Diagnostic performance of coronary magnetic resonance angiography as compared against conventional x-ray angiography: A meta-analysis

DMCAPA

Ienclicpttfd

AD§EIRRM

2

Journal of the American College of Cardiology Vol. 44, No. 9, 2004© 2004 by the American College of Cardiology Foundation ISSN 0735-1097/04/$30.00P

Cardiac Imaging

iagnostic Performance of Coronaryagnetic Resonance Angiography asompared Against Conventional X-Ray AngiographyMeta-Analysis

eter G. Danias, MD, PHD, FACC,*†‡ Arkadios Roussakis, MD,† John P. A. Ioannidis, MD§�thens and Ioannina, Greece; and Boston, Massachusetts

OBJECTIVES This study was designed to define the current role of coronary magnetic resonanceangiography (CMRA) for the diagnosis of coronary artery disease (CAD).

BACKGROUND Coronary magnetic resonance angiography has been proposed as a promising noninvasivemethod for diagnosis of CAD, but individual studies evaluating its clinical value have been oflimited sample size.

METHODS We identified all studies (MEDLINE and EMBASE) that evaluated CAD by both CMRAand conventional angiography in �10 subjects during the period 1991 to January 2004. Werecorded true and false positive and true and false negative CMRA assessments for detectionof CAD using X-ray angiography as the reference standard. Analysis was done at segment,vessel, and subject level.

RESULTS We analyzed 39 studies (41 separate comparisons). Across 25 studies (27 comparisons) withdata on 4,620 segments (993 subjects), sensitivity and specificity for detection of CAD were73% and 86%, respectively. Vessel-level analyses (16 studies, 2,041 vessels) showed sensitivity75% and specificity 85%. Subject-level analyses (13 studies, 607 subjects) showed sensitivity88% and specificity 56%. At the segment level, sensitivity was 69% to 79% for all but the leftcircumflex (61%) coronary artery; specificity was 82% to 91%. There was considerablebetween-study heterogeneity, but weighted summary receiver-operating characteristic curvesagreed with these estimates. There were no major differences between subgroups based ontechnical or population characteristics, year of publication, reported blinding, or sample size.

CONCLUSIONS In evaluable segments of the native coronary arteries, CMRA has moderately high sensitivityfor detecting significant proximal stenoses and may have value for exclusion of significantmultivessel CAD in selected subjects considered for diagnostic catheterization. (J Am Coll

ublished by Elsevier Inc. doi:10.1016/j.jacc.2004.07.051

Cardiol 2004;44:1867–76) © 2004 by the American College of Cardiology Foundation

neicaperhr(rCttStssw

maging of the coronary artery lumen is currently bestvaluated with X-ray angiography with selective intracoro-ary contrast injection. Besides establishing the diagnosis oforonary artery disease (CAD), angiographic data regardinguminal stenoses guide the selection between medical andnterventional/surgical therapies. The number of diagnosticardiac catheterizations has been steadily increasing over theast few years (1). This comes at a considerable cost, andhough the complication rate for each individual is propor-ionally small (2), in total, cardiac catheterizations accountor considerable procedure-related morbidity. A noninvasiveiagnostic alternative is desirable.

From the *2nd Cardiology Clinic and †Cardiac MRI Center, Hygeia Hospital,thens, Greece; Department of Medicine, ‡Cardiovascular Division, Beth Israeleaconess Medical Center and Harvard Medical School, Boston, Massachusetts;

Clinical Trials and Evidence-Based Medicine Unit and Clinical and Molecularpidemiology Unit, Department of Hygiene and Epidemiology, University of

oannina School of Medicine and the Biomedical Research Institute, Foundation foresearch and Technology-Hellas, Ioannina, Greece; and the �Institute for Clinicalesearch and Health Policy Studies, Tufts-New England Medical Center, Boston,assachusetts.

vManuscript received September 21, 2003; revised manuscript received July 26,

004, accepted July 29, 2004.

Among proposed alternatives, coronary magnetic reso-ance angiography (CMRA) offers high spatial resolution,liminates the need for iodinated contrast, and does notnvolve exposure to ionizing radiation. The many technicalhallenges have been addressed with various approaches,nd CMRA has been the focus of clinical research over theast decade. Using two-dimensional (2D) CMRA, initialncouraging results (3) were followed by less optimisticeports (4), thus creating confusion on whether CMRA canave a clinical role in the evaluation of CAD. Subsequenteports with 2D CMRA (5–12) and three-dimensional3D) CMRA (13–45) continued to show a seemingly broadange of sensitivity and specificity for detection of proximalAD. It is unknown whether this variability has been due

o chance or reflects differences in hardware platform,echnique implementation, and subject selection criteria.tudies conducted to date have been of limited size, andhus the confidence intervals for the observed sensitivity andpecificity are large. Given the sample size limitations,ingle studies have even less power for documentinghether CMRA might perform better for specific coronary
essels. To address these issues and to provide an evidence-

bmav

M

Estbsrrsmfaece

acCprcwsasCasuiwmDfapnpm

teciassea

tCTcoaSiddv

ec(pDltasniCrlswoce

ietBa1esspu(a

1868 Danias et al. JACC Vol. 44, No. 9, 2004Magnetic Resonance Angiography to Diagnose CAD November 2, 2004:1867–76

ased evaluation of the clinical utility of this new imagingodality, we performed a comprehensive meta-analysis of

ll currently available studies comparing CMRA with con-entional X-ray coronary angiography.

ETHODS

ligibility criteria and search strategy. We considered alltudies published in English, French, or German evaluatinghe presence of significant CAD in native coronary arteriesy both CMRA and X-ray angiography in the sameubjects. Studies were eligible regardless of whether theyeferred to subjects with suspected or proven CAD andegardless of technique used for CMRA. We excludedtudies with fewer than 10 subjects examined with bothethods, as these would be unlikely to contribute meaning-

ul diagnostic information. Studies evaluating vessel patencylone, phantom-only evaluations, and animal studies werexcluded. Studies limited exclusively to subjects with normaloronaries and studies assessing coronary grafts were alsoxcluded.

We identified eligible studies by searching MEDLINEnd EMBASE (last search January 2004). The searchombined exploded keywords pertaining to CAD andMRA (strategy available upon request). In addition, weerused bibliographies of retrieved articles and reviews. Theetrieved studies were carefully examined to exclude dupli-ation or overlap of subjects. Duplicate/overlapping dataere counted only once in the meta-analysis, unless the

ame subjects had been evaluated at different time pointsnd/or with different CMRA techniques. When a singletudy included more than one well-defined subject group orMRA techniques, each group/approach was considered asseparate comparison. We communicated with all corre-

ponding investigators of the identified reports to clarifynclear data and to obtain key data that were not reportedn the articles. Meeting abstracts were excluded, as theyould not provide adequately detailed data and their resultsight not be final.ata extraction. The following information was extracted

rom each pertinent study: first author, year of publication,nd journal; study population characteristics including sam-le size (number of subjects evaluated with both tests andumber of subjects evaluated with X-ray angiography only);ercentage of subjects with documented CAD; gender;

Abbreviations and AcronymsCAD � coronary artery diseaseCMRA � coronary magnetic resonance angiographySROC � summary receiver-operating characteristic3D � three-dimensional2D � two-dimensional

ean (or median) age (and standard deviation); relative d

iming of the two imaging procedures and blinding ofvaluators of one test to the results of the other and to thelinical condition of the tested subject; technical character-stics of the CMRA (including electrocardiogram gatingnd fat signal suppression [yes/no], respiratory motionuppression method [breath-holding vs. navigator (retro-pective vs. prospective)], mode of acquisition [2D vs. 3D],cho and repetition times, spatial and temporal resolution);nd hardware platform.

We recorded the true positive, false negative, false posi-ive, and true negative CMRA assessments for detection ofAD using X-ray angiography as the reference standard.he 50% diameter stenosis cutoff was selected as the

riterion for significant CAD, unless only a different thresh-ld had been used. Data were recorded separately, whenevervailable, at the level of segments, vessels, and subjects.eparate data were also recorded at the segment level for

ndividual major epicardial vessels (left main, left anteriorescending, left circumflex, and right coronary arteries). Alliagnostic accuracy estimates pertain to evaluable segments/essels/subjects. Detailed data are available upon request.

Two investigators (A.R., P.G.D.) performed the dataxtraction independently. Discrepancies were solved byonsensus and with discussion with the third investigatorJ.P.A.I.). Databases were enhanced with additional datarovided directly by primary investigators of several studies.ata synthesis. The main analysis was performed at the

evel of coronary artery segments, as most studies focused onhis level of information. Secondary analyses combined thevailable vessel-level data and subject-level data. Separateegment-level analyses were performed also for each coro-ary vessel and for prespecified subgroups, defined so as to

nvestigate the potential effect of key characteristics of theMRA technique (acquisition technique [2D vs. 3D] and

espiratory motion compensation technique), temporal evo-ution (year of publication [1991 to 2000 vs. 2001 to 2004]),tudy population characteristics (inclusion or not of subjectsithout CAD), potential for bias (reading of CMRA statedr not to be blind to the results of X-ray angiography and tolinical data; small vs. larger studies), and segments consid-red (including or not distal segments).

Sensitivity and specificity estimates for each analysis werendependently combined across studies using a randomffects model that weighs each proportion by the inverse ofhe sum of its variance and the between-study variance (46).etween-study heterogeneity was assessed for the sensitivitynd specificity estimates using the chi-square test with n �degrees of freedom (n is the number of studies) and Fisher

xact test when numbers were small. Because sensitivity andpecificity are interdependent, independent weighting mayometimes give spurious results and underestimate botharameters. Therefore, we also estimated weighted andnweighted summary receiver-operating characteristicSROC) curves (47) that have been established as the mostppropriate approach to bypass this problem of interdepen-
ence. The SROC curves are estimated from D � a � bS,

wpSisspusipcosc

snwT�(vssSIM(

T

MDMPYNNSWPKMOWHLSSvRSvBGKNNvNPRSvWWBIKKPR

* evel (Sreson

1869JACC Vol. 44, No. 9, 2004 Danias et al.November 2, 2004:1867–76 Magnetic Resonance Angiography to Diagnose CAD

here D is the difference of the logits (log odds) of the trueositives (sensitivity) and false positives (1 � specificity) andis the sum of these logits. When b is 0, the SROC curve

s symmetric around the diagonal. Conversely, when b isignificantly different from 0, the SROC curve is notymmetrical, and this suggests that the overall diagnosticerformance varies in different parts of the curve, withnequal tradeoff between sensitivity and specificity acrosstudies. This indicates significant heterogeneity across stud-es in the thresholds used, the study populations, or otherarameters. Summary receiver-operating characteristicurves should not be extrapolated outside the range ofbserved values. Unweighted SROC curves consider alltudies equally in the calculations, whereas weighted SROC

able 1. Study and Population Characteristics

Author, Year (Ref.)X-Ray

(CMRA)CAD

(%)

Mean Age(SD) or Range

(yrs)M(

anning, 1993 (3) 39 (39) 74 54uerinckx, 1994 (4) 21 (20) 85 61 (8)ohiaddin, 1996 (6) 16 (16) NR 57 (8) N

ost, 1997 (8) 35 (35) NR 58 (10)oshino, 1997 (9) 76 (31) 100 61 (10)itatori, (a) 1998 (10) 70 (70) 100 NR Nitatori, (b) 1998 (10) 43 (43) NR NR Nchmidt, 1999 (11) 29 (29) 100 60 (10)atanuki, 2000 (12) 123 (108) 100 61 (11)

ost, 1996 (13) 20 (20) 75 58essler, 1997 (14) 73 (65) NR 60uller, 1997 (15) 35 (29) 83 61shinski, 1998 (16) 13 (11) NR 57 Noodard, 1998 (17) 10 (10) 100 60uber, 1999 (18) 20 (20) 100 65 (7) Nethimonnier, 1999 (19) 21 (17) NR 57andstede, 1999 (20) 20 (19) 58 61 (9)andstede, 1999 (21) 30 (23) 91 63 (9)an Geuns, 1999 (22) 32 (29) 59 32–73egenfus, 2000 (24) 50 (50) 72 61ardanelli, 2000 (25) 42 (39) 87 65 (9)an Geuns, 2000 (26) 38 (34) 68 43–72unce, 2001 (27) 46 (46) 50 34–84onshior, 2001 (28) 20 (20) 100 65 (8) Nim, 2001 (29) 109 (103) 58 59 (10)ikolaou, (a) 2001 (30) 20 (20) 100 65 (8) Nikolaou, (b) 2001 (30) 20 (20) 100 62 (10) N

an Geuns, 2001 (31) 13 (12) NR 60ikolaou, 2002 (32) 20 (20) 100 63 (8)lein, 2002 (33) 10 (10) 100 61opers, 2002 (34) 118 (87) 17 64ommer, 2002 (36) 112 (107) 59 62 Nan Geuns, 2002 (37) 27 (27) 70 38–76

atanabe, 2002 (38) 16 (12) 100 71eber, 2002 (39) 15 (15) 73 61

ogaert, 2003 (40) 21 (19) 62 62 (5)konen, 2003 (41) 69 (69) 68 58lein, 2003 (42) 12 (10) NR 61 (5)rittayaphong, 2003 (43) 61 (57) 67 63 (9)lein, 2003 (44) 45 (50) 90 59egenfus, 2003 (45) 61 (61) 75 63 (6)

Availability of 2 � 2 tables at the subject level (Su), vessel level (V), and segment lC � clinical data; CAD � coronary artery disease; CMRA � coronary magnetic

urves weigh each study by the variance of D. t

Finally, for the derived estimates of sensitivity andpecificity for CAD at the subject level, we evaluated theegative and positive predictive value among subjectsith 5%, 20%, 50%, and 80% pretest probability of CAD.he positive predictive value is calculated as pretest odds

(sensitivity/[1 � specificity])/(1 � [pretest odds �sensitivity/[1 � specificity])]). The negative predictivealue is calculated as pretest odds � ([1 � sensitivity]/pecificity)/(1 � [pretest odds � ([1 � sensitivity]/pecificity]).oftware. Analyses were performed in SPSS 11.0 (SPSSnc., Chicago, Illinois), Meta-Analyst (Joseph Lau, Boston,

assachusetts), Meta-Test (Joseph Lau) and StatXact 3.0Cytel Inc., Boston, Massachusetts). All p values are

TimeDiff.

(days)Blinding

Stated

Data*(% Evaluable

Segments)Distal

Segments

0–7 T, C Su, V No0–60 T, C S No0–240 T V No0–29 T V No0–7 T, C S NoNR — S NoNR — S No

0–14 — V No0–14 T, C S No0–21 T V No0–7 — S (52%) No0–1 T S No0–21 T, C V No0–30 T V No0–10 T Su, S (100%) No0–9 T S No0–3 T Su, S (73%) No0–4 T, C Su Yes (RCA only)0–30 — Su, S No0–3 T Su, S (77%) No0–14 T, C Su, V, S Yes0–30 T S Yes (RCA only)0–90 T V No0–10 T, C S No0–14 T, C Su, V No0–10 T, C S (100%) No0–10 T, C S (100%) NoNR T, C S No

0–10 T, C S (82%) No90 � 65 T, C Su, V NR1–3 T, C Su, V No0–16 — S No0–30 T, C S (69%) Yes (RCA only)

0 T S (70%) Yes0–28 T Su, S (70%) Yes (LAD,RCA only)0–1 T Su, V, S (70%) Yes0–1 T, C Su, V No

0 T, C S No0–1 T, C Su, V NoNR T, C Su, V Yes

0–3 T S (77%) No

).ance angiography; NR � not reported; T � test (results of X-ray angiography).

ale%)

9095

R7783

RR

7677807571

R50

R9575776380797180

R69

RR

69658077

R597593716292517285

wo-tailed.

R

Wocmsnpda1cCy

s82uctusmtt

(

T

MDMPYNNSWPKMOWHLSSvRSvBGKNNvNPRSvWWBIKKPR

B tive nT nal.


ESULTS

e identified 41 potentially eligible and apparently non-verlapping studies (3–22,24–34,36–45) with 44 differentomparisons of CMRA against X-ray angiography. Supple-ental data were provided by primary investigators for 13

tudies (14 comparisons). For two studies (5,7), we couldot obtain sufficient information for constructing any com-lete 2-by-2 table for any analysis. The remaining 39 studiesescribed 41 comparisons of the two imaging techniques intotal of 1,671 subjects with X-ray angiography, of whom,522 also had CMRA data (91.1%). Twelve studies (12omparisons) with 370 subjects included only subjects withAD and these were not considered in subject-level anal-

able 2. Technical CMRA Characteristics

Author, Year (Ref.) Scanner Seq

anning, 1993 (3) Siemens Magnetom SP 2Duerinckx, 1994 (4) Siemens Magnetom SP 2Dohiaddin, 1996 (6) Picker Vista 2D

ost, 1997 (8) Siemens Magnetom SP 2Doshino, 1997 (9) Siemens Magnetom Vision 2Ditatori, (a) 1998 (10) Siemens Magnetom Vision 2Ditatori, (b) 1998 (10) Siemens Magnetom Vision 2Dchmidt, 1999 (11) Philips Gyroscan NT/ACS 2Datanuki, 2000 (12) Siemens Magnetom Vision 2D

ost, 1996 (13) Siemens Magnetom SP 3Dessler, 1997 (14) Siemens Magnetom Vision 3Duller, 1997 (15) Siemens Magnetom Vision 3Dshinski, 1998 (16) Philips Gyroscan ACS/NT 3Doodard, 1998 (17) Siemens Magnetom Vision 3Duber, 1999 (18) Siemens Magnetom Vision 3Dethimonnier, 1999 (19) GE Signa Echospeed 5–7 3Dandstede, 1999 (20) Siemens Magnetom Vision 3Dandstede, 1999 (21) Siemens Magnetom Vision 3Dan Geuns, 1999 (22) Siemens Magnetom Vision 3Degenfus, 2000 (24) Siemens Magnetom Vision 3D cardanelli, 2000 (25) Philips Gyroscan ACS/NT 3Dan Geuns, 2000 (26) Siemens Magnetom Vision 3D cunce, 2001 (27) Siemens Magnetom Vision 3Donshior, 2001 (28) Picker Edge 3Dim, 2001 (29) Siemens Magnetom Vision 3Dikolaou, (a) 2001 (30) Siemens Magnetom Vision 3Dikolaou, (b) 2001 (30) Siemens Magnetom Vision 3D

an Geuns, 2001 (31) Siemens Magnetom Vision 3Dikolaou, 2002 (32) Siemens Magnetom Vision 3Dlein, 2002 (33) Philips Gyroscan NT Intera CV 3Dopers, 2002 (34) Siemens Magnetom Vision 3D commer, 2002 (36) Philips Gyroscan Intera 3Dan Geuns, 2002 (37) Siemens Magnetom Vision 3D

atanabe, 2002 (38) Philips Gyroscan ACS/NT 3Deber, 2002 (39) Philips Gyroscan ACS/NT 3D

ogaert, 2003 (40) Philips Gyroscan ACS/NT 3Dkonen, 2003 (41) Siemens Magnetom Vision 3Dlein, 2003 (42) Philips Gyroscan ACS/NT 3Drittayaphong, 2003 (43) Philips Gyroscan ACS/NT 3Dlein, 2003 (44) Philips Gyroscan NT Intera CV 3Degenfus, 2003 (45) Siemens Vision 3D c

H � breath holding; MA � motion adaptive; NR � not reported; PN � prospecR/TE � repetition time/echo time; 2D � two-dimensional; 3D � three-dimensio

ses because no specificity could be estimated in these s

tudies. Of the remaining 29 comparisons (1,152 subjects),did not specify how many subjects had CAD, whereas in

0 of the other 21 comparisons, patients with CAD madep half or more of the study population. Study populationharacteristics are shown in Table 1 and technical charac-eristics of the CMRA are shown in Table 2. All studiessed gradient echo (bright blood) sequences and only fourtudies (24,26,34,45) used intravenous contrast enhance-ent. Data acquisition was consistently gated to the elec-

rocardiographic signal and timed to diastole, and nulling ofhe epicardial fat signal was employed.

Data on diagnostic accuracy were available for 25 studies27 comparisons) (4,620 segments of 993 subjects) at the

RMCTR/TE

(ms)

TemporalResolution

(ms)

In-Plane SpatialResolution

(mm)

BH 13/8 78–104 1.4–1.9 � 0.9BH 13/8 117 1.6–2.1 � 0.9–1.2BH 15.7/6.5 126 1.6 � 0.8BH 12.6/7 88 or 113 1.8 � 1.0BH 17/7.9 119–153 NRBH 17/7.9 119–153 NRBH 17/7.9 119–153 NRPN 14/7 140 0.7–0.9 � 1.4–1.9BH 17/7.9 119–153 NRRN 8.4/2.9 260 1.2 � 2.3RN 7.4/2.7 178 1.17 � 1.17RN 7.4/3.6 118 1.95 � 0.98PN 13/7 104 0.7–0.9 � 0.9–1.2RN 8/3 NR 1.2 � 1.2RN 7.4/2.7 178 1.2 � 1.4PN 7.3/2.2 146 1.25 � 1.66RN 6/2.4 NR 1.00 � 1.21RN 7.3/2.7 230 1.17 � 1.17RN 7.4/2.7 128 1.9 � 1.25

t BH 4.2/1.6 294 1.4 � 1.25PN 7.7/2.2 71 0.7 � 1.0

t BH 5.3/2.3 110 1.9 � 1.25RN 7.4/2.7 178 1.2 � 2.3PN 12.9/4.3 NA 1.0–1.2 � 1.4–2.3RN 7.4/2.7 178 1.2 � 1.4RN 7.4/2.7 178 1.0 � 1.5RN 6/2.4 144 1.0 � 1.5BH 5.3/2.3 NR 1.9 � 1.25RN 6/2.4 144 1.0 � 1.5PN 7/2.1 140–200 0.78 � 1.04

t RN 7.4/2.7 125 1.17 � 1.17PN 7.4/2.2 82 0.70 � 0.79RN 8/3 128 1.9 � 1.25PN 9.4/2.7 56 0.65 � 0.92PN 8/4 96 0.7 � 0.7PN 7.1/1.97 �70 0.7 � 1RN 600–800/2.7 118 1.17 � 1.17PN 8/2.3 52 0.9 � 0.7PN 7.7/2.2 NR 0.7 � 1

PN MA 7.4/2.5 78–200 0.78 � 1.04t BH 4.2/1.6 NR 1.4 � 1.25

avigator; RMC � respiratory motion compensation; RN � retrospective navigator;

uence

ontras

ontras

ontras

ontras

egment level, 16 studies (2,041 vessels of 624 subjects) at

tTuui�o

swl1Wcttmtdcswowcsavt

trdwcd(sfaln

itevalompccSTS

dcbstdsp

8saa1

D

Cppypmwdv

sonancects.


he vessel level, and 13 studies (n � 607) at the subject level.he vast majority of studies defined significant CAD bysing the 50% stenosis cutoff. However, two studies (34,44)sed 70% as the cutoff for CAD; one other study (17)ncluded patients with �30% stenoses, and all patients had

40% vessel diameter stenosis by conventional X-ray cor-nary angiography.There was significant between-study heterogeneity in the

ensitivity and specificity estimates in all analyses (Table 3),ith the exception of the sensitivity estimates at the subject

evel that consistently showed high sensitivity (range 73% to00%), whereas specificity varied widely across studies.

eighted random effects independent estimates fell verylose to the weighted SROC curve (Fig. 1), suggesting thathey are appropriate approximations of the overall diagnos-ic performance. Unweighted SROC curves were somewhatore favorable than weighted SROC curves, suggesting

hat smaller studies tended to show slightly more favorableiagnostic performance than larger studies. The SROCurve for the segment-level analysis showed statisticallyignificant asymmetry: across studies, specificity diminishedithout any gains in sensitivity beyond 85% to 90%. Theverall weighted estimate suggested a sensitivity of 73%ith specificity of 86%. However, there seemed to be

lusters of studies: one with low sensitivity (�70%) and highpecificity (�85%), another with high sensitivity (�80%)nd also high specificity (�85%), and a third one withariable sensitivity (60% to 92%) and low specificity (50%o 75%).

At the segment level, the diagnostic accuracy was rela-ively similar for the left main, left anterior descending, andight coronary arteries, with 69% to 79% of the lesions beingetected, against a specificity of 81% to 91%. The sensitivityas considerably lower for lesions in the left circumflex

oronary artery with only slightly over half the lesions beingetected, and without the specificity being any betterTable 3, Fig. 2). There was modest heterogeneity inensitivity estimates across studies. Heterogeneity reachedormal statistical significance for the left anterior descendingnd the right coronary arteries, but not for the left main andeft circumflex coronary arteries, possibly due to smaller

Table 3. Summary of Diagnostic Accuracy of CAngiography in Evaluable Segments, Vessels, a

AnalysisStudies

(n)W

Segment level 27 (4,620)Left main† 19 (802)LAD 21 (1,058)LCx 21 (674)RCA 21 (990)

Subject level 13 (607)Vessel level 11 (1,271)

*Statistically significant (p � 0.10) between-study heterogenwith the proximal segment of the LAD. These data were in

CI � confidence interval; CMRA � coronary magnetic recircumflex; RCA � right coronary artery; RE � random eff

umbers. There was significant between-study heterogene- w

ty across studies for all specificity estimates. In the case ofhe left main coronary artery this heterogeneity was totallyxplained by two studies (18,28) that paradoxically showedery low specificity (27%). These two studies, along withnother report (4) that also showed �50% specificity for theeft anterior descending coronary artery, accounted for thebserved between-study heterogeneity for this vessel’s seg-ents as well. Absolute differences in specificity were less

ronounced between studies for the right and left circumflexoronary arteries. Random effects weighted estimates felllose to the weighted SROC curve, whereas unweightedROC curves provided somewhat more favorable results.here was no statistically significant asymmetry for anyROC curve in these analyses.Subgroup analyses at the segment level showed no evi-

ence that the diagnostic performance might differ statisti-ally significantly across subgroups (p � 0.05 for alletween-subgroup heterogeneity in both sensitivity andpecificity random effects estimates) (Table 4). Nevertheless,here were trends suggesting somewhat paradoxically betteriagnostic performance in earlier studies, 2D studies, andtudies using breath holding for respiratory motion com-ensation.We estimated that for subjects with 5%, 20%, 50%, and

0% pretest probability of CAD, a positive CMRA wouldlightly increase the probability of CAD to 10%, 33%, 66%,nd 89%, respectively. Given the same pretest probabilities,negative test would decrease the risk of CAD to 1.1%, 5%,8%, and 46%, respectively.

ISCUSSION

oronary magnetic resonance angiography has been aromising alternative for invasive X-ray contrast angiogra-hy and has been intensely investigated over the past severalears. Our meta-analysis combined data on the diagnosticerformance of CMRA that represent the experience ofany centers around the world. We showed that in subjectsith intermediate/high likelihood for CAD, CMRA canetect about three-quarters of significant stenoses in theisualized segments of the major epicardial coronary arteries

A Against Conventional X-Rayatients

hted Sensitivity Weighted SpecificityE (95% CI) RE (95% CI)

(69%–77%)* 86% (80%–90%)*(56%–79%) 91% (84%–95%)*(73%–84%)* 81% (71%–88%)*(52%–69%) 85% (78%–90%)*(64%–78%)* 84% (77%–88%)*(82%–92%) 56% (43%–68%)*(68%–80%)* 85% (78%–90%)*

four studies (22,26,37,41) evaluated the left main togetherin the LAD analysis.

e angiography; LAD � left anterior descending; LCx � left

MRnd P

eigR

73%69%79%61%71%88%75%

eity; †cluded

ith a concomitant specificity of 86%. Coronary magnetic

rm(ivfha

fTfia

v

Ftcocb


esonance angiography has very good diagnostic perfor-ance in all vessels except the left circumflex coronary artery

possibly due to the close relationship with the accompany-ng vein and adjacent blood pools of the left atrium andentricle, and lower signal related to the greater distancerom the receiving coil). Across populations with generallyigh prevalence of coronary stenoses, CMRA showed over-

igure 1. Summary receiver-operating characteristic (SROC) curves for thhe gold standard of X-ray coronary angiography using segment-level (Aomparison. Within each panel the diameters of the ellipses are inversely pf specificity (horizontal diameter) of each comparison. Also shown is thealculations (�) as well as the 95% confidence intervals (box). Weighted Sy thin lines.

ll high sensitivity for identifying those with CAD, but gave s

alse positive readings in about half of the healthy subjects.hus, CMRA may be particularly useful when it is negative

or decreasing the probability of CAD below 5%, especiallyn individuals with modest suspicion for CAD, for example,pretest probability for CAD below 20%.One should recognize that the visual evaluation of con-

entional angiography may not be a perfect reference

gnostic performance of coronary magnetic resonance angiography againstssel-level (B), and subject-level (C) data. Each ellipse corresponds to ational to the variance of the sensitivity (vertical diameter) and the variancependently weighted sensitivity and specificity according to random effects

curves are shown with bold lines; unweighted SROC curves are shown

e dia), ve

roporindeROC

tandard (48). Therefore, the diagnostic performance of

Cot

ecsewWp

dpdnftdacm

Fsr


MRA may have been slightly underestimated. Conversely,ther biases such as reporting bias may have led to overes-imation of CMRA performance.

The technical aspects of CMRA are quickly evolving. Forxample, 3D imaging is now routinely performed by mostenters. Nevertheless, we did not observe any statisticallyignificant differences in the diagnostic performance ofarlier studies using less advanced techniques, as comparedith more recent studies and more advanced technologies.e should acknowledge that the meta-analysis was not

igure 2. Summary receiver-operating characteristic curves for the diagnotandard of X-ray angiography using segment-level data according to coroight (D) coronary arteries. Setup as in Figure 1.

owered to detect modest subgroup differences and such s

ifferences might have been missed. Alternatively, thesearameters may not be very important for the overalliagnostic accuracy of CMRA. If anything, we notedon-statistically significant trends for better diagnostic per-ormance in earlier studies with less advanced techniques. Ifhis is not a chance finding, it may reflect intangibleifferences in the subject selection process, center expertisend performance, and changes in the criteria used toharacterize CMRA as “evaluable” or not. The singleulticenter study reported to date demonstrated 93% sen-

erformance of coronary magnetic resonance angiography against the goldessel: left main (A), left anterior descending (B), left circumflex (C), and

stic pnary v

itivity and 42% specificity at the subject level (29). These

vtsusp(m(cw

Cahlnampiecsiflss

dydswespffninea

tccrpc(maebsp

eity.l.


alues reflect a higher sensitivity and lower specificity thanhe main meta-analysis estimates, but the tradeoff is of fairlyimilar magnitude. The anticipated negative predictive val-es based on the results of the multicenter study would belightly more optimistic than those of the meta-analysis atretest CAD probabilities of 20% (4% vs. 5%) or 50%14.3% vs. 18%), but the differences are not large. Theeta-analysis demonstrates in a much larger sample size

approximately a log scale higher than that of the multi-enter study) that CMRA holds clinical promise across aider range of diverse clinical centers.The published reports on the diagnostic performance of

MRA are quite heterogeneous regarding study design andnalytic methodologies. For example, evaluation of CADas been performed in different ways (subject level vs. vessel

evel vs. segment level). However, heterogeneity is notecessarily a limitation in meta-analysis (49), and it provideskey opportunity to show the consistent performance of theethod. We should also acknowledge that not all reports

rovided complete data. Nevertheless, we contacted severalnvestigators to obtain clarifications or additional data thatnhanced our analyses. More rigorous reporting of futurelinical research on coronary artery imaging technologieshould be encouraged. Our study attempted to addressssues pertaining to different analytic approaches by per-orming separate analyses for each level. These results areargely consistent and show the anticipated increase inensitivity and decrease in specificity as one moves fromegments to vessels to subjects.

Table 4. Subgroup Analyses of Evaluable Segm95% CI

Subgroups Studies (

Acquisition technique2-dimensional 4 (605)3-dimensional 23 (4,22

Respiratory motion compensationBreath holding 9 (1,54Retrospective navigator 13 (2,00Prospective navigator 5 (1,07

Publication year�2001 12 (2,11�2000 15 (2,50

Non-CAD subjectsIncluded 18 (3,27Not included 9 (1,34

Distal segmentsConsidered 5 (568)Not considered 22 (4,05

Blinding statedTest and clinical data 12 (1,64Test only 10 (1,60None 5 (1,36

Sample size�200 segments 7 (2,30�200 segments 20 (2,31

*Statistically significant (p � 0.10) between-study heterogenCAD � coronary artery disease; CI � confidence interva

The vast majority of the included studies did not evaluate s

istal segments of the coronary vessels. As CMRA cannotet visualize well the distal coronaries and branch vessels, itsiagnostic performance may be lower for lesions in theseegments. Until CMRA improves technically to the pointhere its performance for distal stenoses can be reliably

valuated, the clinical use of CMRA should only be re-tricted to proximal vessels. This may be adequate for manyatients, as the decision regarding surgical therapy of CADrequently relies on the presence of proximal stenoses. Evenor the proximal coronary arteries, however, current tech-ology only allows for evaluation of roughly 80% of the

maged segments. This limitation, also shared by otheroninvasive approaches (32,50–52), is an important consid-ration for the widespread use of these approaches aslternatives to conventional angiography.

Our findings should be used cautiously when attemptingo support the clinical utility of the method. Large intra-enter differences likely make the technique still very muchenter-dependent and, therefore, larger experience is stillequired. Though various magnetic resonance imaging ap-roaches have been described for coronary artery imaging,urrent published experience only involves gradient echo“bright blood”) techniques. With this methodology, nor-ally flowing blood appears bright and stenoses/occlusions

ppear dark. Such techniques have known pitfalls. Forxample, coronary artery calcifications may cause suscepti-ility artifacts and present as signal voids, erroneouslyuggesting the presence of significant stenoses. Other ap-roaches (black blood imaging, spiral imaging, enhanced

Level Data: Random Effects Estimates and

WeightedSensitivity

WeightedSpecificity

80% (65%–90%)* 91% (76%–97%)*72% (67%–76%)* 86% (80%–90%)*

78% (69%–84%)* 91% (84%–95%)*71% (65%–77%)* 80% (71%–87%)*70% (54%–82%)* 89% (69%–96%)*

71% (63%–77%)* 83% (74%–89%)*74% (68%–80%)* 88% (80%–93%)*

71% (64%–78%)* 88% (82%–93%)*74% (69%–79%)* 79% (66%–88%)*

64% (46%–79%)* 93% (88%–96%)*74% (70%–78%)* 83% (77%–88%)*

71% (65%–77%)* 80% (71%–87%)*75% (67%–83%) 91% (81%–96%)*72% (59%–82%)* 84% (67%–93%)*

76% (69%–82%)* 83% (70%–91%)*71% (65%–77%)* 87% (80%–91%)*

ent-

n)

3)

6)0)4)

2)8)

2)8)

2)

8)4)8)

2)8)

equences, high field imaging) (53–55) may offer additional

vrupoCcgpp

madCvjdsh

ATANRfct

RCah

R

1

1

1

1

1

1

1

1

1

1

2

2

2

2

2

2

2

2


alue. Imaging of the coronary vessel wall and atheroscle-otic plaque (56–58) may also enhance our ability tonderstand the pathophysiology of atherosclerosis and im-act on its course, to improve patients’ outcomes. The utilityf such approaches remains to be proven. The merits ofMRA would also have to be evaluated against other

ompeting noninvasive coronary artery imaging technolo-ies, and in particular computed tomography-based ap-roaches, such as electron-beam and multislice spiral com-uted tomography (32,50–52,59,60).In conclusion, across many studies CMRA has shownoderately high sensitivity and may have satisfactory neg-

tive predictive value for excluding significant major epicar-ial coronary artery stenoses in subjects with suspectedAD. Coronary magnetic resonance angiography may have

alue for exclusion of significant multivessel CAD in sub-ects referred for diagnostic catheterization, but the availableata do not suffice to introduce CMRA as a widely appliedcreening tool, particularly for individuals with low likeli-ood of CAD.

cknowledgmentshe authors wish to thank Drs. Andre Duerinckx,rmin Huber, Aki Ikonen, Warren Manning, Konstantinikolaou, Dudley Pennel, Sven Plein, Franco Sardanelli,obert vanGeuns, Cristoph Weber, and Pamela Woodard

or responding to the request for additional data/larifications and/or for offering their comments regardinghis meta-analysis.

eprint requests and correspondence: Dr. Peter G. Danias,ardiac MR Center, Hygeia Hospital, 4 Erythrou Stavrou Street

nd Kifissias Avenue, Maroussi 15123, Greece. E-mail: [email protected].

EFERENCES

1. American Heart Association. 2004 Heart and Stroke StatisticalUpdate. Dallas, TX: 2004.

2. Johnson LW, Lozner EC, Johnson S, et al. Coronary arteriography1984–1987: a report of the Registry of the Society for CardiacAngiography and Interventions. I. Results and complications. CathetCardiovasc Diagn 1989;17:5–10.

3. Manning WJ, Li W, Edelman RR. A preliminary report comparingmagnetic resonance coronary angiography with conventional angiog-raphy. N Engl J Med 1993;328:828–32.

4. Duerinckx AJ, Urman MK. Two-dimensional coronary MR angiog-raphy: analysis of initial clinical results. Radiology 1994;193:731–8.

5. Davis SF, Kannam JP, Wielopolski P, Edelman RR, Anderson TJ,Manning WJ. Magnetic resonance coronary angiography in hearttransplant recipients. J Heart Lung Transplant 1996;15:580–6.

6. Mohiaddin RH, Bogren HG, Lazim F, et al. Magnetic resonancecoronary angiography in heart transplant recipients. Coron Artery Dis1996;7:591–7.

7. Pennell DJ, Bogren HG, Keegan J, Firmin DN, Underwood SR.Assessment of coronary artery stenosis by magnetic resonance imaging.Heart 1996;75:127–33.

8. Post JC, van Rossum AC, Hofman MB, de Cock CC, Valk J, VisserCA. Clinical utility of two-dimensional magnetic resonance an-
giography in detecting coronary artery disease. Eur Heart J 1997;18:426 –33.
9. Yoshino H, Nitatori T, Kachi E, et al. Directed proximal magneticresonance coronary angiography compared with conventional contrastcoronary angiography. Am J Cardiol 1997;80:514–8.

0. Nitatori T, Yokoyama K, Hachiya J, Seki T, Yoshino H, Ishikawa K.Comparison of 2D coronary MR angiography with conventionalangiography-difference in imaging accuracy according to the severity ofstenosis. Acian Oceanian J Radiol 1998;3:15–9.

1. Schmidt M, Theissen P, Crnac J, et al. [Clinical value of magneticresonance tomography in imaging coronary stenoses. A comparisonwith coronary angiography and myocardial scintigraphy] Dtsch MedWochenschr 1999;124:1294–300.

2. Watanuki A, Yoshino H, Udagawa H, et al. Quantitative evaluation ofcoronary stenosis by coronary magnetic resonance angiography. HeartVessels 2000;15:159–66.

3. Post JC, van Rossum AC, Hofman MB, Valk J, Visser CA. Three-dimensional respiratory-gated MR angiography of coronary arteries:comparison with conventional coronary angiography. Am J Roentge-nol 1996;166:1399–404.

4. Kessler W, Achenbach S, Moshage W, et al. Usefulness of respiratorygated magnetic resonance coronary angiography in assessing narrow-ings �50% in diameter in native coronary arteries and in aortocoronarybypass conduits. Am J Cardiol 1997;80:989–93.

5. Muller MF, Fleisch M, Kroeker R, Chatterjee T, Meier B, Vock P.Proximal coronary artery stenosis: three-dimensional MRI with fatsaturation and navigator echo. J Magn Reson Imaging 1997;7:644–51.

6. Oshinski JN, Hofland L, Dixon WT, Pettigrew RI. Magnetic reso-nance coronary angiography using navigator echo gated real- time slicefollowing. Int J Card Imaging 1998;14:191–9.

7. Woodard PK, Li D, Haacke EM, et al. Detection of coronary stenoseson source and projection images using three-dimensional MR angiog-raphy with retrospective respiratory gating: preliminary experience.Am J Roentgenol 1998;170:883–8.

8. Huber A, Nikolaou K, Gonschior P, Knez A, Stehling M, Reiser M.Navigator echo-based respiratory gating for three-dimensional MRcoronary angiography: results from healthy volunteers and patientswith proximal coronary artery stenoses. Am J Roentgenol 1999;173:95–101.

9. Lethimonnier F, Furber A, Morel O, et al. Three-dimensional coronaryartery MR imaging using prospective real-time respiratory navigator andlinear phase shift processing: comparison with conventional coronaryangiography. Magn Reson Imaging 1999;17:1111–20.

0. Sandstede J, Pabst T, Kenn W, Beer M, Neubauer S, Hahn D.[3-dimensional MR coronary angiography in the navigator technic forthe primary diagnosis of coronary heart disease: a comparison withconventional coronary angiography]. Rofo Fortschr Geb RontgenstrNeuen Bildgeb Verfahr 1999;170:269–74.

1. Sandstede JJ, Pabst T, Beer M, et al. Three-dimensional MR coronaryangiography using the navigator technique compared with conven-tional coronary angiography. Am J Roentgenol 1999;172:135–9.

2. van Geuns RJ, de Bruin HG, Rensing BJ, et al. Magnetic resonanceimaging of the coronary arteries: clinical results from three-dimensional evaluation of a respiratory gated technique. Heart 1999;82:515–9.

3. Ikonen AE, Manninen HI, Vainio P, et al. Repeated 3D coronaryMR angiography with navigator echo gating: technical quality andconsistency of image interpretation. J Comput Assist Tomogr2000;24:375– 81.

4. Regenfus M, Ropers D, Achenbach S, et al. Noninvasive detection ofcoronary artery stenosis using contrast-enhanced three-dimensionalbreath-hold magnetic resonance coronary angiography. J Am CollCardiol 2000;36:44–50.

5. Sardanelli F, Molinari G, Zandrino F, Balbi M. Three-dimensional,navigator-echo MR coronary angiography in detecting stenoses of themajor epicardial vessels, with conventional coronary angiography as thestandard of reference. Radiology 2000;214:808–14.

6. van Geuns RJ, Wielopolski PA, de Bruin HG, et al. MR coronaryangiography with breath-hold targeted volumes: preliminary clinicalresults. Radiology 2000;217:270–7.

7. Bunce NH, Jhooti P, Keegan J, et al. Evaluation of free-breathingthree-dimensional magnetic resonance coronary angiography withhybrid ordered phase encoding (HOPE) for the detection of proximal
coronary artery stenosis. J Magn Reson Imaging 2001;14:677–84.

2

2

3

3

3

3

3

3

3

3

3

3

4

4

4

4

4

4

4

4

4

4

5

5

5

5

5

5

5

5

5

5

6


8. Gonschior P, Pragst I, Valassis G, Vogel-Wiens C, Huber A.High-resolution MR angiography: results in diseased arteries. J Inva-sive Cardiol 2001;13:151–7; discussion, 158–70.

9. Kim WY, Danias PG, Stuber M, et al. Coronary magnetic resonanceangiography for the detection of coronary stenoses. N Engl J Med2001;345:1863–9.

0. Nikolaou K, Huber A, Knez A, Scheidler J, Petsch R, Reiser M.Navigator echo-based respiratory gating for three-dimensional MRcoronary angiography: reduction of scan time using a slice interpola-tion technique. J Comput Assist Tomogr 2001;25:378–87.

1. van Geuns RJ, Wielopolski PA, Wardeh AJ, de Bruin HG, Oudkerk M,de Feyter PJ. Volume coronary angiography using targeted scans(VCATS): a new strategy in MR coronary angiography. Int J CardiovascImaging 2001;17:405–10.

2. Nikolaou K, Huber A, Knez A, Becker C, Bruening R, Reiser M.Intraindividual comparison of contrast-enhanced electron-beam com-puted tomography and navigator-echo-based magnetic resonance im-aging for noninvasive coronary artery angiography. Eur Radiol 2002;12:1663–71.

3. Plein S, Ridgway JP, Jones TR, Bloomer TN, Sivananthan MU.Coronary artery disease: assessment with a comprehensive MR imag-ing protocol—initial results. Radiology 2002;225:300–7.

4. Ropers D, Regenfus M, Stilianakis N, et al. A direct comparison ofnoninvasive coronary angiography by electron beam tomography andnavigator-echo-based magnetic resonance imaging for the detection ofrestenosis following coronary angioplasty. Invest Radiol 2002;37:386–92.

5. Regenfus M, Ropers D, Achenbach S, et al. Comparison of contrast-enhanced breath-hold and free-breathing respiratory-gated imaging inthree-dimensional magnetic resonance coronary angiography. Am JCardiol 2002;90:725–30.

6. Sommer T, Hofer U, Hackenbroch M, et al. [Submillimeter 3Dcoronary MR angiography with real-time navigator correction in 107patients with suspected coronary artery disease]. Rofo Fortschr GebRontgenstr Neuen Bildgeb Verfahr 2002;174:459–66.

7. van Geuns RJ, Oudkerk M, Rensing BJ, et al. Comparison of coronaryimaging between magnetic resonance imaging and electron beamcomputed tomography. Am J Cardiol 2002;90:58–63.

8. Watanabe Y, Nagayama M, Amoh Y, et al. High-resolution selectivethree-dimensional magnetic resonance coronary angiography withnavigator-echo technique: segment-by-segment evaluation of coronaryartery stenosis. J Magn Reson Imaging 2002;16:238–45.

9. Weber C, Steiner P, Sinkus R, Dill T, Bornert P, Adam G.Correlation of 3D MR coronary angiography with selective coronaryangiography: feasibility of the motion-adapted gating technique. EurRadiol 2002;12:718–26.

0. Bogaert J, Kuzo R, Dymarkowski S, Beckers R, Piessens J, Rademak-ers FE. Coronary artery imaging with real-time navigator three-dimensional turbo-field-echo MR coronary angiography: initial expe-rience. Radiology 2003;226:707–16.

1. Ikonen AE, Manninen HI, Vainio P, et al. Three-dimensionalrespiratory-gated coronary MR angiography with reference to X-raycoronary angiography. Acta Radiol 2003;44:583–9.

2. Klein C, Schalla S, Schnackenburg B, et al. Improvement of imagequality of non-invasive coronary artery imaging with magnetic reso-nance by the use of the intravascular contrast agent Clariscan(NC100150 injection) in patients with coronary artery disease. J Magn
Reson Imaging 2003;17:656–62.
3. Krittayaphong R, Mahanonda N, Kangkagate C, Nakyen S, Tanapibun-pon P, Chaithiraphan S. Accuracy of magnetic resonance imaging in thediagnosis of coronary artery disease. J Med Assoc Thai 2003;86 Suppl1:S59–66.

4. Plein S, Jones TR, Ridgway JP, Sivananthan MU. Three-dimensionalcoronary MR angiography performed with subject-specific cardiacacquisition windows and motion-adapted respiratory gating. Am JRoentgenol 2003;180:505–12.

5. Regenfus M, Ropers D, Achenbach S, et al. Diagnostic value ofmaximum intensity projections versus source images for assessment ofcontrast-enhanced three-dimensional breath-hold magnetic resonancecoronary angiography. Invest Radiol 2003;38:200–6.

6. Fleiss JL. The statistical basis of meta-analysis. Stat Methods Med Res1993;2:121–45.

7. Littenberg B, Moses LE. Estimating diagnostic accuracy from multi-ple conflicting reports: a new meta-analytic method. Med DecisMaking 1993;13:313–21.

8. Gurley JC, Nissen SE. Comparison of digital and cine coronaryarteriography. Int J Card Imaging 1990;5:105–18.

9. Lau J, Ioannidis JP, Schmid CH. Summing up evidence: one answeris not always enough. Lancet 1998;351:123–7.

0. Leber AW, Knez A, Becker C, et al. Non-invasive intravenouscoronary angiography using electron beam tomography and multislicecomputed tomography. Heart 2003;89:633–9.

1. Kuettner A, Kopp AF, Schroeder S, et al. Diagnostic accuracy ofmultidetector computed tomography coronary angiography in patientswith angiographically proven coronary artery disease. J Am CollCardiol 2004;43:831–9.

2. Mollet NR, Cademartiri F, Nieman K, et al. Multislice spiralcomputed tomography coronary angiography in patients with stableangina pectoris. J Am Coll Cardiol 2004;43:2265–70.

3. Bornert P, Aldefeld B, Nehrke K. Improved 3D spiral imaging forcoronary MR angiography. Magn Reson Med 2001;45:172–5.

4. Stuber M, Botnar RM, Spuentrup E, Kissinger KV, Manning WJ.Three-dimensional high-resolution fast spin-echo coronary magneticresonance angiography. Magn Reson Med 2001;45:206–11.

5. Stuber M, Botnar RM, Fischer SE, et al. Preliminary report on in vivocoronary MRA at 3 Tesla in humans. Magn Reson Med 2002;48:425–9.

6. Botnar RM, Kim WY, Bornert P, Stuber M, Spuentrup E, ManningWJ. 3D coronary vessel wall imaging utilizing a local inversiontechnique with spiral image acquisition. Magn Reson Med 2001;46:848–54.

7. Botnar RM, Stuber M, Kissinger KV, Kim WY, Spuentrup E,Manning WJ. Noninvasive coronary vessel wall and plaque imagingwith magnetic resonance imaging. Circulation 2000;102:2582–7.

8. Fayad ZA, Fuster V, Fallon JT, et al. Noninvasive in vivo humancoronary artery lumen and wall imaging using black-blood magneticresonance imaging. Circulation 2000;102:506–10.

9. Budoff MJ, Achenbach S, Duerinckx A. Clinical utility of computedtomography and magnetic resonance techniques for noninvasive cor-onary angiography. J Am Coll Cardiol 2003;42:1867–78.

0. Ropers D, Baum U, Pohle K, et al. Detection of coronary arterystenoses with thin-slice multi-detector row spiral computed tomogra-
phy and multiplanar reconstruction. Circulation 2003;107:664–6.

Documents

Diagnostic performance of coronary magnetic resonance angiography as compared against conventional x-ray angiography: A meta-analysis