Author + information
- Received May 23, 2018
- Revision received September 24, 2018
- Accepted October 8, 2018
- Published online January 14, 2019.
- Roel S. Driessen, MDa,
- Ibrahim Danad, MDa,
- Wijnand J. Stuijfzand, MDa,
- Pieter G. Raijmakers, MD, PhDb,
- Stefan P. Schumacher, MDa,
- Pepijn A. van Diemen, MDa,
- Jonathon A. Leipsic, MDc,
- Juhani Knuuti, MD, PhDd,
- S. Richard Underwood, MD, PhDe,
- Peter M. van de Ven, PhDf,
- Albert C. van Rossum, MD, PhDa,
- Charles A. Taylor, PhDg,h and
- Paul Knaapen, MD, PhDa,∗ (, )@VumcAmsterdam
- aDepartment of Cardiology, VU University Medical Center, Amsterdam, the Netherlands
- bDepartment of Radiology, Nuclear Medicine & PET Research, VU University Medical Center, Amsterdam, the Netherlands
- cDepartment of Medicine and Radiology, University of British Columbia, Vancouver, British Columbia, Canada
- dTurku PET Centre, Turku University Hospital and University of Turku, Turku, Finland
- eDepartment of Nuclear Medicine, Royal Brompton Hospital, London, United Kingdom
- fDepartment of Epidemiology and Biostatistics, VU University Medical Center, Amsterdam, the Netherlands
- gHeartFlow, Inc., Redwood City, California
- hDepartment of Bioengineering, Stanford University, Stanford, California
- ↵∗Address for correspondence:
Dr. Paul Knaapen, Department of Cardiology, VU University Medical Center, De Boelelaan 1117, 1081 HV Amsterdam, the Netherlands.
Background Fractional flow reserve (FFR) computation from coronary computed tomography angiography (CTA) datasets (FFRCT) has emerged as a promising noninvasive test to assess hemodynamic severity of coronary artery disease (CAD), but has not yet been compared with traditional functional imaging.
Objectives The purpose of this study was to evaluate the diagnostic performance of FFRCT and compare it with coronary CTA, single-photon emission computed tomography (SPECT), and positron emission tomography (PET) for ischemia diagnosis.
Methods This subanalysis involved 208 prospectively included patients with suspected stable CAD, who underwent 256-slice coronary CTA, 99mTc-tetrofosmin SPECT, [15O]H2O PET, and routine 3-vessel invasive FFR measurements. FFRCT values were retrospectively derived from the coronary CTA images. Images from each modality were interpreted by core laboratories, and their diagnostic performances were compared using invasively measured FFR ≤0.80 as the reference standard.
Results In total, 505 of 612 (83%) vessels could be evaluated with FFRCT. FFRCT showed a diagnostic accuracy, sensitivity, and specificity of 87%, 90%, and 86% on a per-vessel basis and 78%, 96%, and 63% on a per-patient basis, respectively. Area under the receiver-operating characteristic curve (AUC) for identification of ischemia-causing lesions was significantly greater for FFRCT (0.94 and 0.92) in comparison with coronary CTA (0.83 and 0.81; p < 0.01 for both) and SPECT (0.70 and 0.75; p < 0.01 for both), on a per-vessel and -patient level, respectively. FFRCT also outperformed PET on a per-vessel basis (AUC 0.87; p < 0.01), but not on a per-patient basis (AUC 0.91; p = 0.56). In the intention-to-diagnose analysis, PET showed the highest per-patient and -vessel AUC followed by FFRCT (0.86 vs. 0.83; p = 0.157; and 0.90 vs. 0.79; p = 0.005, respectively).
Conclusions In this study, FFRCT showed higher diagnostic performance than standard coronary CTA, SPECT, and PET for vessel-specific ischemia, provided coronary CTA images were evaluable by FFRCT, whereas PET had a favorable performance in per-patient and intention-to-diagnose analysis. Still, in patients in whom 3-vessel FFRCT could be analyzed, FFRCT holds clinical potential to provide anatomic and hemodynamic significance of coronary lesions.
- coronary artery disease
- coronary computed tomography angiography
- fractional flow reserve
- myocardial perfusion imaging
At present, a large armamentarium of noninvasive tests is available to diagnose coronary artery disease (CAD) and subsequently risk stratify patients or guide revascularization options (1,2). The importance of accurately using these noninvasive tests is highlighted by the present-day low diagnostic yield of invasive coronary angiography (ICA) (3) and is emphasized by current guidelines for the management of suspected stable CAD (4,5). Among the available modalities, coronary computed tomography angiography (CTA) is an established diagnostic tool that strongly correlates with ICA (6). However, the estimation of hemodynamic significance by anatomic stenosis severity, as determined with coronary CTA as well as with ICA, has shown to be unreliable (7,8). This shortcoming of visual assessment of lesion severity is emphasized by previous studies, which showed that event-free survival was not improved by revascularization based on invasive angiography alone, but it did improve when invasive angiography was combined with physiological measures in terms of invasive fractional flow reserve (FFR) (9). As such, FFR has emerged as the gold standard for determining lesion-specific ischemia and guiding revascularization decision making. Conversely, myocardial perfusion imaging (MPI) with single-photon emission computed tomography (SPECT) or positron emission tomography (PET) provides physiological repercussions of CAD at the myocardial tissue level, but lacks coronary anatomical information. Recently, fractional flow reserve derived from computed tomography (FFRCT) has emerged as a promising alternative, providing functional significance of CAD derived from a standard coronary CTA (10). Using computational fluid dynamics, pressures during simulated stress can be calculated, which allows for an assessment of physiological significance of CAD expressed as FFR. Previous prospective trials have shown a significant improvement of diagnostic power for FFRCT in comparison with coronary CTA stenosis assessment alone (11–13). Comparative studies investigating FFRCT against other functional imaging modalities, however, are lacking. Therefore, the aim of this PACIFIC (Prospective Comparison of Cardiac PET/CT, SPECT/CT Perfusion Imaging and CT Coronary Angiography With Invasive Coronary Angiography) trial (14) substudy was to evaluate the diagnostic performance of FFRCT and compare it with coronary CTA, SPECT, and PET for the diagnosis of ischemia.
This post hoc substudy comprised all 208 patients from the PACIFIC trial who were suspected of CAD and who underwent coronary CTA, SPECT, and PET with routine interrogation by FFR of all major coronary arteries (NCT01521468) (14). In this single-center study, all patients prospectively underwent all noninvasive imaging and ICA with FFR within 2 weeks, regardless of the imaging results. Participants were characterized by an intermediate pre-test likelihood of stable CAD and normal left ventricular ejection fraction (LVEF). Patients were not eligible if they had previously documented CAD, signs of prior myocardial infarction, atrial fibrillation, renal failure, or contraindications to adenosine. The study complied with the Declaration of Helsinki, the study protocol was approved by the VUmc Medical Ethics Review Committee, and all patients provided written informed consent.
Coronary CTA was performed using a 256-slice CT scanner (Philips Brilliance iCT, Philips Healthcare, Best, the Netherlands) as described previously (14) in accordance with the society of cardiovascular computed tomography (SCCT) guidelines (15). Prior to the scanning protocol, sublingual nitroglycerine spray was administered to all patients and metoprolol only if necessary, aiming for a heart rate <65 beats/min. The scan was triggered using an automatic bolus-tracking technique with a region of interest placed in the descending thoracic aorta. Prospective electrocardiogram gating was used at 75% of the R-R interval. Nevertheless, persistent elevated heart rates in 4 scans required a retrospective helical protocol. An intravenous bolus of 100 ml of iodinated contrast agent was injected. Coronary CTA datasets were transmitted to an independent and blinded core laboratory (St. Paul’s Hospital, Vancouver, British Columbia, Canada) for the evaluation of diameter stenosis severity. All coronary segments ≥2 mm in diameter were visually graded on an intention-to-diagnose basis and classified as 0%, 1% to 24%, 25% to 49%, 50% to 69%, and 70% to 100% diameter stenosis, whereas stenosis ≥50% or unevaluable segments were considered significantly obstructive.
FFR derived from computed tomography
FFRCT technology involves extraction of a patient-specific geometric model of the coronary arteries from coronary CTA data, population-derived physiological models, and computational fluid dynamics techniques to solve the governing equations of blood flow for velocity and pressure under simulated hyperemic conditions (10). In this study, updated FFRCT software was used (HeartFlow FFRCT version 2.7, Redwood City, California), comprising deep-learning artificial intelligence methods to aid in identifying the lumen boundary, physiological models incorporating vessel lumen volume as well as myocardial mass data and hybrid 3-dimensional–1-dimensional computational fluid dynamics methods to improve computational efficiency while maintaining accuracy (16). HeartFlow first performed an image quality check with rejection of uninterpretable cases because of incompletely imaged myocardium or coronary arteries (partly outside of field of view) or severe forms of misalignment, motion, blooming, noise, or other artifacts, followed by FFRCT analysis blinded to the other imaging reads as well as the invasive angiographic and FFR data. Extraction of FFRCT values was performed by an independent researcher (R.D.) with knowledge of the FFR wire position but not invasive FFR values. FFRCT ratios could be obtained along the entire epicardial tree, but were taken at the same position as the FFR wire, which was generally in the distal part of the vessel and not at a specific post-stenosis location, as per PACIFIC study protocol. For identified occluded coronary arteries, a value of 0.50 was assigned.
MPI with SPECT
MPI with SPECT was acquired and analyzed as reported previously (14). In summary, SPECT images were acquired on a dual-head hybrid SPECT/CT scanner (Symbia T2, Siemens Medical Solutions, Erlangen, Germany). All patients underwent a 2-day stress-rest 99mTc-tetrofosmin protocol using intravenous adenosine (140 μg/kg/min) as a hyperemic agent and a weight-adjusted dose of 370 to 550 MBq 99mTc-tetrofosmin as radiotracer. All SPECT images were acquired using electrocardiographic gating and followed by a low-dose CT scan for attenuation correction. Image analysis was performed by a blinded core laboratory (Royal Brompton Hospital, London, England). MPI images were interpreted based on a 17-segment model. Each segment was scored using a 5-point scoring system (0, normal; 1, mildly decreased; 2, moderately decreased; 3, severely decreased; and 4, absence of segmental uptake). Summed rest scores, summed stress scores, and summed difference scores (SDS) were calculated from the segmental scores, with an SDS ≥2 considered abnormal.
MPI with PET
PET scans were performed using a hybrid PET-CT device (Philips Gemini TF 64, Philips Healthcare). Acquisition and analysis of PET imaging were described previously (14). In short, a dynamic PET perfusion scan was performed using 370 MBq of [15O]H2O during resting and adenosine (140 μg/kg/min)–induced hyperemic conditions. Low-dose CT scans allowed for attenuation correction. Reconstructed images were sent to a blinded core laboratory (Turku University Hospital, Turku, Finland), where images with quantitative myocardial blood flow (MBF) were generated. Hyperemic MBF, expressed in ml/min/g of perfusable myocardial tissue, was calculated for all 3 vascular territories derived from standard segmentation: left anterior descending, left circumflex, and right coronary artery. Hyperemic MBF ≤2.30 ml/min/g was defined abnormal (17).
ICA and FFR
ICA was performed using a standard protocol in at least 2 orthogonal directions per evaluated coronary artery segment. For the induction of epicardial coronary vasodilation, 0.2 ml of intracoronary nitroglycerin was administered prior to contrast injection. All major coronary arteries were routinely interrogated by FFR, regardless of stenosis severity, except for occluded or subtotal lesions of more than 90%. Intracoronary (150 μg) or intravenous (140 μg/kg/min) adenosine infusion was used to induce maximal coronary hyperemia. FFR was calculated as the ratio of mean distal intracoronary pressure and mean arterial pressure. A coronary lesion was considered hemodynamically significant in case of FFR ≤0.80, or stenosis severity >90% obtained with quantitative coronary angiography in case of missing FFR. A stenosis with an FFR >0.80 or a stenosis severity <30% (obtained with quantitative coronary angiography) in the absence of FFR measurements was considered not to be functionally relevant. All images and FFR signals were interpreted by experienced interventional cardiologists blinded to imaging results.
Continuous variables are presented as mean ± SD or median (interquartile range) as appropriate. Categorical variables are expressed as frequencies and percentages. Differences between continuous baseline characteristic variables were compared using the 2-sided Student’s t-test, whereas differences between categorical baseline variables were analyzed by the chi-square test or Fisher exact test when appropriate. The study endpoint was the comparison of FFRCT against CT, SPECT, and PET, in terms of sensitivity, specificity, negative predictive value, positive predictive value (PPV), diagnostic accuracy, and area under the receiver operating characteristic curve (AUC), referenced by invasive FFR. In patient-based analysis, these diagnostic measures were calculated as simple proportions with 95% confidence intervals (CIs). In vessel-based analysis, diagnostic measures were calculated using generalized estimating equations (GEE) with an exchangeable correlation structure to account for within-patient correlation. AUCs were generated to quantify the discriminative ability of each modality, and compared using the method of DeLong et al. (18) with MedCalc (MedCalc Software 184.108.40.206, Mariakerke, Belgium). Sensitivity, specificity, and accuracy for diagnosis on a patient level were compared using McNemar's test, whereas PPV and negative predictive value were compared using a marginal regression model using a working independent correlation structure. Patients were considered positive for a modality (reference standard) if at least 1 vessel was considered positive. Vessel-based diagnostic measures were compared using GEE with an exchangeable correlation structure to account for correlation between multiple vessels within the same patient. Primary per-vessel analysis was performed using all vessels that were evaluable by FFRCT, whereas only fully evaluable coronary CTA datasets (i.e., 3-vessel) were used on a per-patient basis. Secondary analysis was performed with all datasets on an intention-to-diagnose (i.e., nonevaluable vessels were deemed positive). The diagnostic performance of combined FFRCT and coronary CTA was explored by including these parameters in a multivariable model using GEE and a subsequent AUC comparison, next to reporting combined FFRCT and CTA results together with other imaging results according to FFR subgroups. The combined FFRCT and coronary CTA was considered positive only when both FFRCT and coronary CTA were positive according to 1 of the 2 coronary CTA thresholds (>25% and >50%). The association between FFRCT and invasive FFR was quantified using Pearson’s and Spearman’s correlation coefficients and agreement was assessed with Bland-Altman analysis. A p value ≤0.05 was considered statistically significant. All statistical analyses were performed using SPSS software package (IBM SPSS Statistics 20.0, Chicago, Illinois), unless stated otherwise.
Among the total of 208 patients included in the PACIFIC study, coronary CTA images of 157 (75%) were fully evaluable by FFRCT (3-vessel) and were used for the primary patient-based analysis. For 180 (87%) coronary CTA datasets, FFRCT could be assessed at least in part. Altogether, 505 (83%) vessels could be evaluated by FFRCT and were used for the primary vessel-based comparative analysis. Baseline characteristics of the study population are previously reported and therefore listed in Online Table 1 (14). In brief, patient age averaged 58.7 ± 8.5 years, 99 (63%) were male, and 71 (45%) patients were found to have significant CAD as defined by ICA with an FFR ≤0.80 or stenosis >90%. Coronary CTA assessment characteristics are shown in Table 1. Whereas failed FFRCT analysis predominantly concerned the right coronary artery (51 of 107 unevaluable vessels, 48%), most frequent reasons for unevaluable datasets were related to relatively higher heart rates such as motion (84%) and misalignment (18%), next to noise (12%) and high calcification burden (10%). Occluded arteries were present in 11% of fully evaluable cases, compared to 4% in non–fully evaluable cases (p = 0.17). Although image quality varied, numbers of unevaluable datasets were 1 (0.5%) for coronary CTA, 2 (1.0%) for SPECT, and 0 (0%) for PET, next to 0 (0%) failed coronary CTA, 2 (1.0%) failed SPECT, and 4 (1.9%) failed PET scanning procedures due to technical issues or claustrophobia. Noninvasive test results were positive in 100 (64%) patients for FFRCT, 90 (57%) patients for coronary CTA, 49 (31%) patients for SPECT, and 73 (47%) for PET (Table 2). Figure 1 illustrates typical imaging findings for the currently tested modalities.
Per-vessel diagnostic performance of FFRCT, coronary CTA, SPECT, and PET for diagnosing hemodynamically significant CAD
Per-vessel diagnostic performances of all imaging modalities for the detection of FFR-defined significant CAD in the primary analysis group (n = 505) are displayed in Table 3. As demonstrated in the Central Illustration and Figure 2A, the AUC for FFRCT was 0.94 (95% CI: 0.92 to 0.96), and significantly higher than coronary CTA alone (0.83; 95% CI: 0.80 to 0.86; p < 0.001), SPECT (0.70; 95% CI: 0.65 to 0.74; p < 0.001), and PET (0.87; 95% CI: 0.83 to 0.90; p < 0.001). Diagnostic accuracy of FFRCT (87%) was also higher than coronary CTA (79%) and PET (80%), but comparable with SPECT (82%). Sensitivity for FFRCT (90%) was higher than any of the other modalities, whereas specificity for FFRCT (86%) was comparable with coronary CTA and PET, yet lower than SPECT. The AUC for combined FFRCT and coronary CTA was 0.95 (p = 0.051 compared with FFRCT alone). Whereas selective FFRCT assessment in case of coronary CTA stenosis >25% and >50% resulted in diagnostic accuracies of 88% and 86%, respectively.
Per-patient diagnostic performance of FFRCT, coronary CTA, SPECT, and PET for diagnosing hemodynamically significant CAD
Detailed diagnostic performances of FFRCT and the comparison with other noninvasive modalities on a per-patient level in the primary analysis group (n = 157) are also shown in Table 3. As shown in Figure 2B, discriminatory power for FFRCT in terms of AUC was 0.92 (95% CI: 0.87 to 0.96), which was significantly greater than coronary CTA alone (0.81; 95% CI: 0.74 to 0.87; p = 0.002) and SPECT (0.75; 95% CI: 0.67 to 0.81; p < 0.001), but comparable to PET (0.91; 95% CI: 0.85 to 0.95; p = 0.559). Diagnostic accuracy of FFRCT (78%), however, was comparable with coronary CTA alone (76%) and SPECT (78%), but was significantly lower than PET (88%). Selective FFRCT assessment following coronary CTA stenosis >25% and >50% resulted in diagnostic accuracies of 84% and 83%, respectively (Table 2). The AUC for combined FFRCT and coronary CTA was 0.95 (p = 0.053 compared with FFRCT alone).
Diagnostic performances of imaging modalities for significant CAD with an intention-to-diagnose
Using the entire cohort of patients (n = 208) and vessels (n = 612), including nonevaluable vessels, an intention-to-diagnose analysis resulted in diagnostic performances as reported in Table 4 and Figure 3. Vessel-based AUC was equal among FFRCT (AUC: 0.83), coronary CTA (AUC: 0.80; p = 0.261), and PET (AUC: 0.86; p = 0.157), but was lower for SPECT (AUC: 0.68; p < 0.001). Per-vessel and -patient AUC for combined FFRCT and coronary CTA were 0.88 and 0.87, respectively (p < 0.001 and p = 0.001 compared with FFRCT alone). There were no significant differences with regard to overall per-vessel diagnostic accuracy. However, FFRCT showed the highest per-vessel sensitivity (92%) as opposed to the lowest specificity (70%). Per-patient AUC and diagnostic accuracy were comparable between FFRCT (0.79 and 70%, respectively), coronary CTA (0.76 and 74%; p = 0.327 and p = 0.341, respectively), and SPECT (0.74 and 76%, p = 0.087 and p = 0.266, respectively), but were outperformed by PET (0.90 and 86%; p = 0.005 and p < 0.001, respectively). Per-patient sensitivity of FFRCT (97%) was similar to coronary CTA, but higher than SPECT and PET. Conversely, specificity of FFRCT (47%) was significantly inferior to each of the other modalities.
Correlation of FFRCT and invasive FFR
Per-vessel FFRCT values showed a good correlation with invasive FFR measures, Pearson's and Spearman’s correlation coefficients were 0.80 and 0.67, respectively, p < 0.001 for both (Figure 4A). FFRCT and FFR values were concordant in 87%, whereas 10% showed FFRCT positive but FFR negative results, and 3% vice versa. Bland-Altman analysis indicated an underestimation of FFRCT compared with FFR (mean difference, 0.05 ± 0.10; p < 0.001) (Figure 4B). According to subgroups of FFRCT <0.60, 0.60 to 0.80, and >0.80, mean difference was 0.03 ± 0.11, 0.04 ± 0.15, 0.05 ± 0.08, respectively, with a significant overall intergroup difference (p < 0.001) using the analysis of variance test. In comparison, the Spearman’s correlation with invasive FFR measures and (semi)quantitative coronary CTA stenosis, SPECT SDS, and PET MBF, were 0.55, 0.42, and 0.50, respectively. Accordingly, Online Figures 1 and 2 illustrate the relationship between PET and SPECT with FFRCT.
In the present PACIFIC substudy of patients with suspected stable CAD, FFRCT values strongly correlated with invasively derived FFR, which resulted in a high diagnostic performance (Central Illustration) even though, in line with previous reports, measured FFRCT values were systematically lower than invasive FFR values. When images were of sufficient quality to analyze FFRCT, it showed an improved per-vessel and -patient diagnostic discriminative ability compared with coronary CTA, SPECT, and PET in terms of AUC, except for per-patient analysis with PET. However, intention-to-diagnose analysis, including coronary CTA images nonevaluable for FFRCT, diluted the incremental value of FFRCT resulting in a lower per-patient AUC than PET. The current results represent the first true head-to-head comparison of the functional FFRCT assessment derived from standard coronary CTA against more traditional functional imaging with SPECT and PET. Our findings support the use of FFRCT in clinical practice, taking into account an anticipated increase of FFRCT analyzability, whereas current multisociety guidelines do not advocate the use of any specific imaging modality (4,5).
Diagnostic performance of FFRCT
This study extends findings from 3 previous major FFRCT studies regarding diagnostic performance as well as feasibility (11–13). Compared with results from the DeFACTO (Determination of Fractional Flow Reserve by Anatomic Computed Tomographic Angiography) study (12), the PACIFIC FFRCT study showed a remarkably higher per-vessel diagnostic performance with an improved diagnostic accuracy from 69% towards 87%, despite only a slightly higher per-patient sensitivity and moderately increased specificity. Conversely, in comparison with results of the more recent NXT (Analysis of Coronary Blood Flow Using CT Angiography: Next Steps) trial (13), per-vessel diagnostic performance results were particularly similar between both studies, resulting in the numerical highest accuracy for PACIFIC (86% vs. 87%). Because of its direct impact on diagnostic parameters in general, some differences between the current study and previous work may in part be due to the prevalence of FFR-defined significant CAD (45% for PACIFIC, against 51% for DeFACTO and 32% for NXT). Another dissimilarity between previous dedicated FFRCT trials may be that patients in the PACIFIC trial underwent all imaging in a prospective manner after study inclusion without any of the noninvasive imaging tests already performed, and all coronary arteries were interrogated by FFR regardless of imaging results. Therefore, the current results reflect a genuine real-world performance of each of the modalities without potential inclusion bias, which might occur when patients are included after an initial test is performed and deemed suitable for study purposes. Furthermore, the present study showed a strong performance of FFRCT analysis, and the superior performance compared with coronary CTA reading was significant on a per-vessel level yet limited on a per-patient level. This reduced incremental value of FFRCT over coronary CTA in comparison with previous studies seems to be more related to the previously reported poor to moderate performance of coronary CTA than the altered performance of FFRCT itself. The reason for the improved diagnostic accuracy of coronary CTA in the PACIFIC study (76%) compared with both the DeFACTO (64%) and NXT (53%) trials remains unknown, but improved scanner characteristics might play a role next to patient preparation and core laboratory reading. Still, most diagnostic parameters improved for FFRCT in the current analysis, even with a proper performance of coronary CTA. In line with DeFACTO but in contrast to the NXT results, there was no marked increase in specificity when FFRCT was compared with coronary CTA alone, yet rather an increase in sensitivity. These altered sensitivity/specificity distributions could possibly be explained by the aforementioned differences in inclusion method and prevalence of disease, which resembles DeFACTO more than NXT, and is in line with the underestimation of FFRCT values (Figure 4B). Another major difference from other studies that could explain some different findings is that in the PACIFIC trial, all vessels were measured with FFR, regardless of visual stenosis severity. As can be appreciated from Table 2, in the distinctly abnormal subgroup with FFR <0.70, all imaging modalities showed high rates of abnormal test results (range 75% to 98%). Interestingly, in the subgroup with minimally abnormal FFR values (0.70 to 0.80), SPECT showed almost no abnormal test results (7%) in contrast to the other imaging modalities (range 73% to 87%). This might partly explain the relatively low sensitivity of SPECT, but should be interpreted in context of previous optimal cutoff value derivation for PET and FFRCT using invasive FFR values, which was not done for SPECT with FFR (10,17). In the subgroup with evidently normal FFR values (>0.90), coronary CTA still showed a relatively high prevalence of significant stenosis, reflecting the relatively low PPV. Furthermore, combined FFRCT and coronary CTA stenosis severity showed a shift of slightly reduced positive test results in abnormal FFR cases but a more pronounced reduction of false positive results, which lead to a slightly enhanced AUC (Figures 2 and 3). As such, there seems to be additional value of FFRCT analysis after visual coronary CTA stenosis assessment, which still might the be most convenient route in clinical practice.
Clinical applicability of FFRCT compared with MPI
In total, 83% of all vessels could be evaluated by FFRCT, resulting in 75% fully evaluable datasets and 87% at least partly evaluable datasets. This drop-out rate resembles previous findings and remains a substantial issue in clinical practice. Interestingly, coronary calcification burden had only a minor impact on evaluability as reflected by small differences of median CAC scores between evaluable and nonevaluable images (Table 1). This is in line with a previous study by Norgaard et al. (19), and supports the potential of FFRCT even in patients with high calcified plaque burden. Conversely, higher heart rates, additional administration of beta-blockers, and retrospective acquisition were shown to be predictive of nonevaluable CT scans. As such, an optimal use of FFRCT in clinical practice appears to be greatly dependent on high-quality images as a result of adequate pre-scan medication and low heart rates, next to high-quality scanners and acquisition protocols. MPI with SPECT or PET, however, is usually not bothered by variable heart rates. A drawback of perfusion imaging, on the other hand, is the need of pharmacological stress or exercise, which can be simply simulated with FFRCT from resting images (10). Of note, results of these tests provide essentially different information, as coronary CTA predominantly assesses epicardial coronary disease, whereas MPI evaluates the entire vascular bed including epicardial and microvascular function. As such, current European and American guidelines do not favor any specific test in general, but leave the choice of modality to patient characteristics, next to local availability, clinical question asked, radiation exposure, and costs (4,5). Still, most guidelines recommend the use of stress imaging in intermediate- to high-risk patients because of the large body of evidence that ischemia-guided revascularization trumps anatomical-guided treatment with regard to improved outcome (9). Recently updated National Institute of Health and Care Excellence guidelines, however, advocate for coronary CTA as the initial test in patients with suspected CAD, regardless of pre-test likelihood of disease (20). In this regard, a large body of evidence exists for the prognostic value of both standard coronary CTA and MPI (21–23). Such data is still lacking for FFRCT, but intuitively, the combination of anatomical and functional data from coronary CTA with FFRCT holds great potential in prognostic value. In fact, the recent PLATFORM (Prospective Longitudinal Trial of FFRct: Outcome and Resource Impacts) trial showed FFRCT to be an effective gatekeeper in clinical work-up as reflected by reducing the number of unnecessary invasive coronary angiography in up to 61% of patients without compromising clinical outcome (24,25). Additionally, the potential of FFRCT-guided revascularization was shown by Curzen et al. (26) with a change in treatment strategy of 36% of patients suspected of having CAD. However, larger studies with a longer follow-up of these low-risk patients are eagerly awaited to provide more insights in the prognostic value of FFRCT. Furthermore, notwithstanding the high diagnostic agreement in terms of concordancy, the distinct scatter of actual estimated FFRCT values as displayed in Figure 4 could be clinically relevant and necessitates careful interpretation.
The present study was a post hoc subanalysis of the previously reported PACIFIC study. Although the drop-out rate is expected to decrease with future improvement of CT scanner possibilities, software updates, and optimized patient preparation, a considerable proportion of 17% of vessels were qualified as nonevaluable by FFRCT and were excluded from the primary analysis. Therefore, a subsequent intention-to-diagnose analysis with all datasets was performed, which resulted in an ample shift of diagnostic performances in an unfavorable way for FFRCT. As such, a truly prospective comparison of imaging modalities is warranted with pre-specified optimized CT acquisition. Of note, in a recent large registry, the acceptance rate for FFRCT was 95% (27). Although the present results would likely not have been substantially different if FFRCT extraction had been performed blinded to the knowledge of the pressure wire position, it could have potentially enhanced the quantitative relationship with FFR. Accordingly, the continuous nature of the results of FFRCT may disadvantage modalities like coronary CTA and SPECT with limited categories of outcome results (Figures 3 and 4). Even so, definitions of ischemia of PET and SPECT were determined based on international guidelines, disregarding, however, the potential extent and depth of ischemia, which are of clinical validity. Using alternative thresholds for SPECT scoring such as 5% reversible perfusion defect, or using different outcome parameters for PET such as coronary flow reserve could alter current results. Whereas a widely available SPECT protocol and tracer were used, PET was performed with the less frequently used [15O]H2O tracer. Hence, current results may not be extrapolated to the more commonly used tracers [13N]NH3 and Rubidium-82. It should, however, be emphasized that the results of the current analysis only apply to patients with a normal LVEF without a prior documented history of CAD. Furthermore, generalizability of present FFRCT findings might be hampered by the relatively small sample size and the single-center study setup with the use of 1 specific coronary CTA acquisition protocol and scanner. Last, although it is believed that coronary CTA generally performs best in low-risk populations, the influence of current patient population with a relatively high prevalence of CAD is unknown, while patients with a documented history of CAD were excluded.
In this head-to-head comparative study, FFRCT showed the highest diagnostic performance for vessel-specific ischemia, provided coronary CTA images were evaluable by FFRCT. On an intention-to-diagnose basis, however, PET displays the highest diagnostic performance due to the relatively high rejection rate of FFRCT. Further improvements in CT acquisition and reconstruction are needed to improve the evaluability rate of FFRCT so the diagnostic performance could be similar to PET. Still, FFRCT would be of value in clinical practice for the noninvasive evaluation of CAD, providing not only anatomic but also hemodynamic significance of coronary lesions.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: In patients with stable CAD and adequate image quality, FFRCT provides superior functional assessment of coronary stenosis compared with coronary CTA, SPECT, or PET imaging.
TRANSLATIONAL OUTLOOK: Future studies should focus on enhancing CT image acquisition and reconstruction to improve the applicability of FFRCT in clinical practice.
Dr. Leipsic has received research grants from GE Healthcare; and serves as a consultant and holds stock options in Circle CVI and HeartFlow. Dr. Knuuti has provided trial consultancy for GE Healthcare and AstraZeneca. Dr. Underwood provides occasional consultancy to GE Healthcare. Dr. Taylor has an equity interest in and is an employee of HeartFlow. Dr. Knaapen has received unrestricted research grants from HeartFlow. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Listen to this manuscript's audio summary by Editor-in-Chief Dr. Valentin Fuster on JACC.org.
- Abbreviations and Acronyms
- computed tomography angiography
- fractional flow reserve
- fractional flow reserve derived from coronary computed tomography angiography
- invasive coronary angiography
- positron emission tomography
- summed difference scores
- single-photon emission computed tomography
- Received May 23, 2018.
- Revision received September 24, 2018.
- Accepted October 8, 2018.
- 2019 American College of Cardiology Foundation
- Danad I.,
- Szymonifka J.,
- Twisk J.W.R.,
- et al.
- Fihn S.D.,
- Blankenship J.C.,
- Alexander K.P.,
- et al.
- Meijboom W.B.,
- Meijs M.F.,
- Schuijf J.D.,
- et al.
- Tonino P.A.,
- Fearon W.F.,
- De B.B.,
- et al.
- Meijboom W.B.,
- van Mieghem C.A.,
- van P.N.,
- et al.
- Taylor C.A.,
- Fonte T.A.,
- Min J.K.
- Koo B.K.,
- Erglis A.,
- Doh J.H.,
- et al.
- Norgaard B.L.,
- Leipsic J.,
- Gaur S.,
- et al.
- Danad I.,
- Raijmakers P.G.,
- Driessen R.S.,
- et al.
- Abbara S.,
- Blanke P.,
- Maroules C.D.,
- et al.
- Danad I.,
- Uusitalo V.,
- Kero T.,
- et al.
- Norgaard B.L.,
- Gaur S.,
- Leipsic J.,
- et al.
- National Institute for Health and Care Excellence
- Min J.K.,
- Berman D.S.,
- Dunning A.,
- et al.
- Shaw L.J.,
- Berman D.S.,
- Maron D.J.,
- et al.
- Douglas P.S.,
- Pontone G.,
- Hlatky M.A.,
- et al.
- Douglas P.S.,
- De B.B.,
- Pontone G.,
- et al.
- Curzen N.P.,
- Nolan J.,
- Zaman A.G.,
- Norgaard B.L.,
- Rajani R.
- Kitabata H.,
- Leipsic J.,
- Patel M.R.,
- et al.