosis ≥50%. Diagnostic performance of FFRCT and CCTA stenosis was assessed with invasive FFR as the reference standard.
Results Fifty-six percent of patients had ≥1 vessel with FFR ≤0.80. On a per-vessel basis, the accuracy, sensitivity, specificity, positive predictive value, and negative predictive value were 84.3%, 87.9%, 82.2%, 73.9%, 92.2%, respectively, for FFRCT and were 58.5%, 91.4%, 39.6%, 46.5%, 88.9%, respectively, for CCTA stenosis. The area under the receiver-operator characteristics curve was 0.90 for FFRCT and 0.75 for CCTA (p = 0.001). The FFRCT and FFR were well correlated (r = 0.717, p < 0.001) with a slight underestimation by FFRCT (0.022 ± 0.116, p = 0.016).
Conclusions Noninvasive FFR derived from CCTA is a novel method with high diagnostic performance for the detection and exclusion of coronary lesions that cause ischemia. (The Diagnosis of ISChemia-Causing Stenoses Obtained Via NoninvasivE FRactional FLOW Reserve; NCT01189331)
Large-scale randomized trials have demonstrated that fractional flow reserve (FFR)—or the ratio of maximal myocardial blood flow through a diseased artery to the blood flow in the hypothetical case that this artery is normal—is a useful physiological test for assessment of lesion-specific ischemia and a valuable adjunct to anatomic assessment of coronary artery disease (CAD) as determined by invasive coronary angiography (ICA) (1–4). This combined anatomic-physiological evaluation of CAD allows for enhanced clinical decision-making that improves event-free survival, reduces unnecessary revascularization, and lowers healthcare costs (5,6).
Coronary computed tomographic angiography (CCTA) has emerged as a noninvasive test that assesses anatomic CAD stenosis severity (7). Although prior multicenter studies have demonstrated favorable diagnostic performance for CCTA identification and exclusion of anatomically obstructive coronary stenoses, CAD stenosis by CCTA demonstrates an unreliable relationship to lesion-specific ischemia, with the majority of high-grade stenoses detected by CCTA not causal of ischemia (8–11). These findings have raised concerns that widespread use of CCTA might result in excess referral of patients to ICA and unnecessary revascularization of nonischemic coronary lesions (12,13).
Computational fluid dynamics (CFD), as applied to CCTA images, represents a novel method that enables prediction of blood flow and pressure fields in coronary arteries and calculation of lesion-specific FFR (14–16). The FFR can be computed from typically acquired CCTA scans (FFRCT) without any modification of CCTA protocols, additional image acquisition, or administration of medications.
We performed a prospective multicenter study to compare the diagnostic performance of FFRCT with CCTA stenosis for the diagnosis of lesion-specific ischemia, as determined by FFR performed at the time of ICA.
The DISCOVER-FLOW (Diagnosis of ISChemia-Causing Stenoses Obtained Via NoninvasivE FRactional FLOW Reserve) study was conducted at 4 sites (Seoul National University Hospital, Seoul, Korea; Pauls Stradins Clinical University Hospital, Riga, Latvia; Inje University Ilsan Paik Hospital, Goyang, Korea; Stanford University Medical Center, Palo Alto, California). The study protocol was approved by the Institutional Review Boards of each center. The study was funded by Heartflow, Inc. (Redwood City, California).
The study population comprised 103 stable patients with suspected or known CAD who underwent CCTA, ICA, and FFR between October 13, 2009, and January 14, 2011. Inclusion criteria were: adult ≥18 years of age, CCTA with ≥50% stenosis in a major coronary artery ≥2.0 mm diameter (determined by clinical site), and undergoing clinically indicated ICA with FFR. Exclusion criteria included individuals unable to provide informed consent; noncardiac illness with life expectancy <2 years; pregnant state; allergy to iodinated contrast; serum creatinine ≥1.7 mg/dl; significant arrhythmia; heart rate ≥100 beats/min; systolic blood pressure ≤90 mm Hg; contraindication to beta blockers, nitroglycerin or adenosine; prior coronary artery bypass surgery (CABG); Canadian Cardiovascular Society class IV angina; or non-evaluable CCTA as determined by the CCTA core laboratory. Importantly, every evaluable CCTA as judged by the CCTA core laboratory was interpreted in independent blinded fashion by the FFRCT core laboratory in an intent-to-diagnose fashion.
Protocol for CCTA
Each center performed CCTA in accordance with the Society of Cardiovascular Computed Tomography Guidelines on Performance of CCTA with a variety of different computed tomography scanner platforms (Lightspeed VCT, GE Healthcare, Milwaukee, Wisconsin; Somatom Sensation and Definition CT, Siemens, Forchheim, Germany; Brilliance 256 and 64, Philips, Surrey, United Kingdom; Aquilion One and 64, Toshiba, Otawara, Japan) (17). Intravenous or oral metoprolol was administered for any patient with a heart rate ≥65 beats/min. Immediately before image acquisition, 0.2 mg sublingual nitroglycerin was administered. During acquisition, 80 to 100 cc of contrast (Isovue 370 mg/dl, Bracco, Princeton, New Jersey; Omnipaque 350 mg/dl, GE Healthcare, Princeton, New Jersey; Visipaque 320 mg/dl, GE Healthcare) was injected, followed by a saline flush. Helical or axial scan data were obtained with retrospective or prospective electrocardiographic gating, respectively. Image acquisition was prescribed to include the coronary arteries, left ventricle, and proximal ascending aorta. The scan parameters were: 64 × 0.625/0.750 mm collimation, tube voltage 100 or 120 mV, effective 400 to 650 mA. Radiation dose reduction strategies were employed when feasible, and doses ranged from 3 to 15 mSv.
Noninvasive coronary artery analysis by CCTA. The CCTAs were analyzed in blinded fashion by an independent core laboratory (CVCTA, San Francisco, California) in accordance with the Society of Cardiovascular Computed Tomography guidelines on CCTA interpretation (18). The CCTA images were evaluated with 3-dimensional workstations (Vital Images, Minneapolis, Minnesota; Ziosoft, Redwood City, California). The CCTAs were visualized by any post-processing method, including axial, multiplanar reformat, maximum intensity projection, and cross-sectional analysis.
Coronary segments were scored in a semi-quantitative manner with an 18-segment Society of Cardiovascular Computed Tomography model. In each segment, atherosclerosis was defined as tissue structures >1 mm2 that existed within the coronary artery lumen or adjacent to the coronary lumen that could be discriminated from pericardial tissue, epicardial fat, or vessel lumen itself. Coronary lesions were quantified for luminal diameter stenosis as none (0%), mild (1% to 49%), moderate (50% to 69%), or severe (≥70%). Anatomically obstructive CAD by CCTA was defined at 2 thresholds (≥50% or ≥70% stenosis severity) on a per-patient and per-vessel basis.
ICA image acquisition and FFR performance
Selective ICA was performed by standard catheterization in accordance with the American College of Cardiology Guidelines for Coronary Angiography (19). Two projections were obtained/major epicardial vessel, with angles of projection optimized on the basis of cardiac position. The FFR was performed in vessels as clinically indicated but was not performed for subtotal (99% stenosis) lesions. After administration of nitroglycerin, a pressure-monitoring guidewire (PressureWire Certus, St. Jude Medical Systems, Uppsala, Sweden; ComboWire, Volcano Corporation, San Diego, California) was advanced past the stenosis. Hyperemia was attained by administration of intravenous (140 μg/kg/min, n = 90) or intracoronary (50 μg, n = 13) adenosine, with route of administration at the discretion of the operator (20). The position of the distal pressure sensor was recorded to enable the FFRCT to be calculated from the same point as the measured FFR. The FFR was calculated by dividing the mean distal coronary pressure by the mean aortic pressure during hyperemia. The FFR was considered diagnostic of ischemia at a threshold of ≤0.80 on a per-patient and per-vessel basis (3).
The FFRCT was performed in blinded fashion by core laboratory scientists at HeartFlow, Inc. (Redwood City, California). Three-dimensional models of the coronary tree and ventricular myocardium were reconstructed with custom methods applied to blinded CCTA data for simulation of coronary flow and pressure (14). Blood was modeled as a Newtonian fluid with incompressible Navier-Stokes equations and solved subject to appropriate initial and boundary conditions with a finite element method on a parallel supercomputer. Because coronary flow and pressure are unknown a priori, a method to couple lumped parameter models of the microcirculation to the outflow boundaries of the 3-D model was used (21). The FFRCT analyses required approximately 5 h/exam; iterative improvements in automation are expected to reduce processing time.
The FFRCT technology is based on 3 key principles. Because patients with rest angina were excluded, the first is that coronary supply meets myocardial demand at rest. Adherence to this principle enabled calculation of total resting coronary flow relative to ventricular mass. The second principle is that resistance of the microcirculation at rest is inversely but not linearly proportional to the size of the feeding vessel (22–26). The third principle is that microcirculation reacts predictably to maximal hyperemic conditions in patients with normal coronary flow (27). On the basis of these principles, a lumped parameter model representing the resistance to flow during simulated hyperemia was applied to each coronary branch of the segmented CCTA model. The FFRCT was modeled for conditions of adenosine-induced hyperemia; an FFRCT ≤0.80 was considered diagnostic of lesion-specific ischemia.
The FFRCT core laboratory scientists were instructed as to the location within a coronary artery where the FFR was reported by an independent scientist who interpreted both the CCTA and ICA and who was not involved in the CCTA, ICA, FFRCT, or FFR analyses. The FFRCT core laboratory scientists interpreted all CCTAs judged evaluable by an independent blinded reader in an intent-to-diagnose fashion.
Sample size calculation and statistical analyses
The diagnostic accuracy of CCTA stenosis ≥50% for the detection of lesion-specific ischemia as compared with FFR was estimated to be 49% (11). To detect a relative improvement in diagnostic accuracy of ≥25% for FFRCT as compared with CCTA stenosis, 150 vessels provided 90% power with a 2-sided alternative hypothesis. For estimation of power, we employed McNemar's test to account for the paired nature of the data and assumed independence of lesion-specific ischemia in vessels and a 22% rate of discordance between FFRCT and FFR.
Categorical variables are presented as frequencies and percentages, with continuous variables as mean ± SD. Diagnostic measures on a per-patient and -vessel basis were calculated, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, positive likelihood ratio, and negative likelihood ratio. The area under the receiver-operator characteristics curve (AUC) was calculated for CCTA stenosis and FFRCT. The AUCs were compared by the DeLong method. Pearson's and Spearman's correlation coefficients were calculated to determine the relationship between FFRCT and FFR. Bland-Altman analysis was performed with FFR as the standard of reference. All analyses were performed with SAS Proprietary Software (version 9.2, SAS Institute, Cary, North Carolina).
One hundred three patients (100%) who underwent ICA, FFR, CCTA, and blinded CCTA core laboratory evaluation comprised the study population. Baseline characteristics are listed in Table 1. The mean interval between the CCTA and FFR was 2.3 days (range 0 to 26 days), with no adverse events or revascularization between exams. Six vessels were sub-totally occluded (99% stenosis), and FFR was not measured in these vessels. Of the 159 vessels for which FFR was performed, maximal stenoses as determined by the CCTA core lab were: 0% (n = 7); 1% to 25% (n = 20); 26% to 50% (n = 18); 50% to 69% (n = 47); and 70% to 98% (n = 67).
Diagnostic performance of FFRCT versus CCTA for diagnosis of ischemia-producing lesions
The FFRCT applied to coronary vessels resulted in 51 true positives (32.1%), 83 true negatives (52.2%), 18 false positives (11.3%), and 7 false negatives (4.4%). Diagnostic performance of FFRCT and CCTA on a patient- and vessel-based evaluation are listed in Table 2. Representative examples of anatomically obstructive stenosis with and without ischemia-producing stenoses are shown in Figures 1 and 2⇓. Applying FFRCT values and CCTA stenosis ranges revealed a higher AUC for FFRCT as compared with CCTA (0.90 vs. 0.75, p = 0.001) on a lesion-specific basis (Fig. 3). Per-patient performance of FFRCT versus CCTA can be seen in Table 2. A higher AUC for FFRCT was observed, similar to per-vessel analyses, as compared with CCTA for per-patient discrimination (0.92 vs. 0.70, p = 0.0001) (Fig. 3). No improvement in AUC was observed on either the per-vessel and per-patient level when CCTA stenosis was added to FFRCT (p = 0.50 and p = 0.64, respectively).
Per-vessel diagnostic performance of FFRCT in vessels with intermediate stenoses by CCTA
Among the 47 vessels with 50% to 69% stenoses by CCTA, 25.5% (n = 12) exhibited ischemia FFR. For these stenoses, FFRCT demonstrated a diagnostic accuracy, sensitivity, specificity, PPV, and NPV of 83.0%, 66.7%, 88.6%, 66.7%, and 88.6%. All 4 “false negative” vessels, where FFRCT did not demonstrate ischemia, had measured FFR values between 0.75 and 0.80.
Correlation of FFRCT to FFR
There was good correlation of per-vessel FFRCT values with FFR values (Spearman's rank correlation = 0.717, p < 0.0001; Pearson's correlation coefficient = 0.678, p < 0.0001), with a slight underestimation of FFRCT as compared with measured FFR (mean difference 0.022 ± 0.116, p = 0.016) (Figs. 4 and 5)⇓⇓ at the per-vessel level, with similar values and no systematic differences at the per-patient level (mean difference 0.019 ± 0.128, p = 0.131).
In this prospective multicenter study comparing FFRCT with CCTA stenosis in patients undergoing ICA and FFR, the diagnostic accuracy of FFRCT was superior and additive to CCTA stenosis for the diagnosis of ischemia-causing lesions as determined by an invasive FFR reference standard. The FFRCT augmented the discriminatory ability to identify lesions that cause ischemia, as compared with CCTA stenosis, largely by reducing the rates of false positive lesions incorrectly classified by stenosis alone.
To our knowledge, this study represents the first of its kind to evaluate CFD principles applied to CCTA images for derivation of patient-specific measures of coronary artery pressure. Importantly, the calculation of FFRCT required no modification of CCTA acquisition protocols, no additional imaging, and no additional administration of medications. The results of the present study establish the feasibility of CFD modeling for noninvasive determination of the physiological consequences of CAD and support a utility for application of FFRCT in patients undergoing CCTA.
Coronary computed tomographic angiography is a noninvasive test for anatomic assessment of CAD severity that has been suggested as an alternative to ICA (8–10,28). Prior CCTA studies employing computed tomography scanners with ≥64-detector rows have observed a generally high diagnostic performance of CCTA, as compared with quantitative coronary angiography as a reference standard. However, significant false positive rates underscore a general overestimation of CAD severity by CCTA, and even among obstructive CAD lesions identified by CCTA that are confirmed at ICA, only a minority of such lesions are causal of ischemia (8,10,11). These findings have also been observed for ICA, and prior multicenter studies support a distinct benefit for adjunctive physiological assessment of anatomic CAD stenoses by FFR (3,5,6,29). In the FAME (Fractional Flow Reserve versus Angiography for Multivessel Evaluation) study of 1,005 patients with multivessel CAD, those that underwent FFR-guided revascularization—as compared with patients undergoing anatomically guided revascularization—experienced lower rates of adverse events, placement of fewer coronary stents, and lower healthcare costs (3,5). In this regard, the addition of FFRCT to CCTA might improve clinical decision-making and promote salutatory outcomes for individuals with CCTA-identified CAD; prospective outcomes studies are warranted.
We employed definitions identical to those used in the FAME study for ischemia by FFR (≤0.80) and obstructive stenosis by angiography (≥50%) (3). Prior studies have demonstrated enhanced specificity for detection of ischemia when employing an FFR threshold of ≤0.75 or a stenosis threshold of ≥70%, albeit at a penalty of increased false negatives (30). We examined the relationship of an FFRCT cutoff of 0.75 and CAD stenosis of 50%, respectively, and found similar results (data not shown).
One distinction between our study and the FAME study is how we dealt with subtotal (99%) occlusions. In the FAME study, these lesions were not subjected to FFR for reasons of safety and were assigned a value of 0.50 (3). We identified 6 vessels with subtotal occlusions in our study and excluded them from the primary analysis to estimate the most conservative diagnostic performance of FFRCT. Assignment of a value of 0.50 to FFRCT would necessarily improve the performance of FFRCT for vessels that would rarely be clinically interrogated by FFRCT for ischemia. Post hoc analyses assigning these vessels a value of 0.50 did expectedly improve diagnostic performance with an accuracy, sensitivity, specificity, PPV, and NPV of 84.8%, 89.1%, 82.2%, 76.0%, and 92.2%, respectively.
Interestingly, 12 of 47 (25.5%) lesions judged to be of 50% to 69% stenosis severity by CCTA caused ischemia, as determined by an FFR ≤0.80. These results are in agreement with prior ICA-FFR studies wherein the magnitude of stenosis is often discordant with the presence of ischemia (31). These patients do not reach an angiographic threshold for revascularization but, nevertheless, experience ischemia (32). In these patients, the diagnostic performance of FFRCT was maintained (accuracy 83.0%), and all 4 false negative lesions exhibited an invasive FFR value between 0.75 and 0.80, suggesting some degree of hemodynamic significance. The ability to noninvasively identify patients with ischemia but without high-grade stenoses might allow for more targeted treatments, and future studies examining this population should be considered.
We enrolled patients who did as well as did not have a prior history of CAD. Diagnostic performance of CCTA for anatomic CAD stenosis for patients with known CAD has been less robust than for patients with no prior history of CAD (8,9). As such, present Appropriate Use Criteria advocate the use of CCTA in the low-intermediate likelihood patient with suspected CAD, but the use of CCTA for patients with known CAD is generally considered uncertain or inappropriate (33). The addition of FFRCT to CCTA measures of anatomic stenosis might enhance the diagnostic accuracy and expand the utility of CCTA in this population.
The findings of this study are of significant importance. Computation of FFRCT from CCTAs might permit an “all-in-one” approach, whereby patient-specific stenosis can be assessed for lesion-specific ischemia. An FFRCT might allow for enhanced detection of patients who might benefit from revascularization in a manner more specific than conventional stress perfusion imaging, which might manifest abnormalities due to epicardial stenosis or microvascular disease or both. Furthermore, radiation doses from cardiac imaging has recently been a topic of great discussion, and the computation of FFRCT from typically acquired CCTA images negates the need for any additional ionizing radiation while providing combined physiological assessment of CAD (34). Finally, future evaluation of FFRCT for prediction of hemodynamic improvement after revascularization or functional assessment of mild diffuse CAD might expand the clinical indications of this FFRCT technique.
This study is not without limitations. First, our study was adequately powered to determine the performance of FFRCT versus CCTA stenosis at the per-vessel rather than the per-patient level. Establishment of per-patient performance of FFRCT versus CCTA will require a greater number of patients, and we are performing a parallel, prospective study to address this hypothesis (35). Second, this study enrolled patients undergoing clinically indicated ICA and FFR, and thus the ability to assess the diagnostic performance of FFRCT in all consecutive patients undergoing CCTA is not possible. However, ethical considerations precluded our performing ICA and FFR in patients for whom no significant CAD was present. Third, our study excluded patients with prior CABG, and thus the utility of FFRCT in this population remains unknown. The FFRCT should permit evaluation of CABG stenosis, by using the same principles applied to native coronary arteries, but future studies will be required to determine its diagnostic performance in this population.
In this prospective multicenter study, noninvasive FFR derived from typically acquired CCTA images offers a novel method that demonstrates high diagnostic performance for detection and exclusion of coronary artery lesions that cause ischemia.
Dr. DeFrance is on the Speaker's Bureau of Toshiba Medical Systems. Dr. Leipsic is on the Speaker's Bureau and medical advisory board of GE Healthcare. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- area under the receiver-operator characteristics curve
- coronary artery bypass surgery
- coronary artery disease
- coronary computed tomographic angiography
- computational fluid dynamics
- fractional flow reserve
- computation of fractional flow reserve from coronary computed tomographic angiography data
- invasive coronary angiography
- Received April 20, 2011.
- Revision received June 21, 2011.
- Accepted June 27, 2011.
- American College of Cardiology Foundation
- DeBruyne B.,
- Baudhuin T.,
- Melin J.A.,
- et al.
- Berger A.,
- Botman K.J.,
- MacCarthy P.A.,
- et al.
- Pijls N.H.,
- van Son J.A.,
- Kirkeeide R.L.,
- DeBruyne B.,
- Gould K.L.
- Fearon W.F.,
- Bornschein B.,
- Tonino P.A.,
- et al.
- Pijls N.H.,
- Fearon W.F.,
- Tonino P.A.,
- et al.
- Min J.K.,
- Shaw L.J.,
- Berman D.S.
- Budoff M.J.,
- Dowe D.,
- Jollis J.G.,
- et al.
- Meijboom W.B.,
- Meijs M.F.,
- Schuijf J.D.,
- et al.
- Meijboom W.B.,
- Van Mieghem C.A.,
- van Pelt N.,
- et al.
- Lauer M.S.
- Scanlon P.J.,
- Faxon D.P.,
- Audet A.M.,
- et al.
- DeBruyne B.,
- Pijls N.H.,
- Barbato E.,
- et al.
- Murray C.D.
- Hutchins G.M.,
- Miner M.M.,
- Boitnott J.K.
- Kamiya A.,
- Togawa T.
- Wilson R.F.,
- Wyche K.,
- Christensen B.V.,
- Zimmer S.,
- Laxson D.D.
- Chow B.J.,
- Abraham A.,
- Wells G.A.,
- et al.
- MacCarthy P.,
- Berger A.,
- Manoharan G.,
- et al.
- Tonino P.A.,
- Fearon W.F.,
- DeBruyne B.,
- et al.
- Taylor A.J.,
- Cerqueira M.,
- Hodgson J.M.,
- et al.
- Einstein A.J.