Endovascular Stenting for Acute Ischemic Stroke Due to Cervical Artery Dissection: Nationwide Study of Stroke Recurrence

Abstract

Background

For patients with acute ischemic stroke (AIS) due to cervical artery dissection (CAD), endovascular stenting can improve and maintain vessel patency for severely stenotic or occlusive lesions. Athough prior reports have suggested equivocal functional outcomes with stenting versus no stenting for AIS‐CAD, there are currently no data on stenting and stroke recurrence.

Methods

This was a retrospective cohort study of the 2016–21 Nationwide Readmissions Database in the United States. Adult patients hospitalized for AIS‐CAD were included. Patients who did and did not undergo endovascular stenting were matched using propensity scores calculated using logistic regression models accounting for demographics, stroke severity, and comorbidities. The primary end point was recurrent ischemic stroke within 180 days. Secondary end points include outcomes of the index AIS‐CAD hospitalization such as routine discharge to home without rehabilitation needs, death, and intracranial hemorrhage.

Results

A total of 20 434 patients were included; 2368 (11.6%) underwent endovascular stenting. After propensity score matching, 2215 patients remained in the no stenting group, and 2244 remained in the stenting group. The stenting group had a significantly lower stroke recurrence risk than the no‐stenting group within 180 days (hazard ratio 0.50 [95% CI: 0.27–0.95], P = 0.034). Among patients with at least 180 days of follow‐up, the rate of recurrent stroke was 1.2% in the stenting group, which was significantly lower than 3.5% in the no‐stenting group (P = 0.017). During the index AIS‐CAD hospitalization, stenting was not associated with a different rate of routine discharge (43.8% versus 44.8%, P = 0.65) or death (8.7% versus 7.3%, P = 0.34); however, it was associated with a higher rate of intracranial hemorrhage (21.9% versus 19.0%, P = 0.027).

Conclusion

For patients with AIS‐CAD, endovascular stenting was associated with a lower rate of stroke recurrence. Although stenting was associated with a higher rate of intracranial hemorrhage, this phenomenon did not culminate in different rates of short‐term morbidity or mortality.

Nonstandard Abbreviations and Acronyms

AIS

acute ischemic stroke

CAD

cervical artery dissection

NRD

Nationwide Readmissions Database

PSM

propensity score matching

SMD

standardized mean difference

Clinical Perspective

• This retrospective nationwide study found that for patients with acute ischemic stroke due to cervical artery dissection, endovascular stenting was associated with a lower rate of stroke recurrence.

• Our findings further support the clinical effectiveness of acute stenting for select patients with acute ischemic stroke due to cervical artery dissection.

Cervical artery dissection (CAD) is a rare cause of acute ischemic stroke (AIS), accounting for approximately 2% of cases. ¹ , ² Although most CAD lesions heal over time, patients with AIS‐CAD are at elevated risk of recurrent stroke. ¹ , ² . More important, unlike other common causes of stroke that have a predilection for the elderly population (eg, atrial fibrillation, atherosclerotic disease, and so forth), AIS‐CAD is a major cause of stroke in younger patients, for whom optimal management is paramount to minimize the burden of lifelong disability ¹ . To date, 2 randomized trials have compared antithrombotic regimens (single antiplatelet therapy, dual antiplatelet therapy, or anticoagulation) for stroke prevention after CAD; however, results have been largely equivocal. ¹ , ³ , ⁴ More recently, the STOP‐CAD (Antithrombotic Treatment for Stroke Prevention in Cervical Artery Dissection) study, a large multicenter cohort study, found that although anticoagulation may reduce the risk of recurrent stroke compared with antiplatelet agents (especially for occlusive lesions), it was associated with an elevated risk of hemorrhage. ⁵ Thus, the optimal management of AIS‐CAD remains unclear.

For patients with CAD with severely stenotic or occlusive lesions, endovascular stenting is a potential treatment option ¹ . Although stenting can restore and maintain vessel patency, prior studies have not demonstrated a significant advantage of stenting in terms of functional outcomes after AIS‐CAD. ⁶ , ⁷ , ⁸ , ⁹ This observation was confirmed in a secondary analysis of the STOP‐CAD study, patients with CAD who underwent endovascular thrombectomy for tandem occlusion strokes. ¹⁰ However, a key tenet of AIS‐CAD management is the prevention of recurrent stroke, and there are currently no data available on the impact of endovascular stenting on stroke recurrence in the population with AIS‐CAD. One general concern of acute stenting in the setting of AIS is the increased risk of hemorrhagic complications associated with the need for immediate postoperative antiplatelet medications to maintain stent patency. ¹¹ , ¹² However, for patients with CAD with severe lesions, anticoagulation or dual antiplatelet therapy is normally required even without stenting to prevent early stroke recurrence. ¹ , ⁵ Thus, it is possible that endovascular stenting for patients with AIS‐CAD may not introduce an added hemorrhagic risk while still benefiting patients by preventing stroke recurrence by ensuring long‐term vessel patency.

In this retrospective study of a nationwide database of hospitalizations in the United States, we investigated whether endovascular stenting was associated with a reduction of stroke recurrence risk compared with no stenting for patients with AIS‐CAD. Furthermore, we assessed the hospitalization outcomes of the index AIS‐CAD, including morbidity, mortality, and intracranial hemorrhage. We hypothesize that stenting is associated with a reduction in recurrent stroke risk without substantially increasing the risk of adverse outcomes.

Methods

All data used in this study are publicly available for purchase at https://hcup-us.ahrq.org.

Database Characteristics

This was a retrospective analysis of the 2016–2021 Nationwide Readmissions Database (NRD), which is part of the Healthcare Cost and Utilization Project. The NRD provides stratified discharge information in the United States. Patients within the NRD are tracked for the remainder of the calendar year using a unique ID to allow for the identification of hospital readmissions. The time between hospital events for each patient can be calculated. Patient identifiers are not included in the NRD; as such, this study was exempt from institutional review board approval under the Health Insurance Portability and Accountability Act or informed consent.

Patient Population

Patients hospitalized for AIS‐CAD were identified. To ensure that included patients did not have chronic infarction as a premorbid diagnosis, we included only patients with available National Institutes of Health Stroke Scale information.

The primary exposure of interest was endovascular stenting, and patients were separated into stenting and no stenting groups. Patient demographics (age, sex) were recorded. National Institutes of Health Stroke Scale score, location of dissection (carotid or vertebral), and concomitant intravenous thrombolysis or endovascular thrombectomy were captured. Stroke risk factors such as atrial fibrillation, hypertension, hyperlipidemia, diabetes, smoking history, myocardial infarction, coagulopathy, and congestive heart failure were also extracted, as were anticoagulant and antiplatelet use. Elixhauser comorbidity index was calculated for each patient to estimate the overall medical comorbidity burden. ¹³

The primary end point was recurrent stroke within 180 days. Hospital readmissions for reasons other than stroke were disregarded unless they led to patient death, in which case the patient was censored for follow‐up at the time of that readmission. Patients readmitted for acute stroke were followed for additional recurrence as a separate index event. The study follow‐up duration was 180 days. Due to the limitations of the NRD in reporting the exact initial admission time and cessation of all follow‐ups at the end of the calendar year, all patients admitted in July through December were censored for follow‐up at the end of the calendar year if a recurrent stroke event had not happened. Secondary end points include perioperative outcomes, including rates of routine discharge (discharge to home with self‐care and no rehabilitation needs), length of hospital stay, intracranial hemorrhage, and patient death. Time‐to‐event analysis of all‐cause mortality was also evaluated during the 180‐day study period. Subgroup analyses stratified by age, sex, stroke severity, concomitant acute stroke treatment, location of dissection, and overall comorbidity burden were also conducted for stroke recurrence risk. All diagnoses and procedures captured in this study were identified using International Classification of Diseases, Tenth Edition (ICD‐ 10) codes (Table S1). The ICD‐10 codes used to identify stroke recurrence have been validated for use in administrative databases in the past. ¹⁴

Statistical Analysis

The number of patients was reported as national estimates after applying discharge weights as per standard practice with studies using Healthcare Cost and Utilization Project databases. Continuous variables were reported as median and interquartile range, and categorical variables were reported as percentages. Comparisons of descriptive data were compared between patient groups using standardized mean differences (SMD, where a value >0.10 is considered a large difference). To adjust for differences in patient characteristics between patient groups, 1‐to‐1 propensity score matching (PSM) was performed with all captured variables in the study (variables listed in Table 1, and associated ICD‐10 codes are listed in Table S1). Binary logistic regressions were used to generate propensity scores, and nearest‐neighbor matching with a caliper of 0.01 without replacement was performed to arrive at the final analytical cohort. PSM with calipers was chosen as the primary method for balancing baseline characteristics in our study as it has been shown to yield robust outcomes comparable to other methods (eg, inverse probability of treatment weighting with restriction) with easier interpretability. ¹⁵ , ¹⁶ Because the numbers of patients are reported as national estimates after applying discharge weights, the sample sizes may be slightly imbalanced after PSM. For the primary analysis, Cox regression was used to calculate hazard ratios for stenting compared to no stenting in terms of stroke recurrence risk within 180 days among the PSM cohorts. For secondary end points, chi‐square and Mann–Whitney U tests were used to compare statistics between the 2 treatment groups. Finally, mediation analyses were performed based on methods described by Baron and Kenny. ¹⁷ Direct effect, total effects, and average causal mediated effects were calculated, and the bootstrapping method described by Preacher and Hayes ¹⁸ with 1000 simulations per analysis was used to generate CIs and test for statistical significance. All statistical analyses accounted for discharge weights. P values <0.05 were deemed statistically significant. All statistical analyses were performed using R, Version 3.6.2.

Table 1.

Patient Characteristics

		Unmatched cohorts			Propensity score‐matched cohorts
Characteristics	Total cohort (N = 20 434)	No stenting (N = 18 066)	Stenting (N = 2368)	SMD	No stenting (N = 2215)	Stenting (N = 2244)	SMD
Age (y)	55 (43–67)	55 (43–68)	56 (46–66)	0.037	56 (45–67)	56 (46–66)	0.010
Female sex	40.4% (8255)	41.2% (7438)	34.5% (817)	0.141	36.0% (797)	35.4% (794)	0.012
National Institutes of Health Stroke Scale score	4 (1–11)	3 (1–10)	10 (4–18)	0.592	10 (3–18)	10 (4–18)	0.012
Acute stroke intervention
Intravenous thrombolysis	18.3% (3731)	16.8% (3026)	29.8% (705)	0.285	30.3% (670)	29.0% (651)	0.028
Endovascular Thrombectomy	19.8% (4037)	14.3% (2579)	61.5% (1457)	0.971	58.7% (1,300)	59.5% (1334)	0.015
Location of dissection
Carotid artery	53.4% (10 906)	49.1% (8869)	86.0% (2037)	1.065	88.3% (1955)	85.2% (1593)	0.087
Vertebral artery	46.6% (9528)	50.9% (9197)	14.0% (331)		11.7% (260)	14.8% (331)
Comorbidities
Hypertension	69.7% (14 233)	69.2% (12 510)	72.7% (1722)	0.078	71.5% (1584)	71.9% (1613)	0.008
Hyperlipidemia	49.9% (10 199)	49.9% (9016)	50.0% (1183)	0.001	49.9% (1105)	50.1% (1124)	0.005
Diabetes	21.6% (4413)	21.6% (3908)	21.3% (505)	0.008	19.2% (425)	20.9% (468)	0.041
Atrial fibrillation	11.2% (2299)	11.3% (2047)	10.6% (251)	0.023	11.2% (247)	10.9% (245)	0.009
Smoking	35.6% (7273)	35.9% (6491)	33.0% (782)	0.062	35.0% (774)	32.7% (735)	0.047
Coagulopathy	5.5% (1128)	5.3% (957)	7.2% (171)	0.075	6.7% (148)	6.9% (155)	0.008
Myocardial infarction	5.3% (1080)	5.4% (969)	4.7% (111)	0.033	5.8% (127)	4.8% (108)	0.045
Chronic heart failure	9.0% (1844)	9.3% (1681)	6.9% (163)	0.096	7.4% (164)	7.2% (161)	0.009
Elixhauser comorbidity index	7 (2–13)	7 (0–13)	12 (7–16)	0.448	12 (6–17)	12 (7–16)	0.003
Antithrombotic medications
Anticoagulant	8.4% (1720)	8.6% (1555)	7.0% (166)	0.063	6.6% (146)	7.1% (159)	0.020
Antiplatelet	9.4% (1929)	8.8% (1592)	14.2% (337)	0.015	13.6% (301)	12.5% (280)	0.033

Data represented as median (interquartile range) or % (n); SMD indicates standardized mean difference.

Results

Patient Population

A total of 36 499 patients hospitalized for AIS‐CAD were identified; 20 434 (56.0%) had available National Institutes of Health Stroke Scale information and were included. Overall, 2368 patients (11.6%) underwent endovascular stenting. Patient characteristics are presented in Table 1. Compared with patients who did not undergo stenting, those who underwent stenting were less likely female (34.5% versus 41.2%, SMD = 0.141), had higher National Institutes of Health Stroke Scale (median 10 versus 3, SMD = 0.592), were more likely treated with intravenous thrombolysis (29.8% versus 16.8%, SMD = 0.285) or endovascular thrombectomy (61.5% versus 14.3%, SMD = 0.971), more likely to have had a carotid artery dissection (86.0% versus 49.1%, SMD = 1.065), and had overall higher comorbidity burden (Elixhauser index median 12 versus 7, SMD = 0.448).

After 1‐to‐1 propensity score matching using all captured variables in the study, 2215 patients remained in the no stenting group, and 2244 remained in the stenting group. Overall, there were no significant imbalances for any clinical variables between the PSM cohorts (all SMD<0.10, Table 1). All subsequent statistical comparisons were made between the PSM cohorts.

Study Outcomes

For the primary analysis, the stroke recurrence risk of the stenting group was compared with the no‐stenting group using time‐to‐event analysis within the 180‐day follow‐up period. Here, we found that the stenting group had a significantly lower stroke recurrence risk than the no‐stenting group (hazard ratio 0.50 [95% CI: 0.27–0.95], P = 0.034; Figure 1, Table 2). In terms of absolute stroke recurrence rates, the stenting group outperformed the no‐stenting group, with significantly lower rates at 30 days (0.8% versus 2.1%, P = 0.018; Table 2), 90 days (1.1% versus 2.5%, P = 0.047; Table 2), and 180 days (1.2% versus 3.5%, P = 0.017; Table 2).

Figure 1.

Stroke recurrence risk following stenting versus no stenting. Treatment cohorts were matched 1 ‐to 1 using propensity scores calculated using all captured clinical variables (Table 1). HR for stenting versus no stenting were calculated using Cox regression models; P values <0.05 were deemed statistically significant. HR indicates hazard ratio.

Table 2.

Study Outcomes of Endovascular Embolization Compared to Conventional Management After Propensity Score‐Matching

Study outcomes	No stenting (N = 2215)	Stenting (N = 2244)	P value
Primary outcome
Hazard ratio (95% CI) for stroke recurrence	0.50 (0.27–0.95)		0.034
Secondary outcomes
Stroke recurrence rates – % [95% CI]
180‐d	3.5 (1.9–5.6)	1.2 (0.5–2.4)	0.017
90‐d	2.5 (1.4–3.9)	1.1 (0.5–1.9)	0.047
30‐d	2.1 (1.3–3.1)	0.8 (0.4–1.5)	0.018
Initial hospitalization outcomes – % (n) or median (interquartile range)
Routine discharge	44.8% (993)	43.8% (983)	0.65
Intracranial hemorrhage	19.0% (420)	21.9% (491)	0.027
Length of stay (d)	6 (3–13)	7 (4–13)	0.030
In‐hospital mortality	7.3% (162)	8.7% (195)	0.23
All‐cause mortality ‐ HR (95%CI)	1.14 (0.87–1.50)		0.34

The HRs represent the relative hazards of the outcome of interest for the stenting group compared to the no stenting group.

HR indicates hazard ratio.

In terms of the initial hospitalization outcomes, stenting was not associated with significant differences in rates of routine discharge (43.8% versus 44.8%, P = 0.65; Table 2); however, it was associated with a significantly higher rate of intracranial hemorrhage (21.9% versus19.0%, P = 0.027; Table 2) and longer length of hospital stay (median 7 versus 6 days, P = 0.030; Table 2). Stenting was not associated with a different risk of all‐cause mortality during the initial hospitalization (8.7% versus 7.3%, P = 0.34; Table 2) or the 180‐day follow‐up period (hazard ratio 1.14 [95% CI: 0.87–1.50], P = 0.34; Table 2). Given that stenting was associated with elevated rates of intracerebral hemorrhage, we performed mediation analyses to determine whether this elevated rate of intracerebral hemorrhage may have affected discharge outcomes. Here, we found that intracerebral hemorrhage did not significantly mediate the relationship between stenting and routine discharge (average causal mediated effect −0.011 [95% CI: −0.017 to 0.003], P = 0.19) or death (average causal mediated effect 0.004 [95% CI: −0.001 to 0.008], P = 0.15).

Finally, we conducted subgroup analyses. Here, we found no significant interaction effects between stenting and patient age, sex, stroke severity, intravenous thrombolysis use, endovascular thrombectomy, location of dissection, or overall comorbidity burden (all interaction P value>0.10, Figure 2). In addition to the listed comparisons, we also performed interaction analyses with age and National Institutes of Health Stroke Scale score as continuous variables; here, also, no significant interactions were observed to suggest the presence of treatment effect heterogeneity (interaction P = 0.32 and 0.29, respectively).

Figure 2.

Subgroup analyses and effect modulation of stenting versus no stenting for stroke recurrence among patients with acute ischemic stroke due to cervical artery dissection. HRs represent the hazard of stroke recurrence for stented patients compared to no stenting; P values <0.05 were deemed statistically significant. HR indicates hazard ratio; and NIH, National Institutes of Health.

Discussion

In this large retrospective cohort study, we provide novel evidence that endovascular stenting for patients with AIS‐CAD was associated with lower rates of stroke recurrence. Our study also corroborates prior investigations in showing that stenting was not associated with a significant difference in short‐term functional outcomes, ⁶ , ⁷ , ⁸ , ⁹ as evidenced by no increased risk of perioperative morbidity or mortality. These data suggest that endovascular stenting, as currently used in real‐world clinical practice, is an effective and safe treatment for the population with AIS‐CAD.

CAD is a rare cause of AIS, and it is associated with an elevated stroke recurrence risk in the short to medium term. ¹ In our study, we found that medically managed patients had a stroke recurrence rate of 2.5% at 90 days and 3.5% at 180 days, which is consistent with prior literature (2.4% ipsilateral stroke within 90 days in CADISS ³ (Cervical Artery Dissection in Stroke Study) and 3.3% stroke risk within 180 days in STOP‐CAD ⁵ among patients treated with antiplatelets). Due to these low baseline stroke recurrence risks, however, numerous studies that have sought to identify the optimal treatment to prevent recurrent stroke have been largely inconclusive. ³ , ⁴ The combination of AIS‐CAD being a rare condition and its association with low rates of recurrent strokes has posed significant statistical challenges for studying this disease. To date, the largest study of patients with CAD is the STOP‐CAD study (N = 3636), which revealed that anticoagulation may be superior to antiplatelet regimens in preventing stroke ⁵ ; however, only 10%–15% of patients received any endovascular treatment (which included stroke thrombectomy). Thus, valuable insight into endovascular stenting's association with stroke recurrence is unlikely to be drawn from STOP‐CAD or any other existing database of patients with CAD. In our study, we present findings from the largest cohort of patients with AIS‐CAD to date. As such, we were able to achieve sufficient statistical power to identify a significant, 2‐fold reduction of stroke recurrence over 6 months associated with endovascular stenting.

Our study findings on short‐term hospitalization outcomes are consistent with the current literature, overall demonstrating equivocal functional outcomes for stenting versus no stenting for patients with AIS‐CAD. ⁶ , ⁷ , ⁸ , ⁹ Despite positive data generally supporting endovascular stenting for carotid lesions in the setting of AIS, ¹⁹ , ²⁰ there remain significant concerns regarding the potentially devastating risk of hemorrhage due to the requirement of anti‐platelet medication use after stent placement. ¹¹ , ¹² Given that many patients with an extensive CAD lesion being considered for stenting would likely require parenteral anticoagulation or dual antiplatelet therapy even if a stent was not placed, ¹ the net additional hemorrhage risk due to stenting may be lower for these patients with AIS‐CAD compared with other patients with stroke. This key feature of standard medical management for CAD likely explained our study showing only a ∼3% increase in the risk of intracranial hemorrhage; furthermore, mediation analyses did not show that this elevated rate translated to an impact on discharge outcomes. Overall, our study findings provide reassuring safety data for stenting among patients with AIS‐CAD.

It is important to recognize that AIS‐CAD is a heterogeneous disease, and providers should take a personalized approach when deciding the optimal treatment strategy. ¹ In our study, patients who underwent endovascular stenting may have had a higher degree of vessel lumen compromise, which may be associated with an elevated risk of distal embolism and cerebral hypoperfusion. Interestingly, despite the stenting group likely being enriched with patients with more severe lesions, the risk of stroke recurrence was still significantly lower than in the no stenting group, which further highlights the potential efficacy of endovascular stenting in select patients with AIS‐CAD. It is likely that the patient selection process for stenting AIS‐CAD lesions has been aided by the widespread use of artificial intelligence‐based platforms for perfusion imaging during acute stroke triage in the United States. ²¹ , ²² , ²³ Future research into the role of perfusion imaging and other tools in selecting patients with AIS‐CAD for acute stenting is needed to further refine and standardize treatment protocols. Of note, our study also found a high degree of treatment effect heterogeneity in the vertebral artery dissection subgroup. This was likely a result of a relatively limited sample size (n = 591); future studies are needed to further explore the efficacy and safety of endovascular stenting for this subpopulation.

Our study has several limitations. First, as a retrospective study of an administrative database, our analyses are subject to numerous biases that could persist despite PSM. Furthermore, many variables associated with CAD (eg, degree of stenosis/occlusion, mechanism of injury, tandem occlusions, and so forth) and AIS (site of occlusion, time from last known well, etc.) were not available, and ICD‐10 code reporting may be inaccurate. However, the large number of patients we were able to identify is a major strength of our study. Given that AIS‐CAD is a rare condition and stenting procedures are uncommon, there is unlikely an alternative method to meaningfully investigate stroke recurrence rates associated with stenting for AIS‐CAD. Second, although prior studies of stroke pathologies using administrative databases such as the NRD have used discharge destinations as surrogates for short‐term neurological outcomes, ²⁴ , ²⁵ this is an imperfect measure, and we do not have detailed information on functional outcomes. Finally, although we used adjudicated ICD‐10 codes to identify ischemic stroke recurrence that led to hospitalizations, ¹⁴ it is possible that some stroke events were clinically milder and did not lead to repeat hospitalization; thus, our study may have underestimated stroke recurrence rates.

Conclusions

For patients with AIS‐CAD, endovascular stenting may be an effective treatment in reducing subsequent stroke risk without increasing the risk of morbidity or mortality. A small but significant increase in intracerebral hemorrhage associated with stenting was observed; however, it did not appear to significantly affect patient outcomes.

Sources of Funding

None.

Disclosures

D.G. receives research funding from the National Institutes of Health, InSightec, the Focused Ultrasound Foundation, MicroVention, the University of Calgary, and the University of Maryland Medical Center and is a consultant for Navigantis. All other authors have no relevant disclosures.

Supporting information

Table S1: ICD‐10 codes used for study.