Geometrical Changes of the Aorta as Predictors for Thromboembolic Events After EVAR With the Anaconda Stent-Graft

Purpose: Thromboembolic events (TE), including limb graft occlusion (LGO) and distal limb embolization (DLE), are common complications after endovascular aneurysm repair (EVAR). The aim of this study was to find predictors for TE in patients treated with the Anaconda stent-graft for infrarenal aneurysms. Materials and Methods: Geometrical and anatomical variables were retrospectively analyzed in a consecutive Anaconda cohort. Pre- and postoperative CT scans were used to derive geometrical parameters length, curvature, torsion, and tortuosity index (TI) from the center lumen lines (CLLs). Limb characteristics, pre-to-post EVAR and mid-term-follow-up changes in the parameters were evaluated for their predictive value for TE. Results: Eighty-four patients (mean age 74±8.3 years, 74 men) were enrolled. The risk of TE was lowered with pre-to-post implant decreasing TI (steps of 0.05: OR: 1.30, 95% CI: 1.01-1.66, p=0.04), pre-to-post implant decreasing mean curvature (OR: 1.08, 95% CI: 1.01-1.16, p=0.03), and a larger degree of circumferential common iliac artery (CIA) calcification (OR: 0.98, 95% CI: 0.97-1.00, p=0.03). The only LGO predictor was the caudal relocation of maximal curvature after EVAR (OR: 1.01, 95% CI: 1.00-1.01, p=0.04). Preventors of DLE were CIA diameter (OR: 0.87, 95% CI: 0.76-0.99, p=0.04), circumferential CIA calcification (OR: 0.97, 95% CI: 0.95-1.00, p=0.03), mean and maximal curvature of the preoperative aortoiliac trajectory (OR: 0.86, 95% CI: 0.79-0.94, p<0.01 and OR: 0.97, 95% CI: 0.95-1.00, p=0.03, respectively) and pre-to-postoperative decrease in mean curvature (OR: 1.11, 95% CI: 1.02-1.21, p=0.02). Midterm TE predictors were length (OR: 0.95, 95% CI: 0.89-1.01, p=0.08) and torsion maximum location (OR: 1.01, 95% CI: 0.99-1.01, p=0.10). Conclusion: The present study confirms that treatment of infrarenal AAA with an Anaconda stent-graft is related to a relatively high TE rate which decreases with a pre-to-postoperative reduction in curvature and TI, and a larger degree of circumferential CIA calcification. In other words, more aortoiliac straightening and more circumferential CIA calcification may prevent TE development after EVAR with this stent-graft.


Study Design
In this single center retrospective study, all consecutive abdominal aortic aneurysm (AAA) patients treated with an off-the-shelf Anaconda stent-graft from December 2014 until December 2018 were included.Patients treated with a single limb or aorto-uni-iliac Anaconda or treated for nonaneurysmal disease of the aortoiliac trajectory were excluded.Patient information, including demographics, comorbidities, and outcome variables were obtained until June 2020 and reported according to the reporting standards for EVAR. 17he study protocol was approved by the institutional review board (K18-43) and the regulations from the Dutch Act on Medical Scientific Research Involving Human Beings (WMO) and General Data Protection Regulation (AVG) were followed when using the medical data.Consent for research participation was waived as it was a retrospective study.
Pre-and postoperative computed tomography (CT) scans were collected for detailed analysis of geometrical changes.The local follow-up protocol included 1 postoperative CT within 3 months after EVAR and thereafter followup by ultrasound.Additional CT scans were also included, regardless the objective of the CT scan, as long as the abdominal aortoiliac trajectory was imaged.When more than 1 preoperative CT scan was available the scan acquired closest to the date of the primary intervention was used for analysis; when more than 2 postoperative CT scans were available the first postoperative scan and the scan closest to 1 year after EVAR were used (midterm follow-up).
The included patients were divided into subgroups.Patients diagnosed with limb occlusion of the device, defined as total device limb occlusion and/or symptomatic stenosis, anytime during the follow-up period were categorized in the limb graft occlusion (LGO) group.Patients that presented with a symptomatic and imaging confirmed peripheral embolism diagnosed to originate from the stent-graft after excluding other potential sources, such as cardiac or thoracic aortic pathology, were included in the distal limb embolization (DLE) group.The LGO and DLE groups were combined in the thromboembolic event (TE) group.Patients without any TE were categorized in the patent-group (PG).
The primary objective of this study was to observe the differences between the PG and TE groups in limb characteristics and the change in geometrical parameters from the preoperative to first postoperative CT scans.Moreover, potential predictive values of these metrics for TE were determined.Midterm geometrical changes in the geometrical variables were compared between groups as the difference from first postoperative CT to the 1-year follow-up CT.Secondary endpoints were to compare PG vs. LGO and PG vs. DLE and evaluate potential predictive variables for LGO and DLE, respectively.

Image Acquisition
The CT scans were performed on a Somatom Definition AS scanner (Siemens, Healthcare, Erlangen, Germany).The scan parameters included rotation time 0.5 seconds, tube potentials 120 kV, tube current 60 to 65 mAs, collimation 64 mm x 0.6 mm, pitch factor 0.9 to 1.4, reconstruction matrix size 512 x 512 pixels and slice thickness ranged between 1.0 and 3.0 mm, obtained using reconstruction kernel I31f, I25f, or I41f Medium Smooth.Most scans were made with contrast administration (Iomeron 300).All scans were performed during inspiration breath hold.

Image Processing
Center lumen lines (CLLs) of all aortoiliac trajectories and preoperative anatomical characteristics were obtained using the EVAR planning software Aquarius Intuition (TeraRecon, Inc, Foster City, CL, USA) with the incorporated workflow protocol, including manual optimization when necessary.
The postoperative CLLs were created from the most proximal stent-ring of the body to the most distal stent-ring of the limb, resulting in 2 CLLs for each patient (right and left).To compare pre-and postoperative CLLs, the start and end points of the preoperative CLLs were determined based on anatomical landmarks, such as bifurcations, vessel branches, and calcifications on the first postoperative CT scan.

Geometrical Parameters
The geometrical parameters length, curvature, torsion, and tortuosity index (TI) were calculated for all CLLs using previously described inhouse written algorithms in the Python programming language (version 3.7). 18First, the CLLs were sampled at 1 mm and filtered to obtain smooth differentiable curves.This was achieved using a Savitzky Golay filter of polynomial order 4 and window length 33, as confirmed by visual inspection.Length of the aortoiliac trajectory was defined as the length of the CLL.TI was calculated as the CLL length divided by the Euclidean distance.Both curvature (Equation 1) 18,19 and torsion (Equation 2) 19 were calculated for each CLL coordinate by numerical computation: 2 ) ) (2) where x , y , and z are the Cartesian coordinates of the CLL, ′ is the first derivative, ″ the second derivative and ″′ the third derivative along the trajectory of the CLL.Curvature can be interpreted as the inverse of the radius of a circle fitted to the CLL for each CLL point, as previously described. 18Torsion can be interpreted as the extent to which curvature involves the third dimension, also for each CLL point.For example, a rollercoaster looping with a curvature of 1/"radius of the loop" ends not in the same plane as it starts, i.e. the loop involves a third dimension (torsion) and the further the start and end point of the this loop/curve are apart, the larger the torsion.For further analysis, the mean and maximal curvature and torsion values for each CLL were used.Since torsion values can also be negative, depending on torsion direction, the absolute torsion values of each CLL coordinate were used.).Time-to-event data were analyzed as survival analysis using the Kaplan-Meier method.Secondary analyses were performed of PG vs.

Statistical Analysis
LGO and PG vs. DLE, similar to the above described GEE method of PG vs. TE, to identify specific predictors for LGO and DLE, respectively.Furthermore, GEE were used to determine the predictive value of the changes in geometrical parameters during follow-up, that is, from the first postoperative CT to midterm follow-up.
Significance was set at p<0.05, except for the parameters resulting from the univariate analysis with a p-value <0.10.These were used as potential input parameters for the multivariate analysis.All statistical analyses were performed using IBM SPSS Statistics (version 25.0;IBM Corporation, Armonk, NY, USA).

Results
A total of 124 EVAR procedures were performed during the study period, of which 84 (mean age 74±8.3 years, 74 men) used the Anaconda stent-graft (Figure 1).Of these patients, 5 (5.9%) were treated outside of the instructions for use (neck angulation >90 o , n=1; neck length <15 mm, n=2; native iliac diameter >21 mm, n=2).Sixteen patients were treated in acute setting (10 symptomatic and 6 ruptured AAAs).All Anaconda stent-grafts were second generation ONE-LOK systems and all were fully ballooned after placement with either a Constellation balloon (PanMedical, New Delhi, India) or Altosa balloon (Andratec GmbH, Koblenz, Germany).The postoperative anticoagulant schedule was single antiplatelet therapy, with a preference for clopidogrel, or the preoperative direct oral anticoagulants or Vitamin K antagonist was continued.Patient characteristics, anatomical characteristics and procedural parameters are summarized in Tables 1-3, respectively.All TE's are listed in appendix A. TE was observed in 20 patients (30 limbs, 17.8%); LGO in 10 patients (14 limbs, 8.3%) and DLE in 10 patients (16 limbs, 9.5%) during a mean follow-up of 39±15 months, (PG: 40±15 months; TE: 38±17 months, ns), see Figure 2. A summary of the TE patients, including event side and treatment, can be found in appendix A. Age was the only statistically significant difference in baseline characteristics between the groups: the PG cohort is on average 7 years older than the TE cohort.The LGO cohort had a higher level of circumferential aortic neck thrombus than the PG cohort (p=0.05) and the DLE cohort was younger (p<0.01) and had a lower ASA grade (p=0.03)than the PG cohort (Appendix B).

Univariate Analysis
Pre-to-post implantation geometrical changes: Univariate analysis identified 3 factors related to the development of Flowchart of patient enrolment and categorization for analysis.The TE group was compared to the PG (i.e.without thromboembolic limb events) on limb parameters, including geometrical changes (green).Secondary analyses are investigation of the specific event types, so LGO versus PG group and DLE versus PG group (orange) and the change in geometrical parameters during follow-up was compared between the groups for the patients that had a second postoperative CT scan available (yellow).AAA, abdominal aortic aneurysm; AUI, aorto-uni-iliac device; bif, bifurcation; CT, computer tomography; DLE, distal limb embolization; EVAR, endovascular aneurysm repair; LGO, limb graft occlusion; PG, patent group; TE, thromboembolic event; vs., versus.

Multivariate Analysis
The prediction model for TE included circumferential CIA calcification (B -0.023, OR: 0.977, 95% CI: 0.960-0.995,p=0.01) and pre-to-post EVAR change in mean curvature (B -0.106, OR: 1.112, 95% CI: 1.025-1.207,p=0.01).Note that the mean curvature changes are negative and   PG and TE were compared using Generalized Estimating Equations (GEE) analysis, also revealing odd's ratio's (OR).Normally distributed continuous data are presented as the mean±standard deviation; skewed distributed continuous data are presented as the median (inter quartile range [IQR]); categorical data are given as number (%).Abbreviations: CIA, common iliac artery; EIA, external iliac artery; mm, millimeter; OR, odds ratio; PG, patent group; TE, thromboembolic event group; 95% CI, 95% confidence interval.a PG group consists of all patent limbs, i.e. not diagnosed with a thromboembolic event.b TE group consists of limbs with a thromboembolic event (limb thrombosis and/or limb embolism) in follow, up.c Measured on the preoperative CT of the center lumen line (CLL) trajectory that will be covered by the stent, graft.thus the mean curvature decreases in both groups, though the decrease is stronger in the PG cohort.A model with circumferential CIA calcification (B -0.027, OR: 0.974, 95% CI: 0.956-0.992,p=0.01) and midterm change in aortoiliac length (B -0.058, OR: 0.944, 95% CI: 0.895-0.995,p=0.03) seemed strong as well (based on the likelihood ratio criteria), though that was based on only the 62 patients that had a midterm CT.The pre-to-postoperative decrease in TI and postoperative length decrease and torsion maximum relocation were excluded as predictors.Furthermore, forcing any of the potential predictors proposed in literature (distal limb diameter, limb extension into the EIA and limb bridging and/or extension) did not lead to improvement of the model.
For LGO prediction, multivariate analyses did not reveal an additional predictor to pre-to-postoperative change in location of maximal curvature.For DLE prediction, the optimal model combined preoperative mean curvature (OR: 0.839, 95% CI: 0.765-0.919,p<0.01), circumferential CIA calcification (OR: 0.958, 95% CI: 0.924-0.994,p=0.02), and CIA diameter (OR: 0.835, 95% CI: 0.715-0.976,p=0.02).Forcing any of the literature proposed predictors into the models did not improve the DLE model.PG and TE were compared using Generalized Estimating Equations (GEE) analysis, also revealing odd's ratio's (OR).Normally distributed continuous data are presented as the mean±standard deviation; skewed distributed continuous data are presented as the median (inter quartile range [IQR]); categorical data are given as number (%).Variables that differ between PG and TE with a p<0.10 were used for multivariate GEE and depicted bold in the table.Abbreviations: CLL, center lumen line; mm, millimeter; OR, odds ratio; PG, patent group; TE, thromboembolic event group; TI, tortuosity index; 95% CI, 95% confidence interval.a PG group consists of all patent limbs, i.e. not diagnosed with a thromboembolic event.b TE group consists of limbs with follow-up thromboembolic events (limb graft thrombosis and/or distal embolization).c The OR(95% CI) and B resulting from the GEE for TI were obtained after rearranging the variable to make steps of .05instead of 1, considering the magnitude of TI. *p-value indicating a difference between PG and TE groups when <0.05.

Discussion
This study showed that the geometric parameters curvature and TI of the aortoiliac trajectory are altered up to 1 year after EVAR.The odds of TE development drop with decreasing mean curvature and TI of the stented aorto-iliac trajectory from pre-to-post implant and with a higher pre-operative percentage of circumferential CIA calcification.Decrease in length during the first year post-EVAR seems to also increase the odds for TE development.Since the aortoiliac trajectory is straightened during EVAR by the stiff guidewire, the stent-graft is deployed in this straightened environment.It is presumed that after EVAR the more calcified CIAs of the PG patients remain in this straightened position and thereby actuate the larger decrease in curvature and TI.Hence, these limbs could experience a more straight blood flow path.The less calcified and thereby more flexible CIAs of the TE patients may tend to go back to their original, tortuous state after removal of the stiff guidewire and could therefore show a smaller decrease in curvature and TI after EVAR.
As the aortoiliac trajectories of the TE patients were straightened when the stent-graft is placed and go back to their original state after removal of the stiff guidewire, the induced curves in the stent graft limbs are thought to be more prone to the Concertina effect (Figure 3).The Concertina effect implies that in tortuous and/or shortened (Anaconda) stent-graft limbs the neighboring nitinol stentrings come closer together, which may cause inward folding of the graft-fabric between these stent-rings.If these resulting plications reduce the blood flow and wall shear stress, thrombus formation may be promoted. 20The results of a previous prospective study conducting in depth analysis of 15 Anaconda patients support the Concertina effect hypothesis as a correlation was found between the decreasing distances between stent rings and decrease in limb length and increase in curvature. 18Moreover, the 2 patients in that cohort who developed limb occlusion showed a severe postoperative increase in curvature and small distances between the individual stent rings.Furthermore, aortoiliac remodeling during the first year post-EVAR reduces the aneurysm size both in transverse and longitudinal direction and thereby shortens this trajectory, which could further attribute to the Concertina effect.Hence, the Concertina effect may be elemental to the elevated LGOrate in Anaconda limbs, 5,14,15 which is also seen in the present cohort where 16 of the 168 limbs occluded (9.5%) on mid-term.
EVAR limb geometry (changes) have been only partially investigated.Tasso et al 21 investigated the aortoiliac geometry and hemodynamics of Endurant and Excluder stentgrafts.They observed a higher curvature in post-EVAR patients than in healthy subjects, especially in the limbs.Moreover, they found a relation between the maximal curvature and torsion values and the percentage of recirculating blood volume computed with computational fluid dynamics (CFD), which indicates that increased curvature and torsion along the aortoiliac trajectory elevate the risk of thrombus formation. 20These findings complement the present results and substantiate the Concertina effect hypothesis.Further investigation into the presence of the Concertina effect in Anaconda and other EVAR-stent-grafts may be conducted in the field of ultrasound particle image velocimetry (ePIV) 22 or computational fluid dynamics (CFD). 21imilar to the findings of Cochennec et al, 23 the patients with a TE event during follow-up were younger than those with patent limbs during follow-up, although we have no Figure 3.The Concertina effect hypothesis in the Anaconda stent-graft: in straightened Anaconda limbs the walls are relatively straight, but with increasing angulation and/or decreasing limb length, the separate nitinol rings are pressed together, inducing infolding of the graft fabric into the lumen.These plications may induce areas of blood stasis (blue) that are suspect for thrombus formation.
explanation for this finding.However, it emphasizes the importance of considering open repair in younger AAA patients.The previously established relation between limb occlusion and distal limb diameter 5 and limb extensions 2 was not found in the present study.This may be attributed to the ballooning of the stent-graft that was performed in each patient, which improves clinical outcomes. 24,25Contrasting to previous studies by Mantas et al 14 and Carroccio et al 15 who found CIA calcification to be a predisposing factor for limb thrombosis, the present study revealed it to be a protective factor.An explanation for this may be found in the study methodologies.Mantas et al 14 categorized the calcification in <50% and ≥50% calcification, while we used the circumferential calcification degree (%) as a continuous variable.In addition, only 12 of the 72 reviewed stentgrafts were Anaconda's.Also, the study of Carroccio et al 15 was published in 2002 and reviewed other (and older generation) devices than the Anaconda.Our hypothesis is that in a stiff calcified iliac trajectory the Anaconda stays more stretched which prohibits the Concertina effect.
The risk of LGO was separately investigated and seemed to increase with caudal relocation of maximal curvature, which could hypothetically induce local flow disturbances.The risk for a DLE was lowered by pre-to-post implant decrease in mean curvature, larger CIA diameter, higher degree of CIA calcification and larger preoperative curvature of the aortoiliac trajectory.A less straightened aortoiliac trajectory, that is, smaller decrease in mean curvature, can be related to the above mentioned Concertina effect.A larger CIA diameter may provide less disturbed outflow and thereby a higher patency rate. 5,6,15,23The difference in parameters influencing LGO and DLE may be due to the number of events and the degree of flow disturbance.Caudal relocation of the maximal curvature may have a more local influence on the blood flow patterns and thereby induce the local problem of LGO, while a smaller CIA diameter, smaller preoperative curvature with a smaller curvature decrease and less CIA calcification may only alter the blood flow to the point where blood starts to clot and thromboembolisms are formed, but not to the extent that the entire limb gets occluded.However, the exact mechanisms behind these observations and hypotheses need further investigation.
The TEs and the performed treatments, listed in appendix A, reveal that realigning the limbs with Gore Excluder limbs avoids re-TE in 9 of the 10 limbs that underwent this TE-treatment.This is in line with the recent publication of Marino et al. 26 Although relining may aid in future durability of the device, note should be made of the additional costs that come with relining.Relining had limited to no influence on the postoperative changes calculated in this study.
8][29][30][31] However, to fully describe the geometrics of a 3D CLL, torsion should be considered as well.Torsion describes the extent to which the induced curvature involves the third dimension.Even though this parameter is very sensitive to the smallest inaccuracies in the CLL, its potential in predicting complications after EVAR is promising. 19,21The retrospective nature of the study may be considered a limitation, though it does provide useful insights in this specific device.

Conclusion
The present study confirms that treatment of infrarenal AAA with an Anaconda stent-graft is related to a relatively high TE rate in the years following implantation.This study revealed that the TE-risk is lowered by a higher degree of pre-intervention circumferential CIA calcification and a larger decrease in curvature and TI pre-to-post EVAR.These findings support the Concertina effect hypothesis, which may also be related to the high angulation cohort which Anaconda devices are used to treat.Further investigation into the mechanisms behind this effect and consequently targeting stent-grafts design and treatment improvements are pivotal.

Appendix A Summary of Events
Table A1 shows the TE's (LGO and DLE) and the treatments of the individual patients.LGO group consists of limbs with a limb thrombosis in follow-up; DLE group consists of limbs with a limb embolism in follow-up.
The OR (95% CI) and B resulting from the GEE for TI were obtained after rearranging the variable to make steps of .05instead of 1, considering the magnitude of TI. *p-value indicating a difference between PG and LGO or DLE groups when <0.05, which are depicted bold in the table.

Figure 1 .
Figure1.Flowchart of patient enrolment and categorization for analysis.The TE group was compared to the PG (i.e.without thromboembolic limb events) on limb parameters, including geometrical changes (green).Secondary analyses are investigation of the specific event types, so LGO versus PG group and DLE versus PG group (orange) and the change in geometrical parameters during follow-up was compared between the groups for the patients that had a second postoperative CT scan available (yellow).AAA, abdominal aortic aneurysm; AUI, aorto-uni-iliac device; bif, bifurcation; CT, computer tomography; DLE, distal limb embolization; EVAR, endovascular aneurysm repair; LGO, limb graft occlusion; PG, patent group; TE, thromboembolic event; vs., versus.

d
The body diameter was included once for each limb (so twice in total) e Measured on first postoperative CT as the CLL length from flow divider to last stent ring *p-value indicating a difference between PG and TE groups when <0.10, which are depicted bold in the table.

Figure 2 .
Figure 2. Survival analysis with Kaplan Meier plots for freedom from a TE within the entire cohort (n=84, purple) and separately for LGO (blue) and DLE (red), all in months after endovascular aneurysm repair (EVAR).DLE, distal limb embolization; LGO, limb graft occlusion; TE, thromboembolic event.

Table 1 .
Baseline Characteristics of the Cohort of AAA Patients Treated With an Anaconda Stent-Graft.p-value indicating a difference between PG and TE groups when <0.10, which are depicted bold in the table.
Normally distributed continuous data are presented as the mean±standard deviation; skewed distributed continuous data are presented as the median (inter quartile range [IQR]); categorical data are given as number (%).Abbreviations: AAA, abdominal aortic aneurysm; ASA, American Society of Anesthesiologists; MI, myocardial infarction; PG, patent group; TE, thromboembolic event group.a PG group consists of all patients with patent limbs, i.e. not diagnosed with a thromboembolic event.b TE group consists of patients with a thromboembolic event (limb graft occlusion and/or distal limb embolism) in follow-up.c Only includes current smokers.d Transient ischemic attack, TIA and/or (ischemic) Cerebrovascular Accident, (i)CVA.*

Table 2 .
Limb Characteristics of All Abdominal Aortic Aneurysm (AAA) Patients Treated With an Anaconda Stent-Graft.

Table 3 .
Changes (Δ, Latter Minus Former Scan Moment) in the Limb Geometry Parameter From Pre-to-Postoperative and During Follow-Up for All Limbs of the Abdominal Aortic Aneurysm (AAA) Patients Treated With an Anaconda Stent-Graft.

Table A1 .
Summary of Thromboembolic Events of the Anaconda Patients, Divided in LGO Group and DLE Groups.

Table B2 .
Limb Characteristics Compared Between the PG, the LGO Group and the DLE Group.
a PG group consists of all patent limbs, i.e. not diagnosed with a thromboembolic event.bLGO group consists of limbs with a limb thrombosis in follow-up; DLE group consists of limbs with a limb embolism in follow-up.c Measured on the preoperative CT of the center lumen line (CLL) trajectory that will be covered by the stent-graft.d The body diameter was included once for each limb (so twice in total).e Measured on first postoperative CT as the CLL length from flow divider to last stent ring.*p-value indicating a difference between PG and LGO or DLE groups when <0.05, which are depicted bold in the table.

Table B3 .
Changes in the Limb Geometry Parameters From Pre-to Postoperative and During Follow-Up, of the PG, the LGO Group and DLE Group.PG and LGO or DLE were compared using Generalized Estimating Equations (GEE) analysis, also revealing odd's ratio's (OR).Normally distributed continuous data are presented as the mean±standard deviation; skewed distributed continuous data are presented as the median (inter quartile range [IQR]); categorical data are given as number (%).Variables that differ between PG and LGO or DLE with a p<0.10 were used for multivariate GEE.Abbreviations: CLL, center lumen line; DLE, distal limb embolization; LGO, limb graft occlusion; MI, myocardial infarction; mm, millimeter; PG, patent group; TI, tortuosity index; 95% CI, 95% confidence interval.
a PG group consists of all patent limbs, i.e., not diagnosed with a thromboembolic event.b