Patient-specific fluid-structure simulations of anomalous aortic origin of right coronary arteries.

Objectives: Anomalous aortic origin of the right coronary artery (AAORCA) may cause ischemia and sudden death. However, the specific anatomic indications for surgery are unclear, so dobutamine-stress instantaneous wave-free ratio (iFR) is increasingly used. Meanwhile, advances in fluid-structure interaction (FSI) modeling can simulate the pulsatile hemodynamics and tissue deformation. We sought to evaluate the feasibility of simulating the resting and dobutamine-stress iFR in AAORCA using patient-specific FSI models and to visualize the mechanism of ischemia within the intramural geometry and associated lumen narrowing.
Methods: We developed 6 patient-specific FSI models of AAORCA using SimVascular software. Three-dimensional geometries were segmented from coronary computed tomography angiography. Vascular outlets were coupled to lumped-parameter networks that included dynamic compression of the coronary microvasculature and were tuned to each patient's vitals and cardiac output.
Results: All cases were interarterial, and 5 of 6 had an intramural course. Measured iFRs ranged from 0.95 to 0.98 at rest and 0.80 to 0.95 under dobutamine stress. After we tuned the distal coronary resistances to achieve a stress flow rate triple that at rest, the simulations adequately matched the measured iFRs (r = 0.85, root-mean-square error = 0.04). The intramural lumen remained narrowed with simulated stress and resulted in lower iFRs without needing external compression from the pulmonary root. Conclusions: Patient-specific FSI modeling of AAORCA is a promising, noninvasive method to assess the iFR reduction caused by intramural geometries and inform surgical intervention. However, the models' sensitivity to distal coronary resistance suggests that quantitative stress-perfusion imaging may augment virtual and invasive iFR studies.

Computational model of anomalous intramural coronary tissue deformation and blood flow.

Patient-specific computational fluid–structure interaction modeling is a promising noninvasive tool to quantify the hemodynamic impact of the anomalous aortic origin of the right coronary artery.

Instantaneous wave-free ratio directly measures ischemia due to the anomalous aortic origin of a right coronary artery at stress. To noninvasively acquire the same values from computed tomography, we developed novel computational models of vascular wall deformation and perfusion. The simulation accuracy demonstrated the potential for fluid–structure interaction modeling to guide surgical management.

See Commentary on page 163.

Anomalous aortic origin of a coronary artery (AAOCA) is a congenital malformation in which the coronary artery arises from the aorta outside of the normal coronary sinus of Valsalva., It is the second-leading cause of sudden death in otherwise-healthy youth., Present in 0.1% to 1.5% of the general population, most AAOCA cases are asymptomatic and presumed to be benign.,

Previous AAOCA autopsies and registries found greater frequencies of ischemia and sudden death among certain anatomic variants., Notably, the cumulative risk of sudden death for young athletes from 15 to 35 years of age is approximately 6.3% for the anomalous left main coronary artery and 0.2% for the anomalous right coronary artery. Therefore, current guidelines recommend surgical repair of all anomalous left main coronary arteries but only anomalous right coronary arteries (AAORCA) in the presence of ischemia or other morphologic risk factors., Greater-risk morphologic features visible on noninvasive imaging include a high ostial takeoff (above the sinutubular junction), slit-like orifice, interarterial course (passing between the great arteries), and intramural course (passing circumferentially within the aortic wall)., Despite these risk factors, many patients with AAOCA survive through late adulthood with an indeterminate ischemic burden and concomitant cardiovascular comorbidities., Therefore, further risk stratification is necessary, especially for adults with AAORCA.

The controversy regarding treatment recommendations is complicated by the uncertainty about the exact mechanism of ischemia. Although a narrowed ostium intuitively restricts blood flow, the effect of the proximal course anatomy is less clear. Some believe that the great arteries compress the interarterial segment during exercise, whereas others doubt pulmonary artery pressures are sufficient., Alternatively, the expanding aorta may push the interarterial coronary against the pulmonary root stemming from the cardiac fibrous skeleton. As the aorta distends outward, the wall is also stretched thinner, potentially narrowing the cross-sectional area of the intramural course. Understanding the exact mechanism is critical to guide the extent of surgical intramural unroofing and predict the potential benefit of pulmonary artery translocation for interarterial coronaries.,

To guide clinical management, some centers use invasive pressure catheters to measure intracoronary hemodynamics at rest and with dobutamine stress to simulate exercise., Similar to fractional flow reserve (FFR), instantaneous wave-free ratio (iFR) measures the averaged poststenotic pressure as a fraction of the upstream reference but focuses only on the wave-free period, the portion of diastole when the coronary microcirculation resistance is minimized. Based on stable coronary artery disease trials, iFR ≤0.89 is often considered hemodynamically significant, at rest or dobutamine stress, indicating a need for intervention., However, these invasive procedures are expensive, difficult with a tight ostium, and carry a small risk of coronary dissection and bleeding.

One noninvasive option is using coronary computed tomography angiography (CTA) for patient-specific computational fluid dynamics modeling, which has simulated the FFR in atherosclerotic coronary artery disease and the hemodynamics affecting bypass graft survival., More recently, computational fluid dynamics has also been used to model the FFR and hemodynamics in AAOCA, but such methods have yet to be validated with direct measurements for anomalous coronaries.18, 19, 20 Also, computational fluid dynamics modeling fails to simulate the coronary lumen compression that may occur during exercise. Meanwhile, advancements in fluid–structure interaction (FSI) modeling enable simultaneous modeling of blood flow and tissue deformation, but to our knowledge, this approach has not been applied to AAOCA.

We hypothesized that patient-specific FSI simulations of AAORCA yield iFRs similar to the invasively measured values in the cardiac catheterization lab during rest and dobutamine-stress. If accurate, such a model of AAORCA may noninvasively quantify the ischemic burden in individual patients and guide surgical management if necessary.

Methods

Patients and Clinical Characteristics

Cleveland Clinic's institutional review board approved the use of data for this clinical cohort study, with patient consent waived (study number 17-1087, approved August 15, 2017, expiration date August 14, 2021). We identified 6 patients with AAORCA with an available coronary CTA who underwent a dobutamine-stress iFR cardiac catheterization between April 2018 and May 2019. Baseline characteristics (including demographics, coronary anatomy, and noninvasive stress tests) were abstracted from the patients' electronic medical records (Table 1). Coronary anatomy was determined based on operative reports (when available) and expert review of the coronary CTA. Noninvasive stress tests included electrocardiography, echocardiography, single-photon emission computed tomography, and positron emission tomography. For each modality, we categorized the results as normal, equivocal, or definite for myocardial ischemia. Electrocardiography results were classified as indicative of ischemia if the ST changes met diagnostic criteria. Similarly, we classified the imaging studies as indicative of ischemia if a wall motion abnormality or perfusion defect correlated with the territory supplied by the AAOCA. Otherwise, abnormal results were defined as equivocal.

Table 1

Baseline patient characteristics

ID	Age, y	Sex	Symptoms	Noninvasive stress	Comorbidities	% Diameter CAD stenosis (LMCA/LAD/LCx/RCA)	Inter-arterial	Intra-mural
1	45	F	Angina, dyspnea, palpitation	Equivocal (ECG)	Asthma, obesity, anxiety	0/30/0/0	Yes	No
2	60	M	Atypical angina	Normal (echocardiogram, ECG)	None	0/0/0/0	Yes	Yes
3	63	M	Atypical angina, dyspnea	Normal (ECG, PET)	HTN	0/0/0/0	Yes	Yes
4	24	F	Atypical angina, dyspnea, palpitation	None	Arrhythmia, asthma, LAD myocardial bridge	0/0/0/0	Yes	Yes
5	42	F	Syncope	Equivocal (ECG)	Arrhythmia	0/0/0/0	Yes	Yes
6	53	F	NSTEMI, angina, dyspnea	Normal (ECG, PET)	HTN, HLD, anxiety	0/0/0/0	Yes	Yes
Average ± SD	48 ± 14					0/5/0/0

Patient characteristics at the time of iFR study. Stress tests results were recorded normal, ischemic, or equivocal. Inducible ischemia was based on stress electrocardiography, echocardiography, or nuclear perfusion imaging. CAD, Coronary artery disease; LMCA, left main coronary artery; LAD, left anterior descending coronary artery; LCx, left circumflex coronary artery; RCA, right coronary artery; F, female; ECG, electrocardiogram; M, male; PET, positron emission tomography; HTN, hypertension; NSTEMI, non-ST-elevation myocardial infarction; HLD, hyperlipidemia; SD, standard deviation.

Patient-Specific FSI Modeling

We constructed the patient-specific FSI models in SimVascular, an open-source software platform for modeling cardiovascular blood flow. Specifically, we used the svFSI solver, which implements an arbitrary Lagrangian–Eulerian method to simultaneously calculate fluid flow and tissue deformation. Material characteristics such as the densities, blood viscosity, and tissue elasticity properties were the same for all patients and described in greater detail in the Appendix 1.

Baseline geometric models of the lumen were segmented from each patient's coronary CTA (Figure 1). The vascular walls were offset from the lumen of the aortic root and coronary arteries by their corresponding thicknesses. The 3-dimensional (3D) models were then discretized into millions of tetrahedral elements to achieve adequate spatial resolution and accuracy.

Figure 1

Fluid–structure interaction (FSI) modeling pipeline with SimVascular. Computed tomography angiographies (CTAs) were segmented by first drawing pathlines and 2-dimensional cross-sections to generate the 3-dimensional lumen model. Vessel walls came from offsetting the lumen by set thicknesses. Both the fluid and solid domains were meshed with local refinement around the proximal anomalous coronary. The simulated prestress within the solid tissue was added as a boundary condition along with the inlet pressure and outlet lumped-parameter networks.

Next, we assigned the boundary conditions, mathematical properties that defined the forces and movement at surfaces. The inlets and outlets were fixed in space whereas the external surfaces freely moved. The aortic fluid inlet was a time-dependent pressure source scaled to the patient's blood pressure and heart rates (Figure E1). For the fluid outlets, the same hemodynamic parameters, along with stroke volume, scaled each component in the lumped-parameter network, which mathematically modeled the downstream physiology. Importantly, the coronary lumped-parameter model included a dynamic pressure source (representing the ventricular compression of the microvasculature). We also simulated low- and high-exercise conditions that targeted volumetric flows at 2 and 3 times resting values, respectively, consistent with previous coronary computational fluid dynamics models. We also incorporated the internal stresses within the vascular wall using a previously described prestress modeling process., The method and justification of scaling all parameters at rest and stress is detailed in Appendix 1.

Figure E1

Simulation boundary condition with transformed pressure waveforms. Representative pressures measured by cardiac catheterization (dotted lines) had high-frequency whip artifacts. The raw waveforms were filtered and scaled to match the patient's blood pressures and heart rates at rest and stress as boundary condition inputs (solid lines) for the fluid–structure interaction models. Pressures shown at rest included the raw catheter measurements in the aorta (Cath_rest) and the simulated values for the aorta (Aorta_rest) and left ventricle (LV_rest). The same waveforms are shown for the stress conditions, respectively (Cath_stress, Aorta_stress, LV_stress).

Ultimately, the svFSI solver calculated the corresponding tissue deformation and blood flow over 5 cardiac cycles. The resulting pressure ratio between the aortic inlet and coronary outlet during the wave-free period was calculated as specified by the iFR_MATLAB algorithm (Figure 2).

Figure 2

Representative flows and pressures in aorta and right coronary artery (RCA). The aortic (red) and RCA (blue) pressures were similar at rest but separated with dobutamine stress, during which the flow (purple) through the RCA increased to 3 times the resting rate. The iFR was calculated during the wave-free period as the mean ratio between the RCA (orange) and aorta (green) pressures. iFR, Instantaneous wave-free ratio.

Sensitivity Analysis: Mesh Resolution and Pulmonary Root Forces

We verified mesh convergence by doubling the resolution of the solid and fluid domain for one model. As an exploratory effort, we added the mechanical effects of the adjacent pulmonary root in another patient. The external surface contacting the pulmonary root was assigned a Robin-type boundary condition, which acted as a spring pushing back against the interarterial coronary. Additional details of the mesh generation and Robin-type boundary condition are described in Appendix 1.

Statistics

Patient clinical characteristics were summarized with averages and standard deviations. Accuracy of FSI-simulated iFR compared with invasively measured values was evaluated with the root-mean-square error (RMSE), and the Bland–Altman plot with 95% confidence limits of the mean difference. We also reported the linear regression of the simulated iFR over the measured iFR and the corresponding Pearson's correlation coefficient, r. All analyses were completed in R, version 4.0.3 (R Core Team).

Results

Patient Characteristics and Association With Measured iFR

The 6 patients in this adult AAORCA cohort had a mean age of 48 ± 14 years (range 24-63 years). Most were female (4/6, 67%). All had atypical cardiac symptoms leading to coronary imaging that diagnosed the AAOCRA. Each anomalous coronary had an interarterial course and arose leftward of the left-right commissure near the sinotubular junction. All but one had an intramural course. There was no atherosclerotic coronary artery disease ≥30% stenosis or any definitive ischemia on noninvasive stress tests. Therefore, the recommendation for surgical intervention for these patients was difficult, prompting further iFR assessment. Dobutamine-stress lowered the average iFR from 0.96 ± 0.02 (range 0.95-0.98) at rest to 0.87 ± 0.06 (range 0.80-0.95). Individual iFR measurements and hemodynamic parameters are summarized in Table 2.

Table 2

Measured hemodynamic parameters

Patient number	Heart rate, bpm	Blood pressure, mm Hg	Cardiac output, L/min	Measured iFR	Simulated iFR
Rest
1	95	137/94	5.9	0.98	0.99
2	58	140/85	4.3	0.97	0.97
3	53	140/78	4.3	0.95	0.99
4	100	91/61	4.9	0.95	0.98
5	99	114/61	4.5	0.98	0.97
6	76	122/63	4.9	0.95	0.96
Average ± SD	80 ± 21	124/74 ± 19/14	4.8 ± 0.6	0.96 ± 0.02	0.98 ± 0.01
Stress
1	142	155/94	N/A	0.95	0.98
2	110	240/150	N/A	0.83	0.87
3	134	142/134	11.0	0.84	0.89
4	162	103/68	N/A	0.86	0.85
5	151	146/74	N/A	0.95	0.86
6	164	138/85	10.2	0.80	0.75
Average ± SD	144 ± 20	154/101 ± 46/34	10.6 ± 0.6	0.87 ± 0.06	0.87 ± 0.07

The invasively measured iFR value, our primary outcome, is listed with the patient-specific parameters in the fluid–structure interaction models for the resting and dobutamine stress conditions. These parameters included the heart rate, blood pressures (systolic/diastolic), and cardiac output. iFR, Instantaneous wave-free ratio; SD, standard deviation; N/A, not available.

For this clinically significant range of dobutamine-stress iFR, we assessed whether any of the easily measured input hemodynamic parameters for the FSI models independently predicted the measured iFR. Neither blood pressures, heart rates, nor cardiac outputs at stress correlated with the measured stress iFR (Figure E2). Unsurprisingly, the patient with the greatest iFRs at both rest and stress was the only one lacking an intramural course.

Figure E2

Association between measured dobutamine-stress iFR and each patient-specific input for the FSI model. No significant correlation was observed between the dobutamine-stress iFR and any of the patient-specific factors: blood pressures (BP; systolic and diastolic), heart rate, and cardiac output. Cardiac output at stress was modeled as either 2 or 3 times the cardiac output at rest. The dotted trendline represents the least-square regression. RCA, Right coronary artery; iFR, instantaneous wave-free ratio; FSI, fluid–structure interaction.

Of the 4 patients who had a dobutamine-stress iFR less than 0.89, 3 underwent coronary unroofing. One surgical patient is still undergoing postoperative evaluation for ischemia as of the last follow-up in 2021. The other 2 had chest pain resolve after surgery and dobutamine-stress iFR improvements from 0.80 to 0.93 and 0.84 to 0.91. The patient with a dobutamine-stress iFR 0.83 had a systolic pressure of 240 mm Hg during the test and was managed medically for the abnormal hypertensive response first.

FSI-Simulated iFR Accuracy

Under resting conditions, the simulated iFR reasonably matched the measured iFR (RMSE = 0.02). For the dobutamine-stress condition, the simulated iFR was more accurate after targeting a multiplier of 3 times rather than 2 times the resting flow (RMSE = 0.05 vs 0.08, respectively) (Figure E3). The lower flow tended to underestimate the iFR reduction (such that the simulated iFR was greater than that measured). The combined iFRs at rest and stress demonstrated a moderately strong linear correlation between the simulated and measured iFRs (r = 0.85, RMSE = 0.04, Figure 3). Bland–Altman analysis demonstrated moderate agreement between the measured and simulated iFR within differences ≤0.05, except for one outlier in which the simulated dobutamine-stress iFR was 0.09 lower than the measured value (Figure 4).

Figure E3

Correlation between FSI-simulated and measured iFR stratified by physiologic stress level. The iFR was simulated by the FSI model of AAORCA at rest and 2 stress conditions, one with 2 times the resting cardiac output (2 × CO) and another with 3 times the resting cardiac output (3 × CO). At rest, the FSI-simulated and measured iFR correlated closely, with a root mean square error (RMSE) of only 0.02. For the stress conditions, the higher cardiac output achieved a greater iFR reduction in the anomalous right coronary and matched the invasively measured iFR more closely (RMSE 0.05 vs 0.05). The solid diagonal gray line represents the ideal perfect correlation and the dotted colored lines represent the linear regressions corresponding to each of the stress levels. FSI, Fluid–structure interaction; iFR, instantaneous wave-free ratio; AAORCA, anomalous aortic origin of a right coronary artery; S, simulated iFR; M, measured iFR.

Figure 3

Correlation between simulated and measured iFR. To quantify the ischemic burden from the anomalous aortic origin of a right coronary artery (AAORCA), we invasively measured the iFR at rest and dobutamine stress. We also computationally simulated the iFRs with fluid–structure interaction (FSI) models segmented from computed tomography. The strong correlation between measured and simulated iFRs suggests that noninvasive FSI modeling may accurately assess anomalous coronaries and guide surgical management. iFR, Instantaneous wave-free ratio; RMSE, root mean squared error.

Figure 4

Bland–Altman plot analyzed the agreement between the FSI-simulated iFR to the iFR measured in the cardiac catheterization lab at both rest and dobutamine stress. The horizontal axis is the invasively measured reference iFR, and the vertical axis is the difference between that and the simulated iFR. Horizontal reference lines represent the mean difference and corresponding 95% confidence interval (CI). FSI, Fluid–structure interaction; iFR, instantaneous wave-free ratio.

In addition to the numeric iFR values, the hemodynamic waveforms from which they were derived also appeared reliable. After a few initial cardiac cycles to wash out transients, the pressures and flows converged to a periodic steady-state during which we calculate the iFR (Figure E4). The flow through the anomalous right coronary had a physiologic bimodal velocity waveform, whereas the left coronary flow peaked only in diastole as expected. After dobutamine stress, all the anomalous right coronaries developed substantial pressure drops relative to the aorta, especially during diastole.

Figure E4

FSI-simulated pressure and flow rates. Representative simulation depicting pressures and flow rate converging to periodic steady-state within 5 cardiac cycles at rest and dobutamine stress. Pressures and flows shown included the inlet, aorta, right coronary artery (RCA), left anterior descending (LAD), and left circumflex (LCx). Pressure also included the left ventricle (LV). The flow rates for the inlet and aorta were scaled down to 1%.

Our sensitivity analyses verified that the mesh resolution and pulmonary root had minimal effects on the results. After the total number of elements was doubled, the simulated pressures and flows were nearly identical with less than a 1% change in iFR (Figure E5). Adding the pulmonary root stiffness caused the contact surface displacement between the coronary and pulmonary artery to fall from 0.72 mm to 0.02 mm, but the iFR changed by less than 1% (Figure E6). The interarterial segment maintained the same cross-sectional shape but was prevented from translating outward by the pulmonary root.

Figure E5

Mesh convergence on hemodynamic parameters. Representative example of an FSI simulation of an anomalous RCA under dobutamine. The total element count increased from 1.4 to 3.1 million, and the resolution across the short-axis of the intramural course increased from 4 to 7 elements wide. The different mesh resolutions yielded minimal change to the flow rates (dotted), pressures (thin), and wave-free period pressures (thick). FSI, Fluid–structure interaction; RCA, right coronary artery; iFR, instantaneous wave-free ratio.

Figure E6

External compression from the pulmonary root. A and B, Pulmonary root (violet), pushing against the anomalous right coronary artery, was modeled as a Robin boundary condition applied to the contact surface (red). C, The displacement of the coronary artery external surface, relative to the original nondisplaced surface (green), exceeded 0.7 mm. D, After the stiffness of the pulmonary root was applied as a spring-like Robin boundary condition, the displacement decreased to less than 0.02 mm while the cross-sectional area of the intramural course and iFR remained the same. iFR, Instantaneous wave-free ratio.

Discussion

FSI Model Accuracy and Uncertainty

For our primary outcome, we compared the resting and dobutamine-stress iFRs from the FSI simulations to those measured in the cardiac catheterization lab. The strong correlation (r = 0.85) in this study is somewhat greater than similarly reported results in successful clinical trials evaluating CTA-derived FFR for coronary artery disease patients with normal coronary origins (r = 0.80). The FSI models also did not show significant systemic bias (eg, overestimation or underestimation) in iFR predictions during rest and stress. The overall error (RMSE = 0.04) was also small. For comparison, the beat-to-beat variation of iFR is about 0.01 to 0.02 and the mean difference between repeated measures of FFR has a standard deviation of 0.04. Thus, our 6 AAORCA FSI models accurately simulated the iFR at rest and dobutamine stress using only blood pressure, heart rate, and noninvasive imaging data.

We also assessed FSI simulation realism by examining the characteristic flow profile of the coronary and aortic outlets (Figure 3). The left coronary flows had a single peak during diastole (due to the systolic compression of the microvascular bed by the left ventricle). In comparison, the ventricular compression of the right coronary system (which predominantly supplies the right ventricle) is weaker and permitted a greater systolic flow resulting in the characteristic bimodal flow waveform.

The accuracy of the simulated iFR also was not explained by the patient-specific hemodynamic input parameters alone. We suspected lower iFR values for patients with greater blood pressures, heart rates, and cardiac outputs, but none of these variables correlated with the measured iFR (Figure E2). Therefore, the full FSI simulation was necessary to integrate the patient-specific 3-dimensional geometry and hemodynamic measurements to accurately estimate the iFR.

The error between the simulated and measured iFR possibly arose from a variety of sources such as the 3-dimensional geometry, solid tissue properties, and boundary conditions. Image segmentation repeatability and intramural course interpretation were limited by the CTA resolution. The uncertainty from the distal coronary resistances was reflected by the different iFR results for the 2 volumetric flow rates for the stress condition. The more accurate results required raising the permitted coronary flow to 3 times the resting rate which is on the higher end of the expected stress-state coronary perfusion. One recently published computational fluid model showed that the range of possible distal coronary resistances has a similar magnitude of FFR change as 70% narrowing of the proximal anomalous coronary. Sensitivity to distal coronary resistances also highlights iFR's potential underappreciation of hemodynamic severity of AAOCA in the setting of microvascular disease. To improve the model accuracy and assess microvascular dysfunction, quantitative myocardial perfusion imaging with positron emission tomography or cardiac magnetic resonance imaging may be helpful.

To systematically quantify the effects of input parameters uncertainty on the iFR results, several approaches for uncertainty quantification suitable for cardiovascular models are available. One FSI study of normal resting coronary flow found that the variables with the strongest influence were the inlet pressure and the intramyocardial pressure (contained within the coronary lumped-parameter model). Coronary wall elasticity had a minimal effect such that one standard deviation change in elasticity resulted in less than a 1% change in coronary flow. Other analyses in rigid-wall computational fluid dynamics studies found significant sources of uncertainty in the coronary lumped-parameter network values, especially the microvascular resistance., Despite the various sources of input variability, the resulting hemodynamic output was acceptably accurate such that the FFR predicted by CTA-based computational fluid dynamics is approved in the United States and the United Kingdom to guide clinical management of coronary artery disease. Although adequately accurate for coronaries with normal origins, the robustness of computation simulations has yet to be proven for anomalous coronaries.

Mechanism of Ischemia Revealed by FSI Simulation

Our FSI simulations, which omitted pulmonary artery effects, demonstrated that the iFR reduction can be accurately explained by the intramural geometry alone. We also observed that the intramural lumen was restricted from expanding during exercise (Video 1). During stress, the lower distal coronary resistance allowed for a higher flow through the anomalous coronary (with a relatively constant cross-section) resulting in a proportionally greater pressure drop and a lower iFR. Thus, these simulation results implicated the intramural course rather than the interarterial course as the cause of ischemia.

Other engineering models and imaging studies of AAOCA also point to a similar ischemia mechanism. Finite element analyses of the intramural course under constant pressure loads (without flow simulation) found that the intramural lumen expands less compared to normal-origin coronary arteries., Meanwhile, computational fluid dynamics simulations with rigid geometries showed how the narrowed intramural course caused significant hemodynamic abnormalities.,, A 3D-printed model of an intramural AAORCA had an abnormally low FFR that normalized in a separate model of the postoperative geometry. A cine CTA of an interarterial (but not intramural) anomalous left coronary found no luminal compression nor perfusion defect during dobutamine-stress. In contrast, intravascular ultrasound of intramural left coronaries showed the intramural segments to be narrowed. Interestingly, 1 of the 3 intravascular ultrasound cases showed a 10% worsening of the intramural stenosis after dobutamine-stress induction. Although the dobutamine-stress imaging studies were of the anomalous left coronary, the intramural course deformation is likely similar for AAORCA.

Although our FSI simulations implicated the intramural geometry as the primary culprit of stress-induced ischemia, the pulmonary root may still play a role in select interarterial cases. Previous AAOCA registries found associations between the interarterial course and ischemia or sudden death, but separating the intramural and interarterial effects through statistical models is difficult because the 2 features are highly correlated., Additional FSI models of any ischemic interarterial cases without intramural courses could provide additional insights.

Implications for Surgical Intervention

Based on the atypical chest pain or indeterminate stress tests present in all patients within this cohort, the decision for surgical correction of the AAORCA was unclear. Using a dobutamine-stress iFR threshold of 0.89, 4 of 6 patients would have been recommended surgery. The FSI-simulated iFR results led to similar surgical recommendations except for 1 patient with a measured iFR of 0.95 but a simulated iFR of 0.86. The simulation would have recommended surgery whereas the invasively measured iFR would not have. Alternatively, this patient may have had undiagnosed microvascular dysfunction resulting in a greater measured iFR. A second patient with a measured iFR of 0.85 had a simulated iFR of 0.89, which would have led to only a weak recommendation for surgical repair. Even with small differences in measured and simulated iFR, the discrepancy can still lead to divergent recommendations for surgical repair. This imperfect dichotomization also afflicts FFR-guided revascularization in coronary artery disease. With further validation, patient-specific FSI models may become a noninvasive and cost-effective tool to recommend surgery versus medical observation.

Mechanistic insight from the FSI simulation also helps inform the surgical technique for AAOCA repair. Since the omission of the pulmonary root from these models still resulted in significantly low iFRs, pulmonary artery translocation likely would not have benefited these specific interarterial cases. Instead, unroofing, reimplantation, or neostium creation would be necessary to address the intramural course. None of the patients in this cohort were good candidates for bypass grafting due to the absence of obstructive coronary artery disease, without which competitive flow through the native vessel would likely lead to early graft failure. Had there been significant coronary artery disease, our simulation results could identify whether the ischemia was due to the anomalous origin or atherosclerotic stenosis. Determining the culprit lesion may improve the selection of surgical repair techniques. We can also preoperatively model the postoperative coronary anatomy and resulting hemodynamics to confirm that adequate coronary perfusion would be restored.,

Limitations

Although dobutamine-stress iFR is increasingly used to guide clinical management, the correlation between iFR and risk of major adverse cardiac events in anomalous coronaries has yet to be validated. Like noninvasive stress tests, iFR has traditionally been used to assess fixed atherosclerotic coronary artery disease., That no patients in our cohort had clear signs of ischemia on nuclear perfusion imaging, despite suspicious symptoms, also underscores the limitations of noninvasive ischemia assessment in AAOCA. Patients with strong evidence of ischemia on noninvasive testing attributable to AAOCA would have undergone surgery without the need for further iFR testing. Thus, our study lacked patients with confirmed ischemia on noninvasive testing. With no single modality representing the gold standard, we were unable to calculate the sensitivity or specificity of our iFR results. Future studies comparing iFR with other ischemia testing modalities and clinical outcomes are necessary.

While our FSI simulations were more complete than previous AAOCA computational models, some aspects may have been oversimplified. For example, all but one case omitted the effects of the pulmonary root. Also, the solid tissue was modeled as a uniform isotropic material despite the frequent proximity of the stiff commissures and our intraoperative observation that the intramural course is often surrounded by more fibrotic tissue. Increasing the model complexity with different material properties may result in less reproducibility between model builders and increase uncertainly of results, but such advancements should be confirmed in future studies.

Given the time-intensive manual 3-dimensional model construction process and significant computational costs of FSI simulations, the sample size of this study was limited to only 6 adult patients with AAORCA containing an interarterial course. To generalize the FSI modeling to all patients with AAOCA, stronger validation of accuracy and robustness is needed with and increased sample size that includes anomalous left coronaries, other morphologic variants, and children.

Conclusions

Patient-specific FSI modeling of iFR is a promising tool to noninvasively quantify the hemodynamic significance of AAORCA with dobutamine stress. Compression from the pulmonary root was unnecessary to achieve accurate iFR results, suggesting that the narrowed intramural course is a sufficient mechanism of ischemia. However, additional models incorporating the pulmonary root may be necessary to rule out ischemia due to the interarterial course for specific patients. Finally, the FSI model's sensitivity to the volumetric flow rate permitted by the distal coronary microvascular resistance suggests a future role for quantitative stress-perfusion imaging to complement iFR for risk stratification of AAOCA.

Conflict of Interest Statement

The authors reported no conflicts of interest.

The Journal policy requires editors and reviewers to disclose conflicts of interest and to decline handling or reviewing manuscripts for which they may have a conflict of interest. The editors and reviewers of this article have no conflicts of interest.

Table E1

Patient-specific parameters for inlet pressures and outlet lumped-parameter network models

Parameter	Rest	Stress
Aorta blood pressures	Cath	Cath
Cardiac output (CO)	Echo	[2-3] × COrest
Aorta
Ra:Rd	9:91	20:80
C (cm5/dyne)	0.001	0.001
Coronary
Ra:Ra-micro:Rv-micro + Rv	32:52:16	55:25:20
Ca:Cim	0.11:0.89	0.11:0.89
CRCA, cm5/dyne	2.5 × 10−5	∝ CO
CLAD, cm5/dyne	2.2 × 10−5	∝ CO
CLCx, cm5/dyne	1.4 × 10−5	∝ CO

Patient-specific aortic pressures originated from the iFR measurements at rest and stress. The target cardiac output (CO) from resting cardiac output (COrest) was doubled or tripled to approximate the stress condition. The remainder of the parameters were based on ratios from previously published coronary simulations. Cath, Cardiac catheterization; echo, echocardiography; Ra:Rd, proximal to distal aortic resistance ratio; Ra:Ra-micro:Rv-micro + Rv, coronary arteriole to microvascular to venule resistance ratios; Ca:Cim, coronary arteriole to intramyocardial capacitance ratios; C, total capacitances for right (RCA), left anterior descending (LAD), and left circumflex (LCx) coronary arteries, respectively.