Author + information
- Received April 14, 2015
- Revision received May 26, 2015
- Accepted June 4, 2015
- Published online September 1, 2015.
- Bjarne L. Nørgaard, MD, PhD∗∗ (, )
- Sara Gaur, MD∗,
- Jonathon Leipsic, MD, PhD†,
- Hiroshi Ito, MD, PhD‡,
- Toru Miyoshi, MD, PhD‡,
- Seung-Jung Park, MD, PhD§,
- Ligita Zvaigzne, MD‖,
- Nikolaos Tzemos, MD, PhD¶,
- Jesper M. Jensen, MD, PhD∗,
- Nicolaj Hansson, MD∗,
- Brian Ko, MBBS, PhD#,
- Hiram Bezerra, MD, PhD∗∗,
- Evald H. Christiansen, MD, PhD∗,
- Anne Kaltoft, MD, PhD∗,
- Jens F. Lassen, MD, PhD∗,
- Hans Erik Bøtker, MD, DMSci∗ and
- Stephan Achenbach, MD, PhD††
- ∗Department of Cardiology, Aarhus University Hospital Skejby, Aarhus, Denmark
- †Department of Radiology, St. Paul’s Hospital, University of British Columbia, Vancouver, British Columbia, Canada
- ‡Department of Cardiology, Okayama University Hospital, Okayama, Japan
- §Heart Institute, University of Ulsan College of Medicine, Asan Medical Center, Seoul, South Korea
- ‖Diagnostic Institute of Radiology, Paul Stradins Clinical University Hospital, Riga, Latvia
- ¶Department of Radiology, Golden Jubilee Hospital, Glasgow, Scotland
- #MonashHeart, Monash Medical Center and Monash University, Victoria, Australia
- ∗∗Department of Cardiology, Harrington Heart and Vascular Institute, University Hospitals, Cleveland, Ohio
- ††Department of Cardiology, Erlangen University Hospital, Erlangen, Germany
- ↵∗Reprint requests and correspondence:
Dr. Bjarne Linde Nørgaard, Department of Cardiology, Aarhus University Hospital Skejby, Skejby DK-8200 Aarhus N, Denmark.
Objectives The goal of this study was to examine the diagnostic performance of noninvasive fractional flow reserve (FFR) derived from coronary computed tomography angiography (CTA) (FFRCT) in relation to coronary calcification severity.
Background FFRCT has shown promising results in identifying lesion-specific ischemia. The extent to which the severity of coronary calcification affects the diagnostic performance of FFRCT is not known.
Methods Coronary calcification was assessed by using the Agatston score (AS) in 214 patients suspected of having coronary artery disease who underwent coronary CTA, FFRCT, and FFR (FFR examination was performed in 333 vessels). The diagnostic performance of FFRCT (≤0.80) in identifying vessel-specific ischemia (FFR ≤0.80) was investigated across AS quartiles (Q1 to Q4) and for discrimination of ischemia in patients and vessels with a low-mid AS (Q1 to Q3) versus a high AS (Q4). Coronary CTA stenosis was defined as lumen reduction >50%.
Results Mean ± SD per-patient and per-vessel AS were 302 ± 468 (range 0 to 3,599) and 95 ± 172 (range 0 to 1,703), respectively. There was no statistical difference in diagnostic accuracy, sensitivity, or specificity of FFRCT across AS quartiles. Discrimination of ischemia by FFRCT was high in patients with a high AS (416 to 3,599) and a low-mid AS (0 to 415), with no difference in area under the receiver-operating characteristic curve (AUC) (0.86 [95% confidence interval (CI): 0.76 to 0.96] vs. 0.92 [95% CI: 0.88 to 0.96]) (p = 0.45). Similarly, discrimination of ischemia by FFRCT was high in vessels with a high AS (121 to 1,703) and a low-mid AS (0 to 120) (AUC: 0.91 [95% CI: 0.85 to 0.97] vs. 0.95 [95% CI: 0.91 to 0.98]; p = 0.65). Diagnostic accuracy and specificity of FFRCT were significantly higher than for stenosis assessment in each AS quartile at the per-patient (p < 0.001) and per-vessel (p < 0.05) level with similar sensitivity. In vessels with a high AS, FFRCT exhibited improved discrimination of ischemia compared with coronary CTA alone (AUC: 0.91 vs. 0.71; p = 0.004), whereas on a per-patient level, the difference did not reach statistical significance (AUC: 0.86 vs. 0.72; p = 0.09).
Conclusions FFRCT provided high and superior diagnostic performance compared with coronary CTA interpretation alone in patients and vessels with a high AS.
- computed tomography angiography
- coronary angiography
- coronary artery disease
- coronary calcification
- fractional flow reserve
Noninvasive anatomic assessment by coronary computed tomography angiography (CTA) shows high diagnostic performance for the detection or exclusion of coronary artery disease (CAD) (1–8). However, coronary CTA has only modest accuracy regarding the quantification of stenosis severity, particularly in the setting of coronary calcification. The presence of calcified coronary lesions often leads to overestimation of stenosis severity due to blooming and beam-hardening artifacts obscuring the vessel lumen. Thus, the presence of coronary calcification is associated with reduced diagnostic specificity of coronary CTA (1–8). Moreover, coronary CTA, in its current form, has a limited capacity to determine the hemodynamic significance of coronary stenosis (9–12). These findings have prompted concern that the widespread utilization of coronary CTA may result in an increase in unneeded downstream diagnostic and therapeutic procedures (13,14).
Recent advances in computational fluid dynamics and individual image-based modeling permit calculation of coronary blood flow and pressure from standard acquired by coronary CTA datasets at rest without the need for additional imaging or medication (10–12,15). This technique allows for noninvasive calculation of fractional flow reserve derived from coronary computed tomography angiography (FFRCT), which assesses the ratio of flow across a stenosis to putative flow in the absence of stenosis. In 3 prospective trials of patients with known or suspected CAD, FFRCT exhibited high and superior diagnostic performance compared with stenosis assessment by coronary CTA in identifying ischemia as revealed by invasive measured fractional flow reserve (FFR) (10–12). Because calcium compromises identification of vessel boundary conditions for physiological modeling, the diagnostic performance of FFRCT may potentially be influenced by coronary calcification. The extent to which the presence and severity of coronary calcification affects the diagnostic performance of FFRCT is not known. Thus, the goal of the present study was to examine the diagnostic performance of FFRCT in relation to the severity of coronary calcification in patients who were enrolled in the prospective NXT (Analysis of Coronary Blood Flow Using CT Angiography: Next Steps) study and for whom coronary calcium scores were available.
Study design and patients
The rationale and design of the NXT trial have been described previously (12,16). In brief, the study was a prospective, multicenter trial designed to assess the diagnostic performance of FFRCT versus stenosis severity as assessed with coronary CTA in patients suspected of CAD by using FFR as the reference standard. Patients had coronary CTA performed <60 days before scheduled, nonemergent, clinically indicated invasive coronary angiography (ICA). Exclusion criteria included previous coronary intervention or coronary bypass surgery, suspected acute coronary syndrome, previous myocardial infarction <30 days before coronary CTA or between coronary CTA and ICA, and contraindications to beta-blockers, nitroglycerin, or adenosine.
The study protocol was approved at each of the 10 participating centers by the local institutional review board. All study subjects provided written informed consent.
Coronary CTA acquisition and analysis
Coronary CTA was performed with the use of ≥64 detector row scanners (temporal resolution 75 to 175 ms) using standard acquisition protocols in accordance with societal recommendations (17). Coronary calcium scores were assessed by local investigators (in 8 of 10 participating centers) according to the Agatston method (18). In brief, the Agatston score (AS) is calculated by using a weighted value assigned to the highest density of calcification in a given coronary artery multiplied by the area of calcium. The calcification score in each vessel is summed to give the AS. A prospective electrocardiogram-triggered scan acquired at 55% to 65% of the RR interval with 3-mm slice thickness was used for the nonenhanced acquisitions. For the contrast-enhanced scans, prospective triggering or retrospective gating was used for scan acquisition. Oral or intravenous beta-blockers, or both, were administered to achieve a heart rate ≤60 beats/min. Sublingual nitroglycerin was administered in all patients. Data acquisition was performed with 100-kV tube voltage in patients ≤70 kg and 120 kV in patients >70 kg. Experienced local investigators assessed luminal diameter stenosis in each segment ≥2 mm in diameter by using an 18-segment coronary model before ICA. Significant stenosis was defined as lumen reduction >50% in a major epicardial coronary artery segment ≥2 mm in diameter.
ICA and FFR
Diagnostic ICA was performed in all patients according to societal guidelines (19). Angiograms were transferred to the Angiography/FFR Core Laboratory (Harrington Heart and Vascular Institute, University Hospitals, Cleveland, Ohio) for blinded quantitative angiography. An ICA stenosis >50% was considered obstructive. An FFR measurement (Pressure Wire, St. Jude Medical, St. Paul, Minnesota) was performed during ICA in at least 1 vessel segment ≥2 mm in diameter and lumen stenosis ≥30%. Hyperemia was achieved by continuous intravenous adenosine infusion (infusion rate 140 to 180 μg/kg per min). Lesion-specific ischemia was defined according to an FFR ≤0.80.
FFR derived from CTA
Integration of patient-specific models of coronary anatomy and physiology to 3-dimensional computational fluid dynamic models enable computation of coronary flow and pressure at each point in the coronary tree under simulated adenosine-mediated hyperemic conditions (12,15,16). FFRCT was calculated by dividing the mean distal coronary pressure by the mean aortic pressure. Lesion-specific ischemia was defined according to an FFRCT ≤0.80. Three-dimensional blood flow simulations in the coronary vasculature were performed at HeartFlow Inc. (Redwood City, California) with the use of updated propriety software (version 1.4) in a blinded fashion, as described previously (12,16).
Integration of coronary CTA and FFR
The Angiography/FFR Core Laboratory received a 3-dimensional computer model of the coronary anatomy without FFRCT values from the Core Laboratory (HeartFlow Inc.). The model indicated the vessel locations in which FFR was measured. The corresponding FFRCT values were used for comparison (12,16).
The per-patient and per-vessel diagnostic performance of FFRCT (≤0.80) and coronary CTA (lumen reduction >50%) for the diagnosis of vessel-specific ischemia (FFR ≤0.80) were assessed according to accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) across AS quartiles (Q1 to Q4) and for discrimination of ischemia by using the area under the receiver-operating characteristic curve (AUC) in patients and vessels with a high AS (Q4) versus a low-mid AS (Q1 to Q3). The pre-test likelihood of CAD was determined by using the updated Diamond-Forrester risk score algorithm (20,21). AUC comparisons were performed on per-patient and per-vessel levels by using the method of DeLong et al. (22). Diagnostic accuracy, sensitivity, specificity, PPV, and NPV were calculated as simple proportions with Wald-type 95% confidence intervals (CIs). The Fisher exact test was used for comparison of FFRCT diagnostic accuracy, sensitivity, and specificity across AS quartiles. Subject-level comparison of diagnostic performance characteristics was performed by using the McNemar test for paired samples or the percentile bootstrap approach with 100,000 resamples as appropriate. Pearson’s correlation coefficient was calculated for assessment of the per-vessel relationship of FFRCT and FFR. Bland-Altman plots were created comparing per-vessel FFRCT and FFR values. Confidence limits were created assuming a parametric distribution, and 1-sample t tests were run to detect any significant fixed bias. All analyses were performed by using SAS version 9.3 (SAS Institute, Inc., Cary, North Carolina).
Of the 254 patients included in the NXT trial, 214 (84%) patients underwent coronary calcium scoring and thus formed the basis of the present study. Baseline characteristics of the study population are presented in Table 1. The present study cohort did not differ significantly from the overall NXT trial population. Per-vessel AS was available in 163 patients (FFR examination was performed in 333 vessels). The per-patient and per-vessel characteristics of coronary CTA, ICA, FFRCT, and FFR are shown in Table 2. Per-patient and per-vessel AS was 302 ± 468 (range 0 to 3,599) and 95 ± 172 (range 0 to 1,703), respectively. The per-patient median AS in Q1, Q2, Q3, and Q4 was 0.5 (range 0 to 26), 63 (range 27 to 147), 259 (range 148 to 415), and 711 (range 416 to 3,599). AS was >1,000 in 13 (6%) patients. The per-vessel median AS in Q1, Q2, Q3, and Q4 was 0 (range 0 to 0), 9 (range 0 to 22), 56 (range 23 to 120), and 248 (range 121 to 1,703). The proportion of patients with a positive test (coronary CTA, FFRCT, and FFR) result across AS quartiles is shown in Table 3.
Diagnostic performance of FFRCT
There was no statistically significant difference in per-patient or per-vessel diagnostic accuracy, sensitivity, and specificity of FFRCT across AS quartiles (Figure 1). There was no difference in discrimination of ischemia by FFRCT in patients with a high AS (416 to 3,599) compared with those with a low-mid AS (0 to 415) (AUC: 0.86 [95% CI: 0.76 to 0.96] vs. 0.92 [95% CI: 0.88 to 0.96]; p = 0.45) (Figure 2A). Similarly, there was no difference in discrimination of ischemia according to FFRCT in vessels with a high AS (121 to 1,703) versus those with a low-mid AS (0 to 120) (AUC: 0.91 [95% CI: 0.85 to 0.97] vs. 0.95 [95% CI: 0.91 to 0.98]; p = 0.65) (Figure 2B). Figure 3 displays a representative case of a patient with severe calcified coronary lesions that did not cause ischemia.
Diagnostic performance of FFRCT versus coronary CTA for ischemia at various levels of calcification
Accuracy, sensitivity, specificity, NPV, and PPV for FFRCT and coronary CTA in patients and vessels according to quartile of AS are shown in Table 4. The per-patient and per-vessel diagnostic accuracy and specificity of FFRCT were higher than for coronary CTA in each AS quartile, with similar high sensitivity. In patients with a low-mid or high AS, FFRCT correctly reclassified 74% and 60% of coronary CTA false-positive to true-negative findings, whereas in vessels, the proportion was 70% regardless of the AS level (Figure 4). FFRCT displayed improved discrimination of ischemia in highly calcified vessels compared with coronary CTA alone (AUC: 0.91 vs. 0.71; p = 0.004); on a per-patient level, the difference did not reach statistical significance (AUC: 0.86 vs. 0.72; p = 0.09) (Figure 2).
Correlation of FFRCT to FFR
There was a good correlation of FFRCT to FFR in vessels with a low-mid AS (0 to 120) or high AS (121 to 1,703) with R values of 0.82 (p < 0.0001) and 0.79 (p < 0.0001), respectively (Figures 5A to 5C). The difference between FFR and FFRCT values in vessels with a low-mid AS was 0.03 ± 0.07 (95% limits of agreement: −0.10 to 0.17) and 0.01 ± 0.08 (95% limits of agreement: −0.14 to 0.16) in vessels with a high AS (Figures 5D and 5E).
The present study examined the diagnostic performance of FFRCT at various levels of coronary calcification for the identification and exclusion of ischemia-causing lesions using FFR as the reference standard. The 2 major findings of this study were: 1) FFRCT provided high per-patient and per-vessel diagnostic performance and discrimination for ischemia over a wide range of coronary calcification severity; and 2) the diagnostic performance of FFRCT was superior to coronary CTA stenosis interpretation regardless of the AS level.
Because the presence of myocardial ischemia is associated with a poor prognosis (23), current guidelines recommend noninvasive functional imaging testing as the first-line strategy in patients with suspected stable CAD (21). However, shortcomings of current noninvasive diagnostic strategies are apparent from the frequent inaccurate selection of patients for ICA (24,25). To date, FFR is the only diagnostic tool shown to improve clinical outcomes and to reduce health care costs in large prospective randomized trials (26,27). As a result, FFR is considered the gold standard for guiding coronary revascularization (28). The use of FFR, however, is inherently limited by its invasiveness and costs. Moreover, FFR cannot always be measured in vessels due to extreme vessel tortuosity and/or coronary calcification. These issues underscore the need for more diagnostically accurate noninvasive testing modalities for gatekeeping to the catheterization laboratory.
In the present study, we considered high levels of coronary calcification to be represented by patients and vessels in the fourth quartile of AS values. An AS threshold of 400 has previously been used to differentiate cardiovascular risk independent of other risk factors (29,30), thus underscoring the relevance in this study of a per-patient AS level of 415 as a surrogate of severe coronary calcification. However, the presence of calcification challenges the diagnostic performance of coronary CTA because of limited spatial resolution with partial volume and beam-hardening related artifacts. Accordingly, several studies have shown that increasing severity of coronary calcification is associated with lower diagnostic specificity of coronary CTA (1–8). Yan et al. (5) recently showed that a per-patient AS ≥1 is an independent predictor of a false-positive coronary CTA result. Moreover, Dewey et al. (6) found that on a per-vessel basis, for each increase of 10 AS units, the risk of a misdiagnosis increased by 3%. Thus, guidelines do not recommend performing coronary CTA in patients with high levels of coronary calcification (21).
In 3 previous prospective multicenter trials including a total of 609 patients and 1,050 vessels with blinded comparison to FFR, computation of FFRCT from standard acquired coronary CTA images produced promising results in identifying lesion-specific ischemia (10–12). Because FFRCT is derived from coronary CTA images, significant computed tomography imaging artifacts may impair the diagnostic performance of FFRCT. Thus, in the recent NXT trial, 12% of the patients were not eligible for FFRCT computation because of the presence of computed tomography–related artifacts (motion/misregistration, high image noise, excessive calcium blooming, or low contrast to noise), which alone or in combination significantly compromised the image quality (12). However, similar to coronary CTA, the diagnostic performance of FFRCT has been shown to improve with the adherence to best practices guidelines for image acquisition, particularly regarding heart rate control and the use of pre-scan nitroglycerin (17,31,32).
In the present study, there was a tendency towards declining diagnostic accuracy and specificity with increasing AS levels. However, even in the highest quartile, with per-patient AS ranging from 416 to 3,599, FFRCT had almost a 2-fold improvement in diagnostic accuracy and a >3-fold improvement in specificity compared with coronary CTA interpretation alone. Accordingly, FFRCT reclassified 60% of those patients in the highest AS quartile with false-positive coronary CTA results to true-negative findings. Moreover, FFRCT per-patient and per-vessel diagnostic accuracy (74% and 83%, respectively), sensitivity (88% and 82%), and specificity (68% and 84%) in the highest AS quartile were equal to or superior to diagnostic performance characteristics reported for most conventional noninvasive ischemia testing modalities when compared with directly measured FFR (33).
The high diagnostic performance of FFRCT in the setting of coronary calcification most likely is a result of the FFRCT computation process. The latter is based on a large pool of data, including the global coronary and myocardial anatomy. Moreover, luminal dimensions are assessed along the entire length of each vessel by using segmentation methods that correct for calcium blooming and physiological models that incorporate both global and local flow factors, which may influence pressure gradients along the course of the vessel. Thus, segmental artifacts per se may not significantly influence the overall FFRCT computation result. In contrast, coronary CTA stenosis assessment relies on identification of segmental reductions in lumen caliber and, thus, the presence of calcification may have greater impact on interpretation and may compromise both the per-vessel and per-patient diagnostic performance. This is supported by findings in a substudy from the DISCOVER-FLOW (Diagnosis of Ischemia-Causing Coronary Stenoses by Noninvasive Fractional Flow Reserve Computed From Coronary Computed Tomographic Angiograms) trial, which showed that even at lower levels of coronary CTA image quality, FFRCT continued to provide significant diagnostic improvement when compared with coronary CTA (34). Specifically, the diagnostic accuracy of FFRCT in 42 patients with calcification-related artifacts was superior to coronary CTA (86% vs. 67%) arising from an increase in specificity (82% vs. 29%) and comparable sensitivity.
The present study extends earlier findings by demonstrating high and superior diagnostic performance of FFRCT compared with coronary CTA interpretation over a wide range of coronary calcification severities in a large cohort of patients (and vessels). The findings in this study support the potential of FFRCT as a reliable gatekeeper to ICA and coronary revascularization across a representative cohort of patients. Moreover, the high diagnostic performance of FFRCT in patients with coronary calcification, together with future improvements in computed tomography spatial resolution and/or FFRCT technology, may potentially expand the eligibility of coronary CTA testing to patients with higher pre-test probability of disease. Supportive of this possibility is recent data from Pontone et al. (7) showing that high spatial resolution compared with standard coronary CTA is associated with improved evaluability and diagnostic performance regardless of calcification severity. The latter issues need further delineation in future studies.
The present trial was a substudy of the prospective, multicenter NXT trial. Inherently, study inclusion was based on pre-specified selection criteria (12); for example, a target heart rate <60 beats/min and body mass index <35 kg/m2. Furthermore, patients with recent myocardial infarction, acute coronary syndrome, or previous coronary revascularization with percutaneous coronary intervention or coronary artery bypass grafting were not included in this study; hence, the findings cannot be generalized to all patients with known CAD. Assessment of AS was not required in the NXT study. Nonetheless, AS was recorded and available in 214 (84%) patients and in 333 vessels, representing a large patient and vessel sample size with a broad range of AS values. The number of patients with an AS ≥1,000 was very low; thus, the diagnostic performance of FFRCT in this patient category needs further investigation.
FFRCT provided high diagnostic performance and discrimination of ischemia in patients and vessels over a wide range of coronary calcification scores. The diagnostic accuracy and specificity of FFRCT was superior to coronary CTA assessment in patients and vessels with low, intermediate, or high levels of calcification.
COMPETENCY IN MEDICAL KNOWLEDGE: Noninvasive FFRCT provided high diagnostic performance and discrimination of ischemia in patients and vessels over a wide range of calcification scores. The diagnostic accuracy and specificity of FFRCT was superior to conventional coronary CTA assessment in patients and vessels with low-mid or high levels of coronary calcification.
TRANSLATIONAL OUTLOOK: The high diagnostic performance of FFRCT in patients with coronary calcification together with future improvements in computed tomography spatial resolution and FFRCT technology may potentially expand the eligibility of coronary CTA testing in real-world practice (e.g., to patients with high pre-test probability of CAD). Additional studies are needed.
Dr. Nørgaard has received research grants from Edwards Lifesciences. Dr. Leipsic has received speaker's honorarium from GE and is a consultant for Edwards Lifesciences and HeartFlow. Dr. Christiansen has received research grants from St. Jude Medical. Dr. Lassen has received research grants from St. Jude Medical, Boston Scientific, Radi, Terumo, and Volcano. Dr. Achenbach has received research grants from Siemens and Abbott Vascular. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- Agatston score
- area under the receiver-operating characteristic curve
- coronary artery disease
- confidence interval
- computed tomography angiography
- fractional flow reserve
- fractional flow reserve derived from coronary computed tomography angiography
- invasive coronary angiography
- negative predictive value
- positive predictive value
- Received April 14, 2015.
- Revision received May 26, 2015.
- Accepted June 4, 2015.
- 2015 American College of Cardiology Foundation
- Budoff M.J.,
- Dowe D.,
- Jollis J.G.,
- et al.
- Abdulla J.,
- Pedersen K.S.,
- Budoff M.,
- Kofoed K.F.
- Arbab-Zadeh A.,
- Miller J.M.,
- Rochitte C.E.,
- et al.
- Yan R.T.,
- Miller J.M.,
- Rochitte C.E.,
- et al.
- Dewey M.,
- Vavere A.L.,
- Arbab-Zadeh A.,
- et al.
- Kruk M.,
- Noll D.,
- Achenbach S.,
- et al.
- Meijboom W.B.,
- Van Mieghem C.A.,
- van Pelt N.,
- et al.
- Koo B.K.,
- Erglis A.,
- Doh J.H.,
- et al.
- Nørgaard B.L.,
- Leipsic J.,
- Gaur S.,
- et al.
- Taylor C.A.,
- Fonte T.A.,
- Min J.K.
- Agatston A.S.,
- Janowitz W.R.,
- Hildner F.J.,
- Zusmer N.R.,
- Viamonte M. Jr..,
- Detrano R.
- Hachamovitch R.,
- Hayes S.W.,
- Friedman J.D.,
- Cohen I.,
- Berman D.S.
- Fearon W.F.,
- Shilane D.,
- Pijls N.H.,
- et al.
- Wijns W.,
- Kolh P.,
- Danchin N.,
- et al.
- Vliegenthart R.,
- Oudkerk M.,
- Hofman A.,
- et al.