Journal of Surgery

Introduction

Thoracic Endovascular Aortic Repair (TEVAR) has been shown to be a safe and effective treatment for different aortic pathologies, outstripping open surgery as the primary approach [1]. The post TEVAR aortic remodeling is due to the natural aging process, disease progression, and endoprosthesis radial forces and it has been confirmed as critical in determining treatment durability [2]. Other investigators have extensively described the importance of the aortic anatomy (angulation, diameter, tortuosity) and their relationship with endoleaks in different aortic pathologies, but these studies mainly considered a short follow-up period [3-5]. Several reports describing heterogeneous pathologic processes indicate that endoleaks are associated with anatomic factors, such as the diameter of the residual aneurysm, the radius of the aortic curve, and the tortuosity of the aorta [2,6,7]. But maximal diameter, maximal diameter evolutions over time as well as endoleak apparition remain the main parameters to unfavorable aortic remodeling after TEVAR. The insertion of the endoprosthesis in an aneurysm sac may promote adverse aortic remodeling because the stent graft is less compliant than the normal aortic tissue [8,9]. Not all patients will experience the same evolution of thoracic aorta remodeling and risk of endoprosthesis failure after TEVAR. Assessing remodeling progression should enable clinicians to better predict which patients are at higher risk of aortic complications, allowing them to provide a better and earlier patient-specific treatment [10].

The aim of this study was, 1) to describe a framework of medical image analysis to quantify geometrical thoracic aorta changes during the first three years after TEVAR 2) To assess the value of several geometrical parameters to predict unfavorable aortic remodeling earlier than reference parameters (i.e., diameter and diameter progression).

Materials and Methods

The study was conducted in accordance with the Declaration of Helsinki and RGPD Law. This restrospective study follows the MR004 rules and local agreement was obtained # PADS23-17 on 6 October 2021.

Study Design and Patients Sample

A retrospective analysis was conducted of twenty-five patients who were treated with thoracic stent-graft at a single center for thoracic aortic aneurysm, with the following inclusion criterions: availability of a preoperative computed tomography angiography (CTA) (T0) and 3 postoperative CTA performed 6 to 12 months (T6), 24 months (T24) and 36 months (T36) after surgery and availability of all clinical information for a follow up covering at least 36 months after surgery. Only patient with atherosclerotic aneurysm were included. Patients with a missing CTA (T0, T6, T24 or T36) or lost of view for medical follow-up or aneurysm of a non-atherosclerotic origin were excluded. Favorable A group (FAG) was define by the association of 2 parameters: patient without significant aortic diameter increase (maximal aortic diameter increase <10mm / 3 years) and patients who did not require an aortic reintervention during follow-up. On the other hand, patient who required aortic reinterventions as additional treatment related to the thoracic aorta or any direct complication of the initial TEVAR procedure were included in an unfavorable B group (UBG). All the TEVAR procedures were performed using the following stents-graft: GORE Conformable TAG (W. L. Gore & Associates, Flagstaff, Arizona), Cook (Cook Medical, Bloomington, IN, USA) and Medtronic (ValiantTM or the Valiant CAPTIVIA generation). The diameter of stent-graft was oversized by 15% to 20% according to pre intervention CTA measurements. The mean of the three diameter measurements for each proximal and distal landing zone was subtracted from the nominal device diameter and then divided by the mean diameter of the landing zone to calculate the endoprosthesis oversizing. The proximal and distal anchoring zone was recorded according to Ishimaru’s classification modified by Criado [6,7]. Proximal and distal sealing zones of at least 20 mm along the aorticcentre line was selected in normal thoracic aorta, definedby an aortic wall with no evidence of thrombus, calcification or excessive angulation with a diameter lower than 40mm. The overlap between 2 stent graft was ≥ 3 cm if second device was one or two sizes larger than the first device. The overlap was ≥ 5 cm if second device was the same size. Prophylactic lumbar drainage and revascularization of the supraaortic trunk vessels was discussed during Multidisciplinary Consultation Meetings (MCMs) before surgery and according to international guidelines and multidisciplinary expertise.

Image Acquisition and 3D Geometric Analysis

CTA (see Supplemental Appendix 1 for details) was performed on a multi-detector 64-row scanner (REVO EVO, General Electric Healthcare, Buc, France) after contrast media injection with standard parameters. All CTAs (T0, T6, T24, T36) were transferred on a workstation equiped with a vascular imaging software (Endosize; Therenva, Rennes, France). The first step of image analysis was the manual extraction of a three dimensional (3D) arterial Lumen Centerline (CL) starting by a manual designation of the proximal and distal ends of the thoracic aorta from sinotubular junction to celiac trunk allowing the automatic creation of a 3D aortic CL (Figure 1A). Spatial coordinates (x,y,z) of CL points were exported every mm. The aorta was divided into five anatomic zones, from zone 0 to zone 4 according to Ishimaru’s classification modified by Criado [6,7,11]. Points of the CL that crosses limits between two successive zones were defined as follows: P0 for the first point at the sinotubular junction, P1 between Z0 and Z1, P2 between Z1 and Z2, P3 between Z2 and Z3, P4 between Z3 and Z4, and P5 at the end of Z4 (Figure 1B).

Figure 1. Image A illustrates the centerline created from the sinotubular junction to the coeliac trunk. Image B shows the points (P0 to P5) used to define the cross-sections of the aortic lumen and length of the successive zones (Z0 to Z4). Image C shows an example of measurements of angles (α3). Image D shows an example of tortuosity measurement (TI) from Z0 to Z2.

Definition of the Geometrical Parameters

Centerline was used to define aortic length, angle, and tortuosity in the 5 predefined zones (Figure 1B).

Lengths: L0, L1, L2, L3 and L4 were defined respectively as the length of Z0, Z1, Z2, Z3, and Z4 along the CL. The length Li_j corresponded to the length from the beginning of zone Zi to the end of zone Zj. The sum of all lengths (L0+L1+L2+L3+L4) defined the length of the thoracic aorta (L_ATot). The lengths of proximal and distal neck (Lpn and Ldn respectively), as well as the length of the entire endoprothesis (L_EP), were measured (Figure 1C).

Aortic Angulation: α0, α1, α2, α3, and α4 were the angles between planes perpendicular to the CL at the points marking the start (Pi) and end (Pi+1) of each zone. The angle αi_j corresponded to the angle from the beginning of zone Zi to the end of zone Zj. Angles between start and end of the proximal and distal neck of the aneurysm (αpn, and αdn respectively), the angulation of the entire endoprothesis (α_EP), and overall aortic angulation of the whole thoracic aorta (α_ATot) were also assessed. (Figure 1C)

Tortuosity Index: TIi  reflected the tortuosity of a Zi zone, defined by dividing the length of the zone Li by the spatial straight

distance di ( ) between the start point (Pi) and the end point (Pi+1) of Zi. The overall tortuosity of the aortic arch (TI0_2) (Figure 1D) and of the descending aorta (TI3_4) were considered for statistical analysis. The tortuosity index of the entire endoprosthesis (TI_EP) was also assessed. Finally, overall tortuosity of the aorta (TI_ATot) was evaluated by dividing L_ ATot by spatial straight distance between proximal and distal ends of the whole CL ().

Diameters: Maximal aortic diameter was assessed perpendicular to the CL and including aortic thrombus and aortic wall. There were automatically computed by the post-processing tool and corrected, if necessary, by the reader. D0, D1, D2, D3 and D4, were the maximal aortic diameters at P0, P1, P2, P3, P4, respectively. Diameters of proximal and distal necks of the aneurysm (Dpn and Dpn respectively), and maximal aneurysm diameter, Daneurysm, were also measured.

Statistical Analysis

Values are expressed in mean and range. Bayesian Gaussian logistic regression and linear parametric approaches were therefore chosen for their ability to treat small populations (see Supplemental Appendix 2 for implementation details and references). The first model (Bayesian logistic regression) is performed to evaluate the associations between unfavorable outcomes and geometrical explanatory variables at preoperative (T0) and postoperative stages (T6, T24, and T36). When an odd ratio (OR) is higher than 1, the risk of poor outcome increases whereas an OR < 1 is predictive of good outcome. The results are significative if the 95% confidence interval (CI_95%) does not include 1 value.

The second model (Bayesian linear) is performed to compare the evolution of each morphological and geometrical parameter between pre and each postoperative time in FAG and UBG. Mean temporal evolution for each group is estimated and the difference between UBG mean temporal evolution value and FAG mean temporal evolution value is reported (β₁ values). The 95% confidence interval (CI_95%) is calculated for the  value of group variable. If the confidence interval does not include the null  value, we conclude that there is a statistically significant difference in the temporal evolution between the groups.

To compare the means between group and temporal evolution of the means between UBG and FAG groups, a two-way analysis of variance (ANOVA) was used with a significance level of 0.05. Anova was performed using Prism (GraphPad, Boston, USA).

Results

Patient Cohort

Twenty-five subjects met the inclusion criteria for this retrospective study. Demographic data of the patients are presented in Table 1. There was no major perioperative morbidity, and no postoperative mortality. The location of the distal landing zone was on Z4 for all patients. A debranching procedure was performed in 13 cases (52%), with 100% transpositions of the left subclavian artery, 46% debranching of the left carotid artery, and 62% transpositions of the brachiocephalic trunk.