Vascular Closure Devices For Axillary Artery Access: A Systematic Review and Meta-Analysis

Abstract

Purpose:

To evaluate the technical success and complication rates of vascular closure devices (VCDs) in the axillary artery.

Materials and Methods:

MEDLINE and Embase were searched independently by two reviewers to identify observational studies from inception through October 2021. The following outcomes were meta-analyzed: technical success, hematoma, dissection, pseudoaneurysm, infection, and local neurological complications. Complications were also graded as mild, moderate, and severe. A logistic regression evaluating the influence of sheath size for the outcome of technical success rate was performed using individual patient–level data.

Results:

Of 1496 unique records, 20 observational studies were included, totaling 915 unique arterial access sites. Pooled estimates were as follows: technical success 84.8% (95% confidence interval [CI]: 78%–89.7%, I²=60.4%), hematoma 7.9% (95% CI: 5.8%–10.6%, I²=0%), dissection 3.1% (95% CI: 1.3%–7.3%, I²=0%), pseudoaneurysm 2.7% (95% CI: 1.3%–5.7%, I²=0%), infection <1% (95% CI: 0%–5.7%, I²=20.5%), and local neurological complications 2.7% (95% CI: 1.7%–4.4%, I²=0%). There was a significant negative association between sheath size and technical success rate (odds ratio [OR]: 0.87 per 1 French (Fr) increase in sheath size, 95% CI: 0.80–0.94, p=0.0005). Larger sheath sizes were associated with a greater number of access-site complications (adjusted odds ratio [aOR]: 1.21 per 1 Fr increase sheath size, 95% CI: 1.04–1.40, p=0.013).

Conclusions:

Off-label use of VCDs in the axillary artery provides an 85% successful closure rate and variable complication rate, depending on the primary procedure and sheath size. Larger sheaths were associated with a lower technical success and greater rate of access-related complications.

Clinical Impact

Safe arterial access is the foundation for arterial intervention. While the common femoral artery is a well established access site, alternative arterial access sites capable of larger sheath sizes are needed in the modern endovascular era. This article provides the largest synthesis to date on the use of vascular closure devices for percutaneous axillary artery access in endovascular intervention. It should serve clinicians with added confidence around this approach in terms of providing a reference for technical success and complications. Clinically, this data is relevant for patient consent purposes as well as for practice quality improvement in setting safety standards for this access site.

Introduction

Vascular closure devices (VCDs) have decreased the need for manual compression, reducing physician time, patient discomfort, and complications such as access site bleeding. ¹ Whereas these devices are indicated for common femoral artery access, the rise of complex endovascular procedures, iliofemoral occlusive disease, and need for superior approaches for visceral arterial segments have increased their use in the axillary artery. ²

There are risks specific to manual compression of the axillary artery. Whereas it has a moderate diameter that can accommodate large sheath sizes, it lacks a bony protuberance for compression and is near the brachial plexus. The axillary artery has become a frequent access site for complex endovascular aortic repair, percutaneously inserted mechanical circulatory support (MCS) devices, and transcatheter valve implantation (TAVI) due to its diameter conducive to accepting larger sheaths. ³

Although benefits exist for the use of closure devices at the axillary artery, hesitation with their use remains among clinicians, given their off-label classification. ⁴ A synthesis of the evidence is warranted to evaluate the safety of these devices and the influence of factors on closure outcomes. Therefore, we performed a systematic review and meta-analysis to estimate the safety and efficacy of VCDs in the axillary artery.

Materials and Methods

A systematic review and meta-analysis was performed according to the Cochrane Handbook for Systematic Reviews of Interventions. ⁵ MEDLINE and Embase databases were searched for randomized controlled trials (RCTs) and observational studies evaluating VCDs in the axillary artery in adult patients. There was no restriction on article language or publication type (full text article or conference abstract). References of included studies were screened for relevant articles. The search is current to October 14, 2021 (Supplemental Table S1). Research ethics board approval was not required as all data were publicly available.

Records were screened by two reviewers independently and in duplicate, with a third reviewer included for resolution of disagreements, if necessary. For non-comparative observational studies (ie, case series), risk of bias was assessed using a version of the Newcastle–Ottawa scale, ⁶ modified for single-arm observational studies, as published previously. ⁷ Five domains were evaluated, as follows: (1) enrollment of a representative study sample, (2) attrition bias (high risk if >10% attrition), (3) outcome adjustment for length of follow-up, (4) selective outcome reporting, and (5) additional biases. For cohort studies, the CLARITY instrument was used. ⁸ Reported adverse events were classified from each study according to the Society of Interventional Radiology (SIR) Standards of Practice Committee: mild complications involved events requiring no therapy or non-substantial therapy; moderate complications involved those requiring “substantial treatment [such as] intervention under conscious sedation, blood product administration, or overnight admission”; and severe complications required “marked escalation of care [such as] prolonged hospital admission and complex intervention performed requiring general anesthesia.” ⁹

Outcomes of meta-analyses were stratified according to the primary procedures, namely, (1) endovascular procedures, and (2) delayed closure of the arteriotomy from percutaneous insertion and removal of MCS device (eg, intra-aortic balloon pump, Impella devices). Endovascular and MCS groups were separate analyses as patients with mechanical circulatory devices may require the support device in situ for more than 1 day; the device in situ time has been associated with increased vascular complications. ¹⁰ For endovascular procedures, we meta-analyzed efficacy outcomes (technical success) and access site safety outcomes (hematoma, stenosis or occlusion, infection, thrombosis, dissection, pseudoaneurysm, limb ischemia, local neurological complications [paresthesias, pain, weakness; excluding stroke, transient ischemic attack [TIA]). Adverse events were meta-analyzed separately as mild, moderate, and severe, using the SIR classification ⁹ ; these were also summed for the outcome of total complications. For endovascular procedures, we also stratified analyses for technical success and total complications by primary indication (endovascular aortic repair [EVAR], transcatheter aortic valve replacement [TAVR], peripheral arterial disease [PAD] intervention, and other). For percutaneous MCS, a meta-analysis was not possible, given insufficient number of studies, so the outcomes technical success rate and total complications were reported qualitatively.

A logistic regression evaluating the influence of sheath size for the outcome of technical success rate was performed using individual patient–level data. These data were obtained by reconstructing publicly available figures and tables from included studies. This analysis was conducted with individual patient data as it permits evaluation of a clinically important subgroup commonly reported as a continuous variable.

Statistical Analysis

Meta-analysis of single proportions was performed using a random-effects model with inverse-variance weighting. ¹¹ Heterogeneity was quantified using the I² statistic. ¹² The unit of analysis for the meta-analyses was the access site. An exploratory meta-regression was performed to evaluate whether sheath size was associated with the total number of complications, adjusting for primary procedure (PAD, TAVI, EVAR, or other procedure). Publication bias was performed for endovascular procedures by first screening outcomes using Egger’s linear regression test. ¹³ Analyses deemed statistically significant were then evaluated using a funnel plot and, if asymmetric, the trim-and-fill method was used to provide bias-adjusted pooled estimates suspected of publication bias. ¹⁴ Publication bias was not performed for percutaneous MCS, given too few studies available. A p value of less than 0.05 was considered statistically significant. Statistical analysis was performed with packages meta ¹⁵ using R statistical software (Version 4.1.2).

Results

Of 1496 unique records identified, 125 were screened as full-text and 20 studies were included (Figure 1).^{10,16 –34} There were 15 non-comparative observational studies (n=9 high risk of bias, 60%) and 5 cohort studies (n=1 high risk of bias, 20%). Pooled mean age of included patients was 73.6 years. There were 915 unique access sites. Regarding the category of primary procedure, 17 studies were endovascular procedures and 3 were percutaneous MCS implantation and removal. Of device types, 9 studies used Perclose ProGlide only, 2 used a combination of Perclose ProGlide or Perclose ProStar XL, 2 used an unspecified Perclose device only, 1 used Angio-Seal only, 1 used Starclose only, and 5 used a combination of devices. Characteristics of each included study are reported in Table 1. Risk of bias rating and rationale for each domain is reported in Supplemental Tables S2 and S3.

Figure 1.

Flow diagram showing study selection. Twenty studies were included in the systematic review and 17 studies were included in meta-analyses.

Table 1.

Characteristics of Included Studies.

Study	Institution, country	Device	Access sites	Sex	Closure technique	Sheath size (n)
Agrusa et al 16	New York-Presbyterian Hospital/Weill Cornell Medical Center, United States	Perclose ProGlide	46	29 M, 17 F	Pre-close with 2 VCDs	8 Fr (1), 9 Fr (1), 10 Fr (1), 12 Fr (42), and 16 Fr (1)
Bertoglio et al 17	IRCCS San Raffaele Scientific Institute, Italy	Perclose ProGlide	60	47 M, 12 F	Pre-close with 2 VCDs	10 Fr (11), 12 Fr (42), 14 Fr (6), and 16 Fr (1)
Branzan et al 18	University Hospital, Germany	Perclose ProGlide	40	27 M, 13 F	Pre-close with 2 VCDs	12 Fr
Gonen et al 19	Turkey	StarClose	7	5 M, 1 F	Standard	6 Fr (3), 7 Fr (4)
Harris et al 20	Albany Medical Center and St. Peter’s Hospital, United States	Perclose ProGlide, Angio-Seal	25	13 M, 6 F	Not specified, likely pre-close with 1 VCD	4 Fr (1), 5 Fr (2), 6 Fr (6), 7 Fr (15), and 12 Fr (1)
Inglese et al 21	Policlinico San Donato, Italy	Angio-Seal	4	NA	Standard	8 Fr (3), 9 Fr (1)
Meertens et al 22	Epworth Hospital and Royal Melbourne Hospital, Australia	Perclose ProGlide	25	16 M, 7 F	Pre-close with 1 VCD for 24 accesses; 1 required 2 VCDs	6 Fr (1), 7 Fr (11), and 12 Fr (13)
Puippe et al 23	University Hospital Zurich, Switzerland	Perclose ProGlide	20	17 M, 1 F	Pre-close with 1 VCD	Mean size 9.4 ± 1.6
Schäfer et al 24	NA	Perclose ProGlide, Perclose ProStar XL	24	16 M, 8 F	Pre-close with 2 VCD (ProGlide) or 1 VCD (ProStar)	10 Fr
Thawabi et al 25	Newark Beth Israel Medical Center, United States	Angio-Seal, Perclose ProGlide, Mynx	41	13 M, 17 F	Standard	6Fr (37), 7Fr (4)
Abid et al 26	United States	Angio-Seal, Mynx	20	8 M, 10 F	NA	5–8 Fr
Bertoglio et al 27	11 centers	Perclose ProGlide	331	229 M, 102 F	Most patients pre-close with 2 VCDs	≤8 (49), 10–14 (160), and >16 (122)
Chouhan et al 28	Medanta Hospital, India	Perclose (specific device not reported)	4	NA	NA	NA
Deuschl et al 29	University Heart Center, Germany	Perclose ProGlide	12	5 M, 7 F	Pre-close with 2 VCD	NA
Kajy et al 30	NA	Perclose (specific device not reported)	29	24 M, 5 F	NA	NA
Kanei et al 31	University of Massachusetts Memorial Medical Center, United States	Perclose ProGlide	23	NA	Pre-close with 2 VCD (ProGlide)	NA
McCabe et al 10	10 centers	Perclose ProGlide, Angio-Seal, ProStar	46	NA	Various, most patients closure with 2 ProGlide	NA
Ooms et al 32	Erasmus University Medical Center, Netherlands	Perclose ProGlide, Manta	34	NA	Closure with 2 VCD (ProGlide) or Manta	NA
Schäfer et al 33	University Hospital Hamburg-Eppendorf and Asklepios Klinik St. Georg, Germany	Perclose ProGlide, Perclose ProStar XL	100	53 M, 47 F	Pre-close with 2 VCD (ProGlide) or 1 VCD (ProStar)	NA
Wilkins et al 34	Copenhagen University Hospital, Denmark	Perclose ProGlide	24	15 M, 9 F	Pre-close with 2 VCD	8 Fr

Abbreviations: N, number; M, male; F, female; VCD, vascular closure device; Fr, French; NA, not available; IRCCS, Istituto di Ricovero e Cura a Carattere Scientifico.

Meta-analysis for technical success rate, using axillary artery access, was 84.8% (95% CI: 78%–89.7%, I²=60.4%, 14 studies, Figure 2). The following rates were identified with meta-analyses of complications: hematoma (7.9%, 95% CI: 5.8%–10.6%, I²=0%, 10 studies; Figure 3), dissection (3.1%, 95% CI: 1.3%–7.3%, I²=0%, 5 studies; Supplemental Figure S1), stenosis or occlusion (4.0%, 95% CI: 1.4%–10.6%, I²=38%, 7 studies), pseudoaneurysm (2.7%, 95% CI: 1.3%–5.7%, I²=0%, 7 studies; Supplemental Figure S2), and infection (<1%, 95% CI: 0%–5.7%, I²=20.5%, 4 studies; Supplemental Figure S3). Local neurological symptoms excluding stroke or TIA was 2.7% (95% CI: 1.7%–4.4%, I²=0%, 12 studies, Supplemental Figure S4), which improved for the majority of patients at follow-up. Total complication rate was 12.7% (95% CI: 8.1%–19.2%, I²=66.2%, 14 studies; Figure 4). All meta-analyzed outcomes are detailed in Table 2. The remaining outcome forest plots are depicted in Supplemental Figures S1 to S10.

Figure 2.

Forest plot of meta-analysis for technical success. Pooled technical success rate was 85% (95% CI: 78%–90%, I²=60%).

Figure 3.

Forest plot of meta-analysis for hematoma. Pooled hematoma rate was 8% (95% CI: 6%–11%, I²=0%).

Figure 4.

Forest plot of meta-analysis for total complication rate. Pooled total complication rate was 13% (95% CI: 8%–19%, I²=66%).

Table 2.

Pooled Efficacy and Complications for Endovascular Procedures Using Axillary Artery Access.

Outcome	N studies	Pooled risk (95% CI)	I2
Technical success	14	85% (78%–90%)	60%
Hematoma	10	8% (6%–11%)	0%
Artery stenosis or occlusion	7	4% (1%–11%)	38%
Infection	4	<1% (0%–6%)	21%
Thrombosis	3	1% (0%–4%)	20%
Dissection	5	3% (1%–7%)	0%
Pseudoaneurysm	7	3% (1%–6%)	0%
Limb ischemia	5	2% (0%–6%)	0%
Local neurological complication	12	3% (2%–4%)	0%
Mild complications	14	8% (6%–11%)	4%
Moderate complications	14	11% (8%–15%)	15%
Severe complications	13	4% (2%–5%)	0%
Total complications	14	13% (8%–19%)	66%

Abbreviation: CI, confidence interval.

The subgroup analysis for endovascular procedures stratified by the primary indication of the procedure revealed no significant difference (p=0.096) in technical success rate between EVAR (84.0%, 95% CI: 77%–89%, 5 studies), TAVI (80.0%, 95% CI: 58%–92%, 4 studies), PAD (98%, 95% CI: 89%–100%, 2 studies), and other primary indications (89.2%, 95% CI: 75%–96%, 3 studies; Supplemental Figure S11). The subgroup analysis for total complication rate revealed a significant difference (p=0.013) between EVAR (21.2%, 95% CI: 14.2%–30.5%, 5 studies), TAVI (12.2%, 95% CI: 4.9%–27.2%, 4 studies), PAD (4.2%, 95% CI: 1.2%–13.5%, 2 studies), and other primary indications (12.2%, 95% CI: 4.9%–27.2%, 3 studies; Supplemental Figure S12).

Our univariate logistic regression of individual patient data extracted from a single multicenter study of 331 patients ¹⁰ demonstrated a significant negative association between sheath size and technical success rate using axillary access (odds ratio [OR]: 0.87 per 1 French (Fr) increase in sheath size, 95% CI: 0.80–0.94, p=0.0005). The study-level meta-regression showed an independent positive association with sheath size and total number of complications (adjusted odds ratio [aOR]: 1.21 per 1 Fr increase sheath size, 95% CI: 1.04–1.40, p=0.013; Supplemental Figure S13).

Three studies used VCDs to close the arteriotomy site following percutaneous MCS device removal. Studies reporting MCS devices included intra-aortic balloon pump for decompensated heart failure, ²⁸ Impella CP (Abiomed, Danvers, MA, USA) for cardiogenic shock, ³⁰ and a combination of the Impella family of ventricular assist devices and intra-aortic balloon pumps. ¹⁰ A meta-analysis was not possible due to insufficient number of studies. Technical failure rate ranged from 0% to 14% and total complication rate ranged from 0% to 28%. Kajy and colleagues ³⁰ reported the following complications: hematoma 3.4% (1/29), need for stent due to technical failure to closure arteriotomy site 10.3% (3/29), dissection 6.9% (2/29), and ipsilateral arm ischemia 10.3% (3/29). McCabe included 102 patients; however, device explantation involved a VCD in 61% of patients. ¹⁰ The majority of outcomes were attributed to the total number of patients instead of only those with a VCD used, and so were not reported in our study. ¹⁰

Publication bias assessment, using Egger’s linear regression test, detected publication bias in the following outcomes: stenosis or occlusion at access site (p<0.001), mild complications (p=0.01), moderate complications (p=0.0001), and total complications (p=0.0002). The visual assessment of funnel plots for these outcomes confirmed plot asymmetry for stenosis or occlusion (Supplemental Figure S14), mild complications (Supplemental Figure S15), moderate complications (Supplemental Figure S16), and total complications (Figure 5). Using the trim-and-fill method, the following bias-adjusted pooled estimates were constructed: mild complications (adjusted 8.7%, 95% CI: 5.7%–13.2%, vs unadjusted 8.0%, 95% CI: 5.8%–10.9%), moderate complications (adjusted 12.9%, 95% CI: 8.9%–18.3%, vs adjusted 10.9%, 95% CI: 8.0%–14.7%), and total complications (adjusted 20.2%, 95% CI: 14.0%–29.7%, vs unadjusted 13.0%, 95% CI: 8.1%–19.2%). The trim-and-fill model for stenosis or occlusion did not converge, and so a bias-adjusted pooled estimate was not possible to construct.

Figure 5.

Funnel plot for total complication rate demonstrating asymmetry.

Discussion

Our study provides a comprehensive overview of the technical success and complications rates of using off-label VCDs in the axillary artery. Endovascular procedures, including aortic aneurysm repair, incurred the highest complication rate, whereas PAD procedures such as balloon angioplasty and stenting incurred the lowest complication rate; however, there may be confounding variables, given the amount of heterogeneity in the data used. Finally, we identified that larger sheath sizes were associated with a lower technical success rate.

The axillary artery is becoming a common access site in cases with hostile iliofemoral arteries and when superior access is needed for downward facing visceral arteries. Access is typically obtained at the lateral border of the pectoralis minor muscle, between the second and third segments. Common indications for axillary access in our meta-analysis included EVAR, TAVI, insertion of temporary MCS, and PAD procedures. When considering the anatomy, operators should be mindful of its close proximity to the brachial plexus, lack of bony protuberance for compression, risk of pneumothorax, and large hematoma compressing the medial brachial fascial compartment. Our meta-analysis found that, in endovascular procedures using the axillary artery for access, the rates of dissection, pseudoaneurysm, and limb ischemia were each less than 5%. Access site neurological complications, including paresthesias and weakness, were 3%, which importantly improved for the majority of patients at follow-up. Whether differences exist between different segments of the axillary artery remains to be formally examined. Bertoglio and colleagues published a multicenter registry of 331 patients who underwent axillary access for endovascular aortic procedures, whereby the first, second, and third segments of the axillary artery were used 69%, 11%, and 20%, respectively. ²⁷ Sheaths ≥16 Fr were used significantly more frequently in the first versus second or third segments (86% vs 59%, p<0.001). ²⁷ Best practices for percutaneous axillary access are being established, but future studies comparing complication rates between segments of the axillary artery may inform future clinicians of the safest segment for puncture. ³⁵

There are several factors that one must consider when evaluating complication rates with off-label VCDs. Operator expertise, sheath size, primary indication, and device used are several examples. Among these, we explored sheath size, which independently increased the risk of complications. These results were consistent with our expectations, given the generally increased difficulty closing larger arteriotomy sites. However, we do note other factors such as type of procedure that may contribute to confounding this result. Endovascular aortic repair generally requires the use of larger sheath sizes than PAD interventions. This was seen visually in our bubble plot, whereby EVAR studies had higher complication rates and larger sheath sizes, whereas PAD studies had lower complication rates and smallest sheath sizes. Therefore, these results can only be interpreted as hypothesis-driving, warranting larger database studies with more homogeneous data that can examine other factors.

To put axillary artery site complications into perspective, one can perform indirect comparisons with other meta-analyses, evaluating complications of VCDs using common femoral artery access.^36,37 Korney and colleagues performed a systematic review in patients undergoing coronary angiography or percutaneous vascular interventions and reported the following ranges for complications using common femoral artery access: hematoma 0% to 36%, pseudoaneurysm 0% to 10%, and leg ischemia 0% to 2%. ³⁶ Biancari and colleagues performed a meta-analysis in percutaneous coronary intervention and found the following rates: groin hematoma 4.4%, pseudoaneurysm 2.6%, groin infection 0.9%, and lower limb ischemia or arterial stenosis 0.3%. ³⁷ With the exception of access site hematoma formation, these complication rates are comparable to our pooled rates of any hematoma (8%), pseudoaneurysm (3%), and limb ischemia (2%). Notably, all hematomas were included in our meta-analysis, including those that were conservatively managed. These similar safety outcomes predict a promising future for VCDs in the axillary artery, particularly when considering that the complications rates of our meta-analysis were in comorbid patients with hostile lower extremity access or in patients undergoing complex endovascular interventions requiring upper extremity access.

This study has several limitations. First, many of the included studies were non-randomized, retrospective, and some did not report whether a consecutive or convenience sample was included, which increases the risk of selection bias. There were also some studies that included a small sample size, which reduces the precision of their outcomes. Second, there were several subgroups that we were unable to evaluate, given insufficient data granularity. These included patient subgroups (body mass index, anticoagulation status), anatomical subgroups (axillary artery segment and side), and device subgroups (Proglide vs Angioseal). Angioseal, Mynx, and Proglide have different mechanisms of action, which may confer different complication rates. Third, the inclusion of smaller, lower quality studies may have increased the risk of publication bias. Publication bias was detected for the outcomes of stenosis or occlusion, mild complications, moderate complications, and total complications. We addressed this limitation by performing the trim-and-fill method, which provides readers with statistically adjusted estimates of complication rates. Fourth, our literature search included only MEDLINE and Embase and did not include a gray literature search, so it may be possible that not all studies were identified. Finally, we were conservative in terms of data extraction of complications. Studies were required to explicitly state the specific complication in order for it to be included in our meta-analysis, rather than only stating that “no complications occurred.” Whereas we understand this likely overestimates complication rates, as the presence of an outcome is more likely to be reported than the absence of an outcome, we believe that this provides standardization in data extraction and ultimately contributes to the objectivity of our analysis.

Conclusion

This systematic review and meta-analysis provides a comprehensive overview of the technical success and complication rates of off-label VCDs in the axillary artery. Larger sheaths were associated with a lower technical success and greater rate of access-related complications. Further studies are needed to confirm these findings and investigate other clinical factors such as device type and axillary artery access site location.