Diagnostic performance of stress perfusion CMR for CAD evaluation is presently limited primarily by a lower specificity (FP results). In a large meta-analysis on 24 studies using perfusion CMR imaging mean specificity was 81%. Our study on 206 consecutive patients yielded a PPV of (80.1%), despite adding rest to stress perfusion imaging for recognition of artefacts as proposed by Thomson et al. [11]. Hence our population was representative for current CMR performance.

To the best of our knowledge, the present study is the first to describe an interrelationship between small caliber coronary vessels and CMR visualization of hypoperfusion in the absence of coronary stenosis (FP results regarding CAD). As the main finding of our study, we observed a correlation between FP CMR perfusion results and the presence of a small caliber coronary vessel supplying the area of ischemia in CMR, opposite to the dominant vessel. Of note, in the subgroup of patients with other stress tests performed prior to CMR these exams had similarly yielded pathological findings indicative of ischemia.

Our analysis showed the presence of such a small caliber vessel to be a possible predictive parameter for discrimination of false versus true positive. By adding this information to CMR perfusion, diagnostic performance improved by potentially avoiding up to half of FP CMR results. ROC curve analysis of a combined approach (CMR perfusion and proximal coronary diameter) yielded a significantly better result compared to CMR perfusion alone. This concept is strengthened by the further increment in CMR diagnostic accuracy achieved when additionally including the ratios of ipsilateral/contralateral coronary diameter for FP versus TP discrimination. In contrast, standard CMR function and perfusion criteria were not discriminatory (LV ejection fraction, LV enddiastolic volume, LV mass, LV wall stress, transmural extent of ischemia and temporal persistence of ischemia).

Anatomic studies have shown that while the irrigation of the sternocostal surface of the heart is extremely regular, the inferior or diaphragmatic surface supplied by the circumflex and right coronary arteries is highly variable and gives each heart its own “distinctive physiognomy” [15]. In this respect, the presence of small caliber coronary arteries has been described early. Concerning perfusion physiology, a study in healthy humans found heterogeneity of resting and hyperaemic myocardial blood flow with a broad range of regional distribution [16].

Although our description of an impact of small caliber coronary arteries for stress perfusion CMR evaluation may be considered disputable, the importance of coronary variability is well known in other cardiovascular imaging conditions. Most recently, Ortiz-Pérez et al. [17] have shown the impact of variations in coronary anatomy on the assignment of LGE viability assessment by CMR imaging to the specific coronary territory. Their study highlighted that in a per-segment analysis, only 76% of segments followed the distribution of coronary arteries empirically attributed and the discordance reached 67% of cases in a per-patient analysis. In the case of nuclear cardiac imaging, Segall et al. [9] concluded that the accurate assessment of coronary stenoses by thallium SPECT imaging requires close correlation with arteriography owing to the significant variability in normal coronary anatomy. A main finding of their study was the great variability in blood supply to the inferior and lateral wall, but also in the apex, which is one of the most variable regions with regard to myocardial perfusion [9]. The impact of coronary variability on ECG findings in acute myocardial infarction was first analyzed by Birnbaum et al. Thus, the occlusion in the same site of a coronary artery in different patients may result in a different size and location of the ischemic area at risk and hence the ECG pattern may not be the same in all cases with similar coronary artery occlusion [18]. These findings were confirmed by means of direct CMR scar visualization (LGE) by Rovai et al. [19].

The coronary distribution and calibers found in our study are in accordance with the literature. Thus, proximal coronary artery calibers measured in our study are in the magnitude of previous studies [14]. Similarly, the higher incidence of right dominant pattern in our study confirms former data [15].

Given the increasing demand for non-invasive assessment of suspected coronary artery disease [20], a method with higher sensitivity and specificity compared to the present approaches is strongly needed. A recent study using myocardial perfusion scintigraphy showed that a non-invasive “gatekeeper” approach could make about half of catheterizations redundant [21]. Furthermore, even in high-risk groups substantial savings were possible, and the risk of overlooking patients with severe disease seemed negligible. Besides clinical relevance, the approach of an ischemia-guided selective catheterization has been shown to potentially contribute to saving increasing health expenses [22], as has been most recently confirmed also for CMR perfusion imaging [23].

As a valuable alternative to perfusion scintigraphy for non-invasive CAD detection, perfusion stress CMR currently has reached very high sensitivity values, but is still limited by a lack in specificity [1, 2]. Our results suggest that the knowledge of proximal coronary diameter provides added diagnostic value and may increase PPV of perfusion CMR up to 90% without relevant loss of sensitivity. Our data provide preliminary support for an extended CMR algorithm in CAD by adding proximal coronary artery diameter information to the current standard approach of Klem et al. (combination of stress perfusion and LGE [24]). Applied as a non-invasive gatekeeper, such an extended CMR approach could further reduce FP CMR results, and consequently, the rate of superfluous diagnostic CA. We are aware that the major limitation for implementation of this CMR concept into clinical routine is the fact that magnetic resonance CA (MRCA) to date is still experimental and inferior to coronary computed tomographic angiography (CCTA) [25, 26]. Thus, a most recent attempt to integrate MRCA and CMR perfusion imaging was not successful [27]. Summarized, the usefulness of this extended CMR approach for clinical practice is limited as long as the vessel caliber is not known at the time of CMR. Currently, it is primarily CCTA that could be applied to achieve this goal, particularly if radiation concerns [28, 29] are addressed with development of techniques leading to significant reduction in CCTA radiation exposure [30, 31].

In our study, nearly half of the FP CMR patients showed no association with coronary caliber variations. The reasons for these FP results were not identifiable either by standard CMR parameters, or by knowledge of the proximal diameters. Possible explanations are hypoperfusion due to variations in regional blood flow distribution [16] or to severe small vessel disease, a known cause of FP CMR results [32, 33]. In addition, the findings might have been caused by unrecognized susceptibility, motion, or ringing artefacts, difficult to distinguish from perfusion deficits despite use of rest perfusion images as in reference [11]. Recent work has shown that one possibility to minimize “dark rim artefacts” was using higher field strength and parallel imaging strategies [34, 35].

The following limitations need to be mentioned for our study. It represents a single center experience with retrospective analysis of prospectively obtained data. Most important, we are aware that the finding of an association between FP CMR results and coronary caliber variations does not prove causal relationship, particularly since we did not perform quantitative perfusion analysis or invasive flow measurements. However, the pathological findings indicative of ischemia encountered in other stress tests support the interpretation of hypoperfusion as a potential cause of the FP results and argue against pure CMR artefacts. Further, we acknowledge that our criterion for classification of a vessel as small caliber may be considered arbitrary. Yet, in the absence of well established definitions the use of the range lesser than one SD below the mean seems plausible. The inclusion criterion of a positive CMR may cause selection bias and thus our results may not reflect the incidence of small caliber coronary arteries in the general population. Finally, since coronary diameter was measured by invasive CA, it remains to be shown whether non-invasive MRCA and CCTA can yield equal results. Thus, our single-center experience should presently be considered as hypothesis-generating result requiring prospective multicenter verification of our findings and evaluation of MRCA’s potential in this concept.