JACC: CARDIOVASCULAR IMAGING VOL. 16, NO.

10, 2023

ª 2023 BY THE AMERICAN COLLEGE OF CARDIOLOGY FOUNDATION

PUBLISHED BY ELSEVIER

ORIGINAL RESEARCH

Cardiac Computed Tomography for

Quantification of Myocardial Extracellular
Volume Fraction
A Systematic Review and Meta-Analysis

Donghee Han, MD,a,* Andrew Lin, MBBS, PHD,b,* Keiichiro Kuronuma, MD, PHD,a Heidi Gransar, MS,a
Damini Dey, PHD,b John D. Friedman, MD,a Daniel S. Berman, MD,a,y Balaji K. Tamarappoo, MD, PHDc,y

ABSTRACT

BACKGROUND Extracellular volume (ECV) is a quantitative measure of extracellular compartment expansion, and an
increase in ECV is a marker of myocardial fibrosis. Although cardiac magnetic resonance (CMR) is considered the standard
imaging tool for ECV quantification, cardiac computed tomography (CT) has also been used for ECV assessment.

OBJECTIVES The aim of this meta-analysis was to evaluate the correlation and agreement in the quantification of
myocardial ECV by CT and CMR.

METHODS PubMed and Web of Science were searched for relevant publications reporting on the use of CT for ECV
quantification compared with CMR as the reference standard. The authors employed a meta-analysis using the restricted
maximum-likelihood estimator with a random-effects method to estimate summary correlation and mean difference. A
subgroup analysis was performed to compare the correlation and mean differences between single-energy CT (SECT) and
dual-energy CT (DECT) techniques for the ECV quantification.

RESULTS Of 435 papers, 13 studies comprising 383 patients were identified. The mean age range was 57.3 to 82 years,
and 65% of patients were male. Overall, there was an excellent correlation between CT-derived ECV and CMR-derived
ECV (mean: 0.90 [95% CI: 0.86-0.95]). The pooled mean difference between CT and CMR was 0.96% (95% CI: 0.14%-
1.78%). Seven studies reported correlation values using SECT, and 4 studies reported those using DECT. The pooled
correlation from studies utilizing DECT for ECV quantification was significantly higher compared with those with SECT
(mean: 0.94 [95% CI: 0.91-0.98] vs 0.87 [95% CI: 0.80-0.94], respectively; P ¼ 0.01). There was no significant dif-
ference in pooled mean differences between SECT vs DECT (P ¼ 0.85).

CONCLUSIONS CT-derived ECV showed an excellent correlation and mean difference of <1% with CMR-derived ECV.
However, the overall quality of the included studies was low, and larger, prospective studies are needed to examine the
accuracy and diagnostic and prognostic utility of CT-derived ECV. (J Am Coll Cardiol Img 2023;16:1306–1317) © 2023 by
the American College of Cardiology Foundation.

From the aDepartment of Imaging and Medicine, Cedars-Sinai Medical Center, Los Angeles, California, USA; bBiomedical Imaging
Research Institute, Cedars-Sinai Medical Center, Los Angeles, California, USA; and the cCardiovascular Institute, Indiana
University School of Medicine, Indianapolis, Indiana, USA. *Drs Han and Lin contributed equally to this work as first authors.
yDrs Berman and Tamarappoo contributed equally to this work as senior authors.
The authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’
institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information,
visit the Author Center.

Manuscript received February 10, 2023; revised manuscript received March 30, 2023, accepted March 30, 2023.

ISSN 1936-878X/$36.00 https://doi.org/10.1016/j.jcmg.2023.03.021

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023 Han et al 1307
OCTOBER 2023:1306–1317 CT vs CMR for ECV Quantification

M yocardial fibrosis, a key pathologic pro- sample size and focus on specific disease ABBREVIATIONS

cess in cardiomyopathies and valvular entities. Hence, we sought to perform a sys- AND ACRONYMS
1
heart disease, is characterized by exces- tematic review and meta-analysis integrating
CMR = cardiac magnetic
sive deposition of collagen-rich extracellular matrix the results of prior studies, to compare ECV resonance
proteins in the interstitial space of the myocardium, quantification by CT with CMR as the refer- CT = computed tomography
resulting in expansion of the extracellular matrix ence standard.
DECT = dual-energy computed
structure, myocardial remodeling, and ultimately, tomography
impaired cardiac function.2 METHODS
ECV = extracellular matrix
Rapid technological advances in cardiac imaging volume

now permit the noninvasive assessment of myocar- This study was performed according to the SECT = single-energy
dial fibrosis with quantification of myocardial extra- Preferred Reporting Items for Systematic Re- computed tomography

cellular matrix volume (ECV) fraction. Cardiac views and Meta-Analyses guidelines. Ethical
magnetic resonance (CMR)-derived ECV is a well- approval was not required for this meta-analysis
established biomarker that correlates with histologic because it was based on published research and did
diffuse interstitial fibrosis, 3
differentiates healthy not recruit patients. All studies included were
from diseased myocardium, 4
and associates with approved by their local Institutional Review Boards.
clinical outcomes. 5,6
This method of measuring T1 The meta-analysis protocol was developed following
relaxation times in the left ventricular myocardium the PROSPERO guidelines. However, it was not
before and after administration of gadolinium registered with PROSPERO as they had announced
contrast on CMR has become the noninvasive refer- their inability to process new protocol registrations
ence standard of myocardial fibrosis. due to the high volume of requests arising from the
More recently, cardiac computed tomography (CT) COVID-19 pandemic at the time of protocol
angiography has emerged as an alternative modality development.
for ECV quantification. Using an iodinated contrast PUBLICATION SEARCH AND STUDY ELIGIBILITY.
equilibrium technique, prior reports have shown We systemically reviewed Web of Science and MED-
good agreement between myocardial ECV values LINE to identify relevant papers published from
determined by CT and CMR in clinical cohorts.7-10 2000, using predefined search criteria (Supplemental
However, these studies are limited by their small Table 1). The last search was performed on October

F I G U R E 1 Flowchart Describing the Publication Search and Selection Algorithm

Records identified through database Additional records identified through

searching other sources
Identification

(PubMed) (Web of Science)

(n = 134) (n = 301)

Total records Duplicates removed

(n = 435) (n = 92)

Records excluded
Records screened (Abstracts screened
(n = 343) for relevance)
(n = 274)
Screening

Full-text articles Full-text articles excluded,

assessed for with reasons (n = 58)
eligibility
(n = 71) Study design (n = 6)
Index test (n = 4)
Publication type (n = 22)
Reference standard (n = 26)
Total number of
Included

studies included in the

meta-analysis
(n = 13)

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
1308 Han et al JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023

CT vs CMR for ECV Quantification OCTOBER 2023:1306–1317

T A B L E 1 Baseline Study Characteristics

First Author, Year Study Design N Age, y Male, % Target Population CT Scanner CMR Scanner

Kim et al,19 2022 Single center, 51 59.9  15.6 51 CMP Somatom Definition flash, 3.0-T Magnetom Trio,
retrospective Siemens Siemens
Dubourg et al,20 2021 Single center, 21 82.0  4.9 48 Severe AS Discover CT 750HD, GE 1.5 T Avanto, Siemens
retrospective
Emoto et al,18 2021 Single center, 21 60.5  13.0 76 AS, amyloidosis, iQon Spectral CT, Philips 3.0-T Ingenia CX, Philips
retrospective CMP
17
Ohta et al, 2020 Single center, 23 65 (55-74) 61 CMP Discover CT 750HD, GE 3.0-T MAGNETOM Skyra,
prospective Siemens
Emoto et al,23 2020 Single center, 34 64.5  14.6 68 Suspected cardiac 320 detector Aquilion Not reported
retrospective disease One, Toshiba
Oda et al,21 2019 Single center, 40 63.6  11.9 65 Myocardial infarction, iQon Spectral CT, Philips Not reported
retrospective CMP, amyloidosis,
heart failure
Wang et al,24 2018 Single center, 35 59 (48-69) 60 Heart failure Dual source Somatom 3.0-T Verio, Siemens
prospective Definition flash,
Siemens
Kurita et al,16 2016 Single center, 8 65.0 37 Myocardial infarction, Dual source Somatom Not reported
prospective CMP Definition flash,
Siemens
Lee et al,10 2016 Single center, 30 57.3  14.8 63 CMP, amyloidosis, Dual source Somatom 3.0-T Ingenia, Philips
prospective sarcoidosis Definition flash,
Siemens
Treibel et al,9 2015 Single center, 53 64  14 amyloid 75 Amyloidosis and AS Somatom Sensation 64, 3.0-T Magnetom Trio,
prospective 68  8 AS Siemens Siemens
Liu et al,22 2013 — 20 63.2 (45-95) — Unspecified 320 detector Aquilion 3.0-T Verio, Siemens
One, Toshiba
8
Bandula et al, 2013 Single center, 23 70.8  8.3 70 AS Somatom Sensation 64, 1.5-T Avanto, Siemens
prospective Siemens
Nacif et al,7 2012 Single center, 24 63.2  10.0 58 11 control, 320 detector Aquilion 3.0-T Verio, Siemens
prospective 13 heart failure One, Toshiba

Values for age are mean  SD, median (IQR), or mean.

AS ¼ aortic stenosis; CMP ¼ cardiomyopathy; CMR ¼ cardiac magnetic resonance; CT ¼ computed tomography.

25th, 2021. Two reviewers (D.H. and A.L.) indepen- QUALITY ASSESSMENT. Methodological quality and
dently searched and performed data extraction. The potential sources of bias for included studies were
inclusion criteria for studies in the analysis were: assessed using the QUADAS-2 (Quality Assessment of
1) diagnostic or case-control studies; 2) ECV evaluated Diagnostic Accuracy Studies) criteria by 2 indepen-
in left ventricular myocardium; 3) human and adult dent reviewers (D.H. and A.L.). 11 Discrepancies were
subjects (age $18 years); 4) ECV evaluated using resolved by consensus.
single-energy or dual-energy delayed CT scans;
STATISTICAL ANALYSIS. We employed a meta-
5) CMR as a reference standard; and 6) correlation or
analysis using the Restricted Maximum-Likelihood
agreement of ECV values between CT and CMR re-
estimator with random-effects method12 to estimate
ported. Extracted studies were screened for eligibility
a summary correlation and mean differences (bias)
by title and abstract review. Selected records were
with 95% CI. Standard errors and corresponding CIs
further reviewed in full text. Discrepancies at each
were estimated in each study using Fisher’s z-trans-
stage of selection were arbitrated by consensus be-
formation.13 The I 2 index was used to assess hetero-
tween the reviewers after a further extensive review
geneity. Significant heterogeneity was defined as I 2
of the full-text papers.
>50% with a value of P < 0.05.14 The Deeks’ funnel
DATA COLLECTION. The following data were plot was used to assess for possible publication bias
collected from each study: study design, sample size, and asymmetry was tested using Egger’s regression
patient demographics, CT and CMR scanner type, CT test.15 Meta-regression analysis was then performed
protocol, radiation dose for postcontrast scan, and to explore the identified factors potentially contrib-
interval between CT and CMR scans. Additionally, uting heterogeneity in correlation and bias between
ECV values by CT and CMR, correlation, and mean CT-derived ECV and CMR-derived ECV. We per-
differences with 95% CIs were collected. formed a subgroup analysis comparing correlation

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023 Han et al 1309
OCTOBER 2023:1306–1317 CT vs CMR for ECV Quantification

T A B L E 2 Baseline Imaging Characteristics

Additional Post
Reported ECV, Radiation Contrast
First Author, CT Type, ECV Correlation Bias Location or Exposure, Timing, Interscan
Year Energy ECV CT Reference Coefficient (95% CI) Method mSv Contrast Dose min Interval

Kim et al,19 DE 34.5  6.9 34.1  6.6 — 0.44 (4.22 to 5.1) Septum — A bolus injection 12 Within 3 d
2022 of 1.0 mL/kg
(iopamidol);
additional
contrast to
attain total
concentration
of 1.6 mL/kg
Dubourg SE 29.9  4.6 29.1  3.9 0.97 2.5 (2.1 to 7.1) Global 1.89  0.38 Total 95 mL of 7 14.9  5.7
et al,20 2021 measurement iohexol; no 3-21
(mean of all additional
segments) contrast for
delayed scan
Emoto et al,18 SE 32.8  14 34.3  9.3 0.84 1.5 (14.3 to 17.3) Global 5.2  1.3 — 7 Within 4 wk
2021 DE 33.4  9.3 34.3  9.3 0.95 0.9 (5.2 to 6.9) measurement
(mean of all
segments)
Ohta et al,17 DE 31.6  9.1 33.2  9.1 0.84 1.3 (9.0 to 11.5) Per-segment 3.42  0.87 A bolus injection 7-8 1 mo
2020 results of 0.9 mL/kg
iopamidol;
additional
contrast for
1 min until the
total volume of
contrast
medium
reached
1.4 mL/kg/scan
Emoto et al,23 SE — — 0.86 0.5 (6.1 to 7.1) Septum 3.9  2.2 Approximately 7 3 mo
2020 total 500 mg
I per kg of
iodine;
additional
contrast for
delayed scan
Oda et al,21 DE — — 0.94 0.1 (5.7 to 5.7) Septum 4.8  1.6 1.5 mL/kg 7 7 (range
2019 (iopamidol) 1-26 d)
Wang et al,24 DE 33 (32-36) 30 (30-32) 0.95 2.6 (0.4 to 5.6) Global 4.21  1.05 Total 60–90 mL 7 24 h
2018 measurement (iopomide); no
(mean of all additional
segments) injection for
delayed scan
Kurita et al,16 SE — — 0.84 3.3 (11.8 to 5.1) Per-segment 1.79 — 7 —
2016 results
Lee et al,10 DE 34.5  9.0 34.2  9.0 — 0.06 (1.79 to 1.19) Global — 1.8 mL/kg 12 1 or 2 d
2016 measurement (iopamidol); no
(mean of all additional
segments) contrast for
delayed scan
Treibel et al,9 SE — — 0.85 0 (11 to 11) Septum 1.56  0.58 1 mL/kg 5 —
2015 (iodixanol); no
additional
contrast
Liu et al,22 SE — — — 1.1 (10.5 to 12.6) Global — — 7-10 —
2013 measurement
(mean of all
segments)
Bandula et al,8 SE 0.31 (0.18-0.44) 0.3 (0.25-0.41) 0.73 1.4 (31.7 to 34.7) Septum — 1 mg/kg 25 48 h
2013 (iohexol);
additional
contrast with
an infusion of
1.88 mL/kg per
hour iohexol
Nacif et al,7 SE 31.6  5.1 28.6  4.4 0.82 3.0 (2.82 to 8.85) Anterior and 1.99  0.16 125  24 mL 10 4h
2012 anterolateral (iopamidol); no
segments additional dose
for delayed
scan

DE ¼ dual energy; ECV ¼ extracellular volume fraction; SE ¼ single energy; other abbreviations as in Table 1.

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
1310 Han et al JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023

CT vs CMR for ECV Quantification OCTOBER 2023:1306–1317

C E NTR AL IL L USTR AT I O N Meta-Analysis for Correlation and Bias

A Author (Year) N Correlation COR (95% CI) Weight

Dubourg et al. (2021) 21 0.97 (0.94-1.00) 15.3%

Emoto et al. (2021) - DECT 21 0.95 (0.91-0.99) 13.7%
Wang et al. (2018) 35 0.95 (0.92-0.98) 14.7%
Oda et al. (2019) 40 0.94 (0.90-0.98) 14.3%
Emoto et al. (2020) 34 0.86 (0.77-0.95) 8.8%
Treibel et al. (2015) 53 0.85 (0.77-0.93) 10.1%
Emoto et al. (2021) - SECT 21 0.84 (0.71-0.97) 5.8%
Ohta et al. (2020) 23 0.84 (0.72-0.96) 6.2%
Kurita et al. (2016) 8 0.84 (0.62-1.06) 2.6%
Nacif et al. (2012) 24 0.82 (0.69-0.95) 5.5%
Bandula et al. (2013) 23 0.73 (0.53-0.93) 3.2%

Random effects model 0.90 (0.86-0.95) 100.0%

2
Heterogeneity: I = 64%, P < 0.01 (0.78-1.02)

0.5 0.6 0.7 0.8 0.9 1

B Author (Year) N Bias Bias (95% CI) Weight

Nacif et al. (2012) 24 3.00 (1.81-4.19) 8.4%

Wang et al. (2018) 35 2.60 (2.09-3.11) 10.2%
Dubourg et al. (2021) 21 2.50 (1.50-3.50) 9.0%
Emoto et al. (2021) - SECT 21 1.50 (–1.95 to 4.95) 3.3%
Bandula et al. (2013) 23 1.40 (–5.54 to 8.34) 1.0%
Ohta et al. (2020) 23 1.30 (–0.83 to 3.43) 5.7%
Liu et al. (2013) 20 1.10 (–1.47 to 3.67) 4.7%
Emoto et al. (2021) - DECT 21 0.90 (–0.41 to 2.21) 8.0%
Emoto et al. (2020) 34 0.50 (–0.63 to 1.63) 8.6%
Kim et al. (2021) 51 0.44 (–0.21 to 1.09) 9.9%
Lee et al. (2016) 30 0.10 (–0.10 to 0.30) 10.6%
Treibel et al. (2015) 53 0.00 (–1.51 to 1.51) 7.4%
Oda et al. (2019) 40 –0.10 (–1.02 to 0.82) 9.2%
Kurita et al. (2016) 8 –3.30 (–6.27 to –0.33) 4.0%

Random effects model 0.96 (0.14-1.78) 100.0%

Heterogeneity: I2 = 89%, P < 0.01 (–1.71 to 3.63)

–5 0 5
Han D, et al. J Am Coll Cardiol Img. 2023;16(10):1306–1317.

Meta-analysis for (A) correlation and (B) bias between cardiac magnetic resonance–derived extracellular matrix volume and computed tomography–derived
extracellular matrix volume. DECT ¼ dual-energy computed tomography; SECT ¼ single-energy computed tomography.

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023 Han et al 1311
OCTOBER 2023:1306–1317 CT vs CMR for ECV Quantification

reported SECT-derived ECV,7-9,16,20,22,23 5 studies re-

T A B L E 3 Meta-Regression Analysis
ported DECT-derived ECV, 10,17,19,21,24 and 1 study re-
Meta-Regression Standard ported both SECT- and DECT-derived ECV. 18 The mean
Covariate Coefficient Error P Value
additional radiation exposure from the postcontrast
Correlation coefficient
between CT-derived ECV scan was 1.89 to 5.2 mSv. The total contrast dose used
and CMR-derived ECV varied from 0.9 to 1.8 mL/kg, or 60 to 90 mL to 125 
Sample size 0.001 0.002 0.916
24 mL. Three of the 13 studies used a protocol with an
Publication year 0.016 0.006 0.032
additional contrast injection for postcontrast scanning
Age 0.002 0.003 0.577
after standard CT acquisition. The timing of the post-
Male, % 0.002 0.002 0.407
CMR ECV 0.002 0.014 0.897 contrast scan ranged from 5 minutes9 to 25 minutes 8
CT ECV 0.004 0.026 0.878 following contrast administration. Details for CMR
Postcontrast timing, min 0.009 0.006 0.128 imaging methods are reported in Supplemental
Bias (mean difference) Table 2.
between CT-derived ECV
and CMR-derived ECV
QUALITY ASSESSMENT AND PUBLICATION BIAS.
Sample size 0.013 0.033 0.700
The overall quality of included studies as per
Publication year 0.022 0.136 0.874
Age 0.061 0.047 0.223
QUADAS-2 was unclear for risk of bias and low for
Male, % 0.009 0.042 0.829 applicability concerns (Supplemental Table 3,
CMR ECV 0.523 0.045 <0.001 Supplemental Figure 1). Of the included studies, 7 had
CT ECV 0.540 0.169 0.015 an unclear risk of bias for the index test and reference
Postcontrast timing, min 0.007 0.046 0.964 standard due to insufficient information regarding
whether interpretation of the index and reference
Abbreviations as in Tables 1 and 2.
standard test results was blinded to clinical infor-
mation and/or other test results. Two studies were
considered to have high bias for flow and timing.
and bias between single-energy CT (SECT) and dual-
There were low concerns regarding the applicability
energy CT (DECT) method for ECV quantification.
of the index test and reference standard across all
Subgroup differences was assessed using the chi-
square test of heterogeneity (Q) between groups. included studies. Using Deeks’ test, there was no

All statistical analyses were performed with R Soft- evidence of publication bias (Supplemental Figure 2)

ware (R Foundation for Statistical Computing) using (P ¼ 0.282).

the ‘Metagen’ and ‘Metacor’ packages. CT VS CMR FOR ECV ASSESSMENT. Mean ECV
ranged from 28.6% to 34.3% by CMR and 29.9% to
RESULTS
34.5% by CT. The pooled correlation and bias are
presented in Central Illustration. Eleven studies re-
STUDY CHARACTERISTICS. A flow diagram of the
ported the correlation coefficient (r), and 13 studies
study selection process is shown in Figure 1. A total of
reported mean differences. The correlation co-
435 potentially relevant papers were retrieved in the
efficients ranged from 0.73 to 0.97. Overall, there was
initial search. We removed 92 duplicated studies and
an excellent correlation between CT-derived ECV and
excluded 274 studies based on title and abstract. After
CMR-derived ECV (pooled correlation coefficient:
performing full-text reviews of 71 studies, 6 studies
0.90 [95% CI: 0.86-0.95]). The pooled mean differ-
were excluded due to study design (5 were animal
ence in ECV was 0.96% (95% CI: 0.14%-1.78%), indi-
studies, 1 phantom study), 4 studies were excluded
cating that ECV values measured by CT were slightly
due to reporting noncardiac ECV (other organs), and 22
higher than those measured by CMR.
due to publication types (7 case report or case series, 3
abstracts, 6 letters or editorials, and 6 review papers) HETEROGENEITY AND META-REGRESSION ANALYSIS.
and 26 studies did not use CMR as the reference. The I 2 index test indicated significant heterogeneity
Finally, 13 papers were included in the meta-analysis. for both correlation and bias among the studies
The baseline characteristics of each study are listed in (Central Illustration) (I 2 ¼ 64% and 89%, respectively;
Tables 1 and 2. A total of 383 patients were included in both P < 0.01). Meta-regression analysis showed
the current analysis. The mean age range was 57.3 to publication year was a significant source of the het-
82.0 years, and 65% of patients were male. The studies erogeneity in correlation (Table 3) (coefficient: 0.016;
included various cardiac conditions, such as cardio- P ¼ 0.032) (Figure 2A). The reported correlation be-
myopathy, 10,16-19 valvular heart disease, 8,9,18,20 and tween CT-derived ECV and CMR-derived ECV in each
cardiac infiltrative disease.9,10,18,21 Seven studies study was higher in more contemporary studies. Mean

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
1312 Han et al JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023

CT vs CMR for ECV Quantification OCTOBER 2023:1306–1317

F I G U R E 2 Meta-Regression Analysis

A
1.00
Dubourg et al. (2021)

0.95

Wang et al. (2018) Emoto et al. (2021) - DECT

Oda et al. (2019)

0.90

Emoto et al. (2020)

Correlation

0.85 Treibel al. (2015)

Kurita et al. (2016)
Emoto et al. (2021) - SECT
Nacif et al. (2012)

0.80

0.75

Bandula et al. (2013)

0.70

2010 2012 2014 2016 2018 2020 2022

Publication Year

Correlation and publication year (A); bias and magnetic resonance imaging (MRI)-derived extracellular matrix volume (ECV) value (B); and bias
and computed tomography (CT)-derived ECV value (C). Abbreviations as in Figure 2.

Continued on the next page

ECV values by CMR or CT also contributed to the quantification of myocardial ECV. We demonstrated
heterogeneity in bias (Table 3) (coefficient: 0.523; excellent correlation in ECV measurements between
P < 0.001 and 0.540; P ¼ 0.015, respectively) the 2 modalities. Agreement between ECV measure-
(Figures 2B and 2C). There was a significant trend of ments by CT and CMR was also high, with an overall
decrease in reported bias in each study with higher mean difference in ECV values of <1%. In a sub-
mean CMR or CT-derived ECV values. There was no analysis of SECT vs DECT, ECV measured using DECT
significant difference in correlation and bias between was more strongly correlated with CMR-derived ECV.
studies with and without a protocol for additional CMR has been widely used for ECV assessment
contrast administration after standard CT (P for during the last decade thus providing a strong evi-
difference ¼ 0.46 and P ¼ 0.32, respectively). dence base for use of ECV in clinical practice.
SECT VS DECT. The pooled correlation and bias be- Although CT-based ECV quantification is not
tween SECT and DECT are presented in Figures 3 commonly used compared to CMR, it offers several
and 4. The pooled correlation from studies that uti- advantages. CT imaging is faster and more widely
lized DECT for ECV quantification was significantly available and clinically used across in the United
higher compared with the pooled correlation States. CT may be more acceptable in patients with
from studies using SECT (0.94 [95% CI: 0.91-0.98] vs claustrophobia and can be performed in patients with
0.87 [95% CI: 0.80-0.94), respectively; P for cardiac devices that might not be MR-compatible. CT
difference ¼ 0.01). There was no significant difference can also provide high-resolution 3-dimensional
in the pooled bias between studies using SECT vs volumetric ECV measurement with whole-heart
DECT (1.00 [95% CI: 0.58 to 2.58]) vs 0.85 acquisition and limited cardiac and patient motion.
[95% CI: 0.25 to 1.95]; P for difference ¼ 0.85). Moreover, ECV measurement with CT can be per-
formed with an additional delayed postcontrast scan,
DISCUSSION allowing its application in routine cardiac CT angi-
ography studies. The ECV measured does not require
This systematic review and meta-analysis evaluated any modification of contrast amount or scan protocol
the correlation and agreement of CT with CMR for for standard cardiac CT angiography. However, it will

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023 Han et al 1313
OCTOBER 2023:1306–1317 CT vs CMR for ECV Quantification

F I G U R E 2 Continued

Nacif et al. (2012)

2 Dubourg et al. (2021) Wang et al. (2018)

Bandula et al. (2013) Emoto et al. (2021) - SECT

Ohta et al. (2020)
Emoto et al. (2021) - DECT
Bias

0 Kim et al. (2021)

Lee et al. (2016)

–2

–4

28 29 30 31 32 33 34 35
MRI Derived ECV

Continued on the next page

increase the total radiation dose and prolong the scan ventricular ejection fraction30 and greater risk of
timing by at least 3 to 5 minutes to achieve equili- heart failure hospitalization and death 31 following
bration of the iodine concentration between plasma transcatheter aortic valve replacement. There has
and myocardium. also been increasing application of myocardial ECV
CT-derived ECV has the potential to provide addi- for the detection of cardiac amyloidosis. Among pa-
tive and complementary information to current tients with severe aortic stenosis referred for trans-
routine clinical CT assessment of various cardiac catheter aortic valve replacement, ECV measured on
diseases. Myocardial fibrosis is commonly present in preprocedural CT has been shown to detect cardiac
patients with nonischemic cardiomyopathies and amyloidosis with an accuracy of 0.87.32 Notably, the
25,26
portends poorer outcome. Since coronary assess- prevalence of concomitant cardiac amyloidosis in
ment is required for excluding obstructive coronary patients with aortic stenosis is estimated to be as high
artery disease in these patients, CT offers a suitable as 16%,33 and CT-derived ECV has the potential to
approach for simultaneously assessing myocardial become a clinically meaningful clinical imaging
fibrosis and coronary stenosis. In addition, the marker in this setting. Patients being evaluated for
delayed postcontrast scan from which ECV is derived transcatheter valve replacement routinely undergo
can be used to characterize focal myocardial scar of cardiac CT for preprocedural planning with CT, and
both acute and chronic infarction,27,28 as well as in measurement of ECV can be made from the single
various nonischemic cardiomyopathies.29 In patients preprocedural procedure. These patients seldom un-
undergoing routine planning CT before transcatheter dergo preprocedural CMR studies thus, CT-based ECV
valve interventions, ECV measurements can be quantification in the TAVR population can reduce the
readily performed with an additional delayed scan, need for additional imaging.
which carries extra radiation exposure. Recent A requirement for ECV quantification on cardiac CT
studies have shown an elevated baseline CT-derived is the additional delayed postcontrast scan. Long
ECV to associate with impaired recovery in left delay times for postcontrast imaging would reduce

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
1314 Han et al JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023

CT vs CMR for ECV Quantification OCTOBER 2023:1306–1317

F I G U R E 2 Continued

Nacif et al. (2012)

Wang et al. (2018)
2 Dubourg et al. (2021)

Bandula et al. (2013) Emoto et al. (2021) - SECT

Kim et al. (2021)
Ohta et al. (2020)
Emoto et al. (2021) - DECT
Bias

0
Lee et al. (2016)

–2

–4

29 30 31 32 33 34 35 36
CT Derived ECV

the efficiency of the CT workflow. However, the correlation and agreement between CT- and CMR-
optimal timing of the postcontrast scan has not been derived ECV. In the future, a standardized scanning
determined. The present study revealed that the protocol with shortened postcontrast timing for the
postcontrast timing varied widely among the delayed scan will likely improve the clinical utility of
included studies, ranging from 5 to 25 minutes after CT-derived ECV quantification.
contrast injection. The later delayed scan allows The measurement of ECV by CT is associated with
greater equilibration of the iodine concentration be- an increase in radiation exposure to the patient due to
tween the myocardial ECV and circulating plasma, at extra acquisitions before contrast and the delayed
a cost of reduced signal due to overall lower iodine scans. DECT is a CT technique that uses 2 separate
concentration. Shorter delay times appear to be x-ray photon energy spectra, allowing the interroga-
adequate for accurate ECV measurements. Triebel tion of materials with different attenuation properties
et al 9 showed a 5-minute postcontrast scan to have at different energies.34 Because CT-derived ECV
superior accuracy and higher signal-to-noise ratio quantification is based on the ratio of iodine distri-
compared with a 15-minute delayed scan. Our group bution between myocardium and blood, the iodine
has also demonstrated ECV quantification at 5 mi- mapping technique, which allows the quantification
nutes postcontrast to be safe and provide prognostic of iodine content in the myocardium and blood, is
value.30,31 More recently, Scully et al32 showed a 3- theoretically the best tool for CT-based ECV quanti-
minute postcontrast image acquisition to have fication. van Assen et al 35 showed that using DECT for
excellent sensitivity for the detection of cardiac ECV assessment provides similar results at a lower
amyloidosis in patients with severe aortic stenosis. radiation exposure compared with the conventional
Finally, the meta-regression analysis performed in SECT-based approach. Additionally, the DECT
the current study demonstrated that the postcontrast delayed scan may allow replacement of the true
timing is not a significant source of heterogeneity for noncontrast scan with a virtual noncontrast DECT

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023 Han et al 1315
OCTOBER 2023:1306–1317 CT vs CMR for ECV Quantification

F I G U R E 3 Subgroup Analysis SECT vs DECT: Correlation

Author (Year) N Correlation COR 95%-CI Weight

SECT
Dubourg et al. (2021) 21 0.97 [0.94; 1.00] 15.3%
Emoto et al. (2020) 34 0.86 [0.77; 0.95] 8.8%
Treibel et al. (2015) 53 0.85 [0.77; 0.93] 10.1%
Emoto et al. (2021) 21 0.84 [0.71; 0.97] 5.8%
Kurita et al. (2016) 8 0.84 [0.62; 1.06] 2.6%
Nacif et al. (2012) 24 0.82 [0.69; 0.95] 5.5%
Bandula et al. (2013) 23 0.73 [0.53; 0.93] 3.2%
Random effects model 0.87 [0.80; 0.94] 51.2%
Prediction interval [0.69; 1.04] --
Heterogeneity: I2 = 75%, P < 0.01

DECT
Emoto et al. (2021) 21 0.95 [0.91; 0.99] 13.7%
Wang et al. (2018) 35 0.95 [0.92; 0.98] 14.7%
Oda et al. (2019) 40 0.94 [0.90; 0.98] 14.3%
Ohta et al. (2020) 23 0.84 [0.72; 0.96] 6.2%
Random effects model 0.94 [0.91; 0.98] 48.8%
Prediction interval [0.90; 0.99] --
Heterogeneity: I2 = 0%, P = 0.39

0.90 [0.86; 0.95] 100.0%

[0.78; 1.02]
0.5 0.6 0.7 0.8 0.9 1

Test for subgroup differences: F21 = 6.55, df = 1 (P = 0.01)

COR ¼ correlation; DECT ¼ dual-energy computed tomography; SECT ¼ single-energy computed tomography.

reconstruction.36 In the present analysis, DECT yiel- high-quality data examining the robustness of the
ded a significantly higher correlation with CMR for CT-ECV technique in the clinical setting. The gold
ECV quantification compared with SECT. Thus, DECT standard for the detection of myocardial fibrosis is
is a promising CT approach for assessment of invasive endomyocardial biopsy. CMR-derived ECV
myocardial fibrosis. has been previously validated against histological
The meta-regression analysis in the present study collagen volume fraction in patients undergoing
indicates that study publication year and mean CT/ cardiac transplantation, with a strong correlation
CMR-derived ECV values are significant sources of between the 2 measurements shown throughout the
heterogeneity. Specifically, the correlation between entire heart.37 The present analysis used
CMR- and CT-derived ECV improved in more CMR-derived ECV as a noninvasive correlate with
contemporary studies. This may reflect the increasing well-established diagnostic and prognostic value in a
standardization of CT-based ECV quantification. variety of cardiac diseases. Finally, this study
Meanwhile, we found agreement (bias) to decrease in focused on assessing the correlation and agreement
studies with higher mean CT/CMR-derived ECV between CMR and CT-derived ECV; the diagnostic
values. These trends may indicate that CT can accu- accuracy or optimal threshold of CT-derived ECV to
rately identify patients with abnormal ECV. Future detect significant myocardial fibrosis or cardiac
studies are warranted to establish a reliable threshold amyloidosis was not within the scope of the current
of CT-derived ECV to rule out the presence of signif- study.
icant myocardial fibrosis.
CONCLUSIONS
STUDY LIMITATIONS. The overall quality of the
included studies was low, given that all were single CT-derived ECV showed excellent correlation with
center with small sample sizes, and over one-half CMR-derived ECV. The overall mean difference be-
were retrospective. This highlights the need for tween the 2 modalities was <1%. DECT yielded a

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
1316 Han et al JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023

CT vs CMR for ECV Quantification OCTOBER 2023:1306–1317

F I G U R E 4 Subgroup Analysis SECT vs DECT: Bias

Author (Year) N Bias 95%-CI Weight

SECT
Nacif et al. (2012) 24 3.00 [1.81; 4.19] 8.4%
Dubourg et al. (2021) 21 2.50 [1.50; 3.50] 9.0%
Emoto et al. (2021) 21 1.50 [–1.95; 4.95] 3.3%
Bandula et al. (2013) 23 1.40 [–5.54; 8.34] 1.0%
Liu et al. (2013) 20 1.10 [–1.47; 3.67] 4.7%
Emoto et al. (2020) 34 0.50 [–0.63; 1.63] 8.6%
Treibel et al. (2015) 53 0.00 [–1.51; 1.51] 7.4%
Kurita et al. (2016) 8 –3.30 [–6.27; –0.33] 4.0%
Random effects model 1.00 [–0.58; 2.58] 46.3%
Prediction interval [–3.13; 5.13] --
Heterogeneity: I2 = 74%, P < 0.01

DECT
Wang et al. (2018) 35 2.60 [2.09; 3.11] 10.2%
Ohta et al. (2020) 23 1.30 [–0.83; 3.43] 5.7%
Emoto et al. (2021) 21 0.90 [–0.41; 2.21] 8.0%
Kim et al. (2021) 51 0.44 [–0.21; 1.09] 9.9%
Lee et al. (2016) 30 0.10 [–0.10; 0.30] 10.6%
Oda et al. (2019) 40 –0.10 [–1.02; 0.82] 9.2%
Random effects model 0.85 [–0.25; 1.95] 53.7%
Prediction interval [–2.19; 3.89] --
Heterogeneity: I2 = 94%, P < 0.01

Random effects model 0.96 [0.14; 1.78] 100.0%

Prediction interval [–1.71; 3.63]
Heterogeneity: I2 = 89%, P < 0.01
–5 0 5

Test for subgroup differences: F21 = 0.04, df = 1 (P = 0.85)

Abbreviations as in Figure 2.

PERSPECTIVES
significantly higher correlation with CMR for ECV
quantification compared to SECT. The overall quality
of the included studies was however low, and larger, COMPETENCY IN MEDICAL KNOWLEDGE:
prospective studies are needed to examine the accu- CT-derived ECV demonstrated excellent correlation
racy and diagnostic and prognostic utility of CT- and very low bias compared with CMR-derived ECV as
derived ECV in a variety of cardiac diseases. the reference standard. DECT technique demon-
strated a significantly higher correlation with CMR for
FUNDING SUPPORT AND AUTHOR DISCLOSURES ECV quantification compared with SECT.

This work was supported in part by the Dr Miriam and Sheldon G. TRANSLATIONAL OUTLOOK: ECV measurement
Adelson Medical Research Foundation. The authors have reported
with CT can be obtainable from an additional delayed
that they have no relationships relevant to the contents of this paper
to disclose. scan to standard cardiac CT angiography. CT-based
ECV may be a useful imaging tool to evaluate
myocardial fibrosis in patients undergoing CT imaging.
ADDRESS FOR CORRESPONDENCE: Dr Daniel S.
Larger, prospective studies are needed to examine the
Berman, S Mark Taper Foundation Imaging Center,
diagnostic and prognostic utility of CT-derived ECV in
Cedars-Sinai Medical Center, 8700 Beverly
a variety of cardiac diseases.
Boulevard, Los Angeles, California 90036, USA.
E-mail: Daniel.Berman@cshs.org.

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.
JACC: CARDIOVASCULAR IMAGING, VOL. 16, NO. 10, 2023 Han et al 1317
OCTOBER 2023:1306–1317 CT vs CMR for ECV Quantification

REFERENCES

1. Anderson KR, Sutton MG, Lie JT. Histopatho- accuracy was assessed. J Clin Epidemiol. 28. Nieman K, Shapiro MD, Ferencik M, et al.
logical types of cardiac fibrosis in myocardial dis- 2005;58(9):882–893. Reperfused myocardial infarction: contrast-
ease. J Pathol. 1979;128(2):79–85. enhanced 64-Section CT in comparison to MR
16. Kurita Y, Kitagawa K, Kurobe Y, et al. Data on
imaging. Radiology. 2008;247(1):49–56.
2. Frangogiannis NG. The extracellular matrix in correlation between CT-derived and MRI-derived
ischemic and nonischemic heart failure. Circ Res. myocardial extracellular volume. Data Brief. 29. Ko SM, Hwang SH, Lee HJ. Role of cardiac
2019;125(1):117–146. 2016;7:1045–1047. computed tomography in the diagnosis of left
ventricular myocardial diseases. J Cardiovasc Im-
3. de Meester de Ravenstein C, Bouzin C, Lazam S, 17. Ohta Y, Kishimoto J, Kitao S, et al. Investiga-
aging. 2019;27(2):73–92.
et al. Histological validation of measurement of tion of myocardial extracellular volume fraction in
diffuse interstitial myocardial fibrosis by myocar- heart failure patients using iodine map with rapid- 30. Han D, Tamarappoo B, Klein E, et al.
dial extravascular volume fraction from Modified kV switching dual-energy CT: segmental compar- Computed tomography angiography-derived
Look-Locker imaging (MOLLI) T1 mapping at 3 T. ison with MRI T1 mapping. J Cardiovasc Comput extracellular volume fraction predicts early re-
J Cardiovasc Magn Reson. 2015;17(1):48. Tomogr. 2020;14(4):349–355. covery of left ventricular systolic function after
transcatheter aortic valve replacement. Eur Heart
4. Puntmann VO, Voigt T, Chen Z, et al. Native T1 18. Emoto T, Oda S, Kidoh M, et al. Myocardial
J Cardiovasc Imaging. 2021;22(2):179–185.
mapping in differentiation of normal myocardium from extracellular volume quantification using cardiac
diffuse disease in hypertrophic and dilated cardiomy- computed tomography: a comparison of the dual- 31. Tamarappoo B, Han D, Tyler J, et al. Prognostic
opathy. J Am Coll Cardiol Img. 2013;6(4):475–484. energy iodine method and the standard subtrac- value of computed tomography-derived extracel-
tion method. Acad Radiol. 2021;28(5):e119–e126. lular volume in TAVR patients with low-flow low-
5. Puntmann VO, Carr-White G, Jabbour A, et al.
gradient aortic stenosis. J Am Coll Cardiol Img.
Native T1 and ECV of noninfarcted myocardium 19. Kim NY, Im DJ, Youn JC, et al. Synthetic 2020;13(12):2591–2601.
and outcome in patients with coronary artery extracellular volume fraction derived using virtual
disease. J Am Coll Cardiol. 2018;71(7):766–778. unenhanced attenuation of blood on contrast- 32. Scully PR, Patel KP, Saberwal B, et al. Identi-
enhanced cardiac dual-energy CT in nonischemic fying cardiac amyloid in aortic stenosis: ECV
6. Treibel TA, Fridman Y, Bering P, et al. Extra-
cardiomyopathy. Am J Roentgenol. 2022;218(3): quantification by CT in TAVR patients. J Am Coll
cellular volume associates with outcomes more
454–461. Cardiol Img. 2020;13(10):2177–2189.
strongly than native or post-contrast myocardial
T1. J Am Coll Cardiol Img. 2020;13(1 Pt 1):44–54. 33. Castaño A, Narotsky DL, Hamid N, et al.
20. Dubourg B, Dacher JN, Durand E, et al. Single-
Unveiling transthyretin cardiac amyloidosis and its
7. Nacif MS, Kawel N, Lee JJ, et al. Interstitial source dual energy CT to assess myocardial extra-
predictors among elderly patients with severe
myocardial fibrosis assessed as extracellular vol- cellular volume fraction in aortic stenosis before
aortic stenosis undergoing transcatheter aortic
ume fraction with low-radiation-dose cardiac CT. transcatheter aortic valve implantation (TAVI).
valve replacement. Eur Heart J. 2017;38(38):
Radiology. 2012;264(3):876–883. Diagn Interv Imaging. 2021;102(9):561–570.
2879–2887.
8. Bandula S, White SK, Flett AS, et al. Measure- 21. Oda S, Emoto T, Nakaura T, et al. Myocardial
34. McCollough CH, Leng S, Yu L, Fletcher JG.
ment of myocardial extracellular volume fraction late iodine enhancement and extracellular volume
Dual- and multi-energy CT: principles, technical
by using equilibrium contrast-enhanced CT: vali- quantification with dual-layer spectral detector
approaches, and clinical applications. Radiology.
dation against histologic findings. Radiology. dual-energy cardiac CT. Radiol Cardiothorac Im-
2015;276(3):637–653.
2013;269(2):396–403. aging. 2019;1(1):e180003.
35. van Assen M, De Cecco CN, Sahbaee P, et al.
9. Treibel TA, Bandula S, Fontana M, et al. Extra- 22. Liu Y, Liu S, Nacif MS, et al. A framework to
Feasibility of extracellular volume quantification
cellular volume quantification by dynamic equi- measure myocardial extracellular volume fraction
using dual-energy CT. J Cardiovasc Comput
librium cardiac computed tomography in cardiac using dual-phase low dose CT images. Med Phys.
Tomogr. 2019;13(1):81–84.
amyloidosis. J Cardiovasc Comput Tomogr. 2013;40(10):103501.
2015;9(6):585–592. 36. Han D, Tamarappoo B, Kwan AC, et al. Ab-
23. Emoto T, Kidoh M, Oda S, et al. Myocardial
stract 13748: Feasibility of quantitation coronary
10. Lee HJ, Im DJ, Youn JC, et al. Myocardial extra- extracellular volume quantification in cardiac CT:
and aortic valve calcification in virtual non-
cellular volume fraction with dual-energy equilib- comparison of the effects of two different itera-
contrast images from dual-energy delayed CT
rium contrast-enhanced cardiac CT in nonischemic tive reconstruction algorithms with MRI as a
scans in patients with severe aortic stenosis. Cir-
cardiomyopathy: a prospective comparison with reference standard. Eur Radiol. 2020;30(2):691–
culation. 2021;144(suppl_1):A13748.
cardiac MR imaging. Radiology. 2016;280(1):49–57. 701.
37. Miller CA, Naish JH, Bishop P, Coutts G,
11. Whiting PF, Rutjes AW, Westwood ME, et al. 24. Wang R, Liu X, Schoepf UJ, et al. Extracellular Clark D, Zhao S, et al. Comprehensive Validation of
QUADAS-2: a revised tool for the quality assess- volume quantitation using dual-energy CT in pa- Cardiovascular Magnetic Resonance Techniques
ment of diagnostic accuracy studies. Ann Intern tients with heart failure: comparison with 3T car- for the Assessment of Myocardial Extracellular
Med. 2011;155(8):529–536. diac MR. Int J Cardiol. 2018;268:236–240. Volume. Circ Cardiovasc Imaging. 2013;6(3):373–
12. Veroniki AA, Jackson D, Viechtbauer W, et al. 25. Gulati A, Jabbour A, Ismail TF, et al. Associa- 383.
Methods to estimate the between-study variance tion of fibrosis with mortality and sudden cardiac
and its uncertainty in meta-analysis. Res Synth death in patients with nonischemic dilated car-
Methods. 2016;7(1):55–79. diomyopathy. JAMA. 2013;309(9):896–908.
13. Olkin I, Finn JD. Correlations redux. Psychol 26. Shanbhag SM, Greve AM, Aspelund T, et al. KEY WORDS cardiac magnetic resonance,
Bull. 1995;118(1):155–164. Prevalence and prognosis of ischaemic and non- computed tomography, extracellular volume,
14. Higgins JP, Thompson SG, Deeks JJ, ischaemic myocardial fibrosis in older adults. Eur meta-analysis
Altman DG. Measuring inconsistency in meta-an- Heart J. 2019;40(6):529–538.
alyses. BMJ. 2003;327(7414):557–560.
27. Gerber BL, Belge B, Legros GJ, et al. Charac-
15. Deeks JJ, Macaskill P, Irwig L. The performance terization of acute and chronic myocardial infarcts A PP END IX For supplemental figures and
of tests of publication bias and other sample size by multidetector computed tomography. Circula- tables, please see the online version of this
effects in systematic reviews of diagnostic test tion. 2006;113(6):823–833. paper.

Downloaded for Anonymous User (n/a) at West China Hospital of Sichuan University from ClinicalKey.com by Elsevier on July
25, 2024. For personal use only. No other uses without permission. Copyright ©2024. Elsevier Inc. All rights reserved.