Comparison between the performance of quantitative flow ratio and perfusion imaging for diagnosing myocardial ischemia

,

T he landmark FAME (Fractional Flow Reserve versus Angiography for Multivessel Evaluation) trial demonstrated fractional flow reserve (FFR)Àguided revascularization to be superior in terms of patient outcome compared with an angiography-based approach (1).As such, guidelines advocate the use of FFR measurements to detect hemodynamically significant coronary artery disease (CAD) (2).However, FFR usage is limited to intervention centers as it requires the use of an intracoronary pressure wire.Quantitative flow ratio (QFR) is a technique that estimates FFR based on a 3-dimensional (3D) coronary model reconstructed from cine contrast images obtained during invasive coronary angiography (ICA) and fast computational fluid dynamics that obviate the need for intracoronary pressure wires or hyperemic conditions (3).The FAVOR-II (Functional Diagnostic Accuracy of Quantitative Flow Ratio in Online Assessment of Coronary Stenosis) trials demonstrated that QFR accurately depicted lesionspecific, FFR-defined significant CAD (4,5).Besides invasive diagnostic tools, myocardial perfusion imaging (MPI) with either single-photon emission computed tomography (SPECT) or positron emission tomography (PET) allows for the functional assessment of CAD in a noninvasive manner (6).An invasive strategy by ICA with functional testing (e.g., QFR) and a noninvasive approach using MPI are both viable diagnostic options for patients with suspected CAD (2).Currently, a head-to-head comparison between the diagnostic value of QFR and MPI when referenced by FFR is lacking.

METHODS
STUDY POPULATION.This is a substudy of the prospective comparison of coronary CT angiography, SPECT, PET, and hybrid imaging for the diagnosis of ischemic heart disease determined by FFR (PACIFIC [Comparison of Cardiac Imaging Techniques for Diagnosing Coronary Artery Disease]; NCT01521468) (7).A total of 208 patients without a history of CAD (i.e., previous coronary revascularization or myocardial infarction) but suspected of having CAD underwent coronary CT angiography, SPECT, and PET before ICA with interrogation of all major coronary arteries by FFR, regardless of stenosis severity.For the present study, QFR computation was attempted in all coronary arteries in which FFR was obtained.
The PACIFIC trial was approved by the institutional Medical Ethics Committee and complied with the Declarations of Helsinki.All participants provided written informed consent.
SPECT.SPECT scans were obtained using a dual-head hybrid SPECT/CT machine (Symbia T2, Siemens Medical Solutions, Erlanger, Germany).As previously described, imaging entailed a 2-day stress (intravenous adenosine 140 mg/kg/min) and rest protocol using a weight-adjusted dose of 370 to 550 MBq of 99 mTc tetrofosmin as a radiopharmaceutical (7).Uptake images were acquired using elec- PET.[ 15 O]H 2 O PET perfusion images were acquired on a hybrid PET/CT device (Philips Gemini TF64, Philips Healthcare, Best, the Netherlands) as previously published (7).Absolute myocardial blood flow in ml/min/g was obtained using a dynamic rest and stress protocol using intravenous adenosine (140 mg/kg/min) as a hyperemic agent and 370 MBq of [ 15 O]H 2 O as a radioactive tracer.Vascular territories were defined according to the standardized 17-segment model of the American Heart Association (8).A blinded core laboratory (Turku University Hospital, Turku, Finland) studied reconstructed PET scans for the presence of ischemia.A hyperemic myocardial blood flow of #2.3 ml/min/g in at least 2 adjacent segments within 1 vascular territory was used to define myocardial ischemia (7).After visual assessment, coronary arteries were interrogated by FFR, regardless of stenosis severity, except for occluded or subtotal lesions in which wire passage was not deemed feasible by the operator.
FFR was calculated as the ratio of mean distal intracoronary pressure and mean arterial pressure.A lesion with an FFR #0.80 was deemed significant.

THREE-DIMENSIONAL QUANTITATIVE CORONARY
ANGIOGRAPHY AND QFR COMPUTATION.QFR computation was performed by a blinded core laboratory (ClinFact Medis Specials bv., Leiden, the Netherlands) using the QAngio XA 3D/QFR V1.2 solution software package (Medis Medical Imaging Systems bv., Leiden, the Netherlands).An end-diastolic frame of 2 projections at least 25 apart from the same coronary artery was used to reconstruct a 3D model of the vessel.This model allowed for 3D quantitative coronary angiography (3D-QCA), which resulted in anatomical lesion information such as diameter stenosis percentage and lesion length.The reference vessel was constructed by fitting the healthy segments proximally and distally to the lesion of interest.An estimation of the contrast velocity was obtained by frame count analysis that indicated the frames where contrast entered and exited the analyzed part of the vessel.The estimated contrast velocity was subsequently converted into a virtual hyperemic flow velocity that allowed for computation of the pressure drop along the vessel, which permitted QFR reading at any point along the vessel.For the present study, QFR computation was attempted in all vessels with documented FFR.Similar to FFR, a QFR of #0.80 was deemed significant.Furthermore, lesions with a diameter stenosis of 30% to 90%, as defined by 3D-QCA, were considered to be of intermediate severity.van Diemen et al.

RESULTS
PATIENT AND LESION CHARACTERISTICS.A total of 624 vessels (208 patients) were evaluated for inclusion in the present study.FFR measurements were absent in 70 (11%) vessels and were therefore excluded, leaving a total of 552 vessels in which QFR computation was attempted (Figure 1).ICA was obtained without using a dedicated QFR acquisition protocol.Therefore, issues related to ICA images (e.g., panning, foreshortening, or vessel overlap) were the main drivers of failure to compute QFR.
Finally, QFR analysis was successful in 286 (52%) vessels among 169 patients.PET imaging was not completed due to claustrophobia in 2 of these patients, whereas SPECT scanning could not be performed due to technical difficulties in 1 patient.
Patient and angiographic characteristics are presented in Tables 1 and 2. Of the vessels included, 173 (60%) exhibited an intermediate lesion as defined by 3D-QCA.FFR and QFR had skewed distributions, with medians of 0.93 (interquartile range: 0.84 to 0.97) and 0.96 (interquartile range: 0.84 to 1.00), respectively.A lesion with an FFR below the cutoff that defined ischemia was present in 21% of the vessels.
OVERALL DIAGNOSTIC PERFORMANCE OF QFR AND MPI.The overall diagnostic performance of QFR, SPECT, and PET is presented in Table 3, whereas diagnostic performance of SPECT and PET among included and excluded vessels is displayed in Supplemental Table 1.QFR exhibited a significantly higher area under the curve (0.94) compared with SPECT (0.63; p < 0.001) and PET (0.82; p < 0.001) (Figure 2).

DISCUSSION
The present study was the first to compare the diagnostic performance of QFR with the MPI modalities of SPECT and PET in a head-to-head fashion against a reference of invasive FFR measurements.QFR and FFR showed good correlation and agreement.Overall, the diagnostic accuracy of QFR was higher than SPECT and PET; however, when solely considering vessels with an intermediate lesion, the accuracy of QFR was superior to SPECT, but similar to PET.
Nevertheless, comparative area under the receiveroperating characteristic curves analyses demonstrated that QFR exhibited a superior performance for diagnosing FFR-defined significant CAD compared with SPECT and PET, overall and among vessels with an intermediate lesion (Central Illustration).

DIAGNOSTIC PERFORMANCE OF QFR AND MPI.
A recent meta-analysis of 819 (969 vessels) prospectively enrolled patients demonstrated QFR to have a sensitivity of 84% and specificity of 88% when referenced by FFR (9).Sensitivity seemed to be higher among studies that computed QFR online (i.e., directly during ICA) compared with off-line computation after obtaining ICA.Online computation might lead to a more favorable QFR and FFR concordance because the wire location can be directly matched by the operator.Diagnostic values of the present study  1 and 2.
van Diemen et al.
Comparison Between Performance of QFR and MPI S E P T E M B E R 2 0 2 0 : 1 9 7 6 -8 5 were in line with the values observed in the WIFI-II (Wire-Free Functional Imaging-II Study) (sensitivity 77%, specificity 86%), which similarly analyzed QFR in an off-line fashion (10).With regard to MPI, the observed per-vessel sensitivity of SPECT (29%) and PET (75%) appeared to be lower compared with the sensitivity reported in the PACIFIC main paper (39% and 81%, respectively), whereas specificity and accuracy were similar.Importantly, the present study excluded vessels with subtotal or chronic coronary total occlusion because QFR analysis would not be clinically relevant; however, these vessels would have been correctly assessed by MPI in most cases, augmenting sensitivity (Supplemental Table 1).Nevertheless, including these vessels would not have changed the comparison between QFR and MPI because these vessels would also be classified correctly by ICA in conjunction with QFR.
QFR VERSUS MPI.QFR appears to have a superior diagnostic value for detecting FFR-defined significant CAD compared with SPECT and quantitative [ 15 PET, which is worth analyzing.The concordance between QFR and FFR is not surprising because both techniques were developed to solely assess epicardial lesion specific significance.MPI takes the whole coronary vasculature into account, assessing both epicardial stenosis and diffuse and/or small-vessel disease, which do not cause focal pressure gradients and therefore go undetected by FFR or QFR (11,12).As such, discordant invasive and noninvasive results do not necessarily depict inaccuracies of either technique, but more likely reflect the ability of the modalities to assess different stages and aspects of the atherosclerotic process (12).For example, impaired FFR with normal perfusion can be observed in patients with focal stenosis but preserved microvasculature; in contrast, normal FFR with diminished perfusion can be seen in patients with small-vessel disease (11,12).Therefore, a simplified diagnostic comparison using binary results (i.e., normal vs. abnormal) does not do justice to the complex relationship between atherosclerosis and myocardial perfusion.Nevertheless, a FFR-guided revascularization strategy is the only approach that leads to a beneficial outcome in patients with stable CAD.Therefore, determining the value of diagnostic techniques to assess the FFR revascularization threshold is of clinical importance (13).The higher accuracy of QFR compared with SPECT is driven by a lower rate of false negative findings, whereas specificity is similar.
A growing body of evidence shows SPECT has a poor sensitivity when referenced by FFR (7,14,15).Sensitivity of SPECT is hampered by the relatively low spatial resolution and unfavorable tracer kinetics, which cause subtle perfusion defects to remain undetected (6).Conversely, quantitative PET results in a higher rate of false positive findings compared with QFR, whereas sensitivity is comparable.These false positive findings are presumably caused by diffuse atherosclerosis and/or small vessel disease, which do not lead to pressure gradients but do result in lower myocardial perfusion (12).
CLINICAL APPLICABILITY.QFR enables sites that obtain ICA to functionally evaluate CAD, because it solely relies on cine contrast images.Contemporary guidelines present an invasive approach by ICA in conjunction with functional assessment (e.g., QFR), as well as a noninvasive approach using MPI (e.g., SPECT/PET) as feasible diagnostic pathways for patients with an abnormal coronary CT angiography or stroke, and even death) (16).The risk of experiencing these detrimental events does not apply to noninvasive imaging techniques, as such patient counseling before choosing a diagnostic pathway, is vital.
Another clinical scenario in which both an invasive approach and a noninvasive diagnostic strategy are feasible is the functional assessment of nonculprit lesions (NCL) in patients with ST-segment elevation myocardial infarction (17).QFR computation of NCLs based on ICA of the primary intervention demonstrates a high diagnostic accuracy when referenced by staged-FFR (18,19), comparable to the diagnostic accuracy observed among patients with stable CAD (18).In addition, immediate QFR also has good agreement with immediate FFR (accuracy: 94%) (20).
Limited data are available on the ability of noninvasive MPI to assess the functional significance of NCLs, which has not been attempted with SPECT or PET, to the best of our knowledge.However, assessment of NCLs with noninvasive MPI by cardiac magnetic resonance has been undertaken and demonstrated a  lines and the applied PET threshold is considered the optimal cutoff to discern FFR-defined significant CAD, results of the present study might differ when alternate thresholds are used (11,14,22).Third, the present study focused on the diagnostic performance in terms of sensitivity, specificity, and accuracy, which are dependent on the precision of the modalities to assess FFR but also dependent on FFR distribution within a population.Therefore, accuracy will be higher in a population with FFR values far away from the threshold and will be lower when FFR is clustered around the threshold (Supplemental Table 2).As such, the present results should be interpreted in light of the studied population, that is, patients with an intermediate probability of CAD.
Last, although the impact of individualized segmentation on the diagnostic performance of PET was negligible, discordance between the standardized American Heart Association 17-segment model used for MPI and true anatomy could not be ruled out (23).
trocardiographic gating and followed by a low-dose CT scan for attenuation correction.A blinded core laboratory (Royal Brompton Hospital, London, United Kingdom) assessed SPECT uptake images.A 17-segment model was used, in which each segment of the rest and stress scans was visually graded for the presence of a perfusion defect scored on a 5-point scale (0: normal; 1: mildly decreased; 2: moderately decreased; 3: severely decreased; 4: absence of uptake) (8).A summed difference score (SDS) was calculated by subtracting the summed rest score from the summed stress score.An SDS $2 within 1 vascular territory was considered indicative of myocardial ischemia.
ICA AND FFR.At least 2 orthogonal projections per evaluated coronary artery were obtained.Epicardial vasodilation was achieved by an intracoronaryA B B R E V I A T I O N S A N D A C R O N Y M S 3D-QCA = 3-dimensional quantitative coronary angiography CAD = coronary artery disease CT = computed tomography FFR = fractional flow reserve ICA = invasive coronary angiography ICC = intraclass coefficients MPI = myocardial perfusion imaging PET = positron emission tomography QFR = quantitative flow ratio SDS = summed difference score SPECT = single-photon emission computed tomography The authors attest they are in compliance with human studies committees and animal welfare regulations of the authors' institutions and Food and Drug Administration guidelines, including patient consent where appropriate.For more information, visit the JACC: Cardiovascular Imaging author instructions page.Manuscript received September 20, 2019; revised manuscript received December 12, 2019, accepted February 5, 2020.injection of 0.2 mg nitroglycerine before contrast injection.ICA images were obtained without adherence to a dedicated QFR acquisition protocol (i.e., no standardized views, varying magnification, collimation, and panning at the discretion of the operator).

STATISTICAL ANALYSIS.
Continuous variables are presented as mean AE SD or median (interquartile range) where appropriate, whereas categorical variables are expressed as frequencies with percentages.Diagnostic performance measures were compared using generalized estimating equations with an exchangeable working correlation structure (sensitivity, specificity, and accuracy) or independent working correlation structure (positive predictive value and negative predictive value).In addition, sensitivity, specificity, and accuracy of QFR, SPECT, and PET were compared using the paired McNemar test.A Bonferroni correction, which multiplied the p value by 2, was applied to account for multiple testing.Areas under the receiver operating characteristic curves were compared with the DeLong

FIGURE 1
FIGURE 1 Study Flowchart

9 , 2 0 2 0
Comparison Between Performance of QFR and MPI S E P T E M B E R 2 0 2 0 : 1 9 7 6 -8 5 method.Associations between QFR and FFR were quantified using Spearman's rank correlations.Agreement between QFR and FFR was assessed using intraclass coefficients (ICCs) and Bland-Altman analyses.A 2-way mixed effects model was used to determine the ICCs for single measures.Lastly, bias between QFR and FFR was assessed using paired Student's t-tests.All analyses were performed using IBM SPSS (SPSS Statistics 26, IBM, Armonk, New York) and MedCalc (MedCalc Software 11.6.0.0,Mariakerke, Belgium).

J
A C C : C A R D I O V A S C U L A R I M A G I N G , V O L .1 3 , N O .9 , 2 0 2 0 van Diemen et al. S E P T E M B E R 2 0 2 0 : 1 9 7 6 -8 5

FIGURE 2 5
FIGURE 2 Diagnostic Performance of QFR, SPECT, and PET, Overall and Among Vessels With an Intermediate Lesion

FIGURE 3 2 B 2 0 5 FIGURE 4
FIGURE 3  Correlation and Agreement Between QFR and FFR, Overall and Among Vessels With an Intermediate Lesion

J 5
A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 1 3 , N O .9 , 2 0 2 0 van Diemen et al. S E P T E M B E R 2 0 2 0 : 1 9 7 6 -8 Comparison Between Performance of QFR and MPI vessel accuracy of 75% when referenced to staged FFR (21).Prospective comparative studies are necessary to determine whether QFR or MPI has superior diagnostic value and improves patient outcome when assessing NCLs in the setting of ST-segment elevation myocardial infarction.STUDY LIMITATIONS.First, ICA in the PACIFIC trial was obtained without adherence to a dedicated QFR acquisition protocol; therefore, QFR could not be analyzed in 48% of the vessels, which hampered a per-patient and intention-to-diagnose analysis.Furthermore, in general, QFR computation is not validated in patients with atrial fibrillation, bifurcation lesions with a medina 1,1,1 classification, ostial left main or ostial right coronary artery stenosis, and grafted arteries.Second, although the definition of ischemia on SPECT is based on international guide-

2 0
82 (0.75-0.88)QFR: 0.94 (0.91-0.97) van Diemen, P.A. et al.J Am Coll Cardiol Img.2020;13(9):1976-85.The ability of quantitative flow ratio (QFR) (#0.80) to assess fractional flow reserved (FFR)Àdefined significant coronary artery disease (FFR #0.80) was compared with single-photon emission computed tomography (SPECT) (summed difference score $2) and [ 15 O]H 2 O positron emission tomography (PET) (hyperemic perfusion #2.3 ml/min/g in at least 2 adjacent segments).Patients (n ¼ 208) underwent SPECT and PET before invasive coronary angiography in conjunction with FFR.QFR computation was retrospectively attempted and successful in 286 (52%) of the vessels in which FFR was obtained (552 vessels); of these vessels, 60% had a lesion of intermediate severity.The areas under the curve (AUC) demonstrate QFR exhibited a superior diagnostic performance in comparison to SPECT and PET for assessing FFR-defined ischemia.CI ¼ confidence interval; CX ¼ left circumflex artery; LAD ¼ left anterior descending artery; RCA ¼ right coronary artery.van Diemen et al.J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 1 3 , N O .9 , 2 0 Comparison Between Performance of QFR and MPI S E P T E M B E R 2 0 2 0 : 1 9 7 6 -8 5

TABLE 3
Diagnostic Performance of QFR and MPI, Overall and Among Vessels With a Lesion of Intermediate Severity Values are % (95% confidence interval).*The p value concerns the comparison with QFR.†The p value calculated using the paired McNemar-test; every p value has been multiplied by 2 to account for multiple testing.MPI ¼ myocardial perfusion imaging; NPV ¼ negative predictive value; PET ¼ positron emission tomography; PPV ¼ positive predictive value; SPECT ¼ single-photon emission computed tomography; other abbreviations as in Tables