CMR Imaging Assessing Viability in Patients With Chronic ...imaging.onlinejacc.org/content/jimg/5/5/494.full.pdf · With Chronic Ventricular Dysfunction Due to Coronary ... myocardial

J A C C : C A R D I O V A S C U L A R I M A G I N G V O L . 5 , N O . 5 , 2 0 1 2

© 2 0 1 2 B Y T H E A M E R I C A N C O L L E G E O F C A R D I O L O G Y F O U N D A T I O N I S S N 1 9 3 6 - 8 7 8 X / $ 3 6 . 0 0

P U B L I S H E D B Y E L S E V I E R I N C . D O I : 1 0 . 1 0 1 6 / j . j c m g . 2 0 1 2 . 0 2 . 0 0 9

CMR Imaging Assessing Viability in PatientsWith Chronic Ventricular Dysfunction Dueto Coronary Artery DiseaseA Meta-Analysis of Prospective Trials

Jorge Romero, MD,* Xiaonan Xue, PHD,* Waddy Gonzalez, MD,† Mario J. Garcia, MD*

Bronx and New York, New York

O B J E C T I V E S The purpose of this study was to compare the diagnostic accuracy of cardiac

magnetic resonance (CMR) assessing myocardial viability in patients with chronic left ventricular (LV)

dysfunction due to coronary artery disease using 3 techniques: 1) end-diastolic wall thickness (EDWT);

2) low-dose dobutamine (LDD); and 3) contrast delayed enhancement (DE).

B A C K G R O U N D CMR has been proposed to assess myocardial viability over the past decade.

However, the best CMR strategy to evaluate patients being contemplated for revascularization has not

yet been determined. Some centers advocate DE CMR due to its high sensitivity to identify scar, whereas

others favor the use of LDD CMR for its ability to identify contractile reserve.

M E T H O D S A systematic review of MEDLINE, Cochrane, and Embase for all the prospective trials

assessing myocardial viability in subjects with chronic LV dysfunction using CMR was performed using

a standard approach for meta-analysis for diagnostic tests and a bivariate analysis of sensitivity,

specificity, positive predictive value (PPV), and negative predictive value (NPV).

R E S U L T S A total of 24 studies of CMR evaluating myocardial viability with 698 patients fulfilled the

inclusion criteria. Eleven studies used DE, 9 studies used LDD, and 4 studies used EDWT. Our

meta-analysis indicates that among CMR methods, DE CMR provides the highest sensitivity as well as the

highest NPV (95% and 90%, respectively) for predicting improved segmental LV contractile function after

revascularization, followed by EDWT CMR, whereas LDD CMR demonstrated the lowest sensitivity/NPV

among all modalities. On the other hand, LDD CMR offered the highest specificity and PPV (91% and

93%, respectively), followed by DE CMR, whereas EDWT showed the lowest of these parameters.

C O N C L U S I O N S DE CMR provides the highest sensitivity and NPV, whereas LDD CMR provides the

best specificity and PPV. In light of these findings, integrating these 2 methods should provide increased

accuracy in evaluating patients with chronic LV dysfunction being considered for revascularization. (J Am

Coll Cardiol Img 2012;5:494–508) © 2012 by the American College of Cardiology Foundation

From the *Division of Cardiology and Montefiore-Einstein Center for Heart and Vascular Care, Montefiore Medical Center,Albert Einstein College of Medicine, Bronx, New York; and the †Division of Cardiology, St. Luke’s Roosevelt Hospital,Columbia University College of Physicians & Surgeons, New York, New York. Dr. Garcia is a consultant to TheHeart.org. Allother authors have reported that they have no relationships relevant to the contents of this paper to disclose. The use ofgadolinium contrast agents for myocardial imaging is not an FDA approved indication.

Manuscript received February 22, 2012; accepted February 23, 2012.

http://TheHeart.org

S

rplaa

tTunmmnAOmdfOOtplbT

uted tomography

J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 5 , N O . 5 , 2 0 1 2

M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.

Magnetic Resonance Assessing Myocardial Viability

495

hortly after the phrase hibernating myocar-dium was introduced in 1986 and the pos-sibility for detection of salvageable myocar-dium in coronary artery disease (CAD) was

identified (1), several different methods of assessingmyocardial viability have been implemented andtested. Viability tests have become a crucial tool inevaluating whether patients with congestive heartfailure related to CAD might benefit from revascu-larization therapy. Revascularization is accomplished

See page 509

in the form of either coronary artery bypass graft(CABG) or percutaneous coronary intervention(PCI) (2), both of which have been proven to besuperior to medical therapy in optimizing cardiaccontractility (3).

The most studied noninvasive techniques forevaluating myocardial viability are dobutaminestress echocardiography (DSE), positron emissiontomography with fluorine-18 deoxyglucose (PET-FDG), single-photon emission computed tomogra-phy (SPECT) with thallium-201 stress–redistribution–reinjection, thallium-201 late redistribution, andtechnetium-99m sestamibi (4). These modalitiesrely on the demonstration of wall motion abnor-malities, preserved myocardial metabolism, cellmembrane integrity, and intact mitochondrial func-tion in assessing the patient’s myocardial viability,respectively (5–7). Bax et al. (8) published 2 meta-analyses evaluating the accuracy of the aforemen-tioned techniques; the first study in 1997 concludedthat DSE had overall the highest predictive accu-racy in assessing myocardial viability. His secondstudy in 2001 concluded that nuclear imagingrendered higher sensitivities and negative predictivevalues (NPV), whereas dobutamine echocardiogra-phy provided higher specificities and positive pre-dictive values (PPV) (9).

Over the past decade, newer techniques such asmagnetic resonance imaging (MRI), electroanatomicmapping, and myocardial contrast echocardiographyhave been proposed to assess myocardial viability(10–12). Cardiac magnetic resonance (CMR) hasgained popularity due to technological innovationssuch as electrocardiographic gating and respiratorymotion suppression methods, which facilitate high-quality cross-sectional images of the heart with supe-rior spatial resolution (13). Unlike other imagingmodalities, CMR has the advantage of detecting
the percentage of transmural involvement in the p
ventricular wall, differentiating transmural fromsubendocardial infarcts (10).

Previous investigations have fundamentally eval-uated 3 CMR methods: 1) resting assessment of leftventricular (LV) end-diastolic wall thickness(EDWT); 2) low-dose (LDD) dobutamine stressassessment of contractile reserve; and 3) delayedcontrast enhancement (DE) to assess for scar tissue(14–16). To date, there have been 3 reviews regard-ing CMR and myocardial viability in which pooleddata from previous original investigatorswere displayed (17–19). More recently,Schinkel et al. (20) updated the work doneby Bax et al. (8,9) and also included CMRas a new technique in their analyses. Nev-ertheless, these included only a limitednumber of studies in each CMR modality,and significant differences among studieswere not accounted for. In the followingmeta-analysis, we scrutinize the accuracyof different techniques using CMR in theevaluation of myocardial viability in theextensive literature on this modality.

M E T H O D S

Search strategy. The objective of the cur-ent analysis was to evaluate the availablerospective trials in which CMR, using at

east 1 of the 3 aforementioned methods,ssessed LV regional and global functionfter revascularization.

We searched PubMed, Embase, andhe Cochrane Central Register of Clinicalrials (Cochrane Library, Issue 1, 2011)sing the terms (MRI OR magnetic reso-ance imaging OR magnetic resonance ORagnetic resonance spectroscopy OR cardiacagnetic resonance OR cardiovascular mag-etic resonance OR contrast-enhanced MRI)ND (viability OR myocardial viabilityR cardiac viability OR viability assess-ent OR viability test OR viable myocar-

ium OR ventricular dysfunction OR myocardial dys-unction OR cardiac dysfunction OR ejection fraction

R dysfunctional myocardium OR functional recoveryR hibernating myocardium). We limited our search

o humans and adults (older than 19 years of age) ineer-reviewed journals from 1966 to June 2011. No

anguage restriction was applied. The reference lists ofibliographies of identified articles were also reviewed.rials in the abstract form without a manuscript

A B B

A N D

CABG

graft

CAD �

CI � c

CMR �

reson

DE �

enhan

DSE �

echoc

EDWT

thickn

LDD �

LV �

MRI �

imagi

NPV �

PCI �

interv

PET-F

tomog

deoxy

PPV �

QUAD

of dia

instru

ROC �

chara

SPECT

comp

ublished were excluded from this analysis.

R E V I A T I O N S

A C R O N YM S

� coronary artery bypass

coronary artery disease

onfidence interval

cardiac magnetic

ance

contrast delayed

cement

dobutamine stress

ardiography

� end-diastolic wall

ess

low-dose dobutamine

left ventricular

magnetic resonance

ng

negative predictive value

percutaneous coronary

ention

GD � positron emission

raphy with fluorine-18

glucose

positive predictive value

AS � quality assessment

gnostic accuracy studies

ment

receiver-operating

cteristic

� single-photon emission

ttCPavpiplst4tct

snrbttrv

cn

tdtsscsr(tamowaomab

oavpRnpsfitaia


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


496

Selection criteria. To be included in the analysis, arial had to fulfill the following criteria: 1) prospec-ive study involving patients with CAD in whom a)MR was performed before revascularization (i.e.,CI or CABG) in order to assess viability, and b)ny current standard evaluation technique for leftentricular regional and/or global function waserformed to assess improvement after revascular-zation; 2) assessment of viability was performed inatients only with chronic stable LV dysfunction at

east 2 weeks after myocardial infarction to avoidtunning myocardium; 3) study allowed for sensi-ivity, specificity, NPV, and PPV calculations; and) there was use of standardized cutoffs for eachechnique, or the study provided enough data toalculate diagnostic and predictive accuracies usinghese cutoffs.Data extraction. Two investigators (J.R. and W.G.)extracted the data independently and in duplicate.Data was extracted using standardized protocol andreporting forms. Disagreements were resolved byarbitration (J.R. or M.J.G.), and consensus wasreached after discussion. We extracted characteris-tics of each trial, interval between revascularizationand follow-up CMR, methods, baseline demo-graphics, and number of viable and nonviable seg-ments predicted at baseline and after the revascu-larization for our analysis. In instances where thesevalues were not readily available, the main investi-gator of that particular trial was approached tosupply the relevant information.Quality assessment. To assess the quality and re-porting of studies, we evaluated 14 items that wereconsidered relevant to the review topic, based onthe quality assessment of diagnostic accuracy studiesinstrument (QUADAS) (21). Two reviewers (J.R.and W.G.) independently assessed the qualityitems, and discrepancies were resolved by consen-sus. These items covered patient spectrum, refer-ence standard, disease progression bias, verificationbias, review bias, clinical review bias, incorporationbias, test execution, study withdrawals, and indeter-minate results.Statistical analysis. Sensitivities (number of viableegments estimated by the test divided by the totalumber of segments with improved function afterevascularization), specificities (number of nonvia-le segments estimated by the test divided by theotal number of segments without improved func-ion after revascularization), PPV (segments withecovery after revascularization divided by test-
iable segments), and NPV (segments without re-
overy after revascularization divided by test-onviable segments) were calculated for every study.Several methods for meta-analysis of diagnostic

ests have been developed lately. Some methods areesigned to be used with individual patient data ofhe studies. Some methods are applicable when onlyensitivity and specificity for each study is available,uch as the situation in this paper, which is mostommonly seen in practice. A commonly used andtandard method for such situation is the summaryeceiver-operating characteristic (ROC) method22,23). This approach coverts each pair of sensi-ivity and specificity values into a single measure ofccuracy, the diagnostic odds ratio. However, sum-ary ROC does not distinguish between the ability

f detecting the sick (sensitivity) and identifying theell (specificity). Discriminating between these

bilities is important to determine the optimal usef a test in clinical practice. Therefore, in thiseta-analysis, we estimated summary sensitivity

nd specificity using a more recently developedivariate random effects model instead (24).The bivariate approach assumed logit transforms

f sensitivity and specificity from individual studiesre from a bivariate normal distribution. The bi-ariate approach is considered to be a better ap-roach as compared with the standard summaryOC approach because first, it assesses heteroge-eity across studies in sensitivity and specificity, androvides a summary estimate of sensitivity andpecificity; second, it models sensitivity and speci-city jointly so that a 95% confidence ellipse aroundhe summary estimate can be calculated; third, itllows one to directly compare sensitivity and spec-ficity between methods; further, several choices arevailable to obtain a summary ROC curve (24,25).

In this paper, the summary ROC curve was ob-tained by transforming the regression line of logitsensitivity on logit specificity into ROC space (25).A similar bivariate approached was used to modelPPV and NPV (26). Publication bias was assessedfor each technique using Egger’s, Macaskill’s, andDeeks’s methods. Deeks et al. (27) recently pointedout that Egger’s and Macaskill’s methods may bemisleading because their type I error rates aretypically inflated and can have low power whendiagnostic odds ratios are heterogeneous.

We assessed between-study heterogeneity visu-ally, by plotting sensitivity and specificity in theROC curves. We also drew summary ROC curvesand confidence regions for summary sensitivity and
specificity (24,28).

i6sCa(yaasaeC


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


497

The analyses were conducted using SAS (SASversion 9.2, 2002 to 2008, SAS Institute, Cary,North Carolina), and the figures were generatedusing R (R version 2.12.2, 2011, The R Foundationfor Statistical Computing, Vienna, Austria).Sensitivity analysis. We further evaluated whetherthe performance of each technique depends onfeatures of the technique and patient characteristics.A logistic regression for each technique was used tomodel the sensitivity on these factors. For DECMR, the standard deviation cutoff value, thefollow-up time after the procedure and the propor-tion of males and average age for the study popu-lation were examined. No factor has been identifiedthat had a significant influence on its sensitivity; forLDD, the follow-up time after the procedure and theproportion of males in the study population had asignificant impact on its sensitivity (p � 0.001 forboth): the longer the follow-up time and the moremen in the study, the higher the sensitivity; forEDWT, the follow-up time, the proportion of malesin the study, and the mean age of the population allhad a significant impact on its sensitivity; however, thelonger the follow-up and the more men in the study,the lower the sensitivity. But older age is associatedwith higher sensitivity for EDWT, and the cutoffvalue (5.5 vs. 6.0) for viability is not associated with itssensitivity.

R E S U L T S

Study selection. We identified 12,200 articles, outof which 8,705 abstracts were retrieved and re-viewed for possible inclusion (Fig. 1). Twenty-fourstudies (Tables 1, 2, and 3) enrolling 698 patients(mean age 62 years; 83% men) and a total of 6,404LV segments fulfilled the inclusion criteria andwere included in the analysis. Fourteen studies wereexcluded from the final analysis because they didnot meet the inclusion criteria: 5 used a differentCMR technique (pixel-tracking–derived myocar-dial deformation imaging [29], regional systolicmyocardial strain [30,31], diffusion-tensor CMR[32], and non–contrast-enhanced myocardial rim[33]); 3 DE CMR studies used a different cutoff(34–36); 3 studies provided no data to calculatediagnostic accuracies (37–39); 2 LDD CMR stud-ies had no data available to calculate sensitivity/specificity (40,41); 1 study used high-dose dobut-amine 15 �g/kg/min (42); and an EDWT studydid not provide data to calculate diagnostic accu-
racies (43).
Baseline characteristics. Of the 24 studies, 11 stud-es (10,44 –53) enrolling 331 patients (mean age4 years; 83% men) and analyzing 4,397 LVegments evaluated myocardial viability using DEMR, 10 studies used cine-CMR for follow up,

nd only 1 used echocardiography. Nine studies36,45,51,54 –59) with 247 patients (mean age 62ears; 79% men) and 1,120 LV segments evalu-ted myocardial viability using LDD CMR withll of them using cine-MRI for follow up, and 4tudies (45,55,59,60) with 120 patients (meange 57 years; 92% men) and 887 LV segmentsvaluated myocardial viability using EDWTMR (Tables 1, 2, and 3).

Quality assessment. Reporting was especially pooron item 11 (“Were the reference standard resultsinterpreted without knowledge of the results of theindex test?”); this refers to blinding and might haveled to inflated measures of diagnostic accuracy,which is known as a review bias. Twelve percent ofthe articles did not explain withdrawals from thestudies, indicating that test performance may intro-duce a bias. Otherwise, all the studies showedhigh-quality scores in the remaining 12 items of

Records identified through data base sources and other methods

(n=12,200)

Records after duplicate removed (n=8,705)

Records Screened(n=8,705)

Full-text assessed for eligibility(n=281)

Studies included in qualitative synthesis(38)

Records excluded on theof title and/or abstra

(n=8,424)

Full article excluded duacute heart failure AMI

Lack of revasc (n=11Lack of post-revasc follow-u

Trials excluded (n=14Incomplete data (n=5

Different cutoff for viabilitDifferent viability method

Studies included (n=24)DE (N=11)LDD (N=9)

EDWT (N=4)

Figure 1. Selection of Studies

Fourteen studies were excluded from the final analysis: 5 used a diCMR technique (pixel-tracking–derived myocardial deformation ima[29], regional systolic myocardial strain [30,31], diffusion-tensor CMRnon–contrast-enhanced myocardial rim [33]), 3 DE CMR studies useferent cutoff (34–36), 3 studies provided no data to calculate diagnaccuracies (37–39), 2 LDD CMR studies had no data available to calsensitivity/specificity (40,41), 1 study used high-dose dobutamine 1min (42), and an EDWT study did not provide data to calculate diagaccuracies (43). AMI � acute myocardial infarction; CMR � cardiacresonance; DE � contrast delayed enhancement; EDWT � end-diasthickness; LDD � low-dose dobutamine.

basisct

e to(48)2)p (n=83)

))

y (n=3) (n=6)

fferentging[32],

d a dif-osticculate5 �g/kg/nosticmagnetictolic wall

QUADAS (Figs. 2 and 3).

left ventricular ejection fra

LDD � low-dose dobutam


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


498

Publication bias. Using Egger’s or Macaskill’s meth-ods, there is no indication of publication bias forany of the 3 techniques. Likewise, Using Deeks’stest, there is no indication of publication bias forDE and EDWT CMRs (p � 0.43 and 0.47,respectively). However, a borderline significanceindicates that there might be some possibility ofpublication bias in LDD CMR (p � 0.05). Thisindicates that some studies reporting negative re-sults for this technique might not have been sub-mitted for publication, and if they were, they werenever published.

racteristics of Studies Included in the Meta-Analysis Using CMR W

StudyDesign n

Male(%)

Age(yrs)

LVEF(%) Revascularization

Follow-UpMRI

(Weeks)

ospective 21 62 59 41 CABG/PCI 36

ospective 36 84 62 39 CABG 12





ospective 12 83 61 NR CABG/PCI 12


ospective 52 NR NR 62 CABG 24

ospective 29 93 68 NR CABG/PCI 12


pass graft; CMR � cardiac magnetic resonance; DE � contrast delayed enhancemection; NR � not reported.

racteristics of Studies Included in the Meta-Analysis Using CMR W

StudyDesign n

Male(%)

Age(yrs)


Prospective 43 93 58 42 CABG/PCI

Prospective 52 48 58 41 CABG/PCI

05 Prospective 20 95 64 29 CABG

00 Prospective 10 80 69 44 PCI

1999 Prospective 25 88 58 NR CABG/PCI

Prospective 10 70 NR NR CABG/PCI

4 Prospective 40 92 57 42 CABG/PCI

04 Prospective 18 56 62 52 CABG/PCI

2004 Prospective 29 93 68 NR CABG/PCI

ine; other abbreviations as in Table 1.

Delayed enhancement CMR. A total of 11 studiesevaluated myocardial viability using DE. Thefollow-up CMR was performed between 6 and 36weeks (mean 19 weeks) after revascularization.This difference in follow-up did not reach statis-tical significance. Also, some studies used a dif-ferent standard deviation to define hyperen-hancement, ranging from 2 to 6 SD (mean of3.28 SD), which did not show any significance.Gadolinium was administered as a contrast in allstudies, and images were obtained 6 to 25 minafter administration.

DE

chnique tosess LVEF

Time AfterGadolinium

Administration(min)

Hyperenhancement(SD Above Normal

Intensity)Cutoff for

Viability (%)

CMR 15 �3 �50

CMR 13 �5 �50

CMR 15 �2 �50

CMR NR �6 �50

CMR 15 �3 �50

CMR 6 �2 �50

CMR 15 NR �50

ECHO 25 NR �50

CMR 10 �2 �50

CMR 13 NR �50

CMR 15 NR �50

SE � dobutamine stress echocardiography; ECHO � echocardiography; LVEF �

LDD

ollow-UpCMR

(Weeks)Technique toAssess LVEF

Dobutamine Dose(�g/kg/min)

Cutoff forViability (mm)

20 CMR 10 �2

NR CMR 5–10 �2

24 CMR 5–10 �2

24 CMR 5 �2

12 CMR 10 �2

6 CMR 5–10 �2

20 CMR 10 �2

36 CMR 10 �2

12 CMR 5–10 �2

Table 1. Baseline Cha ith

First Author(Ref #), Year

TeAs

Becker et al. (44),2008

Pr

Bondarenko et al.(53), 2007

Pr

Gutberlet et al.(45), 2005

Pr

Kim et al. (10),2000

Pr

Kuhl et al. (46),2006

Pr

Pegg et al. (47),2010

Pr

Sandstede et al.(48), 2000

Pr

Schvartzmanet al. (49), 2003

Pr

Selvanayagamet al. (50), 2004

Pr

Wellnhofer et al.(51), 2004

Pr

Wu et al. (52),2007

Pr

CABG � coronary artery by nt; D

Table 2. Baseline Cha ith


F

Baer et al. (55), 1998

Baer et al. (54), 2000

Gutberlet et al. (45), 20

Lauerma et al. (56), 20

Sandestede et al. (57),

Sayad et al. (58), 1998

Schmidt et al. (59), 200

Van Hoe et al. (36), 20

Wellnhofer et al. (51),

tur3

ecL

rt

cEfws9a(C6

aoc


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


499

All the studies included used 50% of LV wallhyperenhancement as a cutoff to determine whetheror not a LV segment was viable (i.e., �50%hyperenhancement was deemed viable and �50%hyperenhancement was deemed nonviable). Theweighted mean sensitivity and specificity were 95%(95% confidence interval [CI]: 93% to 97%) and51% (95% CI: 40% to 62%), whereas the PPV was69% (95% CI: 56% to 80%) and NPV was 90%(95% CI: 85% to 93%) (Table 4). This techniquehad a weighted overall accuracy of 70% (95% CI:69% to 71%).DE CMR using <0%, <25%, and <75% as cutoffs. Ofhe 11 studies evaluating hibernating myocardiumsing DE CMR, only 6 studies reported theiresults by quartiles. A total of 214 patients and,365 LV segments were analyzed.

CUTOFF <0%. The weighted sensitivity and speci-ficity for this cutoff were 0.53 (95% CI: 0.50 to0.55) and 0.87 (95% CI: 0.85 to 0.88), whereas thePPV and NPV were 0.77 (95% CI: 0.75 to 0.80)and 0.68 (95% CI: 0.66 to 0.70), respectively.

CUTOFF <25%. The weighted sensitivity and speci-ficity were 0.78 (95% CI: 0.76 to 0.80) and 0.71(95% CI: 0.69 to 0.73), whereas the PPV and NPVwere 0.70 (95% CI: 0.67 to 0.72) and 0.79 (95% CI:0.77 to 0.81), respectively.

CUTOFF <75%. The weighted sensitivity and speci-ficity for this cutoff were 0.99 (95% CI: 0.99 to1.00) and 0.21 (95% CI: 0.19 to 0.22), whereas thePPV and NPV were 0.52 (95% CI: 0.50 to 0.54)and 0.97 (95% CI: 0.96 to 0.99), respectively.Dobutamine stress CMR. A total of 9 studies werevaluated using this method to evaluate for myo-ardial viability. Each study used a 2-mm change inV wall motion during LDD infusion (5 to 10

�g/kg/min) as a cutoff to classify a segment asviable or not. Each study performed the secondCMR also between 6 and 36 weeks (mean 19weeks) after revascularization. There was no statis-

Table 3. Baseline Characteristics of Studies Included in the Met

First Author(Ref #), Year Study Design n

Male(%)

Age(yrs)

Baer et al. (55), 1998 Prospective 43 93 58

Gutberlet et al. (45), 2005 Prospective 20 95 64

Klow et al. (60), 1997 Prospective 17 88 63

Schmidt et al. (59), 2004 Prospective 40 92 57

EDWT � end-diastolic wall thickness; other abbreviations as in Table 1.

tical difference in terms of revascularization proce-

dure (CABG vs. PCI). In these studies, the meanweighted sensitivity and specificity were 81% (95%CI: 73 to 86) and 91% (95% CI: 84% to 95%),whereas the PPV and NPV were 93% (95% CI:87% to 97%) and 75% (95% CI: 65% to 83%),respectively (Table 5). The weighted overall accu-acy for this technique was 84% (95% CI: 82%o 86%).EDWT CMR. Only 4 studies fulfilled the inclusionriteria for this method. The cutoff used wasDWT of 5.5 to 6.0 mm for each study. The

ollow-up CMR was performed between 20 to 88eeks (mean 38 weeks). The mean weighted sen-

itivity and specificity were 96% (95% CI: 91% to8%) and 38% (95% CI: 23% to 57%), whereas the PPVnd NPV were 71% (95% CI: 49% to 86%) and 85%95% CI: 70% to 93%), respectively (Table 6). EDWTMR had a weighted overall accuracy of 68% (95% CI:6% to 70%).

The bivariate model showing summary diagnosticccuracies and comparing every method versus eachther are shown in Table 7. Forest plots and ROCurves are displayed in Figures 4A to 4C and 5A to 5C,

respectively.

D I S C U S S I O N

In modern medicine, viability tests are routinelyperformed on subjects in whom revascularization isbeing considered. Allman et al. (61) demonstrated astrong association between viable myocardium onnoninvasive testing and increased survival afterrevascularization with a reduction in annual mor-tality of 79.6% compared with medical therapy(3.2% vs. 16%). Hence, the 2009 guidelines for thediagnosis and management of heart failure in adultsby the American College of Cardiology and theAmerican Heart Association recommend noninva-sive imaging in patients with heart failure who haveknown CAD and no angina (Class IIa, Level ofEvidence: B), based on the fact that CABG or PCI

alysis Using CMR With EDWT


Follow-UpCMR

(Weeks)TechniqueAssess LV

42 CABG/PCI 20 CMR

29 CABG 24 CMR

40 CABG 88 CMR

42 CABG/PCI 20 CMR

a-An

toEF

Cutoff forViability (mm)

�5.5

�6

�6

�5.5

is recommended in patients with chest pain regard-

her


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


500

less of the degree of ischemia or viability (62).Nonetheless, no assessment modality is specified as

Baer (EDWT) 1998

Baer (LDD) 1998

Baer (LDD) 2000

Becker (DE) 2008

Bordarenko (DE) 2007

Gutberlet (DE) 2005

Gutberlet (EDWT) 2005

Gutberlet (LDD) 2005

Kim et al. (DE) 2000

Klow (EDWT) 1997

Kuhl (DE) 2006

Lauerma et al (LDD) 2000

Pegg (DE) 2010

Sandstede (DE) 2000

Sandstede (LDD) 1999

Sayad (LDD) 1998

Schmidt (EDWT) 2004

Schmidt (LDD) 2004

Schvartzman (DE) 2003

Selvanayagam (DE) 2004

Van Hoe (LDD) 2004

Wellnhofer (DE) 2004

Wellnhofer (LDD) 2004

Wu (DE) 2007

Rep

rese

nta

tive

sp

ectr

um

?

Sel

ecti

on

cri

teri

a d

escr

ibed

?

Acc

epta

ble

ref

eren

ce s

tan

dar

d?

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

Figure 2. Methodological Quality Summary

Good (�), indeterminate or poor (�) quality.

QUADAS � quality assessment of diagnostic accuracy studies; ot

preferable because most of them have similar diag-

nostic accuracies, and each has shortcomings, withnone of them near to be perfect. Some series have

Par

tial

ver

ific

atio

n a

void

ed?

Dif

fere

nti

al v

erif

icat

ion

avo

ided

?

Inco

rpo

rati

on

avo

ided

?

Ind

ex t

est

des

crib

ed in

det

ail t

o p

erm

it r

eplic

atio

n?

Ref

eren

ce s

tan

dar

d d

escr

ibed

in d

etai

l to

per

mit

rep

licat

ion

?

Ind

ex t

est

resu

lts

blin

ded

?

Ref

eren

ce s

tan

dar

d r

esu

lts

blin

ded

?

Rel

evan

t cl

inic

al in

form

atio

n?

Un

inte

rpre

tab

le r

esu

lts

rep

ort

ed?

Wit

hd

raw

als

exp

lain

ed?

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

–

–

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

abbreviations as in Figure 1.

Acc

epta

ble

del

ay b

etw

een

tes

ts?

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

+

suggested that CMR may be a more accurate

item


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


501

modality, but these claims have not been systemat-ically evaluated. The current study aimed at com-paring the diagnostic and predictive accuracies of 3different assessment methods that have been stud-ied.

Our meta-analysis indicates that among CMRmethods, DE CMR provides the highest sensitivityas well as the highest NPV for predicting improvedsegmental LV contractile function after revascular-ization, followed by EDWT CMR, whereas LDDCMR demonstrated the lowest sensitivity/NPVamong all modalities. On the other hand, LDD

Representativ

Selection criteri

Acceptable referen

Acceptable delay be

Partial verificat

Differential verificat

Incorporat

Index test described in detail to permi

Reference standard described in detail to permi

Index test res

Reference standard res

Relevant clinical

Uninterpretable resu

Withdrawa

Yes (high quality)

Figure 3. Methodological Quality Graph

Reporting was especially poor on item 11 (“Were the reference stanindex test?”); this refers to blinding and might have led to inflatedTwelve percent of the articles did not explain withdrawals from thewise, all the studies showed high-quality scores in the remaining 12

Table 4. Sensitivities/Specificities and Predictive Values of DE C


Sensitivity (%)Segments

Becker et al. (44), 2008 95 (215/227)

Bordarenko et al. (53), 2007 93 (79/85)

Gutberlet et al. (45), 2005 99 (198/200)

Kim et al. (10), 2000 97 (411/425)

Kuhl et al. (46), 2006 98 (94/96)

Pegg et al. (47), 2010 96 (381/397)

Sandstede et al. (48), 2000 97 (39/40)

Schvartzman et al. (49), 2003 94 (95/101)

Selvanayagam et al. (50), 2004 95 (323/340)

Wellnhofer et al. (51), 2004 90 (111/124)

Wu et al. (52), 2007 92 (142/154)

Weighted mean 95

Segments are given as % (n/N).
NPV � negative predictive value; PPV � positive predictive value; other abbreviat
CMR offered the highest specificity and PPV,followed by DE CMR, whereas EDWT showedthe lowest of these parameters. LDD CMR alsoprovided the highest diagnostic odds ratio, suggest-ing that it has the best overall performance; how-ever, the difference with the other 2 methods is notstatistically significant, therefore this result needs tobe interpreted cautiously.

By comparing these values with those reportedfor 4 different imaging modalities in 2 compellingmeta-analyses published by Bax et al. (9) andSchinkel et al. (20), it is clearly seen that DE CMR

pectrum?

scribed?

tandard?

en tests?

avoided?

avoided?

avoided?

lication?

lication?

blinded?

blinded?

rmation?

eported?

plained?

clear

0% 25% 50% 75% 100%

No (low quality)

d results interpreted without knowledge of the results of thesures of diagnostic accuracy, which is known as a review bias.dies indicating that test performance may introduce a bias. Other-s of QUADAS (quality assessment of diagnostic accuracy studies).

Specificity (%)Segments

PPV (%)Segments

NPV (%)Segments

42 (100/236) 61 (215/351) 89 (100/122)

38 (92/237) 35 (79/224) 94 (92/98)

94 (30/32) 99 (198/200) 98 (30/32)

44 (168/379) 66 (411/622) 92 (168/182)

70 (64/91) 78 (94/121) 97 (64/66)

59 (332/560) 63 (381/609) 96 (332/348)

76 (25/33) 83 (39/47) 96 (25/26)

25 (27/106) 55 (95/174) 81 (27/33)

26 (71/272) 62 (323/524) 80 (71/88)

52 (85/164) 58 (111/190) 86 (85/98)

45 (44/98) 73 (142/196) 78 (44/56)

51 69 90

e s

a de

ce s

twe

ion

ion

ion

t rep

t rep

ults

ults

info

lts r

ls ex

Un

darmeastu

MR

ions as in Table 1.


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


502

provides the highest sensitivity and NPV (95% and90%) for predicting functional improvement afterrevascularization of hibernating myocardium of anyother technique in clinical practice (i.e., PET-FDG[92% and 87%], rest-redistribution thallium-201-SPECT [87% and 79%], technetium-99msestamibi-SPECT [83% and 76%], and DSE [80%and 83%]). Similarly, LDD CMR provides thehighest specificity and PPV (91% and 93%) com-pared with any other modality, including PET-FDG (63% and 74%), rest-redistribution thallium-201 SPECT (54% and 67%), technetium-99msestamibi SPECT (65% and 74%), and DSE (78%and 75%).

Having the highest NPV, DE CMR allows forphysician confidence in evaluating the appropriatenessof revascularization therapy in selected patients.

Likewise, having the highest PPV, LDD CMRwould prevent patients from undergoing unneces-sary high-risk revascularization procedures. It hasbeen shown that the annual mortality rate is ap-proximately 3.2% after revascularization in pa-tients with viable myocardium as compared with7.7% in those without evidence of viable myocar-dium. Furthermore, the perioperative mortality isalmost insignificant for patients with viable myo-

Table 5. Sensitivities/Specificities and Predictive Values of Dobu



Baer et al. (55), 1998 89 (24/27)

Baer et al. (54), 2000 86 (24/28)

Gutberlet et al. (45), 2005 88 (183/208)

Lauerma et al. (56), 2000 75 (43/57)

Sandstede et al. (57), 1999 61 (65/106)

Sayad et al. (58), 1998 89 (25/28)

Trent et al. (42), 2000 71 (81/114)

Van Hoe et al. (36), 2004 78 (56/72)

Wellnhofer et al. (51), 2004 75 (93/124)

Weighted mean 81

Segments are given as % (n/N).Abbreviations as in Tables 1 and 4.

Table 6. Sensitivities/Specificities and Predictive Values of EDW



Baer et al. (55), 1998 94 (176/188)

Gutberlet et al. (45), 2005 96 (216/225)

Klow et al. (60), 1997 98 (63/64)

Schmidt et al. (59), 2004 100 (25/25)

Weighted mean 96

Segments are given as % (n/N).
Abbreviations as in Tables 1, 3, and 4.
cardium and as high as 10% in those withoutviability (63– 65).

Identification of hibernating myocardium in pa-tients with CAD and LV dysfunction was per-formed by Shimoni et al. (12), comparing 3 differ-ent modalities head to head: 1) myocardial contrastechocardiography; 2) thallium-201 (Tl201) scintig-raphy; and 3) dobutamine echocardiography. Thesensitivities for functional recovery were 90%, 92%,and 80%, respectively. On the other hand, thereported specificity for each technique was 63%,45%, and 54%, respectively (12). In the ischemiccascade, perfusion abnormalities occur earlier thancontractile abnormalities. This explains why severalstudies have shown higher sensitivity for nuclearperfusion imaging compared with contractile re-serve (12,66). Scar formation, which is measured byDE MRI, is the last manifestation in the ischemiccascade; that is why presence of DE should have thehighest sensitivity to predict absence of recovery.The superiority of CMR to echocardiography andnuclear techniques might be attributed to 2 mainfactors: 1) CMR has higher spatial resolution com-pared with SPECT or PET, by which it can providea more precise delineation of scar tissue; and 2)contrast-enhanced echocardiography may potentially

ine Stress CMR

ecificity (%)Segments

PPV (%)Segments

NPV (%)Segments

94 (15/16) 96 (24/25) 83 (15/18)

92 (22/24) 92 (24/26) 85 (22/26)

89 (32/36) 97 (183/187) 56 (32/57)

00 (29/0) 100 (43/43) 67 (29/43)

90 (91/101) 87 (65/75) 43 (91/132)

93 (14/15) 96 (25/26) 82 (14/17)

70 (163/232) 54 (81/150) 83 (163/196)

82 (37/45) 88 (56/64) 70 (37/53)

93 (152/164) 86 (93/105) 83 (152/183)

91 93 75

R

pecificity (%)Segments

PPV (%)Segments

NPV (%)Segments

52 (113/219) 62 (176/282) 90 (113/125)

35 (11/31) 92 (216/236) 55 (11/20)

19 (23/120) 40 (63/160) 96 (23/24)

53 (8/15) 78 (25/32) 100 (8/8)

38 71 85

tam

Sp

1

T CM

S

in Ta


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


503

identify nonperfused scarred myocardial segments, butthis technique is greatly limited by attenuation andbubble destruction artifacts (67–69).

DE StudiesBecker 2008Bordarenko 2007Gutberlet 2005Kim 2000Kuhl 2006Pegg 2010Sandstede 2000Schvartzman 2003Selvanayagam 2004Wellnhofer 2004Wu 2007

Summary

Sensitivity

0.8 0.85 0.9 0.95 1

LDD StudiesBaer 1998Baer 2000Gutberlet 2005Lauerma 2000Sandstede 1999Sayad 1998Schmidt 2004Van Hoe 2004Wellnhofer 2004

Summary

Sensitivity

0.4 0.5 0.70.6 0.8 0.9 1

EDWT StudiesBaer 1998Gutberlet 2005Klow 1997Schmidt 2004

Summary

Sensitivity

0.7 0.75 0.850.8 0.9 0.95 1

A

B

C

Figure 4. Forest Plots of Sensitivity and Specificity

(A) DE CMR, (B) LDD CMR, and (C) EDWT CMR: the size of the squaHorizontal lines are the 95% confidence intervals, and the summar

Table 7. Summary Estimates for Sensitivity, Specificity, and DO

CMR Mean Sensitivity (95% CI) Mean Specific

DE CMR 0.95 (0.93–0.97) 0.51 (0.4

LDD CMR 0.81 (0.73–0.86) 0.91 (0.8

EDWT CMR 0.96 (0.91–0.98) 0.38 (0.2

p Value DE vs. LDD �0.001 �0.0

p Value DE vs. EDWT 0.89 0.

p Value LDD vs. EDWT �0.001 �0.0

CI � confidence interval; DOR � diagnostic odds ratio; other abbreviations as

approach. Abbreviations as in Figure 1.

On the basis of our analysis, performing a LDDCMR may be useful in high operative risk patientswho have a positive DE CMR in order to obtain a

DE StudiesBecker 2008Bordarenko 2007Gutberlet 2005Kim 2000Kuhl 2006Pegg 2010Sandstede 2000Schvartzman 2003Selvanayagam 2004Wellnhofer 2004Wu 2007

Summary

Specificity

0.2 0.4 0.6 0.8 1

LDD StudiesBaer 1998Baer 2000Gutberlet 2005Lauerma 2000Sandstede 1999Sayad 1998Schmidt 2004Van Hoe 2004Wellnhofer 2004

Summary

Specificity

0.6 0.7 0.90.8 1

EDWT StudiesBaer 1998Gutberlet 2005Klow 1997Schmidt 2004

Summary

Specificity

0.1 0.2 0.40.3 0.5 0.70.6 0.8

lotting symbol is proportional to the same size for each study.nsitivity and specificity are calculated based on the bivariate

om the Bivariate Model

(95% CI) Mean DOR (95% CI) PPV (95% CI) NPV (95% CI)

62) 21.12 (10.98–40.55) 0.69 (0.56–0.80) 0.90 (0.85–0.93)

95) 41.57 (18.25–94.68) 0.93 (0.87–0.97) 0.75 (0.65–0.83)

57) 13.33 (4.16–42.74) 0.71 (0.49–0.86) 0.85 (0.70–0.93)

0.21 �0.001 0.001

0.34 0.87 0.37

0.08 0.01 0.21

bles 1, 2, 3, and 4.

re py se

R Fr

ity

0–0.

4–0.

3–0.

01

25

01

ations as in Figure 1.


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


504

higher specificity and thus higher PPV. Conversely,DE CMR could similarly be added to LDD CMRin low operative risk patients with a negative LDDtest to obtain a higher sensitivity and NPV. SinceDE CMR is easier to perform and carries a lowercomplication risk and since its sensitivity/NPV ishigher than LDD CMR’s specificity/PPV, DECMR should probably be used as the first-lineassessment. This approach has previously suggestedby Kaandorp et al. (70); in their study, DE CMRand LDD CMR were compared head to head onthe same patients in order to evaluate myocardialviability. The authors concluded that LDD CMRmight be an option when patients have an interme-diate extent of scar to ultimately differentiate viablefrom nonviable myocardium (70).

It is worth pointing out that the low specificityand PPV of DE CMR might be due to the fact thatinvestigators have had to use a particular cutoff of50% for research purposes in order to differentiateviable from nonviable myocardium. As describedfirst by Kim et al. (10) and Van Hoe et al. (36) a fewyears later, myocardial viability should not be inter-preted as an all-or-none phenomenon for 2 reasons:1) although segments without any degree of hyper-enhancement have a high PPV (92%), segmentswith partial hyperenhancement (0% to 50%) have aless predictable response to revascularization; and2) different degrees of wall motion abnormalitieshave a major impact on myocardial recovery, withsegments showing akinesia and dyskinesia demon-strating the best recovery (sensitivity 92%, NPV96%, specificity 86%, and PPV 73%) (10,36).

In order to take the aforementioned situationinto account, we carried out a subanalysis on DECMR studies that reported their findings by quar-tiles. Our results mainly showed that using a cutoffof �0% to define viable myocardium significantlyimproves the specificity and PPV of this techniqueas compared with using a cutoff of 50%. On thisbasis, patients without any degree of hyperen-hancement might not require LDD CMR tooptimally differentiate viable myocardium. Yet, itis important to remember that the values ob-tained using a cutoff of �0% for DE CMR arenot as high as those provided by LDD CMR(87%, 77% vs. 91%, 93%).

In addition to providing 1 of the best diagnosticaccuracy values, CMR can also provide valuableinformation in assessing LV function, LV volumes,and the presence of either functional mitral regur-gitation or aneurysm that might be suitable for

Sen

siti

vity

1.0

0.8

0.6

0.4

0.2

0.0

0.0 0.40.2 0.80.6 1.0

1-Specificity

DE CMR

Sen

siti

vity

1.0

0.8

0.6

0.4

0.2

0.0

0.0 0.40.2 0.80.6 1.0

1-Specificity

LDD CMR

Sen

siti

vity

1.0

0.8

0.6

0.4

0.2

0.0

0.0 0.40.2 0.80.6 1.0

1-Specificity

EDWT CMR

A

B

C

Figure 5. ROC Curves

Bivariate summary estimate of sensitivity and specificity for (A) DECMR, (B) LDD CMR, and (C) EDWT CMR and the corresponding95% confidence ellipse around its mean value. The solid square inthe center represents the mean specificity and specificity. Each cir-cle represents an individual study, whereas the size of each circle isproportional to the sample size for each study. The solid curve isthe summary receiver-operating characteristic (ROC) curve. Abbrevi-

repair at the time of CABG (71,72). Improvement

ufmh

B

DTwablameovnmtiprbDsfatcc


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


505

of LV systolic dysfunction is clinically importantbecause LV ejection fraction is a major determinantof survival in patients with CAD (73). Although,no CMR study among the trials in our meta-analysisevaluated global LV function in terms of diagnosticaccuracy, the ejection fraction increased from 41% to47% (p � 0.007) in DE CMR studies, from 42% to48% (p � 0.03) in LDD CMR, and from 37% to 41%(p � 0.34) in EDWT CMR.

Two shortcomings of CMR are worth mention-ing. First, CMR is contraindicated in many patientswith metallic implants. However, devices such asintravascular stents, most prosthetic cardiac valves,and prosthetic joints placed within the last 2 de-cades are considered “MRI safe.” Pacemakers andimplantable cardioverter-defibrillators are still con-sidered a strong relative contraindication to CMRexamination due to the risk of arrhythmia induc-tion, device movement, and especially “lead heat-ing” (74). However, a recent randomized clinicaltrial evaluated CMR safety of a new implantablecardiac device (MRI SureScan pacemaker system,Medtronic, Minneapolis, Minnesota) in 200 pa-tients, with no CMR-related complications re-ported during or after the test (75,76).

Second, nephrogenic systemic fibrosis is a devas-tating (albeit extremely rare) potential complicationin patients exposed to gadolinium-based contrastagents. This complication occurs almost exclusivelyin patients with moderate to severe renal disease,particularly those on dialysis (77,78). The Food andDrug Administration currently defines patientswith an estimated glomerular filtration rate of �30ml/min as “at risk” for this complication.Clinical implications. In light of having no single testto accurately assess myocardial viability, it may beadvisable to combine the assessment of the degree oftransmurality scar tissue with DE CMR and theassessment of the contractile reserve with LDD CMRto obtain the best possible diagnostic and prognosticinformation. Here, it has been clearly demonstratedthat DE CMR provides the highest sensitivity andNPV for predicting LV recovery of functionality afterrevascularization, whereas LDD CMR provides thehighest specificity and PPV of any currently availabletest to evaluate myocardial viability. Supported bythese results, patients with �50% of DE shouldndergo LDD CMR to precisely differentiate viablerom nonviable myocardium. The latter techniqueight not be entirely necessary if patients have no

yperenhancement in the LV wall.The recent multicenter trial of “Coronary-Artery

ypass Surgery in Patients with Left Ventricular

ysfunction” published by the STICH (Surgicalreatment for Ischemic Heart Failure) Investigatorsas carried out in order to evaluate whether there is

ny benefit in terms of morbidity and mortalityetween medical therapy as compared with revascu-arization therapy (79). This trial showed a significant,lthough modest, reduction in the risk of death,yocardial infarction, and other major cardiovascular

vents in patients randomized to CABG (79). More-ver, a subanalysis of the STICH trial assessing thealue of myocardial viability assessment suggested thateither echocardiographic nor nuclear scintigraphicethods helped in selecting patients for revasculariza-

ion (80). Hence, this study highlights the need tonvestigate the role of more reliable tests to evaluateatients with chronic LV dysfunction in whom aevascularization intervention is being planned. It haseen clearly demonstrated in prospective trials thatE CMR has very low intraobserver and interob-

erver variability (81). Moreover CMR can also allowor a more comprehensive evaluation of virtually everyspect of the cardiac anatomy and function. Whetherhis will translate into increased ability to identifyhronic ischemic patients who will benefit from revas-ularization remains to be determined.Study limitations. An appropriate assessment of thediagnostic accuracy for EDWT CMR was notreached due to the limited number of studies andthe high level of heterogeneity in specificity. Therewere few studies using CMR for the prediction ofrecovery of global LV function after revasculariza-tion. Similarly, no test evaluated improvement inheart failure symptoms or exercise capacity. Giventhe fact that a viability test is currently often chosenbased on physician preferences, availability, or ex-perience with the test, it might not be suitable toextrapolate these results widely. Other methods toevaluate myocardial viability using CMR have alsobeen recently proposed. However, due to the smallnumber of studies, they could not be included inthis meta-analysis.

DE CMR and LDD CMR diagnostic accuraciescould not be statistically combined in order toevaluate how adding these 2 methods will improvesensitivity/specificity upon each technique sepa-rately, given the fact that only 2 studies imple-mented both LDD and DE techniques on the samepatients, and patient-level data was not provided inthese articles. Finally, results from LDD CMRshould be cautiously interpreted, given the border-line significance for publication bias demonstrated
by Deeks’s test.

DHM

1


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


506

C O N C L U S I O N S

Among CMR viability methods, DE CMR pro-vides the highest sensitivity and NPV. Likewise,LDD CMR provides the highest specificity andPPV of any other modality. In light of thesefindings, integrating these 2 methods should pro-

resonance imaging to identify revers-

1

1

1

1

1

1

1

1

1

Cardiovascular magnaging for detection

chronic LV dysfunction being considered for revas-cularization.

Reprint requests and correspondence: Dr. Mario J. Garcia,ivision of Cardiology, Montefiore-Einstein Center foreart and Vascular Care, Albert Einstein College ofedicine, 111 East 210th Street, Silver Zone, Bronx,

vide increased accuracy in evaluating patients with New York 10467-2400. E-mail: [email protected].

2

2

2

2

2

2

2

2

2

R E F E R E N C E S

1. Braunwald E, Rutherford JD. Revers-ible ischemic left ventricular dysfunc-tion: evidence for the “hibernatingmyocardium.” J Am Coll Cardiol1986;8:1467–70.

2. Tillisch J, Brunken R, Marshall R, etal. Reversibility of cardiac wall-motion abnormalities predicted bypositron tomography. N Engl J Med1986;314:884–8.

3. Alderman EL, Fisher LD, Litwin P,et al. Results of coronary artery sur-gery in patients with poor left ventric-ular function (CASS). Circulation1983;68:785–95.

4. Dilsizian V, Bonow RO. Current di-agnostic techniques of assessing myo-cardial viability in patients with hiber-nating and stunned myocardium.Circulation 1993;87:1–20.

5. Schelbert HR. Metabolic imaging toassess myocardial viability. J NuclMed 1994;35:8S–14.

6. Weich HF, Strauss HW, Pitt B. Theextraction of thallium-201 by the myocar-dium. Circulation 1977;56:188–91.

7. Takahashi N, Dahlberg ST, GilmoreMP, Leppo JA. Effects of acute isch-emia and reperfusion on the myocar-dial kinetics of technetium 99m-labeled tetrofosmin and thallium-201.J Nucl Cardiol 1997;4:524–31.

8. Bax JJ, Wijns W, Cornel JH, VisserFC, Boersma E, Fioretti PM. Accu-racy of currently available techniquesfor prediction of functional recoveryafter revascularization in patients withleft ventricular dysfunction due tochronic coronary artery disease: com-parison of pooled data. J Am CollCardiol 1997;30:1451–60.

9. Bax JJ, Poldermans D, Elhendy A,Boersma E, Rahimtoola SH. Sensitiv-ity, specificity, and predictive accura-cies of various noninvasive techniquesfor detecting hibernating myocar-dium. Curr Probl Cardiol 2001;26:147–86.

0. Kim RJ, Wu E, Rafael A, et al. Theuse of contrast-enhanced magnetic

ible myocardial dysfunction. N EnglJ Med 2000;343:1445–53.

1. Botker HE, Lassen JF, Hermansen F,et al. Electromechanical mapping fordetection of myocardial viability inpatients with ischemic cardiomyopa-thy. Circulation 2001;103:1631–7.

2. Shimoni S, Frangogiannis NG, AggeliCJ, et al. Identification of hibernatingmyocardium with quantitative intrave-nous myocardial contrast echocardiog-raphy: comparison with dobutamineechocardiography and thallium-201scintigraphy. Circulation 2003;107:538–44.

3. Hundley WG, Meshack BM, WillettDL, et al. Comparison of quantitationof left ventricular volume, ejectionfraction, and cardiac output in patientswith atrial fibrillation by cine mag-netic resonance imaging versus inva-sive measurements. Am J Cardiol1996;78:1119–23.

4. Perrone-Filardi P, Bacharach SL,Dilsizian V, et al. Metabolic evidenceof viable myocardium in regions withreduced wall thickness and absent wallthickening in patients with chronicischemic left ventricular dysfunction.J Am Coll Cardiol 1992;20:161–8.

5. Baer FM, Voth E, LaRosee K, et al.Comparison of dobutamine trans-esophageal echocardiography and do-butamine magnetic resonance imagingfor detection of residual myocardialviability. Am J Cardiol 1996;78:415–9.

6. Simonetti OP, Kim RJ, Fieno DS, etal. An improved MR imaging tech-nique for the visualization of myocar-dial infarction. Radiology 2001;218:215–23.

7. Bax JJ, de Roos A, van der Wall EE.Assessment of myocardial viability byMRI. J Magn Reson Imaging 1999;10:418–22.

8. Kaandorp TA, Lamb HJ, van derWall EE, de Roos A, Bax JJ. Cardio-vascular MR to access myocardial vi-ability in chronic ischaemic LV dys-function. Heart 2005;91:1359–65.

9. Glaveckaite S, Valeviciene N, LauceviciusA, Celutkiene J, Rudys A, Tamosiunas A.

etic resonance im-of myocardial via-

bility in chronic ischemic left ventric-ular dysfunction. Medicina (Kaunas)2009;45:585–99.

0. Schinkel AF, Bax JJ, Poldermans D,Elhendy A, Ferrari R, RahimtoolaSH. Hibernating myocardium: diag-nosis and patient outcomes. CurrProbl Cardiol 2007;32:375–410.

1. Whiting P, Rutjes AW, Reitsma JB,Bossuyt PM, Kleijnen J. The develop-ment of QUADAS: a tool for thequality assessment of studies of diag-nostic accuracy included in systematicreviews. BMC Med Res Methodol2003;3:25.

2. Littenberg B, Moses LE. Estimatingdiagnostic accuracy from multipleconflicting reports: a new meta-analytic method. Med Decis Making1993;13:313–21.

3. Moses LE, Shapiro D, Littenberg B.Combining independent studies of adiagnostic test into a summary ROCcurve: data-analytic approaches andsome additional considerations. StatMed 1993;12:1293–316.

4. Reitsma JB, Glas AS, Rutjes AW,Scholten RJ, Bossuyt PM, ZwindermanAH. Bivariate analysis of sensitivity andspecificity produces informative sum-mary measures in diagnostic reviews.J Clin Epidemiol 2005;58:982–90.

5. Arends LR, Hamza TH, van Hou-welingen JC, Heijenbrok-Kal MH,Hunink MG, Stijnen T. Bivariaterandom effects meta-analysis of ROCcurves. Med Decis Making 2008;28:621–38.

6. Chu H, Nie L, Cole SR, Poole C.Meta-analysis of diagnostic accuracystudies accounting for disease preva-lence: alternative parameterizationsand model selection. Stat Med 2009;28:2384–99.

7. Deeks JJ, Macaskill P, Irwig L. Theperformance of tests of publicationbias and other sample size effects insystematic reviews of diagnostic testaccuracy was assessed. J Clin Epide-miol 2005;58:882–93.

8. Harbord RM, Deeks JJ, Egger M,Whiting P, Sterne JA. A unificationof models for meta-analysis of diag-
nostic accuracy studies. Biostatistics2007;8:239–51.
mailto:[email protected]

3

3

3

3

3

3

3

4

4

4

4

4

4

4

4

4

4

5

5

5

5

5

5

5

5


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


507

29. Hoffmann R, Stempel K, Kuhl H, etal. Integrated analysis of cardiac tissuestructure and function for improvedidentification of reversible myocardialdysfunction. Coron Artery Dis 2009;20:21–6.

30. Maniar HS, Cupps BP, Potter DD, etal. Ventricular function after coronaryartery bypass grafting: evaluation bymagnetic resonance imaging andmyocardial strain analysis. J ThoracCardiovasc Surg 2004;128:76–82.

31. Potter DD, Araoz PA, McGee KP,Harmsen WS, Mandrekar JN, SundtTM 3rd. Low-dose dobutamine car-diac magnetic resonance imaging withmyocardial strain analysis predictsmyocardial recoverability after coro-nary artery bypass grafting. J ThoracCardiovasc Surg 2008;135:1342–7.

32. Wu MT, Su MY, Huang YL, et al.Sequential changes of myocardialmicrostructure in patients postmyo-cardial infarction by diffusion-tensorcardiac MR: correlation with left ven-tricular structure and function. CircCardiovasc Imaging 2009;2:32–40.

3. Rassaf T, Nolte J, Heussen N, et al.Quantitation of the thickness of thenon-enhanced myocardial rim pre-dicts recovery of territorial myocardialfunction in chronic ischemic heart dis-ease: a cardiac magnetic resonanceimaging study. Clin Res Cardiol2010;99:293–300.

4. Krittayaphong R, Laksanabunsong P,Maneesai A, Saiviroonporn P, Udom-punturak S, Chaithiraphan V. Com-parison of cardiovascular magneticresonance of late gadolinium enhance-ment and diastolic wall thickness topredict recovery of left ventricularfunction after coronary artery bypasssurgery. J Cardiovasc Magn Reson2008;10:41.

5. Kirschbaum SW, Baks T, van den EntM, et al. Evaluation of left ventricularfunction three years after percutane-ous recanalization of chronic total cor-onary occlusions. Am J Cardiol 2008;101:179–85.

6. Van Hoe L, Vanderheyden M. Isch-emic cardiomyopathy: value of differentMRI techniques for prediction of func-tional recovery after revascularization.AJR Am J Roentgenol 2004;182:95–100.

7. Beek AM, Bondarenko O, Afshar-zada F, van Rossum AC. Quantifica-tion of late gadolinium enhancedCMR in viability assessment inchronic ischemic heart disease: a com-parison to functional outcome. J Car-diovasc Magn Reson 2009;11:6.

8. Ugander M, Cain PA, Johnsson P,Palmer J, Arheden H. Chronic non-transmural infarction has a delayed
recovery of function following revas-
cularization. BMC Cardiovasc Disord2010;10:4.

9. Knuesel PR, Nanz D, Wyss C, et al.Characterization of dysfunctionalmyocardium by positron emission to-mography and magnetic resonance:relation to functional outcome afterrevascularization. Circulation 2003;108:1095–100.

0. Gunning MG, Anagnostopoulos C,Knight CJ, et al. Comparison of201Tl, 99mTc-tetrofosmin, and dobut-amine magnetic resonance imaging foridentifying hibernating myocardium.Circulation 1998;98:1869–74.

1. Kaandorp TA, Bax JJ, Bleeker SE, et al.Relation between regional and globalsystolic function in patients with isch-emic cardiomyopathy after beta-blockertherapy or revascularization. J Cardio-vasc Magn Reson 2010;12:7.

2. Trent RJ, Waiter GD, Hillis GS,McKiddie FI, Redpath TW, WaltonS. Dobutamine magnetic resonanceimaging as a predictor of myocardialfunctional recovery after revascularisa-tion. Heart 2000;83:40–6.

3. Sharma R, Katz JK. Increased myocar-dial wall thickening as index of viabilityassessment: a preliminary report on de-layed contrast MRI. Contrast MediaMol Imaging 2009;4:37– 41.

4. Becker M, Lenzen A, Ocklenburg C,et al. Myocardial deformation imagingbased on ultrasonic pixel tracking toidentify reversible myocardial dys-function. J Am Coll Cardiol 2008;51:1473–81.

5. Gutberlet M, Frohlich M, Mehl S, etal. Myocardial viability assessment inpatients with highly impaired left ven-tricular function: comparison of de-layed enhancement, dobutamine stressMRI, end-diastolic wall thickness,and TI201-SPECT with functionalrecovery after revascularization. EurRadiol 2005;15:872–80.

6. Kuhl HP, Lipke CS, Krombach GA,et al. Assessment of reversible myocar-dial dysfunction in chronic ischaemicheart disease: comparison of contrast-enhanced cardiovascular magnetic res-onance and a combined positronemission tomography-single photonemission computed tomography im-aging protocol. Eur Heart J 2006;27:846–53.

7. Pegg TJ, Selvanayagam JB, Jennifer J,et al. Prediction of global left ventric-ular functional recovery in patientswith heart failure undergoing surgicalrevascularisation, based on late gado-linium enhancement cardiovascularmagnetic resonance. J CardiovascMagn Reson 2010;12:56.

8. Sandstede JJ, Lipke C, Beer M, et al.Analysis of first-pass and delayed
contrast-enhancement patterns ofdysfunctional myocardium on MR
imaging: use in the prediction of myo-cardial viability. AJR Am J Roent-genol 2000;174:1737–40.

9. Schvartzman PR, Srichai MB, GrimmRA, et al. Nonstress delayed-en-hancement magnetic resonance im-aging of the myocardium predictsimprovement of function after revas-cularization for chronic ischemicheart disease with left ventriculardysfunction. Am Heart J 2003;146:535– 41.

0. Selvanayagam JB, Kardos A, Francis JM,et al. Value of delayed-enhancement car-diovascular magnetic resonance imag-ing in predicting myocardial viabilityafter surgical revascularization. Circu-lation 2004;110:1535–41.

1. Wellnhofer E, Olariu A, Klein C, etal. Magnetic resonance low-dose do-butamine test is superior to SCARquantification for the prediction offunctional recovery. Circulation 2004;109:2172–4.

2. Wu YW, Tadamura E, YamamuroM, et al. Comparison of contrast-enhanced MRI with (18)F-FDGPET/201Tl SPECT in dysfunctionalmyocardium: relation to early functionaloutcome after surgical revascularization inchronic ischemic heart disease. J NuclMed 2007;48:1096–103.

3. Bondarenko O, Beek AM, Nijveldt R,et al. Functional outcome after revas-cularization in patients with chronicischemic heart disease: a quantitativelate gadolinium enhancement CMRstudy evaluating transmural scar ex-tent, wall thickness and periproceduralnecrosis. J Cardiovasc Magn Reson2007;9:815–21.

4. Baer FM, Theissen P, Crnac J, et al.Head to head comparison of dobutamine-transoesophageal echocardiography anddobutamine-magnetic resonance im-aging for the prediction of left ventric-ular functional recovery in patientswith chronic coronary artery disease.Eur Heart J 2000;21:981–91.

5. Baer FM, Theissen P, Schneider CA,et al. Dobutamine magnetic resonanceimaging predicts contractile recoveryof chronically dysfunctional myocardiumafter successful revascularization. J AmColl Cardiol 1998;31:1040–8.

6. Lauerma K, Niemi P, Hanninen H, etal. Multimodality MR imaging assess-ment of myocardial viability: combi-nation of first-pass and late contrastenhancement to wall motion dynam-ics and comparison with FDG PET-initial experience. Radiology 2000;217:729–36.

7. Sandstede JJ, Bertsch G, Beer M, etal. Detection of myocardial viability bylow-dose dobutamine Cine MR imag-
ing. Magn Reson Imaging 1999;17:1437–43.

6

6

6

6

7

7

7

7


M A Y 2 0 1 2 : 4 9 4 – 5 0 8

Romero et al.


508

58. Sayad DE, Willett DL, HundleyWG, Grayburn PA, Peshock RM.Dobutamine magnetic resonance im-aging with myocardial tagging quan-titatively predicts improvement in re-gional function after revascularization.Am J Cardiol 1998;82:1149–51; A10.

59. Schmidt M, Voth E, Schneider CA,et al. F-18-FDG uptake is a reliablepredictory of functional recovery ofakinetic but viable infarct regions asdefined by magnetic resonance imag-ing before and after revascularization.Magn Reson Imaging 2004;22:229–36.

60. Klow NE, Smith HJ, Gullestad L,Seem E, Endresen K. Outcome ofbypass surgery in patients with chronicischemic left ventricular dysfunction.Predictive value of MR imaging. ActaRadiol 1997;38:76–82.

61. Allman KC, Shaw LJ, HachamovitchR, Udelson JE. Myocardial viabilitytesting and impact of revascularizationon prognosis in patients with coronaryartery disease and left ventricular dys-function: a meta-analysis. J Am CollCardiol 2002;39:1151–8.

62. Hunt SA, Abraham WT, Chin MH,et al., American College of CardiologyFoundation, American Heart Associ-ation. 2009 Focused update incorpo-rated into the ACC/AHA 2005Guidelines for the Diagnosis andManagement of Heart Failure inAdults: a report of the American Col-lege of Cardiology Foundation/American Heart Association TaskForce on Practice Guidelines. J AmColl Cardiol 2009;53:e1–90.

63. Pagley PR, Beller GA, Watson DD,Gimple LW, Ragosta M. Improvedoutcome after coronary bypass surgeryin patients with ischemic cardiomyop-athy and residual myocardial viability.Circulation 1997;96:793–800.

64. Weinstein R, Martinez R, HassounH, Palek J. Does a patient with he-reditary spherocytosis qualify for pre-operative autologous blood donation?Transfusion 1997;37:1179–83.

65. Bax JJ, Poldermans D, Elhendy A, etal. Improvement of left ventricularejection fraction, heart failure symptoms
and prognosis after revascularization in
patients with chronic coronary artery dis-ease and viable myocardium detected bydobutamine stress echocardiography.J Am Coll Cardiol 1999;34:163–9.

6. Panza JA, Dilsizian V, Laurienzo JM,Curiel RV, Katsiyiannis PT. Relationbetween thallium uptake and contrac-tile response to dobutamine. Implica-tions regarding myocardial viability inpatients with chronic coronary arterydisease and left ventricular dysfunc-tion. Circulation 1995;91:990–8.

7. Wagner A, Mahrholdt H, Holly TA,et al. Contrast-enhanced MRI androutine single photon emission com-puted tomography (SPECT) perfu-sion imaging for detection of suben-docardial myocardial infarcts: animaging study. Lancet 2003;361:374–9.

8. Garvin AA, Cullom SJ, Garcia EV.Myocardial perfusion imaging usingsingle-photon emission computed to-mography. Am J Card Imaging 1994;8:189–98.

9. Kuikka JT, Yang J, Kiiliainen H.Physical performance of the SiemensE.CAM gamma camera. Nucl MedCommun 1998;19:457–62.

0. Kaandorp TA, Bax JJ, Schuijf JD, etal. Head-to-head comparison betweencontrast-enhanced magnetic reso-nance imaging and dobutamine mag-netic resonance imaging in men withischemic cardiomyopathy. Am J Car-diol 2004;93:1461–4.

1. Lawson MA, Blackwell GG, DavisND, Roney M, Dell’Italia LJ, PohostGM. Accuracy of biplane long-axisleft ventricular volume determined bycine magnetic resonance imaging inpatients with regional and global dys-function. Am J Cardiol 1996;77:1098–104.

2. Athanasuleas CL, Buckberg GD,Stanley AW, et al. Surgical ventricularrestoration in the treatment of conges-tive heart failure due to post-infarction ventricular dilation. J AmColl Cardiol 2004;44:1439–45.

3. The Multicenter Postinfarction Re-search Group. Risk stratification andsurvival after myocardial infarction.
N Engl J Med 1983;309:331–6. v
74. Levine GN, Gomes AS, Arai AE, etal. Safety of magnetic resonance im-aging in patients with cardiovasculardevices: an American Heart Associa-tion scientific statement from theCommittee on Diagnostic and Inter-ventional Cardiac Catheterization,Council on Clinical Cardiology, andthe Council on Cardiovascular Radi-ology and Intervention. Circulation2007;116:2878–91.

75. Sutton R, Kanal E, Wilkoff BL, et al.Safety of magnetic resonance imagingof patients with a new MedtronicEnRhythm MRI SureScan pacingsystem: clinical study design. Trials2008;9:68.

76. Wilkoff BL, Bello D, Taborsky M, etal. Magnetic resonance imaging inpatients with a pacemaker system de-signed for the magnetic resonance en-vironment. Heart Rhythm;8:65–73.

77. Deo A, Fogel M, Cowper SE. Neph-rogenic systemic fibrosis: a populationstudy examining the relationship ofdisease development to gadoliniumexposure. Clin J Am Soc Nephrol2007;2:264–7.

78. Shabana WM, Cohan RH, Ellis JH,et al. Nephrogenic systemic fibrosis: areport of 29 cases. AJR Am J Roent-genol 2008;190:736–41.

79. Velazquez EJ, Lee KL, Deja MA, et al.Coronary-artery bypass surgery in pa-tients with left ventricular dysfunction.N Engl J Med 2011;364:1607–16.

80. Bonow RO, Maurer G, Lee KL, et al.Myocardial viability and survival inischemic left ventricular dysfunction.N Engl J Med 2011;364:1617–25.

81. Bulow H, Klein C, Kuehn I, et al.Cardiac magnetic resonance imaging:long term reproducibility of the lateenhancement signal in patients withchronic coronary artery disease. Heart2005;91:1158–63.

Key Words: cardiac magneticresonance y left ventriculardysfunction y myocardial
iability y revascularization.

Documents

CMR Imaging Assessing Viability in Patients With Chronic ...imaging.onlinejacc.org/content/jimg/5/5/494.full.pdf · With Chronic Ventricular Dysfunction Due to Coronary ... myocardial