Author + information
- Received February 1, 2009
- Revision received April 24, 2009
- Accepted May 21, 2009
- Published online November 1, 2009.
- Peter Bernhardt, MD⁎,†,⁎ (, )
- Jochen Spiess, MD⁎,
- Benny Levenson, MD‡,
- Günter Pilz, MD†,
- Berthold Höfling, MD†,
- Vinzenz Hombach, MD⁎ and
- Oliver Strohm, MD§
Reprint requests and correspondence:
Dr. Peter Bernhardt, Department of Internal Medicine II, University of Ulm, 89077 Ulm, Germany
Objectives We sought to assess the accuracy of an integrated cardiac magnetic resonance (CMR) protocol for the diagnosis of relevant coronary artery or bypass graft stenosis in patients with suspected coronary artery disease (CAD) or with previously performed percutaneous coronary intervention (PCI) or coronary bypass graft surgery (CABG).
Background CMR is suitable for diagnosing inducible myocardial ischemia in patients with suspected CAD and has been proven to be a helpful diagnostic tool for decision of further treatment. However, little is known about its diagnostic accuracy in patients with known CAD who previously were treated by PCI or CABG.
Methods A total of 477 patients with suspected CAD, 236 with previous PCI, and 110 after CABG referred for coronary X-ray angiography (CXA) underwent an integrated CMR examination before CXA. Myocardial ischemia was assessed using first-pass perfusion after vasodilator stress with adenosine (140 μg/kg/min for 3 min) using gadolinium-based contrast agents (0.1 mmol/kg). Late gadolinium enhancement (LGE) was assessed 10 min after a second contrast bolus.
Results CXA demonstrated a relevant coronary vessel stenosis (≥70% luminal reduction) in 313 (38%) patients using quantitative coronary analysis. The combination of CMR perfusion and LGE assessment for detecting a relevant coronary stenosis in patients with suspected CAD yielded sensitivity and specificity of 0.94 and 0.87, in PCI patients 0.91 and 0.90, and in CABG patients 0.79 and 0.77, respectively.
Conclusions A combined CMR protocol for the assessment of myocardial perfusion and LGE is feasible for the detection of relevant coronary vessel stenosis even in patients who previously were treated by PCI or CAG in a routine clinical setting. However, diagnostic accuracy is reduced in patients with CABG. This could be due to different flow and perfusion kinetic. Further studies are needed to optimize the clinical protocols especially in post-surgical patients.
Decision of necessity of reangiography and intervention in patients with known coronary artery disease (CAD) and previous percutaneous coronary artery intervention (PCI) or coronary artery bypass graft (CABG) partly depends on the detection of myocardial ischemia and sometimes remains challenging (1). Cardiac magnetic resonance (CMR) imaging has been shown to be a safe and feasible imaging method (2) to assess myocardial ischemia with high accuracy using vasodilator and/or positive inotropic stressors as adenosine and dobutamine (3–9). Myocardial viability can be assessed noninvasively by late gadolinium enhancement (LGE) techniques with high accuracy (10–12). Recently, it has been shown that the combination of perfusion and LGE assessment increases accuracy of detecting a relevant coronary stenosis in patients with suspected, but not previously documented CAD (13,14).
However, most of the published studies have been performed in patients with suspected CAD. There is little knowledge about the ability of CMR to detect relevant coronary vessel stenosis or occlusion in patients who previously underwent PCI or CABG.
The primary aim of our study was to assess the ability of a combined CMR protocol including adenosine stress perfusion and LGE imaging to identify patients with previous PCI or CABG with a relevant coronary vessel stenosis and to compare the diagnostic performance to patients without previous PCI or CABG.
From May 2004 to October 2007, 823 patients with a clinical indication for X-ray coronary angiography (CXA) were included into the trial in three centers. Exclusion criteria were contraindication for vasodilatory stress-CMR including unstable symptoms, implanted devices, severe obstructive lung disease, and severe valvular disease. All patients were taken off vasodilator drugs and caffeine 24 h prior to CMR, and all patients gave informed consent.
CMR imaging protocol
Imaging was performed using 1.5-T whole-body clinical magnetic resonance scanners (Intera, Philips Medical Systems, Best, the Netherlands, using a cardiac 5-element phased-array receiver coil [University of Ulm]; CV/i, GE Medical System, Milwaukee, Wisconsin, using a cardiac 4-element phased-array receiver coil [St.-Gertrauden-Hospital, Berlin], and Signa Excite, GE Medical System, using a cardiac 8-element phased-array receiver coil [Hospital Agatharied]). Heart rate, blood pressure, and oxygen saturation were monitored noninvasively during the examination.
Functional imaging of the left ventricle was done in 3 long-axis (2-, 3-, and 4-chamber view) orientation and in contiguous short-axis orientation to cover the left ventricle from the base to the apex (steady-state free precession, repetition time [TR]: 5.1 ms, echo time [TE]: 2.2 ms, field of view: 32 to 34 × 32 to 34 cm, matrix: 256 × 192, slice thickness: 8 mm, no interslice gap; acquisition in end-expirational hold).
After 3 min of adenosine infusion at a constant rate of 140 μg/kg/min, 0.1 mmol/kg gadolinium-based contrast agent (Omniscan, GE Medical Systems, Munich, Germany, or Magnevist, Bayer Schering Pharma, Berlin, Germany) was injected intravenously during a first-pass perfusion sequence as previously described (2,3,15). A hybrid echo-planar pulse sequence (TR: 6.4 to 16 ms, TE: 1.2 to 1.4 ms, flip angle: 25°, slice thickness: 10 mm, voxel size: 2.8 × 2.8 mm, 3 to 5 slices acquired at every heart beat) was used on the GE scanners as previously described (4). On the Philips scanner, a balanced fast-field echo sequence (TR: 2.6 ms, TE: 1.3 ms, saturate pre-pulse with 100-ms delay, flip angle: 50°, 40 dynamics, turbo field echo factor: 48, voxel size: 2.8 × 2.9 mm, slice thickness: 10 mm, 3 to 4 slices acquired at every heart beat) was used as previously described (5). Perfusion images were acquired in short-axis orientation using a variable gap to cover the entire left ventricle from base to apex. Breathing commands were trained with the patients prior to the study to minimize breathing-related artifacts.
Ten minutes after a second contrast agent bolus (0.1 mmol/kg), inversion-recovery gradient-echo sequences (LGE) were acquired in the same orientation as the perfusion sequence (TR: 7.1 ms, TE: 3.2 ms, voxel size: 1.2 × 1.2 mm, slice thickness: 10 mm on the GE scanners; TR: 5.6 ms, TE: 2.1 ms, voxel size: 1.2 × 1.2 mm, slice thickness: 10 mm on the Philips scanner). Additionally, a 3D LGE sequence with full coverage of the left ventricle was performed in breath hold (TR: 7.1 ms, TE: 3.2 ms, voxel size: 1.2 × 2.2 mm, slice thickness: 8 mm). TI was individually adjusted for complete nulling of the myocardium.
All CMR studies were evaluated by 1 reader of each center locally (J.S., G.P., O.S.) and a second reader for all centers (P.B.). Investigators were blinded to quantitative coronary analysis results and to patients' history and clinical symptoms.
Functional images were analyzed for left ventricular end-diastolic and end-systolic volumes by planimetry using the multislice approach. Functional images were analyzed to obtain left ventricular functional parameters as left ventricular end-diastolic and end-systolic volume, and left ventricular ejection fraction.
First-pass perfusion and LGE images were assessed visually. A relevant perfusion deficit was defined as a subendocardially beginning hypoenhancement in at least 2 adjacent segments or slices with no attributes for imaging artifacts as previously described (16). Diffuse subendocardial perfusion defects not assignable to a specific coronary supply territory were defined as unspecific; dark rim artifact (17) was not regarded as perfusion deficit using previously described criteria (18,19).
Analysis of first-pass perfusion and LGE was performed using 2 different algorithms (Fig. 1). One algorithm was similar to the one described by Klem et al. (13). The second algorithm aims to consider chronic myocardial infarction in territories supplied by previously revascularized coronary arteries. In this case, we hypothesized that presence of LGE was not considered as presence of clinically relevant stenosis of the supplying artery. Relevant myocardial ischemia was defined as mismatch between hypoenhancement on first-pass perfusion and enhancement on LGE sequences and classified as reversible ischemia. A match between first-pass deficit and LGE was considered as chronic infarction with no additional reversible ischemia. Figure 2 shows examples of stress perfusion as well as of LGE.
Areas of ischemia and areas of LGE were assigned to the presumably supplying coronary artery using the 17-seqment model with exclusion of segment 17 (apex), as this was not visualized on the first-pass perfusion images (20).
The CMR data were analyzed for accuracy of detecting patients with relevant coronary artery or bypass graft stenosis as well as detecting.
All coronary catheterizations and angiographies were performed as recommended by the American College of Cardiology and American Heart Association (1). Blinded analysis of invasive data, including degree of luminal narrowing of the coronary artery, was performed by independent investigators of the respective centre (J.S., B.L., B.H.). Quantitative coronary analysis was performed in all patients therefore. A stenosis ≥70% in a coronary artery or in a bypass graft with a diameter ≥2 mm was considered significant. A stenotic native vessel with corresponding nonstenotic bypass graft was not considered for ischemia.
All statistical analysis was performed with commercially available statistic software (StatView 5, SAS Institute, Cary, North Carolina). Data are reported as the mean ± standard deviation. Continuous variables between groups were compared using the t-test for unpaired observations. Nominal data were compared using the Fisher exact test. Ordinal variables were compared using Wilcoxon signed rank test for matched pairs. The McNemar test was used for comparison of the 2 analysis algorithms. In all cases, a p value < 0.05 was considered statistically significant.
A total of 876 patients were screened for inclusion into the study. Fifty-three patients were not included due to claustrophobia (n = 27), contraindication for adenosine (n = 21) or refusal to give consent (n = 5). Thus, 823 patients (629 [76%] male) formed the study population. 477 (58%) patients with suspected CAD, 236 (29%) with previous PCI, and 110 (13%) with previous CABG formed the different study groups. CMR studies were performed 8 ± 5 days before CXA.
Clinical indication for CXA was relevant angina (74.4%), positive stress test (other than CMR) (31%), and/or other abnormalities (8.1%), as, e.g., dyspnea, considered as angina equivalent, new left bundle brunch block, or new nonsustained ventricular tachycardia.
Mean time between PCI or CABG and CMR examination was 314 ± 231 days and 423 ± 275 days, respectively. Patients' characteristics are given in Table 1.
No major or minor complications occurred; all patients could complete the entire protocol. Transient AV-block was observed in 81 (9%) patients, but did not require specific treatment or cessation of the study.
Quality of perfusion images was sufficient for further analysis in all patients. In 138 (16.8%) patients, noncritical artifacts could be seen on perfusion imaging. These artifacts were graded to be motion artifacts due to breathing in 51 cases, dark rim artifacts in 49 cases, and ghosting artifacts in 38 cases. A total of 185 (0.9%) segments had to be excluded from visual analysis due to severe artifacts. However, no study had to be excluded due to insufficient image quality. In 23 (2.8%) patients, image quality of the LGE sequences was insufficient for analysis.
CMR and diagnostic performance
Volumetric data of the left ventricle are given in Table 1. Patient groups did not differ significantly for ventricle functional parameters. Patients with and without coronary stenosis did not differ for these parameters, either.
Perfusion deficits were detected in 417 (50.7%) patients in the first-pass perfusion. Two hundred of 477 (41.9%) patients with perfusion deficits during adenosine-stress perfusion were found in the patient group with suspected CAD, 129 of 236 (54.7%) in the PCI and 88 of 110 (80.0%) in the CABG group. LGE could be visualized in 312 (37.9%) patients with a transmural extent in 152 patients; 262 of these had a history of myocardial infarction.
CXA revealed relevant coronary artery stenosis or chronic occlusion in 173 (36%) patients with suspected CAD, in 69 (29%) PCI, and in 71 (65%) CABG patients.
Algorithm A performed best in patients with suspected CAD and yielded a sensitivity of 0.93, a specificity of 0.87, a positive predictive value of 0.81, a negative predictive value of 0.96, and an overall accuracy of 0.89. In patients with PCI and CABG, we observed better CMR interpretation, with algorithm B resulting in a sensitivity of 0.88 and 0.73, a specificity of 0.90 and 0.77, a positive predictive value of 0.79 and 0.85, a negative predictive value of 0.95 and 0.61, and an overall accuracy of 0.91 and 0.75, respectively.
Diagnostic performance of CMR with both analysis algorithms for all patient groups and subanalysis for 1- or multiple-vessel disease is given in Table 2. Furthermore, diagnostic performance subanalysis of both CMR algorithms in all patient groups for the left anterior descending, left circumflex, and right coronary artery perfusion territories as well as for arterial, venous, and mixed bypass grafts in the CABG group is provided in Table 2.
Patients with coronary artery or bypass graft stenosis ≥70% had significantly more segments with LGE (p < 0.0001) and had significantly more often transmural extent than patients without coronary artery stenosis (p = 0.0074). There were no differences in age, sex, cardiovascular risk factors, or left ventricular functional parameters between patients with and without angiographically detected stenosis. Comparison of patients with and without coronary artery stenosis as detected by CXA is given in Table 3.
In 3 of 9 (33%) patients with false positive CMR interpretation, the perfusion deficit observed affected only basal myocardial segments. In 8 of 19 (42%) patients with false negative CMR results, we found perfusion deficit and LGE match, thus this was interpreted as nonreversible ischemia in chronic myocardial ischemia.
The primary finding of our study is that an integrated CMR examination using first-pass perfusion and LGE is a feasible method to detect myocardial ischemia and irreversible myocardial injuries in patients with previous coronary artery intervention or CABG with high sensitivity, specificity and overall accuracy.
Comparing the amount of inducible ischemia (assessed by vasodilator stress first-pass perfusion) and the amount of irreversible ischemic damage (as assessed by LGE in the same image orientation) allows predicting a relevant coronary artery stenosis that would benefit from revascularization. A “match” between the amount of inducible ischemia and LGE is considered a completed myocardial infarct and thus would not require revascularization of the supplying vessel. To account for the presence of irreversible injury in segments with inducible ischemia, we developed an analysis algorithm and correlated the predictive values to invasive coronary angiography.
Only the visualization of inducible ischemia in non-LGE myocardium would predict a clinically relevant coronary vessel stenosis and mandate for revascularization. We therefore compiled our second analysis algorithm to identify patients with inducible myocardial ischemia and chronic myocardial infarction. The better performance of this algorithm strengthens the thesis about necessity of an analysis and interpretation algorithm that considers chronic myocardial infarction.
Assessment of a relevant restenosis or new stenosis in an artery after intervention has been described using different diagnostic tools. A recent meta-analysis proved exercise treadmill testing to be a rather poor diagnostic test to identify the presence of graft stenosis, yielding a sensitivity of 45% and a specificity of 82% (21–24). Dobutamine stress echocardiography with contrast-enhanced perfusion imaging has been shown to have a sensitivity of 73% and a specificity of 75% (25) for the evaluation of coronary artery stenosis. Nuclear medicine techniques have also been evaluated for the detection of restenosis in patients after PCI. Single-photon emission computed tomography (SPECT) imaging has been shown to have a sensitivity of 93% and a specificity of 77% for detection of restenosis in patients after PCI (23); however, SPECT is known to suffer from attenuation artifacts (26) and exposes patients to unwanted radiation.
CMR offers the ability to assess different aspects of CAD in a combined protocol; it has been shown to quantify ventricular functional parameters (27,28), assess first-pass perfusion during stress (2–8,13–15), and allow for noninvasive assessment of myocardial viability (10–15,20,29) in a single study. Recent studies have shown that CMR provides a high accuracy in detecting myocardial perfusion defects in patients with suspected CAD (3–6,8,13,14).
We could demonstrate that patients after bypass can be assessed with CMR, but with a lower diagnostic yield. Delayed contrast appearance in myocardial tissue during stress perfusion study could have different etiologies, such as post-stenotic pressure reduction in the supplying coronary artery, altered coronary artery anatomy with changed flow properties, slower flow and/or longer transit time of the contrast bolus through bypass grafts. A comparison between radial artery grafts and longer saphenous venous bypass grafts found significant different flow characteristics and flow response (29). These data strengthen the hypothesis of different flow characteristics in different bypass types that could possibly cause different myocardial perfusion (30) and thus be a limitation in the correct assessment of graft patency.
This may be a universal limitation for all stress methods after bypass grafts and may mandate a baseline-stress study immediately after surgery. Supporting this theory is a study by Fenchel et al. (31) that provided information on the success of interventional procedures in 18 patients performing stress-CMR as a baseline after intervention. The lack of a “baseline” study could explain the lower specificity of CMR performance in CABG patients in our study.
The presence of significant collateral flow in patients with multivessel CAD could account for the lower sensitivity in our patient group after CAPB; these patients had a higher rate of multiple-vessel disease and presumably more often developed system of collateral flow. A study by Becker et al. (32) showed that stenoses of the coronary vessels supplying collateral flow to ischemic regions are necessary for the occurrence of coronary steal phenomenon. Another recent study detected coronary steal in only one-third of patients with chronic total coronary occlusion (33). These findings strengthen the higher rate of false negative findings in our CABG group and thus could explain the lower sensitivity of CMR performance.
In 33% of our false positive CMR interpretations, we observed myocardial ischemia affecting only the basal myocardial segments. A possible explanation could be a distal anastomosis of the bypass graft, restricting flow to the basal segments in case of a proximal stenosis or occlusion. Thus, in case of proximal stenosis or occlusion of the native coronary artery, a perfusion of the basal segments could be restricted.
The patient groups included in the study had a high pre-test probability for the presence of CAD, as they all had a clinical indication for invasive coronary angiography. Thus, the results of our study may not be easily transferrable to the general population or to patients with low or medium pre-test probability.
The use of different CMR systems and perfusion sequences in the 3 centers could be a possible limitation to our study. However, we did not observe significant differences regarding CMR performance between our centers.
Another possible limitation is the presence of artifacts in 16.8% patients. However, less than 1% of stress perfusion images and less than 3% of LGE images had to be excluded from analysis due to insufficient image quality.
Our data show that CMR is suitable for noninvasive detection of relevant coronary artery stenosis in patients with suspected CAD, including those with previous PCI or CABG. The combined assessment of inducible ischemia and reversible ischemia in an integrated CMR protocol allows predicting the presence of a clinically relevant coronary artery and/or bypass graft stenosis. In patients with previous CABG, sensitivity, specificity, and accuracy of CMR is reduced in comparison to the other patient groups. Further studies are warranted to determine whether a quantitative assessment in this patient group could improve the accuracy of the test.
Dr. Strohm is a shareholder in and consultant for Circle Cardiovascular Imaging Inc., Calgary, Canada.
- Abbreviations and Acronyms
- coronary artery bypass graft
- coronary artery disease
- cardiac magnetic resonance
- coronary X-ray angiography
- late gadolinium enhancement
- percutaneous coronary artery intervention
- Received February 1, 2009.
- Revision received April 24, 2009.
- Accepted May 21, 2009.
- American College of Cardiology Foundation
- Scanion P.J.,
- Faxon D.P.,
- Audet A.M.,
- et al.,
- ACC/AHA guidelines for coronary angiography
- Giang T.H.,
- Nanz D.,
- Coulden R.,
- et al.
- Schwitter J.,
- Nanz D.,
- Kneifel S.,
- et al.
- Nagel E.,
- Klein C.,
- Paetsch I.,
- et al.
- Lee D.C.,
- Simonetti O.P.,
- Harris K.R.,
- et al.
- Kwong R.Y.,
- Schussheim A.E.,
- Rekhraj S.,
- et al.
- Klein C.,
- Nekolla S.G.,
- Bengel F.M.,
- et al.
- Klem I.,
- Heitner J.F.,
- Shah D.J.,
- et al.
- Cerqueria M.D.,
- Weissman N.J.,
- Dilsizian V.,
- et al.
- Honan M.B.,
- Bengston J.R.,
- Pryor D.B.,
- et al.
- Elhendy A.,
- Tsutsui J.M.,
- O'Leary E.L.,
- Xie F.,
- Majeed F.,
- Porter T.R.
- Hendel R.C.,
- Berman D.S.,
- Cullom S.J.,
- et al.
- Bellenger N.G.,
- Burgess M.I.,
- Ray S.G.,
- et al.
- Kim R.J.,
- Fieno D.S.,
- Parrish T.B.,
- et al.
- Becker L.C.
- Werner G.S.,
- Figulla H.R.