Anesthesia technique and outcomes of mechanical thrombectomy in patients with acute ischemic stroke

Abstract

Background and Purpose

The impact of anesthesia technique on the outcomes of mechanical thrombectomy for acute ischemic stroke remains an issue of debate. We investigated the association of general anesthesia with outcomes in patients undergoing mechanical thrombectomy for ischemic stroke.

Methods

We performed a cohort study involving patients undergoing mechanical thrombectomy for ischemic stroke from 2009-2013, who were registered in the New York Statewide Planning and Research Cooperative System (SPARCS) database. An instrumental variable (hospital rate of general anesthesia) analysis was used to simulate the effects of randomization and investigate the association of anesthesia technique with case-fatality and length of stay (LOS).

Results

Among 1,174 patients, 441 (37.6%) underwent general anesthesia, and 733 (62.4%) underwent conscious sedation. Employing an instrumental variable analysis, we identified that general anesthesia was associated with a 6.4% increased case-fatality (95% CI, 1.9% to 11.0%), and 8.4 days longer LOS (95% CI, 2.9 to 14.0) in comparison to conscious sedation. This corresponded to 15 patients needing to be treated with conscious sedation to prevent one death. Our results were robust in sensitivity analysis with mixed effects regression, and propensity score adjusted regression models.

Conclusions

Using a comprehensive all-payer cohort of acute ischemic stroke patients undergoing mechanical thrombectomy in New York State, we identified an association of general anesthesia with increased case fatality and LOS. These considerations should be taken into account when standardizing acute stroke care.

INTRODUCTION

The evolution of mechanical thrombectomy has revolutionarized the treatment of acute ischemic stroke. Despite initial reservations for first generation devices,^{1, 2} subsequent clinical trials^3-7 have demonstrated that newer clot retrievers are associated with improved mortality and functional outcomes, in appropriately selected patients. The efficacy of this intervention is contingent upon timely revascularization of the occluded vessels.⁸ Significant efforts are currently underway to optimize emergency medical services associated with stroke, streamline transfers, centralize care, and minimize door-to-needle time. However, most of these initiatives focus on either pre-hospital care pathways, or in-hospital protocols aimed at expediting patient transfer to the angiography suite. Although, the method of anesthesia has significant timing implications, the understanding of its impact on the outcomes of mechanical thrombectomy is limited.^9-12 General anesthesia is the preferred method due to the perceptions of improved procedural safety and efficacy.⁹ However, conscious sedation, or no sedation, allow continuous neurologic monitoring, minimal delays, and potentially improved hemodynamic stability.¹⁰

Previous observational studies attempting to answer this question have shown mixed results.^13-21 The main limitation of such investigations is not accounting for unmeasured confounding. Patients included in prior retrospective studies have been selected for either anesthesia technique in advance. This selection reflects the different preferences and background of the treating physicians, as well as specific patient characteristics. Administrative databases lack such granularity, limiting the ability to control for these confounders. There has been no prior study attempting to account for these limitations through different analytic approaches in an adult cohort of all ages.

We used the New York Statewide Planning and Research Cooperative System (SPARCS)²² to study the association of general anesthesia with case-fatality, and length of stay (LOS) for patients undergoing mechanical thrombectomy for acute ischemic stroke. An instrumental variable analysis was used to control for unmeasured confounding and simulate the effect of randomization.

METHODS

New York Statewide Planning and Research Cooperative System (SPARCS)

This study was approved by the Dartmouth Committee for Protection of Human Subjects. All patients undergoing mechanical thrombectomy for acute ischemic stroke who were registered in the SPARCS (New York State Department of Health, Albany, NY)²² database between 2009 and 2013 were included in the analysis. For these years, SPARCS contains patient-level details for every hospital discharge, ambulatory surgery, and emergency department admission in New York State as coded from admission and billing records. More information about SPARCS is available at https://www.health.ny.gov/statistics/sparcs/.

Cohort Definition

In order to establish the cohort of patients, we used International Classification of Disease-9-Clinical Modification (ICD-9-CM) codes to identify patients in the database who underwent mechanical thrombectomy (ICD-9-CM code 39.7) for acute ischemic stroke (ICD-9-CM code 433.x1, 434.x1) between 2009 and 2013.

Outcome variables

The primary outcome variable was case-fatality during the initial hospitalization after mechanical thrombectomy for ischemic stroke. Secondary outcome was LOS during the initial hospitalization.

Exposure variables

The primary exposure variable was the anesthesia technique (general versus conscious sedation). General anesthesia required the patient being intubated, whereas conscious sedation included intravenous sedation (with or without local anesthetic) without the use of a breathing tube. Hospitals voluntarily report data on anesthesia care to SPARCS.

Covariates (Supplemental Table I) used for risk-adjustment were age, gender, race (African-American, Hispanic, Asian, Caucasian, other), insurance (private, Medicare, Medicaid, uninsured, other), and administration of IV-tPA (intravenous tissue plasminogen activator) (ICD-9-CM 99.10, V45.88). The comorbidities used for risk adjustment were diabetes mellitus (DM), smoking, chronic lung disease, hypertension, hypercholesterolemia, peripheral vascular disease (PVD), congestive heart failure (CHF), coronary artery disease (CAD), history of transient ischemic attack (TIA), alcohol abuse, obesity, chronic renal failure (CRF), and coagulopathy. Only variables that were defined as “present on admission” were considered part of the patient’s preadmission comorbidity profile.

Statistical analysis

The association of anesthesia technique with our outcome measures was examined in a multivariable setting. Patients undergoing general anesthesia, or conscious sedation in our cohort were non-randomly selected for either technique based on provider and patient characteristics. In order to account for this unmeasured confounding, and to simulate the effect of randomization, we used an instrumental variable analysis, an econometric technique (Supplementary Methods).²³ The regional rate of general anesthesia (hospital level general anesthesia rate) was used as an instrument for the technique received. This advanced observational technique has been used before by clinical researchers, to answer comparative effectiveness questions for different interventions. The goal is to simulate randomization, especially when the baseline functional characteristics of the patients (including the functional status of stroke patients, NIHSS etc.) are unknown (similar to our application). This is an established technique in prior literature for ischemic stroke, anesthesia technique, and other pathologies when a number of variables are missing from the dataset.^24-28

A good instrument is not associated with the outcome other than through the exposure variable of interest (a requirement known as the exclusion restriction criterion).²⁹ In our case it is unlikely that the regional rates of general anesthesia would be associated with case-fatality in any way other than the choice of treatment. A two stage least squares (2SLS) method was used for the calculation of the coefficients. The value of the F statistic in the first stage of the 2SLS approach was 125, which is consistent with a strong instrument (F statistic>10), based on a practical rule.²³

A probit regression was used for the categorical outcomes (case-fatality),³⁰ and a linear regression for the linear outcomes (LOS). Other models such as general logit were considered. However, we elected to use probit because this is the most widely used and studied model in instrumental variable analysis.^{31, 32} The covariates used for risk adjustment in these models were: age, gender, race, insurance, and all the comorbidities mentioned previously. Since the coefficients produced by the probit function are not interpretable, we used the marginal effects of our independent variables instead. The marginal effects are the partial derivatives of the coefficients, and reflect the change in the probability of the dependent variable, for 1 unit change in the independent variable, at the average value of all other covariates.

In order to demonstrate the robustness of our data in a sensitivity analysis, we used standard techniques to account for measured confounding, while accounting for clustering at the hospital level. For categorical outcomes we used a probit regression model with hospital ID as a random effects variable, while controlling for all the covariates mentioned previously. For continuous outcomes, we performed similar analyses using linear models. LOS demonstrated a positively skewed distribution, and a logarithmic transformation was additionally used in sensitivity analysis. The direction of the observed associations did not change, and therefore we elected to present the untransformed data for ease of interpretation. In an alternative way to control for confounding for categorical outcomes, we used a propensity adjusted (with deciles of propensity score) probit regression model. We calculated the propensity score of general anesthesia with a separate probit regression model, using all the covariates mentioned previously. The results were identical (Supplemental Table II). In post-hoc analyses, we investigated the association of anesthesia technique with the risks of pneumonia, or hemorrhagic transformation using the previously specified instrumental variable analysis.

Regression diagnostics were used for all models. Number needed to treat (NNT) were calculated when appropriate. All results are based on two sided tests, and the level of statistical significance was set at 0.05. This study, based on 1,174 patients, has sufficient power (80%) at a 5% type I error rate to detect differences in case-fatality, as small as 7.8%. Statistical analyses were performed using Stata version 13 (StataCorp, College Station, TX).

RESULTS

Patient characteristics

In the selected study period there were 1,174 patients undergoing mechanical thrombectomy for acute ischemic stroke (mean age was 67.3 years, with 52.4% females) who were registered in SPARCS. 441 (37.6%) underwent general anesthesia, and 733 (62.4%) underwent conscious sedation. The characteristics of the two cohorts at baseline can be seen in Table 1.

Table 1.

Patient characteristics

	Total		General anesthesia		Conscious sedation
	N=1,174		N=441		N=733
	Mean	SD	Mean	SD	Mean	SD
Age	67.3	15.0	66.5	15.2	67.6	14.9
	N	%	N	%	N	%
Female Gender	615	52.4%	219	49.7%	394	53.8%
Race
Caucasian	782	66.6%	309	70.1%	522	71.2%
African-American	149	12.7%	57	12.9%	83	11.3%
Hispanic	143	12.2%	40	9.0%	73	10.0%
Asian	67	5.7%	21	4.8%	38	5.2%
Others	33	2.8%	14	3.2%	17	2.3%
Insurance
Medicare	633	53.9%	233	52.8%	400	55.3%
Private insurance	419	35.7%	156	35.4%	263	34.9%
Medicaid	76	6.5%	31	7.0%	45	6.3%
Uninsured	26	2.2%	15	3.4%	11	1.6%
Others	20	1.7%	⦸	⦸	⦸	⦸
IV-tPA	613	52.2%	219	49.7%	394	53.8%
Comorbidities
Diabetes	289	24.6%	105	23.8%	183	25.0%
Smoking	151	12.9%	40	9.1%	104	14.2%
Obesity	97	8.3%	9	8.8%	59	8.0%
Transient ischemic attack	⦸	⦸	⦸	⦸	⦸	⦸
Coronary artery disease	322	27.4%	125	28.3%	206	28.1%
Congestive Heart Failure	313	26.7%	113	25.6%	199	27.1%
Chronic Lung Disease	151	12.9%	62	14.1%	97	13.2%
Coagulopathy	48	4.1%	23	5.2%	27	3.7%
Chronic Renal Failure	114	9.7%	37	8.4%	78	10.6%
Hypertension	791	67.4%	297	67.3%	497	67.8%
Hypercholesterolemia	475	40.5%	152	34.5%	335	45.7%
Alcohol	38	3.2%	⦸	⦸	⦸	⦸
Peripheral Vascular Disease	104	8.9%	32	7.3%	73	10.0%

SD: Standard Deviation; IV t-PA: intravenous tissue plasminogen activator

⦸ Output suppressed to respect the SPARCS reporting limit of 11 patients

Inpatient case-fatality

Overall, 126 (25.6%) inpatient deaths were recorded after general anesthesia and 147 (18.1%) after conscious sedation. General anesthesia was associated with increased case-fatality in comparison to conscious sedation (Difference, 7.2%; 95% CI, 2.6% to 12.0%) in unadjusted analysis. Likewise, using a probit regression with instrumental variable analysis, we identified that general anesthesia was associated with a 6.4% increased case-fatality (95% CI, 1.9% to 11.0%), in comparison to conscious sedation (Table 2). This persisted in a mixed effects probit regression model (Adjusted difference, 7.5%; 95% CI, 2.9% to 12.1%). This corresponded to 15 patients needed to be treated with conscious sedation to prevent one death.

Table 2.

Models examining the association of general anesthesia with outcomes

	Inpatient Mortality°		Length-of-stay§
	Difference (95% CI)	P-value	Difference (95% CI)	P-value
Unadjusted analysis	7.2% (2.6% to 12.0%)	<0.001	7.9 (5.1 to 10.7)	<0.001
	Adjusted difference (95% CI)	P-value	Adjusted difference (95% CI)	P-value
Instrumental variable analysis*	6.4% (1.9% to 11.0%)	<0.001	8.4 (2.9 to 14.0)	<0.001
Mixed effects regression⌘	7.5% (2.9% to 12.1%)	<0.001	7.3 (4.6 to 10.1)	<0.001

CI: confidence intervals; NA: not applicable

^*Hospital level general anesthesia rate was used as an instrument of anesthesia technique

^⌘Hospital ID was used as a random effects variable

^°All regressions were based on probit models

^§All regressions were based on linear models

Length-of-stay

The average LOS was 19.6 days (SD 35.0) after general anesthesia, and 11.7 days (SD 12.5) after conscious sedation. General anesthesia was associated with increased LOS in comparison to conscious sedation (Difference, 7.9; 95% CI, 5.1 to 10.7) in the unadjusted analysis. Using a linear regression with instrumental variable analysis we demonstrated (Table 2) that general anesthesia was associated with 8.4 days longer LOS in comparison to conscious sedation (95% CI, 2.9 to 14.0). We found similar results in a mixed effects linear regression model (Adjusted difference, 7.3; 95% CI, 4.6 to 10.1).

Post-hoc analyses

In post-hoc analysis, using an instrumental variable analysis, anesthesia technique was not associated with the risk of pneumonia (Adjusted difference, 2.3%; 95% CI, −1.2% to 5.7%), or hemorrhagic transformation (Adjusted difference, 1.5%; 95% CI, −2.4% to 5.6%).

DISCUSSION

Using a comprehensive all-payer cohort of patients in New York State with acute ischemic stroke we identified an association of general anesthesia with increased case-fatality and LOS after mechanical thrombectomy. Our results were robust when considering several advanced observational techniques to account for measured and unmeasured confounders. Mechanical thrombectomy has seen explosive growth in recent years, and is currently performed by multiple specialties, without standardized peri-operative protocols, including anesthesia choice.^{11, 12} This is contributing to an ongoing debate about the relative effectiveness of different anesthesia techniques during mechanical thrombectomy for acute ischemic stroke.^{9, 10}

Several observational studies have compared the outcomes of general anesthesia and conscious sedation for this population. Most of the studies have been retrospective analyses of single institution experiences, demonstrating conflicting results with limited generalization, given their inherent selection bias.^{15-18, 20, 21, 33} The interpretation of larger multi-center studies is equally limited. McDonald et al,¹⁹ utilizing the MarketScan database, demonstrated that conscious sedation is associated with a survival benefit. They employed propensity score matching to balance the covariates among anesthesia groups. However, participation in this commercial database was voluntary, and therefore it is likely that hospitals incentivized to achieve higher quality standards would be overrepresented. This self-selection introduces significant unmeasured confounding, which the authors did not account for. In another study, Abou-Chebl et al¹⁴ analyzed a national registry of a single thrombectomy device and demonstrated superior results for conscious sedation. The generalizability of their results in the population at large is limited only to the device used, and the few hospitals incentivized to participate in this national registry. A post-hoc analysis of the anesthesia method used in a randomized controlled trial, examining the effectiveness of mechanical thrombectomy, demonstrated that the benefit of the trial was only seen with conscious sedation.³⁴ This analysis had the same bias as all retrospective designs. Three randomized trials are currently underway, specifically looking into the comparative effectiveness of different anesthesia techniques.

These prior analyses have some common methodologic limitations. Multicenter studies are vulnerable to clustering at the hospital level. Most previous authors did not evaluate or adjust for this bias. Most importantly, all the analytical methods used accounted, to some degree, for known confounders. Although this may be adequate in some studies, the selection of patients for either anesthesia technique prior to the analysis introduces significant unmeasured confounding. Patients may be selected for general anesthesia because of worse functional status, more severe stroke, or heavier comorbidity burden. Physician or patient preference, as well as provider training and specialty might affect that decision too. Not accounting for this dimension of confounding puts the robustness of their findings into question.

Our study purposefully addresses many of these methodologic limitations. First, we created a cohort of all patients in a major state, giving a true picture of practice in the community. Second, we used advanced observational techniques to control for confounding. Propensity score stratification was used to adjust our analyses for known confounders. The possibility of clustering, which can bias the results of multi-center national studies, was accounted for by using mixed effects methods. Most importantly, an instrumental variable analysis was used to control for unmeasured confounders (mainly the a priori selection of anesthesia technique), and simulate the effects of randomization. The instrumental variable analysis is expected to control for such factors and report results for patients of similar functional status. Results were consistent across techniques, supporting the validity of the observed associations.

Further research into the factors contributing to the superiority of conscious sedation for this pathology is warranted. Theoretical reasons include the neurotoxicity of certain anesthetic agents,³⁵ the lack of continuous neurologic assessment during general anesthesia,¹³ and perioperative hypotension.³⁶ In the latter case, the induction of general anesthesia commonly leads to a reduction in blood pressure, which may be associated with worse outcomes.³⁶

Our study has several limitations. Residual confounding could account for some of the observed associations. However, this is minimized to the extent that we are using a good instrument for anesthesia technique. The F statistic in our analysis suggests a strong instrument. In addition, coding inaccuracies will undoubtedly occur and can affect our estimates. However, several reports have demonstrated that coding for stroke has shown nearly perfect association with medical record review.^{37, 38} Although SPARCS includes all hospitals from the entire New York State, the generalization of this analysis to the US population is uncertain. SPARCS does not provide any clinical information on the functional status of the patients (National Institutes of Health Stroke Scale), which can affect the choice of anesthesia. However, the use of the instrumental variable analysis is attempting to control for unknown confounders such as these, and has been used before in stroke patients of this database.^24-26 By comparing our point estimates for the instrumental variable model and the multivariable model without instrumental variable we can identify a small difference. That indicates that there is likely a small degree of selection bias in our data, before the application of the instrumental variable analysis.

Additionally, we were lacking post-hospitalization, and long-term data on our patients. Quality metrics (i.e. modified Rankin score) are also not available through SPARCS, and therefore we cannot compare the two treatment techniques on these outcomes. The definitive comparison of the two techniques on functional outcomes can only be done in prospective registries. In this direction, the NeuroPoint Alliance has created the first module for a cerebrovascular registry, with results expected in the near future.³⁹ We do not have any information on the availability of neuro-anesthesia in the institutions we are studying. Our results might reflect to some degree the difference between neuro-anesthesia and conscious sedation, although it is likely that some centers might preferentially offer conscious sedation despite the availability of neuro-anesthesia services. Finally, causality cannot be definitively established based on observational data, despite the use of advanced techniques, such as the instrumental variable analysis.

Conclusions

The impact of anesthesia techniques on the outcomes of mechanical thrombectomy for acute ischemic stroke remains an issue of debate. Using a comprehensive all-payer cohort of patients in New York State with acute ischemic stroke, we identified an association of general anesthesia with increased case-fatality and LOS after mechanical thrombectomy. Our results were robust when considering several advanced observational techniques to account for measured and unmeasured confounders. These considerations should be taken into account when standardizing acute stroke care.