Accuracy of cardiac magnetic resonance imaging to rule out significant coronary artery disease in patients with systolic heart failure of unknown aetiology: Single-centre experience and comprehensive meta-analysis.

(1)

ArchivesofCardiovascularDisease(2018)111,686—701

Availableonlineat

ScienceDirect

www.sciencedirect.com

CLINICAL

RESEARCH

Accuracy

of

cardiac

magnetic

resonance

imaging

to

rule

out

signiﬁcant

coronary

artery

disease

in

patients

with

systolic

heart

failure

of

unknown

aetiology:

Single-centre

experience

and

comprehensive

meta-analysis

Intérêt

de

l’imagerie

par

résonance

magnétique

cardiaque

pour

exclure

la

présence

d’une

coronaropathie

sous-jacente

chez

les

patients

avec

insufﬁsance

cardiaque

systolique

d’étiologie

inconnue

:

expérience

monocentrique

et

méta-analyse

Aurélie

Manchuelle

a

,

Franc

¸ois

Pontana

b,c,d,e

,

Pascal

De

Groote

a,d,f

,

Paul

Lebert

b

,

Marie

Fertin

a

,

Marine

Baijot

a

,

Christopher

Hurt

a

,

Nicolas

Lamblin

a,c,d,f

,

Nicolas

Debry

a

,

Guillaume

Schurtz

a

,

Anju

Duva

Pentiah

a

,

Arnaud

Sudre

a

_,

_Martine

_Remy-Jardin

b,c

_,

Patrizio

Lancellotti

g,h

,

Eric

Van

Belle

a,c,d,e

,

Christophe

Bauters

a,c,d,f

,

Gilles

Lemesle

a,c,d,e

,

Cédric

Delhaye

a,∗

a_Service_de_cardiologie,_institut_cœur_poumon,_centre_hospitalier_régional_et_{universitaire}_de Lille,59000Lille,France

b_Service_de_radiologie,_institut_cœur_poumon,_centre_hospitalier_régional_et_{universitaire}_de Lille,59000Lille,France

c_Faculté_de_médecine_de_Lille,_université_de_Lille,₅₉₀₀₀_Lille,_France d_Institut_Pasteur_de_Lille,₅₉₀₀₀_Lille,_France

e_Inserm_U1011,₅₉₀₀₀_Lille,_France

Abbreviations:CA,coronaryangiogram/angiography;CAD,coronaryarterydisease;CMR,cardiacmagneticresonanceimaging;HF,heart

failure;LGE,lategadoliniumenhancement;LVEF,leftventricularejectionfraction;NLR,negativelikelihoodratio;PLR,positivelikelihood ratio;st-LGE,subendocardialortransmurallategadoliniumenhancement.

∗_{Corresponding}_author._Institut_cœur_poumon,_centre_hospitalier_régional_et_{universitaire}_de_Lille,_boulevard_{Prof.-Leclercq,}₅₉₀₃₇_Lille

cedex,France.

E-mailaddress:[email protected](C.Delhaye).

https://doi.org/10.1016/j.acvd.2018.04.004

(2)

CMRandHFaetiology 687

f_Inserm_U1167,₅₉₀₀₀_Lille,_France

g_Department_of_Cardiology,_Heart_Valve_Clinic,_University_of_Liège_Hospital,_GIGA

CardiovascularSciences,CHUSartTilman,4000Liège,Belgium

h_Gruppo_Villa_Maria_Care_and_Research,_Anthea_Hospital,₇₀₁₂₄_Bari,_Italy

Received20January2018;receivedinrevisedform6April2018;accepted9April2018 Availableonline1June2018

KEYWORDS Coronaryartery disease; Magneticresonance imaging; Lategadolinium enhancement; Coronaryangiogram; Heartfailure Summary

Background.—Coronaryarterydisease(CAD)istheleadingcauseofsystolicheartfailure(HF). Cardiacmagneticresonanceimaging(CMR)isanon-invasivetechniquethatdetectsamyocardial infarctionscarassubendocardialortransmurallategadoliniumenhancement(st-LGE). Aim.—Wesoughttoevaluatewhetheralackofst-LGEcouldruleoutCADinnew-onsetsystolic HFofunknownaetiology.

Methods.—Weincluded232consecutivepatientswithnew-onsetHFandleftventricular ejec-tionfraction ≤35%whounderwentbothcoronaryangiographyandCMRtoassessHFaetiology. CADwasdeﬁnedasthepresenceofcoronaryarterystenosis_≥50%onacoronaryangiogram. Weassessedsensitivity,speciﬁcity,andpositiveandnegativelikelihoodratios(PLRandNLR)of thepresenceofst-LGEtodetectunderlyingCAD.Acomplementarymeta-analysisof11studies (includingours)wasalsoperformed.

Results.—Inourstudy,49(21.1%)patientshadCAD.Thesensitivityandspecificityofthe pres-ence ofst-LGE to detect CADwere 69 and92%, respectively. PLRand NLR were 8.47and 0.33,respectively.Inthemeta-analysis,1227patientswereincluded,andtheprevalenceof CADrangedfrom19.2to68.3%.Sensitivity,specificity,PLRandNLRwere87%(95%confidence interval[CI]0.80—0.92),93%(95%CI0.89—0.96),12.91(95%CI7.70—21.64)and0.14(95%CI 0.09—0.22),respectively.Altogether,55patientspresentedCADwithnost-LGE;inversely,75 patientspresentedst-LGEwithnoCAD.

Conclusion.—WithaCMRspeciﬁcityof93%,theabsenceofst-LGErulesoutsigniﬁcant under-lyingCADinpatientswithsystolicHFofunknownaetiologyinmostcases.

MOTSCLÉS Coronaropathie; Imageriepar résonance magnétique; Rehaussement tardif; Angiographie coronaire; Insufﬁsance cardiaque Résumé

Contexte.—Lacoronaropathieestlaprincipalecaused’insufﬁsancecardiaque(IC)systolique. L’imagerie parrésonancemagnétiquecardiaque (IRM-C)estune techniquenoninvasivequi permetdedétecterlescicatricesd’infarctusdumyocardesousla formed’unrehaussement tardifsous-endocardiqueoutransmuralaprèsinjectiondegadolinium(st-RT).

Objectif.—L’objectifdecetteétudeestd’évaluersil’absencedest-RTpeutexclurela coro-naropathiechezlespatientsprésentantuneICsystoliqued’étiologieinconnue.

Méthodes.—Autotal,232patientsconsécutifsprésentantuneICavecunefractiond’éjection ventriculairegauche _≤35%etquionteuàlafoisuneangiographiecoronaireetuneIRM-C ontétéinclus.Lacoronaropathieétaitdéfinieparlaprésenced’unesténosecoronaire≥50% enangiographie.Lasensibilité,laspécificité,lesrapportsdevraisemblancepositifsetnégatifs (RVPetRVN)delaprésenced’unst-RTàl’IRM-Cpourdétecterlacoronaropathieontétéévalué. Uneméta-analysecomplémentairede11études(dontlanôtre)aégalementétéréalisée. Résultats.—Dansnotreétude,49(21,1%)patientsavaientunecoronaropathie.Lasensibilité etlaspécificitédelaprésencedest-RTpourdétectercettedernièreétaientde69et92%, respectivement.Les RVPetRVNétaient respectivementde8,47et0,33.Laméta-analysea inclus1227patients.Laprévalencedelacoronaropathievariaitde19,2à68,3%.Lasensibilité, laspécificité,leRVPetleRVNétaientrespectivementde87%(IC95%0,80—0,92),93%(IC95% 0,89—0,96),12,91(IC95 %7,70—21,64)et0,14(IC95%0,09—0,22).Autotal,55 patientsne présentaientpasdest-RTmaisunecoronaropathie;inversement75patientsprésentaientune st-LGEsanscoronaropathie.

Conclusion.—Avecunespéciﬁcitédel’IRMcardiaquede93%,l’absencedest-RTexclutune coronaropathie chez les patients présentantune IC systoliqued’étiologie inconnue dans la plupartdescas.

(3)

688 A.Manchuelleetal.

Background

Coronaryarterydisease(CAD)iscurrentlytheleadingcause ofsystolicheart failure(HF),andimpactspatient progno-sisand care in practice [1]. In this context, the need to systematicallyperform a coronary angiogram (CA), which is an invasive procedure, in patients with systolic HF of unknownaetiologyhasbeenchallengedbynewtechniques. Cardiac magnetic resonance imaging (CMR) has emerged asa non-invasive technique that provides high-resolution imagesoftheheartinanydesiredplaneandwithout radi-ation; it assesses cardiac morphology and function, and the presence and extent of myocardial infarction scar, by showing subendocardial or transmural late gadolinium enhancement(st-LGE)[2].Recentsmall-scaleinvestigations havesuggestedthatlategadoliniumenhancement(LGE)can distinguish whether HF is related to CAD [3—14]. There-fore,the inclusion of CMR in the clinical management of patientswithsystolicHFofunknownaetiologymightavoid unnecessaryinvasivediagnosticcoronaryangiography(CA) procedures.

Inthepresentstudy,wesoughttoassessthediagnostic accuracyofst-LGEonCMR toruleoutsigniﬁcant underly-ingCADinpatientswithnew-onsetsystolicHFofunknown aetiology,andtoincludeourdatainacomprehensive meta-analysisofitsdiagnosticaccuracy comparedwithinvasive CAasareferencestandard.

Methods

Patient

population

Anongoingregistryofcatheter-basedcoronaryprocedures is maintained at our institution. From February 2005 to December 2016, all consecutive patients who underwent CAfornew-onsetsystolicHF(leftventricularejection frac-tion[LVEF]≤35%ontransthoracicechocardiography)were screened.Thosewhohadnoclinicalorelectricalevidence suggestiveofCAD,andforwhomCMRwasavailable,were included inthe present analysis.Duringthe study period, patients presenting withnew-onset HF and reduced LVEF wereroutinely referredforbothCAandCMR.Lackof evi-dence suggestive of CAD was considered in cases of no clinical evidence (no history of documented CAD [history ofunstableangina and/orhistory ofmyocardialinfarction and/orpreviouspercutaneouscoronaryinterventionand/or previouscoronarybypassand/orpreviousCAshowing>50% diameterluminalstenosisinanyepicardialcoronaryartery] orpresenceoftypicalchestpain),andnoelectrocardiogram evidence(pathologicalQwaves).Regionalwallmotion alter-ationontransthoracicechocardiographywasnotconsidered tobediscriminatingforCAD,becausethisvariableisknown tobenon-speciﬁc[15].Theﬁnalstudypopulationconsisted of232patientswithnewsystolicHF.

CA

Conventional techniques were used to perform the CA. Patients were considered to have CAD when≥50% diam-eter luminal stenosis in any epicardial coronary artery (with a diameter>2mm) was observed. Two experienced

interventional cardiologists (A.M. and C.D.) reviewed the CA inblinded conditions.Incases of discrepancy, consen-sus was reached by discussion with a third experienced interventionalcardiologist(G.L.).

CMR

All patients were examined using a 1.5 Tesla MR sys-tem(AchievaTM_;_Philips_Healthcare,_Best,_The_{Netherlands).}

DetailedinformationabouttheCMRisavailableinAppendix B. LGE imaging was performed 10minutes after an intravenous bolus of 0.4mL/kg (0.2mmol/kg) of gadoter-atemeglumine (Dotarem®; Guerbet, Roissy, France), with optimization of inversion times using inversion recovery pulses to null the signal from the normal myocardium. Images wereevaluatedindependently bytwoexperienced radiologists(F.P.andP.L.),whowereblindedtotheclinical variablesandtheCAresults.Incasesofdiscrepancy, consen-suswasreachedbydiscussionwithathirdradiologist(M.R.). End-diastolicandend-systolicvolumeswerecalculatedfrom manualcontouring.Theleftmyocardiumwasanalysedin17 segmentsaccordingtostandardizedreportedsegmentation. Tobevalidated,LGEmustbepresentinthesamemyocardial segmentinatleasttwodifferentplanes.Whenpresent, st-LGEwasthendescribedusingathree-pointscale,according tothemyocardialthicknessinvolved(1=<25%;2=25—75%; 3=>75%).Onlyst-LGEwasconsideredforthepresent anal-ysis to assess the accuracy of CMR to detect underlying CAD in patients with HF of unknown aetiology. However, subepicardialorintramyocardialenhancementwithoutany coronaryarterydistributionwasalsoreportedandincluded withpatientswithnoLGEinthegroupofpatientswithno st-LGE,accordingtothepreviousliterature[16].

Meta-analysis

The PubMed database was searched for eligible stud-ies, with no time restriction, on 5 April 2017, by using the combined medical subject headings for ‘‘coronary artery disease, left ventricular dysfunction, heart fail-ure,ischaemic,lategadoliniumenhancementandmagnetic resonance imaging’’. The complete search used for PubMed was: (coronary artery disease[Title/Abstract] OR left ventricular dysfunction[Title/Abstract] OR heart fail-ure[Title/Abstract]ORischaemic[Title/Abstract])ANDlate gadoliniumenhancement[Title/Abstract]ORmagnetic res-onance imaging[Title/Abstract]) AND english[Language]. Two investigators (G.L. and C.D.) independently checked retrieved titles and abstracts for eligibility, and relevant fulltextsweresystematicallyretrievedforfurtherdetailed assessment. Major reviews regarding the place of CMR in HFwerealsohand searched.Cross-referencesand quoted paperswerechecked,andexpertswerecontactedto iden-tify other relevant studies. The retrieved studies were examinedtoexcludeduplicateoroverlappingdata. Unpub-lished data were not considered for the present analysis, because results could not be considered as certain and deﬁnitive.Meeting abstracts werealsoexcluded, because theycouldnotprovideadequatelydetaileddataandtheir resultsmightnothavebeenﬁnal.

(4)

CMRandHFaetiology 689 • iftheyincludedpatientswithsystolicHFofunknown

aeti-ology;

• iftheyreportedst-LGEbythepresenceofsigniﬁcantCAD statusonCA;

• iftheyreportedacleardeﬁnitionofsigniﬁcantCAD; • iftheabsolutenumbersoftruepositives,falsepositives,

truenegativesandfalsenegativeswerereportedorcould bederived.

Studieswereexcluded:

• iftheyfocusedonpatientswithknownspeciﬁc cardiomy-opathy as post-partum cardiomyopathy, hypertrophic cardiomyopathy, amyloid, sarcoidosis or mitochondrial cardiomyopathy,rightventriculardysplasiaorcongenital heartdisease;

• iftheyfocusedonpatientswithCADonly; • iftheyfocusedonpatientswithoutCAD; • iftheyfocusedonpatientswithLVEF>50%;

• ifacleardeﬁnitionofsigniﬁcantCADwasnotreported; • ifst-LGEwasnotreportedbyCADstatus;

• if they were performed with no ﬁnal report and only abstractsavailable;

• iftheywereperformedinanimals.

The information that was systematically searched for and extracted from each study is presented in Appendix B.Threeinvestigators(G.L.,C.D.andA.M.)performedthe data extraction independently; discrepancies were solved by consensus. Two investigators (G.L. and C.D.) indepen-dentlyassessedthequalityofallincludedstudiesusingthe revised quality assessment of diagnostic accuracy studies (QUADAS-2) tool [17]. Any disagreement wasresolved by consensus.

Statistical

analysis

Continuous variables are expressed as means±standard deviations, except for delays, which are expressed as medians[25th—75thpercentiles].Categoricalvariablesare expressedasnumbers (percentages). Student’s ttest was used to compare continuous variables, and the ␹2 _test

or Fisher’s exact test was used to compare categorical variables.Delaysbetweengroupswerecomparedusingthe Mann-WhitneyUtest.

Sensitivity and specificity were calculated using true positive, true negative, false positive and false nega-tive rates. Sensitivity was calculated as the number of patients with CAD and st-LGE divided by the number of patients with CAD; specificity was calculated as the number of patients without CAD and no st-LGE divided by the number of patients without CAD. Also calcu-latedwere thelikelihoodratios,which expresshowmuch the odds of CAD change in the presence of st-LGE (positivelikelihoodratio[PLR]=sensitivity/[1—specificity]) or in the absence of st-LGE (negative likelihood ratio [NLR]=[1—sensitivity]/specificity). Post-test probability oddsexpressed inpercentages with95% confidence inter-vals(CIs)werecalculated,definedbytheprobabilityofCAD afterapositivetestandafteranegativetest,represented inaFagandiagram.

All these measures of diagnostic accuracy were also calculatedforeach individualstudyincludedin the meta-analysis,andwerereportedaspointestimateswith95%CIs.

Pooledresults for the meta-analysiswerecalculated with STATA software,version 14.0 (StataCorp, College Station, TX,USA)usingtheMIDAScommand.Between-study statis-ticalheterogeneitywasassessedbyusingtheCochranQ␹2

testandtheI2_test._The_study_was_performed_according_to

establishedmethods,andincompliancewiththequalityof reportingofmeta-analyses(QUORUM)guidelines[18].

AllstatisticalanalyseswereperformedwithSTATA soft-wareversion14.1 (StataCorp).Statistical signiﬁcance was assumedataPvalue<0.05.

Results

Single-centre

experience

Population

Thebaselinecharacteristicsofthepatientsaresummarized

in Table 1. In the overall population, the mean age was

57.3±12.3 years. Altogether, 79.3% of cases were male, and15.1% had diabetes. The mean LVEF was22.5±7.5%, asassessedbytransthoracicechocardiography.Themedian delaybetweenCAandCMRwas2[1—6]days.

Signiﬁcant obstructive CAD was found in 49 patients (21.1%). Patientswith CAD were older (62.4 vs 56 years;

P=0.001) and more often had diabetes (26.5% vs 12.0%;

P=0.012)andhypertension(49.0%vs32.2%;P=0.03)than thosewithoutCADatinclusion.

Regarding CMR data, st-LGE was found in 49 patients (21.1%)intheoverallpopulation.Patientswithst-LGEwere moreoftenmale(91.8%vs76.0%;P=0.015)thanthose with-outst-LGEatinclusion.

Diagnostic

accuracy

of

st-LGE

Altogether,st-LGEwasfoundin34(truepositives,69.4%)of the49patients withCAD,andin15(falsepositives,8.2%) ofthe183patientswithoutCAD.CMRdidnotreportany st-LGEin 15(falsenegatives,30.6%) ofthe 49patients with

CAD(Table2).Repartitionofthepatients accordingto

st-LGEandCADstatusisdepictedinFig.1.Therefore,inour population,thesensitivityandspeciﬁcityofthepresenceof st-LGEonCMRtodetectCADwere69%and92%,respectively. PLRandNLRwere8.47and0.33,respectively.The descrip-tionoftheunderlyingCADinthe15patientswithsigniﬁcant obstructiveCADbutnost-LGEonCMRisavailableinTable3. Ofnote,11patientshadsingle-vesseldisease(includingonly onewithsevere [>70%]stenosis), threepatients had two-vesseldiseaseandonepatienthadthree-vesseldisease.

Meta-analysis

Search

results

and

study

selection

Wefound3232citationsinPubMedandotherdatasources. Therewere113relevantstudies,forwhichadetailed assess-mentofthefulltextwasperformed.Weﬁnallyincluded11 studies(includingourown)[3—12]andexcluded102others. The reasonsfor exclusionwere study focused onpatients withCAD only (n=23); study focused onpatients with no CAD(n=53);studydidnotfocusonpatientswithsystolicHF (n=9);studydidnotreportst-LGEbyCADstatus(n=17);and studyhadduplicatedata(n=9).Astudycouldbeexcluded

(5)

690 A.Manchuelleetal. Table1 Baselinecharacteristicsofourpopulation.

Total population

WithCAD Without CAD P With st-LGE Without st-LGE P (n=232) (n=49) (n=183) (n=49) (n=183) Demographics Age(years) 57.3±12.3 62.4±11.9 56±12 0.001 57.3±10.9 57.3±12.6 0.987 Malesex 184(79.3) 43(87.8) 141(77.0) 0.100 45(91.8) 139(76.0) 0.015 Riskfactors Bodymassindex (kg/m2₎ 26.6±5.8 26±4.5 26.8±6.1 0.399 25.9±4.1 26.9±6.2 0.291 Diabetesmellitus 35(15.1) 13(26.5) 22(12.0) 0.012 10(20.4) 25(13.7) 0.248 Hypertension 83(35.8) 24(49.0) 59(32.2) 0.030 19(38.8) 64(35.0) 0.622 Activesmoker 149(64.2) 35(71.4) 114(62.3) 0.236 35(71.4) 114(62.3) 0.236 Hypercholestero-laemia 60(25.9) 15(30.6) 45(24.6) 0.393 11(22.4) 49(26.8) 0.539 Familialhistoryof CAD 40(17.2) 8(16.3) 32(17.5) 0.849 8(16.3) 32(17.5) 0.849 Peripheralarterial diseaseorcarotid disease 20(8.6) 6(12.2) 14(7.7) 0.309 7(14.3) 13(7.1) 0.112 Initialclinical presentation Acutepulmonary oedema 42(18.1) 12(24.5) 30(16.4) 0.191 13(26.5) 29(15.8) 0.085 Othercongestive symptomsa 190(81.9) 37(75.5) 153(83.6) — 36(73.5) 154(84.2) — Electrocardiogram ﬁndings Atrialﬁbrillation 42(18.1) 9(18.4) 33(18.0) 0.970 5(10.2) 37(20.2) 0.103 Leftbundlebranch

block

49(21.1) 14(28.6) 35(19.1) 0.156 12(24.5) 37(20.2) 0.527 LVEFby

echocardiography(%)

22.5±7.5 23.3±7.6 22.3±7.5 0.407 21.7±7 22.7±7.7 0.382 DelaybetweenCAand

CMR(days) 2[1—6] 2[1—4] 3[1—6] 0.036 1[1—5] 3[1—6] 0.099 CAdata ObstructiveCAD 49(21.1) 49(100) — — 34(69.4) 15(8.2) <0.0001 One-vesselCAD 28(12.1) 28(57.1) — — 16(32.7) 12(6.6) <0.0001 MultivesselCAD 21(9.0) 21(42.9) — — 18(36.7) 3(1.6) — Delaybetween

symptomonsetand CA(days) 12[9—21] 12[9—19] 12[9—21] 0.911 12[9—21] 12[9—21] 0.815 CMRdata st-LGE 49(21.1) 34(69.4) 15(8.2) <0.0001 49(100) — — Subepicardialor nodular(atypical) LGE 14(6.0) 1(2.0) 13(7.1) 0.184 — 14(7.7) — Delaybetween symptomonsetand CMR(days)

11[7—15] 6[11—14] 7[11—15] 0.305 6[11—15] 7[11—15] 0.511

Dataareexpressedasmean±standarddeviation,number(%)ormedian[25th—75thpercentiles].CA:coronaryangiogram;CAD:coronary arterydisease;CMR:cardiacmagneticresonanceimaging;LGE:lategadoliniumenhancement;LVEF:leftventricleejectionfraction; NYHA:NewYorkHeartAssociation;st-LGE:subendocardialortransmurallategadoliniumenhancement.

(6)

CMRandHFaetiology 691 Table2 Numbersoftruepositive,falsepositive,truenegativeandfalsenegativepatientsinthe11studiesincludedin themeta-analysis.

Study All

patients

TPs(patientswith CADandst-LGE)

FNs(patientswith CADandnost-LGE)

FPs(patientswith noCADandst-LGE)

TNs(patientswithno CADandnost-LGE) McCrohonetal. (2003) 90 27 0 8 55 Casoloetal. (2006) 60 40 1 3 16 Sorianoetal. (2007) 123 39 11 8 65 Schietingeretal. (2007) 26 2 3 0 21 LePolainde Warouxetal. (2008) 71 24 7 5 35 Valle-Munoz etal.(2009) 100 18 3 6 73 Krittayaphong etal.(2011) 98 53 5 0 40 Gaoetal.(2012) 124 59 0 8 57 DiBellaetal. (2016) 187 77 9 6 95 Thompsonetal. (2017) 116 73 1 16 26 Manchuelleetal. (2018) 232 34 15 15 168

Dataareexpressedasnumber.CAD:coronaryarterydisease;FN:falsenegative;FP:falsepositive;st-LGE:subendocardialortransmural lategadoliniumenhancement(atleastonesegment);TN:truenegative;TP:truepositive.

Figure1. Studypopulationaccordingtothepresenceof suben-docardialortransmurallategadoliniumenhancement(st-LGE)and coronaryarterydisease(CAD).HF:heartfailure;LVEF:left ventric-ularejectionfraction.

for severalreasons. The ﬂow chartof thestudy selection processisshowninFig.2.

Study

and

patient

characteristics

Altogether,11studiesand1227patientswereincludedinthe meta-analysis.Detailedinformationaboutthe11included studies is presented in Table 3 andTable 4: eightstudies were prospective registries and three were retrospective studies.Theyearofpublicationrangedfrom2003to2017. Themedianageofincludedpatientsrangedfrom49to66.4 years.Therateofmalesexrangedfrom65to80.6%,andthe

Figure2. Flowchartofthestudyselectionprocess.CAD: coro-naryarterydisease; HF:heartfailure;st-LGE:subendocardialor transmurallategadoliniumenhancement.*:asinglestudycouldbe excludedforseveralreasons.

rateofdiabetesmellitusrangedfrom11.5to42.9%amongin theeightstudiesthatreportedthisinformation.Themedian LVEFrangedfrom25to40%.ThemediantimebetweenCAD assessmentbyCAandst-LGEassessmentbyCMRrangedfrom 2daysto3.4yearsintheeightstudiesthatreportedthese data.ThedeﬁnitionofobstructiveCADdifferedacrossthe studies.TheprevalenceofCADrangedfrom19.2to68.3%.

(7)

692

A.

Manchuelle

et

al.

Table3 Mainpatientcharacteristicsinthe11studiesincludedinthemeta-analysis.

Study Patients (n) Age (years) Men Diabetes mellitus LVEF Prevalenceof obstructiveCAD Prevalence ofst-LGE Prevalence ofatypical LGE

DescriptionofunderlyingCADin patientswithCADandnost-LGE McCrohonetal. (2003) 90 54inNIDCM vs67in IDCM 41(65.0) NA 39%in NIDCMvs 33%in IDCM assessed byCMR

27(30.0) 35(38.9) 18(20.0) NopatientwithCADandnost-LGE

Casoloetal. (2006)

60 66.4 44(73.3) 12(20.0) 35%

assessed byTTE

41(68.3) 56(93.0) 0(0) Onepatient:severelesionoftheLM

Sorianoetal. (2007)

123 60 99(80.5) 32(26.0) 28%

assessed byTTE

50(40.7) 47(38.2) NA 11patients:ninewithsinglevessel

diseaseandtwowithtwo-vessel disease,notinvolvingtheLMor proximalLADinallcases Schietingeretal.

(2007)

26 49 19(73.1) 3(11.5) 25%

assessed byCMR

5(19.2) 2(7.7) 2(7.7) Threepatients:allwithsigniﬁcant

stenosisoftheLADandatleastone othermajorcoronaryartery LePolainde Warouxetal. (2008) 71 Mean 56—63 50(70.0) NA Mean 25—29% assessed byCMR

31(43.6) 28(39.4) 11(15.5) Sevenpatients:onewithsingle proximalLADstenosis;onewith proximalLADanddistalLCxstenosis; fourwiththree-vesseldisease;one withdominantRCAstenosis Valle-Munoz

etal.(2009)

100 60.4 68(68.0) 33(33.0) 29.1%

assessed byCMR

21(21.0) 24(24.0) 7(7.0) Threepatients:onewith80%stenosis ofthemidLADand70%intheﬁrst marginalbranch;onewith80%inthe proximalnon-dominantRCAand70% stenosisofthedistalsegmentofthe LCx;onewith70%stenosisofthemid LAD

(8)

CMR and HF aetiology 693 Table3(Continued) Study Patients (n) Age (years) Men Diabetes mellitus LVEF Prevalenceof obstructiveCAD Prevalence ofst-LGE Prevalence ofatypical LGE

DescriptionofunderlyingCADin patientswithCADandnost-LGE Krittayaphong etal.(2011) 98 59.4in NIDCMvs 61.5in IDCM 66(67.3) 42(42.9) 32.8%in IDCMvs 27.1%in NIDCM assessed byCMR

58(59.2) 53(54.1) 10(10.2) Fivepatients:noinformationonCAD pattern

Gaoetal.(2012) 124 61 100(80.6) 30(24.2) 26%

assessed byCMR

59(47.6) 67(54.0) 38(30.6) NopatientwithCADandnost-LGE

DiBellaetal. (2016)

187 61 135(72.2) 37(19.8) 37%

assessed byCMR

86(46.0) 83(44.4) 48(25.7) Ninepatients:noinformationonCAD pattern Thompsonatal. (2017) 116 64 90(77.6) NA 40% assessed byCMR

73(62.9) 89(76.7) 15(12.9) Onepatient:totalocclusionofan epicardialvessel,notinvolvingtheLM orLAD Manchuelleetal. (2018) 232 57 184(79.3) 35(15.1) 22.5% assessed byTTE

49(21.1) 49(21.1) 14(6.0) 15patients:11withsinglevessel disease(including50—70%stenosisof thedistalRCA[n=1],midLAD[n=5] andLCx[n=3];80—90%stenosisofthe midRCA[n=2]);threewithtwo-vessel disease(including60%stenosisofthe midLADand50—70%stenosisofthe LCx[n=2];80%stenosisofthemid LADandCTOofthemidRCA[n=1]); onewiththree-vesseldisease(LAD, 90%;LCx,80%;andCTOoftheRCA) Dataareexpressedasnumber(%)unlessotherwiseindicated.CAD:coronaryarterydisease;CMR:cardiacmagneticresonanceimaging;CTO:chronictotalocclusion;IDCM:ischaemic cardiomyopathyLAD:leftanteriordescendingartery;LCx:leftcircumﬂexartery;LGE:lategadoliniumenhancement;LM:leftmain coronaryartery;LVEF: leftventricularejection fraction;NA: not available;NIDCM: non-ischaemicdilated cardiomyopathy; RCA:right coronary artery; st-LGE: subendocardialor transmural late gadoliniumenhancement; TTE: transthoracicechocardiography.

(9)

694

A.

Manchuelle

et

al.

Table4 Descriptionofthe11studiesincludedinthemeta-analysis.

Study Design Periodof

inclusion Consecutive patients Main inclusion criteria Mainexclusion criteria Techniqueto assessLGE Techniqueto assessunderlying CAD Definitionof obstructive CAD Intervalbetween CADevaluation andCMR McCrohonetal. (2003) Prospective registry NA Yes Clinical presentationof HF,documented LVdilationand dysfunctionby TTEor radionucleotide imaging Suspected infiltrativeheart disease, hypertrophic cardiomyopathy, previous revascu-larization, significantvalve disease,history ofmyocarditis Assessed blindly CAinall patients, assessedblind byasingle cardiologist >50% stenosisof any epicardial vessel 2.2yearsin NIDCMvs 3.4yearsin IDCM Casoloetal. (2006) Retrospective study NA Yes Clinical symptoms, LVEF<40%byTTE andLVdilatation byTTEwith volume>2SDsof thesex/body surfacearea Severe congestive symptoms, previous revascu-larization, significant valvulardisease, hypertrophic cardiomyopathy, infiltrative cardiomyopathy, myocarditis Two operators, blindedto clinical data CAinall patients ≥75% stenosisin atleastone major epicardial vessel Maximum 30days Sorianoetal. (2007) Prospective registry

NA Yes Clinicalevidence

ofHF,LV dysfunction documentedby TTE<50% ACSinprevious3 months, secondarycauses ofHF,including valvedisease, constrictive pericarditis, hypertrophic cardiomyopathy, restrictive cardiomyopathy, myocarditis Two physicians blindedto CAdata CAinall patients, assessedblind byasingle expert ≥50% stenosisof any epicardial vessel 5months

(10)

CMR and HF aetiology 695 Table4(Continued)

inclusion Consecutive patients Main inclusion criteria Mainexclusion criteria Techniqueto assessLGE Techniqueto assessunderlying CAD Deﬁnitionof obstructive CAD Intervalbetween CADevaluation andCMR Schietingeretal. (2007) Prospective registry NA NA New-onsetHF andLVsystolic dysfunction, symptom duration<3 months Age<18years, knownCAD, previouspositive stresstest, previousACS, previous revascu-larization, valvularheart disease Onereader blindedto allclinical andCA data CAinall patients >50% stenosisof any epicardial vessel 14days LePolainde Warouxetal. (2008) Prospective registry NA Yes LVEF<50%, referredforHF aetiology Previously established diagnosisofLVSD, haemodynamic instability,atrial ﬁbrillation Onereader blindedto allclinical andCA data CAinall patients >50% stenosisof any epicardial vesselwith diame-ter>1.5mm Maximum1 month Valle-Munoz etal.(2009) Prospective registry

NA Yes Acutenew-onset

HFwithLVSD (LVEF<40%and increased LVEDD>95th percentileon TTE),withno previoushistory ofCAD,noQ wavesonECG andnoclinical dataattimeof diagnosisto suggestCAD Clinicaldata suggesting hypertrophic cardiomyopathy, inﬁltrativeheart diseaseor myocarditis Two inde-pendent observers; discrepan-cies resolvedby consensus CAperformed byone cardiologist blindedto CMRresults ≥70% stenosisof any epicardial vessel NA

(11)

696 A. Manchuelle et al. Table4(Continued)

inclusion Consecutive patients Main inclusion criteria Mainexclusion criteria Techniqueto assessLGE Techniqueto assessunderlying CAD Deﬁnitionof obstructive CAD Intervalbetween CADevaluation andCMR Krittayaphong etal.(2011) Prospective registry NA NA Age>30years, historyofHF within6months, LVEF<50%on TTE,CMRor ventriculogram Clinically unstable,history of revasculariza-tion,HFfrom valvularor pericardial causes Experienced cardiologist unawareof CAresults CAinall patients ≥50% stenosisof any epicardial vessel Maximum1 year

Gaoetal.(2012) Prospective registry

NA Yes LVEF ≤35%on

TTE,stablefor3 monthsunder maximal tolerated therapies, referredforICD implantation - Assessed blindbya single expert CAor computed tomography angiography ≥70% stenosisof any epicardial vessel NA DiBellaetal. (2016) Prospective registry June2007 toJanuary 2013 Yes Newly-diagnosed LVdysfunction (LVEF<45%by ambulatoryTTE evaluation)and NYHA≤2 Chestpain, historyofCAD, historyof secondarycauses ofHF(primary valvedisease, constrictive pericarditis, hypertrophic cardiomyopathy, restrictive cardiomyopathy, myocarditis), cardiac hospitalization, atrialﬁbrillation Consensus ofthree CMRexpert cardiolo-gists CAinall patients ≥50% steno-sisintheLM and/or≥70% inanyother epicardial vessel NA

(12)

CMR and HF aetiology 697 Table4(Continued)

inclusion Consecutive patients Main inclusion criteria Mainexclusion criteria Techniqueto assessLGE Techniqueto assessunderlying CAD Deﬁnitionof obstructive CAD Intervalbetween CADevaluation andCMR Thompsonatal. (2017) Retrospective study 2006to April2013 NA LVEF<50%or LVEDV≥97mL/m2 onCMRor previousLV dysfunctionon TTE Historyof revascularization Assessed indepen-dentlyof CA Assessed independently ofCMR ≥50% steno-sisintheLM and/or≥70% inanyother epicardial vessel 42days Manchuelleetal. (2018) Retrospective study Feb2005to Nov2016 Yes New-onsetHF withLVEF ≤35% measuredbyTTE ANDnoclinical, electricalor biological evidenceofCAD HistoryofCADor presenceof typicalchest pain, pathologicalQ wavesonECG, troponin elevation Two readers blindedto clinical variables andCA results; consensus withathird if discrepancy CAinall patients;blind reviewofCA bytwo experienced interventional cardiologists; consensus withathirdif discrepancy ≥50% stenosisin any epicardial coronary arterywith diame-ter≥2mm 2days

ACS:acute coronarysyndrome;CA:coronaryangiogram;CAD:coronaryarterydisease; CMR:cardiacmagnetic resonanceimaging; ECG:electrocardiogram;HF: heartfailure;ICD: implantablecardioverterdeﬁbrillator;IDCM:ischaemiccardiomyopathy;LGE:lategadoliniumenhancement;LM:leftmaincoronaryartery;LV:leftventricular;LVEF:leftventricular ejectionfraction;LVEDD;leftventricularend-diastolicdiameter;LVEDV;leftventricularend-diastolicvolume;LVSD:left ventricularsystolicdysfunction;NA:notavailable;NIDCM: non-ischaemicdilatedcardiomyopathy;NYHA:NewYorkHeartAssociation;SD:standarddeviation;TTE:transthoracicechocardiography.

(13)

698 A.Manchuelleetal.

TableA.1showstheoverallqualityoftheincludedstudies,

accordingtoQUADAS-2recommendations[17].

Diagnostic

accuracy

of

no

st-LGE

AsshowninTable2,55(falsenegative)patientspresented nost-LGEonCMR, butsigniﬁcantCAD;inversely,75 (false positive) patients presented st-LGE on CMR and no CAD. The diagnostic performance of the presence of st-LGE is

summarized in Fig.3, withpooledresults.Sensitivity and speciﬁcity were 87% (95% CI 0.80—0.92) and 93% (95% CI 0.89—0.96), respectively. PLR and NLR were 12.91 (95% CI7.70—21.64) and0.14(95% CI0.09—0.22),respectively. According to the Fagan diagram, and using a pre-test probabilityofCAD(prevalenceofthediseaseinthetested population) of25%, thepost-testprobabilityofCAD when st-LGE was present on CMR was 81%, and the post-test probability of CAD when no st-LGE was found on CMR

Figure3. Diagnosticaccuracy ofthe presence of subendocardialor transmural late gadoliniumenhancement on cardiacmagnetic resonanceimagingtodetectcoronaryarterydisease.CI:conﬁdenceinterval;NLR:negativelikelihoodratio;PLR:positivelikelihoodratio.

(14)

CMRandHFaetiology 699 was4% (Fig.A.1). When reported in original studies, the

description of theunderlying CAD in the 55patients with signiﬁcantCADbutnost-LGEonCMRisgiveninTable3.

Discussion

In patients with new-onset systolic HF, CAD is known to be a leading cause, and may have important therapeutic and prognosticimplications [1,19]. In some cases, angina togetherwithelectrocardiogramsigns(Qwaves) makethe diagnosisofCADverylikely;CAisthusrecommended,with theobjectivesofadaptingmedicaltreatmentand identify-ingpatientssuitablefor coronaryarteryrevascularization, both of which may favour the recovery of myocardium contractile function affected by hibernation or stunning

[20,21]andimprovethepatient’scondition andoutcomes

[22].

To date, guidelines propose performing CA in patients with systolic HF in case of angina or when a non-invasive stress test is positive for ischaemia/viability in asymptomaticpatients[1].However,inpractice,the inter-pretation of a non-invasive stress test can be highly challenging in suchpatients withdepressed left ventricu-lar systolicfunction and a remodelledleftventricle [23]. BecausetheaetiologyofsystolicHFhasbeenshowntobe CADinabout 30%ofangina-freepatientswithnoprevious coronaryevent,severalteamsroutinelycontinuetoperform CA,whichisconsideredtobethegoldstandardforassessing underlyingCADinthisspeciﬁcsubset[24].

CA is,however, an invasivetechniqueand mayleadto complications—althoughthesearerarewhenno percuta-neouscoronary interventionis performed (<1%in current practice)[25,26].Therefore,severalauthorshaveproposed alternativestrategies forCADdetectionin thesepatients, andCMRhasbeenshowntobeaninterestingtechniquein thiscontext.Inaddition,CMRalsoprovidesuseful informa-tion about cardiac morphology and function. The present study is the largest in the field; it included 232 patients withnew-onsetsystolicHFofunknownaetiology,who under-wentbothCAandCMRevaluationatamedianintervalof2 days,whichistheshortesteverreported.Inprevious liter-ature,thisdelayoftenlastedseveralmonthsoryears,and sometimeswasnotevenavailable[8,10].Inourstudy,the presenceofst-LGEonCMRhadasensitivityof69%,a speci-ficityof92%,aPLRof8.47andanNLRof0.33todetectthe presenceofsignificantunderlyingCAD.

Importantly, and as in previous studies, the definition weusedforsignificantCADmayhaveaffectedourresults; significant CAD was defined, as in most previous litera-ture focusingon CAD [27],as thepresence of a coronary arterystenosis≥50%inanyepicardialvesselwitha diame-terof>2mm.Ourobjectiveherewastonotunderestimate the presenceof underlying CAD by choosingtoo stringent a definition. Of note, all CAs were assessed blind by two experiencedinterventionalcardiologists,andconsensuswas reached witha third in case of discrepancies. Among the 15 patients identified as having no st-LGE but significant CAD in our study, 11 had single-vessel disease (including twowithseverenarrowing[>80%])andfourhadmultivessel CAD.The presenceofasinglemildand/or distalcoronary arterystenosisprobablycannotsolelyexplaintheseverity

andextentoftheleftventriculardysfunctioninsomecases

[21].Indeed,assuggestedbySorianoetal.[10],who pre-viouslyreportedacorrelationbetweentheextentofst-LGE andastandardizeddeﬁnitionofischaemiccardiomyopathy (historyofmyocardialinfarctionand/or coronary revascu-larization;orstenosis≥75%intheleftmaincoronaryartery and/or in the anteriordescending coronary artery before theﬁrstdiagonalbranch;ortwoormorestenoses≥75%in epicardial coronaryarteries) [19], it could be speculated thatthiswasa fortuitousassociation insomecases. How-ever, beyond the therapeutic and prognostic implications of the presence of a ‘‘real’’ ischaemic cardiomyopathy, we believe that the presence of any CAD may alsocarry relevant individual information for patients with systolic HF (medical treatment, follow-up, prognostic evaluation, etc.).

Todefinitivelyassesswhethertheabsenceofst-LGEcan ruleoutsignificantCAD,wealsoperformedacomprehensive meta-analysisof the previous literature, which reported, overall, veryconsistent results with ourstudy (sensitivity 69vs 87,specificity 92 vs93, PLR8.47 vs 12.9,NLR 0.33 vs0.14).ThedefinitionusedforsignificantobstructiveCAD variedacrossthe11included studies,buttheoneusedin ourstudy wasthe most commonly used(Table 4)[3—12]. Accordingtotheresultsofthismeta-analysis,thepresence ofst-LGEonCMR hadasensitivityof 87%,a specificityof 93%, a PLR of 12.91 and an NLR of 0.14 to detect CAD in patients withsystolic HFof unknown aetiology.Of the 1227includedpatients,55withnost-LGEhad,infact, sig-nificant CAD. PLR (‘‘good at ruling in the disease’’) and NLR (‘‘good at ruling out the disease’’) describethe dis-criminatorypropertiesofpositiveandnegativetestresults, respectively (i.e. the presence or absence of st-LGE). Of note, both likelihood ratios are roughly independent of prevalence rates, and state how many times more likely particulartest resultsareinpatients withdisease thanin thosewithout disease (i.e. CAD). There is consensusthat aPLR>10andanNLR<0.1providehighlyconvincing diag-nosticevidence,whereasaPLR>5andanNLR<0.2provide strongdiagnosticevidence[28].Whenappliedtoourresults, CMRlooksusefulforexcludingsignificantCADinmostcases (NLR 0.14). Interestingly, some authors have additionally suggested that coupling CMR to a stress test or to mag-neticresonancecoronaryangiographymayfurtherimprove itsdiagnosticaccuracyinthiscontext[13].Inthesameway, couplingCMRtoacoronarycomputedtomographyscanmay alsoprovideusefulinformation,andmayoffer,inourview, thebestnon-invasivestrategy[7].

Finally, it should be emphasized that LGE might be related to numerous causes other than CAD, and that LGE discovery on CMR may also carry relevant diagnos-tic(sarcoidosis,inflammatorydisease,etc.)andprognostic informationoutsidethecontextofCAD.Asspecifiedinour Methodssection,wefocusedonst-LGE,whichishighly spe-cificformyocardialinfarctionscar.

Ourstudypopulationbelongstoanobservationalcohort of patients with new-onset HF. Nevertheless, it is the largestseriesincludingpatientswhounderwentboth CMR andCA.Whenfocusingonourmeta-analysis,apublication bias could not be ruled out, as only small-scale studies havefocused onthistopicin thepast. Therefore,studies withlowsensitivity and/or speciﬁcity maynot have been

(15)

700 A.Manchuelleetal. submittedbyinvestigatorsoracceptedbyeditorsforfutility

reasons.However,Fig.A.2isreassuringregardingthisissue. Thepresenceofst-LGEonCMR hadasensitivityof87% andaspecificityof93%todetectthepresenceofCADina selectedpopulationofpatientswithnew-onsetHFwithLVEF ≤35%andnoclinicaland/orelectricalevidenceofCAD.Our results therefore suggest that CMR is a rathergood non-invasivealternative tothesystematic useof CAin sucha population.However,CMRwasunabletodetectsignificant CAD in a few cases, and couplingCMR to magnetic reso-nanceangiographyortoacoronarycomputedtomography scanmayfurtherimprovediagnosticperformance.Ourdata andthecost-effectivenessofthisapproachshould,however, beconfirmedinalargercohortofpatients.

Funding

Thisresearchdidnotreceiveanyspeciﬁcgrantfromfunding agenciesinthepublic,commercialornot-for-proﬁtsectors.

Appendix

A. Supplementary

data

Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.

1016/j.acvd.2018.04.004.

Disclosure

of

interest

Theauthorsdeclarethattheyhavenocompetinginterest.

References

[1]PonikowskiP,VoorsAA,AnkerSD,etal.2016ESCGuidelines for thediagnosis and treatmentofacute and chronic heart failure: The Task Forcefor the diagnosis and treatmentof acute and chronic heart failure of theEuropean Societyof Cardiology(ESC).Developedwiththespecialcontributionof theHeart FailureAssociation (HFA)oftheESC.EurHeart J 2016;37:2129—200.

[2]AljizeeriA,SulaimanA,AlhulaimiN,AlsaileekA,Al-MallahMH. Cardiacmagneticresonanceimaginginheartfailure:wherethe alphabetbegins!HeartFailRev2017;22:385—99.

[3]Casolo G, Minneci S, Manta R, et al. Identiﬁcation of the ischemicetiologyofheartfailurebycardiovascularmagnetic resonance imaging: diagnostic accuracy of late gadolinium enhancement.AmHeartJ2006;151:101—8.

[4]Di Bella G, Pingitore A, Piaggi P, et al. Usefulness of late gadoliniumenhancementMRIcombinedwithstressimagingin predictivesigniﬁcantcoronarystenosisinnew-diagnosedleft ventriculardysfunction.IntJCardiol2016;224:337—42. [5]GaoP,YeeR, GulaL,etal. Predictionofarrhythmicevents

in ischemic and dilated cardiomyopathy patients referred for implantable cardiac deﬁbrillator: evaluation ofmultiple scar quantiﬁcation measures for late gadolinium enhance-ment magnetic resonance imaging. Circ Cardiovasc Imaging 2012;5:448—56.

[6]KrittayaphongR,BoonyasirinantT,SaiviroonpornP, Udompun-turakS.Lategadoliniumenhancementfromcardiacmagnetic resonanceinischemicandnon-ischemiccardiomyopathy.JMed AssocThai2011;94(Suppl1):S33—8.

[7]le Polain de Waroux JB, Pouleur AC, Gofﬁnet C, Pas-quet A, Vanoverschelde JL, Gerber BL. Combined coronary and late-enhanced multidetector-computed tomography for delineation of the etiology of left ventricular dysfunction: comparisonwithcoronaryangiographyandcontrast-enhanced cardiac magnetic resonance imaging. Eur Heart J 2008;29: 2544—51.

[8]McCrohon JA,Moon JC, PrasadSK, et al. Differentiation of heartfailurerelatedtodilatedcardiomyopathyandcoronary arterydiseaseusinggadolinium-enhancedcardiovascular mag-neticresonance.Circulation2003;108:54—9.

[9]SchietingerBJ,VorosS,IsbellDC,MeyerCH,ChristopherJM, KramerCM.Canlategadoliniumenhancementby cardiovascu-larmagneticresonanceidentifycoronaryarterydiseaseasthe etiologyofnewonsetcongestiveheartfailure?IntJCardiovasc Imaging2007;23:595—602.

[10]Soriano CJ, Ridocci F, Estornell J, et al. Late gadolinium-enhanced cardiovascular magnetic resonance identiﬁes patients with standardized deﬁnition of ischemic car-diomyopathy: a single centre experience. Int J Cardiol 2007;116:167—73.

[11]ThompsonAC,CrilleyJG,WilsonDW,HunginAP,FuatA,Murphy JJ.Reliableexclusionofprognosticallysigniﬁcantcoronary dis-easeinleftventriculardysfunctionbycardiacMRI.ClinRadiol 2017;72:159—64.

[12]Valle-Munoz A, Estornell-Erill J, Soriano-Navarro CJ, et al. Late gadolinium enhancement-cardiovascular magnetic res-onance identiﬁes coronary artery disease as the aetiol-ogy of left ventricular dysfunction in acute new-onset congestive heart failure. Eur J Echocardiogr 2009;10: 968—74.

[13]Assomull RG, Shakespeare C, Kalra PR, et al. Role of car-diovascularmagnetic resonance asa gatekeeper toinvasive coronaryangiographyinpatientspresentingwithheartfailure ofunknownetiology.Circulation2011;124:1351—60.

[14]WonE,Donnino R,SrichaiMB,etal.Diagnosticaccuracy of cardiacmagneticresonanceimagingintheevaluationofnewly diagnosedheartfailurewithreducedleftventricularejection fraction.AmJCardiol2015;116:1082—7.

[15]DiazRA,Nihoyannopoulos P,Athanassopoulos G,Oakley CM. Usefulnessof echocardiographytodifferentiatedilated car-diomyopathyfromcoronary-inducedcongestiveheartfailure. AmJCardiol1991;68:1224—7.

[16]SatohH,SanoM,SuwaK,etal.Distributionoflategadolinium enhancement in various types of cardiomyopathies: signiﬁ-canceindifferentialdiagnosis,clinicalfeaturesandprognosis. WorldJCardiol2014;6:585—601.

[17]Whiting PF, Rutjes AW, Westwood ME, et al. QUADAS-2: a revisedtoolforthequalityassessmentofdiagnosticaccuracy studies.AnnInternMed2011;155:529—36.

[18]Moher D, Cook DJ, Eastwood S, Olkin I, Rennie D, Stroup DF.Improvingthequalityofreportsofmeta-analysesof ran-domisedcontrolledtrials:theQUOROMstatement.Qualityof reportingofmeta-analyses.Lancet1999;354:1896—900. [19]FelkerGM,ShawLK,O’ConnorCM.Astandardizeddeﬁnition

ofischemiccardiomyopathyforuseinclinicalresearch.JAm CollCardiol2002;39:210—8.

[20]WijnsW,VatnerSF,CamiciPG.Hibernatingmyocardium.NEngl JMed1998;339:173—81.

[21]Elsasser A, Schlepper M, Klovekorn WP, et al. Hibernating myocardium:anincompleteadaptationtoischemia. Circula-tion1997;96:2920—31.

[22]CarsonP,WertheimerJ,MillerA,etal.TheSTICHtrial(Surgical TreatmentforIschemicHeartFailure):mode-of-deathresults. JACCHeartFail2013;1:400—8.

[23]Sicari R, Nihoyannopoulos P, Evangelista A, et al. Stress EchocardiographyExpertConsensusStatement–Executive

(16)

Sum-CMRandHFaetiology 701 mary: European Association of Echocardiography (EAE) (a

registeredbranchoftheESC).EurHeartJ2009;30:278—89. [24] SilvaF,BorgesT,RibeiroA,etal.Heartfailurewithreduced

ejectionfraction:shouldwesubmitpatientswithoutanginato coronaryangiography?IntJCardiol2015;190:131—2. [25] LimMJ,WhiteCJ.Coronaryangiographyisthegoldstandard

forpatientswithsigniﬁcantleftventriculardysfunction.Prog CardiovascDis2013;55:504—8.

[26] WestR,EllisG,BrooksN,JointAuditCommitteeoftheBritish Cardiac S, Royal College of Physicians of L. Complications of diagnostic cardiac catheterisation: results from a

conﬁ-dential inquiry into cardiac catheter complications. Heart 2006;92:810—4.

[27]Lemesle G, Tricot O, Meurice T, et al. Incident myocar-dialinfarctionand verylatestentthrombosis inoutpatients with stable coronary artery disease. J Am Coll Cardiol 2017;69:2149—56.

[28]JaeschkeR,GuyattGH,SackettDL.Users’guidestothe medi-calliterature.III.Howtouseanarticleaboutadiagnostictest. B.Whataretheresultsandwilltheyhelpmeincaringformy patients?TheEvidence-BasedMedicineWorkingGroup.JAMA 1994;271:703—7.