Stroke after Carotid Stenting and Endarterectomy in the Carotid Revascularization Endarterectomy versus Stenting Trial (CREST)

Abstract

Background

Stroke occurs more commonly after carotid artery stenting than carotid endarterectomy. Details regarding stroke type, severity, and characteristics have not been previously reported. We describe the strokes occurring in the Carotid Revascularization Endarterectomy versus Stenting Trial (CREST).

Methods and Results

CREST is a randomized, open-allocation, controlled trial with blinded endpoint adjudication. Stroke was a component of the primary composite outcome. Patients who received their assigned treatment within 30 days of randomization are included. Stroke was adjudicated by a panel of board-certified vascular neurologists with secondary central review of clinically-obtained brain images. Stroke type, laterality, timing, and outcome are reported. A periprocedural stroke occurred among 81 of the 2502 patients randomized and among 69 of the 2272 in this analysis. Strokes were predominantly minor (81%, n=56), ischemic (90%, n=62), in the anterior circulation (94%, n=65), and ipsilateral to the treated artery (88%, n=61). There were seven hemorrhages, occurring 3-21 days post-procedure, and five were fatal. Major stroke occurred in 13 (0·6%) of the 2272 patients. The estimated four-year mortality after stroke was 21·1% compared to 11·6% for those without stroke. The adjusted risk of death at four years was higher after periprocedural stroke (HR = 2·78, CI₉₅ 1·63-4·76).

Conclusions

Stroke, particularly severe stroke, was uncommon after carotid intervention in CREST, but stroke was associated with significant morbidity and was independently associated with a near threefold increased future mortality. The delayed timing of major and hemorrhagic stroke after revascularization suggests that these strokes may be preventable.

Introduction

Stroke is a more frequent complication of carotid stenting (CAS) compared to endarterectomy (CEA).^1-4 In the Carotid Revascularization Endarterectomy versus Stenting Trial (CREST), this greater occurrence of stroke in the CAS arm was offset by the greater occurrence of myocardial infarction (MI) in the CEA arm.⁵ However, stroke was found to have a greater impact on quality of life based upon assessment of the SF-36 physical and mental health subscales, compared to MI. In the International Carotid Stenting Surgery (ICSS) trial, a higher occurrence of minor stroke in the CAS arm was balanced by a higher occurrence of cranial nerve palsy in the CEA arm.⁴ Further, a MRI sub-study of ICSS confirmed that CAS was associated with a greater number of ischemic cerebral infarctions, both symptomatic and asymptomatic, observed on diffusion-weighted imaging.⁶ Age and sex may modify this risk with an observed higher risk of stroke with CAS at older ages^{7, 8} and among women.⁹

Prevention of stroke is the goal of carotid revascularization, yet differences in interpretation of periprocedural stroke as an endpoint in randomized trials of CAS and CEA have generated controversy. Well-informed discussants have advocated MI and cranial neuropathy as important complications of CEA, important enough to consider CAS an equivalent procedure with regard to safety.¹⁰ Others have advocated that patients are more affected by stroke (than MI or cranial neuropathy) on measures of disability and quality of life, and therefore that CEA should be the preferred procedure.^{11, 12} European guidelines favor CEA as the primary modality of choice; North American guidelines allow either procedure.¹³ The discussions and guidelines have been limited by a dearth of detailed information describing the strokes.

We describe the nature, localization, severity, and outcome of periprocedural stroke in CREST. We also report on the imaging characteristics of these strokes, the timing of stroke after carotid revascularization, and the association of periprocedural stroke with long-term outcomes including late mortality.

Methods

CREST is a randomized, open-allocation, controlled trial with blinded endpoint adjudication comparing CEA to CAS among both symptomatic and asymptomatic patients with atherosclerotic carotid artery stenosis. Stroke was a component of the primary composite outcome for the 2502 patients randomized. The analysis of the stroke outcome was a pre-specified analysis. Patients who received their assigned treatment within 30 days of randomization (n=2272) are included in these analyses. Institutional/Ethics review boards at all participating centers approved the protocol. All patients provided written informed consent. The authors designed the study, gathered and analyzed the data, wrote the manuscript, made the decision to publish the findings, vouch for the completeness and accuracy of the data, and attest to the fidelity of the report to the study protocol. The primary results and full description of the methodology have been previously reported.⁵

Patients were enrolled at 108 centers in the United States and nine in Canada. The CEA surgeons were credentialed by a surgical management committee. The Interventional Management Committee was responsible for training and credentialing the CAS interventionists. Patients were classified as symptomatic if they had experienced a recent transient ischemic attack (TIA), stroke, or transient monocular blindness ipsilateral to the study artery in the preceding 180 days prior to randomization. Otherwise, they were classified as asymptomatic.

Stroke was defined as an acute neurologic event with focal symptoms and signs, lasting for 24 hours or more, that were consistent with focal cerebral ischemia or hemorrhage, and was considered as a complication of carotid revascularization if it occurred within 30 days of the procedure. Initially, a broad net was cast to try to identify stroke outcome events. One or both of the following could be used as confirmatory evidence of stroke: a one-point increase on the National Institutes of Health Stroke Scale (NIHSS), or an appropriate new or extended abnormality seen on computed tomography or MRI. The stroke adjudication process was initiated by the clinical sites after detection of a clinically significant neurologic event, any positive response on the TIA–Stroke Questionnaire, or an increase in the NIHSS score of ≥ 2 points; this second phase process was more detailed and deterministic. The occurrence and severity of stroke were determined by the Stroke Adjudication Committee. Stroke was minimally defined as major on the NIHSS score if the score was greater or equal to 9 at 90 days after the procedure and minor otherwise. Stroke was considered non-disabling if the modified Rankin scale score was ≤ two at 30 days. TIA and amaurosis fugax were not considered in this analysis. However, the final determination of whether a stroke was major or minor was based upon a combination of narrative clinical reports, the NIHSS, imaging reports and outcome data.

The Stroke Adjudication Committee consisted of six board-certified stroke neurologists. All adjudicators were blinded to the randomized procedure. Events were reviewed by at least two adjudicators. Enlistment of a third reviewer occurred if the first two adjudicators disagreed with one or more of the following variables: stroke outcome, date of occurrence, vascular distribution and stroke severity. In the event of an ongoing disagreement after the third review, the Stroke Adjudication Committee met via conference call to resolve the disagreement. A total of 300 potential events were screened. Of these, 46 were deemed to be TIA or amaurosis fugax (i.e. symptoms lasted in duration less than 24 hours) and per internal protocol were not sent for physician review. Potential periprocedural events were submitted for physician adjudication for 254 suspected events. For events adjudicated as strokes, the laterality, timing, stroke type, and outcome were determined.

Post-operative images of the brain were not collected routinely as part of the study protocol. Following publication of the primary outcomes of CREST in 2011, we asked sites to provide digital copies of brain images (computed tomography or MRI) to further characterize the strokes occurring as periprocedural complications of carotid revascularization. All images were reviewed centrally, blinded to treatment allocation (MDH, WFM). Imaging characteristics of stroke are described quantitatively, by stroke type (hemorrhage or ischemia), by arterial territory among ischemic strokes, and qualitatively by imaging pattern of infarction. Where images were unavailable, we requested neuroradiologists’ imaging reports and used these to estimate imaging characteristics. Volumes of strokes were estimated using standard planimetry. The brain imaging review was a post-hoc analysis; it was not pre-specified.

For this analysis we primarily considered only the periprocedural period, which lasted 30 days from the date of the procedure. In addition, we considered only patients who underwent a carotid revascularization procedure and only those patients who had a stroke during or after their procedure; therefore, this is a per-protocol analysis and not an intention-to-treat analysis. We conducted a secondary analysis to examine long-term mortality after stroke using an intention-to-treat approach including all patients by their assigned treatment and all periprocedural strokes that occurred after randomization. Data are reported using standard descriptive statistics. Proportions were compared using Fisher’s exact test or χ² test and normally distributed continuous variables were compared using a t-test. We used Kaplan-Meier survival analysis to estimate long-term mortality. Hazard Ratios for mortality up to four years, adjusted for age, sex, treatment and symptomatic status, were calculated using a Cox proportional hazards model. For the purposes of this latter analysis, we considered the periprocedural stroke event, which by definition occurred within 30 days of the procedure, as if it had occurred at time 0.

Results

A total of 122 subjects had a stroke; 81 occurred during the periprocedural period. Three subjects had a stroke after randomization but before undergoing CEA or CAS and were excluded from the present analysis. Nine additional subjects did not have their assigned procedure within 30 days of randomization and were excluded. Therefore, 69 (3·0% of 2272) patients who received assigned treatment within 30 days of randomization had a stroke within 30 days of their procedure and comprise the primary cohort for the current analysis. Patients who had a periprocedural stroke compared to those who did not had similar baseline clinical characteristics (Table 1).

Table 1.

Characteristics of patients with and without periprocedural stroke.

	CAS			CEA
	Stroke (n=48)	Non-Stroke (n=1075)	p- value	Stroke (n=21)	Non-Stroke (n=1128)	p- value
Age (mean ± SD)	73·0 ± 7·7	68·5 ± 9·0	<0·001	70·2 ± 9·7	69·1 ± 8·7	0·57
Male, n (%)	25 (52·1)	700 (65·1)	0·06	14 (66·7)	756 (67·0)	0·97
White, n (%)	44 (91·7)	1006 (93·6)	0·60	21 (100·0)	1057 (93·7)	0·24
Prior cardiovascular disease, n (%)	19 (40·4)	452 (43·7)	0·66	9 (45·0)	482 (44·6)	0·97
Risk factor status, n (%)
Hypertension	45 (93·8)	905 (84·3)	0·07	18 (85·7)	970 (86·1)	0·96
Diabetes	18 (37·5)	320 (29·9)	0·26	5 (23·8)	349 (31·0)	0·48
Dyslipidemia	37 (77·1)	902 (84·3)	0·18	18 (85·7)	963 (85·8)	1·0
Current smoker	7 (14·9)	296 (27·9)	0·05	6 (28·6)	289 (26·0)	0·79
Symptomatic arteries, n (%)	34 (70·8)	562 (52·3)	0·01	15 (71·4)	597 (52·9)	0·09
Type of Qualifying Event (Symptomatic Patients), n (%)
Stroke	16 (47·1)	241 (42·9)	0·61	5 (33·3)	257 (43·1)	0·75
TIA	15 (44·1)	237 (42·2)		7 (46·7)	243 (40·7)
Amaurosis Fugax	3 (8·8)	84 (14·9)		3 (20·0)	97 (16·3)
Medical treatment pre-procedure, n (%)
Antiplatelet therapy 48 hours	45 (95·7)	1010 (96·4)	0·69	21 (100·0)	1035 (91·8)	0·40
Cholesterol medication	33 (94·3)	810 (92·4)	0·67	16 (94·1)	849 (91·0)	0·66

Strokes were most commonly minor (81%, n=56). The NIHSS score determined within one month after detection of the stroke was available for 57 of the 69 strokes. The median NIHSS was 2, (interquartile range [IQR] 6) (Figure 1). The median NIHSS for the minor strokes was 2 (available for 50 of 56 minor strokes) and for the major strokes 8 (available for 7 of 13 major strokes). The strokes were overwhelmingly ischemic (90%, n=62), in the anterior circulation (94%, n=65), and ipsilateral to the treated artery (88%, n=61) (Table 2). Two in each group involved the posterior circulation and included posterior cerebral artery territory, splenial and pontine infarcts.

Figure 1.

Box-and-whisker plots showing the distribution of NIHSS following periprocedural stroke for CEA (red) and CAS (blue) patients. The horizontal axis shows the time period, and provides the number of patients in the CEA and CAS treatment group respectively (i.e. n = 21/48 implies 21 patients in the CEA group and 48 in the CAS group). The bottom of the box is the 25^th percentile, the line in the center the 50^th percentile (median), the top of the box the 75^th percentile, and the top of the whisker the 90^th percentile. In the cases where the median is not shown, it is equal to zero (0) and is plotted with the 25^th percentile (also zero). Data points outside the 90^th percentile are shown as dots. Because there is no NIHSS score for patients who have died, these patients are shown in the gray shaded box at the top of the figure (representing an outcome that is worse than any patient surviving). Note that this display shows the entire distribution of the NIHSS outcomes as a function of time, allowing the reader to see that those patients alive at the PostProc period are doing worse than the patients alive at “1 Mth” (note the lower median and 75^th percentiles); however, at “1 Mth” there were deaths and a small proportion of patients with a very poor outcome. PreProc=Pre-procedure, PostProc=Post-procedure, Mth=Month, and n=number.

Table 2.

Stroke characteristics reported clinically.

	CAS	CEA
	Stroke (n=48) % (n)	Stroke (n=21) % (n)
Ischemic	91·7 (44)	85·7 (18)
Ipsilateral	90·9 (40)	83·3 (15)
Contralateral	4·6 (2)	11·1 (2)
Vertebrobasilar	4·6 (2)	5·6 (1)
Stroke Severity
Major Stroke	20·8 (10)	14·3 (3)
Minor Stroke	79·2 (38)	85·7 (18)
Intracerebral Hemorrhage	8·3 (4)*	14·3 (3)*
mRS (30 days) (median, IQR)**	1 (1·5)	1 (2)
Timing (days from randomization to stroke) (median, IQR)	7 (10)	8 (14)
Timing (days from procedure to stroke) (median, IQR)	0 (3·5)	1 (7)
Death (12 months)	14·6 (7)	14·3 (3)

^*All four of the CAS intracerebral hemorrhages were ipsilateral, and two of the three CEA intracerebral hemorrhages were ipsilateral (one being vertebrobasilar).

^**mRS scales were unavailable on 5 of the 69 patients.

The median time from the date of procedure to stroke was 0 days (IQR 4 days). The median time to minor stroke was 0 days (IQR 3) and the median time to major stroke was 3 days (IQR 12 days). Figure 2 shows the distribution of strokes relative to post-procedural time interval for the two procedures. Stroke was disabling (mRS > 2 at 30 days) in 23·4% (n=15) of the 64 stroke patients with a modified Rankin Scale score at one month. Mortality among all 69 patients who had strokes was 14.5% at 1 year.

Figure 2.

Timing of stroke after carotid revascularization.

Among 59 subjects reviewed (49 primary imaging data, 10 neuroradiology reports), 40 (68%) had MRIs completed within a week of their event and 19 (32%) had computed tomography only (Table 3). Nine patients had no evidence of a new stroke on imaging. There were three common patterns in anterior circulation infarcts, distributed approximately in thirds and equally divided proportionately between the CAS and CEA groups (Figure 3): (1) scattered or a shower of emboli in the distribution of the revascularized artery; (2) typical wedge-shaped cortical infarcts, and; (3) small subcortical and lacunar infarcts. The mean volume of cortical infarcts was 22·5 ml (sd 28·6 ml). It was judged impractical to attempt to measure the sum of multiple scattered emboli, particularly when MRI was not uniformly conducted. Among the CAS subject images, one ischemic stroke was contralateral, two were in the brainstem or cerebellum, and three were bilateral or multi-territory. Among the CEA subjects with imaging studies, one ischemic stroke was contralateral to the treated artery. No bilateral, multi-territory, or posterior ischemic strokes were identified in the CEA subjects upon imaging review.

Table 3.

Imaging characteristics of periprocedural stroke in 59 patients with available imaging who underwent carotid artery stenting (CAS) versus endarterectomy (CEA).

	CAS	CEA
	Stroke (n=40/48) % (n/N)	Stroke (n=19/21) % (n/N)
Ischemic (n=42)	73 (29/40)	68 (13/19)
Hemorrhage (n=5)	8 (3/40)	11 (2/19)
No stroke on imaging (n=9)	13 (5/40)	21 (4/19)
UTD (n=3)*	8 (3/40)	---
Ischemia Distribution (n=42)
Posterior circulation	3 (1/29)	0 (0/13)
Anterior circulation**	83 (24/29)	100 (13/13)
Bilateral or Multi-territory	14 (4/29)	0 (0/13)
Ischemia Pattern (n=42)
Scattered embolic	38 (11/29)	38 (5/13)
Cortical infarct	31 (9/29)	31 (4/13)
Subcortical infarct	17 (5/29)	31 (4/13)
Bilateral or Multi-territory	14 (4/29)	0 (0/13)

There were: 69 total strokes, of which 59 had imaging data reviewed.

^*UTD = Unable to determine from available images due to poor quality scans preventing review (n=1) or imaging was available only pre-stroke (n=2).

^**Two ischemic anterior strokes, one each in the CAS and CEA arms, anterior circulation were contralateral to the targeted carotid lesion.

Figure 3.

Ischemic stroke patterns. Representative ischemic stroke cases. First row, 3 patients with scattered emboli in the distribution of the revascularized artery. Second row, 3 patients with cortical infarction in the territory of the revascularized artery. Third row, 4 patients with subcortical infarction.

There were seven intracerebral hemorrhages (ICH), four in the CAS arm and three in the CEA arm. Five of these subjects died within the periprocedural period (Supplemental Figure 1). One of the ICHs after CEA was contralateral and occurred at 14 days. This ICH was located in the right posterior parieto-occipital region, resulted in intraventricular rupture, and the patient died on day 16. All other cases were ipsilateral. ICH occurred on day two, three, four, eight, 14, and 21 (two patients) after intervention.

In the CAS arm, plaque characteristics such as eccentricity and ulceration were numerically more common among patients who had a stroke (Table 4). Intra-procedural factors were different in the CAS arm; patients who had a stroke more commonly required blood transfusion. At baseline, within the CAS arm, patients who had a stroke were more likely to be older and recently symptomatic but less likely to be current smokers. Within the CEA arm, there were no substantial differences in baseline characteristics (Table 5).

Table 4.

Baseline characteristics of carotid atherosclerotic lesions for patients who underwent carotid artery stenting (CAS) versus endarterectomy (CEA).

	CAS			CEA
	Stroke (n=48)	Non-Stroke (n=1075)	p- value	Stroke (n=21)	Non-Stroke (n=1128)	p- value
Minimal residual lumen (mm) (mean ± SD, n)*	1·2 ± 0·6 (46)	1·3 ± 1·8 (1035)	0·76	1·2 ± 0·9 (6)	1·5 ± 5·0 (427)	0·50
Diameter stenosis (mm) (mean ± SD, n) *	76·3 ± 9·4 (47)	76·0 ± 11·2 (1069)	0·84	72·0 ± 16·4 (7)	73·8 ± 10·8 (456)	0·67
Lesion length (mm) (mean ± SD, n)	20·9 ± 7·6 (48)	17·6 ± 8·5 (1070)	0·01	---	---
Eccentric lesion (%, n)	70·8 (34)	56·6 (608)	0·051	---	---
Ulcerated lesion (%, n)	54·2 (26)	36·0 (387)	0·01	---	---
Left carotid treated (%, n)	62·5 (30)	49·4 (531)	0·08	17·1 (12)	52·4 (591)	0·67

^*Taken from procedural angiogram for CAS patients and from baseline angiogram for CEA patients (where available).

Table 5.

Procedure information among patients who underwent carotid stenting (CAS) versus endarterectomy (CEA).

	CAS			CEA
	Stroke (n=48)	Non-Stroke (n=1075)	p- value	Stroke (n=21)	Non-Stroke (n=1128)	p- value
Length of procedure (min) (mean ± SD, n)*	77·7 ± 30·5 (47)	69·0 ± 41·2 (1072)	0.07	184·9 ± 51·7 (19)	170·9 ± 59·2 (1008)	0.31
Hypertension requiring treatment (%, n)	4·2 (2)	1·3 (14)	0.15	14·3 (3)	4·2 (47)	0.06
Hypotension requiring treatment (%, n)	10·4 (5)	4·4 (47)	0.07	0·0 (0)	2·1 (24)	1.00
Bradycardia requiring treatment* (%, n)	2·1 (1)	1·4 (15)	0.51	0·0 (0)	0·6 (7)	1.00
Transfusion (%, n)	12·5 (6)	1·6 (17)	<0.00 01	0·0 (0)	1·0 (11)	1.00

^*As reported on procedure form.

Figure 1 describes the stroke severity across time for both groups using the NIHSS. Prior to the procedure (PreProc) the distribution of the NIHSS was similar (p_Wilcoxon = 0·43) and uniformly below a score of 5. Stroke was more severe for CAS-treated patients after the procedure (P_Wilcoxon=0.15) with the 75^th and 90^th percentile, respectively, for CEA-treated patients being 3 and 6 compared with 5 and 12 for CAS-treated patients. However, at one month and beyond the majority of stroke patients returned to near pre-procedure neurological deficits. There was little evidence of differences in the severity of strokes between treatment groups (p > 0·27). The chance of death following periprocedural strokes was also similar for CEA-treated and CAS-treated patients. Among CEA-treated patients there were two of 19 (10·5%) deaths by one month, three of 17 (17·6%) deaths by six months, and three of 17 (17·6%) deaths by 12 months compared to CAS where there were four of 46 (8·7%), six of 43 (14·0%), and seven of 45 (15·6%) deaths respectively. Hence, while there were slightly more than twice as many strokes among the CAS-treated than the CEA-treated patients, there was not strong evidence that the distribution of severity differed by treatment group.

In the intention-to-treat analysis, during long-term follow-up (median, 2·5 years; range, 1-4 years), there were 177 deaths, with an estimated four-year mortality of 11·9%. Periprocedural stroke occurred in 81 patients and long-term mortality was higher if a stroke occurred. Mortality was typically acute, occurring soon after the event. The estimated mortality rate at four years was 11·6% in the stroke-free group and 21·2% in the stroke group (age, sex, treatment, and symptomatic status adjusted HR = 2·78, CI₉₅ 1·63-4·76) (Figure 4). We conducted a sensitivity analyses by considering time 0 beginning at 30 days post-procedure and by considering stroke as a time-varying covariate; this resulted in an adjusted HR ranging from 2.76 to 2.84.

Figure 4.

Mortality after periprocedural stroke (intention-to-treat analysis). Survival curve of mortality by stroke (n=81) or nonstroke (n=2421) status. This analysis includes all 69 periprocedural strokes, 3 strokes that occurred after randomization but before the procedure, and 9 strokes that occurred after the 30-day periprocedural period. Log-rank test P<0.0001. Estimated mortality rate at 4 years was 11.6% (±1.0; 162/2421) in the stroke-free group and 21.2% (±5.2, 15/81) in the stroke group.

Discussion

Stroke occurred infrequently after carotid intervention in CREST. The rates of periprocedural stroke for symptomatic patients are the lowest reported from recent randomized trials comparing CAS and CEA (Supplemental Table 1). The rates of periprocedural stroke following CAS and CEA compare favorably to those reported in the Asymptomatic Carotid Atherosclerosis Study (ACAS) (1·4%) and the Asymptomatic Carotid Surgery Trial (ACST) (2·5%).^{14, 15} The periprocedural strokes in CREST were most commonly minor, ipsilateral to the treated artery, and ischemic in type and occurred twice as frequently in the CAS arm. Major stroke occurred in 0·6% (13/2272), indicative of the very low overall complication rate observed in the trial.

Review of the available computed tomographies and MRIs suggests three patterns of periprocedural stroke: scattered emboli, cortical, and small subcortical (Table 3). Scattered emboli in the distribution of the treated artery are commonly seen after intervention and may also be seen spontaneously without intervention, which suggests an arteroembolic mechanism. Cortical infarcts, such as wedge-shaped cortical infarcts, may be seen from an arteroembolic source or a cardioembolic source. We do not know if patients who developed wedge-shaped cortical infarcts had alternate co-existent cardioembolic sources that arose peri-operatively. Patients with known chronic or paroxysmal atrial fibrillation were not included in the trial. Further, because we do not have serial MRIs with diffusion weighted sequences and because many patients were treated within a few days of randomization, we do not know if the scattered emboli pattern seen on the post-procedural MRI was spontaneous from the initial stroke or TIA event secondary to the symptomatic carotid artery lesion, or arose directly from the procedure. The limitations of our analysis emphasize the importance of conducting pre-planned image analysis as a component outcome of stroke clinical trials.

Not all strokes were related to the artery being addressed. Strokes that were posterior, contralateral or multi-territory occurred in both CAS and CEA arms but quantitatively more commonly with CAS. It is straightforward to envision catheter-related disruption of aortic arch plaque causing posterior, contralateral, or multi-territory anterior circulation strokes. It is less clear how this occurs with CEA; metachronous atherosclerotic plaque instability in the aortic arch, contralateral carotid artery, intracranial circulation, or an alternate cardioembolic source are possible explanations.¹⁶

Hemorrhage was severe and devastating and was not more common in the CAS arm. We cannot necessarily conclude that the use of double-antiplatelet therapy in the CAS arm predisposes to hemorrhage. The timing of hemorrhage suggests that these cases may have been related to hyperperfusion syndrome with underlying disordered auto-regulation of cerebral blood flow ipsilateral to the revascularized artery.¹⁷ Reperfusion hemorrhage has been proposed as a mechanism of hemorrhage after intracranial artery stenting completed in the SAMMPRIS trial (Stenting and Aggressive Medical Management for Preventing Recurrent Stroke in Intracranial Stenosis) and after thrombolysis for acute stroke thrombolysis but has been observed too infrequently in randomized trials to justify conclusions.¹⁸ Because the CREST hemorrhages occurred days after intervention, we hypothesize that there is an opportunity to prevent them and speculate that careful and tight blood pressure control could be lifesaving.

The timing of stroke after revascularization is important. Minor strokes occurred early, typically on the same day of the procedure. Qualitatively we know that few were observed intra-procedurally. Minor stroke was only identified on careful examination later in the day or the following day. Major strokes, including hemorrhages, tended to occur several days after the procedure. While we do not know the exact mechanism of each of these strokes nor the details of post-operative management, we can infer that there is a substantial opportunity for prevention of these major strokes. For example, stringent blood pressure control might conceivably mitigate the risk of both hemorrhage and major ischemic stroke. Unerring use of antiplatelet medication, statins and good diabetic management, similar to management in SAMMPRIS, could reduce the risk of major ischemic stroke.¹⁸

Stroke implied a poorer long-term mortality compared to those who underwent revascularization without incident. The risk of death was nearly three-fold higher (HR = 2·78, CI₉₅ 1·63-4·76) and this relative increase is very similar in magnitude to those who had a perioperative MI (HR = 3·67, CI₉₅ 1·71-7·90) (Figure 4).¹⁹ The question of whether or not major stroke was the driver of this relationship of periprocedural stroke to long-term mortality could not be addressed. The adjudication of a given stroke as major took place following the occurrence of death, and with knowledge of the death by the adjudicators, in every case. Stratifying the analysis into major and minor stroke, and then looking at an outcome (death) that was in part used to determine the classification would yield a tautological result.

The present study has limitations. Imaging data were collected and analyzed on a post hoc basis and were not complete, and imaging was performed as indicated clinically rather than at prespecified time points with specified modalities. The number of stroke outcomes was low which is good for patients but reduced our sample size enough to make some of our conclusions hypothesis-generating.

Overall, stroke, particularly severe stroke, was uncommon after carotid intervention in the CREST trial but was associated with significant morbidity and mortality. The timing of major stroke after revascularization suggests that major stroke is potentially preventable. Minor stroke occurred most commonly and temporally at the time of CAS suggesting that CAS has potential for further improvement from expected advances in technology, technique, and training.

Supplementary Material

01

Supplemental Figure 1. Intracerebral hemorrhage after carotid intervention. Four cases of severe hemorrhage occurred periprocedurally, all of which were fatal. Intracerebral hemorrhage occurred at 2–21 days post-procedure.

Supplemental Table 1. Rates of periprocedural stroke for symptomatic and asymptomatic patients (actual treatment analysis) from selected clinical trials.

Stroke is a feared complication of carotid endarterectomy (CEA) and carotid stenting (CAS). CREST and European trials have shown that CAS is associated with a greater risk of stroke compared to CEA. CREST also showed that CEA was associated with a greater risk of myocardial infarction (MI) compared to CAS. The greater risk of MI numerically balanced the greater risk of stroke, so that the composite primary outcome (periprocedural stroke, MI, or death, and ipsilateral stroke out to 4-years) was similar for CEA and CAS. This result has invited criticism because of the differing directions of stroke and MI within the composite outcome.

To understand further, we examined the strokes that occurred as a complication of the procedure. Stroke was still more common after CAS, but overall the risk of severe stroke was <1% and similar for CEA and CAS. The delayed timing of some major strokes, particularly intracerebral hemorrhage occurring a few days post-operatively, makes it plausible that these post-operative strokes are preventable, perhaps with careful attention to blood pressure control. Minor stroke occurred most commonly on the same day as CAS, suggesting that the technical aspects of the procedure could be improved to minimize stroke as a complication. Previously we reported that MI, including biomarker-only MI, was associated with an increased risk in long-term mortality. Here we report that stroke, including minor stroke, was also associated with an increased risk in long-term mortality.

Carotid intervention with CEA or CAS is safe. Periprocedural stroke incurred significant morbidity and mortality.