Retrospective Studies

Observational, predominantly retrospective studies, which have attempted to evaluate the association between anesthetic strategy and clinical outcomes among patients with AIS managed with EVT, have reported mixed results (see Table 2). CS has been associated with improved functional outcomes ^{Reference Sugg, Jackson, Holloway, Martin, Akhtar and Rymer3–Reference Feil, Herzberg and Dorn14} and reduced mortality ^{Reference Abou-Chebl, Lin and Hussain4–Reference McDonald, Brinjikji, Rabinstein, Cloft, Lanzino and Kallmes7,Reference Just, Rizek, Tryphonopoulos, Pelz and Arango9,Reference Powers, Dornbos and Mlynash13,Reference Feil, Herzberg and Dorn14} compared with GA; others have reported no difference in outcomes between CS and GA. ^{Reference Li, Deshaies and Singla26–Reference Byrappa, Lamperti, Ruzhyla, Killian, John and St Lee31} LA has similarly been associated with better functional outcomes compared with GA, ^{Reference Davis, Menon and Baghirzada15–Reference Pop, Severac and Happi Ngankou18} often with reduced mortality, ^{Reference Davis, Menon and Baghirzada15,Reference Abou-Chebl, Yeatts and Yan16} although these findings are contrasted by those from a subanalysis of Trial and Cost Effectiveness Evaluation of Intra-arterial Thrombectomy in Acute Ischemic Stroke ^{Reference Bracard, Ducrocq and Mas32} and a prospective study by Wu et al. ^{Reference Wu, Jadhav and Zhao33} Only two studies have directly compared outcomes in those managed with CS versus LA. Work from 1034 patients in the Endovascular Treatment in Ischemic Stroke registry in France reported higher rates of good functional outcome with CS versus LA (mRS 0–2 at 3 months CS 52% vs LA 40%, p = 0.028). ^{Reference Benvegnù, Richard and Marnat34} A retrospective review from one of the MR CLEAN centers reported contrasting results (mRS 0–2 at 3 months CS 22% vs LA 47%, OR 0.4 [0.2–0.8]). ^{Reference van de Graaf, Samuels and Mulder35} Both Goldhoorn et al. and Cappellari et al. retrospectively evaluated outcomes in patients with LA, CS, or GA. In both studies, those managed with LA had improved functional outcomes (Goldhoorn mRS 0–2 GA 35%, CS 25%, LA 41%, p ≤ 0.01; Cappellari mRS 0–2 GA 42.5%, CS 46.6%, LA 52.4%, p ≤ 0.001) and reduced mortality (Goldhoorn GA 32%, CS 36%, LA 27%, p = 0.04; Cappellari GA 21.5% CS 19.7% LA 14.8%, p ≤ 0.001). ^{Reference Goldhoorn, Bernsen and Hofmeijer36,Reference Cappellari, Pracucci and Forlivesi37} The range of rates of good outcome among these various studies is large implying likely differences in the populations under study. Interpretation of results from these nonrandomized studies is limited (inherent bias in trial design, combination of LA and CS into single comparator group); however, findings suggest improved outcomes in those managed without GA and point toward better outcomes in patients managed with LA rather than CS. In addition to improved outcomes, patients managed without GA have a shorter length of stay as reported by powers (GA 8.02 days (5.35–12.18 days), CS 5.93 days (3.31–8.85 days), p = 0.03), ^{Reference Powers, Dornbos and Mlynash13} and Bekelis et al. (GA 19.6 days, CS 11.7 days, unadjusted difference 7.9 (CI 5.1–10.7) ^{Reference Bekelis, Missios, MacKenzie, Tjoumakaris and Jabbour38} and a reduced cost of hospitalization (GA $USD 34,903 ($25,530–55,444), CS $26,775 ($18,790–39,935), p ≤ 0.0001). ^{Reference McDonald, Brinjikji, Rabinstein, Cloft, Lanzino and Kallmes7}

Table 2: Nonrandomized studies

adOR=adjusted odds ratio; CS=conscious sedation; CV=conversion from LA or CS to GA; GA=general anesthetic; ICH=intracranial hemorrhage; LA=local anesthesia; MAC=minimum alveolar concentration; mRS=modified Rankin scale; NA=nonapplicable – propensity matched; NS=not stated; OR=odds ratio; P-GA=planned GA; SICH=symptomatic intracranial hemorrhage; TICI=thrombolysis in cerebral infarction; UA=undetermined anesthesia; U-GA=unplanned GA.

mRS scores recorded at 3 months unless otherwise stated. Only significant p values provided.

* GA associated with lower rate of functional independence when mRS corrected for differences in baseline characteristics (OR 0.32; p = 0.05).

**T hese patients were excluded.

Anesthesia-Specific RCTs

Five single center RCTs have examined the influence of anesthetic strategy on outcomes among patients with AIS due to anterior circulation LVO (see Tables 3–5). Overall rates of functional independence (mRS 0–2 at 90 days) were variable, ranging from 27.3% in SIESTA ^{Reference Schönenberger, Uhlmann and Hacke25} to 59.4% in GOLIATH. ^{Reference Simonsen, Yoo and Sørensen21} Similar variability was seen in the HERMES trials, with the most comparable (by anesthetic strategy) reporting functional independence in 32.6% patients in MR CLEAN (38% GA), ^{Reference Berkhemer, Fransen and Beumer41} 60.0% in Solitaire With the Intention For Thrombectomy as PRIMary Endovascular Treatment (37% GA), ^{Reference Saver, Goyal and Bonafe42} and 71.0% in Extending the Time for Thrombolysis in Emergency Neurological Deficits – Intra-Arterial (36% GA). ^{Reference Campbell, Mitchell and Kleinig43} Only two of the anesthesia-specific RCTs, ANSTROKE and the work by Ren et al., specifically evaluated the impact of anesthetic outcome on functional status 3 months poststroke. ^{Reference Ren, Xu, Liu, Liu, Wang and Gao23,Reference Löwhagen Hendén, Rentzos and Karlsson24}

Table 3: Anesthesia-specific RCTs – baseline characteristics

ANSTROKE=Anesthesia During Stroke Trial; ASPECTS=Alberta Stroke Programme Early CT Score; CANVAS=Choice of ANesthesia for EndoVAScular Treatment of Acute Ischemic Stroke; GOLIATH=General Or Local anesthesia in Intra Arterial Therapy; HERMES=Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke trials; mRS=modified Rankin scale; NIHSS=National Institutes of Health Stroke Scale; NS=not stated; SIESTA=Sedation versus Intubation for Endovascular Stroke Treatment; TPA=Tissue Plasminogen Activator.

Table 4: Anesthesia-specific RCTs – anesthetic protocols

BIS=bispectral index; CS=conscious sedation; DEX=dexmedetomidine; ET CO₂=end tidal carbon dioxide; GA=general anesthetic; MAP=mean arterial pressure; NS=not stated; RASS=Richmond Agitation Sedation Scale; SaO₂=arterial oxygen saturation; SBP=systolic blood pressure.

Table 5: Anesthesia-specific RCTs – outcomes

ANSTROKE=Anesthesia During Stroke Trial; CANVAS=Choice of ANesthesia for EndoVAScular Treatment of Acute Ischemic Stroke; GA=general anesthesia; GOLIATH=General Or Local anesthesia in Intra Arterial Therapy; HERMES=Highly Effective Reperfusion Evaluated in Multiple Endovascular Stroke trials; MAP=mean arterial pressure; mRS=modified Rankin scale; NS=not stated; SIESTA=Sedation versus Intubation for Endovascular Stroke Treatment; TICI=thrombolysis in cerebral infarction.

*p ≤ 0.05, **p ≤ 0.001.

# SIESTA reported critical hypo (SBP < 120 mmHg) and hypertension (SBP > 180 mmHg) rather than MAP values.

⁺GOLIATH used time MAP < 70 mmHg as similar index.

^HERMES reported onset to randomization and randomization to reperfusion rather than puncture to recanalization.

∼Ren et al. reported vasopressor use for each individual agent. For CS range 11.9%–23.8%. For GA range 6.25%–22.9%. No significant difference between groups.

SIESTA (Sedation versus Intubation for Endovascular Stroke Treatment) randomized 150 patients (73 GA, 77 CS) from the University of Heidelberg in Germany. GA was the standard of care prior to trial commencement; CS was the intervention in the trial. The CS protocol, implemented 6 months prior to study commencement, targeted a Richmond Agitation Sedation Scale of −3 to −2. There was no difference in primary outcome, change in NIHSS at 24 hours, between groups (GA mean −3.2, CS mean −3.6, difference −0.4 (−3.4–2.7) p = 0.82), although patients managed with GA were significantly more likely to achieve functional independence at 3 months (GA 37%, CS 18.2%, difference −18.8 (−32.8 to −4.8), p = 0.01). ^{Reference Schönenberger, Uhlmann and Hacke25}

ANSTROKE (Anesthesia During Stroke Trial) randomized 90 patients (45 GA, 45 CS) from Sahlgrenska University Hospital in Sweden who presented within 8 hours of onset. GA was considered the intervention, suggesting CS was the standard of care prior to trial enrollment. Neither the primary endpoint of percentage of patients that achieved an mRS of 0–2 at 3 months (GA 42.2%, CS 40%, p = 1.00) nor the median mRS at 3 months (GA 3 (1–4), CS 3 (1–5.5), p = 0.5001) differed by anesthetic strategy. ^{Reference Löwhagen Hendén, Rentzos and Karlsson24}

In GOLIATH (General Or Local anesthesia in Intra Arterial Therapy), both GA and CS were routinely used at the Aarhus University Hospital, Denmark prior to trial commencement. Similar to ANSTROKE, no target level of sedation was provided in the trial protocol. The primary outcome measure, infarct growth on MRI at 48–72 h, did not differ by anesthetic strategy (GA 8.2 ml (2.2–38.6), CS 19.4 ml (2.4–79.0), p = 0.10), although final infarct volume was lower in those managed with GA (GA 22.3 ml (8.1–64.5), CS 38.0 ml (16.7–128.0), p = 0.04), potentially as a result of higher successful recanalization rates (TICI 2b-3) with GA (GA 76.9%, CS 60.3%, p = 0.04). Those managed with GA had better functional outcomes at 3 months (mRS 0–2 GA 66.2%, CS 52.4%; median mRS GA 2 (1–3), CS 2 (1–4), p = 0.04). ^{Reference Simonsen, Yoo and Sørensen21}

The CANVAS (Choice of ANesthesia for EndoVAScular Treatment of Acute Ischemic Stroke) trial assessed recruitment rates and rates of conversion from CS to GA to facilitate a larger trial evaluating the influence of anesthetic strategy on clinical outcomes (in progress). The target level of sedation in the CS group was a bispectral index of 70 or more, so that patients maintained normal, purposeful responses to stimuli, and were able to protect their own airway. Only 31.4% of patients who presented to Beijing Tiantan Hospital Capital Medical University with LVO were enrolled in the trial (40 patients, GA 20, CS 20); the most common reason for exclusion was LVO of the posterior circulation. A fifth (18.2%) of CS patients were converted to GA. Functional outcomes at 3 months were similar across groups (mRS 0–2 GA 55%, CS 50%), although mortality was higher with CS (GA 5%, CS 30%). ^{Reference Sun, Liang and Wu22}

Most recently, Ren et al. randomized 90 patients (GA 48, CS 42), using the same depth of CS as SIESTA and demonstrated no difference in outcomes by anesthetic strategy (median mRS at discharge 2 for GA and CS, p = 0.890; mRS 3 months 2.5 for GA and CS, p = 0.796). ^{Reference Ren, Xu, Liu, Liu, Wang and Gao23}

Rates of symptomatic ICH (SICH) did not differ by anesthetic strategy in the anesthesia-specific RCTs, weakening the argument that GA is necessary to prevent excessive patient movement to reduce the risk of vessel perforation. With the exception of Ren et al., rates of SICH in the anesthesia-specific RCTs were comparable to results from the HERMES collaboration (SICH Ren et al. GA 18.8%, CS 16.7%; HERMES GA 4%, CS 4%). ^{Reference Campbell, van Zwam and Goyal20,Reference Ren, Xu, Liu, Liu, Wang and Gao23} GA was associated with higher rates of pneumonia in SIESTA (GA 13.7%, CS 3.9%, p = 0.03) ^{Reference Schönenberger, Uhlmann and Hacke25} and the work by Ren et al. (GA 20.8%, CS 4.8%, p = 0.031). ^{Reference Ren, Xu, Liu, Liu, Wang and Gao23} Mortality rates were also similar to the HERMES collaboration results, ranging from 10% to 20% at 3 months. ^{Reference Campbell, van Zwam and Goyal20–Reference Löwhagen Hendén, Rentzos and Karlsson24} SIESTA had the highest mortality rate at 24.7% in both groups. ^{Reference Schönenberger, Uhlmann and Hacke25} There was a trend toward increased mortality in the CS arm in ANSTROKE, GOLIATH, and the CANVAS pilot; however, these differences were not statistically significant. ^{Reference Simonsen, Yoo and Sørensen21,Reference Sun, Liang and Wu22,Reference Löwhagen Hendén, Rentzos and Karlsson24}

Meta-Analyses

Campbell et al., summarized data using meta-analysis from SIESTA, ANSTROKE, GOLIATH, and CANVAS concluding that those managed with GA were more likely to achieve a favorable functional outcome at 3 months (mRS 0–2 GA 49.3%, CS 36.6%, OR 1.71 (CI 1.13–2.59), p = 0.01), a result which may have been partially attributable to higher recanalization rates in those managed with GA (GA 86.2%, CS 74.6%, OR 2.14 (CI 1.26–3.62), p = 0.0050). ICH rates were similar across groups (GA 2.5%, CS 4.9%, OR 0.61 (CI 0.2–1.85), p = 0.38). ^{Reference Campbell, Diprose, Deng and Barber44} Work by Zhang et al., following analysis of results from SIESTA, ANSTROKE, and GOLIATH, also demonstrated improved functional outcomes in those managed with GA (OR 1.87 (CI 1.15–3.03)). Increased successful recanalization (TICI 2b-3) was also seen with GA (OR 1.94 (CI 1.13–3.35)), with no difference in SICH or periprocedural complications. ^{Reference Zhang, Jia, Fang, Ma, Cai and Faramand45}

Findings from the anesthesia-specific RCTs and meta-analyses are in contrast to those from the HERMES collaboration. In HERMES, with the exception of MR CLEAN, anesthetic agents and protocols were not prespecified, and as such were more likely to reflect “real-world” practice. In HERMES, those managed without GA had significantly better functional outcomes compared to those managed under GA (covariate adjusted cOR 1.53 (CI 1.14–2.04), p = 0.0044), with 40% in the GA group achieving an independent functional outcome at 3 months compared with 50% in the no GA group (OR 1.65 (CI 1.14–2.38), p = 0.0078). An excellent functional outcome (defined as an mRS 0–1) was also more likely in those managed without GA (GA 23%, no GA 32%, OR 1.68 (CI 1.12–2.52, p = 0.013). These between-group differences were significant in that for every 100 patients managed with GA (rather than no GA), 18 would have poorer functional outcomes and 10 would not achieve functional independence. ^{Reference Campbell, van Zwam and Goyal20}

Validity and generalizability of results

Results from nonrandomized and randomized studies evaluating the influence of anesthetic strategy on clinical outcomes in patients with AIS managed with EVT are inconsistent. The outcomes from the nonrandomized studies suffer from selection bias and confounding by indication, but by how much is unknown and whether this is enough to account for differences between these cohort studies and the single center RCTs remains undetermined. There may also be residual confounding due to unmeasured sources of unknown bias. Those with more severe neurologic impairment on presentation (higher NIHSS scores, altered conscious state, agitation) or multiple preexisting comorbidities are more likely to be allocated to GA. This is supported by findings from a post hoc analysis of the International Management of Stroke III trial; patients with a medical indication for GA were less likely to have a good outcome (mRS 0–2 medically indicated GA 19.7%, LA 48%, adRR 0.49 (CI 0.30–0.81), p = 0.005) and had increased mortality (medically indicated GA 33.8%, LA 7.4%, adRR 3.93 (CI 2.18–7.10), p ≤ 0.0001) compared with LA, while there was no difference in probability of good outcome (mRS 0–2 routine GA 40.8%, LA 48%, adRR 0.80 (CI 0.60–1.06), p = 0.12) and mortality (routine GA 13.2%, LA 7.4%, adRR 1.82 (CI 0.87–3.77), p = 0.11) between LA and routine intubation. Higher in-hospital mortality was also seen among the medically indicated GA compared with routine intubation cohort (medically indicated GA 33.8%, routine GA 13.2%, adRR 2.16 (CI 1.09–4.29), p = 0.0274). ^{Reference Abou-Chebl, Yeatts and Yan16}

While findings from the anesthesia-specific RCTs suggest that GA is noninferior to CS or LA in patients with AIS undergoing EVT, there are a number of caveats which limit the applicability of these results to other EVT capable centers. Firstly, GA was the standard of care, and CS the intervention, in the majority of the anesthesia-specific RCTs, which may have made it more difficult to demonstrate a benefit of CS over GA. Anesthesia was provided by highly specialized neuroanesthesia and neurocritical care teams able to provide 24 hours in-hospital EVT coverage. Delays due to GA were minimal as teams were fast: arrival to puncture was only 11 minutes in the GA and CS group in the study by Ren et al. ^{Reference Ren, Xu, Liu, Liu, Wang and Gao23} and delays due to GA were 10 minutes or less in SIESTA, GOLIATH, and ANSTROKE, ^{Reference Simonsen, Yoo and Sørensen21,Reference Löwhagen Hendén, Rentzos and Karlsson24,Reference Schönenberger, Uhlmann and Hacke25} in contrast to 20 minutes longer with GA in HERMES (GA 105 minutes (80–149 minutes) vs non-GA 85 minutes (51–118 minutes) p ≤ 0.0001). ^{Reference Campbell, van Zwam and Goyal20} The type and depth of CS varied within and across trials, and a target depth for CS was not provided in some protocols (GOLIATH, ANSTROKE), making comparison across studies difficult. In some cases, the “noGA” group included those managed with both LA and CS (which could be deep), potentially masking a benefit of LA over GA. The small sample size used in each trial further limits the widespread applicability of these results; the largest anesthesia-specific RCT, SIESTA, enrolled only 150 patients. ^{Reference Schönenberger, Uhlmann and Hacke25} Lastly, functional outcomes were only evaluated as a primary outcome measure in ANSTROKE and the study by Ren et al. ^{Reference Ren, Xu, Liu, Liu, Wang and Gao23,Reference Löwhagen Hendén, Rentzos and Karlsson24} Taken together, these limitations suggest that the functionality and expertise of the team in highly experienced centers is a critical factor. Highly experienced, fast teams can use GA effectively and without loss of clinical efficacy of EVT, but this may not be generalizable.