Use of Transcarotid Artery Revascularization, Transfemoral Carotid Artery Stenting, and Carotid Endarterectomy in the US From 2015 to 2019

This cohort study examines the pattern of use, outcomes, and patient and disease factors associated with transcarotid artery revascularization (TCAR), transfemoral carotid artery stenting (TFCAS), and carotid endarterectomy (CEA) in the US from 2015 to 2019.


Introduction
Carotid endarterectomy (CEA) has historically been considered the standard operative therapy for carotid artery stenosis for both symptomatic 1-3 and asymptomatic patients. 4,5 As best medical therapy has evolved, however, the role of carotid revascularization in asymptomatic patients has become less clear. [6][7][8] In 2004, Yadav et al 9 published the results of the SAPPHIRE (Stenting and Angioplasty With Protection in Patients at High Risk for Endarterectomy) randomized clinical trial, which established that transfemoral carotid artery stenting (TFCAS) with an embolic protection device was not inferior to CEA among patients with severe carotid artery stenosis and coexisting conditions, and continued safety was preserved in long-term follow-up. 10 Although the evolution of best medical therapy was associated with decreasing overall rates in carotid revascularization between 2000 and 2016, TFCAS as a proportion of overall revascularization rates increased since the SAPPHIRE trial was published in 2004. 11 In 2015, Kwolek et al 12

published the results of the ROADSTER (Safety and Efficacy Study for
Reverse Flow Used During Carotid Artery Stenting Procedure) multicenter trial, which established that transcarotid artery revascularization (TCAR) was also safe and effective, with a low overall stroke rate at 30 days (1.4%) and durable outcomes at 1-year postoperative follow-up. 13,14 These results led to the first US Food and Drug Administration (FDA)-approved TCAR device. 15 Since this approval, there has been a rapid increase in TCAR adoption, 7 particularly given that contemporary data from the Vascular Quality Initiative (VQI) have shown excellent results. [16][17][18][19] The goal of this cohort study was to quantify the temporal changes in the operative approach to carotid revascularization (CEA vs TFCAS vs TCAR) since the approval of the first TCAR device, and to identify patient and disease characteristics commonly associated with each approach. A secondary aim was to evaluate the association of outcomes with year of surgery, with additional subanalyses in high-volume vs low-volume TCAR centers.

Study Population
Using the VQI database, we identified patients with carotid artery stenosis who underwent CEA, TFCAS, or TCAR from January 1, 2015, through December 31, 2019. Institutional review board approval was obtained from The Johns Hopkins University School of Medicine, which waived the patient informed consent requirement because this study was a retrospective analysis of publicly available deidentified data. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 20 At the time of study design, VQI data were available through September 2020. Because we aimed to evaluate the temporal changes of carotid revascularization in the context of disease and patient characteristics, we truncated the analysis period to before the COVID-19 pandemic. Patients with missing baseline characteristics (n = 488) or operative approach (n = 418) were excluded.

Definitions
The patient comorbidities reported were defined by the VQI.

Evaluation of Temporal Changes
The primary outcomes of this analysis were the number and proportion of carotid revascularizations by operative approach (CEA, TFCAS, or TCAR). Patients were grouped by operative approach and analyzed by the month and year of surgery. Changes in monthly and annual case volumes were captured graphically using heat maps and scatterplots. Because different institutions participate in the VQI each year, the data cannot be considered to be a true cross-sectional representation of the national patterns overall. Therefore, the proportions of all carotid revascularization procedures by operative approach were evaluated. These proportions were captured graphically using stacked bar charts and scatterplots with linear splines for all included patients and for those in the high-risk category alone. Overall yearly temporal changes in carotid revascularization approach were then compared using a nonparametric Wilcoxon-type test for 3 independent groups (Cuzick test), and individual years were compared using χ 2 statistics. A similar approach was used to explore temporal changes in a subgroup of patients at high risk for stroke, cranial nerve injury, or cardiovascular events as a sensitivity analysis.

Statistical Analysis
Patient characteristics (age, self-reported race and ethnicity [which were reported by the centers and included American Indian or Alaska Native, Asian, Black, Hispanic, Native Hawaiian or Other Pacific Islander, White, more than 1 race, and unknown or other]), sex, insurance payer, functional status, smoking status, and comorbidities) and carotid disease characteristics (degree of stenosis and symptomatic status) were divided by operative approach (CEA, TFCAS, or TCAR) and compared using χ 2 statistics. Multinomial logistic regression was used to identify the independent patient and disease characteristics associated with each approach. These characteristics are reported as relative risk ratios (RRRs) for an approach given 3 possible approaches, with CEA considered to be the reference. We identified covariates for the multinomial model a priori, and after confirming no collinearity, we included the following covariates in the multivariable model: age, sex, race and ethnicity, insurance status, comorbidities (hypertension, coronary artery disease, congestive heart failure, chronic obstructive pulmonary disease, diabetes, and chronic kidney disease [CKD] or hemodialysis), functional status (fully functional or not fully functional), smoking status (never, previous, current), high-risk vs standard-risk status, degree of stenosis (moderate grade or high grade), symptomatic status, and year of surgery.
We created clusters by hospital to account for possible hospital-based practice variation. We then performed a sensitivity analysis including only patients who underwent carotid revascularization procedures in 2019 to capture the contemporary state of therapy choice to assess whether associations existed immediately after FDA approval of TCAR. Covariates included in the sensitivity analysis were the same as those used in the main multivariable model with the exception of year of surgery, which was removed because all of these procedures were performed in 2019.
Additional analysis was performed to compare overall in-hospital outcomes of carotid revascularization (stroke, MI rate, and mortality) over time, and these outcomes were then stratified by TCAR volume status. In the overall analysis, simple linear regression with computation of 95% CIs was performed to compare these outcomes during the study period, and the F test was used to test a slope that was not equal to 0. To ascertain whether a center was designated as high or low volume, a frequency analysis was performed that tallied the number of TCARs that each center performed from 2015 to 2019. A histogram and a violin plot were used to represent these data graphically.
Centers that performed more than 40 TCARs from 2015 to 2019 were identified as high volume.
In-hospital outcomes for high-volume and low-volume TCAR centers were then compared over time.
Analyses consisted of testing the association between year of surgery, TCAR volume status, and in-hospital outcome using unpaired, 2-tailed t tests for multiple comparisons. We also compared in-hospital complications directly using t tests without including year of surgery.
A 2-sided P < .05 was considered to be significant for all analyses. Statistical analysis was performed with Stata, version 17 (StataCorp LLC), and figures were created using GraphPad Prism, version 9.3.0 (GraphPad Software). Data were analyzed from January to April 2022.

Temporal Changes in All Operative Approaches
The total number of carotid revascularization procedures included in the VQI database per year increased significantly over the study period (16 754  In a sensitivity analysis limited to patients considered to be high risk for the CEA approach on the basis of the Centers for Medicare & Medicaid Services criteria, the temporal changes observed for CEA, TFCAS, and TCAR were greater. In 2015, CEA was used in 51.6% of all carotid revascularization cases, followed by TFCAS (45.6%) and TCAR (2.8%). In 2019, the prevalence of CEA use decreased to 20.3% of cases, TFCAS use decreased to 24.6%, and TCAR use increased significantly to 53.2% (P < .001) ( Figure 3A). Expressed on a per year basis, the proportional use of CEA for patients with high risk decreased by 7.8% (95% CI, −11.9% to −3.8%), TFCAS decreased by 4.8% (95% CI, −9.5% to −0.14%), and TCAR increased by 12.6% (95% CI, 7.1%-18.1%) ( Figure 3B).

Patient Covariates Associated With Each Operative Approach
Multinomial logistic regression was performed to identify independent patient covariates associated with each operative approach (

Temporal Changes in Patient Outcomes
Unadjusted MI, stroke, and mortality rates were tabulated across all operative approaches in the VQI database from January 1, 2015 to December 31, 2019 (eFigure 1 in the Supplement). Linear regression revealed no cohortwide association between year of surgery and in-hospital MI, stroke, or mortality.
Centers were dichotomized into a high-volume group or low-volume group, and 45 centers were identified as having performed more than 40 TCAR procedures from 2015 to 2019 after the frequency distribution of TCAR procedures by center was analyzed (eFigure 2 in the Supplement). Overall, in-hospital carotid revascularization outcomes (including TFCAS and CEA) of these highvolume and low-volume centers were compared over the study period. In this cohort-level analysis, there was no association between TCAR volume status and in-hospital outcomes. Specifically, there was no association between year of surgery; TCAR volume status; and mortality, stroke, or MI risk (eFigure 3A in the Supplement). There was also no association detected between frequencies of these in-hospital complications and TCAR volume status when outcomes were analyzed in aggregate (eFigure 3B in the Supplement).

Discussion
This study found a 13% annual increase in the number of carotid revascularization procedures performed from 2015 to 2019 within the VQI database. In 2018, TCAR surpassed TFCAS as the dominant stenting modality and, in 2019, it was used 65% more often than TFCAS (21.9% vs 13.3%).
Even after controlling for patient disease characteristics using multinomial logistic regression, the year of surgery remained associated with the operative approach, suggesting that this change was associated with external (nonpatient) temporal factors. Although the proportion of cases that used   TCAR increased each year, patient risk status was the single most important characteristic associated with a stenting approach (ie, TCAR and TFCAS), highlighting the perceived importance of carotid stenting therapies in high-risk patient populations.
The temporal changes in CEA, TCAR, and TFCAS use reported in this study are largely in accordance with the findings in contemporary literature. We observed a higher number of revascularization procedures with time, which is aligned with the increasing age of the US population, 22 the disproportionate burden of carotid artery stenosis in older adults, and a greater number of institutions subscribing to the VQI. 23 Because the number of participants in the VQI changes each year, we focused on the proportion of included cases instead of the crude totals. We observed a 5.0% annual decrease in the proportion of carotid revascularization procedures performed with CEA. Multiple studies have illustrated decreasing CEA rates from 1999 to 2015, with 1 study demonstrating a decrease in the national CEA rate from 298 per 100 000 beneficiary-years to 128 per 100 000 beneficiary-years between 1999 and 2014. 24 [26][27][28][29] which is often believed to be the main source of atheroembolic events during TFCAS. 30 In retrospective analyses, TCAR has demonstrated comparable rates of stroke and death but half the risk of in-hospital transient ischemic attack, stroke, and mortality compared with TFCAS. 14,31,32 A meta-analysis of 30-day mortality rates with TCAR and TFCAS reported that TCAR had lower perioperative stroke and death rates compared with CEA. 33 However, there are currently no randomized clinical trials comparing outcomes for TCAR vs TFCAS or CEA, and long-term (>1 year) data are lacking. Overall, TCAR has been rapidly adopted by physicians who report data to the VQI because of its favorable safety profile compared with TFCAS, and this approach is becoming important for patients at high risk for stroke, cranial nerve injury, or cardiovascular events. We expect that the use of TCAR will continue to expand now that its application in patients with normal risk was approved by the FDA in May 2022. 8,34 Chronic kidney disease and hemodialysis were associated with CEA over either stenting approach after adjustment. Patients with CKD or hemodialysis were typically excluded from the major clinical revascularization trials. To our knowledge, the only randomized clinical trial that specifically evaluated the association of CKD with carotid revascularization outcomes in a post hoc analysis was the North American Symptomatic Carotid Endarterectomy Trial, which demonstrated that patients with CKD had similar rates of perioperative stroke and death but higher rates of cardiac events compared with patients with preserved kidney function. 35 Given that the major benefit from TFCAS is reduced risk of MI, 36,37 these findings suggest that carotid stenting may be beneficial for patients with CKD. There are no data to support this theory, however. In contrast, data from the Nationwide Inpatient Sample have shown that TFCAS vs CEA is associated with a higher risk of major adverse cardiovascular or cerebrovascular events in patients with CKD, 38 although these data are based on in-patient observations and do not account for degree of stenosis. In addition, the results from the present analysis suggest that the perceived benefit associated with carotid stenting in patients with CKD is low. It is possible that use of preoperative computed tomography angiography limits the application of TFCAS and TCAR in populations with CKD, as contrast carries a potential risk of contrast-induced acute kidney injury. 39 It is also possible that patients with CKD have more calcific

JAMA Network Open | Surgery
Use of TCAR, TFCAS, and CEA in the US From 2015 to 2019 carotid disease that limits the use of stenting technology. Future analyses of VQI data or similar robust data are necessary to better understand the risks and benefits associated with TFCAS and TCAR for patients with CKD.
The temporal changes in operative approaches that we observed were not associated with overall changes in major in-hospital outcomes. These data are aligned with increasing reports that TCAR performs relatively similarly to CEA. 18,32,[40][41][42] Given that TCAR use increased and CEA use decreased in the 2015 through 2019 period, it is encouraging to see a lack of change in major complications. We attempted to analyze center-level major complications by stratifying according to TCAR use, but we found no difference in outcomes overall or on an annual basis for high-volume or low-volume TCAR centers. A similar analysis evaluating major adverse cardiovascular events after TCAR and CEA found that centers that adopted TCAR had a 10% decrease in the likelihood of major adverse cardiovascular events at 12 months after TCAR adoption compared with centers that continued to perform CEA alone. 40 The results of the present analysis and previous studies 18,32,[40][41][42] suggest that the introduction of TCAR is associated with stable or slightly improved changes in major in-hospital outcomes after carotid revascularization. Thus, in the future, we expect to see TCAR incorporated into major professional guidelines about the management of extracranial carotid disease.

Limitations
This study has limitations. The analysis focused on pre-COVID-19 temporal changes to understand practice patterns in a nonpandemic era. The VQI includes a subset of hospitals and surgical centers that choose to participate; thus, this analysis does not represent a true cross-sectional cohort of carotid revascularization practices in the US. The number of carotid revascularization procedures increases year after year, but this increase may be partly attributed to the inclusion of more centers performing carotid revascularization in the study. For this reason, we focused on the proportion of each approach overall and on patients at high risk for stroke, cranial nerve injury, or cardiovascular events, but the general application of the findings across centers that do not participate in the VQI is unclear. In addition, it is impossible the know a surgeon's rationale for choosing a procedure over another. We evaluated patient covariates that we believed may play a role in this decision, but the analysis may be subject to residual confounding.
There are multiple other studies that report on the outcomes of CEA, TFCAS, and TCAR. 40,43 We assessed outcomes, including MI, stroke, and mortality, over the study period among all participants in the VQI, and we then dichotomized the centers by high or low volume to investigate the outcomes at centers with early TCAR adoption. Although this analysis did not reveal a detectable association at the cohort level, power was limited by the rarity of these outcomes and was dependent on the number of years in the study, which was small. Future work may include an interrupted time series analysis of the association between choice of operative approach and outcomes given the recent FDA label expansion 44

Conclusions
This cohort study showed that since the first TCAR device received FDA approval in 2015, TCAR has overtaken TFCAS as the primary carotid stenting approach in all patients, and TCAR has overtaken TFCAS and CEA as the dominant operative approach in patients at high risk for stroke, cranial nerve injury, or cardiovascular events. The main patient characteristic associated with carotid stenting