Coronary Artery Disease

Diagnostic Accuracy of Stress Myocardial Perfusion

Imaging Compared to Invasive Coronary Angiography
With Fractional Flow Reserve Meta-Analysis
Richard A.P. Takx, MD; Björn A. Blomberg, MD; Hamza El Aidi, MD; Jesse Habets, MD, PhD;
Pim A. de Jong, MD, PhD; Eike Nagel, MD, PhD; Udo Hoffmann, MD, MPH; Tim Leiner, MD, PhD

Background—Hemodynamically significant coronary artery disease is an important indication for revascularization.

Stress myocardial perfusion imaging is a noninvasive alternative to invasive fractional flow reserve for evaluating
hemodynamically significant coronary artery disease. The aim was to determine the diagnostic accuracy of myocardial
perfusion imaging by single-photon emission computed tomography, echocardiography, MRI, positron emission
tomography, and computed tomography compared with invasive coronary angiography with fractional flow reserve for
the diagnosis of hemodynamically significant coronary artery disease.
Methods and Results—The meta-analysis adhered to the Preferred Reporting Items for Systematic Reviews and Meta-
analyses statement. PubMed, EMBASE, and Web of Science were searched until May 2014. Thirty-seven studies,
reporting on 4721 vessels and 2048 patients, were included. Meta-analysis yielded pooled sensitivity, pooled specificity,
pooled likelihood ratios (LR), pooled diagnostic odds ratio, and summary area under the receiver operating characteristic
curve. The negative LR (NLR) was chosen as the primary outcome. At the vessel level, MRI (pooled NLR, 0.16; 95%
confidence interval [CI], 0.13–0.21) was performed similar to computed tomography (pooled NLR, 0.22; 95% CI, 0.12–
0.39) and positron emission tomography (pooled NLR, 0.15; 95% CI, 0.05–0.44), and better than single-photon emission
computed tomography (pooled NLR, 0.47; 95% CI, 0.37–0.59). At the patient level, MRI (pooled NLR, 0.14; 95% CI,
0.10–0.18) performed similar to computed tomography (pooled NLR, 0.12; 95% CI, 0.04–0.33) and positron emission
tomography (pooled NLR, 0.14; 95% CI, 0.02–0.87), and better than single-photon emission computed tomography
(pooled NLR, 0.39; 95% CI, 0.27–0.55) and echocardiography (pooled NLR, 0.42; 95% CI, 0.30–0.59).
Downloaded from http://ahajournals.org by on October 5, 2019

Conclusions—Stress myocardial perfusion imaging with MRI, computed tomography, or positron emission tomography
can accurately rule out hemodynamically significant coronary artery disease and can act as a gatekeeper for invasive
revascularization. Single-photon emission computed tomography and echocardiography are less suited for this
purpose.  (Circ Cardiovasc Imaging. 2015;8:e002666. DOI: 10.1161/CIRCIMAGING.114.002666.)

Key Words: fractional flow reserve, myocardial ◼ meta-analysis ◼ myocardial perfusion imaging

S tress myocardial perfusion imaging (MPI) has been pro-

posed as an important gatekeeper for invasive coronary
angiography (ICA) and percutaneous coronary interventions
(PCI) in patients evaluated for hemodynamically significant
coronary artery disease (CAD). MPI can be performed by sin-
gle-photon emission computed tomography (SPECT), echo-
cardiography, MRI, positron emission tomography (PET), and
computed tomography (CT).
See Clinical Perspective
Traditionally, hemodynamically significant CAD is deter-
mined by ICA with fractional flow reserve (FFR) measure-
ments.1,2 FFR is quantified as the ratio between the maximum
achievable myocardial blood flow in the case of a stenosis
and the maximum achievable myocardial blood flow in the
absence of a stenosis.2 Practically, FFR is measured during
ICA by calculating the ratio between the distal (poststenotic)
and proximal (presteonotic) coronary artery pressure during
maximal myocardial hyperemia. FFR has a circumscript isch-
emic threshold value between 0.75 and 0.80.1 Previous studies

Received July 29, 2014; accepted December 15, 2014.

From the Departments of Radiology (R.A.P.T., B.A.B., H.E.A., J.H., P.A.d.J., T.L.) and Cardiology (H.E.A.), University Medical Center Utrecht,
Utrecht, The Netherlands; Cardiac MR PET CT Program, Department of Radiology, Massachusetts General Hospital, Harvard Medical School, Boston
(R.A.P.T., U.H.); and Division of Imaging Sciences and Biomedical Engineering, St. Thomas’ Hospital, London, United Kingdom (E.N.).
Guest Editor for this article was David A. Bluemke, MD, PhD.
The Data Supplement is available at http://circimaging.ahajournals.org/lookup/suppl/doi:10.1161/CIRCIMAGING.114.002666/-/DC1.
Correspondence to Richard Takx, MD, Department of Radiology, UMC Utrecht, Heidelberglaan 100, P.O. Box 85500, 3584 CX Utrecht, The Netherlands.
E-mail rtakx@umcutrecht.nl
© 2015 American Heart Association, Inc.
Circ Cardiovasc Imaging is available at http://circimaging.ahajournals.org DOI: 10.1161/CIRCIMAGING.114.002666

1
 2   Takx et al   Diagnostic Accuracy of Stress MPI to ICA With FFR  
have demonstrated that an FFR≤0.75 reliably identifies induc-
ible myocardial ischemia, whereas an FFR of >0.80 reliably
excludes myocardial ischemia.2,3 Benefits of FFR over other
quantifiers of hemodynamically significant CAD are that it is
unaffected by fluctuations in heart rate, blood pressure, and
myocardial contractility.4,5
This meta-analysis aimed to determine the diagnostic accu-
racy of MPI compared with ICA with FFR for the diagnosis of
hemodynamically significant CAD in patients with suspected
or known CAD. Because an accurate and consistent reference
standard is essential for a reliable comparison of the different
MPI techniques, only studies using FFR as the reference stan-
dard were considered.
Methods
This meta-analysis was conducted in adherence to the Preferred
Reporting Items for Systematic Reviews and Meta-analyses
statement.6 The protocol was published in the PROSPERO
database (www.crd.york.ac.uk/PROSPERO; PROSPERO
2014:CRD42014008737). Using predefined search criteria PubMed,
EMBASE, and Web of Science were systematically searched from
inception to May 2014. The search syntax is presented in Table 1.
No search restrictions were applied. A manual reference check of
included articles was performed to identify potential studies missed
by our search strategy.
Eligibility Criteria
Two independent reviewers used predefined criteria to determine
article eligibility. Relevant articles were included based on the fol-
lowing criteria: study domain—patients with suspected or known
CAD; index test—MPI SPECT, echocardiography, MRI, PET, and
Downloaded from http://ahajournals.org by on October 5, 2019
CT; reference standard—FFR for intermediate coronary lesions
as defined by the individual studies; study results—agreement be-
tween index and reference standard; study design—cross-sectional
study. When studies stemmed from overlapping populations, the
study with the largest population was included. Discordances
between reviewers were resolved by consensus discussion.
Probability-adjusted agreement (κ statistic) evaluated inter-rater
agreement. Animal studies, phantom studies, and case-reports
(n<10) were excluded.
Critical Appraisal
Two independent reviewers assessed the relevance and validity of the
included studies using the revised Quality Assessment of Studies of
Diagnostic Accuracy Included in Systematic Reviews criteria.7 The
item blinding was presented separately to stress its relevance. The
term flow and timing described the time interval between index test
and reference standard and a period of ≤30 days was rated positive.
Probability-adjusted agreement (κ statistic) was used to evaluate
inter-rater agreement. Discrepancies were resolved by consensus
discussion.
Table 1.  Search Syntax
Database Search Term
PubMed (adenosine stress [title/abstract] or adenosine induced
stress [title/abstract] or stress perfusion [title/abstract] or
stress myocardial [title/abstract] or myocardial perfusion
[title/abstract] or dobutamine stress [title/abstract])
and (fractional flow reserve [title/abstract] or FFR [title/
abstract])
EMBASE Replaced (title/abstract) with ti,ab
Web of Science Replaced (title/abstract) with topic
FFR indicates fractional flow reserve.
Data Extraction
Data extraction was performed by one reviewer (R.A.P.T.) and sub-
sequently verified by a second reviewer (B.A.B.). The following data
categories were extracted from the included studies: patient charac-
teristics, stressor used for index test, index test characteristics, and
FFR threshold. True positive, false-positive, false-negative, and true
negative numbers were extracted. The findings were summarized in
2 by 2 contingency tables. Empty cells were filled with 0.5 events to
allow the calculation of the outcome measures of interest. In case of
multiple techniques/thresholds, the predefined primary end point was
chosen. If the primary end point was not reported, the best performing
technique/threshold parameter was chosen.
Data Synthesis and Analysis
All analyses were performed at both the vessel and the patient level. For
each study, the sensitivity, specificity, positive and negative likelihood
ratio (NLR), and the diagnostic odds ratio, along with the 95% confi-
dence interval (CI), were calculated to express the diagnostic accuracy
of the MPI techniques in diagnosing hemodynamically significant CAD
when compared with ICA with FFR. Because methodological hetero-
geneity between included studies was anticipated, a random-effects
(DerSimonian and Laird) model was used for pooling the data.8 Per
MPI technique, post-test probabilities were calculated based on various
pretest probabilities and the pooled likelihood ratio. Also, pooled results
were used to determine the area under the receiver operating characteris-
tic curve and the Q* statistic. Assuming that MPI can act as a gatekeeper
for PCI, ruling out hemodynamically significant CAD is the most impor-
tant. Therefore, the NLR was chosen as the most important diagnostic
test characteristic. Heterogeneity among study result was quantified by
calculating the I2 statistic. The degree of heterogeneity was considered
low (I2<50%), moderate (I2=50%–75%), or high (I2>75%).9 Per imaging
technique, possible publication bias was assed graphically by drawing
funnel plots and statistically via the Egger test for funnel plot asym-
metry.10 Statistical analysis was performed using R version 2.10.2 com-
bined with the package meta, version 3.5-0 (R Foundation for Statistical
Computing, Vienna, Austria) and the dedicated meta-analysis software
Meta-DiSc version 1.4 (Universidad Complutense, Madrid, Spain).11
Results
After exclusion of duplicates, the systematic search retrieved
488 potentially relevant articles (Figure 1). Thirty-seven arti-
cles (excluding 4 with overlapping patient populations)12–15
met our predefined criteria and were selected for inclusion
with an inter-rater agreement of κ=0.89. Included studies
were published between 1996 and 2014. Population charac-
teristics (Table I in the Data Supplement), imaging protocols
(Table II in the Data Supplement), reference test characteris-
tics including preselection criteria for FFR evaluation (Table
III in the Data Supplement), and diagnostic accuracy (Tables
IV and V in the Data Supplement) are listed. Eight studies
reported on SPECT, 2 on echocardiography, 2 on SPECT and
echocardiography combined, 15 on MRI, 3 on PET, and 7 on
CT. Methodological quality of included studies was assessed
using the modified Quality Assessment of Studies of Diagnos-
tic Accuracy Included in Systematic Reviews-2 criteria (Table
VI in the Data Supplement), with an inter-rater agreement of
κ=0.95. Overall, studies rated poorly on the item blinding the
index test result from the assessor of the reference standard
and standardized selection of the study population (Figure 2).
Pooled Diagnostic Accuracy
For MRI, PET, and CT, pooled estimates of diagnostic accuracy
(including sensitivity and specificity) were substantially higher
when compared with SPECT or echocardiography (Tables 2 and
 3   Takx et al   Diagnostic Accuracy of Stress MPI to ICA With FFR 
Downloaded from http://ahajournals.org by on October 5, 2019
3). This finding was observed at both the vessel and the patient
level. The summary receiver operating characteristic curve also
demonstrated superior diagnostic accuracy for MRI, PET, and
CT when compared with SPECT at the vessel (Figure 3A) and
SPECT and echocardiography at patient level (Figure 3B). In
line with these findings, a large change in post-test probability
was observed for MRI, PET, and CT, whereas a modest change
in post-test probability was observed for SPECT and echocar-
diography (Figure 4; Table VII in the Data Supplement).
Negative Likelihood Ratio
At the vessel level, the NLR of MRI (pooled NLR, 0.16; 95%
CI, 0.13–0.21), CT (pooled NLR, 0.22; 95% CI, 0.12–0.39), and
 
Figure 1. Flow chart. Search performed on May
9, 2014. *Two studies evaluated both echocar-
diography (ECHO) and single-photon emission
computed tomography (SPECT). CT indicates
computed tomography; and PET, positron emission
tomography.
PET (pooled NLR, 0.15; 95% CI, 0.05–0.44) were more favor-
able when compared with the NLR of SPECT (pooled NLR, 0.47;
95% CI, 0.37–0.59). At the patient level, the NLR of MRI (pooled
NLR, 0.14; 95% CI, 0.10–0.18) was similar to CT (pooled NLR,
0.12; 95% CI, 0.04–0.33) and PET (pooled NLR, 0.14; 95% CI,
0.02–0.87), but substantially lower when compared with the NLR
of SPECT (pooled NLR, 0.39; 95% CI, 0.27–0.55) and echocar-
diography (pooled NLR, 0.42; 95% CI, 0.30–0.59). At the ves-
sel level, heterogeneity was high for CT (I2=82.3%) and PET
(I2=87.9%), moderate for SPECT (I2=66.3%), and low for MRI
(I2=15.8%). At the patient level, heterogeneity was high for PET
(I2=84.1%), moderate for CT (I2=69.8%) and SPECT (I2=55.3%),
and low for echocardiography (I2=0.0%) and MRI (I2=0.0%).
Figure 2. Assessment of methodologi-
cal quality of included studies using the
Quality Assessment of Studies of Diag-
nostic Accuracy Included in Systematic
Reviews-2 criteria. Stacked bars repre-
sent the proportion of studies with a low
risk of bias (green), unclear risk of bias
(yellow), or high risk of bias (red). IT indi-
cates index test; and RT, reference test.
 4   Takx et al   Diagnostic Accuracy of Stress MPI to ICA With FFR  

Table 2.  Outcome Summary at Vessel Level

Index No. of
Test Vessels TP FP FN TN Sensitivity Specificity PLR NLR DOR AUC Q*-Statistic
SPECT 924 221 88 139 476 0.61 0.84 3.76 0.47 8.17 0.83 0.76
(0.56–0.66) (0.81–0.87) (2.74–5.16) (0.37–0.59) (4.99–13.37) (0.67–0.98) (0.62–0.90)
MRI 1830 457 123 70 1178 0.87 0.91 8.27 0.16 66.86 0.95 0.89
(0.84–0.90) (0.89–0.92) (4.93–13.87) (0.13–0.21) (43.16–103.56) (0.93–0.97) (0.86–0.91)
PET 870 182 71 38 579 0.83 0.89 7.43 0.15 48.53 0.95 0.89
(0.77–0.88) (0.86–0.91) (5.03–10.99) (0.05–0.44) (14.02–167.97) (0.91–0.99) (0.83–0.95)
CT 1074 238 108 69 659 0.78 0.86 5.74 0.22 28.90 0.91 0.85
(0.72–0.82) (0.83–0.88) (3.48–9.46) (0.12–0.39) (13.52–61.81) (0.86–0.96) (0.79–0.90)
95% confidence interval in parentheses. AUC indicates area under the receiver operating characteristic curve; CT, computed tomography; DOR, diagnostic odds
ratio; FN, false-negative; FP, false-positive; NLR, negative likelihood ratio; PET, positron emission tomography; PLR, positive likelihood ratio; TN, true negative; TP, true
positive; and SPECT, single-photon emission computed tomography.

Publication Bias hemodynamically significant CAD in patients with chest pain

Publication bias assessment was only performed for MRI
because it was the only modality that had >10 included studies.
At the vessel level, the funnel plot showed no sign of publication
bias (Figure IA in the Data Supplement). This finding was sup-
ported by the Egger test for funnel plot asymmetry (P=0.82).
At the patient level, the funnel plot suggested publication bias,
with smaller studies yielding better odds ratios (Figure IB in
the Data Supplement). This observation was not supported by
the Egger test for funnel plot asymmetry (P=0.11).
Discussion
This meta-analysis provides an overview of various noninva-
sive MPI techniques to diagnose hemodynamically significant
Downloaded from http://ahajournals.org by on October 5, 2019
syndromes.
During the past decade, the treatment of CAD has radically
changed. Studies such as Fractional Flow Reserve Versus
Angiography for Multivessel Evaluation (FAME) and Clinical
Outcomes Utilizing Revascularization and Aggressive Drug
Evaluation (COURAGE) initiated a shift from PCI as the treat-
ment of choice based on anatomic CAD to a more nuanced
use of PCI guided by the hemodynamic significance of CAD
and by recognizing that optimal medical therapy is equal or
favorable in many patients with CAD.18,19 Furthermore, FAME
demonstrated that the routine measurement of FFR results
in significantly reduced mortality and myocardial infarc-
tion rates at 2 years of follow-up in patients with multivessel

CAD and found that MRI, PET, and CT are accurate in rul-
ing out hemodynamically significant CAD in patients with
suspected or known CAD. Therefore, MPI with MRI, CT, or
PET has the potential to serve as a gatekeeper for invasive
assessment of hemodynamic significance by ICA and FFR.
Compared with CT and PET, MRI can determine myocardial
blood flow without exposure to ionizing radiation. Further-
more, MRI, together with PET, allows absolute quantifica-
tion of myocardial blood flow, which is important to detect
perfusion defects in patients with 3-vessel disease.16,17 Hence,
MRI combines high accuracy with low risk and could be
considered as the method of choice to noninvasively rule out
CAD undergoing PCI with drug eluting stents.18 Noninvasive
assessment of hemodynamic CAD with stress MPI may fur-
ther benefit patient outcome and costs.
A previously published meta-analysis compared stress MPI
with MRI, PET, and SPECT to ICA without FFR and reported
similar diagnostic accuracy, in terms of pooled sensitivity and
specificity, at the patient level for PET (sensitivity, 0.84; 95%
CI, 0.81–0.87; specificity, 0.81; 95% CI, 0.74–0.87). However, it
reported lower accuracy for MRI (sensitivity, 0.89; 95% CI, 0.88–
0.91; specificity, 0.76; 95% CI, 0.73–0.78) and reported higher
sensitivity for SPECT (sensitivity, 0.88; 95% CI, 0.88–0.89;
specificity, 0.61; 95% CI, 0.59–0.62) compared with our results.20

Table 3.  Outcome Summary at Patient Level

Index No. of
Test Patients TP FP FN TN Sensitivity Specificity PLR NLR DOR AUC Q*-Statistic
SPECT 533 162 67 58 246 0.74 0.79 3.13 0.39 9.63 0.82 0.75
(0.67–0.79) (0.74–0.83) (2.09–4.70) (0.27–0.55) (4.57–20.31) (0.73–0.91) (0.68–0.83)
ECHO 177 46 18 21 92 0.69 0.84 3.68 0.42 12.04 0.83 0.76
(0.56–0.79) (0.75–0.90) (1.89–7.15) (0.30–0.59) (5.25–27.65) (0.74–0.93) (0.68–0.85)
MRI 798 355 52 43 348 0.89 0.87 6.29 0.14 50.94 0.94 0.88
(0.86–0.92) (0.83–0.90) (4.88–8.12) (0.10–0.18) (32.45–79.97) (0.92–0.96) (0.85–0.90)
PET 224 73 18 14 119 0.84 0.87 6.53 0.14 47.26 0.93 NA 0.87 NA
(0.75–0.91) (0.80–0.92) (2.83–15.06) (0.02–0.87) (4.17–536.29)
CT 316 143 30 20 123 0.88 0.80 3.79 0.12 43.36 0.93 0.87
(0.82–0.92) (0.73–0.86) (1.94–7.40) (0.04–0.33) (20.26–92.78) (0.89–0.97) (0.82–0.92)
95% confidence interval in parentheses. AUC indicates area under the receiver operating characteristic curve; CT, computed tomography; DOR, diagnostic odds
ratio; FN, false-negative; FP, false-positive; NA, not applicable; NLR, negative likelihood ratio; PET, positron emission tomography; PLR, positive likelihood ratio; TN, true
negative; TP, true positive; and SPECT, single-photon emission computed tomography.
 5   Takx et al   Diagnostic Accuracy of Stress MPI to ICA With FFR  

A Summary ROC curve B Summary ROC curve

Sensitivity Sensitivity
1 1
SPECT SPECT
AUC = 0.83 AUC = 0.82
s.e. (AUC) = 0.08 s.e. (AUC) = 0.04
0.9 0.9 Q* = 0.75
Q* = 0.76
s.e. (Q*) = 0.07 s.e. (Q*) = 0.04

0.8 MRI 0.8 Echocardiography

AUC = 0.95 AUC = 0.83
s.e. (AUC) = 0.01 s.e. (AUC) = 0.05
Q* = 0.87 0.7 Q* = 0.76
0.7
s.e. (Q*) = 0.01 s.e. (Q*) = 0.04

PET MRI
0.6 AUC = 0.95 0.6 AUC = 0.94
s.e. (AUC) = 0.02 s.e. (AUC) = 0.01
Q* = 0.89 Q* = 0.88
0.5 s.e. (Q*) = 0.03 0.5 s.e. (Q*) = 0.01

CT PET
AUC = 0.91 AUC = 0.93
0.4 s.e. (AUC) = 0.03
0.4 s.e. (AUC) = NA
Q* = 0.85 Q* = 0.87
s.e. (Q*) = 0.03 s.e. (Q*) = NA
0.3 0.3
CT
AUC = 0.93
0.2 0.2 s.e. (AUC) = 0.02
Q* = 0.87
s.e. (Q*) = 0.02

0.1 0.1

0 0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

1 - specificity 1 - specificity

Figure 3. Summary receiver operating characteristic (ROC) curve plotting the true positive rate (sensitivity) against the false-positive rate
(1 specificity) for per-vessel (A) and per-patient analysis (B). Each symbol represents an individual study in the meta-analysis, with the size
of the symbol proportional to the sample size of the study. The Q* statistic represents the point where sensitivity and specificity are equal.
AUC indicates area under the summary receiver operating characteristic curve. CT indicates computed tomography; PET, positron emis-
sion tomography; and SPECT, single-photon emission computed tomography.

These differences may be because of the use of an anatomic refer-
ence standard instead of a hemodynamic reference standard.
The pooled results of our meta-analysis need to be consid-
ered in the context of the included studies, which were limited in
number for some modalities. Although many more studies have
compared MPI with ICA, most of these were noneligible because
Downloaded from http://ahajournals.org by on October 5, 2019
provided an accurate summary diagnostic accuracy estimate
largely unachievable by standalone studies.24 Therefore, the cal-
culated pooled estimates of diagnostic accuracy provided ample
information to decide on the technique of preference for noninva-
sive evaluation of hemodynamically significant CAD.
An important finding of this meta-analysis is that several

they did not perform FFR as reference standard. However, we noninvasive diagnostic tests can be used to select patients accu-
considered any comparison with anatomic CAD by ICA as inap- rately whom do not benefit from coronary revascularization.
propriate because ICA alone has poor diagnostic performance The clinical use of the different techniques was evaluated using
for intermediate coronary stenoses and the decision to perform a the average pretest probability for hemodynamically significant
PCI based on hemodynamic significance improves patient out-
1
comes.18,21,22 Moreover, recent evidence demonstrates that FFR
SPECT
has a continuous and independent relationship with subsequent 0.9 ECHO
major adverse cardiac events and lower FFR values benefit
MRI
more from revascularization.23 However, there is an ongoing 0.8
PET
discussion on the true reference standard for the measurement
CT
of myocardial ischemia. Although this problem remains unre- 0.7
solved, an answer will be provided by outcome data as obtained
Post-test probability

in the International Study of Comparative Health Effectiveness 0.6

With Medical and Invasive Approaches (ISCHEMIA) trial

0.5
(ClinicalTrials.gov Identifier: NCT01471522).
Significant in-between study heterogeneity was observed for
0.4
CT, PET, and SPECT. Differences in study methodology, includ-
ing hardware, acquisition protocols, stress protocols, and differ-
0.3
ences in the time interval between index and reference standard,
likely account for this observation. Furthermore, the subjective 0.2
nature of visual and semiquantitative assessment of MPI and
imaging artifacts could result in an in-between study heterogene- 0.1
ity. Discrepancies in referral patterns and clinical thresholds for
noninvasive perfusion imaging, reflected by the heterogeneity in 0
0 0.2 0.4 0.6 0.8 1
prevalence of hemodynamically significant CAD among included
Pre-test probability
studies ranging between 19% and 61% and 20% and 77% on a
per-vessel and per-patient basis, respectively, are another potential Figure 4. Absolute change in post-test probability of hemody-
source of an in-between study heterogeneity. Although substan- namically significant coronary artery disease by various pretest
probabilities. CT indicates computed tomography; ECHO, echo-
tial in-between study heterogeneity was observed among included cardiography; PET, positron emission tomography; and SPECT,
studies for most MPI techniques, the random-effects model single-photon emission computed tomography.
 6   Takx et al   Diagnostic Accuracy of Stress MPI to ICA With FFR  
CAD (46%). The post-test probability given a negative test
result was reduced to 9% to 11% for MRI, PET, and CT. Hence,
a negative MRI, PET, and CT decreases the probability of dis-
ease by ≥35%. In case of a positive result, MRI and PET per-
formed best with a post-test probability of hemodynamically
significant CAD of 84% to 85%. Apart from the diagnostic
performance per se, however, several other considerations are
relevant to decide which test to use in clinical practice, such as
availability, costs, and exposure to ionizing radiation.
Radiation exposure is an important issue to consider when
using SPECT, PET, and CT. Nevertheless, efforts such as
iterative reconstruction algorithms combined with improve-
ments in imaging hardware continue to reduce the radiation
exposure of such imaging techniques. Also the use of 99mTc
(effective dose, 5.0–20.3 mSv) for SPECT examinations com-
pared with short-lived radioisotope, such as 15O- water (effec-
tive dose, 0.7–1.4 mSv), for PET, allow for reduced radiation
dosages.4,5,25 Moreover, high-quality anatomic information is
not compulsory for PET MPI, reducing the radiation dosage
from coregistered CT images. MRI does not expose patients
to ionizing radiation. However, MRI cannot be performed in
patients with iatrogenic devices, which are incompatible with
the MRI environment. Nonetheless, the benefits of MPI still
exceed the harmful effects of radiation exposure in patients
considered for invasive revascularization.
An important strength of the present meta-analysis is that we
compared all commonly used MPI techniques with FFR as a ref-
erence standard for the evaluation of intermediate stenosis. Our
meta-analysis had several limitations. First, the meta-analysis
Downloaded from http://ahajournals.org by on October 5, 2019
focused on the diagnostic performance of MPI as a standalone
technique. As such, we did not evaluate the added value of other
diagnostic factors, such as age, sex, ventricular function, or steno-
sis degree. Combining multiple diagnostic variables could possi-
bly yield even higher diagnostic accuracy in case of MRI and CT
techniques. Second, it is unclear whether all stenoses were evalu-
ated by FFR and studies used different FFR thresholds to define
an intermediate anatomic stenosis. This is important because
low-grade anatomic stenoses can be associated with a pathologi-
cal FFR.26 Third, the results are likely affected by expert center
bias because most studies included were conducted at experi-
enced imaging centers. Expert center bias could overestimate
the diagnostic accuracy of the index test. Fourth, we observed
broad confidence intervals. The width of the confidence interval
depends on the precision of individual study estimates and on the
total number of pooled vessels/patients. On the basis of random-
effects model, the confidence interval increases accordingly with
an increase in in-between study heterogeneity. Because of the lim-
ited number of studies, determining the effect of an in-between
study heterogeneity (including differences in patient demograph-
ics, differences in pretest probability of CAD, and differences
in imaging technology) by meta-regression techniques could
not be performed. Nonetheless, eyeballing of Tables IV and V
in the Data Supplement revealed no obvious correlation between
publication year and diagnostic accuracy of MPI techniques
with large in-between study heterogeneity. Therefore, it seems
unlikely that advances in imaging technology account for the
observed in-between study heterogeneity. Fifth, studies evaluat-
ing the performance of echocardiography were based on a differ-
ent classification of ischemia, namely wall motion abnormalities.
Because wall motion abnormalities occur later in the ischemic
cascade, echocardiography should be less sensitive, but more
specific when compared with other MPI techniques. Finally,
only studies using ICA with FFR as the reference standard were
eligible for inclusion. Therefore, only 3 PET studies (all using
15
O-water) met the inclusion criteria, with substantial differences
in diagnostic accuracy, resulting in high in-between study hetero-
geneity on both per-vessel and per-patient level. Because of the
short half-life (≈2 minutes) of 15O-water, the production of this
PET-tracer requires an onsite cyclotron. As a result, most PET
MPI examinations are performed with 82Rb-chloride instead.
82
Rb-chloride has less favorable flow-extraction characteristics
compared with 15O-water. Therefore, it is not possible to extrapo-
late the findings from 15O-water PET studies to clinical settings
in which 82Rb-chloride is used. No studies evaluated the use of
82
Rb-chloride compared with FFR for intermediate stenoses.
Conclusions
The presence of hemodynamically significant CAD can be
accurately ruled out by stress MPI with MRI, CT, or PET,
but less accurate with SPECT or echocardiography when
compared with invasive FFR. Therefore, MPI with MRI, CT,
or PET can act as an important gatekeeper for ICA and PCI.
Considering the merits and shortcomings of the different non-
invasive techniques, MRI could be regarded as the technique
of choice because it does not involve exposure to ionizing
radiation and achieves similar diagnostic performance com-
pared with PET and CT. In contrast, the more widely available
SPECT and echocardiography imaging techniques are less
suited for excluding hemodynamically significant CAD.
Disclosures
Dr Nagel has received significant grant support from Bayer Healthcare
and Philips Healthcare. The other authors report no conflicts.
References
1. De Bruyne B, Sarma J. Fractional flow reserve: a review: invasive imag-
ing. Heart. 2008;94:949–959. doi: 10.1136/hrt.2007.122838.
2. Pijls NH, De Bruyne B, Peels K, Van Der Voort PH, Bonnier HJ, Bartunek
J Koolen JJ, Koolen JJ. Measurement of fractional flow reserve to as-
sess the functional severity of coronary-artery stenoses. N Engl J Med.
1996;334:1703–1708. doi: 10.1056/NEJM199606273342604.
3. Petraco R, Sen S, Nijjer S, Echavarria-Pinto M, Escaned J, Francis DP, Davies
JE. Fractional flow reserve-guided revascularization: practical implications
of a diagnostic gray zone and measurement variability on clinical decisions.
JACC. Cardiovasc Interv. 2013;6:222–225. doi: 10.1016/j.jcin.2012.10.014.
4. de Bruyne B, Bartunek J, Sys SU, Pijls NH, Heyndrickx GR, Wijns W.
Simultaneous coronary pressure and flow velocity measurements in hu-
mans. Feasibility, reproducibility, and hemodynamic dependence of coro-
nary flow velocity reserve, hyperemic flow versus pressure slope index,
and fractional flow reserve. Circulation. 1996;94:1842–1849.
5. Pijls NH, Van Gelder B, Van der Voort P, Peels K, Bracke FA, Bonnier
HJ, el Gamal MI. Fractional flow reserve. A useful index to evaluate the
influence of an epicardial coronary stenosis on myocardial blood flow.
Circulation. 1995;92:3183–3193.
6. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gøtzsche PC, Ioannidis JP,
Clarke M, Devereaux PJ, Kleijnen J, Moher D. The PRISMA statement
for reporting systematic reviews and meta-analyses of studies that evaluate
healthcare interventions: explanation and elaboration. BMJ. 2009;339:b2700.
7. Whiting PF, Rutjes AW, Westwood ME, Mallett S, Deeks JJ, Reitsma
JB, Leeflang MM, Sterne JA, Bossuyt PM; QUADAS-2 Group.
QUADAS-2: a revised tool for the quality assessment of diag-
nostic accuracy studies. Ann Intern Med. 2011;155:529–536. doi:
10.7326/0003-4819-155-8-201110180-00009.
 7   Takx et al   Diagnostic Accuracy of Stress MPI to ICA With FFR  

8. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin
Trials. 1986;7:177–188.
9. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency
in meta-analyses. BMJ. 2003;327:557–560. doi: 10.1136/bmj.327.7414.557.
10. Sterne JA, Egger M. Regression methods to detect publication and
other bias in meta-analysis In: Rothstein HR, Sutton AJ, Borenstein M,
eds. Publication Bias in Meta-Analysis: Prevention, Assessment, and
Adjustments. Chichester: Wiley; 2005:99–110.
11. Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A. Meta-DiSc: a
software for meta-analysis of test accuracy data. BMC Med Res Methodol.
2006;6:31. doi: 10.1186/1471-2288-6-31.
12. Bettencourt N, Ferreira N, Chiribiri A, Schuster A, Sampaio F, Santos
L, Melica B, Rodrigues A, Braga P, Teixeira M, Leite-Moreira A, Silva-
Cardoso J, Portugal P, Gama V, Nagel E. Additive value of magnetic
resonance coronary angiography in a comprehensive cardiac magnetic
resonance stress-rest protocol for detection of functionally significant cor-
onary artery disease: a pilot study. Circ Cardiovasc Imaging. 2013;6:730–
738. doi: 10.1161/CIRCIMAGING.113.000280.
13. Erhard I, Rieber J, Jung P, Hacker M, Schiele T, Stempfle HU, König A,
Baylacher M, Theisen K, Siebert U, Klauss V. The validation of fractional
flow reserve in patients with coronary multivessel disease: a comparison
with SPECT and contrast-enhanced dobutamine stress echocardiography.
Z Kardiol. 2005;94:321–327. doi: 10.1007/s00392-005-0213-6.
14. Futamatsu H, Wilke N, Klassen C, Shoemaker S, Angiolillo DJ, Siuciak
A, Morikawa-Futamatsu K, Suzuki N, von Ziegler F, Bass TA, Costa MA.
Evaluation of cardiac magnetic resonance imaging parameters to detect
anatomically and hemodynamically significant coronary artery disease.
Am Heart J. 2007;154:298–305. doi: 10.1016/j.ahj.2007.04.024.
15. Joutsiniemi E, Saraste A, Pietilä M, Mäki M, Kajander S, Ukkonen H,
Airaksinen J, Knuuti J. Absolute flow or myocardial flow reserve for the
detection of significant coronary artery disease? Eur Heart J Cardiovasc
Imaging. 2014;15:659–665. doi: 10.1093/ehjci/jet274.
16. Ziadi MC, Dekemp RA, Williams K, Guo A, Renaud JM, Chow BJ, Klein
R, Ruddy TD, Aung M, Garrard L, Beanlands RS. Does quantification of
myocardial flow reserve using rubidium-82 positron emission tomography
facilitate detection of multivessel coronary artery disease? J Nucl Cardiol.
2012;19:670–680. doi: 10.1007/s12350-011-9506-5.
Downloaded from http://ahajournals.org by on October 5, 2019
19. Shaw LJ, Berman DS, Maron DJ, Mancini GB, Hayes SW, Hartigan
PM, Weintraub WS, O’Rourke RA, Dada M, Spertus JA, Chaitman
BR, Friedman J, Slomka P, Heller GV, Germano G, Gosselin G, Berger
P, Kostuk WJ, Schwartz RG, Knudtson M, Veledar E, Bates ER,
McCallister B, Teo KK, Boden WE; COURAGE Investigators. Optimal
medical therapy with or without percutaneous coronary intervention to
reduce ischemic burden: results from the Clinical Outcomes Utilizing
Revascularization and Aggressive Drug Evaluation (COURAGE) trial
nuclear substudy. Circulation. 2008;117:1283–1291. doi: 10.1161/
CIRCULATIONAHA.107.743963.
20. Jaarsma C, Leiner T, Bekkers SC, Crijns HJ, Wildberger JE, Nagel E,
Nelemans PJ, Schalla S. Diagnostic performance of noninvasive myo-
cardial perfusion imaging using single-photon emission computed
tomography, cardiac magnetic resonance, and positron emission tomog-
raphy imaging for the detection of obstructive coronary artery disease:
a meta-analysis. J Am Coll Cardiol. 2012;59:1719–1728. doi: 10.1016/j.
jacc.2011.12.040.
21. Kern MJ, Lerman A, Bech JW, De Bruyne B, Eeckhout E, Fearon WF,
Higano ST, Lim MJ, Meuwissen M, Piek JJ, Pijls NH, Siebes M, Spaan JA;
American Heart Association Committee on Diagnostic and Interventional
Cardiac Catheterization, Council on Clinical Cardiology. Physiological
assessment of coronary artery disease in the cardiac catheterization
laboratory: a scientific statement from the American Heart Association
Committee on Diagnostic and Interventional Cardiac Catheterization,
Council on Clinical Cardiology. Circulation. 2006;114:1321–1341. doi:
10.1161/CIRCULATIONAHA.106.177276.
22. Tobis J, Azarbal B, Slavin L. Assessment of intermediate severity coronary
lesions in the catheterization laboratory. J Am Coll Cardiol. 2007;49:839–
848. doi: 10.1016/j.jacc.2006.10.055.
23. Johnson NP, Tóth GG, Lai D, Zhu H, Açar G, Agostoni P, Appelman Y, Arslan
F, Barbato E, Chen SL, Di Serafino L, Domínguez-Franco AJ, Dupouy P,
Esen AM, Esen OB, Hamilos M, Iwasaki K, Jensen LO, Jiménez-Navarro
MF, Katritsis DG, Kocaman SA, Koo BK, López-Palop R, Lorin JD, Miller
LH, Muller O, Nam CW, Oud N, Puymirat E, Rieber J, Rioufol G, Rodés-
Cabau J, Sedlis SP, Takeishi Y, Tonino PA, Van Belle E, Verna E, Werner GS,
Fearon WF, Pijls NH, De Bruyne B, Gould KL. Prognostic value of fractional
flow reserve: linking physiologic severity to clinical outcomes. J Am Coll

17. Morton G, Chiribiri A, Ishida M, Hussain ST, Schuster A, Indermuehle
A, Perera D, Knuuti J, Baker S, Hedström E, Schleyer P, O’Doherty M,
Barrington S, Nagel E. Quantification of absolute myocardial perfusion
in patients with coronary artery disease: comparison between cardiovas-
cular magnetic resonance and positron emission tomography. J Am Coll
Cardiol. 2012;60:1546–1555. doi: 10.1016/j.jacc.2012.05.052.
18. Pijls NH, Fearon WF, Tonino PA, Siebert U, Ikeno F, Bornschein B, van’t
Veer M, Klauss V, Manoharan G, Engstrøm T, Oldroyd KG, Ver Lee PN,
MacCarthy PA, De Bruyne B; FAME Study Investigators. Fractional
flow reserve versus angiography for guiding percutaneous coronary in-
tervention in patients with multivessel coronary artery disease: 2-year
follow-up of the FAME (Fractional Flow Reserve Versus Angiography for
Multivessel Evaluation) study. J Am Coll Cardiol. 2010;56:177–184. doi:
10.1016/j.jacc.2010.04.012.
Cardiol. 2014;64:1641–1654. doi: 10.1016/j.jacc.2014.07.973.
24. Higgins JP. Commentary: Heterogeneity in meta-analysis should be ex-
pected and appropriately quantified. Int J Epidemiol. 2008;37:1158–1160.
doi: 10.1093/ije/dyn204.
25. Flotats A, Knuuti J, Gutberlet M, Marcassa C, Bengel FM, Kaufmann PA,
Rees MR, Hesse B; Cardiovascular Committee of the EANM, the ESCR
and the ECNC. Hybrid cardiac imaging: SPECT/CT and PET/CT. A joint
position statement by the European Association of Nuclear Medicine
(EANM), the European Society of Cardiac Radiology (ESCR) and the
European Council of Nuclear Cardiology (ECNC). Eur J Nucl Med Mol
Imaging. 2011;38:201–212. doi: 10.1007/s00259-010-1586-y.
26. Rioufol G, Finet G. Functional versus anatomical stenosis evaluation:
fractional flow reserve defeats intravascular ultrasound. JACC. Cardiovasc
Interv. 2011;4:812–813. doi: 10.1016/j.jcin.2011.06.003.

CLINICAL PERSPECTIVE
This study provides important evidence on the relative merits and shortcomings of different noninvasive myocardial perfu-
sion imaging modalities to detect hemodynamically significant, clinically relevant stenoses. Imaging techniques studied in
this meta-analysis include positron emission tomography (PET), computed tomography (CT), MRI, single-photon emission
computed tomography, and echocardiography. All of these tests aim to measure myocardial perfusion to assess the presence
of relevant coronary artery stenoses indirectly. In current clinical practice, invasive coronary angiography with fractional
flow reserve is the best method to select patients with coronary artery disease in need of revascularization. In this meta-
analysis, we demonstrated that PET, CT, and MRI can accurately rule out hemodynamically significant coronary artery
disease when compared to invasive coronary angiography with functional flow reserve measurements. Single-photon emis-
sion computed tomography and echocardiography are clearly less accurate for this purpose. Our findings lay an important
foundation for more widespread use of PET, CT, or MRI myocardial perfusion imaging as gatekeepers for invasive coronary
angiography and percutaneous coronary intervention and a suggests a more limited role for single-photon emission com-
puted tomography and echocardiography. Considering the merits and shortcomings of PET, CT, and MRI, MRI could be
regarded as the technique of choice because it does not expose patients to ionizing radiation and achieves similar diagnostic
performance when compared with PET and CT.