Comparison of Hemodynamics After Fenestrated, Branched, and Chimney Endovascular Aneurysm Repair Employing Computational Fluid Dynamics

1. Introduction

Endovascular repair using fenestrated and branched endografts is the first line treatment for patients with juxtarenal and pararenal abdominal aortic aneurysms (AAAs) [1,2] while off-the-shelf solutions may may be used in urgent setting or when the anatomy does not permit a custom-made endograft application [3]. Endovascular treatment for complex aneurysms has emerged as a preferred approach in contemporary medical practice. The advancements in endovascular solutions, including the use of custom-made grafts that incorporate branches and fenestrations, alongside off-the-shelf devices, have greatly enhanced the feasibility and effectiveness of these interventions. Despite the high technical success, low early mortality and morbidity recorded in large retrospective cohorts of high-volume aortic centers, the durability of complex endovascular approaches is still questioned due to the higher reintervention rates [4,5,6]. Especially for complex endovascular aortic repair, target vessel adverse events are the leading cause for reintervention during the mid-term follow-up [5,6]. Previous analyses on fenestrated versus branched devices showed that fenestrations, when applicable, are superior to branches in terms of target vessel instability [7].

Recent studies show that postoperative hemodynamics in the visceral branches critically influence F / BEVAR and chEVAR success [8,9]. Abnormal flow patterns and shear-stress distributions have been linked to branch stenosis, impaired organ perfusion, and increase target vessel instability events. Given the variability in endograft and techniques, understanding the impact of configuration on target vessels blood flow remains an important unresolved issue [1,10,11,12]. Due to the lack of patient-specific hemodynamic data at the inlet and outlets of the 3D vascular models, advanced multiscale cardiovascular models, 3D–1D, can generate the required waveforms for the 3D simulations and assisted understanding of complex hemodynamic features [13].

This study aims to compare the blood-flow characteristics in target vessels after complex endovascular aortic aneurysm repair using different approaches (FEVAR, BEVAR and chEVAR) using patient-specific computational models.

2. Materials and Methods

Study design and patient selection. A retrospective analysis was conducted on blood flow features pre- and post- FEVAR, BEVAR and chEVAR, see Supplementary Figure 1, among patients that were managed for degenerative complex aortic aneurysms between 2018 and 2021 in a single-tertiary center. Computed tomography angiography (CTA) scans were used to define the pre- and post-repair blood flow lumens of patients with complex aortic aneurysms using computational fluid dynamics simulations, with special focus on the hemodynamic properties of the visceral vessels [8,13].

Patients with juxta-renal, para-renal and thoracoabdominal aortic aneurysms that were managed using FEVAR, BEVAR or chEVAR under elective setting and completed the one-month CTA (slice thickness <1mm and arterial phase) imaging follow-up were considered eligible. A revascularization and bridging of three (for chEVAR) or four target vessels was a criterion for inclusion. The decision for the treatment approach relied on patient-specific characteristics, anatomy of the aorta and target vessels and availability of devices and materials. Patients managed for aortic dissections (acute or chronic), pseudoaneurysms or had an anamnesis of previous endovascular or open aortic aneurysm repair were excluded. No surgeon-modified endografts were used during the study period.

Pseudonymized pre- and post-operative computed tomography angiographies (CTA) of 3 cases for each technique (FEVAR, BEVAR and chEVAR) were collected and analyzed. This study complied with the declaration of Helsinki. No Ethics Committee approval was required due to its retrospective nature and pseudonymized information.

Endovascular approaches. FEVAR was used for the treatment of juxta- and para-renal aortic aneurysms, if the apposition of the device on the aortic wall and especially the level of the target vessels was expected. All FEVAR cases were managed with company-manufactured custom-made devices relying on the Zenith Platform (Cook Medical, Bloomington, IN, USA) [14]. BEVAR was chosen for patients with more extensive aortic pathologies, including thoracoabdominal aneurysms, when a device apposition on the aortic wall was not expected. Patients managed with BEVAR were treated using off-the-shelf branched devices (T-Branch device, William Cook Europe, Bjaeverskov, Denmark); all relying on the Zenith platform. ChEVAR was used in urgent cases (symptomatic or ruptured juxta- or pararenal aortic aneurysms) or aneurysms with a diameter over 70mm, when a custom-made was not available due to the long-time interval for device design and manufacture. For patients managed with chEVAR, the Endurant IIs (Medtronic, Santa Rosa, CA, USA) device was used as aortic component. All aortic devices were oversized by 20%, except for chEVAR, where the aortic device was oversized by 30%, respecting that three target vessels were revascularized [15].

All target vessels were bridged using the same bridging stents (BeGraft Peripheral, Bentley, Hechingen, Germany). The bridging stent diameter and length was decided upon the technical features of the main device and the distal landing zone and anatomy of the target vessels. Details on anatomic parameters and devices are presented in Table 1.

Modeling. The 3D model of the FEVAR, chEVAR, BEVAR system, including the Celiac Artery (CA), Superior mesenteric Artery (SMA), Left Renal Artery (LRA), Right Renal Artery (RRA), was constructed using the CTA scan for each patient Supplementary Figure 1A. The reconstruction of the DICOM images into a 3D lumen model was performed using the methodology mentioned in previous studies [8,9,13].

The reconstruction of geometry under consideration started just above the CA and ended just below the iliac bifurcation, while the region of interest was the visceral aorta and specifically the CA, SMA, RRA and LRA. For the FEVAR and BEVAR cases, the segmentation was performed as a combined structure. Regarding chEVAR cases, the lumen of the covered abdominal aorta and stented visceral arteries (SMA, RRA, LRA) were segmented separately using the same software, Supplementary Figure 1. The optimal smoothing factor value (SFV) was determined to be 0.05, based on a smoothing parameter study performed on renal and mesenteric diameters using the Vascular Modeling Toolkit (VMTK) [8].

The following hemodynamic parameters: Time Average Wall Shear Stress (TAWSS), Oscillatory Shear Index (OSI), Relative residence time (RRT) and Local Normalized Helicity (LNH) were calculated. The postoperative results presented correspond to a representative patient from each group, as no significant differences between patients were observed.

Numerical simulations were performed with the software package Ansys Fluent (Ansys Inc., Canonsburg, Pennsylvania, United States). Blood was considered a Newtonian fluid with a density of ρ = 1050 kg/m^-3 and a kinematic viscosity of v=3.2 × 10^-6 m³ s^-1. The zero-velocity condition was applied on the surface of the endograft, following the rigid wall assumption. Over two million tetrahedral elements were discretized from the fluid domain, with an additional three-layer boundary the inner element size was approximately 1 mm. The boundary conditions at the inlet and outlets were defined using the flow and pressure waveforms obtained from the multiscale model (0D-3D) developed in our previous studies [9,13]. Specifically, a flow waveform representing physiological (healthy) conditions, obtained from the literature, was applied to the input [9,13]. At the outlets, resistance-based boundary conditions (pressure waveforms) derived from the 1D mathematical model were imposed for each respective artery. The velocity profile at the outputs was almost parabolic during the systolic phase [8,13,16,17,18,19,20]. Convergence to the solution at each step was considered achieved when the computational error fell below 10^-4. We used exclusively the results of the last cardiac cycle, avoiding any dynamic behavior of the first pulses [10]. Grid independence tests were performed in regions of predicted disturbed flow to determine the optimal mesh size. The simulation was considered grid-independent when the hemodynamic parameters (velocity and pressure) varied by less than 2% between successive mesh refinements [21]. In all 3D cases, a parabolic velocity waveform is imposed at the inlet, no-slip condition at the wall (i.e. on the stent-graft surface) and pressure waveforms at all outlets (Supplementary Figure 2). The latter (pressure waveforms) were derived from a 1D hemodynamic model of the entire arterial network [13,16,18,19,20,22,23,24,25]. Paraview (Kitware Inc.) software was used for the visualization of hemodynamic parameters.

Statistical analysis. Differences in hemodynamic parameters (RRT, OSI, TAWSS) among the three endograft groups were assessed in the target vessels (SMA, LRA, RRA) using one-way ANOVA, with p < 0.05 considered statistically significant.

3. Results

Wall Shear Stress (WSS). Postoperatively, during the peak systole, WSS values showed the maximum value of 3Pa for all designs (Figure 1). Higher WSS values across the whole structure and especially in the visceral arteries were observed for BEVAR cases, followed by intermediate values for FEVAR and the lowest values for chEVAR as shown in Table 2.

Local Normalized Helicity (LNH). The helical structures in all three endograft groups are distributed evenly throughout the device (Figure 2). In FEVAR cases (Figure 2A), the helical patterns appear more compact and smoother compared to those observed in BEVAR (Figure 2C). Postoperatively, the helical structures in BEVAR were more pronounced and exhibited a clearer rotational configuration (|LNH| < 0.3). According to the literature, a threshold value of 0.3 is typically considered indicative of rotational flow structures [26,27,28]. In contrast, in FEVAR cases (Figure 2A), the local normalized helicity (LNH) shows less rotational behavior compared to chEVAR (Figure 2B). Additionally, in the visceral arteries of all groups, a relatively uniform distribution of helical structures was observed postoperatively.

Time Averaged Wall Shear Stress (TAWSS). The results were evaluated assuming a maximum TAWSS value of 3Pa. As shown in Table 2, BEVAR exhibited the highest TAWSS values, compared to the other two techniques. BEVAR and FEVAR demonstrated relatively high TAWSS, with only minor deviations between them, whereas chEVAR cases showed significantly lower TAWSS (Figure 3). In BEVAR cases, increased TAWSS was distributed across nearly the entire surface of the structure, while in FEVAR cases, moderate TAWSS values were observed primarily in the region below the renal arteries and above the iliac bifurcation (Figure 3A).

Oscillatory Shear Index (OSI) and Relative Residence Time (RRT). There were areas of low OSI values, in the main aortic endograft for all groups. More regions of low value of localized OSI, mainly occurred in the FEVAR and BEVAR cases. OSI showed intense oscillations in individual locations of the visceral arteries in the chEVAR group. In postoperative BEVAR cases, the OSI showing increased values among the mesenteric and renal arteries Supplementary Figure 3C. In BEVAR target vessels, low levels of OSI in proximal end and moderate values in the distal end were recorded, in contrary to chEVARs, where high value localized OSI prevails [Supplementary Figures 3C and B]. In all cases, the RRT values were mainly driven by the OSI patterns following the opposite behavior (Supplementary Figures 3 and 4).

Flow comparison in FEVAR, chEVAR and BEVAR. The flow at peak systole in FEVAR, chEVAR and BEVAR groups is depicted in Figure 4. The maximum flow velocity in the visceral arteries was recorded in BEVAR ~2.5 m/s versus ~2.2 m/s in chEVAR and ~1.7m/s in FEVAR cases (Table 2). No statistically significant differences in visceral vessel flow were observed between FEVAR and chEVAR. The endograft segment between the CA and renal arteries showed increased flow in BEVAR, while the flow was smoother in the main body of the other two techniques (Figure 4) than in chEVAR [Supplementary Figure 5B], during the systolic phase.

Mean visceral hemodynamics. No statistically significant differences were found in RRT and OSI of the target vessels among the techniques. Comparing the mean values of hemodynamic parameters (RRT, OSI and TAWSS), only slight differences were observed between FEVAR and BEVAR. In contrast, chEVAR showed deviation from the other two groups Figure 5.

Statistically significant differences (p < 0.05) were observed for TAWSS, for the SMA (p = 0.0162) and RRA (p = 0.0255) in the post-operative setting. Regurgitant flow in the visceral arteries was computed and compared across FEVAR, chEVAR, and BEVAR cases. chEVAR exhibited the highest mean value, FEVAR intermediate, and BEVAR the lowest, Table 3 [8,17].

4. Discussion

FEVAR / BEVAR expanded their targeted patient population and currently represents the recommended approach in patients with adequate anatomy and complex aortic aneurysms while chimney technique received a lot of criticism regarding the risk for failure of the proximal landing zone and higher endoleak rates and represents mainly a bailout approach [29,30,31]. FEVAR and chEVAR endovascular aortic repair have been applied in anatomically suitable complex aortic aneurysms. Regardless of the technique, endovascular management seems to provide a higher benefit in terms of mortality and morbidity perioperatively while the need for reintervention still represents a major issue, with most of them though being minor and endovascularly performed, without impact on survival [29,30,31]. Except for the main device, target-vessels’ fate plays a significant role in FBEVAR and chEVAR clinical success, with endoleaks, occlusions and instability affecting the durability of the procedure and being related mainly to branches [32]. In this study a multiscale computational approach composed of simple holistic mathematical models combined with advanced 3D CFD simulations could provide clinicians with important information about patients’ hemodynamics. Further clinicians can evaluate the applications of the 3 methods, follow the appropriate treatment strategy for each patient in the treatment of complex aneurysms and predict any postoperative complications. In this analysis evaluating the hemodynamic performance of fenestrated, branched devices and chimney approaches, it was shown that all techniques lead to a reconstruction of the aortic and visceral vessel geometry and hemodynamic results close to normal levels [8,10,11,12,13]. Higher-pressure values during the systolic phase were obtained proximally in the main device regarding target-EVAR cases followed by chEVAR while the lowest values were recorded in FEVAR cases. Overall, FEVAR and BEVAR seem to provide similar smoother flow with less regurgitation than the chimney technique.

FEVAR and chEVAR are usually applied in anatomically similar complex abdominal aortic aneurysms, but their application may lead to distinct local hemodynamic changes. An overall improvement in hemodynamics was detected after repair with either technique, with improved hemodynamics towards normal values and reduced recirculation zones in the main graft and target vessels. Preoperatively, a disturbed pro-thrombotic WSS profile was recorded in several zones of the sac in both FEVAR and chEVAR cases. The LNH results showed a better organization of the helical structures at post-operative setting, decreasing thrombus formation, with either modality. Similarly, TAWSS increased and OSI decreased post-operatively, signaling non-disturbed blood flow. The RRT was locally reduced. The flow in target vessels tended to be more streamlined in chEVAR, compared to evident recirculation regions at renal and superior mesenteric artery fenestrations (p = 0.06); a finding probably related to the configuration of the bridging stent, with chimney grafts running vertically while the fenestration bridging stents run transversely. However, FEVAR showed less intense flow regurgitation in bridging stents.

Regarding the comparative findings on BEVAR and FEVAR, the hemodynamic characteristics after the repair were clearly improved regardless of the approach. The WSS profiles of BEVARs show slightly higher values, compared to FEVARs. LNH in both cases improved by suppressing the disturbed blood flow but providing a more balanced distribution in FEVARs cases. The distribution of TAWSS in BEVAR was prominently increased in the entire structure, which leads to low values of OSI and RRT. The flow in target vessels seems to be more streamlined in FEVAR but presents an instant increase, compared to BEVAR where the flow develops gradually, with BEVAR showing slightly less recirculation zones in the branches, in contrast to FEVAR where the flow is more stable. These findings justify the fact that in FEVAR, we have less occlusion and more stable technical outcomes in target vessels as also reported in the literature [33,34].

Limitations. The elastic properties of the graft material were not considered, and the surface of the lumen was modeled as rigid. Blood could be modeled as non-Newtonian fluid, separating red blood cells from plasma, which were considered in the current study as a single continuum medium. The geometric and structural parameters employed in the 1D arterial model were based on bibliographic data. A future arterial model could be constructed with a geometry and elasticity database derived from patient-based measurements, and the model predictions would be compared with non-invasive measurements. The selection of the patients utilized in the present study was made after anatomical control, so that our sample is morphologically comparable. In the comparative results of FEVAR, chEVAR and BEVAR, the different preoperative anatomical characteristics should be considered. To validate the hemodynamic predictions, we could impose patient-specific boundary conditions on each individual patient. An extended study on a larger cohort of patients could shed light on the hemodynamics of patients undergoing complex endovascular repair. In addition, patient-specific boundary conditions could be imposed on each individual patient for obtaining more accurate hemodynamic predictions.

5. Conclusions

In this study, a multiscale modeling framework was employed, coupling low-order computational models with three-dimensional CFD simulations to investigate postoperative aortic hemodynamics. Key flow-related quantities, including velocity, local normalized helicity (LNH), wall shear stress (WSS), and time-averaged wall shear stress (TAWSS), were analyzed in patient-specific aortic geometries following implantation of advanced endovascular devices, namely BEVAR, chEVAR, and FEVAR.

All three endovascular designes led to an overall improvement in postoperative blood flow characteristics toward near-physiological conditions, with a reduction in adverse flow patterns and recirculation zones. FEVAR demonstrated more stable visceral artery perfusion, BEVAR exhibited higher flow rates with fewer recirculation regions, while chEVAR showed more streamlined visceral artery flow accompanied by reduced regurgitation at the bridging stent entries. Despite these differences, all configurations effectively preserved visceral artery perfusion.

Overall, all EVAR designs produced comparable hemodynamic outcomes across most parameters and consistently improved flow conditions relative to the pathological state. FEVAR was associated with lower velocity magnitudes and more uniform, smoother helicity distributions compared to BEVAR and chEVAR. In contrast, the multiscale model predicted higher visceral artery flow values in the BEVAR configuration relative to FEVAR. The use of three-dimensional, patient-specific aortic models enhances the clinical relevance of the present findings and supports their potential application in personalized treatment planning. These results may further contribute to the future technological refinement of branch endografts tailored for use in F/BEVAR and chEVAR procedures.