Select 2

new england
The
journal of medicine
established in 1812 April 6, 2023 vol. 388 no. 14
Trial of Endovascular Thrombectomy for Large Ischemic Strokes

A. Sarraj, A.E. Hassan, M.G. Abraham, S. Ortega‑Gutierrez, S.E. Kasner, M.S. Hussain, M. Chen, S. Blackburn,
C.W. Sitton, L. Churilov, S. Sundararajan, Y.C. Hu, N.A. Herial, P. Jabbour, D. Gibson, A.N. Wallace, J.F. Arenillas,
J.P. Tsai, R.F. Budzik, W.J. Hicks, O. Kozak, B. Yan, D.J. Cordato, N.W. Manning, M.W. Parsons, R.A. Hanel,
A.N. Aghaebrahim, T.Y. Wu, P. Cardona‑Portela, N. Pérez de la Ossa, J.D. Schaafsma, J. Blasco, N. Sangha,
S. Warach, C.D. Gandhi, T.J. Kleinig, D. Sahlein, L. Elijovich, W. Tekle, E.A. Samaniego, L. Maali, M.A. Abdulrazzak,
M.N. Psychogios, A. Shuaib, D.K. Pujara, F. Shaker, H. Johns, G. Sharma, V. Yogendrakumar, F.C. Ng,
M.H. Rahbar, C. Cai, P. Lavori, S. Hamilton, T. Nguyen, J.T. Fifi, S. Davis, L. Wechsler, V.M. Pereira, M.G. Lansberg,
M.D. Hill, J.C. Grotta, M. Ribo, B.C. Campbell, and G.W. Albers, for the SELECT2 Investigators*
a bs t r ac t
BACKGROUND
Trials of the efficacy and safety of endovascular thrombectomy in patients with The authors’ full names, academic de‑
large ischemic strokes have been carried out in limited populations. grees, and affiliations are listed in the
Appendix. Dr. Sarraj can be contacted at
METHODS amrou.sarraj@uhhospitals.org or at the
Department of Neurology, University
We performed a prospective, randomized, open-label, adaptive, international trial Hospitals Cleveland Medical Center–
involving patients with stroke due to occlusion of the internal carotid artery or the Case Western Reserve University, 11100
first segment of the middle cerebral artery to assess endovascular thrombectomy Euclid Ave., Hanna House No. 504,
Cleveland, OH 44106.
within 24 hours after onset. Patients had a large ischemic-core volume, defined
as an Alberta Stroke Program Early Computed Tomography Score of 3 to 5 (range, *The SELECT2 investigators are listed in
0 to 10, with lower scores indicating larger infarction) or a core volume of at least the Supplementary Appendix, available
at NEJM.org.
50 ml on computed tomography perfusion or diffusion-weighted magnetic reso-
nance imaging. Patients were assigned in a 1:1 ratio to endovascular thrombec- Drs. Hassan and Abraham and Drs. Ribo,
Campbell, and Albers contributed equally
tomy plus medical care or to medical care alone. The primary outcome was the to this article.
modified Rankin scale score at 90 days (range, 0 to 6, with higher scores indicat-
This article was published on February
ing greater disability). Functional independence was a secondary outcome. 10, 2023, and last updated on January 18,
2024, at NEJM.org.
RESULTS
The trial was stopped early for efficacy; 178 patients had been assigned to the N Engl J Med 2023;388:1259-71.
DOI: 10.1056/NEJMoa2214403
thrombectomy and 174 to medical care. The median ischemic-core volumes were Copyright © 2023 Massachusetts Medical Society.
74 ml and 77 ml in the two groups, respectively. The generalized odds ratio for a
shift in the distribution of modified Rankin scale scores toward better outcomes CME
in favor of thrombectomy was 1.51 (95% confidence interval [CI], 1.20 to 1.89; at NEJM.org
P<0.001). A total of 20% of the patients in the thrombectomy group and 7% in the
medical-care group had functional independence (relative risk, 2.97; 95% CI, 1.60
to 5.51). Mortality was similar in the two groups. In the thrombectomy group, arte-
rial access-site complications occurred in 5 patients, dissection in 10, cerebral-vessel
perforation in 7, and transient vasospasm in 11. Symptomatic intracranial hemorrhage
occurred in 1 patient in the thrombectomy group and in 2 in the medical-care group.
CONCLUSIONS
Among patients with large ischemic strokes, endovascular thrombectomy resulted
in better functional outcomes than medical care but was associated with vascular
complications. Cerebral hemorrhages were infrequent in both groups. (Funded by
Stryker Neurovascular; SELECT2 ClinicalTrials.gov number, NCT03876457.)
n engl j med 388;14 nejm.org April 6, 2023 1259
The New England Journal of Medicine
Downloaded from nejm.org on June 7, 2024. For personal use only.
No other uses without permission. Copyright © 2023 Massachusetts Medical Society. All rights reserved.
The n e w e ng l a n d j o u r na l of m e dic i n e
E
ndovascular thrombectomy has ed end-point assessment.17 An academic steering
been shown to be more effective in reduc- committee oversaw the conduct of the trial. The
ing disability than medical therapy alone first author designed the trial and wrote the first
in selected patients with ischemic stroke due to draft of the manuscript, and three of the authors
A Quick Take
a large cerebral vessel occlusion.1-7 However, pa- performed the data analyses. The trial protocol
is available at tients with large strokes on noncontrast com- (available with the full text of this article at
NEJM.org puted tomography (CT) or perfusion imaging NEJM.org) was approved by the local institu-
have been underrepresented in thrombectomy tional review board at each participating site
trials, despite that such strokes account for ap- before enrollment began. All enrolled patients or
proximately one fifth of large-vessel occlusion their legally authorized representatives provided
strokes.8 Consequently, the safety and efficacy of written informed consent. The trial was con-
thrombectomy in patients with a larger ischemic ducted in accordance with the principles of the
burden have not been well established.9-11 These Declaration of Helsinki and the International
patients generally have poor neurologic out- Council for Harmonisation Good Clinical Prac-
comes, including progression of stroke symp- tice guidelines. Adverse-event monitoring and
toms, brain edema, and death. The results of a adjudication was performed by an independent
trial conducted in Japan, post hoc analyses from medical safety monitor. The data and safety
previous trials, and a prospective cohort study monitoring board (which was composed of inde-
have suggested that endovascular thrombectomy pendent vascular neurologists, neurointervention-
may improve functional outcomes in patients alists, and statisticians) oversaw patient safety
with large strokes.8,12-14 and the interim analyses. The authors vouch for
The estimated extent of ischemic change in the accuracy and completeness of the data and
acute stroke differs depending on the imaging for the fidelity of the trial to the protocol.
method that is used to measure the volume of The trial was funded by a grant from Stryker
infarcted tissue.15 Ischemic changes on noncon- Neurovascular to the University Hospitals Cleve-
trast CT appear as areas of hypodensity and are land Medical Center and University of Texas
assessed with the use of the semiquantitative Houston McGovern Medical School. Stryker Neu-
measure of the Alberta Stroke Program Early rovascular did not provide trial equipment, and
Computed Tomography Score (ASPECTS).16 Per- the design of the trial did not mandate the use
fusion imaging identifies quantitative brain vol- of Stryker Neurovascular products. Stryker Neu-
ume with critically reduced blood flow that is rovascular had no role in the design or execution
considered irreversibly damaged, whereas diffu- of the trial, the analysis of the data, or the writ-
sion-weighted magnetic resonance imaging (MRI) ing or reviewing of the manuscript. Confidenti-
detects the volume of brain tissue affected by ality agreements were in place between Univer-
cytotoxic edema. sity Hospitals Cleveland Medical Center (one of
In a randomized, controlled trial involving the participating centers) and Stryker Neurovas-
patients with acute ischemic stroke with a large cular.
ischemic-core volume, we aimed to evaluate
whether endovascular thrombectomy within 24 Patients
hours after stroke onset (defined as the time the The trial was conducted at 31 sites across the
patient was last known to be well) leads to better United States, Canada, Europe, Australia, and
functional outcomes than standard medical care New Zealand. Eligible patients were 18 to 85
alone. We used several imaging methods to de- years of age and had acute ischemic stroke due
termine the size of the core infarction. to occlusion of the internal carotid artery (either
cervical or intracranial) or the M1 segment
(main trunk) of the middle cerebral artery or
Me thods
both. All patients underwent a standardized
Trial Design and Oversight imaging evaluation with noncontrast CT and,
SELECT2 (Randomized Controlled Trial to Opti- depending on site preference, either CT perfusion
mize Patient’s Selection for Endovascular Treat- imaging or diffusion-weighted MRI. Eligible
ment in Acute Ischemic Stroke) was a phase 3, patients had to have a large ischemic core on
international, randomized, open-label clinical noncontrast CT (defined as an ASPECTS value of
trial with adaptive enrichment design and blind- 3 to 5; on a scale from 0 to 10, with lower values
1260 n engl j med 388;14 nejm.org April 6, 2023

Endovascular Thrombectomy for Large Ischemic Strokes
indicating larger infarction) or an estimated imaging characteristics: age, National Institutes

ischemic-core volume of 50 ml or greater on CT of Health Stroke Scale (NIHSS) score at presen-
perfusion imaging (defined as a relative cerebral tation (with scores ranging from 0 to 42 and
blood flow of <30% as determined with RAPID higher scores indicating worse neurologic defi-
automated software [iSchemaView], which was cits), occlusion location, time window (the inter-
installed at each trial site). For sites that used val beween the time that the patient was last
noncontrast CT and diffusion-weighted MRI as known to be well and randomization), ischemic-
the standard for baseline imaging, the apparent core volume estimate, ASPECTS value, presence
diffusion coefficient value was used to deter- or absence of target perfusion–diffusion mis-
mine the ischemic-core volume. A large ischemic match profile (mismatch ratio [the ratio of criti-
core was considered to be a lesion with a volume cally hypoperfused tissue to the ischemic-core
of 50 ml or greater with an apparent diffusion estimate] of ≥1.8 with a mismatch volume [the
coefficient value of less than 620×10−6 mm2 per volumetric difference between critically hypo-
second as determined with the use of RAPID perfused tissue and the ischemic-core estimate]
software. There was no upper limit for ischemic- of ≥15 ml), affected brain hemisphere, and par-
core volume. The ischemic core volume for pa- ticipating center.
tients enrolled with CT perfusion was based on Endovascular thrombectomy was performed
automated measurement of cerebral blood flow with stent retrievers, aspiration devices, or both
volume. It was appreciated during the trial that in of various manufacturers, depending on trial site.
some cases, the region of hypodensity on non- All the patients received medical care according
contrast CT was larger than the CT perfusion to institutional protocols in accordance with the
core volume. Some ischemic core volumes were guidelines of the American Heart Association–
therefore substituted by planimetric assessment American Stroke Association, European Stroke
of the volume of hypodensity on noncontrast CT Organization, and the Stroke Foundation (Aus-
by core laboratory personnel who were unaware tralia and New Zealand), including guidelines
of the treatment assignments. regarding blood-pressure management, critical
Endovascular thrombectomy was expected to care, and in-hospital and outpatient rehabilita-
begin within 24 hours after the onset of stroke. tion.9-11 Decisions to proceed with decompressive
Other eligibility criteria included a prestroke hemicraniectomy in patients with severe brain
score on the modified Rankin scale of 0 or 1 swelling were made in accordance with local
(indicating no disability), ascertained at the time practices, which could include intravenous throm-
of randomization, and no evidence of intracra- bolytic and oral antiplatelet agents. Patients with
nial hemorrhage on neuroimaging. Scores on tandem occlusions or isolated cervical internal
the modified Rankin scale range from 0 to 6, carotid artery occlusions were allowed in the
with higher scores indicating greater disability. trial; details regarding treatment guidance for
Intravenous thrombolytic drugs (alteplase or these occlusions are provided in the protocol.
tenecteplase) were administered to eligible pa-
tients who were first assessed within 4.5 hours Outcomes and Analyses
after the onset of stroke. A screening log was The primary outcome was the ordinal score on
kept at all trial sites for patients who had a the modified Rankin scale at 90 days. Scores of
large-vessel occlusion and met the criteria for a 6 (indicating death) and 5 (indicating that the
large ischemic core on any imaging method but patient is bedridden and constant care is needed)
were not enrolled in the trial; the log included were merged for purposes of the analysis to
documentation of the reason for exclusion. avoid considering a shift from a score of 6 to
5 as a substantial improvement in functional
Trial Interventions status. The initial plan included, at the request
Patients were randomly assigned in a 1:1 ratio to of regulatory agencies, a modified Rankin scale
undergo endovascular thrombectomy and re- score of 0 to 2 as a primary outcome; however,
ceive standard medical care or to receive stan- the power analysis was based on a single pri-
dard medical care alone. Randomization was mary outcome, and a modified Rankin scale
performed with the use of a central, Web-based score of 0 to 2 was changed to be the first sec-
module with an adaptive (minimization) proce- ondary outcome in the previously published trial
dure to balance groups across key clinical and protocol.17

Secondary outcomes were functional inde- means of telephone interviews19 with the patient
pendence (a score on the modified Rankin scale or a surrogate.
of 0 to 2) at 90 days after randomization (with a All patients underwent noncontrast CT, CT
window of ±15 days), independent ambulation angiography or magnetic resonance angiogra-
(a score on the modified Rankin scale of 0 to 3) phy, and CT perfusion imaging or perfusion–
at 90 days after randomization (with a window diffusion MRI, with automated processing with
of ±15 days), procedural complications, success- the use of RAPID software at baseline to assess
ful reperfusion (defined as grade 2b to 3 on the trial eligibility. All imaging (including imaging
modified Treatment in Cerebral Ischemia Scale; performed at baseline, angiography performed
range, 0 to 3, with higher grades indicating in- during thrombectomy, and follow-up neuroim-
creased reperfusion [grade 2b indicates reperfu- aging) was reviewed by the central imaging core
sion of ≥50% of the occluded middle cerebral laboratory at the University of Texas McGovern
artery territory and grade 3 indicates reperfusionMedical School and the University of Melbourne
of 100% of the occluded middle cerebral artery for central adjudication of ASPECTS values, oc-
territory at the end of the thrombectomy proce- clusion location, perfusion measures, angio-
dure]), discharge location, early neurologic im- graphic procedure success, hemorrhages, and
provement (defined as a reduction of ≥8 points follow-up infarct volumes. The imaging methods
in the NIHSS score from baseline to 24 hours are described in the Supplementary Appendix,
after presentation or a score of 0 to 1), and qual-
available at NEJM.org.
ity of life as assessed with the use of domain- The original trial design included two pre-
specific Neuro-QoL measures (mobility, depres- specified interim analyses, which would be per-
sion, social participation, and cognitive aspects)formed when 200 patients and 380 patients en-
transformed to respective T scores (with higher rolled in the trial had completed the 90-day
scores indicating better performance in a given follow-up. The analyses included an efficacy
domain, except for depression, for which higher boundary (z score of >2.604) and a futility bound-
scores indicate worse performance). Safety out- ary (z score of ≤1.897). The data and safety
comes included symptomatic intracranial hem- monitoring board reviewed the results of the
orrhage,18 death, neurologic worsening (an in- first interim analysis of the data for 200 patients
crease of ≥4 points in the NIHSS score within 24 and recommended that the trial continue. In
hours after presentation), and procedural com- light of the publication of the RESCUE-Japan
plications. Additional details regarding outcome LIMIT trial,14 which showed the efficacy of en-
definitions are provided in the protocol. dovascular thrombectomy in patients with a large
Prespecified subgroups were defined accord- stroke who had been selected for enrollment
ing to age, NIHSS score at presentation, time to primarily on the basis of MRI results, the board
randomization, mismatch profile, affected hemi- requested to review the data for the efficacy and
sphere, occlusion site, and geographic location safety outcomes after 300 patients enrolled in
(U.S. vs. non-U.S. sites). The subgroup analyses the trial had completed the 90-day follow-up.
assessed the odds that the trial patients in the The review by the data and safety monitoring
thrombectomy group would have better func- board of these data showed that the prespecified
tional recovery at 90 days than patients assigned efficacy boundary had been crossed in favor of
to the medical-care group, but the trial was not endovascular thrombectomy. Therefore, on Sep-
powered to allow conclusions from these data. tember 9, 2022, the board recommended that
enrollment be stopped. The statistical analysis
Trial Conduct plan was finalized on November 20, 2022, be-
Clinical assessments were performed at baseline fore the outcome data was released by the inde-
and at 24 hours, 5 to 7 days (or at discharge, if pendent data management core on November
discharge occurred earlier), 30 days, and 90 days 21, 2022.
after randomization. Trained, certified assessors
who were unaware of trial-group assignments Statistical Analysis
and imaging results collected 30-day and 90-day On the basis of data from the Optimizing Pa-
outcomes. The score on the modified Rankin tient Selection for Endovascular Treatment in
scale was assessed at in-person visits or by Acute Ischemic Stroke (SELECT) prospective co-
hort study,15 the trial was designed to have a

maximum sample of 560 patients and 90% signed to the thrombectomy group, and 174
power at a two-sided alpha level of 5% to detect were assigned to the medical-care group (Fig. S1
a mean standardized difference of 0.34 for a in the Supplementary Appendix). Approximately
shift across outcomes on the modified Rankin 60% of the thrombectomy procedures were per-
scale between groups. The primary analysis was formed with the patient under general anesthe-
performed on the basis of the intention-to-treat sia. A total of 3 patients were lost to follow-up
principle and used a two-sided Wilcoxon–Mann– before 90 days, and 1 patient withdrew consent
Whitney test for superiority to assess the distri- at the 90-day follow-up. Two patients crossed
bution of scores on the modified Rankin scale over from medical care to endovascular throm-
at the 90-day follow-up. The effect size was de- bectomy; thrombectomy was attempted in all the
termined with the use of the Wilcoxon–Mann– patients who were assigned to the thrombectomy
Whitney measure (probability) of superiority and group. Eleven patients with an ASPECTS value
assumption-free generalized odds ratios with greater than 5 as adjudicated by the core labo-
95% confidence intervals. A Wilcoxon–Mann– ratory and an ischemic-core volume of less than
Whitney probability of superiority value greater 50 ml who were enrolled on the basis of a site-
than 0.5 and a generalized odds ratio greater assessed ASPECTS value of 3 to 5 were excluded
than 1 indicated improvement of outcomes in from the per-protocol analysis.
the thrombectomy group.20,21 The analyses were Baseline demographic, clinical, and imaging
conducted to control the overall two-sided type characteristics were similar in the two trial groups
I error rate at 0.05, accounting for a cumulative (Table 1). The median age was 66.5 years (inter-
alpha of 0.018 spent at the interim analyses. quartile range, 58 to 75), the median NIHSS
Because the analyses used an assumption-free score was 19 (interquartile range, 15 to 23), the
generalized odds ratio, no test of proportionality median interval between the time the patient
was performed. All effects for binary secondary was last known to be well and randomization
and safety outcomes were estimated as relative was 9.31 hours (interquartile range, 5.66 to
risks and are reported with respective 95% con- 15.33), the median ASPECTS value was 4 (inter-
fidence intervals. Because there was no pre- quartile range, 3 to 5). In 48 of 352 patients,
specified plan for the adjustment of the widths planimetric measurement of noncontrast CT
of confidence intervals for the secondary out- hypodensity by the core laboratory was used to
comes or for multiple comparisons in the sub- estimate ischemic core volume rather than the
group analyses, the confidence intervals should prespecified method; 6 patients fulfilled the pre-
not be used for hypothesis testing. Missing data specified reason of suboptimal perfusion and 42
were imputed with the use of multiple imputa- had underestimated core volume based on CT
tion under the missing-at-random assumption perfusion, as judged by the core laboratory. The
with chained equations. Additional details re- mean volumes for each group reported here are
garding statistical analyses and handling of from the prespecified method of measurement
missing data are provided in the statistical and are smaller than if the substituted planimet-
analysis plan (available with the protocol) and in ric method had been used. The median estimated
the Supplementary Appendix. ischemic-core volume was 74 ml (interquartile
range, 50 to 111.5) in the thrombectomy group
and 77 ml (50.3 to 104) in the medical-care group.
R e sult s
A total of 145 patients (41.2%) were women, and
Patient Characteristics 50 patients (14.2%) were Black. A total of 97 of
From September 2019 through September 2022, 339 patients (28.6%) awoke with symptoms of
at the time the trial was stopped, 958 patients stroke, and 211 of 352 patients (59.9%) were
had been screened; 352 of these patients were transferred from outside hospitals or emergency
eligible, provided informed consent, and were en- departments to participating trial centers. Intra-
rolled. The main reasons for exclusion were venous thrombolysis was administered in 37 pa-
multiple or ineligible clot locations, ineligible tients (20.8%) in the thrombectomy group and
age of the patient, a time since the onset of in 30 patients (17.3%) in the medical-care group;
stroke of more than 24 hours, or the presence of alteplase was used in 32 of the 37 patients in the
imaging abnormalities in addition to the cere- thrombectomy group and in 28 of the 30 patients
bral infarction. A total of 178 patients were as- in the medical-care group. The representativeness

Table 1. Baseline Characteristics of the Patients (Intention-to-Treat Population).*
Endovascular Thrombectomy Medical Care

Characteristic (N = 178) (N = 174)
Median age (IQR) — yr 66 (58–75) 67 (58–75)
Female sex — no. (%) 71 (39.9) 74 (42.5)
Race or ethnic group — no. (%)†
American Indian or Alaska Native 0 1 (0.6)
Asian 5 (2.8) 3 (1.7)
Black 26 (14.6) 24 (13.8)
White 132 (74.2) 130 (74.7)
Native Hawaiian or Pacific Islander 2 (1.1) 0
Other or unknown 13 (7.3) 16 (9.2)
Previous ischemic stroke — no. (%) 19 (10.7) 13 (7.5)
Right hemisphere affected — no. (%) 98 (55.1) 98 (56.3)
Occlusion location — no. (%)‡
Internal carotid artery 80 (44.9) 66 (37.9)
M1 segment 91 (51.1) 100 (57.5)
M2 segment 7 (3.9) 8 (4.6)
Tandem occlusions — no. of patients (%) 56 (31.5) 44 (25.3)
Transfer to center with endovascular thrombectomy capabilities — no. (%) 106 (59.6) 105 (60.3)
Intravenous thrombolysis — no./total no. (%) 37/178 (20.8) 30/173 (17.3)
Median NIHSS score at hospital arrival (IQR)§ 19 (15–23) 19 (15–22)
General anesthesia performed — no. (%) 104/177 (58.8) —
Median interval between time that patient was last known to be well and 9.07 (5.27–15.33) 9.79 (5.82–15.32)
randomization (IQR) — hr
Median interval between hospital arrival and imaging (IQR) — min
CT 16 (9–27) 15 (7–24)
CT perfusion or MRI 26 (17–41) 25 (13–36)
Median interval between hospital arrival and arterial puncture (IQR) — min 109 (76–138) —
Median interval between arterial puncture and reperfusion or end of pro‑ 38 (25–61) —
cedure (IQR) — min
Median ASPECTS value on baseline CT imaging (IQR)¶ 4 (3–5) 4 (4–5)
Imaging method used to estimate ischemic-core volume — no./total no. (%)
CT perfusion 174/177 (98.3) 169/174 (97.1)
Diffusion-weighted MRI 3/177 (1.7) 5/174 (2.9)
Median estimated ischemic-core volume (IQR) — ml‖
Overall 74 (50–111.5) 77 (50.3–105)
CT perfusion 81.5 (59–119) 79 (62–111)
Diffusion-weighted MRI 82 (56–89) 86 (84–104)
Median volume of critically hypoperfused lesion (IQR) — ml** 171 (127–226) 169 (127–216)
Median volume of tissue with Tmax of >10 sec (IQR) — ml 107 (70.5–152.5) 111 (67–147)
* Percentages may not total 100 because of rounding. CT denotes computed tomography, IQR interquartile range, and Tmax time to maxi‑
mum contrast arrival.
† Race and ethnic group were reported by the patients.
‡ The M1 segment was defined as the horizontal segment of the middle cerebral artery, terminating at the genu adjacent to the limen in
sulae. The M2 segment was defined as the segment of the middle cerebral artery distal to the genu adjacent to the limen insulae.
§ Scores on the National Institutes of Health Stroke Scale (NIHSS) range from 0 to 42, with higher scores indicating worse neurologic
deficits.
¶ Alberta Stroke Program Early Computed Tomography Scores (ASPECTS) range from 0 to 10, with lower values indicating larger infarction.
ASPECTS values were adjudicated by the core laboratory.
‖ The ischemic-core volume (irreversibly injured brain tissue) was defined as the volume of tissue with relative cerebral blood flow of less
than 30% of that of the contralateral hemisphere or an apparent diffusion coefficient of less than 620×10−6 mm2 per second.
** The critically hypoperfused lesion volume was defined as the tissue volume with a Tmax of more than 6 seconds on CT perfusion imaging.

Score on Modified Rankin Scale

1.1 0 1 2 3 4 5 6
(N=2)
5.1
(N=9)
Endovascular Thrombectomy 14.0 17.4 15.2 8.4 38.2

(N=178) (N=25) (N=31) (N=27) (N=15) (N=68)
Medical Care 11.5 20.7 18.4 40.8

(N=174) (N=20) (N=36) (N=32) (N=71)
5.2
(N=9)
1.7
0 (N=3)
(N=0)
0 20 40 60 80 100
Percentage of Patients
Figure 1. Distribution of Scores on the Modified Rankin Scale at 90 Days (Intention-to-Treat Population).
A modified Rankin scale score of 0 indicates no symptoms; a score of 1, no clinically significant disability (patients
are able to perform usual work, leisure, and school activities); a score of 2, slight disability (patients are able to look
after their own affairs without assistance but are unable to carry out all previous activities); a score of 3, moderate
disability (patients require some help but are able to walk unassisted); a score of 4, moderately severe disability; a
score of 5, severe disability (patients are bedridden and require constant care); and a score of 6, death. Percentages
may not total 100 because of rounding.
of the trial population to patients with acute 1.60 to 5.51). Independent ambulation (a score
ischemic stroke in the countries in which the on the modified Rankin scale of 0 to 3) at 90
trial was conducted is provided in Table S13. days occurred in 37.9% of the patients in the
thrombectomy group and in 18.7% of the patients
Primary and Secondary Clinical Outcomes in the medical-care group (relative risk, 2.06;
At 90 days, the median score on the modified 95% CI, 1.43 to 2.96). Results for the other sec-
Rankin scale was 4 (interquartile range, 3 to 6) ondary outcomes are shown in Table 2 and were
in the thrombectomy group and 5 (interquartile generally supportive of those of the primary
range, 4 to 6) in the medical-care group (Fig. 1). analysis. There was no prespecified plan for ad-
The Wilcoxon–Mann–Whitney probability of su- justment of the widths of confidence intervals
periority was 0.60 (95% confidence interval [CI], for multiplicity in comparing secondary outcomes
0.55 to 0.65), and the generalized odds ratio fa- between trial groups, and no definite conclu-
voring endovascular thrombectomy was 1.51 sions can be drawn from these data. Successful
(95% CI, 1.20 to 1.89; P<0.001). In the prespeci- reperfusion of the target vessel occurred in 142
fied tipping-point analysis, in which missing patients (79.8%) in the thrombectomy group.
scores on the modified Rankin scale were im-
puted as 6 in the thrombectomy group and as Safety
0 in the medical-care group, the Wilcoxon–Mann– Symptomatic intracranial hemorrhage occurred
Whitney probability of superiority was 0.59 (95% in 1 patient (0.6%) in the thrombectomy group
CI, 0.53 to 0.64), and the generalized odds ratio and in 2 patients (1.1%) in the medical-care
was 1.42 (95% CI, 1.13 to 1.78) (Table S11). group (relative risk, 0.49; 95% CI, 0.04 to 5.36)
Functional independence at 90 days (a score (Table 3). Parenchymal hematoma was observed
on the modified Rankin scale of 0 to 2) was in 5 patients (2.8%) in the thrombectomy group
observed in 20.3% of the patients in the throm- and in 3 patients (1.7%) in the medical-care
bectomy group and in 7.0% of the patients in the group (relative risk, 1.63; 95% CI, 0.39 to 6.73).
medical-care group (relative risk, 2.97; 95% CI, At 90 days, 68 of 177 patients (38.4%) in the

Table 2. Clinical Outcomes (Intention-to-Treat Population).*
Endovascular
Thrombectomy Medical Care Effect Size
Variable (N = 178) (N = 174) (95% CI)
Primary outcome
Median score on modified Rankin scale at 90 days (IQR)† 4 (3–6) 5 (4–6) 1.51 (1.20 to 1.89)‡
Secondary clinical outcomes
Functional independence at 90 days — no./total no. (%)§ 36/177 (20.3) 12/171 (7.0) 2.97 (1.60 to 5.51)¶
Independent ambulation at 90 days — no./total no. (%)‖ 67/177 (37.9) 32/171 (18.7) 2.06 (1.43 to 2.96)¶
Successful reperfusion — no. (%)** 142 (79.8) —
Discharge location — no. (%)
Home 19 (10.7) 10 (5.7)
Acute care facility 11 (6.2) 16 (9.2)
Inpatient rehabilitation facility 72 (40.4) 65 (37.4)
Skilled nursing facility 23 (12.9) 20 (11.5)
Hospice or home hospice 11 (6.2) 19 (10.9)
In-hospital death 42 (23.6) 44 (25.3)
Early neurologic improvement — no./total no. (%)†† 20/174 (11.5) 13/172 (7.6) 1.47 (0.76 to 2.87)¶
Median quality-of-life scores (IQR)‡‡
Mobility domain 35.2 (23.9 to 43.9) 25.1 (16.5 to 33.0) 10.10 (5.02 to 15.18)§§
Depression domain 47.9 (43.1 to 54.3) 53.6 (46.8 to 57.4) −5.70 (−8.83 to −2.57)‖
Social domain 37.1 (32.7 to 42.0) 33.5 (27.7 to 37.8) 3.60 (1.11 to 6.09)‖
Cognitive domain 41.9 (35.0 to 49.6) 37.9 (30.9 to 42.9) 4.00 (0.51 to 7.49)‖
* The widths of the confidence intervals for the secondary outcomes were not adjusted for multiple comparisons, and the reported confi‑
dence intervals should not be used for hypothesis testing.
† Scores on the modified Rankin scale range from 0 to 6, with higher scores indicating greater disability.
‡ The value is the generalized odds ratio (P<0.001). The Wilcoxon–Mann–Whitney probability of superiority was 0.60 (95% CI, 0.55 to 0.65).
§ Functional independence was defined as a score on the modified Rankin scale of 0 to 2.
¶ The value is the relative risk estimate and 95% confidence interval.
‖ Independent ambulation was defined as a score on the modified Rankin scale of 0 to 3.
** Successful reperfusion was defined as grade 2b to 3 on the modified Thrombectomy in the Cerebral Ischemia system; range, 0 to 3, with
higher grades indicating increased reperfusion (grade 2b indicates reperfusion of ≥50% of the occluded middle cerebral artery territory,
and grade 3 indicates reperfusion of 100% of the occluded middle cerebral artery territory at the end of the thrombectomy procedure).
†† Early neurologic improvement was defined as a reduction of at least 8 points in the NIHSS score from the time of presentation to a center
with endovascular-thrombectomy capabilities or a NIHSS score of 0 to 1 at the 24-hour follow-up.
‡‡ Quality-of-life scores were evaluated with the use of domain-specific Neuro-QoL assessments. The values reflect T scores, with higher
scores indicating better performance in a given domain, except for depression, for which higher scores indicate worse performance.
§§ The value is the coefficient and 95% confidence interval for the median value, calculated with the use of quantile regression.
thrombectomy group and 71 of 171 patients Patients who had neurologic worsening had
(41.5%) in the medical-care group had died larger ischemic-core lesions at baseline (median
(relative risk, 0.91; 95% CI, 0.71 to 1.18). volume, 107 ml [interquartile range, 67 to 158]
Early neurologic worsening occurred in 44 among patients with neurologic worsening vs.
patients (24.7%) in the thrombectomy group and 77 ml [interquartile range, 60 to 100] among
in 27 patients (15.5%) in the medical-care group patients without neurologic worsening).
(relative risk, 1.59; 95% CI, 1.03 to 2.45). In a Procedural complications occurred in 33 pa-
post hoc analysis, from which no conclusions tients (18.5%) in the thrombectomy group. Com-
can be drawn, early neurologic worsening was plications at the arterial access site included
associated with worse functional outcomes at occlusion (in 3 patients [1.7%]), hematoma (in
90 days (Wilcoxon–Mann–Whitney probability 1 patient [0.6%]), and infection (in 1 patient
of superiority, 0.37 [95% CI, 0.31 to 0.43]; gener- [0.6%]). In addition, 10 patients (5.6%) had vas-
alized odds ratio, 0.58 [95% CI, 0.45 to 0.74]). cular dissections, 7 (3.9%) had arterial perfora-

Table 3. Safety Outcomes and Procedural Complications (Intention-to-Treat Population).*
Endovascular
Thrombectomy Medical Care Relative Risk
Outcome (N = 178) (N = 174) (95% CI)
Symptomatic intracranial hemorrhage within 1 (0.6) 2 (1.1) 0.49 (0.04 to 5.36)
24 hr — no. (%)†
Early neurologic worsening — no. (%)‡ 44 (24.7) 27 (15.5) 1.59 (1.03 to 2.45)
Death from any cause within 90 days — no./total 68/177 (38.4) 71/171 (41.5) 0.91 (0.71 to 1.18)
no. (%)
Arterial access-site complications — no. (%)
Occlusion 3 (1.7) —
Hematoma 1 (0.6) —
Infection 1 (0.6) —
Vascular injury — no. (%)
Dissection 10 (5.6) —
Perforation 7 (3.9) —
Vasospasm 11 (6.2) —
Other 2 (1.1) —
* The widths of the confidence intervals for the safety outcomes were not adjusted for multiple comparisons, and the
reported confidence intervals should not be used for hypothesis testing.
† Symptomatic intracranial hemorrhage was defined as parenchymal hemorrhage type 2 or remote parenchymal hemor‑
rhage associated with an increase of 4 or more points in the NIHSS score at the 24-hour follow-up (according to Safe
Implementation of Thrombolysis in Stroke–Monitoring Study criteria18).
‡ Early neurologic worsening was defined as an increase of 4 or more points in the NIHSS score from the time of presen‑
tation to a center with endovascular-thrombectomy capabilities to the 24-hour follow-up.
tion, and 11 (6.2%) had intraprocedural vaso- sis, as were the results among patients who had
spasm. Two patients had both arterial access-site been enrolled on the basis of having large isch-
and intracranial vascular complications. The re- emic-core volumes (Fig. 2). The results in the sub-
sults of a post hoc analysis of clinical and safety group of patients with both an ASPECTS value of 5
outcomes among patients who had or did not have or less and an estimated ischemic-core volume of
procedural complications are reported in Table S4. 70 ml or greater were also similar to the overall
results (Wilcoxon–Mann–Whitney probability of
Subgroup and Per-Protocol Analyses superiority, 0.61 [95% CI, 0.54 to 0.68]; generalized
The results of subgroup analyses are shown in odds ratio, 1.58 [95% CI, 1.19 to 2.09]). This direc-
Figure 2 and were generally supportive of the tion of effect persisted in patients with an isch-
primary analysis. The results of the per-protocol emic-core volume greater than 100 ml and greater
analysis (which included 336 patients) and the than 150 ml, although functional independence
as-treated analysis (which included 352 patients) (a score on the modified Rankin scale of 0 to 2)
were consistent with those of the primary inten- was less frequent in these patients than in those
tion-to-treat analysis (Tables S6, S7, S9, and S10). with smaller ischemic cores.
A sensitivity analysis that used site-rated
ASPECTS categories showed similar treatment Discussion
effects to those based on core laboratory ratings
(Table S12). The results of the SELECT2 trial, which involved
patients from broad geographic regions, showed
Efficacy and Safety of Thrombectomy that endovascular thrombectomy plus medical
According to Different Imaging Methods care resulted in better clinical outcomes than
The results among the patients who had been medical care alone in patients with a large ische
enrolled on the basis of having a low ASPECTS mic core who presented within 24 hours after
value were similar to those of the primary analy- the time they were last known to be well. The

Subgroup No. of Patients Generalized Odds Ratio (95% CI)
Age
<70 yr 203 1.66 (1.22–2.27)
≥70 yr 149 1.36 (1.01–1.84)
NIHSS score at presentation
<20 197 1.53 (1.12–2.10)
≥20 155 1.52 (1.12–2.07)
Occlusion location
Internal carotid artery 146 1.31 (0.93–1.85)
Middle cerebral artery 206 1.68 (1.24–2.27)
Interval between time that patient was last
known to be well and randomization
<12 hr 211 1.48 (1.12–1.96)
≥12 hr 141 1.58 (1.09–2.28)
<6 hr 100 1.63 (1.09–2.46)
≥6 hr 252 1.49 (1.14–1.94)
Ischemic-core volume
<70 ml 124 1.39 (0.93–2.07)
≥70 ml 228 1.62 (1.25–2.12)
<100 ml 235 1.57 (1.18–2.09)
≥100 ml 117 1.55 (1.11–2.16)
<150 ml 308 1.51 (1.19–1.93)
≥150 ml 44 1.73 (1.02–2.94)
ASPECTS value
0–2 20 1.40 (0.91–2.16)
3–5 290 1.61 (1.25–2.07)
6–10 42 1.24 (0.65–2.37)
Mismatch ratio ≥1.8 and mismatch volume ≥15 ml
Yes 194 1.36 (1.00–1.84)
No 124 1.83 (1.30–2.58)
Mismatch ratio ≥1.2 and mismatch volume ≥10 ml
Yes 298 1.44 (1.13–1.83)
No 31 2.54 (1.26–5.14)
Subgroup A 328 1.54 (1.22–1.94)
Subgroup B 202 1.58 (1.19–2.09)
Affected hemisphere
Left 156 1.42 (1.02–1.95)
Right 196 1.60 (1.18–2.19)
Geographic region
United States 280 1.63 (1.26–2.11)
Other 72 1.13 (0.72–1.75)
0.5 1.0 1.5 2.0 2.5 5.0
Medical Care Endovascular Thrombectomy

Better Better
results for the secondary outcomes were gener- ing to both noncontrast CT (an ASPECTS value
ally in the same direction as those of the pri- of ≤5) and ischemic-core volume (≥50 ml). The
mary analysis, with the possible exception of results across the prespecified subgroups of pa-
early neurologic improvement. The incidence of tients defined according to imaging criteria were
symptomatic intracranial hemorrhage was low similar to those of the primary analysis.
in both trial groups, but 18% of the patients in A Japanese trial14 showed the efficacy of
the thrombectomy group had complications as- thrombectomy in patients with an ASPECTS
sociated with the procedure. value of 3 to 5, primarily assessed on MRI, when
A total of 85% of the enrolled patients had an performed within the first 6 hours after onset or
ASPECTS value of 5 or less, whereas 75.3% had within 6 to 24 hours after onset when initial
an ischemic-core volume of 50 ml or greater; fluid-attenuated inversion recovery imaging
66.8% of the patients had a large stroke accord- showed no signal change. Other ongoing or re-

Figure 2 (facing page). Analyses According to Prespecified increase of 4 or more points on the NIHSS, was
Subgroups. numerically more frequent in the thrombectomy
Shown is the subgroup analysis of the Wilcoxon–Mann– group than in the medical-care group and was
Whitney generalized odds ratio, indicating the odds associated with worse functional outcomes at 90
that the trial patients assigned to undergo endovascu‑ days in a post hoc analysis. Symptomatic hemor-
lar thrombectomy and receive standard medical care
rhage occurred in one patient who had had early
would have better functional recovery at 90 days (as
reflected by a shift in the distribution of scores on the neurologic worsening. A potential cause of dete-
modified Rankin scale toward more favorable outcomes) rioration in some of these patients was brain
than patients assigned to receive standard medical care edema associated with reperfusion. However,
only. The widths of the confidence intervals were not overall, endovascular thrombectomy was associ-
adjusted for multiple comparisons, and the reported
ated with better outcomes than medical care
confidence intervals should not be used for hypothesis
testing. Scores on the National Institutes of Health alone. Further research may identify treatments
Stroke Scale (NIHSS) range from 0 to 42, with higher to reduce edema and infarct progression. Previ-
scores indicating worse neurologic deficits. Alberta ous studies have reported rates of symptomatic
Stroke Program Early Computed Tomography Scores intracranial hemorrhage in patients with large
(ASPECTS) range from 0 to 10, with lower values indi‑
ischemic-core lesions that are higher than those
cating larger infarction. The mismatch ratio is the ratio
of critically hypoperfused tissue to the ischemic-core in our trial. Therefore, the low percentage of
estimate, and the mismatch volume is the volumetric patients with symptomatic intracranial hemor-
difference between critically hypoperfused tissue and rhage observed in both trial groups was unex-
the ischemic-core estimate. Subgroup A refers to the pected. In the thrombectomy group, 3% of pa-
trial population with the exclusion of patients for whom
tients had arterial access-site complications, 4%
the interval between the time that the patient was last
known to be well and randomization was less than had vessel perforation, and 6% had dissection.
6 hours and who had an ASPECTS value of 6 to 10 and In post hoc analyses, clinical efficacy and safety
with the exclusion of patients for whom the interval be‑ outcomes were similar among patients in the
tween the time that the patient was last known to be thrombectomy group who had or did not have
well and randomization was 6 to 16 hours and who had
procedural complications, but no definite con-
an ASPECTS value of 6 to 10 and an ischemic-core vol‑
ume of less than 70 ml. Subgroup B refers to the sub‑ clusions can be drawn from the results of these
group of patients who had an ASPECTS value of less analyses.
than 6 and an ischemic-core volume of 70 ml or great‑ Approximately 20% of large-vessel occlusion
er. The sizes of the boxes in the plot correspond to the strokes are shown to have a large core. Many
number of patients in each subgroup. The arrow indicates
patients with a large ischemic-core volume have
that the 95% confidence interval was beyond the scale.
not been considered candidates for endovascular
thrombectomy and may not be transferred to
cently completed trials (e.g., ClinicalTrials.gov centers with endovascular thrombectomy capa-
numbers, NCT03805308, NCT03094715, and bilities for intervention. Our results, which re-
NCT03811769) involving patients with extensive flect outcomes in geographic populations that
ischemic injury have based eligibility on noncon- are different from those in previous trials, may
trast CT or MRI ASPECTS criteria, with limited support extending the indication for thrombec-
enrollment of patients on the basis of perfusion tomy to patients with a large ischemic core on
imaging criteria. The SELECT2 trial design in- baseline imaging.
cluded patients on the basis of a low ASPECTS Limitations of the trial include its early termi-
value (a value of <6) or a large ischemic-core nation, which could have caused the treatment
volume on CT perfusion or diffusion-weighted effect to be overestimated. The sample size was
MRI (an ischemic-core volume of ≥50 ml) and therefore smaller than anticipated and under-
was informed by the results of a previous phase powered for subgroup analyses. As with all
2 prospective cohort study (SELECT).15 A trial of thrombectomy trials, treatment was open label.
endovascular treatment for large strokes con- However, outcome assessment was performed by
ducted in China that enrolled patients with an assessors who were unaware of trial-group as-
ASPECTS value of 3 to 5 or a more limited signments. Some patients who were enrolled on
ischemic-core volume range of 70 to 100 ml has the basis of low ASPECTS values had lower
shown results that are generally similar to those ischemic-core volumes than intended for enroll-
of our trial.22 ment. The benefit of endovascular thrombecto-
Early neurologic worsening, defined as an my as compared with standard medical care,

however, persisted numerically in prespecified stroke due to a proximal large-vessel occlusion

analyses after these patients were excluded. Only and with a large ischemic core, endovascular
approximately 20% of the patients in the trial thrombectomy in addition to standard medical
received intravenous thrombolytic agents before care resulted in better functional outcomes than
randomization, partly because of the inclusion medical care alone. Thrombectomy was associ-
of patients who presented more than 4.5 hours ated with procedural vascular complications.
after onset and potentially because of physi-
Supported by an investigator-initiated grant from Stryker
cians’ concerns regarding the use of thrombo- Neurovascular to University Hospitals Cleveland Medical Center
lytic agents in patients with extensive ischemic and the University of Texas McGovern Medical School.
changes. Disclosure forms provided by the authors are available with
the full text of this article at NEJM.org.
Among patients in North America, Europe, A data sharing statement provided by the authors is available
Australia, and New Zealand with acute ischemic with the full text of this article at NEJM.org.
Appendix
The authors’ full names and academic degrees are as follows: Amrou Sarraj, M.D., Ameer E. Hassan, D.O., Michael G. Abraham, M.D.,
Santiago Ortega‑Gutierrez, M.D., Scott E. Kasner, M.D., M. Shazam Hussain, M.D., Michael Chen, M.D., Spiros Blackburn, M.D.,
Clark W. Sitton, M.D., Leonid Churilov, Ph.D., Sophia Sundararajan, M.D., Yin C. Hu, M.D., Nabeel A. Herial, M.D., Pascal Jabbour,
M.D., Daniel Gibson, M.D., Adam N. Wallace, M.D., Juan F. Arenillas, M.D., Ph.D., Jenny P. Tsai, M.D., Ronald F. Budzik, M.D., Wil-
liam J. Hicks, M.D., Osman Kozak, M.D., Bernard Yan, M.B., B.S., Dennis J. Cordato, Ph.D., Nathan W. Manning, M.B., B.S., Mark W.
Parsons, Ph.D., Ricardo A. Hanel, M.D., Amin N. Aghaebrahim, M.D., Teddy Y. Wu, Ph.D., Pere Cardona‑Portela, M.D., Natalia
Pérez de la Ossa, M.D., Ph.D., Joanna D. Schaafsma, M.D., Jordi Blasco, M.D., Ph.D., Navdeep Sangha, M.D., Steven Warach, M.D.,
Chirag D. Gandhi, M.D., Timothy J. Kleinig, Ph.D., Daniel Sahlein, M.D., Lucas Elijovich, M.D., Wondwossen Tekle, M.D., Edgar A.
Samaniego, M.D., Laith Maali, M.D., M. Ammar Abdulrazzak, M.D., Marios N. Psychogios, M.D., Ashfaq Shuaib, M.D., Deep K. Pu-
jara, M.B., B.S., Faris Shaker, M.B., Ch.B., Hannah Johns, Ph.D., Gagan Sharma, M.C.A., Vignan Yogendrakumar, M.D., Felix C. Ng,
Ph.D., Mohammad H. Rahbar, Ph.D., Chunyan Cai, Ph.D., Philip Lavori, Ph.D., Scott Hamilton, Ph.D., Thanh Nguyen, M.D., Johan-
na T. Fifi, M.D., Stephen Davis, M.D., Lawrence Wechsler, M.D., Vitor M. Pereira, M.D., Maarten G. Lansberg, M.D., Michael D. Hill,
M.D., James C. Grotta, M.D., Marc Ribo, M.D., Bruce C. Campbell, Ph.D., and Gregory W. Albers, M.D.
The authors’ affiliations are as follows: the Departments of Neurology (A. Sarraj, S.S., D.K.P.) and Neurosurgery (Y.C.H.), University
Hospitals Cleveland Medical Center–Case Western Reserve University, and the Cerebrovascular Center, Cleveland Clinic (M.S.H., J.P.T.,
M.A.A.), Cleveland, and the Departments of Neurointerventional Radiology (R.F.B.) and Neurology (W.J.H.), OhioHealth–Riverside
Methodist Hospital, Columbus — all in Ohio; the Neuroscience Institute, Valley Baptist Medical Center, Harlingen (A.E.H., W.T.), the
Departments of Neurosurgery (S.B., F.S.), Diagnostic and Interventional Imaging (C.W.S.), and Internal Medicine (M.H.R., C.C.), Mc-
Govern Medical School at UTHealth, and the Mobile Stroke Unit, Memorial Hermann Hospital (J.C.G.), Houston, and the Department
of Neurology, Dell Medical School at the University of Texas at Austin, Austin (S.W.) — all in Texas; the Department of Neurology,
University of Kansas Medical Center, Kansas City (M.G.A., L.M.); the Departments of Neurosurgery and Radiology (S.O.-G.) and Neu-
rology (E.A.S.), University of Iowa Hospitals and Clinics, Iowa City; the Division of Vascular Neurology, University of Pennsylvania
(S.E.K.), the Department of Neurosurgery, Thomas Jefferson University Hospital (N.A.H., P.J.), and the Department of Neurology,
Hospital of the University of Pennsylvania (L.W.), Philadelphia, and Neurovascular Associates of Abington, Jefferson Health, Abington
(O.K.) — all in Pennsylvania; the Department of Neurosurgery, Rush University Medical Center, Chicago (M.C.); Melbourne Medical
School, University of Melbourne (L.C., H.J.), the Melbourne Brain Centre, Royal Melbourne Hospital, and the Department of Medicine,
University of Melbourne (B.Y., G.S., V.Y., F.C.N., S.D., B.C.C.), and the Florey Institute of Neuroscience and Mental Health (L.C.,
B.C.C.), Parkville, VIC, the Departments of Neurology (D.J.C., M.W.P.) and Neurosurgery (N.W.M.), Liverpool Hospital, and the Depart-
ment of Neurology, University of New South Wales (M.W.P.), Liverpool, and the Neurology Service, Royal Adelaide Hospital, Adelaide,
SA (T.J.K.) — all in Australia; the Department of Neurosurgery, Ascension Columbia St. Mary’s Hospital, Milwaukee (D.G., A.N.W.);
the Department of Internal Medicine, Hospital Clínico Universitario de Valladolid, Valladolid (J.F.A.), the Department of Neurology,
Bellvitge University Hospital (P.C.-P.), the Department of Interventional Radiology, Hospital Clínic de Barcelona (J.B.), and the Depart-
ment of Neurology, Hospital Vall d’Hebrón (M.R.), Barcelona, and the Department of Neurology, Hospital Universitari Germans Trias
i Pujol, Badalona (N.P.O.) — all in Spain; Neurosurgery, Corewell Health, Grand Rapids, MI (J.P.T.); Lyerly Neurosurgery, Baptist
Medical Center Jacksonville, Jacksonville, FL (R.A.H., A.N.A.); the Department of Neurology, Christchurch Hospital, Christchurch, New
Zealand (T.Y.W.); the Divisions of Internal Medicine and Neurology, Toronto Western Hospital (J.D.S.), and the Division of Neurology,
St. Michael’s Hospital (V.M.P.), Toronto, the Divisions of Internal Medicine and Neurology, University of Alberta, Edmonton (A. Sh-
uaib), and the Department of Clinical Neurosciences, University of Calgary, Calgary, AB (M.D.H.) — all in Canada; Neurological Ser-
vices, Kaiser Permanente Southern California, Los Angeles (N.S.), the Departments of Biomedical Data Science (P.L.) and Neurology
(M.G.L., G.W.A.), Stanford University, Stanford, and MAPS Public Benefit Corporation, San Jose (S.H.) — all in California; the Depart-
ment of Neurosurgery, Westchester Medical Center and New York Medical College, Valhalla (C.D.G.), and the Department of Neurol-
ogy, Icahn School of Medicine at Mount Sinai, New York (J.T.F.) — both in New York; Interventional Neuroradiology, Goodman
Campbell Brain and Spine, Carmel, IN (D.S.); Neurology, Semmes Murphey Clinic, Memphis, TN (L.E.); Neuroradiology, University
Hospital Basel, Basel, Switzerland (M.N.P.); and the Neurology Department, Boston Medical Center, Boston (T.N.).
References
1. Berkhemer OA, Fransen PSS, Beumer 2. Saver JL, Goyal M, Bonafe A, et al. 3. Jovin TG, Chamorro A, Cobo E, et al.
D, et al. A randomized trial of intraarte- Stent-retriever thrombectomy after intra- Thrombectomy within 8 hours after symp-
rial treatment for acute ischemic stroke. venous t-PA vs. t-PA alone in stroke. N Engl tom onset in ischemic stroke. N Engl J Med
N Engl J Med 2015;372:11-20. J Med 2015;372:2285-95. 2015;372:2296-306.

4. Goyal M, Demchuk AM, Menon BK, pean Society for Minimally Invasive Neu- quantitative computed tomography score
et al. Randomized assessment of rapid rological Therapy (ESMINT) guidelines in predicting outcome of hyperacute stroke
endovascular treatment of ischemic stroke. on mechanical thrombectomy in acute before thrombolytic therapy. Lancet 2000;
N Engl J Med 2015;372:1019-30. ischemic stroke. J Neurointerv Surg 2019 355:1670-4.
5. Campbell BCV, Mitchell PJ, Kleinig TJ, February 26 (Epub ahead of print). 17. Sarraj A, Hassan AE, Abraham M,
et al. Endovascular therapy for ischemic 11. Living stroke guidelines. Melbourne, et al. A randomized controlled trial to op-
stroke with perfusion-imaging selection. VIC, Australia:Stroke Foundation, 2022 timize patient’s selection for endovascu-
N Engl J Med 2015;372:1009-18. (https://strokefoundation.org.au/what-we lar treatment in acute ischemic stroke
6. Nogueira RG, Jadhav AP, Haussen DC, -do/for-health-professionals/living-stroke (SELECT2): study protocol. Int J Stroke
et al. Thrombectomy 6 to 24 hours after -g uidelines). 2022;17:689-93.
stroke with a mismatch between deficit 12. Román LS, Menon BK, Blasco J, et al. 18. Wahlgren N, Ahmed N, Dávalos A,
and infarct. N Engl J Med 2018;378:11-21. Imaging features and safety and efficacy et al. Thrombolysis with alteplase for acute
7. Albers GW, Marks MP, Kemp S, et al. of endovascular stroke treatment: a meta- ischaemic stroke in the Safe Implementa-
Thrombectomy for stroke at 6 to 16 hours analysis of individual patient-level data. tion of Thrombolysis in Stroke-Monitor-
with selection by perfusion imaging. Lancet Neurol 2018;17:895-904. ing Study (SITS-MOST): an observational
N Engl J Med 2018;378:708-18. 13. Campbell BCV, Majoie CBLM, Albers study. Lancet 2007;369:275-82.
8. Sarraj A, Hassan AE, Savitz S, et al. GW, et al. Penumbral imaging and func- 19. Janssen PM, Visser NA, Dorhout Mees
Outcomes of endovascular thrombectomy tional outcome in patients with anterior SM, Klijn CJM, Algra A, Rinkel GJE. Com-
vs medical management alone in patients circulation ischaemic stroke treated with parison of telephone and face-to-face as-
with large ischemic cores: a secondary endovascular thrombectomy versus medi- sessment of the modified Rankin scale.
analysis of the Optimizing Patient’s Selec- cal therapy: a meta-analysis of individual Cerebrovasc Dis 2010;29:137-9.
tion for Endovascular Treatment in Acute patient-level data. Lancet Neurol 2019;18: 20. Agresti A. Generalized odds ratios for
Ischemic Stroke (SELECT) study. JAMA 46-55. ordinal data. Biometrics 1980; 36:
59-67
Neurol 2019;76:1147-56. 14. Yoshimura S, Sakai N, Yamagami H, (https://www.jstor.org/stable/2530495).
9. Powers WJ, Rabinstein AA, Ackerson et al. Endovascular therapy for acute stroke 21. Churilov L, Arnup S, Johns H, et al.
T, et al. Guidelines for the early manage- with a large ischemic region. N Engl J An improved method for simple, assump-
ment of patients with acute ischemic Med 2022;386:1303-13. tion-free ordinal analysis of the modified
stroke: 2019 update to the 2018 guide- 15. Sarraj A, Hassan AE, Grotta J, et al. Rankin scale using generalized odds ra-
lines for the early management of acute Optimizing Patient Selection for Endovas- tios. Int J Stroke 2014;9:999-1005.
ischemic stroke: a guideline for health- cular Treatment in Acute Ischemic Stroke 22. Huo X, Ma G, Tong X, et al. Trial of
care professionals from the American (SELECT): a prospective, multicenter co- endovascular therapy for acute ischemic
Heart Association/American Stroke Asso- hort study of imaging selection. Ann Neu- stroke with large infarct. N Engl J Med
ciation. Stroke 2019;50(12):e344-e418. rol 2020;87:419-33. 2023;388:1272-83.
10. Turc G, Bhogal P, Fischer U, et al. Eu- 16. Barber PA, Demchuk AM, Zhang J, Copyright © 2023 Massachusetts Medical Society.
ropean Stroke Organisation (ESO)–Euro- Buchan AM. Validity and reliability of a


Select 2

Uploaded by

Copyright:

Available Formats

Select 2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Select 2

Uploaded by

Copyright:

Available Formats

new england

Trial of Endovascular Thrombectomy for Large Ischemic Strokes

1260 n engl j med 388;14 nejm.org April 6, 2023

The New England Journal of Medicine

indicating larger infarction) or an estimated imaging characteristics: age, National Institutes

n engl j med 388;14 nejm.org April 6, 2023 1261

1262 n engl j med 388;14 nejm.org April 6, 2023

The New England Journal of Medicine

n engl j med 388;14 nejm.org April 6, 2023 1263

Table 1. Baseline Characteristics of the Patients (Intention-to-Treat Population).*

Endovascular Thrombectomy Medical Care

1264 n engl j med 388;14 nejm.org April 6, 2023

The New England Journal of Medicine

Score on Modified Rankin Scale

Endovascular Thrombectomy 14.0 17.4 15.2 8.4 38.2

Medical Care 11.5 20.7 18.4 40.8

n engl j med 388;14 nejm.org April 6, 2023 1265

Table 2. Clinical Outcomes (Intention-to-Treat Population).*

1266 n engl j med 388;14 nejm.org April 6, 2023

The New England Journal of Medicine

Table 3. Safety Outcomes and Procedural Complications (Intention-to-Treat Population).*

n engl j med 388;14 nejm.org April 6, 2023 1267

Subgroup No. of Patients Generalized Odds Ratio (95% CI)

Medical Care Endovascular Thrombectomy

1268 n engl j med 388;14 nejm.org April 6, 2023

The New England Journal of Medicine

n engl j med 388;14 nejm.org April 6, 2023 1269

however, persisted numerically in prespecified stroke due to a proximal large-vessel occlusion

1270 n engl j med 388;14 nejm.org April 6, 2023

The New England Journal of Medicine

n engl j med 388;14 nejm.org April 6, 2023 1271

You might also like