Background
Early detection of coronary artery disease (CAD) and in particular of myocardial ischemia remains a major challenge even with the advent of novel non-invasive imaging techniques and further development of existing modalities. An increasing number of cardiovascular magnetic resonance (CMR) studies documented a high diagnostic performance of perfusion-CMR vs conventional x-ray coronary angiography (CXA) [
1‐
9] and showed its prognostic value [
10,
11]. In comparison with CXA, for both, perfusion-CMR [
12,
13] as well as for SPECT, cost-effectiveness was demonstrated [
14,
15]. However, for several sub-groups of patients the diagnostic performance of perfusion-CMR and its potential superiority over SPECT is not well established. The first study of the MR-IMPACT program [
2] designed for dose-finding was the largest perfusion-CMR trial at its time and demonstrated equal performance vs SPECT in the head-to-head comparison, and demonstrated superiority of CMR when compared versus the entire SPECT population. The MR-IMPACT II was designed to compare the diagnostic performance of CMR vs SPECT for the detection of CAD (defined as ≥50% diameter reduction of coronary vessels in CXA) in a large international multicenter, multivendor design at a fixed contrast medium (CM) dose. The primary end-point of MR-IMPACT II was the comparison of sensitivity and specificity of perfusion-CMR to detect CAD on CXA vs SPECT based on a single-point threshold reading. In this comparison, perfusion-CMR was more sensitive, but less specific for the detection of CAD in comparison with SPECT [
16]. This single-threshold reading assesses diagnostic performance on a single point on the ROC curve, thus, rendering results susceptible for the reading threshold [
17]. The comparison of test performances by means of the areas under the ROC curves (AUC) avoids such potential bias [
17]. Therefore, we analyzed as a pre-defined secondary end-point of the MR-IMPACT II the AUCs for perfusion-CMR and SPECT for the entire study population. Additional sub-group analyses assessed the diagnostic performance in patients studied by gated-SPECT, in patients without prior myocardial infarctions (MI) with single- or multi-vessel disease, as well as in men and in women. In addition, the primary end-point was also recalculated for a single-point reading at the optimum threshold as derived from the AUC analyses.
Discussion
The main results of the trial can be summarized as follows: 1). The diagnostic performance of perfusion-CMR assessed as the area under the ROC curve was superior over SPECT in detecting CAD when assessed in the entire study population, and 2). Perfusion-CMR was also superior (higher AUC) for CAD detection in the sub-groups analyzed such as in patients with gated-SPECT and non-gated-SPECT, in patients with multi-vessel disease, in men as well as in women, and in the patients without prior MI.
Perfusion-CMR and SPECT comparison
The current MR-IMPACT II results are well in line with a previous perfusion-CMR multicenter study [
3]. The mean AUC for the CM doses of 0.05 and 0.1 mmol/kg in a previous smaller multicenter study was 0.79 which is close to the 0.75 in MR-IMPACT II. Diagnostic performance in the MR-IMPACT II with an AUC of 0.75 was slightly lower than in MR-IMPACT I with an AUC of 0.86. This might be related to the larger number of participating sites in MR-IMPACT II, by which less experienced centers could have contributed to the data. Also, in MR-IMPACT II a slightly lower CM dose of 0.075 mmol/kg was used than the most effective dose in MR-IMPACT I of 0.10 mmol/kg (for regulatory purposes the lowest efficacious dose was tested). With 33 participating sites in US and Europe, this MR-IMPACT II is to our knowledge the CMR study on myocardial ischemia with the highest number of contributing centers conducted so far. Since these centers participated with various CMR systems and drop-out rate was kept as low as 5.6%, the study results are assumed to adequately reflect the performance and robustness of perfusion-CMR in the day-to-day clinical setting. Results of the so far largest single center perfusion-CMR trial called CE-MARC were published recently [
9]. In CE-MARC 752 patients were recruited to undergo 3 tests, i.e. a stress perfusion-CMR, a gated-SPECT, and CXA. In the 628 patients who completed all three tests, diagnostic performance of CMR was superior to SPECT to detect >50% diameter stenosis on CXA with AUCs of 0.84 versus 0.69 (p < 0.0001). Of note, in CE-MARC performances for both, perfusion-CMR and gated-SPECT were similar to those obtained in the current multicenter MR-IMPACT II. These results of the MR-IMPACT II underline the strengths of the perfusion-CMR technique. Correction of cardiac and breathing motion with ECG-triggering and breath-holding, respectively, appears to reliably minimize motion-related artifacts as less than 6% of data had to be excluded from analyses due to inadequate quality. This approach preserves the nominally high spatial resolution of perfusion-CMR and thereby allows detecting small even subendocardial perfusion deficits. Moreover, as with any MR acquisition, the MR perfusion images are not compromised by signal attenuation artifacts which could be perceived as perfusion deficits.
SPECT imaging also proved to be robust over the 33 centers and the SPECT results of this trial are in close match which those of MR-IMPACT I with AUC of 0.65 vs 0.67, respectively. The SPECT results are also well in line with CE-MARC and other previous multicenter SPECT trials [
18,
21‐
23] with sensitivities (ranging from 77% to 87% vs CXA) and specificities (ranging from 36% to 58% vs CXA) being located very closely to the ROC curve of the current MR-IMPACT II. Of note, also higher specificities for the SPECT technique were reported in single center studies.
When comparing the AUCs for perfusion-CMR and SPECT, superiority of CMR was achieved versus the entire SPECT population (p = 0.0004, n = 425) as well as for the gated-SPECT population (p = 0.018, n = 253). When assessing the diagnostic performances (=AUCs) of both, CMR and SPECT, it should be kept in mind that the readings were performed in a fully blinded fashion without integrating any clinical information of the patients which is likely to result in underestimation of the clinical diagnostic performance of these imaging techniques.
As patient prognosis is coupled to the extent of ischemia, it is important to note, that the diagnostic superiority of perfusion-CMR was also preserved in the MVD patients as shown in Figure
2B.
In the study by Lin and coworkers [
24], women were less likely to undergo stress testing prior to PCI compared with men and other studies demonstrated, that younger women had a worse outcome after acute MI compared to age-matched men [
25]. These findings go along with particular difficulties in women for ischemia detection, as breast tissue can cause suboptimal imaging conditions. Also, women are more susceptible to radiation sequelae in terms of cancer incidence [
26,
27]. Therefore, a radiation-free test for ischemia detection in women is particularly valuable. The results of MR-IMPACT II demonstrate equal diagnostic performance of CMR irrespective of gender, which was superior to SPECT in both, men and women (Figure
3A/B). While multicenter data for SPECT in women are rare, one study [
21] yielded 95%-confidence intervals for sensitivity and specificity of 69-100% and 10-61%, respectively, which fall onto the SPECT-ROC curve for women of the current study. Similarly, a single center study using thallium SPECT found a lower diagnostic performance in women versus men, which was related to smaller heart sizes in women [
28]. As spatial resolution of perfusion-CMR is higher than in SPECT and attenuation artifacts do not occur, it is not surprising that two earlier single center studies found a similar performance of CMR in women and men [
29,
30], which is well in line with the current multicenter findings.
For CMR image interpretation, only the stress perfusion images were used, but the readers were exposed to the viability images obtained by applying the late enhancement CMR technique. To test, whether viability information could have played a major role in the recognition of ischemia, a sub-analysis was performed which excluded the patients with prior MI. This reduced the sample size to 174, but the difference in AUC persisted in favor of CMR (with a p-value of 0.0054).
Limitations of the study
The definition of CAD applied in this study was primarily dependent on coronary anatomy (in most patients by fulfilling the criterion of a ≥50% diameter stenosis in a ≥2 mm vessel and in a minority of patients by the history of MI, even when the infarct-related vessel was non-stenosed after PCI). This anatomy-based definition does not consider collateral flow and microcirculatory factors that modify the hemodynamic relevance of epicardial stenoses. Nevertheless, this definition was deemed best as it is relatively easy to measure, is frequently used in such comparative studies, and it often sets the basis for patient management in clinical routine. In this context is should be noted that an optimal patient management should always consider the patient’s symptoms, his risk factor profile, and the prognosis (derived from risk factors, symptoms, and imaging information).
The aim of this study was to assess the diagnostic performance of CMR and SPECT to identify patients with CAD through the detection of perfusion abnormalities. Once CAD is identified, a further evaluation is recommended to assess scar tissue. As the late enhancement CMR technique is well documented as a robust and precise method to detect scar [
31,
32], this additional work-up was not tested in this trial. We would like to stress, however, that perfusion testing should be accompanied in general by a viability testing, particularly in patient with reduced LV function, to allow for an optimal patient management. To explore a potential influence of scar tissue on the study results, the ROC analysis was repeated in the patient population without prior MI and superiority of CMR over SPECT was preserved (p = 0.0054).
In this trial patients with decompensated heart failure, after bypass surgery, and with relevant arrhythmias were excluded, and thus, the findings of this study should not be applied to these patient groups. Due to the inclusion criteria, the frequency of CAD was relatively high with 48.8% in this trial. While this rather rigorous recruitment strategy represents a strength of the trial, the trial results cannot be extrapolated to other populations with lower disease prevalence, e.g. to asymptomatic screening populations.
While this study shows that perfusion-CMR is useful for the detection of CAD, further studies will be needed to address the question, whether this perfusion information when used to guide revascularizations also improves outcome in comparison to other non-invasive methods that test ischemia.
List of participating sites (in alphabetical order)
H. Ahlstroehm, Akademiske Sjukhuset, Uppsala, Sweden; N. Al-Saadi, Universitätsklinikum Charité, Berlin, Germany; J. Barkhausen, Elisabeth Krankenhaus, Essen, Germany; D.S. Berman, Cedars-Sinai Medical Center, Los Angeles, CA, USA; J. C. Carr, NW University Medical Center, Chicago, IL, USA; T. Dill, Kerckhoff-Klinik Bad Nauheim, Germany; S. Flamm, St. Luke’s Episcopal Hospital, Houston, TX, USA; H. Frank, Donauklinikum Landeskrankenhaus Tulln, Austria; T. Fuisz, Washington Hospital Center, Washington, DC, USA; D. Hahn, Universität Wuerzburg, Germany; K. Huettl, Semmelweis University, Budapest, Hungary; C.M. Kramer, University of Virginia Health System, Charlottesville, VA, USA; H. Kuehl, Universitätsklinikum Aachen, Germany G. Layer, Klinikum der Stadt Ludwigshafen, Germany; M. Lombardi, Istituto di Fisiologia Clinica, CNR, Pisa, Italy; A. Luchner, Klinikum der Universität Regensburg, Germany; E.T. Martin, Oklahoma Heart Institute, Tulsa, OK, USA; G. P. Meyer, Medizinische Hochschule Hannover, Germany; A. Mosterd, Meander Medisch Centrum, Amersfoort, The Netherlands; E. Mousseaux, European Hospital G. Pompidou, Paris, France; S. Orn, Stavanger Hospital Trust, Stavanger, Norway; R.M. Peshock, SW Medical Center at Dallas, Dallas, TX, USA; G. Raff, William Beaumont Hospital, Royal Oak, MI, USA; N. Reichek, St. Francis Hospital, Roslyn, NY, USA; B. Rensing, SA Ziekenhuis, Nieuwegein, The Netherlands; E. Sauer, Krankenhaus Landshut, Germany; S. Schoenberg, Klinikum der Universität Muenchen-Grosshadern, Germany; J. Schwitter, University Hospital Zurich, Switzerland; U. Sechtem, Robert-Bosch Krankenhaus, Stuttgart, Germany; T. Simor, University of Science, Pecs, Hungary; O. Strohm, St. Getrauden Hospital, Berlin, Germany; A. van Rossum, VU Medical Center, Amsterdam, The Netherlands; C. Wacker, Medizinische Universitätsklinik Wuerzburg, Germany; N. Wilke, University of Florida Jacksonville, Jacksonville and Gainsville, FL, USA; P. Woodard, Mallinckrodt Institute of Radiology, St. Louis, MO, USA.
Competing interests
J. Schwitter, MD and N. Al-Saadi, MD served as consultants for GE Healthcare (former Amersham Health) and received honoraria. O. Strohm, MD, is a consultant of Circle (a software company not involved in data analysis in this trial). N. Hoebel, MSc, is an employee of GE Healthcare and was responsible for the statistical analyses.
Authors’ contributions
JS: is responsible for the conception and design of the study, the data acquisition and interpretation, and he drafted the manuscript. CMW: contributed to the acquisition and interpretation of data and critically revised the intellectual content of the draft. NW: contributed to the collection and assembly of data and critically revised the intellectual content of the draft. NA: contributed to conception and design of the study, the acquisition and interpretation of data, and he critically revised the intellectual content of the draft. ES, KH, SOS, KD, OS, HA, TD: contributed to the collection of data and critically revised the intellectual content of the draft. NH: contributed to conception and design of the study, she performed the statistical analyses, contributed to data interpretation, and she critically revised the intellectual content of the draft. TS: contributed to conception and design of the study, the acquisition and interpretation of data, and he critically revised the intellectual content of the draft. All authors read and approved the final manuscript.