General Anesthesia Versus Conscious Sedation for Mechanical Thrombectomy in Acute Anterior Circulation Ischemic Stroke

Abstract

Background

Retrospective studies suggested that general anesthesia (GA) for mechanical thrombectomy has worse outcomes compared with conscious sedation (CS). However, randomized single‐center trials suggested noninferiority of GA to CS. We investigated the impact of anesthesia techniques on thrombectomy, and hypothesized that the routine use of GA with a defined protocol would not adversely affect thrombectomy delivery or outcomes.

Methods

A total of 451 consecutive patients receiving mechanical thrombectomy for anterior circulation ischemic stroke from 2016 to 2019 were identified from the local registry. Patients were divided into cohort A when both GA and CS were used, and cohort B (from October 2017) when GA became the default method. Favorable functional outcome was defined as modified Rankin scale of 0 to 2 at 3 months. Intraprocedural blood pressures were audited annually.

Results

In cohort A, compared with patients receiving CS, patients with GA had prolonged median arrival to arterial puncture time (26 versus 18 minutes; P<0.001) and comparable favorable functional outcome at 3 months (37.7% versus 45.1%; P=0.355). In cohort B, the median arrival to arterial puncture was reduced to 10 minutes, with comparable favorable functional outcome of 46.7%, and reduced mortality compared with cohort A (14.2% versus 22.7%; P=0.024). Yearly audits demonstrated good adherence to the protocol. Binary logistic regression analysis showed only old age (odds ratio [OR], 1.04; 95% CI, 1.02–1.07 [P=0.003]), high National Institute of Health Stroke Scale at presentation (OR, 1.17; 95% CI, 1.08–1.26 [P<0.001]), and poor collateral status (OR, 0.29; 95% CI, 0.12–0.72 [P=0.008]) were independent factors predicting for poor prognosis, not GA (OR, 0.71; 95% CI, 0.32–1.60 [P=0.408]).

Conclusions

Patients treated under GA for mechanical thrombectomy achieved comparable functional outcome at 3 months compared with those under CS. Through practice and a defined protocol, GA for mechanical thrombectomy can achieve sustainable good functional outcomes. Large clinical trials are needed to confirm these findings.

Nonstandard Abbreviations and Acronyms

AnStroke

Anesthesia During Stroke

CS

conscious sedation

DAWN

Clinical Mismatch in the Triage of Wake Up and Late Presenting Strokes Undergoing Neurointervention With Trevo

DEFUSE 3

Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke 3

GA

general anesthesia

GOLIATH

General or Local Anesthesia in Intra Arterial Therapy

HERMES

Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke Trials

NIHSS

National Institute of Health Stroke Scale

OOH

out of hour

SBP

systolic blood pressure

SIESTA

Sedation versus Intubation for Endovascular Stroke Treatment

WH

working hour

Clinical Perspective

• This study shows that by using a defined anesthetic protocol, patients receiving general anesthesia for thrombectomy can achieve comparably good 90‐day functional outcome and mortality compared with those who receive conscious sedation.

• We provide new evidence that, in real‐world practice, the use of a defined protocol, optimizing operational workflow and training, general anesthesia for thrombectomy can achieve sustainably good outcomes and reduced mortality without delay or hemodynamic instability.

• This study demonstrates a continuous reduction in arrival to recanalization times, improving successful reperfusion, and consistently good clinical outcomes on introduction of a dedicated anesthetic team, a continuous quality improvement program and training.

Mechanical thrombectomy is a highly effective treatment for patients with acute ischemic stroke caused by large vessel occlusion. ¹ However, the optimum choice of sedation for the procedure (general anesthesia [GA] or non‐GA techniques including conscious sedation [CS] or local anesthesia) is still subject to ongoing research. ² , ³ , ⁴ , ⁵

The effect of GA on thrombectomy outcome has produced conflicting results. Several retrospective studies and post hoc analyses report worse outcomes with GA compared with CS. ² , ³ , ⁴ However, these results are confounded by selection bias such that patients with high stroke severity or dominant hemisphere strokes were more likely to be treated under GA. ⁶ Conversely, meta‐analyses of the only 4 single‐center trials ⁷ , ⁸ , ⁹ , ¹⁰ where patients were randomized to either CS or GA for thrombectomy, under protocol‐based techniques and defined targets for intraprocedural blood pressure (BP) management, demonstrated better functional outcome at 3 months in patients with GA. ¹¹ , ¹²

Concerns regarding the safety of GA in thrombectomy remain. There are primarily 2 plausible mechanisms being proposed as to why GA may be associated with a poorer outcome. First, delays accrued through GA provision may contribute to worsening brain damage through irreversible ischemia. Second, GA may cause BP instability; significant hypotension during induction of anesthesia could potentially compromise the collateral flow crucial in maintaining ischemic penumbra.

As the first hospital in the United Kingdom to introduce a dedicated 24/7 thrombectomy service, our practice has evolved since inception, moving from CS as the default to the use of GA as our standard anesthetic technique. By comparing outcomes before and after introduction of this change, we can provide additional information as to the safety of GA in real‐world thrombectomy practice. The primary aim of this study was to examine the safety and efficacy of GA in our thrombectomy service compared with non‐GA techniques. The secondary aim was to assess changing quality in the implementation of GA by comparing processing times and hemodynamic stability before and after the practice change. We hypothesized that the routine use of GA would not adversely affect thrombectomy delivery or outcomes with careful anesthetic management.

Methods

Study Design

The data that support the findings of this study are available from the corresponding author on reasonable request. The data were from the local thrombectomy registry with the approval of the institutional audit committee (AUDI000377). The registry is an ongoing prospective registry of consecutive patients with acute ischemic stroke accepted for thrombectomy treatment.

The thrombectomy service became a 24/7 facility in July 2016 covering a population of 3.5 million. A protocol for anesthesia was produced in line with consensus guidance published in 2014. ¹² This favored the use of CS for thrombectomy and consequently this became the default method unless there were specific indications for GA. Anesthesia support to the thrombectomy service was initially on an ad hoc basis. For occasional cases, thrombectomy would commence with local anesthesia if no anesthetist was immediately available as a result of competing demands in the operating rooms. However, some cases were abandoned because of patient agitation with only CS or local anesthesia. Dedicated anesthetic support was introduced in January 2017, allowing GA to be available for all cases. With increasing experience and the availability of new evidence, the anesthetic protocol evolved, and, in October 2017, GA became the default method.

Considering this, for this study, we used data from July 2016 to September 2017 (cohort A, the use of CS as default) to examine the safety and efficacy of GA for thrombectomy compared with non‐GA. We further analyzed data from October 2017 to December 2019 (cohort B, use of GA as default) and compared it with the GA group in cohort A and the whole of cohort A to investigate changes in thrombectomy processing times and hemodynamic monitoring, and whether these changes translated into better clinical outcomes over time.

The local thrombectomy protocol has been previously reported. ¹³ In brief, inclusion criteria for this study were age ≥18 years, with intracranial proximal arterial occlusion in the anterior circulation, including internal carotid artery terminus, middle cerebral artery M1 and proximal M2, demonstrated by computed tomography (CT) angiography or digital subtraction angiography. There was no systematic alteration in the actual procedural methodology between the 2 periods.

Patients with posterior circulation stroke were excluded from this study to avoid bias as the efficacy of thrombectomy in these cases is still debated, and a significant number of large vessel occlusion in patients with posterior circulation stroke require intubation clinically before thrombectomy.

Anesthesia Protocol

Anesthesia is provided by neuroanesthesists who are present for all thrombectomy procedures. There was no change in anesthetic protocol regarding anesthetic or sedative agents used, BP, and other parameters’ targets throughout the study period, except for introducing a dedicated anesthetic team and the change of default anesthestic methods, from CS to GA. All patients receive the following noninvasive monitoring with or without the presence of an anesthetist: ECG, BP at 2‐minute cycles, oxygen saturation, and capnography for end‐tidal carbon dioxide. Central or arterial lines are not routinely placed. Patients treated under GA have their airway secured; anesthesia is maintained with intravenous or inhalational drugs at the discretion of the anesthetist. The hemodynamic and respiratory parameter targets were set according to anesthetic consensus¹²: systolic BP (SBP) targets of 140 to 160 mm Hg, with an updated protocol permitting limits between 120 and 180 mm Hg; end‐tidal carbon dioxide targets of 4.5 to 5.5 kPa (34–41 mm Hg); and oxygen saturation targets of >94%. Patients receive a vasopressor (metaraminol) infusion from induction of anesthesia titrated when SBP is <140 mm Hg to maintain BP control. Patients treated under CS receive low‐dose short‐acting analgesics and sedatives as needed. After the completion of the procedure, patients treated under GA are extubated in the angiography suite. All patients are transferred to the hyperacute stroke unit as soon as possible after a period of close observation at the postoperative recovery area. Only patients with low Glasgow Coma Scale (GCS) caused by severe stroke, respiratory failure, or an ongoing requirement for BP control are transferred to the neurological intensive care unit (ICU).

As part of ongoing quality improvement, regular audits were conducted to examine the adherence to the anesthetic protocol. Procedural timings, conduct of anesthesia, hemodynamic management, and adverse outcomes from the audit period in each year were recorded. The reviewer, a senior neuroanesthetist, was blinded to clinical outcome.

Outcome Measures and Other Variables of Interest

The primary outcome was favorable functional outcome as defined by a modified Rankin scale score of 0 to 2 at 90 days following procedure, which was collected by telephone interview using a standardized questionnaire. Secondary outcomes included procedural timings, successful recanalization (defined as a modified thrombolysis in cerebral ischemia score of 2b‐3), procedural‐related complications (including vessel dissection or rupture at the arterial puncture site, carotid, and intracranial vessels), and symptomatic intracerebral hemorrhage (classified according to ECASS [European Cooperative Acute Stroke Study II) criteria). ¹⁴ Compliance with the anesthesia protocol was assessed including poor hemodynamic control, as defined by SBP <140 mm Hg for longer than 10 minutes.

Stroke severity was measured using the National Institute of Health Stroke Scale (NIHSS) score before thrombectomy. Collateral status on single‐phase CT angiogram was independently assessed retrospectively by 2 neuroradiologists blinded to clinical outcomes. Collaterals were dichotomized into good (≥50% filling) or poor (<50% filling) in the ischemic hemisphere by CT angiogram in the equilibrium phase. The definition of thrombectomy was arterial puncture in the angiography suite. The choice of endovascular treatment modality was at the discretion of the treating interventional neuroradiologist as previously described. ¹³ Primarily large bore aspiration catheters and stent retrievers of the modern era were used. During the study period, there was no significant change in the interventional device or technique in our practice, although it may be varied among individual interventional neuroradiologists. Arrival time was taken as the time of arrival at the angiography suite for both local and external presenters. Thrombectomy start time was defined as the time of arterial puncture. Reperfusion time was only recorded in patients who achieved successful recanalization with thrombectomy for anesthesia service provision. Working hours (WHs) were defined as cases presenting between 8 am and 8 pm Monday to Friday and 8 am and 6 pm Saturday. Presentation outside these time frames was considered out of hours (OOHs).

Statistical Analysis

For data analysis, medians with interquartile ranges (IQRs) and means with SDs were reported for continuous variables. Percentages and counts were reported for categorical variables. Differences were assessed using independent t test for continuous variables and chi‐square tests for categorical variables. Binary logistic regression was performed to analyze outcome predictors. In all analyses, a 2‐tailed P<0.05 was considered statistically significant. All statistical analyses were performed using SPSS version 24 (IBM).

Results

Outcomes in Patients Treated Under GA Versus Non‐GA

Between July 2016 and September 2017, 163 patients were included in cohort A: 61 treated under GA and 102 under non‐GA (Supplemental Figure S1). Two patients in the non‐GA group were converted to GA during the procedure because of patient agitation. One patient in the GA group was intubated before the thrombectomy because of seizure at stroke onset.

To evaluate whether GA is as safe and effective as non‐GA, we compared the 2 groups in cohort A (Tables 1 and 2). Both groups had similar age, sex distribution, prestroke function (modified Rankin scale), stroke severity at presentation, vascular risk factors, occlusion sites, thrombolysis treatment, and collateral status (Table 1). Compared with the non‐GA group, there were significant delays in the time from arrival at the angiography suite to arterial puncture in the GA group, mainly because of GA induction (median, 26 minutes [IQR, 19–37] in GA versus 18 minutes [IQR, 11–23] in non‐GA; P<0.001) (Table 20). There was no significant difference in the times from onset to arrival (P=0.735), arrival to recanalization (P=0.104), and onset to recanalization (P=0.564). In the non‐GA group, 16 cases (15.7%) recorded patient agitation during the procedure. Of these, 2 cases were converted to GA during the procedure, and 9 cases reported early termination because of safety concerns as a result of agitation. Both groups achieved high rates of successful reperfusion (81.0% in GA versus 75.5% in non‐GA; P=0.694). Both groups had similar good functional outcomes (modified Rankin scale 0–2: 37.7% in GA versus 45.1% in non‐GA; P=0.355), and mortality rates (21.3% in GA versus 23.5% in non‐GA; P=0.774) at 90 days (Table 2).

TABLE 1.

Baseline characteristics for cohorts A and B

	Cohort A
	Non‐GA n=102	GA n=61	P value Cohort A GA vs non‐GA	Cohort B GA n=273	P value GA cohort A vs cohort B	P value Cohort A vs cohort B
Median age (IQR), y	70 (57–78)	65 (52–76)	0.465	72 (59–79)	0.150	0.206
Men, %	44.1	59.0	0.066	48.7	0.146	0.844
Good prestroke function mRS 0 or 1, %	83.3	85.2	0.747	86.4	0.806	0.491
Medical history, %
Hypertension	53.9	50.8	0.701	54.6	0.594	0.731
Atrial fibrillation	38.2	31.1	0.360	37.4	0.361	0.709
Diabetes	15.7	16.4	0.905	14.7	0.730	0.714
Left side occlusion, %	47.1	50.8	0.642	50.2	0.887	0.785
Occlusion site, %
ICA terminus	23.6	32.7	0.393	24.5	0.070	0.151
MCA M1	57.8	54.1	0.555	60.4	0.362	0.487
MCA M2	18.6	13.2	0.360	15.2	0.653	0.744
Median NIHSS score, IQR	18 (14–23)	19 (14–23)	0.545	16 (10–19)	<0.001	<0.001
Collateral status, good, %	66.3	71.2	0.632	73.9	0.320	0.195
Prior intravenous thrombolysis, %	77.5	75.4	0.766	70.3	0.428	0.149

GA indicates general anesthesia; ICA, Internal carotid artery; IQR, interquartile range; MCA, middle cerebral artery; mRS, modified Rankin scale; and NIHSS, National Institute of Health Stroke Scale.

TABLE 2.

Thrombectomy processing times and outcomes for cohorts A and B

	Cohort A
	Non‐GA n=102	GA n=61	P value Cohort AG A vs non‐GA	Cohort B GA n=273	P value GA cohort A vs cohort B	P value Cohort A vs cohort B
Processing times, min
Median onset to arrival time (IQR)	232 (187–274)	240 (173–280)	0.735	274 (200–336)	0.004	<0.001
Median arrival to arterial puncture (IQR)	18 (11–23)	26 (19–37)	<0.001	19 (12–28)	<0.001	0.362
Median arterial puncture to recanalizaton (IQR)	35 (21–57)	40 (19–66)	0.995	25 (15–42)	<0.001	<0.001
Median arrival to recanalizaton, (IQR)	56 (38–75)	69 (48–86)	0.104	48 (34–68)	<0.001	<0.001
Median onset to recanalizaton (IQR)	272 (240–323)	290 (237–327)	0.564	307 (226–370)	0.005	<0.001
Mean no. of pass, ±SD	1.8±1.4	2.3±1.7	0.053	2.0±1.4	0.194	0.804
Successful recanalizaton mTICI 2b/3, % (n)	75.5 (74)*	81.0 (47)*	0.694	86.4 (235)*	0.007	0.002
Good function at 90 d mRS 0–2, % (n)	45.1 (46)	37.7 (23)	0.355	46.7 (122)	0.201†	0.374†
Mortality at 90 d, % (n)	23.5 (24)	21.3 (13)	0.744	14.2 (37)	0.166†	0.024†
sICH, % (n)	0.0 (0)	3.1 (5)	0.117	2.2 (6)	0.783	0.575
Complication, % (n)	6.8 (7)	8.2 (5)	0.528	5.1 (14)	0.161	0.859
ICU admission post‐MT, % (n)	10.8 (11)	29.5 (18)	0.002	17.9 (49)	0.042	0.967

ICU indicates intensive care unit; IQR, interquartile range; mRS, modified Rankin scale; MT, mechanical thrombectomy; mTICI, modified treatment in cerebral ischemic score; NIHSS, National Institute of Health Stroke Scale; and sICH, symptomatic intracerebral hemorrhage.

^*Seven patients in cohort A (3 in the general anesthesia (GA) group, 4 in the non‐GA group), and 11 patients in cohort B did not have thrombectomy because of lack of target on angiogram or inaccessibility to reach the clot.

†Twelve patients did not have clinical outcome at 90 days.

There were 7 cases in the non‐GA group where complications were recorded (4 femoral artery, 2 carotid artery, and 1 intracranial artery), and 5 cases in the GA group (1 femoral artery and 4 intracranial artery). There were more groin complications in the non‐GA group, whereas more intracranial complications occurred in the GA group. There was no symptomatic intracerebral hemorrhage in the non‐GA group versus 2 cases in the GA group (P=0.117) (Table 2).

In cohort A, 29 patients required ICU admission (18 [29.5%] in GA and 11 [10.8%] in non‐GA groups) (Table 2). The majority of cases were caused by low GCS secondary to severe stroke syndromes (14 in GA versus 6 in non‐GA). Only 5 patients in the GA group had delayed extubation (>24 hours): 4 attributable to low GCS and only 1 attributable to ventilatory failure secondary to aspiration pneumonia (Supplemental Table S1). Binary logistic regression analysis showed that only old age (odds ratio [OR], 1.04; 95% CI, 1.02–1.07 [P=0.003]), higher NIHSS score at presentation (OR, 1.17; 95% CI, 1.08–1.26 [P<0.001]), and poor collateral status (OR, 0.29; 95% CI, 0.12–0.72 [P=0.008]) were independent factors predicting poor prognosis; GA was not an independent predicting factor (OR, 0.71; 95% CI, 0.32–1.60 [P=0.408]).

BP Control and Anesthesia

In line with consensus guidelines, our anesthetic protocol targets an SBP ≥140 mm Hg with failure to meet a hemodynamic target, defined as SBP <140 mm Hg for >10 minutes. The audit in the period of 2016 included 36 patients, with 4 (11.1%) undergoing GA. The mean SBP range was 134 to 161 mm Hg, with 2 patients (5.5%) failing to meet the hemodynamic target at some stage during the procedure. Three (75%) of the patients with GA required the use of a vasopressor. The second audit in the period of 2017 included 81 patients, with 50 (61.7%) receiving GA. The mean SBP range was 116 to 165 mm Hg, with 7 (8.6%) patients failing to meet the target. A total of 47 (94%) patients in the GA group and 2 (6%) patients in the non‐GA group required a vasopressor (Table 3). These 2 audits were presented at national meetings. ¹⁵ During the 2018 period, 22 patients were audited, all of whom received GA. The mean SBP range was 126 to 171 mm Hg, and 7 (31.8%) patients failed to meet the target at some stage during the procedure. Vasopressor use occurred in 14 (63.5%) patients. During 2019, 30 patients were audited, with 29 (96.7%) treated under GA. The mean SBP range was 111 to 154 mm Hg, and 7 (23.3%) patients failed to meet the target. A total of 25 (86.2%) patients received vasopressors (Table 3). Inhalational anesthesia was used in all audited patients.

TABLE 3.

Yearly hemodynamic management audits during the study period

Audit y	GA: non‐GA, n (%)	Mean SBP range, mm Hg	Failure to meet hemodynamic target, n (%)	Vasopressor use GA: non‐GA, n (%)
2016 N=36	4 (11.1): 32 (88.9)	134–161	2 (5.5)	3 (75.0): 0
2017 N=81	50 (61.7): 31 (38.3)	116–165	7 (8.6)	47 (94.0): 2 (6.5)
2018 N=22	22 (100)	126–171	7 (31.8)	14 (63.5): N/A
2019 N=30	29 (96.7): 1 (3.3)	111–154	7 (23.3)	25 (86.2): 0

GA indicates general anesthesia; and N/A, not available.

Failure to reach hemodynamic target defined by systolic blood pressure (SBP) <140 mm Hg for >10 minutes.

GA in Real‐World Practice

Between October 2017 and December 2019, 273 patients who had thrombectomy under GA were included in cohort B (Supplemental Figure S1). Compared with patients in cohort A, cohort B patients had comparable baseline function, risk factors, occlusion sites, collateral status, and thrombolysis treatment, except that cohort B had a lower NIHSS score at presentation compared with patients in cohort A (median of 16 [IQR, 10–19] in cohort B versus median of 19 [IQR, 14–23] in cohort A; P<0.001) and GA only group in cohort A (versus median of 19 [IQR, 14–23] in GA cohort A; P<0.001) (Table 1). Cohort B patients achieved a significantly higher successful recanalization rate of 86.4% when compared with cohort A (77.6%; P=0.002) and patients with GA in cohort A (81.0%; P=0.007). Cohort B had comparable rates of favorable functional outcome (modified Rankin scale, 0–2) compared with cohort A (46.7% in cohort B versus 42.3% in cohort A; P=0.374), but a significantly lower mortality rate (14.2% in cohort B versus 22.7% in cohort A; P=0.024) at 90 days (Table 2). There were 14 (5.1%) complications in cohort B (6 groin, 4 carotid artery, and 4 intracranial artery). There was no significant difference in complication and symptomatic intracerebral hemorrhage rates among all groups (Table 2). There was no significant difference in ICU admission between cohorts A and B (Table 2).

The times of onset to recanalizaton were significantly longer in cohort B compared with cohort A (median of 307 minutes [IQR, 226–370] in cohort B versus median of 276 minutes [IQR, 240–326] in cohort A; P<0.001), which was primarily driven by the prolonged onset to arrival times in cohort B (median of 274 minutes [IQR, 200–336] in cohort B versus median of 236 minutes [IQR, 184–278] in cohort A; P<0.001). Although there was no significant difference in arrival to arterial puncture between cohort A and B, it was significantly shorter in cohort B compared with GA only in cohort A (median of 19 minutes [IQR, 12–28] in cohort B versus median of 26 minutes [IQR, 19–37] in GA only, cohort A; P<0.001). The times from arterial puncture to recanalization were also significantly reduced in cohort B compared with cohort A (median of 25 minutes [IQR, 15–42] in cohort B versus median of 37 minutes [IQR, 19–61] in cohort A; P<0.001) (Table 2).

The quarterly median processing times of our thrombectomy service are shown in Figure 1. There was a downtrend in times of arrival to recanalization, primarily driven by the continuous reduction in time of arrival to arterial puncture: from a median of 26 minutes (IQR, 19–37) in the GA group in cohort A to a median of 10 minutes (IQR, 8–16) in October to December 2019 in cohort B, shorter than the non‐GA group in cohort A (median 18 minutes [IQR, 11–23]) (Figure 1). Procedural times (arterial puncture to recanalization) fluctuated over time. Overall, there was a significant reduction in the times of arrival to recanalization (Table 2). Comparisons of processing times during WHs versus OOHs are shown in Figure 2. In cohort A, 77 (47.2%) thrombectomy cases were performed during OOHs. There were nonsignificant delays in arrival to arterial puncture (P=0.701), arterial puncture to recanalization (P=0.150), and arrival to recanalization (P=0.148) (Figure 2A); whereas in cohort B, 133 (48.7%) cases were performed outside WH. It was comparable between OOHs and WHs among the processing times in cohort B (Figure 2B).

FIGURE 1.

Continuous improvement in processing times in cohorts A and B. AP‐recan indicates arterial puncture to recanalization; Arr‐AP, arrival to arterial puncture; Arr‐recan, arrival to recanalization; and GA, general anesthesia.

FIGURE 2.

Thrombectomy processing times in working hours (WHs) versus out of hours (OOHs; cohort A and cohort B). Comparing processing times (arrival to arterial puncture, arterial puncture to recanalization, and arrival to recanalization) between WHs versus OOHs in cohort A (panel A) and cohort B (panel B). Bars are medians with interquartile range bars. There was no significant change between WHs versus OOHs.

Discussion

This study showed that patients with anterior circulation stroke who received GA for thrombectomy achieved comparably good 90‐day functional outcome and mortality rates compared with those who received CS. In real‐world practice, through experience, improved operational workflow, and a defined protocol, GA for thrombectomy can achieve sustainably good outcomes and reduced mortality without delay or hemodynamic instability.

These findings are in contrast to the HERMES (Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke Trials) post hoc analysis ⁴ of trials designed to examine the efficacy of thrombectomy versus standard medical care. In these studies, the need for GA was not part of the selection criteria or randomization, nor were there any unified anesthetic protocols. In the HERMES analysis, ⁴ patients receiving GA were more likely to have a higher NIHSS score (P=0.09), implying potential patient selection bias. The HERMES findings were supported by most of the retrospective studies, ² , ³ , ⁴ which may have been subject to similar selection bias. In our cohort A, the anesthetic method was dictated by the current consensus and availability of anesthetists. Although this practice could not claim to be free from selection bias, there were no significant differences in baseline characteristics in patients undergoing GA versus non‐GA in cohort A. Our data are consistent with the only 4 randomized controlled trials available to date. ⁷ , ⁸ , ⁹ , ¹⁰ Although these clinical trials were single‐center small studies, they provide direct evidence that GA is at least as good as CS in the outcomes of patients with stroke undergoing thrombectomy, confirmed by meta‐analyses. ¹¹ , ¹²

One of 2 potential disadvantages of GA is the delay in procedural starting time attributable to GA induction. However, one of GA's advantages is reduced patient motion allowing for quicker, potentially safer, and possibly more complete recanalization. In our study cohort A (Table 2, cohort A), there was a delay from angiography suite arrival to arterial puncture in the GA group compared with the non‐GA group. The processing times were further delayed in OOHs compared with WHs (Figure 2A), which were mostly driven by the prolonged procedural time (arterial puncture to recanalization) before having a dedicated neuroanesthetic team available for thrombectomy. By October 2017, when GA became our default method, we had set up a clear workflow, including detailed anesthetic‐related information on referral, using a thrombectomy procedure–specific World Health Organization checklist at handover on arrival, and following a clear anesthetic protocol. Thrombectomy teaching, including simulation training, was implemented to further streamline this highly time‐pressured service. ¹⁶ With practice, our median arrival to arterial puncture time was reduced to 10 minutes (Figure 1). Under GA, the average procedural times were significantly reduced, resulting in significantly reduced arrival to recanalization times. This was maintained during OOHs in cohort B (Figure 2B).

In this study, although GA reduced average procedural times, it fluctuated over time (Table 2, Figure 1), as it was mostly dictated by the complexity of the cases. Our data showed that under GA, the interventional neuroradiologist would perform more attempts to achieve optimal results (number of passes cohort A, Table 2). However, multiple passes might have contributed to a nonsignificant increase in intracranial complications in the GA group compared with the non‐GA group in cohort A, which improved with experience afforded by time (cohort B). In this study, patient agitation in the non‐GA group resulted in premature termination of the thrombectomy procedure in 9 cases, which may have adversely affected those patients’ clinical outcome. Problematic movements during thrombectomy were also reported in a substantive percentage of the CS group (6%–33%) in 3 clinical trials, ⁷ , ⁸ , ⁹ which could be avoided under GA.

The other potential drawback of GA is that it may cause intraprocedural hypotension, resulting in hastening of the progression of ischemia before recanalization. ¹⁷ In previous multicenter thrombectomy trials and retrospective studies, there were no reports of using consistent anesthetic protocols or a clear set of physiological parameter targets, such as BP. By contrast, 3 randomized clinical trials ⁷ , ⁸ , ⁹ had specific protocols for intraprocedural BP control that are consistent with the consensus guideline. ¹⁴ BP was lower in the GA arm of both the AnStroke (Anesthesia During Stroke) and GOLIATH (General or Local Anesthesia in Intra Arterial Therapy) trials, with no difference seen in the SIESTA (Sedation versus Intubation for Endovascular Stroke Treatment) trial. ⁷ , ⁸ , ⁹ Our anesthetic protocol includes clear BP parameters with a detailed treatment plan to ensure adequate BP maintenance. All episodes of hypotension were promptly treated with vasopressor boluses. Yearly audits during the study period demonstrated good adherence to targets achieved with increased use of vasopressors, comparable to the results in the clinical trials. ⁷ , ⁸ , ⁹ Our data and the randomized clinical trials showed that, with a defined protocol, BP can be well controlled during thrombectomy.

In our cohort B study period, 15 of the 288 (5.2%) consecutive patients who received thrombectomy for anterior circulation stroke were performed under CS as they were deemed unsuitable for GA (Supplemental Figure S1). By contrast, ≈40% of patients with thrombectomy were excluded from the 3 randomized trials. ⁷ , ⁸ , ⁹ , ¹⁸ In 3 clinical trials, 28 patients (11.5%) randomized to the sedation group were later converted to GA on an emergency basis because of agitation or aspiration. ⁵ In cohort A, we had 2 patients who required conversion to GA. The change of anesthesia may cause delay or interruption of the procedure and may be associated with worse outcome. ¹⁹ It has been reported that at least 20% of patients with anterior circulation strokes are not favorable candidates for CS only. ⁷ , ⁹ , ²⁰

In our study, the NIHSS scores on admission in cohort B were lower than for patients in cohort A; however, they were comparable with meta‐analysis data. ⁵ , ¹¹ The times from symptom onset to recanalization were significantly longer in cohort B compared with cohort A, which was mainly attributed by the prolonged times of onset to arrival in cohort B. The prolonged onset to arrival is attributable to the relaxation of the inclusion criteria on the time from onset to groin puncture after the publication of the DAWN (Clinical Mismatch in the Triage of Wake Up and Late Presenting Strokes Undergoing Neurointervention With Trevo) and DEFUSE 3 (Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke 3) studies. ²¹ , ²² Despite the prolonged symptom onset to recanalization, patients in cohort B achieved higher favorable clinical outcomes and a lower mortality rate at 90 days compared with cohort A (Table 2). The clinical outcomes at 90 days in cohort B were comparable with those in the GA groups of the randomized clinical trials. ⁵ , ¹¹ Our data support the employment of GA in thrombectomy with caveats around the provision of a comprehensive GA protocol enshrined within a robust operational procedure as a feasible approach in real‐world practice.

It has been suggested that intubation in GA during thrombectomy may increase the risk of pneumonia and ICU admission (where postthrombectomy ICU admission is not part of routine practice) as compared with CS anesthesia. Of the 3 clinical trials, 49% of patients treated under GA in SIESTA were still intubated 2 hours after the completion of thrombectomy, with the other trials recording lower percentages (AnStroke [7%] and GOLIATH [3%]). ⁷ , ⁸ , ⁹ In our practice, we aimed for extubation at the end of thrombectomy in the angiography suite. Our cohort A data showed higher ICU admission in patients undergoing GA compared with non‐GA (Supplemental Table S1). These ICU admissions were mostly attributable to low GCS secondary to the severity of stroke syndromes. Only 5 cases had prolonged intubation, again mostly attributable to low GCS. We did not have complete data for postthrombectomy pneumonia, owing to rapid repatriation of externally referred patients, which could be as early as 30 hours after thrombectomy. However, only 1 patient in the GA group required prolonged intubation because of ventilation failure secondary to aspiration pneumonia

This study has limitations. Its retrospective methodology is inherently subject to possible selection and information bias, including the anesthetic choices in cohort A. Although we do not have complete hemodynamic data, our audits and continuous quality improvement work demonstrate the consistent adherence to our protocol. Finally, this is a single‐center study; however, being a single‐center study, akin to all 4 randomized controlled clinical trials, ⁷ , ⁸ , ⁹ , ¹⁰ may have its advantages: data from the same intervention team with a specific anesthetic protocol may reduce procedural‐related discrepancy. Although we are a large regional center with great ethnic diversity in the population, our practice may not be generalizable to centers without a team dedicated to thrombectomy. This work demonstrates that it is a combination of tight BP control and the overall experience of the teams involved enabling timely reperfusion that are important, rather than the choice of anesthetic techniques, GA versus CS. Further trials are needed to confirm these findings.

Conclusions

Despite previous reports of worse outcome with GA, more recent evidence from randomized clinical trials suggests that protocol‐driven general anesthesia is no worse and may even be associated with better outcomes. This study demonstrates that patients with anterior circulation stroke who received GA for thrombectomy achieved comparable functional outcome at 3 months compared with those receiving CS. Our real‐world practice demonstrated that through practice, improved workflow, and defined protocols, thrombectomy under general anesthesia can be performed without delay or hemodynamic instability and achieve sustainably good outcomes.

Author Contributions

L.Z. and M.E. designed the study, had full access to all data in the study, and took responsibility for the integrity and accuracy of the data and data interpretation. L.Z. performed data analysis and drafted the article. All authors contributed to data interpretation and reviewed and approved the final version of the article.

Funding

None of the authors received any funding in connection with this project.

Disclosures

J.L. receives consultancy agreements with Medtronic, Stryker, and Microvention Terumo.

Supporting information

Table S1 Figure S1

Supporting information.