Clinical Quantification of Myocardial Blood Flow Using PET: Joint Position Paper of the SNMMI Cardiovascular Council and the ASNC

Writing Group
Venkatesh L. Murthy (cochair)*
Timothy M. Bateman†
Rob S. Beanlands‡
Daniel S. Berman§
Salvador Borges-Neto∥
Panithaya Chareonthaitawee¶
Manuel D. Cerqueira#
Robert A. deKemp‡
E. Gordon DePuey**
Vasken Dilsizian††
Sharmila Dorbala‡‡
Edward P. Ficaro§§
Ernest V. Garcia∥∥
Henry Gewirtz¶¶
Gary V. Heller##
Howard C. Lewin***
Saurabh Malhotra†††
April Mann‡‡‡
Terrence D. Ruddy‡
Thomas H. Schindler§§§
Ronald G. Schwartz∥∥∥
Piotr J. Slomka§
Prem Soman¶¶¶
Marcelo F. Di Carli (cochair)‡‡

1. *Frankel Cardiovascular Center, Division of Cardiovascular Medicine, Department of Internal Medicine, University of Michigan, Ann Arbor, MI; ^†Mid America Heart Institute, Kansas City, MO; ^‡National Cardiac PET Centre, Division of Cardiology, University of Ottawa Heart Institute, Ottawa, Ontario, Canada; ^§Departments of Imaging and Medicine, Cedars-Sinai Medical Center, Los Angeles, CA; ^∥Division of Nuclear Medicine, Department of Radiology, and Division of Cardiology, Department of Medicine, Duke University School of Medicine, Duke University Health System, Durham, NC; ^¶Department of Cardiovascular Medicine, Mayo Clinic, Rochester, MN; ^#Department of Nuclear Medicine, Cleveland Clinic, Cleveland, OH; **Division of Nuclear Medicine, Department of Radiology, Mt. Sinai St. Luke’s and Mt. Sinai West Hospitals, Icahn School of Medicine at Mt. Sinai, New York, NY; ^††Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, Baltimore, MD; ^‡‡Cardiovascular Imaging Program, Brigham and Women’s Hospital, Boston, MA; ^§§Division of Nuclear Medicine, University of Michigan, Ann Arbor, MI; ^∥∥Department of Radiology and Imaging Sciences, Emory University, Atlanta, GA; ^¶¶Massachusetts General Hospital and Harvard Medical School, Boston, MA; ^##Gagnon Cardiovascular Institute, Morristown Medical Center, Morristown, NJ; ***Cardiac Imaging Associates, Los Angeles, CA; ^†††Division of Cardiovascular Medicine, Jacobs School of Medicine and Biomedical Sciences, University at Buffalo, Buffalo, NY; ^‡‡‡Hartford Hospital, Hartford, CT; ^§§§Division of Nuclear Medicine, Department of Radiology, Johns Hopkins School of Medicine, Baltimore, MD; ^∥∥∥Cardiology Division, Department of Medicine, and Nuclear Medicine Division, Department of Imaging Sciences, University of Rochester Medical Center, Rochester, NY; and ^¶¶¶Division of Cardiology, Heart and Vascular Institute, University of Pittsburgh Medical Center, Pittsburgh, PA

Expert Content Reviewers
Andrew Einstein###
Raymond Russell****
James R. Corbett††††
SNMMI Cardiovascular Council Board of Directors, and ASNC Board of Directors

1. ^###Division of Cardiology, Department of Medicine, and Department of Radiology, Columbia University Medical Center and New York–Presbyterian Hospital, New York, NY; ****Warren Alpert Medical School, Brown University, Providence, RI; and ^††††Frankel Cardiovascular Center, Division of Cardiovascular Medicine, Department of Internal Medicine, and Division of Nuclear Medicine, Department of Radiology, University of Michigan, Ann Arbor, MI

Preamble

Radionuclide myocardial perfusion imaging (MPI) is among the most commonly performed diagnostic tests in cardiology. Although the diagnostic and prognostic applications of radionuclide MPI are supported by a wealth of observational and clinical trial data, its performance is limited by two fundamental drawbacks. First, conventional MPI by SPECT and PET measures relative perfusion, that is, the assessment of regional myocardial perfusion relative to the region with the highest perfusion tracer uptake. This means that a global reduction in myocardial perfusion (“balanced” reduction of flow) may remain undetected and that, in general, the extent of coronary artery disease (CAD) is underestimated, as has been demonstrated with both ²⁰¹Tl- and ^99mTc-labeled perfusion tracers.1,2,3 For example, Lima et al. found that in patients with severe 3-vessel CAD, ^99mTc-sestamibi SPECT MPI showed perfusion defects in multivessel and typical 3-vessel-disease patterns in only 46% and 10% of patients, respectively.2 Similarly, it has been reported that only 56% of patients with left main CAD are identified as being at high risk by having more than 10% of the myocardium abnormal on stress SPECT MPI.4 Second, the ^99mTc flow tracers available for SPECT MPI are inherently limited by a relatively low first-pass extraction fraction at high flow rates, thus limiting the precision and accuracy of these tracers for estimation of regional myocardial blood flow (MBF) during stress.5 Clinical studies have shown that even small differences in extraction fraction can result in a clinical difference in the detection and quantification of myocardial ischemia by SPECT.6,7

These drawbacks of SPECT are addressed by PET, with its ability to quantify global and regional MBF (in mL/minute/g of tissue), assess regional perfusion abnormalities with relative MPI, and assess contractile function abnormalities and chamber dimensions with gated imaging. The purpose of this document is, first, to consolidate and update technical considerations for clinical quantification of MBF and myocardial flow reserve (MFR) from earlier documents8 and, second, to summarize and update the scientific basis for their clinical application.9,10

Technical Considerations

Perfusion Tracers

The available PET tracers for conventional MPI and quantitative MBF imaging are shown in Table 1. The most commonly used tracers are ⁸²Rb-chloride and ¹³N-ammonia, with a small number of centers worldwide using ¹⁵O-water. ¹⁸F-flurpiridaz is currently under investigation, with one phase III trial completed and a second trial awaiting initiation. Because of their short half-lives, ¹³N-ammonia and ¹⁵O-water require an on-site cyclotron. In contrast, ¹⁸F-flurpiridaz, because of its longer isotope half-life (~2 h), can be produced at regional cyclotron or radiopharmacy facilities and distributed as a unit dose. ⁸²Rb has a short half-life and is produced from an ⁸²Sr/⁸²Rb generator lasting 4–8 weeks,11,12 depending on initial activity and desired radiotracer activity. The short half-lives of ⁸²Rb and ¹⁵O-water enable fast rest–stress imaging protocols (~20–30 minute), but count statistics and standard MPI quality can be limited by the rapid isotope decay. ⁸²Rb also has a long positron range, but this does not limit the achievable spatial resolution in practice, because of image reconstruction post filtering and cardiorespiratory motion. The radiation effective dose (mSv/GBq) is an order of magnitude lower for the short-lived isotopes than for ¹⁸F-flurpiridaz; however, the dose absorbed by the patient can be lowered by reducing the total injected activity at the expense of longer imaging times for conventional MPI.

Table 1 Properties of radiotracers used for PET MBF quantification

The physiologic properties of an ideal perfusion tracer for MBF quantification would include 100% extraction from blood to tissue, and 100% retention (no washout), resulting in a linear relationship between MBF and the measured tracer activity over a wide range. The currently available PET perfusion tracers, however, have limited (< 100%) extraction and retention, resulting in a nonlinear (but still monotonic) relationship between MBF, tracer uptake, and retention rates as illustrated in Figure 1. ¹⁵O-water and ¹³N-ammonia have close to 100% initial (unidirectional) extraction over a wide range of MBF values, resulting in a tracer uptake rate (K ₁) that is close to the true MBF (Figure 1C). Rapid early washout reduces the tracer retention of ¹³N-ammonia to approximately 50%–60% at peak stress MBF values. ¹⁵O-water washes out so rapidly that there is effectively no tracer retention in cardiac tissue above the blood background level (Figure 1D). ⁸²Rb has a substantially lower extraction fraction (~35% at peak stress) and tracer retention than does ¹³N-ammonia. Although only limited data are available, ¹⁸F-flurpiridaz appears to have extraction and retention values similar to or slightly higher than those of ¹³N-ammonia.13,14 These physiologic properties of the particular perfusion tracer have a direct bearing on the optimal choice of kinetic model for image analysis and MBF quantification, as illustrated in Figure 2. Limited spatial resolution causes spillover or blurring of uptake signals from adjacent organs—an effect that varies somewhat between tracers and is a potential concern for accurate MBF quantification (Table 1).

Figure 1

Radiotracer unidirectional extraction fractions (A) used with compartmental modeling of tracer uptake rates K ₁ (C), and radiotracer retention fractions (B) used with simplified retention modeling of tracer net uptake (D). Underlying data were obtained from previous publications.14,22,221,228,229 Limited data suggest that properties of ¹⁸F-flurpiridaz are similar to those of ¹³N-ammonia. Shaded regions represent variability in reported values

Figure 2

Polar maps demonstrating MBF, uptake, and retention along with their relationship to traditional relative MPI in example using ¹³N-ammonia. Uptake of tracer is determined by local MBF. However, because most PET tracers have incomplete extraction at higher MBFs, tracer uptake in high-MBF regions may be reduced (note that intense red regions on MBF image are less intense on uptake image). Furthermore, tracer retention is usually limited in high-MBF regions. Consequently, contrast between high- and low-MBF regions is further reduced on retention images. Standard myocardial perfusion images are produced by normalizing retention images such that regions of greatest retention are scaled to 100%. This does not restore contrast between defect and normal regions. MBF quantification restores contrast and adds absolute scale (mL/minute/g)

Scanner Performance

Contemporary PET scanners operate in 3-dimensional (3D) acquisition mode, as opposed to the older 2-dimensional (2D) (or 2D/3D) systems that were constructed with interplane septa designed to reduce scatter. 3D systems generally require lower injected activity, with a concordant reduction in patient radiation effective dose. For the short-lived tracers ⁸²Rb and ¹⁵O-water, injected activities of as high as 2220–3330 MBq (60–90 mCi) were commonly used with 2D PET systems. However, this amount of activity will cause detector saturation on 3D PET systems; therefore, the injected activity must be reduced to avoid these effects.15 Weight-based dosing may help to provide consistent image quality and accurate MBF quantification, but the maximum tolerated activity can vary greatly between 3D PET systems.15 Careful consideration should be given to optimizing injected doses to avoid detector saturation during the blood pool first-pass uptake phase while also preserving sufficient activity in the tracer retention (tissue) phase to allow high-quality images for MPI interpretation. The ultrashort half-life of ⁸²Rb is particularly challenging in this regard (Figure 3). Importantly, detector saturation will generally result in falsely elevated MBF assessments due to underestimation of the blood input function. Newer solid-state detectors should further increase the dynamic range of 3D PET systems, reducing the need to trade off MPI quality for MBF accuracy.

Figure 3

Decay of typical 370-MBq (10 mCi) dose of ¹³N-ammonia (solid black line) and 1,665-MBq (45 mCi) dose of ⁸²Rb (dash line). Because of the ultrashort half-life of ⁸²Rb, higher activities must be administered to ensure reasonable counting rates during delayed tissue-phase imaging (blue region) for generation of gated and static images for MPI interpretation. However, this results in high counting rates during blood-pool phase (green region) and the potential for detector saturation. Actual threshold for detector saturation will vary with scanner performance

Image Acquisition and Analysis

Quantification of MBF requires accurate measurement of the total tracer activity transported by the arterial blood and delivered to the myocardium over time. Measurements of arterial isotope activity versus time (time–activity curves) are typically acquired using image regions located in the arterial blood pool (e.g., left ventricle, atrium, or aorta). As only the tracer in plasma is available for exchange with the myocardial tissues, whole-blood–to–plasma corrections may be required to account for tracer binding to plasma proteins, red blood cell uptake, hematocrit, and appearance of labeled metabolites in the blood. For example, ¹³N-labeled metabolites (urea, glutamine, glutamate) accumulate in the blood and account for 40%–80% of the total activity as early as 5 minute after injection of ¹³N-ammonia.16

With older 2D PET systems, a single static scan may be adequate for accurate integration of the blood time–activity data,17 because dead-time losses and random rates are low and change relatively slowly over time. However, with current 3D PET systems, dead-time losses and random rates are much higher and more rapidly changing during the bolus first-pass transit; therefore, dynamic imaging with reconstruction of sequential short time-frames is typically required for accurate sampling and integration of the arterial blood activity. Some standardization of image acquisition and reconstruction protocols for accurate MBF quantification has occurred, but it is not universally applied. Dynamic frame-rates typically vary from 5 to 10 s during the first-pass transit through the heart and from 1 to 5 minutes during the later tissue phase. Minimal postreconstruction smoothing should be applied on the dynamic image series. Excess filtering increases adjacent organ spillover effects and can bias the MBF measurements.

In practice, list-mode acquisition is recommended because it allows flexibility in the timing and reconstruction of dynamic images for MBF, static images for MPI, electrocardiography-gated images for left ventricular ejection fraction, and respiration-gated images for quality assurance assessment of breathing artifacts. Further discussion can be found in the “Image Acquisition and Reconstruction Parameters” section. Scatter from intense or focal activity near the edge of the field of view can also bias the 3D scatter correction, leading to artifacts.18 Therefore, when using 3D PET, it is important to flush the tracer injection line with a volume of saline high enough to clear the tracer activity out of the cephalic, axillary, and subclavian veins.

To estimate MBF from dynamic PET images, time–activity curves are fit to a mathematic model describing the tracer kinetics over time.19 Various models have been proposed and evaluated, but the two most commonly used for ⁸²Rb and ¹³N-ammonia are the 1-tissue-compartment model20 and the simplified retention model.17 Both models have the same conceptual property of normalizing the late-phase myocardial activity to account for the total amount of tracer that was delivered by the arterial blood. An example analysis of a 1-tissue-compartment model is shown for a stress ¹³N-ammonia PET scan in Figure 2. The MBF polar map is estimated using an assumed tracer-specific unidirectional extraction fraction dependent on MBF (EF = 1 − e^−PS/MBF, where PS is the permeability–surface area product) and the measured uptake rate constant (K ₁/EF = MBF), as well as regional corrections for total blood volume (TBV) and partial-volume underestimation (1 − TBV) of the myocardial activity.

The simplified retention model can be considered as a special case of the 1-tissue-compartment model (neglecting the effects of tracer washout), in which case MBF must be estimated using the assumed tracer retention fraction (RF), together with the late-phase tissue activity (retention) measured after the first-pass transit (retention/RF = MBF). As shown in Figure 1, the extraction and retention fractions for ⁸²Rb are fairly similar, whereas the extraction of ¹³N-ammonia is much higher (near unity) than the myocardial retention. The effects of tracer extraction, washout, and retention on image contrast in abnormally perfused myocardium (defects) are illustrated in Figure 2. A further simplification has been proposed to measure an index of stress–rest MFR using ¹⁸F-flurpiridaz SUVs only.14 SUVs are unitless and measured simply as the late-phase myocardial activity divided by the total injected dose/kg of body weight. This method still requires additional validation but could simplify the stress–rest protocols substantially by removing the need for first-pass transit imaging and tracer kinetic modeling analysis.

Under resting conditions, autoregulation of myocardial tissue perfusion occurs in response to local metabolic demands. Resting MBF has been shown to vary linearly according to the product of heart rate and systolic blood pressure.21 Adjustment of resting MBF to account for changes in the heart rate–pressure product (RPP) should be considered as part of the interpretation of stress–rest MFR values, which can otherwise appear abnormal despite adequate stress MBF. Adjusted values are computed as MBF_ADJ = MBF_REST/RPP_REST × RPP_REF, where RPP_REF is a reference value such as 8500 reported for a typical CAD population (discussed in detail in the “Resting MBF” section).22 Interpretation of the stress MBF together with the MFR is a complementary method to account for the confounding effects of resting hemodynamics on measured MFR.23

To ensure accurate estimates of MBF and MFR, it is critical to verify that each dynamic series is acquired and analyzed correctly, with thorough review of quality assurance information as illustrated in Figure 4. Dynamic time–activity curves must include at least one background (zero-value) frame to ensure adequate sampling of the complete arterial blood input function. Assessment and correction of patient motion between the first-pass transit phase and the late-phase myocardial retention images are essential, as this can otherwise introduce a large bias in the estimated MBF values.24 The peak height of blood pool time–activity curves at rest and stress should be comparable (or slightly lower at stress) if similar radiotracer activities are injected. If there are substantial differences, extravasation or incomplete delivery of tracer may have occurred and may result in inaccurate MBF estimates (Figure 5). The shape of the blood input function should also be standardized as much as possible (e.g., 30-s square wave), as variations in tracer injection profile have been shown to adversely affect MBF accuracy25 and test–retest repeatability, in particular when using the simplified retention model.26 Blood pool time–activity curves should also be visually examined for multiple peaks or broad peaks, which may suggest poor-quality injections due to poor-quality intravenous catheters, arm positioning, or other factors. Goodness-of-fit metrics such as residual χ ² and coefficient of determination, R ², should be consistently low and high, respectively. Standardization of software analysis methods has been reported for ¹³N-ammonia27 and ⁸²Rb,28,29,30 but significant variation remains among some vendor programs. Further standardization of image acquisition and analysis methods will have the benefit of allowing reliable pooling of MBF data as part of large, multicenter clinical trials.

Figure 4

Example ⁸²Rb stress PET study quality assurance for PET quantification of MBF, including orientation of left ventricular long axis (A), sampling of myocardium and arterial blood regions (B), motion detection, dynamic time–activity curves and kinetic modeling curve-fit (C), regional MBF (FLOW) and total blood volume (TBV) maps, as well as χ ² and R ² goodness-of-fit metrics (D)

Figure 5

Test-retest dynamic ⁸²Rb PET MBF scans acquired at 3 and 13 minute after dipyridamole stress. Typical injection profile (A) is shown with single peak of blood input curve (red) at ~30 s after scan start time. Poor-quality injection profile (B) shows delayed rise and double-peak of blood input curve, suggesting partial obstruction of intravenous line during tracer administration. Tracer uptake curves (dark blue) and polar maps (activity) are similar after 3–6 min, suggesting that full ⁸²Rb dose was eventually delivered. However, inconsistent curve shapes result in markedly different MBF estimates (3.7 vs. 2.3 mL/minute/g), as derived from blood-pool-spillover- and partial-volume–corrected tissue curves (cyan)

Key Points

• Accurate and reproducible quantification of MBF is possible with both ¹³N-ammonia and ⁸²Rb (both of which are Food and Drug Administration–approved).

• Consistent tracer injection profiles improve the reproducibility of MBF measurements.

• The administered dose must be adjusted to avoid detector saturation during the blood pool phase, which can be particularly challenging with ⁸²Rb.

• List-mode acquisition enables reconstruction of static, gated, and dynamic datasets. Dynamic datasets are used for blood flow quantification with compartmental modeling.

Protocols

Planning or Protocoling

This important step optimizes image quality, diagnostic accuracy, and safety. A personalized protocol for each patient considers the clinical history, reason for the test, patient preferences, and contraindications for stress agent. Reproducibility of the stress agent is critical for quantitative MBF studies to evaluate disease progression or response to therapy and requires the same stress agent, radiotracer, and software program.

Stress Test Procedure

The choice of hyperemic stress protocol is an important consideration for measurement and interpretation of MFR. Pharmacologic stress is generally required for MBF imaging because dynamic first-pass images must be acquired with the patient on the scanner bed. Although exercise stress may be preferred in some patients because of the added prognostic value of exercise capacity and electrocardiographic changes, the measured increase in rest-to-stress MBF is generally lower with exercise than with pharmacologic stress using adenosine, regadenoson, or dipyridamole. Exercise stress also reduces uptake by and spillover from adjacent organs such as the stomach and thus could reduce a potential source of artifact from MBF measurements. The use of supine bicycle exercise MBF imaging has been reported, but some detrimental effects of patient body motion may be expected. Further, this approach may be difficult to implement with the current generation of PET/CT scanners with longer imaging gantries.

Patient preparation for pharmacologic stress with PET is the same as for ^99mTc SPECT MPI.31 Patients fast for a minimum of 4 hour, avoid smoking for at least 4 hour, and avoid caffeine intake for at least 12 hour before vasodilator stress.32,33,34 Vasodilator stress with adenosine,35 dipyridamole,36 and regadenoson37 has been evaluated using ¹³N-ammonia, ⁸²Rb-chloride, and ¹⁵O-water. After excluding contraindications, a stress agent is infused on the basis of standard protocols (Table 2). The timing of isotope injection varies for each stress agent. There is no advantage to using modified protocols such as high-dose dipyridamole or hand grip (attenuated hyperemic MBF) during dipyridamole stress and MBF imaging with PET.38 If vasodilator stress is contraindicated, dobutamine combined with atropine stress is an alternate and provides maximal hyperemia equivalent to that with dipyridamole,39,40,41 although there are some data indicating the contrary.42,43,44 Hyperemia from pharmacologic stress may be reversed for significant ischemic electrocardiography changes or symptoms about 3–4 minute after the start of imaging, without jeopardizing quantitative MBF information.

Table 2 Stress pharmaceuticals used in PET MPI

Imaging Protocols

Typically, rest imaging is followed by stress imaging on the same day. Stress-first or stress-only imaging is feasible, but it is not routine practice with quantitative PET. Although several studies have suggested that peak hyperemic MBF is superior to MFR,45,46,47 most studies have concluded that MFR is more powerful for risk stratification,48,49,50,51,52,53 perhaps because of decreased variability compared with peak hyperemic MBF.54 Whether postischemic stunning affects resting MBF with stress-first imaging has not been well studied. Importantly, if regadenoson is used, reversal with 150 mg of aminophylline may not be sufficient to restore resting perfusion conditions.55 More data are needed before a transition to routine stress-only imaging for quantitative PET MBF imaging can be recommended.

Radiotracer Protocols

Table 1 lists doses of clinically used PET radiotracers for MBF imaging. The “Perfusion Tracers” section covers radiotracer properties in greater detail. Adjustment of injected activity for patient weight, body mass index, or attenuation is preferable to optimize trade-offs between the quality of delayed images and the potential for detector saturation with 3D PET. Use of automatic injectors will facilitate uniform delivery of the radiotracer and standardize the input function for MBF quantitation. Consistent tracer injection profiles may have advantages for reliable quantification of MBF,26 although additional clinical data will be helpful.25

Image Acquisition and Reconstruction Parameters

Images are acquired and reconstructed using standard vendor-specific parameters. Briefly, after low-dose CT or a radionuclide localizing scan to position the heart, a dynamic or preferably list-mode acquisition is obtained in 2D or 3D mode. List-mode acquisition provides comprehensive data for static images, gated images for left ventricular volumes and ejection fraction, and dynamic images for MBF quantitation. It is important to keep the patient positioned consistently between the transmission and emission scans. Misalignment of the attenuation CT and PET emission images, potentially exacerbated by patient and respiratory motion during hyperemic stress, may introduce moderate to severe artifacts56 in as many as 1 in 4 studies and can result in significant changes in MBF quantification.57 Camera vendors offer software to manually confirm and adjust alignment of the retention-phase PET images with the attenuation CT scan during image reconstruction. However, patient motion during the first-pass transit can produce inconsistent alignment of the dynamic image series, leading to attenuation artifacts and severe bias in MBF.24 Differences in reconstruction methods may have a substantial impact on measured MBF,58 and standardization is critical. Iterative reconstruction per manufacturer recommendations is preferred for dynamic image series. Minimal smoothing of the images is preferred for MBF quantitation.

Key Points

• To estimate MFR, maximal hyperemia is usually induced with dipyridamole, adenosine, or regadenoson.

• Typical imaging protocols for quantitative PET imaging involve rest imaging followed by stress imaging on the same day, although stress-only protocols may have a role.

• Quality control of dynamic images and time–activity curves is essential and should include inspection for emission–transmission misregistration, patient motion, and evidence of detector saturation.

Preferred Nomenclature and Physiologic Reference Ranges

Nomenclature

A variety of terms have been used in the quantitative PET literature, including coronary flow reserve (CFR), MFR, MBF reserve, and myocardial perfusion reserve. Additionally, in the invasive and echocardiography literature, coronary flow velocity reserve is used. Finally, relative quantification of increased perfusion, without formal quantification of underlying MBF at rest and stress, has been referred to as myocardial perfusion reserve index in the cardiac MRI literature and has more recently been applied to quantification of SPECT images. The use of many different terms in the literature has the potential for confusion. Going forward and for this document, the preferred nomenclature is to refer to quantitative measures at rest or stress as MBF and the ratio of stress/rest MBF as MFR. Although this value generally correlates well with invasively determined CFR,59,60,61,62,63,64 PET methods do not measure volume of blood flow in the epicardial coronary arteries directly but rather blood flow in myocardial tissue. Thus, the term MFR is more appropriate. The standard units of MBF are milliliters·minute⁻¹·gram⁻¹, most commonly denoted as mL/minute/g.

Resting MBF

Resting MBF as measured with PET and various positron-flow radiotracers has been reported to range from 0.4 to 1.2 mL/minute/g.65,66,67,68,69,70,71 Apart from methodologic differences in radiotracer characteristics, tracer kinetic models, and image analysis that may introduce some variations between different studies, the variability of the reported resting MBF values may be attributed in part to differences in myocardial workload and thus the myocardial oxygen demand of the left ventricle.66,67,72,73,74 Sex and genetic variations, including mitochondrial function, are also important determinants contributing to the variability in resting MBF values.75

MBF at rest and during some forms of stress is physiologically coupled with myocardial oxygen demand and thus correlates with indices of myocardial workload (e.g., rate–pressure product, defined as the product of systolic blood pressure and heart rate).76,77,78 Consequently, resting MBF is commonly higher in patients with higher arterial blood pressure or heart rate.67,70,79,80 Age-related increases in resting MBF can be explained by rate–pressure product correction of increased systolic blood pressures.67,81 Most of the reported PET-determined resting MBF values have been higher in women than in men.66,68,82,83 Although the causes of this sex difference are not completely defined, hormonal effects on coronary circulatory function in women with CAD, and sex-dependent lipid profile changes, may be important contributors.66,68,82,83 Finally, in individuals with advanced obesity, resting MBF may also be elevated as induced by a more enhanced activation of the sympathetic nervous system and the renin–angiotensin–aldosterone axis, resulting in higher resting heart rate and arterial blood pressure.70,83,84

Physiologic Ranges for MBF and MFR with ¹³N-Ammonia and ⁸²Rb-Chloride

In 23 studies involving a total of 363 healthy subjects undergoing ¹³N-ammonia PET, the weighted mean MBF values at rest and stress were 0.71 mL/minute/g (range 0.61–1.1) and 2.58 mL/g/minute (range 1.86–4.33), respectively (Table 3). Weighted mean MFR was 3.54 (range 3.16–4.8). The corresponding values for 382 healthy subjects from 8 studies using ⁸²Rb PET are a weighted mean resting MBF of 0.74 mL/g/minute (range 0.69–1.15), a weighted mean stress MBF of 2.86 mL/g/minute (range 2.5–3.82), and a weighted mean MFR of 4.07 (range 3.88–4.47) (Table 4). It is critical to realize that these values represent physiologic ranges derived from young, healthy volunteers without coronary risk factors. In clinical populations, which are generally older and have a substantial burden of coronary risk factors, values below these ranges may frequently be seen and may not represent obstructive epicardial CAD. Instead, modest reductions in stress MBF or MFR below these reference ranges are often due to the effects of diffuse CAD and microvascular disease. A detailed discussion of abnormal thresholds for reporting and clinical action is found in the “Interpretation and Reporting” section.

Table 3 MBF and MFR reference ranges for ¹³N-ammonia PET

Table 4 MBF and MFR reference ranges for ⁸²Rb PE

Key Points

• Although many terms have been used, MBF and MFR are the preferred terms for describing quantitative measures of blood flow.

• Physiologic reference ranges for rest and stress MBF and MFR vary by tracer and may be slightly higher for ⁸²Rb than for ¹³N-ammonia.

Indications and Applications

CAD Diagnosis

A relationship between the severity of epicardial coronary artery stenoses and PET measures of both peak hyperemic stress MBF and MFR has been established for more than 2 decades.85 Though initially established using ¹⁵O-water, this finding was quickly replicated using ¹³N-ammonia86,87,88 and more recently using ⁸²Rb.89,90 The application of stress MBF and MFR for improving the diagnostic accuracy of PET MPI with clinical protocols has been investigated by many groups with both ¹³N-ammonia47,48,88 and ⁸²Rb.52,91 Although these studies have consistently demonstrated improved diagnostic sensitivity (case example in Figure 6), at least 2 large studies have raised concerns about potential for decreased specificity (Figure 7),52,91 possibly due to the contributions of diffuse atherosclerosis and microvascular disease to stress MBF and MFR measurements. Consequently, the positive predictive value of even severely depressed MFR (< 1.5) is only modest.52,91 Conversely, preserved MFR (> 2.0) has an excellent negative predictive value for high-risk CAD (i.e., left main and 3-vessel disease), and high-risk disease is extremely uncommon with an MFR of more than 2.552,91 (see the “Interpretation and Reporting” section 6 for a more detailed discussion).

Figure 6

Clinical utility of blood flow quantification. In this example, from 81-y-old man with hypertension and dyslipidemia, relative MPI (A) with ⁸²Rb PET demonstrated only mild, reversible perfusion abnormality involving distribution of left anterior descending coronary artery. However, MFR was severely reduced globally at 1.11. Nearly entire heart had severely reduced MFR except for inferior and inferolateral walls, where it was only moderately reduced. Coronary angiography (B) showed severe stenosis of mid portion of left main coronary artery

Figure 7

Receiver-operator characteristic curves for detection of severe CAD using MFR. As the threshold for abnormal MFR is decreased from 3.0 to 0.5, sensitivity for high-risk CAD (2-vessel disease including proximal left anterior descending artery, 3-vessel disease, and left main coronary artery) decreases (blue line). Conversely, with lower thresholds for defining abnormal MFR, specificity progressively increases (red line). (Adapted from Naya et al.91)

Prognostic Assessment

The incremental prognostic value of PET measures of stress MBF and MFR in patients with known or suspected CAD referred for clinical stress testing has also been extensively evaluated (Table 5).46,49,50,53,92,93,94 Consistently, patients with more severely reduced stress MBF and MFR are at higher risk than patients with preserved values or modest reductions. An analysis of the relationship between MFR and cardiac mortality suggests an excellent prognosis for an MFR of more than 2 and a steady increase in cardiac mortality for an MFR of less than 2 (Figure 8).54 The largest of these studies has demonstrated that as many as half of intermediate-risk subjects may be reclassified on the basis of MFR, even after accounting for clinical characteristics, relative MPI interpretation, and left ventricular ejection fraction.95 Consequently, in patients at higher clinical risk, for whom even a low-risk relative assessment of MPI may be insufficiently reassuring (i.e., those likely to remain at intermediate posttest risk), referral for stress PET with quantification of MBF may be preferable as an initial test over relative MPI alone, such as with SPECT imaging.

Table 5 Clinical studies of prognostic value of quantitative PET blood flow estimates

Figure 8

Relationship between MFR and risk of cardiac death. Regardless of which ⁸²Rb tracer kinetic model is used, similar pattern of rising risk with MFR < 2 is seen. 1:1 indicates fictitious 100% extraction (MBF = K ₁), which approximates assumptions for myocardial perfusion reserve index. (Adapted from Murthy et al.54)

Treatment Guidance

At present there are no randomized data supporting the use of any stress imaging modality for selection of patients for revascularization or for guidance of medical therapy. Observational data have established a paradigm that patients with greater degrees of ischemia on relative MPI are more likely to benefit from revascularization.96 This paradigm has been conceptually extended to include MFR and stress MBF97 but has not yet been evaluated prospectively. Although observational data are limited to one single-center study with relatively small sample sizes, there is some evidence that early revascularization is associated with a more favorable prognosis only in patients with a low global MFR and that patients with a low MFR may benefit more from coronary artery bypass grafting than from percutaneous revascularization.98

Special Populations

Diabetes Mellitus

Patients with diabetes mellitus are at significantly increased risk of CAD and its complications.99 Furthermore, diabetic patients may have extensive, high-risk CAD even with low-risk relative MPI findings,100 and diabetic patients with low-risk relative MPI findings may still be at significantly elevated risk of CAD complications.101 Important contributors to these concerning findings may be increased rates of diffuse epicardial CAD and microvascular disease among diabetic patients. Consequently, the improved performance of quantitative measures with PET compared with relative MPI is likely to be of particular value. In a large series of 1172 patients with diabetes compared with 1611 patients without diabetes, incorporation of MFR into PET assessment allowed identification of the 40% of diabetic patients who were at high risk (at equivalent risk to those with clinically recognized CAD) compared with the remainder, who experienced event rates comparable to individuals without diabetes.102 Given the important limitations of relative MPI among diabetic patients, PET with quantification of blood flow is preferable to SPECT among patients with diabetes mellitus.

Chronic Kidney Disease

Cardiovascular disease is the leading cause of death among patients with moderate to severe renal dysfunction,103 and early referral for revascularization may be beneficial in patients with suitable disease.104 However, patients with underlying renal dysfunction are also at increased risk of complications after angiography and revascularization.105,106,107 Unfortunately, as with diabetic patients, traditional relative MPI is unable to identify truly low-risk patients.108 Two series from one center have shown that PET measures of MFR have greater prognostic value than do clinical and relative MPI parameters in patients with chronic kidney disease109 and patients requiring renal replacement therapy.110

Cardiomyopathy and Heart Failure

In many cases, relative MPI lacks sufficient negative predictive value to adequately rule out an ischemic etiology in patients with severe reductions in systolic function.4 However, patients with heart failure are also at increased risk of complications from invasive coronary angiography. Consequently, the excellent negative predictive value of preserved MFR may be of particular value in excluding severe multivessel CAD in patients with cardiomyopathy.52,91 Furthermore, in patients with both ischemic and nonischemic cardiomyopathies, impaired MFR is associated with markedly increased rates of major adverse cardiac events and cardiac death.111 However, it is important to note that abnormalities in MFR have been identified in cardiomyopathies of numerous etiologies.112,113,114,115,116 Consequently, whereas a low MFR does not necessarily imply an ischemic etiology, ischemic cardiomyopathy is extremely unlikely with well-preserved MFR. Nonetheless, the prognostic value of MFR is likely to be important regardless of etiology.111,113,116

Heart Transplantation

Patients who have undergone heart transplantation may develop coronary allograft vasculopathy (CAV), a pathologic entity distinct from atherosclerotic CAD. In CAV, intimal fibromuscular hyperplasia and intimal–medial hyperplasia cause smooth narrowing of the coronary arteries with an attendant decrease in vasodilator capacity and MBF.117,118 Because arteries are usually smoothly narrowed, traditional noninvasive diagnostic techniques such as stress SPECT MPI and stress echocardiography may be limited compared with invasive imaging of the vessel wall using intravascular ultrasound or optical coherence tomography.119,120,121,122,123 Smooth narrowing of all vessels may result in normal relative MPI findings or only modest distal perfusion deficits despite global reductions in perfusion and vasodilator capacity. Invasive measures of MFR have been related to adverse outcomes.124 PET measures of MBF or MFR have been shown to correlate with invasive measures of CAV125 and to identify patients at risk of developing CAV.126 Recently, a relatively large study of 140 patients with prior heart transplantation demonstrated that impaired MFR identified those at risk of developing clinical events.127 Indeed, investigational therapies for CAV have demonstrated an ability to improve PET measures of MFR.128 Of note, early after transplantation, decreases in MFR may not reflect early CAV,129,130 possibly because of resting hyperemia. In this early stage, stress MBF may have greater value. Despite this limitation, quantification of MBF in patients with prior heart transplantation has substantial well-established advantages over competing noninvasive methods of CAV diagnosis.

The Elderly

Older patients, by virtue of age alone, are at increased risk of mortality. However, among those of extremely advanced age, cancer rather than cardiovascular disease is the leading cause of mortality. Furthermore, whereas CAD is highly prevalent, the increased risks of invasive investigation and revascularization may shift the balance in some cases toward medical therapy rather than invasive approaches. One unpublished study has demonstrated that MFR assessment with PET may be able to identify patients aged 75 and older with excellent prognosis for survival free of cardiac death.131 Further investigation is of great interest.

Women

There is much debate in the literature132,133 over optimal strategies for evaluation of known or suspected CAD in women. An important consideration is that a sizeable proportion of symptomatic women may have no evidence of obstructive CAD but are nonetheless at increased risk of cardiac complications.134,135 In part, this may be due more to impaired vasomotor function or microvascular disease than to epicardial obstructive stenoses in women compared with men.136 PET assessment of MFR has been demonstrated to be effective in both sexes and can readily identify evidence of epicardial obstructive disease, as well as diffuse CAD and microvascular function, noninvasively.137

Chest Pain with Normal Findings on Coronary Angiography

In both men and women with CAD risk factors but without overt epicardial CAD, coronary vasomotor dysfunction is highly prevalent and can be identified with PET.137 This is likely due to the presence of diffuse disease and microvascular dysfunction and may be present even in the absence of coronary artery calcium.138 In one study of 901 patients referred for suspected CAD who had normal relative MPI results, patients with an MFR of less than 2 experienced a 5.2%/year rate of major adverse cardiac events, even with a coronary artery calcium score of zero. Consequently, assessment of MFR with PET has significant prognostic value even in patients believed to be at low risk on the basis of relative MPI.

Key Points

• Use of stress MBF and MFR for diagnosis is complex, as diabetes, hypertension, age, smoking, and other risk factors may decrease stress MBF and MFR without focal epicardial stenosis.

• Patients with preserved stress MBF and MFR are unlikely to have high-risk epicardial CAD.

• Severe reductions in global MFR (< 1.5) are associated with a substantially increased risk of adverse outcomes and merit careful clinical consideration.

• A preserved global MFR of more than 2.0 has an excellent negative predictive value for high-risk CAD (i.e., left main and 3-vessel disease).

Interpretation and Reporting

Reporting Quantitative MBF Data

One of the practical applications of measuring MBF and MFR with PET is the potential utility of these quantitative physiologic measures in improving the accuracy with which angiographic CAD is detected and its physiologic severity characterized, thereby allowing more informed decisions on referrals for cardiac catheterization and, potentially, revascularization. The decision on when and how to report MBF and MFR values in the context of MPI PET studies requires understanding of what is being measured, as well as the strengths and relative weaknesses of such physiologic parameters for clinical decision making.

The rationale for using quantitative MBF data for uncovering epicardial CAD is based on the relationship between peak hyperemic MBF and MFR and the severity of coronary lesions on coronary angiography demonstrated in experimental models of coronary stenosis139,140 and in humans with atherosclerosis.85,86,87,88 The findings of human studies that have measured MBF and MFR noninvasively by PET, as well as angiographic stenosis severity, can be summarized as follows:

• In humans, resting MBF remains relatively preserved across a wide range of coronary stenosis severity,85,86 which is largely related to the gradual autoregulatory vasodilation of resistive vessels to maintain resting myocardial perfusion in the setting of upstream stenosis. Resting MBF falls only in the presence of critical subocclusive stenosis and poorly developed collateral blood flow.

• The activation of the compensatory autoregulatory changes described above results in a progressive loss in maximum vasodilator capacity with increasing stenosis severity, which is manifested by gradual reductions in hyperemic MBF and MFR as measured by PET.85,86,87

• In general, hyperemic MBF and MFR are relatively preserved for coronary lesions with less than 70% angiographic stenosis or with preserved fractional flow reserve (FFR) (> 0.8).45,47,51,52,85,86,87,88,89,91,141,142 However, both may be reduced even in the absence of overt obstructive stenosis, especially in higher-risk subgroups (e.g., diabetes and prediabetic states,143,144,145,146,147,148,149,150,151,152,153,154 hypertension,155,156,157,158 dyslipidemia,159,160,161,162,163 and chronic kidney disease109,110,164,165).

• Hyperemic MBF and MFR are consistently reduced in lesions with greater than 70% luminal narrowing or those with abnormal FFR.45,47,51,52,85,86,87,88,89,91,141,142

• Coronary stenosis of intermediate severity (e.g., 40%–90%) is associated with significant variability in hyperemic MBF and MFR. For any degree of luminal stenosis, the observed physiologic variability is likely multifactorial and includes the following: geometric factors of coronary lesions not accounted for by a simple measure of minimal luminal diameter, including shape, eccentricity, length, and entrance and exit angles, all of which are known to modulate coronary resistance166,167; development of collateral blood flow166,167; and presence of diffuse coronary atherosclerosis and microvascular dysfunction (combination of endothelial and smooth muscle cell dysfunction in resistive vessels, and microvascular rarefaction),168 all of which are consistent findings in autopsy and intravascular ultrasound studies of patients with CAD.169,170

Is there a physiologic threshold of hyperemic MBF or MFR that can be routinely used to accurately predict obstructive stenosis on coronary angiography? The simple answer is no. The available data from the published literature include a mix of patients with suspected or known CAD (e.g., prior myocardial infarction or percutaneous coronary intervention) and used different endpoints for defining lesion severity (e.g., visual or quantitative coronary stenosis severity, angiographic risk scores, or FFR) and methodologies for measuring MBF (e.g., ¹⁵O-water, ¹³N-ammonia, or ⁸²Rb using different quantitative approaches), resulting in multiple different thresholds being proposed to improve detection of obstructive angiographic CAD. Nonetheless, there are a few areas of agreement that have potentially important practical implications for including quantitative flow data in clinical PET MPI reports:

• A preserved global hyperemic MBF and MFR consistently reduce the probability of high-risk angiographic CAD (i.e., obstructive proximal stenosis in all 3 major coronary arteries, or left main disease). A global hyperemic MBF of more than 2 mL/minute/g and MFR of more than 2 reliably exclude the presence of high-risk angiographic CAD (negative predictive value > 95%).51,52,91

• A severely reduced global hyperemic MBF and MFR identify patients at high risk for major adverse cardiovascular events, including death. Although thresholds may vary in different labs using different software, in general an MFR of less than 1.5 should be considered a high-risk feature on MPI PET46,49,50,53,92,94,98,102 and is associated with an increased likelihood for multivessel obstructive CAD.51,52,91 In these patients, angiographic evaluation may be necessary to exclude disease that can potentially by revascularized.98

• A severe reduction in hyperemic MBF (< 1.5 mL/minute/g) or MFR (< 1.5) in a single vascular territory in a patient with normal MPI PET results by semiquantitative visual analysis should raise the possibility of flow-limiting CAD.

It is important to understand that these thresholds may vary in different labs using different software, and consequently, this should be viewed as a guide. Although individual labs may adopt variations of these thresholds, the general principle that coronary anatomy may need to be defined in patients with severely reduced MFR remains important.

Hyperemic MBF, MFR, or Both?

Hyperemic MBF and MFR provide useful information on coronary vasodilator flow capacity and characterization of flow-limiting CAD. Both parameters also share the same limitation for differentiating predominant focal obstructive stenosis from diffuse atherosclerosis and microvascular dysfunction. For most patients, the information from these two parameters is concordant (normal or abnormal).171 However, in a minority of patients the information may be discordant. Since MFR is a ratio between hyperemic and resting MBF, unusually low or high resting MBF will affect MFR and result in discrepant findings compared with the hyperemic MBF value. For example, patients with prior myocardial infarction may show relatively preserved MFR in infarct-related territories because of low resting MBF. Conversely, patients with normal hyperemic MBF but unusually high resting MBF (e.g., women and heart transplant recipients) may show a relatively reduced MFR. Consequently, both parameters should be considered in the interpretation of the test results.

In studies that have examined the incremental value of MBF quantification for predicting obstructive coronary stenosis on angiography, both hyperemic MBF and MFR have performed similarly.45,47 This suggests that stress-only imaging may be effective in selected patients, especially those without known CAD and normal left ventricular ejection fraction in whom resting MPI may be unnecessary to assess defect reversibility.

From a prognostic perspective, MFR provides better incremental risk stratification than hyperemic MBF alone.50,53 Furthermore, patients on medical therapies that reduce resting MBF, such as β-blockers, may have reduced hyperemic MBF due to disease but may be asymptomatic because of adequate MFR and thus not be in need of intervention,10 further justifying the need to measure both resting and hyperemic MBF to derive the MFR.

Complementary Role of Coronary CT Angiography

The addition of coronary CT angiography can be quite helpful to differentiate patients with extensive obstructive CAD from those with predominantly microvascular dysfunction.172,173,174,175,176 The addition of CT angiography information can improve the specificity of PET, especially in the setting of abnormal MBF values.35

Special Considerations for Reporting MBF and MFR

MBF and MFR studies should be conducted and interpreted by experienced labs. The interpretation must consider the clinical context and the question being asked by the referring provider—for example, whether the question is specifically regarding myocardial ischemia, the hemodynamic significance of disease, microvascular disease, transplant vasculopathy, or some combination of these. The interpretation must also consider the findings of other imaging studies, including electrocardiography changes, coronary calcium score, and coronary anatomy (if CT angiography is performed), as well as high-risk features such as transient ischemic dilation, right ventricular uptake, and lack of augmentation of systolic function with stress.

The reporting physician needs to consider how the information will add value to the diagnostic information and potentially affect decision making so as not to lead to unnecessary testing or undertesting. Conditions known to be associated with diffuse atherosclerosis or microvascular dysfunction that would impair global MFR need to be considered, such as renal failure, prior bypass surgery, and global left ventricular dysfunction. As noted, conditions under which accurate measurement of MFR may not be possible, as in large regions of myocardial infarction, should also be considered. Because these conditions are often already associated with an increased risk of events, the added value of MBF and MFR measurements for prognostication may be limited under these circumstances (Table 6).

Table 6 Reporting MFR in clinical practice*

Special consideration must be made when there is no flow augmentation. Typically, there is some type of change even for severe MFR impairment, and the change is often heterogeneous; that is, some regions may decrease, suggesting steal, and some may increase. Likewise, such severe impairments are often accompanied by other findings, such as transient ischemic dilation, right ventricular uptake, electrocardiography changes, or regional ischemia on relative MPI. When these are not present, when perfusion appears normal, and when errors in stress-agent administration have been excluded—yet MFR is uniform at 1.0 or very close to 1.0—the possibility should be considered that the patient has ingested caffeine or is not responsive to vasodilator stress. The test may need to be repeated with a different stress agent such as dobutamine (Table 6).177,178

Key Points

• Preserved stress MBF of more than 2 mL/minute/g and MFR of more than 2 reliably exclude the presence of high-risk angiographic disease (negative predictive value > 95%) and are reasonable to report when used in clinical interpretation.

• A severely decreased global MFR (< 1.5 mL/minute/g) should be reported as a high-risk feature for adverse cardiac events but is not always due to multivessel obstructive disease. The likelihood of multivessel obstructive disease may be refined by examination of the electrocardiogram, regional perfusion, coronary calcification, and cardiac volumes and function.

• Regional decreases in stress MBF (< 1.5 mL/minute/g) and MFR (< 1.5) in a vascular territory may indicate regional flow-limiting disease.

Physiologic Relationships Among MFR, FFR, And Relative Flow Reserve

Traditionally, treatment decisions on medical therapy, percutaneous coronary intervention, or coronary artery bypass grafting have been based on the visual interpretation of the coronary angiogram, despite extensive evidence that subjective grading of luminal stenosis correlates poorly with hemodynamic significance—particularly for coronary stenoses between 30% and 80% of luminal diameter.179,180,181 Quantitative noninvasive and invasive techniques are now available that go beyond standard interpretation of anatomic coronary stenosis in making this functional assessment. These include noninvasive assessment of maximum MBF and MFR with PET, as well as invasive measurement of CFR and FFR. Noninvasive estimation of FFR using CT has also recently been described.182 Although both FFR and MFR can be used to assess the functional significance of stenosis, what they actually measure, their physiologic basis, and their clinical implications are distinct.

FFR

Invasive FFR has become a well-studied and increasingly used technique providing a surrogate measure of flow limitation and lesion-level ischemia. FFR assesses large-vessel coronary stenosis and is defined as the ratio of maximal blood flow in a stenotic artery relative to maximal flow in the same artery in the theoretic absence of any stenosis (Figure 9).183,184,185,186 FFR is calculated as the ratio of distal coronary pressure and aortic pressure, typically measured using an intracoronary pressure wire during adenosine-induced maximal hyperemia, based on the assumption that during maximal vasodilation, coronary resistance is negligible.

Figure 9

Comparison of physiologic basis of FFR and MFR. FFR is affected by focal stenosis and diffuse atherosclerosis of coronary macrocirculation, whereas index of microcirculatory resistance (IMR) reflects disease of smaller vessels. However, because intact arteriolar microcirculation is required for action of adenosine, FFR may be falsely reassuring in setting of microvascular dysfunction. MFR and CFR integrate entire coronary circulation. (Derived from De Bruyne et al.230)

An FFR of less than 0.75 was originally shown to detect reversible ischemia, defined by noninvasive stress testing (thallium SPECT and PET, dobutamine stress echocardiography, or exercise stress testing), whereas an FFR of more than 0.8 excludes ischemia with a predictive value of over 95%.184 Randomized trials—including Fractional Flow Reserve versus Angiography for Multivessel Evaluation (FAME) and FAME-2, which used an FFR cutoff point of 0.8187,188—have provided evidence that the use of FFR to guide clinical decisions on coronary revascularization results in reduced cardiac events. On the basis of these findings, the use of FFR is now incorporated into guidelines on management of patients with stable ischemic heart disease.187,188,189

FFR, however, has multiple limitations.190 In the presence of serial stenoses, a distal lesion artificially reduces the pressure gradient across the proximal lesion, leading to an overestimation of the proximal lesion’s ratio of distal coronary pressure to aortic pressure, thus underestimating its functional significance.191,192 Conversely, the presence of a proximal lesion artificially lowers this ratio for the distal lesion. Further, FFR assumes an intact microcirculation because this is the site of action of adenosine. FFR can appear falsely normal in the presence of microvascular dysfunction or disease, since elevated pressure distal to a critical stenosis, associated with increased resistance due to a microvascular abnormality, may result in a normal pressure drop across a hemodynamically significant lesion.193,194 Further, in the presence of diffuse atherosclerosis, FFR may be abnormal even without focal stenosis.195 Finally, in the setting of excellent flow capacity, the clinical significance of a reduced FFR across a moderate lesion may be overestimated if peak flow is still sufficient to meet myocardial oxygen demand. In this circumstance, symptoms are unlikely to improve with revascularization despite the reduced FFR.

More recently, the invasively measured instantaneous wave-free ratio has been advanced as a quantitative metric—which can be measured without use of a vasodilator—of the hemodynamic significance of a lesion. Although there has been only limited exploration of the relationships between the instantaneous wave-free ratio and MFR assessed by PET,196 inconsistencies between the instantaneous wave-free ratio and FFR are common.197,198,199 Nonetheless, two randomized trials have demonstrated that a strategy using an instantaneous wave-free ratio of more than 0.89 to defer revascularization yielded noninferior outcomes to a strategy using an FFR of more than 0.8.200,201

Assessments of MBF and Flow Reserve

Quantification of MBF using PET, allowing assessment of peak hyperemic MBF as well as noninvasive calculation of MFR, is physiologically distinct from FFR.202 Unlike FFR, MFR evaluates the effects of abnormality over the entire coronary circulation (Figure 9). It therefore allows assessment not only of the effects of focal epicardial coronary stenosis but also of diffuse coronary atherosclerosis and microvascular dysfunction. As discussed above, an important clinical limitation of blood flow quantitation compared with FFR is that it is difficult to distinguish abnormality due to epicardial artery stenosis from that due to diffuse atherosclerosis, microcirculatory dysfunction, or both. Relative flow reserve—the ratio of stress MBF in regions subtended by stenotic arteries to stress MBF in regions subtended by nonstenotic arteries—has been proposed as one potential solution. However, as with relative assessments of stress perfusion defects by PET, computation of relative flow reserve requires an assumed or defined normal zone for comparison.

CFR can also be measured invasively on a per-territory basis at the time of cardiac catheterization, using an intracoronary wire that assesses flow velocity.203 For invasive CFR, each vessel must be assessed separately, with repeated runs of vasodilator for maximal hyperemia. Importantly, for assessments of coronary physiology during cardiac catheterization, FFR and CFR can now be measured simultaneously with combined pressure sensor– and flow sensor–tipped guidewires.204 More recently, quantitative estimates of myocardial perfusion from Doppler echocardiography205,206 and contrast echocardiography207 have emerged as having clinical value.

Discrepancies Between FFR and MFR

The different physiologic basis of FFR and MFR measurements explains how discrepancies between FFR and assessments of MBF and MFR may arise. FFR, a lesion-based index, assumes uniform endothelial function on either side of the lesion and an intact microcirculation, whereas MBF and MFR consider the entire vascular system of the heart as a totality (Figure 9). Myocardial ischemia associated with diffuse coronary atherosclerosis or microvascular disease in the absence of significant epicardial stenosis will therefore affect MFR and FFR differently.208 Of note, a current multicenter randomized clinical trial—DEFINE-Flow (Distal Evaluation of Functional Performance with Intravascular Sensors to Assess the Narrowing Effect–Combined Pressure and Doppler Flow Velocity Measurements)—is assessing whether, in the presence of an invasive CFR of more than 2 and coronary lesions with an FFR of less than 0.80, percutaneous coronary intervention can be safely deferred.209 Estimates of the functional significance of coronary stenoses by FFR and the noninvasive or invasive CFR techniques usually agree. Concordantly normal studies imply the absence of hemodynamically significant epicardial or microvascular disease. Concordantly abnormal studies imply the presence of significant epicardial stenosis, with or without additional diffuse atherosclerotic or microvascular disease. However, a study by Johnson et al., assembling all combined invasive CFR and FFR measurements throughout the literature (a total of 438 cases), reported only a modest linear correlation between CFR and FFR (r = 0.34, P < 0.001), with 30%–40% of lesions showing discordance.210 Discordance is largely explained by the mechanisms discussed above. When the discordance is that low FFR is seen in regions with normal CFR, a flow decrement that is insufficient to cause ischemia may be the most likely cause, and percutaneous coronary intervention would be unlikely to improve symptoms. The discordance of low MFR with normal FFR is most commonly due to microvascular disease in the setting of diffuse nonobstructive epicardial disease or in isolation.193,194

Thus, both FFR and MFR provide valuable physiologic information for patient management but assess different pathophysiologic processes. Knowledge of these differences is important in understanding the frequently observed discordance between these measurements. For invasive assessment, these considerations lend impetus to increasing the use of physiologic measurements and combining the results of FFR, MFR, and stenosis for a unified interpretation. For noninvasive testing, they point to the value of combining absolute quantitative and regional assessments of perfusion with anatomic assessment—using coronary artery calcification scans or angiography (either invasive or noninvasive)—in settings in which overall clinical assessment based on the physiologic approach alone is not definitive.

Key Points

• PET MFR/invasive CFR and invasive FFR are related but are not interchangeable measures, with discordance in 30%–40% of lesions.

• MFR and invasive CFR measure the combined hemodynamic effects of epicardial stenosis, diffuse disease, and microvascular dysfunction. FFR measures the combined hemodynamic effects of focal and diffuse atherosclerosis. Microvascular dysfunction increases coronary resistance and blunts the pressure gradient across a stenosis and may sometimes lead to falsely negative FFR readings of flow-limiting lesions. The latter may explain some of the discrepancies between FFR and MFR/CFR.

Future Challenges and Conclusions

Quantification of MBF and MFR represents a substantial advance for diagnostic and prognostic evaluation of suspected or established CAD. These methods are at the cusp of translation to clinical practice. However, further efforts are necessary to standardize measures across laboratories, radiotracers, equipment, and software. Most critically, data are needed supporting improved clinical outcomes when treatment selection is based on these measures.