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 . 1 0 , N O . 5 , 2 0 1 7

ª 2 0 1 7 B Y T H E A M E R I C A N CO L L E G E O F C A R D I O L O G Y F O U N DA T I O N

P U B L I S H E D B Y E L S E V I E R

I S S N 1 9 3 6 - 8 7 8 X / $ 3 6 . 0 0

h t t p : / / d x . d o i . o r g / 1 0 . 1 0 1 6 / j . j c m g . 2 0 1 7 . 0 2 . 0 0 6

Stress Perfusion CMR in Patients WithKnown and Suspected CADPrognostic Value and Optimal Ischemic Threshold forRevascularization

Gabriella Vincenti, MD,a,b Pier Giorgio Masci, MD, PHD,a,b Pierre Monney, MD,a,b Tobias Rutz, MD,a,b

Sarah Hugelshofer, MD,a Mirdita Gaxherri, MD,a Olivier Muller, MD, PHD,a Juan F. Iglesias, MD,a

Eric Eeckhout, MD, PHD,a Valentina Lorenzoni, PHD,c Cyril Pellaton, MD,b,d Christophe Sierro, MD,a,b,e

Juerg Schwitter, MDa,b

ABSTRACT

Fro

Ho

of

Va

Br

dis

Ma

OBJECTIVES This study sought to determine the ischemia threshold and additional prognostic factors that identify

patients for safe deferral from revascularizations in a large cohort of all-comer patients with known or suspected

coronary artery disease (CAD).

BACKGROUND Stress-perfusion cardiac magnetic resonance (CMR) is increasingly used in daily practice for ischemia

detection. However, there is insufficient evidence about the ischemia burden that identifies patients who benefit from

revascularization versus those with a good prognosis who receive drugs only.

METHODS All patients with known or suspected CAD referred to stress-perfusion CMR for myocardial ischemia

assessment were prospectively enrolled. The CMR examination included standard functional adenosine stress first-pass

perfusion (gadobutrol 0.1 mmol/kg Gadovist, Bayer AG, Zurich, Switzerland) and late gadolinium enhancement (LGE)

acquisitions. Presence of ischemia and ischemia burden (number of ischemic segments on a 16-segment model), and of

scar and scar burden (number and transmurality of scar segments in a 17-segment model) were assessed. The primary

endpoint was a composite of cardiac death, nonfatal myocardial infarction (MI), and late coronary revascularization (>90

days post-CMR); the secondary endpoint was a composite of cardiac death and nonfatal MI.

RESULTS During a follow-up of 2.5 � 1.0 years, 86 and 32 of 1,024 patients (1,103 screened patients) experienced the

primary and secondary endpoints, respectively. On Kaplan-Meier curves for the primary and secondary endpoints,

patients without ischemia had excellent outcomes that did not differ from patients with <1.5 ischemic segments. In

multivariate Cox regression analyses of the entire population and of the subgroups, ischemia burden (threshold:

$1.5 ischemic segments) was consistently the strongest predictor of the primary and secondary endpoints with hazard

ratios (HRs) of 7.42 to 8.72 (p < 0.001), whereas age ($67 years), left ventricular ejection fraction (#40%), and scar

burden (LGE score $0.03) contributed significantly, but to a lesser extent, in all models with HRs of 2.01 to 3.48, 1.75 to

1.96, and 1.66 to 1.76, respectively.

CONCLUSIONS In a large all-comer patient cohort with known and suspected CAD, an ischemia burden of $1.5

ischemic segments on stress-perfusion CMR was the strongest predictor of the primary and secondary endpoints.

Patients with zero or 1 ischemic segment can be safely deferred from revascularizations.

(J Am Coll Cardiol Img 2017;10:526–37) © 2017 by the American College of Cardiology Foundation.

m the aDivision of Cardiology University Hospital of Lausanne (CHUV), Lausanne, Switzerland; bCardiac MR Center, University

spital of Lausanne (CHUV), Lausanne, Switzerland; cInstitute of Management, Scuola Superiore Sant’Anna, Pisa, Italy; dDivision

Cardiology, Department of Internal Medicine, Neuchâtel, Switzerland; and the eDivision of Cardiology, Centre Hospitalier du

lais Romand (CHVR), Sion, Switzerland. Dr. Schwitter has received funding for clinical studies from Bayer Healthcare and

acco Healthcare. All other authors have reported that they have no relationships relevant to the contents of this paper to

close. Drs. Vincenti and Masci are joint first authors.

nuscript received September 26, 2016; revised manuscript received February 21, 2017, accepted February 23, 2017.

AB BR E V I A T I O N S

AND ACRONYM S

CAD = coronary artery disease

CMR = cardiac magnetic

resonance

EDVi = end-diastolic volume

index

EF = ejection fraction

ESVi = end-systolic volume

index

HR = hazard ratio

LGE = late gadolinium

enhancement

LV = left ventricle

MI = myocardial infarction

ROC = receiver-operating

curve

WMSI = wall motion score

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 . 1 0 , N O . 5 , 2 0 1 7 Vincenti et al.M A Y 2 0 1 7 : 5 2 6 – 3 7 Stress-Perfusion CMR in Patients With CAD

527

A denosine stress-perfusion cardiac magneticresonance (CMR) is increasingly used indaily practice in patients with known or sus-

pected coronary artery disease (CAD) (1), and itsdiagnostic accuracy has been well established (2–7).According to current guidelines in the United Statesand Europe (8–10), stress-perfusion CMR is recom-mended as a first-line test in patients with a proba-bility of intermediate CAD as a Class I indication(Level of Evidence: B) in Europe (9,10) or a ClassIIb indication (Level of Evidence: B) in the UnitedStates (11) if patients have an uninterpretable elec-trocardiogram at rest. In this context, CMR offersthe advantage to obtain information not only onmyocardial ischemia but also on cardiac functionand the presence and extent of myocardial scarwith no radiation exposure. Nevertheless, there re-mains some skepticism about the capacity of thispharmacological stress test in predicting the clinicaloutcome. In particular, there are insufficient dataabout which ischemic threshold, as detected bystress-perfusion CMR, is able to correctly stratify pa-tients into those at high risk and patients who wouldmost likely benefit from revascularizations versuspatients with an uneventful follow-up who can bedeferred safely from revascularization. The annualevent rate for cardiac death and nonfatal myocardialinfarction (MI) was reported consistently as <1% inseveral studies for patients with suspected or knownCAD and a normal stress-perfusion CMR (12–14). Thisrate remained <1% even for patient populations witha high prevalence of CAD (15,16). The excellent prog-nostic power of stress-perfusion CMR was recentlyconfirmed in a large registry that reflected dailyroutine CMR performance, which demonstratedexcellent outcomes in 2,886 patients with suspectedCAD when the stress-perfusion CMR test was normal(17). However, these studies focused on the presenceor absence of ischemia as the outcome predictor,whereas the question which ischemic burden shouldprompt revascularization has not been resolved asyet.

SEE PAGE 538

Therefore, the aim of this study was to correlatethe ischemic burden with the outcome in a largeunselected cohort of consecutive patients withsuspected or known CAD who were referred tostress-perfusion CMR, and, in particular, to identifywhich “ischemic threshold” allowed effective strati-fication of patients. Because ischemia and myocardialscar are known predictors of outcome, we alsoinvestigated the relative contribution of both of theseto outcome prediction.

METHODS

STUDY POPULATION. Between January 2012and May 2015, all-comer consecutive pa-tients with known or suspected CAD referredto adenosine stress-perfusion CMR forassessment of myocardial ischemia wereprospectively enrolled in a pre-defined reg-istry at the CMR Center of the UniversityHospital Lausanne. Informed consent anddetailed medical history were taken at thetime of the CMR examination. Exclusioncriteria were: 1) age younger than 18 years;2) contraindication to CMR; 3) contrain-dication to adenosine (severe asthma orchronic obstructive pulmonary disease, sec-ond- or third-degree atrioventricular block);4) known cardiomyopathy (e.g., hypert-rophic, dilated, or infiltrative) and acute or

chronic myocarditis; 5) known allergy to gadolinium-based contrast medium; and 6) glomerular filtrationrate #30 ml/min/1.73 m2.

The study protocol was approved by the localethics committee and informed written consent wasobtained from all participating patients.

PATIENT FOLLOW-UPANDCLINICALOUTCOME. Patientswere followed up for a minimum of 1 year after theindex stress-perfusion CMR. Clinical outcome wasevaluated by review of hospital and outpatient med-ical records, contacting the patients’ referring physi-cian, or a prescribed telephone interview. Theprimary endpoint was a composite of cardiac death,nonfatal MI, and late coronary revascularization(either percutaneous or surgical occurring >90 daysafter adenosine-stress CMR). The secondary endpointwas a composite of cardiac death and nonfatal MI.Cardiac death was defined as any death preceded byacute MI, acute or exacerbation of heart failure,documented fatal arrhythmias, or unexpected deathwithout a noncardiac cause. Noncardiac deaths werecensored at the time of event. Nonfatal MI wasdefined by symptoms and elevation of serumtroponin to $2-fold the upper normal limit. Patientswho underwent coronary revascularization #90 daysafter the index examination (early coronary revascu-larization) were not censored (18). In patients withmultiple events, only the first event was consideredfor event-free survival analysis.

CMR PROTOCOL. The CMR protocol was performedon 1.5-T (Magnetom, Aera, Erlangen, Germany) or 3-TMR scanners (Skyra and Verio, Siemens Healthcare,Erlangen, Germany) and consisted of: 1) evaluation ofcardiac function by steady-state, free-precession

index

FIGURE 1 Flowchart

1,103 patients referred forstress adenosine-CMR

1,090 patients successfully underwentstress adenosine-CMR

13 patients excluded due to

claustrophobia (n = 11) or premature

CMR exam interruption (n = 2)

1,043 patients

19 patients lost to follow-up

47 patients excluded due to

primary cardiomyopathies (e.g.dilated

or hypertrophic cardiomyopathies)

or myocarditis

1,024 patients with clinical follow-upforming the study cohort

Flowchart of patient recruitment into the trial. CMR ¼ cardiac magnetic resonance.

Vincenti et al. 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 . 1 0 , N O . 5 , 2 0 1 7

Stress-Perfusion CMR in Patients With CAD M A Y 2 0 1 7 : 5 2 6 – 3 7

528

acquisitions of short-axis cine images covering theentire left ventricle (LV); 2) first-pass, adenosine-stress perfusion imaging for the detection ofmyocardial ischemia and ischemic burden quantifi-cation (Online Table 1) (19–21); and 3) standard lategadolinium enhancement (LGE) imaging for myocar-dial scar detection and scar burden quantification(Online Table 1) (22).

CMR IMAGE ANALYSIS. LV volumes and functionwere quantified on the short-axis cine stack applyingthe Simpson’s rule (Argus VF Software, Siemens AG).A left ventricular ejection fraction (LVEF) $50% wasconsidered normal (23).

Stress-perfusion CMR images were evaluatedaccording to the 16-segment model (American HeartAssociation 17-segment model minus the apicalsegment) (24). The analysis of perfusion images wasdone visually by applying criteria similar to thoseused in earlier studies (2–4,21) to identify myocardialischemia (Online Table 1). The total number ofischemic segments was calculated for each patient.

LGE images were analyzed using the AmericanHeart Association 17-segment model (24). Similar togradings used previously (22), the extent of hyper-enhanced tissue within each segment was gradedvisually on a 4-point scale (Online Table 1). AnLGE score was calculated as the sum of the scores

of all segments divided by 17 (yielding a maximumscore of 3).

STATISTICAL ANALYSIS. Continuous variables wereexpressed as mean � SD or median (25th to 75thpercentiles), and categorical variables were expressedas frequency with percentages. All survival analyseswere stratified by early coronary revascularization.Event rates were annualized by dividing theevent rate for the mean duration of follow-up(percentage).

Survival curves were obtained by Kaplan-Meieranalysis and compared by log-rank test. The time toevents was calculated from the adenosine-stress CMRdate. Univariate Cox regression analysis was used toassess the association between baseline covariatesand the primary endpoint. Variables with p < 0.05 atunivariate analysis were included as covariates in themultivariate Cox regression models to test whichvariables were independently associated with thecomposite endpoints. Two multivariate analyseswere performed: 1) in multivariate analysis 1, age, LVvolumes, LVEF, and LV wall motion score index(WMSI) were included as continuous variables inaddition to the presence and/or absence of ischemiaand scar; 2) in multivariate analysis 2, age, LVvolumes, LVEF, LV WMSI, and ischemic and scarburden were introduced as dichotomous variablesaccording to the Youden index after using receiver-operating characteristic (ROC) curves to determinethe best cutoff for the primary endpoint prediction.A stepwise selection procedure was applied (p < 0.05for entry and p > 0.10 for removal) for both multi-variate analyses. Considering the strong correlationbetween LVEF and LV end-systolic volume index(ESVi) (r ¼ �0.852; p < 0.001), between LVend-diastolic volume index (EDVi) and LV ESVi(r ¼ 0.910; p < 0.001), and between LVEF and LVWMSI (r ¼ �0.841; p < 0.001), these covariates wereintroduced separately in the multivariate analyses.On the basis of the result of the multivariate model 2,the additive value of each category of variables wasevaluated on the basis of the increase in thechi-square of the model. The adenosine-stressCMR score was built according to the chi-squarevalue of each added variable. The discriminatory ca-pacity of the score was assessed by the area-under-the ROC curve (c-statistic) and tested in a subset ofpatients (testing subset) obtained by a randomlychosen sample of 60% of the study population. Formultivariate analysis 2, the incremental value forpredicting composite endpoints by stepwise inclusionof CMR parameters and clinical data was assessedby the chi-square test using the Omnibus test of

TABLE 1 Demographic Characteristics of the Patients (N ¼ 1,024)

Age, yrs 63 � 12

Male 720 (70)

Body surface area, m2 1.9 � 0.2

Body mass index, kg/m2 27.8 � 9.1

Cardiovascular risk factors

Family history of CAD 113 (11)

Hypertension 625 (61)

Hypercholesterolemia 608 (59)

Diabetes mellitus 238 (23)

Smoking 312 (31)

Previous myocardial infarction 431 (42)

Previous coronary revascularization 461 (45)

Ten-yr risk for fatal CAD 2.2 (0.7–5.2)*

Indications to adenosine stress CMR

Typical chest pain 200 (20)

Atypical chest pain 192 (19)

Dyspnea 140 (14)

Doubtful exercise stress test 117 (11)

Others 375 (36)

Values are mean � SD, n (%), or % (25th–75th percentile). *Calculated in patientswithout previous myocardial infarction or previous coronary revascularization(n ¼ 563) based on a modified SCORE project (http://www.escardio.org/Education/Practice-Tools/CVD-prevention-toolbox/SCORE-Risk-Charts) that didnot take into account the total cholesterol level.

CAD ¼ coronary artery disease; CMR ¼ cardiac magnetic resonance.

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 . 1 0 , N O . 5 , 2 0 1 7 Vincenti et al.M A Y 2 0 1 7 : 5 2 6 – 3 7 Stress-Perfusion CMR in Patients With CAD

529

model coefficients. All tests were 2-tailed, and p <

0.05 was considered statistically significant. SSPSversion 21 was used for the analyses (IBM, Armonk,New York).

TABLE 2 CMR Characteristics (N ¼ 1,024)

Cardiac rhythm

Sinus rhythm 879 (86)

Sinus rhythm with extrasystoles 98 (10)

Atrial fibrillation/supraventriculararrhythmias

47 (4)

LV EDVi, ml/m2 73 � 26

LV ESVi, ml/m2 35 � 23

LVEF, % 55 � 13

LVEF #50% 279 (27)

LV WMSI 0.06 (0.05–8.06)

LGE 498 (49)

LGE score 0.24 (0.12–0.41)

Adenosine stress CMR

Presence of ischemia 341 (33)

Segments with ischemia 2 (14)

Ischemic score 0.12 (0.06–0.25)

RPP at baseline, mm Hg/beats/min 9,344 (7,781–10,909)

RPP at stress, mm Hg/beats/min 10,366 (8,760–12,136)

Values are n (%), mean � SD, or median (interquartile range).

CMR ¼ cardiac magnetic resonance; EF ¼ ejection fraction; ESVi ¼ end-systolicvolume index; EDVi ¼ end-diastolic volume index; LGE ¼ late gadoliniumenhancement; LV ¼ left ventricular; RPP ¼ rate�pressure product; WMSI ¼ wallmotion score index.

RESULTS

PATIENT CHARACTERISTICS. We screened 1,103patients referred to adenosine stress CMR forischemia evaluation. Reasons for study exclusionsare given in Figure 1, resulting in 1,024 patientswho completed the clinical follow-up and constitutedour study cohort. The main baseline characteristicsare listed in Table 1. The study cohort consisted ofmainly middle-aged male patients. Approximatelyone-quarter of patients were diabetic, and approxi-mately one-half of them had a history of CAD orcoronary revascularization.

ADENOSINE STRESS PERFUSION CMR FINDINGS. Themain adenosine-stress CMR findings are listed inTable 2. One hundred forty-five (14%) patients hadatrial fibrillation or supraventricular arrhythmias atthe time of examination. Less than one-third ofpatients presented with LV systolic dysfunction(i.e., EF #50%), and approximately one-half ofpatients had subendocardial or transmural myocar-dial LGE consistent with a previous MI. One-thirdshowed myocardial ischemia with a median of 2ischemic segments (25th to 75th percentiles: 1 to 4segments). For examples, see Figures 2 and 3.

OVERALL CLINICAL OUTCOME. During a meanfollow-up of 2.5 � 1.0 years (range 1.0 to 4.3 years),the primary endpoint occurred in 86 patients (8.4%).We identified 9 cardiac deaths (5 sudden cardiacdeaths, 4 decompensated heart failures), 23 nonfatalMIs, and 54 late coronary revascularizations (9 bypasssurgery) corresponding to annual event rates of 0.4%,0.9%, and 2.1%, respectively. The annual event ratesfor the primary and the secondary endpoints were3.3% and 0.9%, respectively. Seventy-two patients(7%) had early coronary revascularization (18 bypasssurgeries), and they were more likely men, had ahigher cardiovascular risk profile, and a higheroccurrence of myocardial scar, as well as ischemiathan patients without early revascularization (OnlineTable 2). The ischemic burden was also greater inpatients with early coronary revascularizationcompared with those without early coronary revas-cularization (Online Table 2). Sixty-six percent ofthe patients with ischemia, but without early revas-cularization, showed ischemia in $2 segments, butthey were not referred for planned revascularization,because the clinical indication about coronary angi-ography and coronary revascularization was left tothe clinical judgment of the referring physician.Importantly, none of patients with early coronaryrevascularization experienced cardiac events duringfollow-up.

FIGURE 2 Example of Large Perfusion Deficit

A 65-year-old patient with a large perfusion deficit (14 segments) in the (A to D) territory of the left main artery (E to H) without scar on late gadolinium enhancement

imaging. Coronary angiography shows a severe stenosis of the left main artery (90% to 99%) and the proximal circumflex artery (I, pink arrows); mild nonsignificant

right coronary wall irregularities (<50%) (J, pink dotted arrow).

FIGURE 3 Example of Small Perfusion Deficit

A 60-year-old patient with a small perfusion deficit in the anterolateral wall (2 segments positive in A to D) corresponding to a stenosis of the first marginal branch of

the circumflex coronary artery in (I, arrow) (without scar on late gadolinium enhancement imaging in E to H). (J) Shows a normal right coronary artery.

Vincenti et al. 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 . 1 0 , N O . 5 , 2 0 1 7

Stress-Perfusion CMR in Patients With CAD M A Y 2 0 1 7 : 5 2 6 – 3 7

530

TABLE 3 Univariate Cox Regression Analysis for the Primary Endpoint

HR (95% CI) p Value

Age, yrs 1.038 (1.018–1.058) <0.001

Age $67 yrs 2.884 (1.854–4.487) <0.001

Male 0.888 (0.565–1.397) 0.608

Previous myocardial infarction 1.625 (1.062–2.486) 0.025

Previous coronary revascularization 2.277 (1.460–3.554) <0.001

Family history of CAD 0.934 (0.450–1.939) 0.855

Diabetes 1.574 (0.993–2.495) 0.054

Hypertension 2.112 (1.280–3.487) 0.003

Dyslipidemia 1.281 (0.817–2.010) 0.280

Smoking 1.016 (0.643–1.606) 0.946

LV EDVi, ml/m2 1.009 (1.002–1.016) 0.017

LV EDVi $71 ml/m2 1.786 (1.158–2.755) 0.009

LV ESVi, ml/m2 1.011 (1.004–1.019) 0.003

LV ESVi $46 ml/m2 2.379 (1.508–3.754) <0.001

LVEF, % 0.982 (0.968–0.997) 0.017

LVEF #40% 0.429 (0.267–0.689) <0.001

LV WMSI (1; dimensionless) 1.030 (1.001–1.059) 0.040

LV WMSI $1.56 1.751 (1.121–2.623) 0.013

Presence of ischemia 7.118 (0.435–11.660) <0.001

Number of ischemic segments 1.423 (1.347–1.503) <0.001

Ischemic burden $1.5 segments 8.570 (5.449–13.478) <0.001

Presence of LGE 2.583 (1.629–4.096) <0.001

LGE score (0.1; dimensionless) 4.151 (1.665–10.346) 0.002

LGE score $0.03 2.583 (1.629–4.096) <0.001

CI ¼ confidence interval; HR ¼ hazard ratio; other abbreviations as in Table 2.

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 . 1 0 , N O . 5 , 2 0 1 7 Vincenti et al.M A Y 2 0 1 7 : 5 2 6 – 3 7 Stress-Perfusion CMR in Patients With CAD

531

CLINICAL OUTCOME AND THRESHOLDS FOR

ISCHEMIA AND SCAR BURDEN. The ROC analysisshowed that an ischemic burden of $1.5 segments(corresponding to w9% of LV myocardium) was thebest cutoff for predicting the primary endpoint with asensitivity and specificity of 67% and 81%, respec-tively (area under the curve: 0.77; 95% confidenceinterval: 0.71 to 0.73; p < 0.001). Similarly, a scar scoreof 0.03 was the best cutoff for predicting the primaryendpoint with a sensitivity and specificity of 70% and53%, respectively (area under the curve: 0.61; 95%confidence interval: 0.55 to 0.67; p ¼ 0.001).

UNIVARIATE COX REGRESSION ANALYSIS FOR THE

PRIMARY ENDPOINT. Advanced age, history of coro-nary revascularization or MI, history of hypertension,larger LV volumes, reduced regional and global LVsystolic function, and the presence and severity ofischemia and scar were significantly associated withthe occurrence of the primary endpoint during thefollow-up (Table 3). Similarly, the dichotomizedvariables of age, LV volumes, LVEF, LV WMSI, andischemic and scar burden were unadjusted predictorsof the primary endpoint. An ischemic burden of $1.5segments was associated with an 8-fold increased riskof developing cardiac death, nonfatal MI, or latecoronary revascularization during the follow-up.

FIGURE 4 Event-Free Survival Curves for the Primary and the Secondary Endpoints

1.0

0.8

0.6

0.4

0.2

0.0

0.00 1.00 2.00 3.00 4.00 5.00

Time (years)

*

#

Eve

nt-

free

Su

rviv

al

Log rank: 135P-value < 0.001

P = ns

Primary Endpoint

A

1.0

0.8

0.6

0.4

0.2

0.0

0.00 1.00 2.00 3.00 4.00 5.00

Time (years)

Eve

nt-

free

Su

rviv

al

Log rank: 64P-value < 0.001

Ischemic Burden Absence of ischemia Ischemia < 1.5 segments Ischemia ≥ 1.5 segments

Secondary Endpoint

B

Kaplan-Meier event-free survival curves for the (A) primary and (B) secondary endpoints. *p ¼ 0.121 (blue and green); #p < 0.001 (green and pink).

TABLE 4 Multivariate Cox Regression Analysis for the

Primary Endpoint

HR (95% CI) p Value

Multivariate analysis 1

Model a*

Age, yrs 1.038 (1.018–1.058) <0.001

Presence of LGE 1.892 (1.118–3.014) 0.007

Presence of ischemia 6.855 (4.139–11.354) <0.001

Model b†

Age, yrs 1.035 (1.015–1.055) 0.001

Presence of LGE 1.894 (1.189–3.017) 0.007

Presence of ischemia 6.841 (4.130–11.331) <0.001

Model c‡

Age, yrs 1.035 (1.015–1.056) 0.001

Presence of LGE 1.894 (1.189–3.017) 0.007

Presence of ischemia 6.855 (4.139–11.354) <0.001

Multivariate analysis 2

Model d§

Age $67 yrs 2.321 (1.466–3.676) <0.001

LVEF #40% 1.752 (1.040–2.950) 0.035

Ischemic burden $1.5 segments 8.347 (5.267–13.229) <0.001

LGE score $0.03 1.655 (1.006–2.723) 0.047

Model ekAge $67 yrs 2.418 (1.535–3.808) <0.001

LV ESVi $46 ml/m2 2.195 (1.380–3.491) 0.001

Ischemic burden $1.5 segments 8.722 (5.515–13.794) <0.001

Model f¶

Age $67 yrs 2.423 (1.542–3.808) <0.001

LV EDVi $71 ml/m2 1.631 (1.043–2.549) 0.032

Ischemic burden $1.5 segments 8.376 (5.286–13.273) <0.001

LGE score $0.03 1.764 (1.098–2.836) 0.019

*Model a: Covariates in the model: age, previous coronary revascularization,previous myocardial infarction; history of hypertension, LV EDVi; LVEF; presenceof LGE; presence of ischemia. †Model b: Covariates in the model: age, previouscoronary revascularization, previous myocardial infarction, history of hypertension,LV ESVi; presence of LGE; presence of ischemia. ‡Model c: Covariates in the model:age, previous coronary revascularization, previous myocardial infarction, history ofhypertension, LV EDVi; LV WMSI; presence of LGE; presence of ischemia. §Modeld: Covariates in the model: age $67 years, previous coronary revascularization,previous myocardial infarction, history of hypertension, LV EDVi $71 ml/m2;LVEF #40%; LGE score $0.03; ischemic burden $1.5 segments. kModel e:Covariates in the model: age $67 years, previous coronary revascularization,previous myocardial infarction, history of hypertension, LV ESVi $46 ml/m2; LGEscore $0.03; ischemic burden $1.5 segments. ¶Model f: Covariates in the model:age $67 years, previous coronary revascularization, previous myocardial infarc-tion, history of hypertension, LV EDVi $71 ml/m2; LV WMSI $1.56; LGEscore $0.03; ischemic burden $1.5 segments.

Abbreviations as in Tables 2 and 3.

TABLE 5 Multivariate Cox Regression Analysis (Model d) for the

Primary Endpoint in the Subgroups

HR (95% CI) p Value

LVEF >50% (n ¼ 745)*

Age $67 yrs 3.359 (1.888–5.975) <0.001

Ischemic burden $1.5 segments 8.128 (4.579–14.430) <0.001

LGE score $0.03 1.956 (1.132–3.379) 0.016

LVEF #50% (n ¼ 279)

Ischemic burden $1.5 segments 9.905 (4.571–21.461) <0.001

Absence of LGE (n ¼ 526)†

Age $67 yrs 3.481 (1.524–7.949) <0.001

Ischemic burden $1.5 segments 7.725 (3.419–17.455) <0.001

LVEF #40% 1.956 (1.132–3.379) 0.016

Presence of LGE (n ¼ 498)

Age $67 yrs 2.072 (1.213–3.539) 0.008

Ischemic burden $1.5 segments 7.612 (4.361–13.286) <0.001

No previous coronary revascularization (n ¼ 563)‡

Age $67 yrs 3.160 (1.455–6.864) 0.004

Ischemic burden $1.5 segments 7.415 (3.555–15.464) <0.001

LVEF #40% 2.678 (0.986–7.278) 0.053

Previous coronary revascularization (n ¼ 461)

Age $67 yrs 2.013 (1.151–3.522) 0.014

Ischemic burden $1.5 segments 7.960 (4.438–14.278) <0.001

*Covariates in the model: age $67 years; ischemic burden $1.5 segments; LGEscore $0.03. †Covariates in the model: age $67 years; ischemic burden $1.5segments; LVEF #40%. ‡Covariates in the model: age $67 years; ischemicburden $1.5 segments; LGE score $0.03; LVEF #40%.

Abbreviations as in Tables 2 and 3.

Vincenti et al. 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 . 1 0 , N O . 5 , 2 0 1 7

Stress-Perfusion CMR in Patients With CAD M A Y 2 0 1 7 : 5 2 6 – 3 7

532

SURVIVAL ANALYSES. The Kaplan-Meier curveshowed that patients with an ischemic burdenof$1.5 segments had higher likelihood of experiencingthe primary endpoint during the follow-up comparedwith patients without ischemia or with an ischemicburden of <1.5 segments (Figure 4A). Similarly,Kaplan-Meier analysis showed a higher likelihood ofcardiac death or nonfatal MI (secondary endpoint) inpatients with an ischemic burden of $1.5 segmentscompared with patients without ischemia or with anischemic burden of <1.5 segments (Figure 4B).

MULTIVARIATE COX REGRESSION ANALYSES FOR

CLINICAL OUTCOME. In multivariate analysis 1,older age, the presence of ischemia, and infarct scarwere predictors of the primary endpoint aftercorrection for the other covariates (Table 4; Model a).This result was confirmed when LV ESVi replacedLVEF and LV EDVi (Table 4; Model b) or when LVWMSI replaced LVEF in the multivariate analysis(Table 4; Model c). In multivariate analysis 2, anischemic burden of $1.5 segments was the strongestpredictor (hazard ratio [HR]: 8.374 to 8.722; p < 0.001for all) for the primary endpoint after the correctionfor the other covariates (Table 4; models d to f), and itwas also the strongest independent predictor of theprimary endpoint in the subgroups (reduced and/ornormal LV systolic function, presence and/or absenceof scar, or a history of coronary revascularization(Table 5). Finally, the ischemic burden of $1.5 seg-ments was the strongest predictor of the secondaryendpoint (HR: 9.688; 95% confidence interval: 4.470to 20.990; p < 0.001) after correction for age $67years LVEF #40%. and scar score $0.03.

STRESS-PERFUSION CMR RISK SCORE. On the basisof the global chi-square score (127) of the multivariateanalysis 2 (Model d), a risk score that consisted of 12.5points was created. The points were assigned to theindependent variables according to their additive

TABLE 6 Stepwise Inclusion Procedure for the Multivariate Analysis 2 (Model d)

HR (95% CI) p Value Step Chi-Square Model p Value

Step 1

Ischemic burden $1.5 9.003 (5.692–14.240) <0.001 95.425 95.425 <0.001

Step 2

Ischemic burden $1.5 8.644 (5.468–13.665) <0.001

Age $67 yrs 2.701 (1.722–4.235) <0.001 19.511 114.936 <0.001

Step 3

Ischemic burden $1.5 8.733 (5.521–13.814) <0.001

Age $67 yrs 2.557 (1.488–3.735) <0.001

LVEF #40% 2.155 (1.320–3.515) 0.002 8.485 123.421 0.004

Step 4

Ischemic burden $1.5 8.347 (5.267–13.229) <0.001

Age $67 yrs 2.321 (1.466–3.676) <0.001

LVEF #40% 1.752 (1.040–2.950) 0.035

LGE score $0.03 1.655 (1.006–2.723) 0.047 4.042 127.463 0.044

Covariates in the model: age $67 years; previous coronary revascularization; previous myocardial infarction;history of hypertension; LV EDVi $71 ml/m2; LVEF #40%; ischemic burden $1.5 segments; LGE score $0.03.

Abbreviations as in Tables 2 and 3.

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 . 1 0 , N O . 5 , 2 0 1 7 Vincenti et al.M A Y 2 0 1 7 : 5 2 6 – 3 7 Stress-Perfusion CMR in Patients With CAD

533

chi-square value to each step of the stepwise inclu-sion procedure (Table 6). Nine points were assignedto an ischemic burden of $1.5 segments, 2 points toage $67 years, 1 point to LVEF #40%, and 0.5 pointsfor a scar score $0.03, yielding a median of 2.0 (range0.0 to 12.5). The score showed a good predictivecapacity for the primary (c-statistic ¼ 0.810) andsecondary endpoints (c-statistic ¼ 0.846). The scorepermitted creation of a gradient of the annualizedevent rate across the study population for the primaryand secondary endpoints (Figure 5), and this prog-nostic information persisted throughout the follow-up (Figure 6). This score allowed identification of alarge proportion of patients at low risk (score 0,n ¼ 741 [72%]) (blue line in Figure 6), with anannualized event rates of 1.19% and 0.36% for theprimary and the secondary endpoints, respectively.Conversely, a score $9.0 identified a subset ofhigh-risk patients (n ¼ 183; 18%; black and red lines in

FIGURE 5 CMR Score Performance

12.50

10.00

7.50

5.00

2.50

.00≤ 0 0.5-3.0 3.5-9.0 9.5-12.0 ≥ 12.5

SCORE

Eve

nt

Rat

e fo

rS

eco

nd

ary

En

d-P

oin

t (%

)

Ptrend < 0.001

D

Testing Subset

30.00

20.00

10.00

0.00≤ 0 0.5-3.0 3.5-9.0 9.5-12.0 ≥ 12.5

SCORE

Eve

nt

Rat

e fo

rP

rim

ary

En

d-P

oin

t (%

)

Ptrend < 0.001

C

Testing Subset

12.50

10.00

7.50

5.00

2.50

00.00≤ 0 0.5-3.0 3.5-9.0 9.5-12.0 ≥ 12.5

SCORE

Eve

nt

Rat

e fo

rS

eco

nd

ary

En

d-P

oin

t (%

)

Ptrend < 0.001

Secondary EndpointB

Study Population

25.00

20.00

15.00

10.00

5.00

0.00≤ 0 0.5-3.0 3.5-9.0 9.5-12.0 ≥ 12.5

SCORE

Eve

nt

Rat

e fo

rP

rim

ary

En

d-P

oin

t (%

)

Ptrend < 0.001

Primary Endpoint

Secondary EndpointPrimary Endpoint

A

Study Population

Predictive performance of the score for the primary and secondary endpoint in the (A and B) entire population and the (C and D) test

population. Abbreviation as in Figure 1.

FIGURE 6 Event-Free Survival Curves for the CMR Score

1.0

0.8

0.6

0.4

0.2

0.0

0.00 1.00 2.00 3.00 4.00 5.00

Log rank: 189P-value < 0.001

Score ≤0 0.5-3.0 3.5-9.0 9.5-12.0 ≥12.5

Primary Endpoint

Time (years)

1.0

0.8

0.6

0.4

0.2

0.0

0.00 1.00 2.00 3.00 4.00 5.00

Log rank: 105P-value < 0.001

Secondary Endpoint

Time (years)

Eve

nt-

free

Su

rviv

al

A

Eve

nt-

free

Su

rviv

al

B

Kaplan-Meier curve analysis showing the risk stratification performance of adenosine stress CMR score over the entire duration of follow-up.

Abbreviation as in Figure 1.

TABLE 7 Category of Risk According to Adenosine Stress

CMR-Based Risk Score

Risk Risk score

Annualized Event Rate (%)

Primary Endpoint Secondary Endpoint

Low risk #3.0 1.26 0.38

Moderate risk $3.5 to <9.0 3.14 1.39

High risk $ 9.0 12.29 4.78

Abbreviation as in Table 2.

Vincenti et al. 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 . 1 0 , N O . 5 , 2 0 1 7

Stress-Perfusion CMR in Patients With CAD M A Y 2 0 1 7 : 5 2 6 – 3 7

534

Figure 6) who presented with annualized event ratesof 13.72% and 5.34% for the primary and the second-ary endpoints, respectively (Table 7).

No major differences were observed between thestudy population and the testing subset (n ¼ 630randomly chosen patients), with the exception of aslightly lower incidence of previous coronary revas-cularization and previous MI in the testing subset(Online Tables 2 and 3). In the testing subset, thescore held its prognostic information, yielding agradient of risk for the primary and the secondaryendpoints (c-statistic ¼ 0.823 and ¼ 0.877, respec-tively), which confirmed the prognostic value of thisscore in the testing subset.

INCREMENTAL VALUE OF THE THRESHOLD FOR

ISCHEMIC BURDEN FOR PREDICTION OF CLINICAL

OUTCOMES. Information on an ischemic burdenof $1.5 segments made a sizeable difference in theprediction of the primary and the secondary end-points as shown by the improvement in the chi-square analyses (Figure 7). Although the addition ofthe CMR functional parameter and a scar score $0.0.3slightly improved the capacity of the model for pre-dicting the clinical outcome, the addition of an

ischemic burden of $1.5 segments to the multivariatemodel strongly improved the prediction of both pri-mary and secondary endpoints.

DISCUSSION

The study results can be summarized as follows. 1)For both the primary and secondary endpoints,myocardial ischemia was the most important predic-tor. Ischemia extended to $2 myocardial segmentswas associated with a worse outcome, whereas anischemic burden below that threshold predicted afavorable outcome on medical treatment that did notdiffer from those patients with normal CMR perfusion

FIGURE 7 Model Performances for Primary and Secondary Endpoint Prediction

Model-Clin:Clinical Data

Model-Fn:Model-Clin +

CMR FunctionalParameters

Model-Scar:Model-Fn +

Scar BurdenModel-Isch:

Model-Scar +IschemicBurden

200.00

150.00

100.00

50.00

0.00

P < 0.001

P = 0.015

P = 0.042

Ch

i-sq

uar

e

APrimary Endpoint

Model-Clin:Clinical Data

Model-Fn:Model-Clin +

CMR FunctionalParameters

Model-Scar:Model-Fn +

Scar BurdenModel-Isch:

Model-Scar +IschemicBurden

120.00

100.00

80.00

60.00

40.00

20.00

P < 0.001

P = 0.014

P = 0.013Ch

i-sq

uar

e

BSecondary Endpoint

Incremental chi-square analyses showing the strong incremental value of the ischemic burden $1.5 segments in predicting (A and B) primary

and secondary endpoints (Model-Isch) compared with other models, including clinical data alone (Model-Clin), clinical data in combination

with CMR functional parameters (Model-Fn) or clinical data in combination with CMR functional parameters and scar score $0.03

(Model-Scar). CMR ¼ cardiac magnetic resonance.

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 . 1 0 , N O . 5 , 2 0 1 7 Vincenti et al.M A Y 2 0 1 7 : 5 2 6 – 3 7 Stress-Perfusion CMR in Patients With CAD

535

studies. 2) The presence of myocardial scar was alsopredictive of outcome but was much less powerfulthan ischemia burden. 3) A risk score obtained bycombining clinical and CMR data allowed determi-nation of individual risk of the patient, which mightbe useful in supporting personalized treatmentdecisions.

THE IMPACT OF ISCHEMIA BURDEN ON OUTCOME:

THE ISCHEMIC THRESHOLD. An ischemic burdenof $1.5 segments was independently associated witha nearly 9-fold increased risk of cardiac death,nonfatal MI, or late coronary revascularization duringa mean follow-up of 2.5 years (see also Kaplan-Meiercurves) (Figure 4). Remarkably, this ischemicthreshold remained the strongest independent pre-dictor when the analysis was limited to hard end-points (i.e., cardiac death and nonfatal MI). Thissecondary endpoint analysis (which did not includerevascularizations) was important to eliminate anybias that the CMR results might have introduced intothe decision-making process on revascularizations.These findings were in line with previous stress-perfusion CMR studies that demonstrated the pres-ence of ischemia as an important predictor of clinicaloutcome (12–16,25) and extended their results by

defining a threshold for ischemic burden that clearlydifferentiated patients with a favorable outcome fromthose with a complicated one. In addition, thisischemic burden of $1.5 segments was also a strongand independent predictor of the primary endpoint insubgroups with normal or reduced LVEF, with orwithout scar, or in patients with or without a historyof coronary revascularization. To recommend revas-cularization in patients with at least 2 ischemic seg-ments (based on the 17-segment model) correspondedto w12% of ischemic myocardium. Interestingly,this CMR-based threshold was close to the reported12.5% for scintigraphic techniques (26). However,regarding ischemia detection by dobutamine stressCMR, guidelines recommend a threshold of up to 3segments in analogy to stress echocardiography (9),whereas some dobutamine stress CMR studiesdemonstrated a worse outcome in patients with only1 ischemic segment, corresponding to w6% ofischemic myocardium (27,28). This finding might beexplained by the ischemic cascade that describes theinducible wall motion abnormalities that occur as aconsequence of hypoperfusion (i.e., that occur laterduring the stress test). Hence, the extent of hypo-perfusion (as tested by adenosine stress-perfusionCMR) might extend beyond the borders of a

PERSPECTIVES

COMPETENCY IN PATIENT CARE AND

PROCEDURAL SKILLS: Although stress-perfusion

CMR is increasingly used in daily practice for ischemia

detection, there is insufficient evidence on the

ischemia burden that identifies patients who benefit

from revascularization versus those with a good

prognosis who only receive drugs. This prospective

study included 1,024 patients with suspected or known

coronary artery disease who underwent stress-

perfusion CMR and were followed for 2.5 � 1.0 years.

In multivariate analyses of the entire population and in

subgroups, ischemia burden was consistently the

strongest predictor of the primary endpoint (cardiac

death, nonfatal MI, late [>90 day] revascularization)

and secondary endpoint (cardiac death, nonfatal MI)

with hazard ratios of 7.4 to 8.7, whereas age, LVEF,

and scar burden contributed to a lesser extent in all

models. These results indicated that patients with no

or minor ischemia (<2 segments) can be safely

managed by drugs.

TRANSLATIONAL OUTLOOK: An individual risk

score for patients who underwent noninvasive CMR

diagnostic testing is proposed for combining clinical

information with CMR functional information, as well

as scar and ischemia information. This risk score per-

formed well in a testing cohort (60% of patients

randomly chosen from the entire study population)

and allowed for a comprehensive individual risk

assessment. Larger studies are needed to prove the

robustness of the proposed score.

Vincenti et al. 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 . 1 0 , N O . 5 , 2 0 1 7

Stress-Perfusion CMR in Patients With CAD M A Y 2 0 1 7 : 5 2 6 – 3 7

536

hypocontractile ischemic segment as detected bydobutamine stress CMR.

The ischemic burden was also the most importantpredictor of outcome in patients with reduced LVEF.Thus, ischemia testing appears important in patientswith reduced LVEF, which is in line with actualguidelines on heart failure management that recom-mend ischemia testing as a Class IIb indication (Levelof Evidence: B) (23).

INTEGRATING CLINICAL, FUNCTIONAL, SCAR, AND

ISCHEMIA INFORMATION TO PREDICT OUTCOME. Inclinical practice, decisions on treatment should bal-ance benefits and risks of a proposed procedure.Therefore, it appears helpful to derive an individualrisk estimate for a patient having undergone nonin-vasive diagnostic testing. The proposed risk scorecombines clinical information with CMR functionalinformation as well as scar and ischemia informationto predict outcome. Although univariate analysesdemonstrated a correlation of most clinical parame-ters with outcome (Table 3), in the multivariate ana-lyses, functional and scar parameters, and inparticular ischemia parameters, were strongly asso-ciated with outcome, so that only age as clinicalparameter persisted in the model. This score appearseasy to calculate and performed well when applied toa subgroup of the study patients (testing subgroup).

STUDY LIMITATIONS. This study was based on aregistry including and following up prospectivelyunselected patients who underwent stress-perfusionCMR. Accordingly, no control group was available,and we could not draw conclusions whether patientswith extensive ischemia burden ($1.5 segments)benefited from revascularizations. This importantaspect will be addressed by the large multicenterISCHEMIA (International Study of ComparativeHealth Effectiveness With Medical and InvasiveApproaches) trial (29). Nevertheless, the excellentoutcomes in the patients with early revasculariza-tions (no cardiac deaths and no nonfatal MI)suggested an advantage for revascularizations inthese patients in the present study. No quantitative(30) or semi-quantitative (20) perfusion analysiswas applied, which may be superior to a visualassessment (30). However, the presented visualanalysis approach represents current clinical practice,whereas absolute perfusion quantification has not yetreached general clinical acceptance. The proposedrisk score performed well in this cohort, althoughfuture testing of multicenter data is warranted toprove its robustness. Finally, this study did notcompare the diagnostic yield of noninvasive tech-niques others than CMR. These aspects will be

investigated in the ongoing MR-INFORM (MR Perfu-sion Imaging to Guide Management of Patients WithStable Coronary Artery Disease) trial (31).

CONCLUSIONS

In a large all-comer patient populationwith known andsuspected CAD, an ischemia burden of $1.5 ischemicsegments on stress-perfusion CMR was by far thestrongest predictor of the primary and secondaryendpoints. Patients with zero or 1 ischemic segmentcan be safely deferred from revascularizations. Age,LVEF, and scar burden contributed to a lesser extent tothe outcome prediction.

ADDRESS FOR CORRESPONDENCE: Dr. JuergSchwitter, University Hospital Lausanne, Division ofCardiology, Rue de Bugnon 46, Lausanne 1011,Switzerland. E-mail: jurg.schwitter@chuv.ch.

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 . 1 0 , N O . 5 , 2 0 1 7 Vincenti et al.M A Y 2 0 1 7 : 5 2 6 – 3 7 Stress-Perfusion CMR in Patients With CAD

537

RE F E RENCE S

1. Bruder O,Wagner A, Lombardi M, et al. EuropeanCardiovascular Magnetic Resonance (EuroCMR)registry–multi national results from 57 centers in 15countries. J Cardiovasc Magn Reson 2013;15:9.

2. Schwitter J, Wacker CM, van Rossum AC, et al.MR-IMPACT: Magnetic Resonance Imaging forMyocardial Perfusion Assessment in CoronaryArtery Disease Trial: comparison of perfusion CMRwith single photon emission computed tomogra-phy for the detection of coronary artery disease ina multicenter, multivendor, randomized trial. EurHeart J 2008;29:480–9.

3. Schwitter J, Wacker CM, Wilke N, et al. Superiordiagnostic performance of perfusion-CMR versusSPECT to detect coronary artery disease: thesecondary endpoints of the multicenter multi-vendor MR-IMPACT II. J Cardiovasc Magn Reson2012;14:61–71.

4. Schwitter J, Wacker CM, Wilke N, et al. MR-IMPACT II: Magnetic Resonance Imaging forMyocardial Perfusion Assessment in CoronaryArtery Disease Trial: perfusion-cardiac magneticresonance vs. single-photon emission computedtomography for the detection of coronary arterydisease: a comparative multicentre, multivendortrial. Eur Heart J 2012;34:775–81.

5. Greenwood J, Maredia N, Younger J, et al.Cardiovascular magnetic resonance and single-photon emission computed tomography for diag-nosis of coronary heart disease (CE-MARC): aprospective trial. Lancet 2012;379:453–60.6. Greenwood JP, Motwani M, Maredia N, et al.Comparison of cardiovascular magnetic resonanceand single-photon emission computed tomogra-phy in women with suspected coronary arterydisease: from the Clinical Evaluation of MagneticResonance Imaging in Coronary Heart Disease(CE-MARC) Trial. Circulation 2014;129:1129–38.

7. Nagel E, Klein C, Paetsch I, et al. Magneticresonance perfusion measurements for thenoninvasive detection of coronary artery disease.Circulation 2003;108:432–7.

8. Smith SC Jr., Feldman TE, Hirshfeld JW Jr.,et al. ACC/AHA/SCAI 2005 Guideline Update forPercutaneous Coronary Intervention—SummaryArticle: a report of the American College ofCardiology/American Heart Association Task Forceon Practice Guidelines. J Am Coll Cardiol 2006;47:216–35.

9. Montalescot G, Sechtem U, Achenbach S, et al.ESC guidelines on the management of stablecoronary artery disease: the Task Force on themanagement of stable coronary artery disease ofthe European Society of Cardiology. Eur Heart J2013;34:2949–3003.

10. Windecker S, Kolh P, Alfonso F, et al. 2014ESC/EACTS Guidelines on myocardial revasculari-zation: the Task Force on Myocardial Revasculari-zation of the ESC and the EACTS developed withthe special contribution of the EAPCI. Eur Heart J2014;35:2541–619.

11. Fihn SD, Gardin JM, Abrams J, et al. 2012ACCF/AHA/ACP/AATS/PCNA/SCAI/STS Guidelinefor the Diagnosis and Management of PatientsWith Stable Ischemic Heart Disease: A Report ofthe ACC Foundation/AHA Task Force on PracticeGuidelines, and the American College of Physi-cians, American Association for Thoracic Surgery,Preventive Cardiovascular Nurses Association,Society for Cardiovascular Angiography andInterventions, and Society of Thoracic Surgeons.J Am Coll Cardiol 2012;60:e44–164.

12. Coelho-Filho OR, Seabra LF, Mongeon F-P,et al. Stress myocardial perfusion imaging by CMRprovides strong prognostic value to cardiac eventsregardless of patient’s sex. J Am Coll Cardiol Img2011;4:850–61.

13. Krittayaphong R, Chaithiraphan V, Maneesai A,Udompanturak S. Prognostic value of combinedmagnetic resonance myocardial perfusion imagingand late gadolinium enhancement. Int JCardiovasc Imaging 2011;27:705–14.

14. Abbasi SA, Heydari B, Shah RV, et al. Riskstratification by regadenoson stress magneticresonance imaging in patients with known orsuspected coronary artery disease. Am J Cardiol2014;114:1198–203.

15. Jahnke C, Nagel E, Gebker R, et al. Prognosticvalue of cardiac magnetic resonance stress tests:adenosine stress perfusion and dobutamine stresswall motion imaging. Circulation 2007;115:1769–76.

16. Buckert D, Dewes P, Walcher T, Rottbauer W,Bernhardt P. Intermediate-term prognostic valueof reversible perfusion deficit diagnosed byadenosine CMR: a prospective follow-up study in aconsecutive patient population. J Am Coll CardiolImg 2013;6:56–63.

17. Moschetti K, Petersen SE, Pilz G, et al. Cost-minimization analysis of three decision strategiesfor cardiac revascularization: results of the“suspected CAD” cohort of the European Cardio-vascular Magnetic Resonance Registry.J Cardiovasc Magn Reson 2016;18:1–10.

18. Boden WE, O’Rourke RA, Teo KK, et al. Optimalmedical therapy with or without PCI for stablecoronary disease. N Engl J Med 2007;356:1503–16.

19. Schwitter J, Nanz D, Kneifel S, et al. Assess-ment of myocardial perfusion in coronary arterydisease by magnetic resonance: a comparison withpositron emission tomography and coronaryangiography. Circulation 2001;103:2230–5.

20. Giang T, Nanz D, Coulden R, et al. Detection ofcoronary artery disease by magnetic resonancemyocardial perfusion imaging with variouscontrast medium doses: first European multicenterexperience. Eur Heart J 2004;25:1657–65.

21. Plein S, Kozerke S, Suerder D, et al. Highspatial resolution myocardial perfusion cardiacmagnetic resonance for the detection of coronaryartery disease. Eur Heart J 2008;29:2148–55.

22. Kim RJ, Albert TSE, Wible JH, et al. Perfor-mance of delayed-enhancement magnetic reso-nance imaging with gadoversetamide contrast forthe detection and assessment of myocardialinfarction. an international, multicenter, double-blinded, randomized trial. Circulation 2008;117:629–37.

23. Ponikowski P, Voors AA, Anker SD, et al. 2016ESC Guidelines for the diagnosis and treatment ofacute and chronic heart failure. Eur Heart J 2016;37:2129–200.

24. Cerqueira MD, Weissman NJ, Dilsizian V,Jacobs AK, Kaul S, Laskey WK. Standardizedmyocardial segmentation and nomenclature fortomographic imaging of the heart. Circulation2002;105:539–42.

25. Pontone G, Andreini D, Bertella E, et al.Prognostic value of dipyridamole stress cardiacmagnetic resonance in patients with known orsuspected coronary artery disease: a mid-termfollow-up study. Eur Radiol 2015;26:1–11.

26. Hachamovitch R, Rozanski A, Shaw LJ, et al.Impact of ischaemia and scar on the therapeuticbenefit derived from myocardial revascularizationvs. medical therapy among patients undergoingstress-rest myocardial perfusion scintigraphy. EurHeart J 2011;32:1012–24.

27. Hundley WG, Morgan TM, Neagle CM,Hamilton CA, Rerkpattanapipat P, Link KM. Mag-netic resonance imaging determination of cardiacprognosis. Circulation 2002;106:2328–33.

28. Giusca S, Kelle S, Nagel E, et al. Ischemicburden and clinical outcome: is one ‘culprit’ischemic segment by dobutamine stress magneticresonance predictive? PLoS ONE 2014;9:e115182.

29. International Study of Comparative HealthEffectiveness With Medical and Invasive Ap-proaches (ISCHEMIA). Available at: https://clinicaltrials.gov/ct2/show/NCT01471522. AccessedJanuary 2, 2017.

30. Mordini FE, Haddad T, Hsu L-Y, et al. Diag-nostic accuracy of stress perfusion CMR incomparison with quantitative coronary angiog-raphy: fully quantitative, semiquantitative, andqualitative assessment. J Am Coll Cardiol Img2014;7:14–22.

31. Hussain ST, Paul M, Plein S, et al. Design andrationale of the MR-INFORM study: stress perfu-sion cardiovascular magnetic resonance imaging toguide the management of patients with stablecoronary artery disease. J Cardiovasc Magn Reson2012;14:65.

KEY WORDS cardiac magnetic resonance,coronary artery disease, ischemia burden,outcome, prognosis, scar burden

APPENDIX For supplemental tables, pleasesee the online version of this paper.