Abstract
Objective
To evaluate folding in infrarenal stent grafts due to oversizing, barb angle, and barb length using CT images of stent grafts deployed in explanted bovine aortas.
Methods
CT data from an in vitro investigation on the effect of oversizing (4–45%; N=19), barb length (2–7mm; N=11), and barb angle (10–90°; N=7) on device fixation were examined for instances of folding. Folding was classified as circumferential or longitudinal and quantified on an ordinal scale based on codified criteria. Cumulative fold ranking for each deployment were used as the measure of folding observed.
Results
Of the 37 cases, cumulative fold ranking for stent grafts oversized greater than 30% (N=5) was significantly greater than the rest (mean±SD = 3.4±1.7 vs. 0.5±1.2 respecively; Mann Whitney U test p value < 0.005). When barb length was varied from 2 to 7 mm (oversizing held at 10–20%), folding was noted in 1 of 11 cases. Similarly, when barb angle was varied from 0° (vertical) to 90° (horizontal), folding was not noted in any of the 7 cases. The pullout force was not significantly different between stent grafts with and without folding (5.4±1.95 N vs. 5.12±1.89 N respectfully; p>0.5). Of 37 cases, at least one instance of folding was noted in all 7 of 7 stent grafts with oversizing greater than 23.5% and only 5 of 30 stent grafts with oversizing less than 23.5%.
Conclusions
Stent graft folding was prevalent when oversized greater than 30%. Large variations in barb length and angle did not aggravate folding risk when oversized within the recommended range of 10–20%.
INTRODUCTION
Mechanisms involved in the fixation of stent grafts to the proximal aorta are poorly understood. The effectiveness of fixation is likely affected by the characteristics of the aorta at the deployment site (e.g., neck angulation and length, calcification, etc.), choices made by the physician (e.g., oversizing), and characteristics of the device itself (presence or absence of barbs, infrarenal versus suprarenal fixation devices, etc.). Oversizing enhances the frictional forces at the aortic wall to help anchor the device. The oversizing of these stent grafts in patients has been reported to range from 5 to >30%1 with manufacturers recommending 10–20%. Sternbergh et. al reported a 14% device migration rate after 12 months for oversizing greater than 30% compared to 0.9% for oversizing less than 30% in Zenith stent grafts2. Other groups did not find a strong correlation between oversizing and migration3–6. Some studies have suggested that the addition of barbs may decrease likelihood of migration7, 8. But the long-term outcome and benefits of barbs is also poorly understood. Our group performed bench-top studies using multiple Gore Excluder grafts with varying oversizings that were fit with controlled barb lengths and barb angles. They were deployed into biosynthetic AAA phantoms in a physiological flow loop, scanned under high-resolution CT and followed up with mechanical pullout testing. We reported that >30% oversizing results in poor short-term fixation strength and poor barb penetration9. On anecdotal visual observations of CT scans of the deployed graft-aorta complex, folding of the stent graft was noted at high oversizing ranges. Folding leading to poor apposition at the proximal attachment site was proposed as a possible explanation for the poor fixation strength under high oversizing. In a study of 100 consecutive patients with the AneuRx graft, Wolf et al. reported incidences of `eccentric compression' (or folding, as we call it here) in 5 cases10. When these patients were followed for 3 to 38 months (mean, 12 months), graft migration occurred in 4 of the 5 patients (80%) with eccentric compression as opposed to only 5 of remaining 95 patients (5%) without eccentric compression.
A causative relationship between high oversizing and folding is plausible, but remains unclear. Other factors such as barbs may play a role as well. It is conceivable that poor barb penetration – a phenomenon that is all too possible9 – may aggravate folding risk. The objective of this study is to investigate the impact of graft oversizing, barb length and barb angles on stent graft folding at the proximal attachment site using mechanical testing and imaging data collected in a previous in vitro bench-top investigations.
METHODS
An in vitro bench top study was conducted to assess the effect of three stent graft variables on the proximal fixation strength: oversizing (4 to 45% range; N=20), proximal barb length (0 to 7 mm range; N=13) and barb angle (10 to 90° to graft axis; N=7). Oversizing was defined as the percent difference between the stent graft diameter and the systolic diameter of the aorta it was deployed in. Barb length was defined as the length of the barb protruding downward from the stent (see Figure 1). The chosen range of study for barb length of 0 to 7 mm captured the range of barb lengths in commercial devices. Barb angle was defined as the angle that the barb makes with the axis of the graft when freely deployed. The range studied covered the entire range possible from 0° (vertical, hugging the stent) to 90° (horizontal). When one of these three variables was perturbed, the other two variables were maintained constant.
Figure 1.
Illustration of stent graft barb characteristics studied - barb length (BL) and barb angle (BA)
Gore Excluder grafts with custom nitinol barbs varying in barb length and barb angle were fabricated in-house. The barbs were sutured on the second stent layer of the device. The stent grafts were then deployed into a mock flow loop with physiological flow rates and pressures (time-averaged flow rate of ~2.5L/min, pressure of ~80/140 mmHg at 80 beats/min), and an idealized bio-synthetic aneurysm phantom. Descending thoracic bovine aorta obtained from an abattoir was then attached to the synthetic aneurysm phantom. The device was compressed into a catheter and introduced to the flow loop via the iliac bifurcation. The proximal end of the device was deployed into fresh bovine aortic tissue to ensure realistic deployment and barb penetration conditions (Figure 2). Deployments were conducted under continuous flow and devices remained under flow for about 30 seconds to 2 minutes after deployment to ensure stabilization of flow and pressures post deployment. The flow loop was then stopped and the graft-aorta complex was removed and stored in saline at body temperature. Within an hour after removal from the flow loop, high resolution CT scans on a 64-slice Siemens Cardiac Sensation scanner with scan parameters of 100 effective mAs, 140 kilovoltage, pitch of 1, and a rotation speed of 0.5sec were then performed on the graft-aorta complex. A collimation of 64 × 0.6 mm was selected with a reconstructed slice of 0.6 mm width and recon increment of 0.4 mm to obtain the best resolution and achieve isotropic voxel images. Subsequently, the graft aorta complex was subjected to mechanical pullout tests to document the strength of graft-aorta fixation. Detailed description of the apparatus and test methods may be found in our earlier report.9
Figure 2.
Schematic of the pulsatile flow loop. Compliant tubing and pinch valves serve as compliance and resistance chambers for pulse pressure and flow control. The bypass tube helps eliminate unrealistic spiking of pressure when the flow is partially or fully obstructed during EVG deployment.
CT scans of the proximal stent graft deployed in the bovine aorta/AAA phantom were evaluated with rigor for the level of folding. Two forms of visual assessments of the stent graft apposition and deployment characteristics were performed.
-
(1)
The image volume was obliquely sliced based on local graft axis at the fixation site using Syngo CT (Siemens Medical, Malvern, PA), an image processing platform, to visualize the stent and barbs at the proximal end of the graft along its true cross section (Figure 2a). A coronal section was also used where needed.
-
(2)
3D reconstructions of the stent and barbs were generated using Mimics (Materialise Inc., Plymouth, MI), a 3D solid modeling software (Figure 2b). Data visualization software, (Tecplot Inc., Bellevue,WA), was then used for further visualization of the 3D objects. All data analysis was performed on the proximal region of the device (first few stent layers, especially where the barbs generally attach). Folding in the proximal region of the stent graft is of particular concern because it directly impacts its proximal apposition with the aortic wall. Where necessary, 3D models were sliced along the transverse, sagittal, and coronal planes as well to better visualize and confirm folding along the perimeter of the stent.
For control, a 26 mm stent graft with barb length = 7 mm and barb angle = 20° was deployed freely in air and imaged in a similar manner. The control deployment represents the best case scenario with no deformation or folding (see Figure). The images obtained and processed from the control deployment served as the gold standard while studying the stent grafts deployed within the bovine aortic neck of the biosynthetic AAA phantom.
An initial evaluation of the CT data revealed two types of folding: longitudinal and circumferential (see Figure). Longitudinal folding occurs when the stent graft folds along the length of itself (Figure b). Circumferential folding occurs when the stent graft folds into the lumen of the device (Figure c), or when there is overlapping of neighboring stent struts. Folds observed by image processing and visualization were ranked on a scale of 0–3 for severity (see Table 1 for criteria used) by comparing to the control case. Since some stent grafts may have multiple folds, rank values for all folds within a graft (regardless of fold type) were summed to obtain a cumulative folding rank. Ranks for all cases in this study were assigned by the same investigator based on the criteria in Table 1. The investigator was blinded to all oversizing and barb characteristic information during the ranking process.
Table 1.
| Fold Rank | Criteria |
|---|---|
| 0 | No evidence of folding in stent. |
| 1 | Minimal folding. Slight overlap of neighboring stents. General circular shape of stent still retained. Slight bend in length of device. |
| 2 | Moderate folding. More noticeable overlap of neighboring stents. Circular stent configuration has been compromised in certain areas and is now irregular in shape. More noticeable bend in length of device. |
| 3 | Severe folding. Stent configuration is completely deformed. Severe stent overlap of neighboring stents. Extreme protrusion into lumen or against aortic wall of stent. Severe bend in length of device. |
RESULTS
Forty stent grafts were deployed with various oversizings, barb lengths and barb angles followed by imaging and pullout testing. Of these, we failed to obtain CT scans in three cases resulting in 37 cases used in this analysis. A summary of CT imaging data of deployments used to assess folding is provided in Table 2. In these cases, the CT scan volumes were processed to characterize graft folding on an ordinal scale as laid out in Table 1. In the study population, folding of most ranks (0 to 3) and types (circumferential and longitudinal) were noted. Representative examples for various ranks and types of folding may be found in Figure.
Table 2.
| Parameter Perturbed | Range | Sample Size |
|---|---|---|
| Oversizing (with BL=5mm, BA=20°) | 4–10% | 4 |
| 11–20% | 5 | |
| 21–30% | 6 | |
| >30% | 4 | |
| Barb Length (BL) (with oversizing=10–20%, BA=20°) | 2 | 2 |
| 3 | 2 | |
| 4 | 2 | |
| 5 | 1 | |
| 6 | 2 | |
| 7 | 2 | |
| Barb Angle (BA) (with oversizing= 10–20%, BL=2mm) | 20 | 2 |
| 30 | 1 | |
| 40 | 1 | |
| 50 | 1 | |
| 70 | 1 | |
| 80 | 1 | |
| 90 | 1 |
The relationship between cumulative fold level and oversizing, barb length and barb angle are shown in Figure. Of the 37 cases, 12 deployments showed one or more folds (4 mild cases, 5 moderate cases, 3 severe cases). Eleven of these 12 were from the study where oversizing was perturbed. Three out of the 12 fold cases (38%, 13%, 38% oversizing) experienced 2 folds and 1 (30% oversizing) experienced 3 folds. In total, 17 folds were identified in 12 stent graft deployments with 4 (23%) of these longitudinal and 13 (77%) circumferential. Among the 37 deployments, at least one instance of folding was noted in all 5 of 5 stent grafts with oversizing greater than 30% There was one case of severe folding (cumulative fold level of 6) in a 13% oversized stent graft, a suspected outlier in the data (see Figure). Cumulative fold rankings for oversizing greater than 30% (N=5) were higher than oversizing less than 30%(N=7) with statistical significance (mean±SD = 3.4±1.7 Versus 0.5±1.2 respecively; Mann Whitney U test p value < 0.005). The results were similar with statistical significance when comparing experiments where only oversizing was perturbed were used for comparison between oversizing greater and less than 30%. (3.4±1.7 Versus 1.1±2; p < 0.05). When barb length was varied from 2 to 7 mm (oversizing held at 10–20%), folding was noted in 1 of 11 cases (mild folding in a stent graft with a 7 mm barb). When barb angle was varied from 0° (vertical) to 90° (horizontal), folding was not noted in any of the 7 cases. Pullout force data was available for assessment of the relationship between graft-aorta attachment strength and folding. Pullout force was not significantly different between stent grafts with and without folding (5.4±1.95 N Versus 5.12±1.89 N respectfully; p>0.5).
DISCUSSION
Graft folding may result in poor apposition between the stent graft and aortic wall, a phenomenon whose impact on short-term fixation strength and long-term endoleak and/or migration risk remains poorly understood. Wolf et al. reported 80% incidence of migration in subjects with AAA stent graft folding post-op (they referred to it as `eccentric compression') compared to 5% incidence in those without.10 Incidences of folding and their association with oversizing has also been noted in thoracic endografts. In a review of published cases by Jonker et al.13, 60 subjects with incidences of folding post deployment were identified (1day – 79 months). Oversizing of the endografts was 26.7% ±12.0% (range 8.3%– 60%) with excessive oversizing the cause for 20% of folding observed. Canaud et al.14 also reported on four cases of folding in a retrospective study involving 285 thoracic endovascular repairs. Of the four patients with endovascular graft folding, three were oversized greater than 20%. Empirical data from patient outcome studies are indispensable for gaining a true understanding of graft folding and its effects on migration. But measurements made in such studies are unlikely to be detailed enough to reveal the mechanisms involved in folding or be sufficiently controlled for a deterministic understanding of this process and its association with fixation strength. In vitro bench-top studies, on the other hand may allow for controlled investigations, although admittedly suffer from limited realism to the physiological situation. Further, bench-top studies at best only provide insights into short-term aspects of deployment rather than long-term outcomes. Thus, bench-top studies compliment empirical patient outcome studies by driving hypotheses that may be tested prospectively and/or clarifying findings in retrospective data. This bench-top investigation was motivated by patient outcome reports noting empirical associations between stent graft oversizing, folding and migration rates10, 13, 14. We have assessed some of the factors that may conceivably be associated with and causative of folding and its effects on attachment strength in a retrospective analysis of data obtained from our bench-top investigation. Quantifying folding levels is beneficial since it captures the level of apposition to the host vessel and may have implications on device migration or endoleak. Folds may form a “ledge” in the perimeter of the device resulting in an elevated axial force on the device from blood flow which can aggravate migration risk.
A link between oversizing and folding in stent grafts is to be expected. For an oversized stent graft to fit inside an aorta, it needs to either expand the aorta or compress itself, or both. If the level of compression breaches some threshold, it will likely fold at one or more spots. Thus the precise relationship between oversizing and folding is likely nonlinear. Anecdotal observations in an earlier study by our group suggested that folding occurs particularly beyond 30% oversizing. This study was undertaken to more definitively verify those observations with some level of quantification. Our analysis of detailed image volumes of the deployed stent graft show that oversizing beyond 30% does result in folding. Indeed, folding was observed in all grafts oversized greater than 23.5%. But for one suspected outlier where a 13% oversized graft showed severe folding (cumulative fold level of 6; also see Figure d & g and Figure), folding was minimal or non-existent in stent grafts oversized less than 23.5%. This suggests that the aorta is not expanding to accommodate the stent graft, but rather forces it to fold when oversized greatly. This is only understandable when one considers the nonlinear stiffening of the aorta at higher pressures and that the additional radial pressure from the stent graft onto the aortic wall is about 30 mmHg even when oversized greatly. We performed a quick finite element simulation of the dilation of a circular 2D cross-section of a human aorta (1.9 cm inner diameter; 1.5 mm thick; finite elastic tissue constitutive model) under mean aortic pressure (100 mmHg) followed by the additional radial pressure by 15% and 40% oversized grafts. The aorta inflates by an additional 0.8% from mean aortic pressure to deployment of a 15% oversized graft and by an additional 1.1% for a 40% oversized graft. Clearly, oversizing greatly will invariably result in folding of the stent graft as noted in our experiments. In an in vitro study, Schurink et. al. also reported some positive correlation between oversizing of Gianturco and Palmaz stent grafts and the size of the largest fold in these grafts15.
It is conceivable that barbs interfere with and perhaps aggravate folding risk especially when they do not penetrate the wall, but rather protrude from the wall and push against it. However, results here suggest that barbs in stent grafts oversized in the 10–20% range, irrespective of their length (2–7 mm) or angle (10°–90° with respect to graft axis) do not aggravate folding risk. This is irrespective of whether and by how much the barbs penetrated into the aortic wall. The study did not investigate whether barbs are similarly of no consequence in folding at other oversizing ranges. Interestingly, stent grafts that had folding did not also have lower pullout strength when compared to those with no folds. Caution is warranted in interpreting this to mean that folding is unlikely to affect short-term attachment strength mainly because pullout strength was also affected by other factors such as oversizing and active fixation effects from barbs that were perturbed independently in this study. Some limitations in this study are worth noting. Fold was quantified on an ordinal scale by visual observation by one investigator. However, all observations were performed by leveraging advanced 3D reconstruction methods by a single investigator and fold severity determination was based on comparative level to the condition of all stent grafts within the 37 devices, especially the control deployment. Further, the investigator was blinded from information on oversizing, barb length and angle for a given stent graft whose images were being assessed. Therefore, we are confident that the findings are valid, especially for comparative purposes.
The results observed are also only representative of short-term cases as the devices remained in the flow loop for a very short period of time. If the devices remained under flow conditions for a longer period of time, infolding may become more severe but barb embedment may be more efficient. Although we did not find a statistically significant difference in pullout forces between devices with and without folds, there may be long term implications of this occurrence. As previously mentioned, devices with folding do not have proper apposition to the host vessel and may lead to a lower threshold attachment strength immediately post deployment. Overtime, this could put the patient at higher risk for complications associated with migration. Migration is also a very complex occurrence because of all the factors that influence it (oversizing, barb characteristics, etc.). No conclusive statement can be made regarding the role of folding in migration from this study. Furthermore, because this study was also a retrospective analysis of data gathered from experiments designed to assess how pullout strength was affected by oversizing, folding was not varied in a highly control manner. Studies focused on perturbing other potential causative factors in a controlled manner may provide further insights.
CONCLUSION
In this in vitro investigation, stent graft folding was found to be prevalent when oversized greater than 30%. Within the recommended range of 10–20%, large variations in barb length and angle did not aggravate folding risk.
Figure 3.
(a) Visualization of CT slices on Syngo CT using rotations and oblique slicing; (b) 3D reconstruction, volume rendering and visualization of CT image volume using Mimics software.
Figure 4.
(a) Cross sectional view of 26mm diameter stent graft deployed freely in air. (b) Front half of stent graft after coronal slice. (c) Back half of stent graft after coronal slice.
Figure 5.
Schematic of (a) proper apposition of stent graft in aorta with no folding, (b) longitudinal folding, and (c) circumferential folding
Figure 6.
Representative examples of various ranks and types of folding observed. There were no cases of mild longitudinal folding. Dotted lines on long views indicate the location of cross sectional slice shown above them. Block arrows indicate regions of folding observed. Note that the deployed stent configuration for rank 0 is as good as that in the control (Figure). Circumferential folding is characterized by overlap in the perimeter highlighted by the stent struts on cross-sectional views. Longitudinal folding is characterized by a gap in the perimeter highlighted by the stent struts on cross-sectional views.
Figure 7.
Folding levels observed due to variations in oversizing, barb length and barb angle. When one variable was perturbed, the others were maintained constant. Note that among the three variables, folding was mainly affected by increased oversizing, but for one outlier at 13% oversizing. When oversized within 10–20%, barb length and angle do not cause folding.
ACKNOWLEDGMENTS
This study was funded by NIH/NHLBI grant # R15 HL087642-01: Study of design variables of endovascular graft. The authors are grateful to W.L. Gore and Associates for donating stent graft samples that were modified and used in this study. The authors also thank Dr Zhonghua Li (Cordis Endovascular) and Dr Jafar Golzarian (University of Minnesota) for technical and clinical insights during the study.
Supported by NIH/NHLBI # R15 HL087642-01: Study of design variables of endovascular graft (to Raghavan)
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
REFERENCES
- 1.van Prehn J, Schlosser FJ, Muhs BE, Verhagen HJ, Moll FL, van Herwaarden JA. Oversizing of aortic stent grafts for abdominal aneurysm repair: a systematic review of the benefits and risks. J Vasc Surg. 2009;39(1):20–6. doi: 10.1016/j.ejvs.2009.03.025. [DOI] [PubMed] [Google Scholar]
- 2.Sternbergh WC, 3rd, Money SR, Greenberg RK, Chuter TA. Influence of endograft oversizing on device migration, endoleak, aneurysm shrinkage, and aortic neck dilation: results from the Zenith Multicenter Trial. J Vasc Surg. 2004;39(1):20–6. doi: 10.1016/j.jvs.2003.09.022. [DOI] [PubMed] [Google Scholar]
- 3.Mohan IV, Harris PL, Van Marrewijk CJ, Laheij RJ, How TV. Factors and forces influencing stent-graft migration after endovascular aortic aneurysm repair. J Endovasc Ther. 2002;9(6):748–55. doi: 10.1177/152660280200900606. [DOI] [PubMed] [Google Scholar]
- 4.Zarins CK, Bloch DA, Crabtree T, Matsumoto AH, White RA, Fogarty TJ. Stent graft migration after endovascular aneurysm repair: importance of proximal fixation. J Vasc Surg. 2003;38(6):1264–72. doi: 10.1016/s0741-5214(03)00946-7. discussion 72. [DOI] [PubMed] [Google Scholar]
- 5.Sampaio SM, Panneton JM, Mozes G, Andrews JC, Noel AA, Kalra M, et al. AneuRx device migration: incidence, risk factors, and consequences. Ann Vasc Surg. 2005;19(2):178–85. doi: 10.1007/s10016-004-0166-7. [DOI] [PubMed] [Google Scholar]
- 6.Conners MS, 3rd, Sternbergh WC, 3rd, Carter G, Tonnessen BH, Yoselevitz M, Money SR. Endograft migration one to four years after endovascular abdominal aortic aneurysm repair with the AneuRx device: a cautionary note. J Vasc Surg. 2002;36(3):476–84. doi: 10.1067/mva.2002.126561. [DOI] [PubMed] [Google Scholar]
- 7.Resch T, Malina M, Lindblad B, Malina J, Brunkwall J, Ivancev K. The impact of stent design on proximal stent-graft fixation in the abdominal aorta: an experimental study. Eur J Vasc Endovasc Surg. 2000;20(2):190–5. doi: 10.1053/ejvs.1999.0991. [DOI] [PubMed] [Google Scholar]
- 8.Malina M, Lindblad B, Ivancev K, Lindh M, Malina J, Brunkwall J. Endovascular AAA exclusion: will stents with hooks and barbs prevent stent-graft migration? J Endovasc Surg. 1998;5(4):310–7. doi: 10.1177/152660289800500404. [DOI] [PubMed] [Google Scholar]
- 9.Kratzberg JA, Golzarian J, Raghavan ML. Role of graft oversizing in the fixation strength of barbed endovascular grafts. J Vasc Surg. 2009;49(6):1543–53. doi: 10.1016/j.jvs.2009.01.069. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Wolf YG, Hill BB, Lee WA, Corcoran CM, Fogarty TJ, Zarins CK. Eccentric stent graft compression: an indicator of insecure proximal fixation of aortic stent graft. J Vasc Surg. 2001;33(3):481–7. doi: 10.1067/mva.2001.112322. [DOI] [PubMed] [Google Scholar]
- 11.Antiga L, Ene-Iordache B, Remuzzi A. Computational geometry for patient-specific reconstruction and meshing of blood vessels from MR and CT angiography. IEEE Trans Med Imaging. 2003;22(5):674–84. doi: 10.1109/TMI.2003.812261. [DOI] [PubMed] [Google Scholar]
- 12.Piccinelli M, Veneziani A, Steinman DA, Remuzzi A, Antiga L. A framework for geometric analysis of vascular structures: application to cerebral aneurysms. IEEE Trans Med Imaging. 2009;28(8):1141–55. doi: 10.1109/TMI.2009.2021652. [DOI] [PubMed] [Google Scholar]
- 13.Jonker FH, Schlosser FJ, Geirsson A, Sumpio BE, Moll FL, Muhs BE. Endograft collapse after thoracic endovascular aortic repair. J Endovasc Ther. 2010;17(6):725–34. doi: 10.1583/10-3130.1. [DOI] [PubMed] [Google Scholar]
- 14.Canaud L, Alric P, Desgranges P, Marzelle J, Marty-Ane C, Becquemin JP. Factors favoring stent-graft collapse after thoracic endovascular aortic repair. J Thorac Cardiovasc Surg. 2010;139(5):1153–7. doi: 10.1016/j.jtcvs.2009.06.017. [DOI] [PubMed] [Google Scholar]
- 15.Schurink GW, Aarts NJ, van Baalen JM, Schultze Kool LJ, van Bockel JH. Stent attachment site-related endoleakage after stent graft treatment: An in vitro study of the effects of graft size, stent type, and atherosclerotic wall changes. J Vasc Surg. 1999;30(4):658–67. doi: 10.1016/s0741-5214(99)70105-9. [DOI] [PubMed] [Google Scholar]