A new proof-of-concept study, published as an open-access Original Article in Annals of Biomedical Engineering, reports a preliminary attempt to predict which patients will develop the two most mechanically driven forms of this complication, type I and type III endoleaks, before the operation even begins. Type I endoleaks arise when the seal between the proximal or distal landing zone and the vessel wall is ineffective, while type III endoleaks stem from structural failure of the graft itself, typically the separation of the modular parts of a multi-component device. The researchers, based at institutions in Milan including Politecnico di Milano and the Fondazione I.R.C.C.S. Ca’ Granda Ospedale Maggiore Policlinico, propose a novel hemodynamic risk factor, called R, computed entirely from simulations of the patient’s aorta before intervention.

The team applied a fluid–structure interaction, or FSI, computational framework to ten patient-specific cases of thoracic aortic aneurysm, all treated at the Milan center, for which the post-TEVAR outcomes were already known from follow-up CT angiography scans. Unlike pure computational fluid dynamics, FSI models the coupled motion of blood and the deformable aortic wall, an approach the authors argue offers a more accurate representation of the mechanisms at play, even though its high computational cost has kept it underused in TEVAR research. Indeed, according to the authors, no prior studies had computed drag forces using FSI to explore their association with TEVAR outcomes. Each aortic lumen was reconstructed from pre-operative CTA images using level-set segmentation, from the ascending thoracic aorta down to the origin of the coeliac tripod, and meshed with radius-dependent tetrahedral fluid elements and a three-layered boundary layer, with structural meshes extruded outward from the fluid surface.

The modeling choices were detailed and patient-specific in geometry but partly generic in physics. Blood was treated as a Newtonian, incompressible fluid governed by the Navier–Stokes equations in an Arbitrary Lagrangian–Eulerian formulation, with a large eddy simulation turbulence model, the σ-model, adopted to capture the transition to turbulence considered particularly relevant in aneurysmal disease. The aortic wall was modeled as a nearly incompressible linear elastic material, with a Young’s modulus of 0.8 megapascals in healthy tissue and 1.2 megapascals in the aneurysmal region. A physiological pulsatile pressure waveform was prescribed at the inlet, calibrated resistances were applied at the brachiocephalic, left common carotid and left subclavian outlets, and a three-element Windkessel model governed the descending aortic outlet. Simulations ran over ten cardiac cycles on 112 cores of the LEONARDO supercomputer at Italy’s CINECA center, with the first three cycles discarded to let the flow fields stabilize.

From these simulations, the researchers quantified the hemodynamic drag forces that blood exerts on the aortic wall, computed at the systolic peak as an area-normalized integral of pressure and wall shear stress over each landing zone. The drag force was decomposed into a component perpendicular to the vessel axis and one parallel to it, and each component was assigned a mechanical rationale. A large perpendicular component, the authors hypothesize, could press the stent graft against one side of the wall, potentially detaching it from the opposite wall and causing a type I endoleak, or could drive excessive vessel dilation that reduces the device’s radial force and compromises sealing. Pronounced curvature or tortuosity can also induce lateral graft movements linked to adverse events. A large parallel component, meanwhile, signals a force along the blood-flow direction that could promote device migration, a mechanism considered especially relevant to type III endoleaks through downstream migration of the most distal stent graft.

Tortuosity entered the formula explicitly. Drawing on a published retrospective study by Ueda and colleagues that statistically quantified the correlation between aortic tortuosity and endoleak formation, the team defined a probability of type I and III endoleak occurrence as a function of a tortuosity index, calculated as the sum of curvature values greater than 0.2 per centimeter measured every millimeter along the landing-zone centerline. The local risk factor r was then defined as the product of the squared perpendicular drag component, the parallel drag component and this tortuosity probability, scaled by a constant. The authors describe this multiplicative structure as a deliberate heuristic: it ensures that elevated risk arises primarily when multiple unfavorable conditions coincide, so that a high value in a single component does not necessarily produce a high overall risk, avoiding the compensatory effects an additive formulation would introduce.

Turning the local measure into a patient-level number required a careful workflow. For each patient, the actual proximal, distal and overlap landing zones of the implanted stent grafts were identified on post-operative scans, and the local risk factor was computed for an initial two-centimeter segment of each zone, the minimum length guidelines require for secure fixation. That segment was then progressively extended in contiguous half-centimeter subsegments, cut perpendicular to the vessel centerline, until the full landing zone was covered, and the resulting values were scaled to the interval from zero to one on a patient-specific, zone-specific basis to yield the global risk factor R. The ten-patient cohort was split into an eight-patient calibration set, four without endoleaks and four with them, and a two-patient test set held back for a blind preliminary check.

The calibration produced a striking separation. Every landing zone affected by an endoleak showed an R value above 0.9, with a single exception at 0.71, while nearly all complication-free zones scored below 0.20. On that basis the team assigned low risk to values of R at or below 0.33, moderate risk to values between 0.33 and 0.67, and high risk to values above 0.67. Only one false positive emerged, an overlap zone in a patient whose neighboring regions did develop leaks, and no false negatives were observed. The two moderate-risk cases were dissected in detail: in one, elevated drag forces concentrated over a short proximal landing zone on the aortic arch were tempered by mild tortuosity, while in the other, relatively low drag combined with maximal tortuosity to push the score into the moderate band, illustrating the modulating role of each component.

Applied blindly to the two held-out patients, the classification performed as hoped. In one test patient who had suffered a distal type I endoleak, the risk factor returned its single high-risk value, R equal to 1.00, precisely at the distal landing zone, while assigning low risk to the proximal zone and both overlap zones, matching the follow-up findings. The other test patient, who had no complications, received low-risk scores across all zones. The authors also observed patterns with planning implications: extending the distal landing zone away from the aneurysmal sac, a region of elevated drag, consistently reduced the computed risk, whereas a very long proximal landing zone did not necessarily lower risk, because proximal fixation near the aortic arch sits in hemodynamically hostile territory where added length must be weighed against added force.

The researchers are careful to frame these results as exploratory rather than definitive. The inlet and outflow boundary conditions were drawn from the literature rather than measured in each patient, blood was modeled as Newtonian despite its shear-thinning behavior, the aortic wall was treated as linear elastic, and the risk factor ignored structural quantities such as von Mises stresses. Most importantly, ten patients, with thresholds calibrated on only eight, are far too few for statistically significant conclusions, and the authors explicitly state that the three-level risk stratification should not be interpreted as final. Still, the study identifies all endoleak cases analyzed, and the team plans to expand the cohort and conduct a prospective study in which stent-graft landing zones would be selected using the analysis. If validated at scale, a pre-operative simulation that flags high-risk anatomy before the first incision could give clinicians a genuinely new tool for planning thoracic aortic repairs.

Subject of Research: A computational hemodynamic risk factor derived from pre-operative fluid–structure interaction simulations to predict endoleaks after thoracic endovascular aortic repair

Article Title: Predicting Post-TEVAR Endoleaks: A Pre-operative Hemodynamic Risk Factor from Patient-Specific Fluid–Structure Interaction Simulations

Article References: Duca, F., Tavarone, S., Domanin, M., Bissacco, D., Trimarchi, S., Vergara, C., & Migliavacca, F. (2026). Predicting Post-TEVAR Endoleaks: A Pre-operative Hemodynamic Risk Factor from Patient-Specific Fluid–Structure Interaction Simulations. Annals of Biomedical Engineering. https://doi.org/10.1007/s10439-026-04337-w

Image Credits: AI Generated

DOI: 10.1007/s10439-026-04337-w

Keywords: TEVAR, thoracic aortic aneurysm, endoleak, fluid–structure interaction, hemodynamics, drag forces, aortic tortuosity, stent graft, patient-specific simulation, risk factor, pre-operative planning, computational modeling