An Updated Comparison of Conventional Coronary Angiography With Cardiac Magnetic Resonance Imaging to Diagnose the Origin and Proximal Course of Anomalous Coronary Arteries

Background. Anomalous coronary arteries (ACAs) may present increased risk for adverse cardiac events. We sought to evaluate the accuracy of conventional coronary angiography (CCA), as it is currently used in clinical practice, compared with expert interpretation and cardiac magnetic resonance imaging (CMR) in determining the site of origin and proximal course of ACAs. Methods. Fifty consecutive patients without concomitant congenital heart disease, who were referred for CMR to diagnose the course of an ACA, were retrospectively evaluated. Original CCA reports were reviewed. Angiography images were available in all patients and were interpreted by 2 experts blinded to the prior interpretation and CMR results. The accuracy of interpretation in each group was then compared to the current gold standard of CMR. Results. Identification of the site of origin (ie, aortic sinus) by referring angiographers was similar to that of expert angiographers (sensitivity, 89% vs 98%, respectively; P=.10). However, referring angiographers were less likely to correctly identify the proximal course as compared with expert angiographers (sensitivity, 27% vs 98%, respectively; P<.001). Conclusions. As it is used in current practice, CCA does not provide sufficient diagnostic accuracy for identifying the proximal course of an ACA. Review by expert angiographers added sensitivity, improving the accuracy to nearly 100%. Expert consultation may be nearly as accurate as advanced imaging, and should be considered in cases of ACA in which there is diagnostic uncertainty.

J INVASIVE CARDIOL 2021;33(9):E681-E686. Epub 2021 August 8.

Key words: coronary angiography, coronary vessel anomalies, magnetic resonance imaging

The presence of an anomalous coronary artery (ACA) — specifically, anomalous aortic origin of a coronary artery (AAOCA) from the opposite sinus of Valsalva — may result in an increased risk of myocardial ischemia or infarction, syncope, or sudden death.¹ ACAs are usually first noted when patients undergo cardiac catheterization with conventional coronary angiography (CCA). Correct diagnosis of the proximal course of an ACA is crucial in determining proper treatment; AAOCA from the opposite sinus with an interarterial course (ie, passing between the aorta and pulmonary trunk), particularly those in which the proximal portion is intramural, is associated with an increased risk of adverse events, while the other possible courses are usually considered “benign.”^1,2 A concise review of the characteristics and pathophysiology of the many types of AAOCA has been previously published.³

While the presence of an ACA is usually recognized during CCA, many cardiologists are not confident in diagnosing its proximal course, often leading to referral for further imaging studies, such as cardiac magnetic resonance imaging (CMR) or cardiovascular computed tomography angiography (CCT). These techniques can provide excellent visualization of the ACA and adjacent structures to allow diagnosis of the origin and proximal course of the ACA.^4-10 However, these modalities are not universally available, and their use results in increased cost and inconvenience for the patient, potentially delays an important diagnosis, and (at least with CCT) involves risk from contrast media and radiation exposure. Furthermore, it is unclear whether these studies are necessary, since the proximal course of ACAs may be identifiable from CCA alone.^11,12 Therefore, the purpose of this study was to evaluate the accuracy of CCA as it is currently used in clinical practice (ie, in a “real-world setting” in patients referred to a dedicated CMR center, from both a large academic medical center and from multiple referring cardiologists in community practice) compared with expert CCA interpretation and CMR in determining the origin and proximal course of ACAs.

This retrospective analysis of CMR and coronary angiograms was approved by the institutional review board of the University of Alabama at Birmingham.

Subjects. Fifty consecutive patients (27 women and 23 men; median age, 56 years) without concomitant congenital heart disease referred to our  CMR facility for evaluation of the proximal course of an ACA had CCA images available for our review. Table 1 (Part 1; Part 2) summarizes the rate of identification of the origin and proximal course of the ACA by imaging modality. In 37 of these patients, the original CCA report was available for review. In 48 of the 50 patients, the ACA was identified initially by CCA; in one 13-year-old patient (#22), transthoracic echocardiography suggested an anomalous right coronary artery (RCA) arising from the left sinus of Valsalva (LSV) and passing between the great vessels (following CMR, non-selective aortography demonstrated the ACA). In a 32-year-old patient (#37) with syncope and lateral T-wave inversion on electrocardiogram, the referring cardiologist requested screening CMR out of clinical suspicion of ACA; subsequent CCA was performed at the time of an electrophysiology study.

Conventional coronary angiography. CCA was performed by the referring cardiologists for clinically indicated reasons. The projections obtained were at their  discretion, as we had access to the angiograms only in retrospect. Angiograms were reviewed by a highly trained cardiothoracic radiologist with over 30 years of experience in coronary angiography. Angiograms were also reviewed by a second reader (a cardiothoracic radiologist with over 40 years of experience in coronary angiography for patients #1-#26, and a cardiologist with over 10 years of experience in coronary angiography and intensive training in interpretation of anomalous coronary angiograms for patients #27-#50). Each reader recorded the site of origin and proximal course of the ACA and was blinded to the CMR and the other reader’s diagnosis. Analysis was based on the methods of Ishikawa and Brandt¹¹ and Serota et al.¹²

Cardiac magnetic resonance imaging. Initial CMR was performed on a 1.5 T scanner (CV/i; GE Healthcare) designed for cardiac studies in 24 of the patients. The other 26 patients underwent initial CMR on a 3 T scanner (Intera; Philips Medical Systems) designed for cardiac imaging. Electrocardiographic gating and a combination of spin-echo black blood and gradient-echo bright blood methods (using Cartesian and/or spiral k-space coverage schemes) in multioblique planes were used as needed to visualize the ACA. Scans performed on the 1.5 T system made use of single-breathhold methods to acquire the images, while those on the 3 T scanner also employed a respiratory navigator gating method allowing free breathing, with a T2 preparation and fat suppression. Typical scan parameters for the black-blood sequences were slice thickness of 6 mm, matrix size of 256 x 256, in-plane resolution of 1.4 mm/pixel, flip angle of 90°, repetition time of 2 heartbeats, echo time of 18 ms, and bandwidth of 488 Hz/pixel; those for the bright-blood sequences were slice thickness of 1.5 mm, matrix size of 384 x 384, in-plane resolution of 0.53 mm/pixel, flip angle of 20°, repetition time of 5.5 ms, echo time of 1.7 ms, and bandwidth of 135 Hz/pixel. An experienced CMR cardiologist or radiologist was present during the performance of these scans to optimize the examinations. The diagnosis of the site of origin and proximal path of the ACA from the initial clinical report was then compared with CCA and with CMR findings.

Statistical analysis. Analyses began by summarizing descriptive statistics for each outcome. Frequency of correct and incorrect interpretation by method were summarized. To account for repeated measures for a subject across different modalities, generalized linear mixed models were used to estimate the sensitivity of each modality. Generalized linear mixed models adjust the standard errors of the estimates to account for potential correlation among outcomes for the same patient. All analyses were conducted using SAS, version 9.4 and utilized a type I error rate of 0.05. For pairwise comparisons among modalities, a Bonferroni correction was applied.

The rate of identification of the origin and proximal course of the ACA determined by the CCA interpretation by the referring angiographers (the Referring group), expert angiographers (the Expert group), and by CMR are summarized in Table 1 (Part 1; Part 2). The results of the comparison of the imaging modalities are summarized in Table 2.

A total of 37 original catheterization reports were available for review; we requested (but did not receive) the official report in the other 13 cases. The diagnosis was considered “correct” if the origin/proximal course was clearly identified in the official report. The diagnosis was considered “incorrect” if the origin/proximal course was either not identified or identified incorrectly, as compared with CMR. Referring angiographers made the correct diagnosis of the origin of the ACA in 30 of 37 patients (89%) and of the proximal course of the ACA in 10 of 37 patients (27%). The expert interpreters of CCA made the correct diagnosis of the origin of the ACA in 49 of the 50 patients (98%) and of the proximal course of the ACA in 49 of the 50 patients (98%). There was no significant difference between the different interpreters in determination of the origin. However, the rate of correctly identifying the proximal course of the ACA was significantly lower in the Referring group (P<.001). Notably, 3 patients were ultimately determined to have no ACA by CMR. In 2 of these patients, the Expert group correctly identified the cases as normal. In 1 patient (#29) (Table 1; Part 1, Part 2 and Figure 1), the Expert group identified an anomalous RCA arising from the LSV with an interarterial course, while CMR showed a normal origin of the vessel from the right sinus of Valsalva (RSV), although near to the commissure with the LSV. In this case, a standard left anterior oblique (LAO) view in the catheterization session was not obtained, and it was felt in retrospect by the Expert group that inclusion of this usual view would have increased the likelihood that they would have interpreted the study as “normal.” In comparison, Figure 2 shows non-selective aortogram images in patient #22, in whom the RCA arising from the LSV with interarterial course was seen on the aortograms, and confirmed on cardiac MRI in a 3 T scanner; the LAO view clearly shows that the vessel does not arise from the RSV.

As the gold standard, CMR provided excellent images and allowed diagnosis of the site of origin and proximal course of the ACA in all patients. CMR did not reveal evidence of other congenital cardiac disease in any patient. This is an important factor, since the incidence of ACA is increased in patients with associated congenital heart disease.

This study represents the largest comparison conducted to date between CCA and CMR to assess the origin and proximal course of ACAs. While prior studies have demonstrated the accuracy of CMR compared with CCA,^4-7 the current study is the first to compare initial real-world analysis with both review by expert angiographers and with CMR. Our results demonstrate that the identification of the site of origin (ie, aortic sinus) of ACAs by real-world initial interpretation is similar to that obtained with review by expert angiographers, using CMR as the gold standard. However, the proximal courses of ACAs are frequently left undetermined or incorrectly interpreted by referring angiographers. In our study, initial interpretation was correct in only 10 of 37 cases (27%) in which the initial angiography report was available for review. After expert review of the same angiographic images, the proximal course of the ACA was correctly identified in 49 of 50 cases (98%).

Thus, the diagnostic accuracy of CCA was significantly enhanced, simply by expert review of the angiographic images, adding incremental value to the already-performed CCA test.  Furthermore, CCA interpretation by the trained angiographers was highly reproducible, with complete agreement between the 2 blinded observers, using well-established rules for interpretation.^11,12 As the gold standard, CMR had excellent visualization of the ACA origin and pathway in all cases. Our protocol made use of widely available spin-echo and gradient-echo sequences used for general cardiac imaging, as well as sequences specifically designed for coronary arterial imaging; this method allowed excellent visualization of the proximal ACA in all cases.

Utilization of 3-dimensional (3D) imaging modalities, including CMR and CCT, is known to be useful in the diagnosis of the origin and proximal course of ACA, as these techniques provide direct visualization of important structures surrounding the ACA (ie, the ascending aorta and pulmonary trunk).^4-7,10 With CCA, the position of these structures and their relationship to the ACA must be inferred, since they are not directly visualized. It seems that this limitation, along with the increasing availability of 3D imaging, has led to a heavier reliance on additional imaging (eg, CMR or CCT) to define the proximal course of an ACA. However, based on our findings, further imaging may not be needed in the majority of patients if the cardiologist follows a systematic method of performance and interpretation of CCA. An accurate diagnosis by CCA alone could result in decreased costs, decreased inconvenience for the patient, and (if CCT is employed as an additional imaging modality) decreased risk of iodinated contrast and radiation exposure to the patient. Furthermore, and perhaps most importantly, proper diagnosis at the time of initial CCA would allow appropriate treatment to be initiated sooner. Real-time diagnosis would also allow additional imaging modalities to be employed in the catheterization lab, such as intravascular ultrasound (IVUS), which is useful in demonstrating a potential intramural (ie, within the aortic wall) course of AAOCA from the opposite sinus,¹³ which may impact the strategy for any possible surgical management.

We found that the most useful angiographic views for depicting the course of the ACA were the right anterior oblique (RAO) and lateral projections for a prepulmonary vessel; the LAO and RAO for a retroaortic vessel; the LAO, RAO, and lateral projections for an interarterial vessel; and the RAO, lateral, and anteroposterior projections for an intraseptal vessel. In all cases, we advise performing these views without cranial or caudal angulation, as these can make it difficult to discern whether the proximal part of the vessel takes a superior or inferior course.

Although our expert CCA readers were able to correctly diagnose the course of the vast majority of ACAs here, their level of confidence would have been even higher and the diagnosis even faster if CCA had been performed in all of the optimal projections on all patients. It is possible that their accuracy could have been even higher if a standard LAO view had been obtained on patient #29. There are several reasons why cardiologists might not perform all of the optimal views. These include the following: inability to selectively engage the ACA; technical difficulties (eg, patient body habitus); pressure to finish the procedure quickly; and justifiable efforts to minimize radiation exposure and iodinated contrast dose. The first two of these reasons may be unavoidable, but in such cases where ACA is recognized, efforts to end the exam quickly to minimize exposure of radiation and contrast may well be counterproductive. Finally, even with a well-done CCA study, many cardiologists cannot confidently diagnose the origin and proximal course of an ACA, probably due to the fact that some of these variants are quite uncommon.¹⁴ However, because the anomalies described in this paper are not exceedingly rare, occurring in the range of 1 in 100 to 1 in 10,000 patients, all busy angiographers can expect to encounter them in practice, making a working familiarity with these ACAs and the angiographic techniques needed to describe their anatomy important.¹⁵

Study limitations. An important limitation of our study is the predisposition to sampling bias. We included only patients who were referred for CMR to further define the origin and proximal course of a suspected ACA. Given our patient selection, the diagnostic accuracy of CCA in the true population may be underestimated, as patients in which an ACA was easily identified were likely not referred for additional imaging. Therefore, the true diagnostic accuracy of CCA in all unselected patients who are diagnosed with an ACA at the time of CCA cannot be determined by our study. Despite this, we found that even in cases of ACA who were referred for further testing to determine the origin and proximal course, expert interpretation of the same images improved the accuracy, similar to that of CMR. Another limitation was that only 37 of 50 original catheterization reports were available for review. Therefore, only 37 individuals have all non-missing outcomes. However, statistical procedures, such as generalized estimating equations and generalized mixed linear models, which capitalize on all available information and do not require complete non-missing data for observations to be included in the analyses, were utilized in our analysis.

Because CMR and CCT are known to be useful in diagnosing the origin and proximal course of an ACA, many cardiologists perform only limited CCA after identifying the presence of an ACA, with the intention of referring the patient for further imaging methods. Even CMR and CCT have limitations, however, as there may be anatomical ambiguities in the relation of the ACA to surrounding structures. In some cases, there may be patient factors (eg, claustrophobia, significant arrhythmia, renal dysfunction) or technical difficulties in performing CMR or CCT that lead to images of insufficient quality to allow diagnosis.¹⁶ Finally, although CMR and CCT are capable of providing the 3D relationship of an ACA to the relevant surrounding structures and thus should be very accurate in delineating the proximal ACA course, there are reports of these advanced imaging methods yielding an erroneous diagnosis; in 1 reported case, inappropriate cardiac surgery was planned based on incorrect CMR interpretation.¹⁷ Furthermore, although CCT can non-invasively detect coronary atherosclerosis, CMR is not ideal for evaluating coronary stenosis. Selective CCA of the ACA with adequate views would allow determination of the presence of atherosclerotic disease in the vessel.

This represents the largest study comparing CCA with CMR in ACAs and the first comparing the accuracy of CCA in current practice with expert CCA and with CMR. Given our findings, we conclude that as it is used in current practice, CCA does not provide sufficient diagnostic accuracy for identifying the proximal course of an ACA. However, review by expert angiographers and CMR added incremental benefit, improving the accuracy to nearly 100%. If properly performed and interpreted, CCA allows the diagnosis in most if not all cases of ACA, at least in patients without concomitant congenital heart disease. Further imaging, such as with CMR or CCT, may be helpful if the coronary anatomy is not clearly discernible by CCA, to assess for other cardiovascular abnormalities, or for surgical planning. Given these findings, expert consultation or advanced imaging should be considered in cases of ACA in which there is diagnostic uncertainty, as identification of the proximal course is crucial to determining the risk of the anomaly and in planning treatment. Furthermore, additional exposure in interpreting these pathways would be beneficial for cardiovascular disease trainees.

Acknowledgments. The authors of this paper would like to recognize the contribution of Benigno Soto, MD (deceased 2012), who served as an expert reviewer for cases #1-#26. Without Dr. Soto’s input and inspiration, this work would not have been successful.