RESEARCH ARTICLE

Tmax Volumes Predict Final Infarct Size

and Functional Outcome in Ischemic
Stroke Patients Receiving
Endovascular Treatment
Enrico Fainardi, MD, PhD ,1 Giorgio Busto, MD,2 Andrea Rosi, MD,3 Elisa Scola, MD,2
Ilaria Casetta, MD,4 Andrea Bernardoni, MD,5 Andrea Saletti, MD,5 Francesco Arba, MD,6
Patrizia Nencini, MD,6 Nicola Limbucci, MD,7 Salvatore Mangiafico, MD,8
Andrew Demchuk, MD,9,10,11 Mohammed A Almekhlafi, MD, MSc,9,10,11
Mayank Goyal, MD,9,10,11 Ting Y. Lee, PhD,12 Bijoy K. Menon, MD, MSc,9,10,11 and
13
Andrea Morotti, MD

Objective: The objective of this paper was to explore the utility of time to maximum concentration (Tmax)-based target
mismatch on computed tomography perfusion (CTP) in predicting radiological and clinical outcomes in patients with acute
ischemic stroke (AIS) with anterior circulation large vessel occlusion (LVO) selected for endovascular treatment (EVT).
Methods: Patients with AIS underwent CTP within 24 hours from onset followed by EVT. Critically hypoperfused tissue
and ischemic core volumes were automatically calculated using Tmax thresholds >9.5 seconds and >16 seconds, respec-
tively. The difference between Tmax > 9.5 seconds and Tmax > 16 seconds volumes and the ratio between
Tmax > 9.5 seconds and Tmax > 16 seconds volumes were considered ischemic penumbra and Tmax mismatch ratio,
respectively. Final infarct volume (FIV) was measured on follow-up non-contrast computed tomography (CT) at 24 hours.
Favorable clinical outcome was defined as 90-day modified Rankin Scale 0 to 2. Predictors of FIV and outcome were
assessed with multivariable logistic regression. Optimal Tmax volumes for identification of good outcome was defined
using receiver operating curves.
Results: A total of 393 patients were included, of whom 298 (75.8%) achieved successful recanalization and
258 (65.5%) achieved good outcome. In multivariable analyses, all Tmax parameters were independent predictors of FIV
and outcome. Tmax > 16 seconds volume had the strongest association with FIV (beta coefficient = 0.596 p <0.001)
and good outcome (odds ratio [OR] = 0.96 per 1 ml increase, 95% confidence interval [CI] = 0.95–0.97, p < 0.001).
Tmax > 16 seconds volume had the highest discriminative ability for good outcome (area under the curve [AUC] = 0.88,

View this article online at wileyonlinelibrary.com. DOI: 10.1002/ana.26354

Received Oct 20, 2021, and in revised form Mar 1, 2022. Accepted for publication Mar 7, 2022.

Address correspondence to Dr Fainardi, Struttura Organizzativa Dipartimentale di Neuroradiologia, Dipartimento di Scienze Biomediche,
Sperimentali e Cliniche “Mario Serio,” Università degli Studi di Firenze, Ospedale Universitario Careggi, Largo Brambilla 3, 50134 Firenze.
E-mail: enrico.fainardi@unifi.it

From the 1Neuroradiology Unit, Department of Experimental and Clinical Biomedical Sciences “Mario Serio”, University of Florence, Florence;
2
Neuroradiology Unit, Department of Radiology, Careggi University Hospital, Florence, Italy; 3Division of Diagnostic and Interventional Neuroradiology,
Department of Diagnostics, Geneva University Hospitals, Geneva, Switzerland; 4Section of Neurological, Psychiatric, and Psychological Sciences,
Department of Biomedical and Specialist Surgical Sciences, University of Ferrara, Ferrara, Italy; 5Neuroradiology Unit, Department of Radiology,
Arcispedale S. Anna, Ferrara, Italy; 6Stroke Unit, Careggi University Hospital, Florence, Italy; 7Interventional Neuroradiology Unit, Department of Radiology,
Careggi University Hospital, Florence, Italy; 8Neuroradiologia Diagnostica ed Interventisitca, IRCCS Neuromed, Istituto Neurologico Mediterraneo, Pozzilli,
Italy; 9The Calgary Stroke Program, Department of Clinical Neurosciences, University of Calgary, Calgary, Alberta, Canada; 10Department of Radiology,
University of Calgary, Calgary, Alberta, Canada; 11Hotchkiss Brain Institute, University of Calgary, Calgary, Alberta, Canada; 12Lawson Health Research
Institute and Robarts Research Institute, London, Ontario, Canada; and 13Department of Clinical and Experimental Sciences, Neurology Unit, University of
Brescia, Brescia, Italy

Additional supporting information can be found in the online version of this article.

878 © 2022 The Authors. Annals of Neurology published by Wiley Periodicals LLC on behalf of American Neurological Association.
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution
and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.
 Fainardi et al: Tmax Predicts Infarct Volume and Outcome

95% CI = 0.842–0.909). A Tmax > 16 seconds volume of ≤67 ml best identified subjects with favorable outcome (sensi-
tivity = 0.91 and specificity = 0.73).
Interpretation: Tmax target mismatch predicts radiological and clinical outcomes in patients with AIS with LVO receiv-
ing EVT within 24 hours from onset.
ANN NEUROL 2022;91:878–888

A growing body of data suggest that computed tomogra-

phy perfusion (CTP) is a useful tool for the selection of
patients with acute ischemic stroke (AIS) with anterior circu-
considered as penumbra. Tmax is a relatively robust parameter
to measure on CTP. The aim of the present study was to
assess the utility of this Tmax based mismatch paradigm in
lation large vessel occlusion (LVO) for endovascular treat- predicting imaging and clinical outcomes. Additionally, this
ment (EVT).1 A favorable CTP profile was successfully used analysis sought to identify the optimal Tmax volume thresh-
to establish the eligibility for EVT of patients with AIS with olds that would identify patients with favorable functional
LVO in 2 late time window (6–24 hours) randomized con- outcome on follow-up.
trolled trials (RCTs), namely DEFUSE 3 (Endovascular
Therapy Following Imaging Evaluation for Ischemic Stroke)2
Patients and Methods
and DAWN (Triage of Wake-up and Late Presenting Stro-
This cohort study was approved by the local ethics boards of
kes Undergoing Neurointervention With Trevo),3 and in
each participating site in which CT scans were performed
2 early time window (0–6 hours) RCTs, namely SWIFT
and clinical information were recorded during routine clinical
PRIME (Solitaire With the Intention for Thrombectomy as
activity. Written informed consent was obtained from each
Primary Endovascular Treatment)4 and EXTEND-IA
(Extending the Time for Thrombolysis in Emergency Neu- patient or from their legally authorized representatives at
rological Deficits-Intra-Arterial).5 In all these four CTP- admission or waived by the institutional review board.
guided RCTs,2–5 patients with AIS were selected for therapy
if meeting inclusion criteria consisting of a combination of Patient Selection
parameters collectively called target mismatch. Although cut- This was a retrospective analysis conducted on a prospec-
off values of target mismatch varied across these studies, they tively collected cohort of consecutive patients with AIS with
were always automatically calculated with a dedicated soft- anterior circulation LVO treated with EVT and admitted
ware (RAPID; Rapid Processing of Perfusion and Diffusion; from January 2013 to July 2017 at 2 academic Italian cen-
iSchemaView, Menlo Park, CA), based on the assumption ters: S. Anna University Hospital of Ferrara and Careggi
that critically hypoperfused tissue is indicated by absolute University Hospital of Florence. At both institutions, all
time to the peak of the residual function (time to maximum patients presenting with suspected AIS with LVO, and no
concentration [Tmax]) threshold values more than 6 seconds history of renal failure or contrast allergy routinely undergo
(Tmax > 6 seconds) and ischemic core corresponds to relative NCCT, CT angiography (CTA) of the cervical and intra-
cerebral blood flow (rCBF) threshold values less than 30% cranial vessels and CTP at admission if they arrive at the
of normally perfused tissue (rCBF < 30%). In this model, hospital within 24 hours of symptom onset. If diagnosis of
the difference between Tmax > 6 seconds and rCBF < 30% AIS is confirmed by neuroimaging findings, patients with
volumes (Tmax-CBF mismatch) represents ischemic penum- AIS receive EVT according to the current recommended
bra, whereas the ratio between Tmax > 6 and rCBF < 30% guidelines.8,9 Patients were included if they presented to
volume is the mismatch ratio. Of note, CTP-based and non the emergency department with the following criteria:
CTP-based RCTs shared the same limitation, a suboptimal (1) NCCT Alberta Stroke Programme Early Computed
patient selection as reflected by rates of functional depen- Tomography Score (ASPECTS) ≥ 6; (2) diagnosis of AIS
dency of around 50% despite successful recanalization.2,3,6 within 24 hours from witnessed symptom onset or time last
Recent publications that use the GE CTP 4D show that a seen well; (3) evidence of internal carotid artery (ICA)
Tmax value >16 seconds optimally identified follow-up infarct and/or middle cerebral artery (MCA) M1 or M2 segment
when non-contrast CT (NCCT) to recanalization time was occlusion on CTA; (4) CTP performed at admission;
within 90 minutes versus a Tmax value >9.5 seconds that (5) selected for receiving EVT; and (6) follow-up NCCT
identified follow-up infarct best when early reperfusion was imaging performed at 24 hours. Exclusion criteria were:
not achieved.7 Thus, in patients who achieved recanalization (1) NCCT ASPECTS < 6; (2) age < 18 years; (3) preg-
>90 minutes after NCCT Tmax > 16 seconds can be pre- nancy; (4) severe pre-stroke disability defined as modified
sumed to be the critically hypoperfused tissue. The difference Rankin Scale (mRS) ≥3; (5) detection of intracerebral hem-
between Tmax > 9.5 seconds and Tmax > 16 seconds volumes orrhage on admission NCCT; (6) contraindications to
(Tmax–Tmax mismatch) can therefore be operationally iodinated contrast agent; (7) poor quality of CT acquisition

June 2022 879

ANNALS of Neurology

due to motion artifacts; and (8) inability to complete multi- (maximum 90 ml), 5- to 10-second delay from injection
modal CT protocol at baseline and/or 24-hour follow-up to scanning, 120 kV, 270 mA, 1 second/rotation,
NCCT. NCCT ASPECTS (ASPECTS ≥ 6) was used for 1.25-mm thick slices, and table speed 3.75 mm/rotation.
establishing patient eligibility for EVT before publication of CTA covered from the carotid bifurcation to vertex.
the 2015 American Hospital Association / American Soci- CTP studies were obtained with a dynamic first-pass
ety of Anesthesiologists (AHA/ASA) guidelines,9 based on bolus-tracking methodology according to a 2-phase imag-
the analysis by Puez and colleagues.10 As suggested in the ing protocol, to avoid the truncation of time density cur-
2013 AHA/ASA guidelines,8 CTP was used to provide ves, with axial shuttle mode. The 2-phase acquisition
additional information regarding diagnosis. No patient was consisted of a first phase every 2.8 seconds for 60 seconds
excluded from the study due to a CTP unfavorable profile and an additional second phase every 15 seconds for
because CTP was not performed in patients with low 90 seconds, which started 5 seconds after the automatic
ASPECTS (NCCT ASPECTS < 6). Figure S1 illustrates injection of 40 ml of non-ionic contrast agent followed
the cohort selection process. by a saline flush of 40 ml at the rate of 4 ml/s. Sections
of 8 cm thickness were acquired at 5 mm slice thickness.
Clinical Assessment The other acquisition parameters were 80 kV, 140 mAs,
Demographic and clinical variables were collected by and 0.5 rotation time. All CTP source images were
investigators blinded to the outcomes of interest. In par- reconstructed with the standard filter and display field of
ticular, we obtained data on age, sex, pre-stroke functional view (DFOV) of 25 cm.
status (mRS), known versus unknown stroke onset time,
intravenous fibrinolysis with recombinant tissue plasmino- Imaging Processing and Analysis
gen activator (r-TPA), time from onset to baseline CT, The extent of early ischemic changes was evaluated on
and time from baseline CT to EVT conclusion. Baseline baseline NCCT using the ASPECTS methodology.10
stroke severity was measured with the National Institute Each CTP study was processed by commercially available
of Health Stroke Scale (NIHSS). Clinical outcome was delay-insensitive deconvolution software (CT Perfusion
measured using mRS at 3 months. The mRS ≤ 2 and >2 4D; GE Healthcare), as described elsewhere.7 In-plane
were defined as good and poor outcome, respectively. patient motion was corrected using an automated registra-
tion program included in the software, and the images
Imaging Acquisition with extreme motion at specific time points were manually
All imaging was conducted on 64-slice scanners removed, as needed, by visual inspection of the cine series
(GE Healthcare, Waukesha, WI). NCCT helical scans were and time density curve (TDC). For each study, the TDC
performed from the skull base to the vertex using the fol- for the arterial input function (AIF) and for venous output
lowing imaging parameters: 120 kV, 340 mA, 4  5-mm function (VOF) were measured from the basilar artery,
collimation, 1 second/rotation, and table speed of ICA, or anterior cerebral artery and from the superior sag-
15 mm/rotation. CTA of the cervical and intracranial ittal sinus, respectively using a 2 voxel  2 voxel (in-slice)
vessels was performed as follows: 0.7 ml/kg contrast regions of interest (ROIs). The VOF-TDC was used to

FIGURE 1: Automatic segmentation of critically hypoperfused tissue and infarct core volumes based on Tmax >9.5seconds and
>16 seconds threshold values. Panel A: Color coded CTP Tmax map with scale set at 9.6 to 25 seconds. Panel B: Color coded CTP
Tmax map with scale set at 16.1 to 25 seconds. Panel C: Tmax >9.5 seconds volume automatically segmented on CTP averaged
images. Panel D: Tmax >16 seconds volume automatically segmented on CTP averaged images. CTP = computed tomography
perfusion; Tmax = time to maximum concentration.

880 Volume 91, No. 6

 Fainardi et al: Tmax Predicts Infarct Volume and Outcome

TABLE 1. Study Population Characteristics

All mRS 0–2 mRS 3–6
n = 393 n = 258 n = 135 p

Age, median (IQR), yr 73 (65–78) 72 (64–78) 75 (68–80) <0.001

Sex, males, n (%) 215 (54.7) 136 (52.7) 79 (58.5) 0.272
Pre-stroke mRS
0, n (%) 324 (82.4) 220 (85.3) 104 (77.0) 0.047
1, n (%) 56 (14.2) 33 (12.8) 23 (17.0)
2, n (%) 13 (3.3) 5 (1.9) 8 (5.9)
Admission NIHSS, median (IQR) 14 (10–20) 12 (9–15) 21 (17–23) <0.001
ASPECTS score, median (IQR) 9 (8–10) 9 (8–10) 8 (7–10) <0.001
Occlusion site
Cervical ICA, n (%) 86 (21.9) 44 (17.1) 42 (31.1) <0.001
Terminal ICA, n (%) 14 (3.6) 4 (1.6) 10 (7.4)
MCA M1, n (%) 266 (67.7) 188 (72.9) 78 (57.8)
MCA M2, n (%) 27 (6.9) 22 (8.5) 5 (3.7)
r-TPA before EVT, n (%) 126 (32.1) 88 (34.1) 38 (28.1) 0.229
Time from onset to NCCT, median (IQR) minutes 279 (118–333) 281 (118–327) 257 (116–354) 0.961
Time from NCCT to reperfusion, median (IQR), minutes 124 (100–163) 112 (96–138) 151 (121–213) <0.001
Collateral score
0, n (%) 14 (3.6) 0 (0.0) 14 (10.4) <0.001
1, n (%) 106 (27.0) 17 (6.6) 89 (65.9)
2, n (%) 155 (39.4) 134 (51.9) 21 (15.6)
3, n (%) 118 (30.0) 107 (41.5) 11 (8.1)
mTICI score
0, n (%) 34 (8.7) 8 (3.1) 26 (19.3) <0.001
1, n (%) 24 (6.1) 1 (0.4) 21 (17.0)
2a, n (%) 37 (9.4) 4 (1.6) 33 (24.4)
2b, n (%) 63 (16.0) 40 (15.5) 23 (17.0)
3, n (%) 235 (59.8) 205 (79.5) 30 (22.2)
Tmax >9.5 seconds volume, median (IQR), ml 102 (55–160) 69 (41–112) 175 (121–218) <0.001
Tmax >16 seconds volume, median (IQR), ml 30 (11–77) 20 (6–35) 93 (50–110) <0.001
Tmax mismatch volume, median (IQR), ml 56 (35–91) 48 (29–73) 81 (51–117) <0.001
Tmax mismatch ratio, median 2.7 (2.0–5.2) 3.7 (2.4–6.7) 2.0 (1.6–2.5) <0.001
Final infarct volume, median (IQR), ml 29 (10–92) 16 (5–29) 103 (87–135) <0.001
Hemorrhagic transformation ECASS II
None, n (%) 252 (64.1) 195 (75.6) 57 (42.2) <0.001
HT1, n (%) 63 (16.0) 41 (15.9) 22 (16.3)
HT2, n (%) 32 (8.1) 11 (4.3) 21 (15.6)
PH1, n (%) 20 (5.1) 7 (2.7) 13 (9.6)
PH2, n (%) 26 (6.6) 4 (1.6) 22 (16.3)
sICH, n (%) 39 (9.9) 7 (2.7) 32 (23.7) <0.001
Any hemorrhagic transformation, n (%) 141 (35.8) 78 (57.8) 63 (24.4) <0.001

ASPECTS = Alberta Stroke Program Early Computed Tomography Score; DSA = digital subtraction angiography; ECASS = European Cooperative
Acute Stroke Study; EVT = endovascular treatment; HI1 = hemorrhagic infarction type 1; HI2 = hemorrhagic infarction type 2; ICA = internal
carotid artery; IQR = interquartile range; MCA = middle cerebral artery; mRS = modified Rankin Scale; mTICI = modified treatment in cerebral
infarction score; NCCT = non-contrast computed tomography; NIHSS = National Institute of Health Stroke Scale; PH1 = parenchymal hemorrhage
type 1; PH2 = parenchymal hemorrhage type 2; r-TPA = recombinant tissue plasminogen activator; sICH = symptomatic intracerebral hemorrhage;
Tmax = time to maximum concentration.

June 2022 881

ANNALS of Neurology

TABLE 2. Multivariable Predictors of Good

Functional Outcome
OR (95% CI) p

MODEL 1
Age, yr 0.96 (0.92–1.01) 0.085
Admission NIHSS 0.84 (0.78–0.91) <0.001
Collateral score 9.74 (4.90–19.34) <0.001
mTICI score 5.03 (3.17–7.98) <0.001
Tmax >9.5 seconds 0.98 (0.97–0.99) <0.001
volume, ml
MODEL 2
Admission NIHSS 0.83 (0.76–0.90) <0.001
Collateral score 7.29 (3.76–14.13) <0.001
mTICI score 4.76 (3.06–7.42) <0.001

Tmax >16 seconds 0.96 (0.95–0.97) <0.001

volume, ml
MODEL 3
Admission NIHSS 0.82 (0.76.0.88) <0.001 FIGURE 2: Tmax volumes and Tmax mismatch ratio optimal
values for recognizing patients with AIS with good clinical
Collateral score 11.28 (5.91–21.54) <0.001
outcome as calculated using ROC curves. AIS = acute ischemic
mTICI score 4.53 (2.976.92) <0.001 stroke; AUC = area under the curve; CI = confidence interval;
ROC = Receiver Operating Characteristic; SE = standard error;
Tmax mismatch 0.98 (0.97–0.99) <0.001 Tmax = time to maximum concentration; Tmax 9.5 seconds, Tmax
volume, ml >9.5 seconds volume; Tmax 16 seconds, Tmax >16 seconds
volume. AUC comparison was performed with DeLong test.
MODEL 4 Outcome of interest: modified Rankin Scale 0 to 2 at 90 days.
[Color figure can be viewed at www.annalsofneurology.org]
Admission NIHSS 0.80 (0.74–0.87) <0.001
Collateral score 8.58 (4.69–15.72) <0.001 images over the duration of the first pass of contrast. These
mTICI score 4.18 (2.83–6.18) <0.001 average CTP images were used to exclude cerebrospinal
fluid (CSF) and skull from analysis using Hounsfield
Tmax mismatch ratio 1.17 (1.02–1.35) 0.026
unit (HU) thresholds. Large blood vessels were automatically
Logistic regression with backward elimination at p < 0.1. Variables excluded from calculation by the software. Critically
entered into the model: age, pre-stroke mRS, and NIHSS, ASPECTS, hypoperfused tissue and ischemic core volumes were defined
time from CT to recanalization, mTICI, collateral score, occlusion
as ischemic brain regions with Tmax values >9.5 seconds and
site; Tmax parameters entered separately into different models.
ASPECTS = Alberta Stroke Programme Early Computed Tomogra-
>16 seconds, respectively, and were automatically segmented
phy Score; CI = confidence interval; CT = computed tomography; and calculated by the software (Figure 1). According to the
mRS = modified Rankin Scale; mTICI = modified treatment in Tmax target mismatch paradigm,2,4,5 the difference between
cerebral infarction score; NIHSS = National Institute of Health Tmax > 9.5 seconds and Tmax > 16 secondsvolumes (Tmax mis-
Stroke Scale; OR = odds ratio; Tmax = time to maximum match) was considered as ischemic penumbra, and the ratio
concentration.
between Tmax > 9.5 seconds and Tmax > 16 seconds volumes
was defined as the Tmax mismatch ratio. Occlusion sites were
correct for partial volume averaging in the AIF. CBF, cerebral identified on CTA and classified as cervical ICA, terminal
blood volume (CBV), and Tmax maps were generated for each ICA, middle cerebral artery (MCA) M1 segment, or MCA
patient by deconvolving the AIF from tissue TDCs. CBF, M2 segment occlusions. CTA collateral supply was graded on
CBV, and Tmax values were expressed in ml
min 1
(100 g) 1, a 4-point scale according to a previously published scoring sys-
ml
(100 g) 1 and seconds, respectively. Average CTP maps tem in which collaterals were categorized as absent (score 0),
were created by averaging the cine (dynamic) CTP source >0% but ≤50% (score 1), >50% but <100% (score 2) and

882 Volume 91, No. 6

 Fainardi et al: Tmax Predicts Infarct Volume and Outcome

TABLE 3. Test Characteristics of Tmax Volumes and Tmax Mismatch Ratio

Sensitivity(95% CI) Specificity (95% CI) PPV (95% CI) NPV (95% CI)

Tmax > 9.5 seconds volume ≤ 111.6 ml 0.76 (0.70–0.81) 0.82 (0.74–0.88) 0.89 (0.84–0.92) 0.64 (0.59–0.69)
Tmax > 16 seconds volume ≤ 67.0 ml 0.91 (0.87–0.94) 0.73 (0.64–0.80) 0.86 (0.83–0.89) 0.81 (0.74–0.87)
Tmax mismatch volume ≤ 58.3 ml 0.63 (0.57–0.69) 0.70 (0.61–0.77) 0.80 (0–75-0.84) 0.50 (0.45–0.55)
Tmax mismatch ratio > 2.5 0.70 (0.64–0.76) 0.79 (0.71–0.85) 0.86 (0.82–0.90) 0.58 (0.53–0.63)
CI = confidence interval; NPV = negative predictive value; PPV = positive predictive value; Tmax = time to maximum concentration. Outcome of
interest: modified Rankin Scale 0 to 2 at 90 days.

100% (score 3) of the occluded territory.11 Recanalization was heteroskedasticity (with log-transformation of FIV).
assessed on conventional digital subtraction angiography Models were adjusted for age, admission NIHSS score,
(DSA) at the end of endovascular therapy using the modified ASPECTS score, collateral score, reperfusion status
treatment in cerebral ischemia (mTICI) scale.12 Patients with (defined as mTICI score 2b/3) and variables with p < 0.1
mTICI score of 2b or 3 were considered as successfully in univariable analysis. Backward elimination was used in
recanalized, whereas patients with mTICI score ranging from both models to reach to a final parsimonious model that
0 to 2a were classified as not. Hemorrhagic transformation avoids model overfitting. Interaction terms were used in
(HT) was classified on NCCT at 24 hours from symptom regression models to test the interaction among Tmax
onset/last known well according to the European Cooperative parameters, reperfusion status, and time to reperfusion
Acute Stroke Study (ECASS)-II criteria into four different cat- (defined as time from baseline NCCT to DSA end). The
egories: hemorrhagic infarction type 1 (HI1), HI type utility of different Tmax volumes and Tmax mismatch ratio
2 (HI2), Parenchymal hemorrhage type 1 (PH1), and PH for identification of patients with good functional out-
type 2 (PH2).13 Symptomatic intracranial hemorrhage (sICH) come was analyzed using area under the curve (AUC)
was considered as any intracranial hemorrhage associated with Receiver Operating Characteristic (ROC) curves and opti-
a ≥4-point increase in NIHSS. Final infarct volume (FIV) was mal sensitivity, specificity, positive predictive value, and
measured on follow-up NCCT at 24 hours after symptom negative predictive values identified using the Youden
onset/last known well with a multislice planimetric method by Index. Comparison of models with Tmax > 9.5 seconds,
summation of the hypodense areas, manually traced on each Tmax > 16 seconds, mismatch volume and mismatch ratios
slice in which they were detectable, multiplied by slice as independent variables and the mRS 0 to 2 at 90 days as
thickness.14 dependent variable was performed using the DeLong
test.15 A secondary analysis was focused on subjects pre-
senting after 6 hours from symptom onset or time last
Statistical Analysis
seen well, with the same outcomes of interest explored in
Continuous variables were summarized as median (inter-
the main analysis. In this subgroup of patients, the model
quartile range [IQR]) or mean (standard deviation [SD])
building strategy was the same as in the main analyses.
as appropriate based on their distribution assessed with
Finally, a sensitivity analysis explored the diagnostic per-
the Shapiro–Wilk test. Mann–Whitney test and Student’s
formance of Tmax parameters in patients who achieved
t test were used to compare continuous variables with
mTICI (2b/3) post EVT versus those achieving
non-normal and normal distributions, respectively. Cate-
mTICI ≤ 2a. All analyses were performed with the statisti-
gorical variables were summarized as count (percentage)
cal packages SPSS version 21.0 (www.spss.com) and
and compared using the chi-square test. Good functional
MedCalc (www.medcalc.org). Statistical significance was
prognosis (defined as mRS 0–2) at 90 days from stroke
set at two-sided p < 0.05.
onset and FIV and were the main outcomes of interest.
Variables associated with good functional outcome were
assessed using multivariable logistic regression, adjusting Results
for age, admission NIHSS, ASPECTS score, collateral We screened 477 potentially eligible patients with AIS
score, reperfusion status, and any variable showing signifi- with anterior circulation LVO, of whom 84 patients were
cance at p < 0.1 in univariable analysis. Variables associ- excluded due to the presence of low ASPECTS (<6) on
ated with FIV were explored with multivariable linear baseline NCCT, baseline intracerebral hemorrhage
regression after testing for normality of residuals and (n = 19), inability to complete multimodal CT protocol

June 2022 883

ANNALS of Neurology

TABLE 4. Univariable Predictors of Final Infarct TABLE 5. Multivariable Predictors of Final Infarct
Volume Volume
B (SE) p B (SE) p

Age, yr 0.151 (0.003) 0.003 MODEL 1

Sex, male 0.021 (0.066) 0.675 Admission NIHSS 0.189 (0.004) <0.001
Admission NIHSS 0.581 (0.004) <0.001 Time from NCCT to DSA 0.077 (0.000) 0.017
end, minutes
ASPECTS score 0.239 (0.021) <0.001
Collateral score 0.090 (0.027) 0.012
Carotid occlusion 0.209 (0.073) <0.001
mTICI 2b/3 0.169 (0.049) <0.001
Unknown onset 0.082 (0.092) 0.106
Tmax > 9.5 seconds 0.544 (0.000) <0.001
r-TPA before EVT 0.022 (0.071) 0.664
volume, ml
Time from onset to 0.035 (0.000) 0.492
MODEL 2
NCCT, minutes
Admission NIHSS 0.203 (0.081) <0.001
Time from NCCT to 0.331 (0.000) <0.001
DSA end, minutes Time from NCCT to DSA 0.075 (0.000) 0.017
end, minutes
Collateral score 0.482 (0.034) <0.001
Collateral score 0.146 (0.048) <0.001
mTICI 2b/3 0.428 (0.069) <0.001
mTICI 2b/3 0.596 (0.001) <0.001
Tmax > 9.5 volume, ml 0.747 (0.000) <0.001
Tmax > 16 seconds volume, ml 0.596 (0.001) <0.001
Tmax > 16 volume, ml 0.773 (0.000) <0.001
MODEL 3
Tmax mismatch volume, 0.523 (0.001) <0.001
ml Admission NIHSS 0.282 (0.004) <0.001
Tmax mismatch ratio 0.460 (0.002) <0.001 Time from NCCT to DSA 0.099 (0.000) 0.007
end, minutes
ASPECTS = Alberta Stroke Program Early Computed Tomography
Score; B = beta coefficient; DSA = digital subtraction angiography; Collateral score 0.199 (0.030) <0.001
EVT = endovascular treatment; mTICI = modified treatment in
cerebral infarction score; NCCT = non-contrast computed tomogra- mTICI 2b/3 0.203 (0.056) <0.001
phy; NIHSS = National Institute of Health Stroke Scale; r- Tmax mismatch volume, ml 0.333 (0.001) <0.001
TPA = recombinant tissue plasminogen activator; SE = standard
error. MODEL 4
Admission NIHSS 0.320 (0.004) <0.001

at baseline and/or 24-hour follow-up NCCT (n = 15), Time from NCCT to DSA 0.110 (0.000) 0.003
poor quality of CT acquisition due to motion artifacts end, minutes
(n = 12), contraindications to iodinated contrast agent Collateral score 0.195 (0.031) <0.001
(known contrast allergy, renal failure) (n = 8), pre-stroke
mTICI 2b/3 0.205 (0.057) <0.001
mRS >3 (n = 5), pregnancy (n = 3), and < 18 of age
(n = 2). The study selection flowchart is illustrated in Tmax mismatch ratio 0.307 (0.002) <0.001
Figure S1. Logistic regression with backward elimination at p < 0.1. Variables
Overall, 393 patients (median age = 73, entered into the model: age, NIHSS, ASPECTS, carotid occlusion,
IQR = 65–78, 54.7% men, median NIHSS = 14, time from CT to recanalization, collateral score, mTICI 2b/3, Tmax,
Tmax parameters entered separately into different models.
IQR = 10–20) met study eligibility criteria. Of these,
B = beta coefficient; CT = computed tomography; DSA = digital
314 were selected for EVT in the early (<6 hours) and subtraction angiography; EVT = endovascular treatment;
79 in the late (6–24 hours) time windows. Most of the mTICI = modified treatment in cerebral infarction score;
included patients had a middle cerebral artery occlusion in NCCT = non-contrast computed tomography; NIHSS = National
the M1 segment and around one third of the study popu- Institute of Health Stroke Scale; r-TPA = recombinant tissue plas-
minogen activator; Tmax = time to maximum concentration;
lation received r-TPA before EVT. Table 1 summarizes
SE = standard error.
the patient characteristics and shows the comparison

884 Volume 91, No. 6

 Fainardi et al: Tmax Predicts Infarct Volume and Outcome

FIGURE 3: Tmax volumes in 2 patients with AIS and anterior circulation large vessel occlusion. Panel A: NCCT at admission, color
coded CTP Tmax map with scale set at 9.6 to 25 seconds, Color coded CTP Tmax map with scale set at 16.1 to 25 seconds, Tmax
> 9.5 seconds volume automatically segmented on CTP averaged images, Tmax > 16 seconds volume automatically segmented on
CTP averaged images and NCCT at 24 hours in a 48 year-old patient with AIS and ischemic lesion in left MCA territory who
presented the following Tmax parameters: critically hypoperfused tissue = 39.5 ml; core volume = 8.7 ml; penumbra
volume = 30.8 ml; and mismatch ratio = 4.5. FIV was 7.2 ml and mRS at 3 months was 0. Panel B: NCCT at admission, color coded
CTP Tmax map with scale set at 9.6 to 25 seconds, Color coded CTP Tmax map with scale set at 16.1 to 25 seconds, Tmax >9.5 seconds
volume automatically segmented on CTP averaged images, Tmax >16 seconds volume automatically segmented on CTP averaged
images and NCCT at 24 hours with hemorrhagic transformation in a 67 year old patient with AIS and ischemic lesion in left MCA
territory who presented the following Tmax parameters: critically hypoperfused tissue = 158.0 ml; core volume = 87.4 ml; penumbra
volume = 70.6 ml; mismatch ratio = 1.8. FIV was 131.2 ml and mRS at 3 months was 4. AIS = acute ischemic stroke; CTP = computed
tomography perfusion; FIV = final infarct volume; MCA = middle cerebral artery; mRS = modified Rankin Scale; NCCT = non-contrast
computed tomography; Tmax = time to maximum concentration.

between patients with good and poor functional outcome When Tmax volumes and Tmax mismatch ratio were analyzed
at 3 months. In univariable analyses, patients with good as dichotomous variables, the optimal cutoff points for
prognosis had lower Tmax > 9.5 seconds volumes, predicting favorable 90-day clinical outcome were ≤111.6 ml
Tmax > 16 seconds volumes, Tmax mismatch volumes, and for Tmax > 9.5 seconds, ≤67.0 ml for Tmax > 16 seconds,
higher Tmax mismatch ratio. As shown in Table 2, all these ≤58.3 ml for Tmax mismatch volume, and >2.5 for Tmax mis-
associations between Tmax parameters and clinical outcome match ratio. Tmax > 16 seconds volume ≤67.0 ml had the
remained significant in multivariable logistic regression analy- highest sensitivity (0.91; 95% CI = 0.87–0.94) for identifica-
sis after adjustment for potential confounders. No statistically tion of subjects with favorable 90-day clinical outcome, whereas
significant interaction was noted among Tmax > 9.5 seconds Tmax > 9.5 seconds volume ≤111.6 ml showed the highest spec-
volume, Tmax > 16 seconds volume, and reperfusion time or ificity (0.82; 95% CI = 0.74–0.88). The test characteristics of
reperfusion status in association with 90 day clinical outcome different Tmax volumes are summarized in Table 3. In a second-
(all p values for interaction > 0.1). Conversely, the association ary analysis stratified by mTICI status, results remained consis-
between Tmax mismatch volume and outcome was significant tent. In particular, Tmax > 16 secondsvolume ≤67.0 ml
only in patients with time from CT to reperfusion remained the most sensitive (>0.90) parameter for the identifica-
<165 minutes (p value for interaction < 0.001). Moreover, tion of patients with good prognosis, regardless of the degree of
Tmax mismatch ratio predicted functional outcome only in recanalization (Table S1).
patients achieving a mTICI score of 2b/3 (p for interac- As illustrated in Table 4, Tmax > 9.5 seconds vol-
tion = 0.017). The ROC curve analysis shown in Figure 2 ume, Tmax > 16 seconds volume, Tmax mismatch volume
demonstrated that Tmax > 16 seconds volume had the and Tmax mismatch ratio were all associated with the
highest discriminative ability (AUC 0.878; 95% confidence extent of FIV in unadjusted analysis. All Tmax variables
interval [CI] = 0.842–0.909) for 90 day clinical outcome. remained independently associated with FIV (p < 0.001)

June 2022 885

ANNALS of Neurology

in multivariable linear regression (Table 5). When the of these parameters to predict radiological and clinical
analysis was restricted to 79 patients presenting 6 hours outcomes.
from stroke onset, of whom 47 (59.5%) had favorable Our results are in line with previous investigations
functional outcome, the discriminative ability of Tmax vol- showing that in patients with AIS who underwent
umes and Tmax mismatch volume remained good EVT < 6 hours16 and 6–16 hours17 after symptom onset/
(Tmax > 9.5 seconds volume, AUC 0.86, p < 0.001; last known well time, ischemic core and hypoperfusion
Tmax > 16 seconds volume AUC 0.88, p < 0.001; Tmax volumes were associated with FIV. In addition, as in a
mismatch volume AUC 0.74, p < 0.001; Tmax mismatch prior work on patients receiving EVT < 6 hours from
ratio AUC 0.77, p < 0.001). In this subgroup of patients, onset,18 we also confirm that the 2 variables indicating
the optimal value for predicting good outcome was similar penumbra size, namely, Tmax mismatch volumes and Tmax
to that obtained in overall analyses with Tmax > 9.5 seconds mismatch ratio, predict FIV. The inclusion in our analysis
volume (≤114.4 ml, sensitivity 0.72, specificity 0.84), of both patients who achieved reperfusion, in whom pen-
Tmax > 16 secondsvolume (≤67.4 ml, sensitivity 0.89, and umbral tissue is presumed to be saved, and patients who
specificity 0.75) and Tmax mismatch ratio (>2.8, sensitivity did not achieve reperfusion, in whom penumbra evolves
0.62 and specificity 0.84), whereas Tmax mismatch volume into infarct, could in part explain our results. More inter-
was lower than overall analysis (≤47.1 ml, sensitivity 0.53, esting was the demonstration that Tmax > 9.5 seconds vol-
and specificity 0.84). Tmax > 9.5 seconds volume ≤ 114.4, umes, Tmax > 16 seconds volumes, Tmax mismatch
ml and Tmax > 16 seconds volume ≤ 67.4 ml were the most volumes, and Tmax mismatch ratio were associated 90-day
reliable parameters for identification of patients with favor- good outcome. This association was in line with data from
able clinical outcome, with highest sensitivity (89%) and RCTs1–4 and other studies18–20 suggesting that patients
specificity (84%), respectively. Logistic and linear regression with AIS with favorable target mismatch profile achieved a
models showed that all Tmax parameters remained indepen- good response to EVT. In this setting, the high predictive
dently associated with functional outcome and FIV in these value of Tmax > 9.5 seconds and the poor performance of
late presenters, except for the lack of association between Tmax mismatch ratio for favorable outcome are not unex-
Tmax mismatch ratio and outcome at 90 days (odds ratio pected. Whereas Tmax > 9.5 seconds volume represents
[OR] = 1.11; 95% CI = 0.86–1.43; p = 0.431). Two critically hypoperfused tissue, including both ischemic
illustrative cases showing the application of Tmax volumes core and ischemic penumbra, which strongly correlates
based on the established thresholds of >9.5 seconds and with good prognosis, prior evidence suggests that mis-
>16 secondsare depicted in Figure 3. match ratios cannot be considered as robust markers of
favorable outcome.21 It is important to highlight that
Tmax volumes and mismatch ratio were associated with
Discussion FIV and 90 day clinical outcomes independently from
The main finding of this study is that multiple CTP Tmax NIHSS, recanalization, and collateral score, which are
parameters are independently associated with functional considered strong predictors of radiological and clinical
outcome and with final infarct extent in patients with AIS outcomes. In particular, Tmax parameters and collaterals
undergoing EVT for LVO. In the context of RCTs show- predicted outcome independently from each other,
ing the effectiveness of EVT in patients with AIS with suggesting that Tmax is not an epiphenomenon of collat-
LVO, CTP provided additional value in the early time win- eral extent, but a complementary marker of delay in vessel
dow1,4 and was mandatory in the late time window2,3 for filling of ischemic tissue.22
patients’ selection. Inclusion criteria in these RCTs were The optimal cutoffs for Tmax parameters that predict
based on the assumption that Tmax > 6 seconds and good clinical outcome are similar to those in previous
rCBF < 30% threshold values represented critically studies: a hypoperfusion volume ≤ 111.6 ml was compara-
hypoperfused tissue and ischemic core, respectively, and ble to the optimal value selected in DEFUSE
could be used to select patients for EVT. However, the 2 (<100 ml),23 whereas an infarct volume ≤ 67.0 ml was
identification of patients likely to benefit from EVT similar to a cutoff point used in EXTEND-IA5 and
remains suboptimal as 50% of patients with AIS receiving DEFUSE 32 (<70 ml). A penumbra volume ≤ 58.3 ml
successful reperfusion do not achieve a good functional out- was very different from the cut-off values used in
come.2,3,6 A recent study suggested that critically EXTEND-IA5 (>10 ml), and in DEFUSE 32 and SWIFT
hypoperfused tissue and ischemic core could be better PRIME4 (>15 ml), but in agreement with the concept
defined by Tmax > 9.5 seconds and Tmax > 16 seconds that a RAPID Tmax > 8 seconds volume > 85 mL24 indi-
threshold values (GE CTP4D), respectively.7 Our findings cates a malignant profile reflecting severely hypoperfused
expand this previous observation demonstrating the ability tissue destined to progress into infarction. A Tmax

886 Volume 91, No. 6

 Fainardi et al: Tmax Predicts Infarct Volume and Outcome

mismatch ratio >2.5 in this analysis was higher than the actual applicability of Tmax target mismatch in selecting
thresholds used in SWIFT PRIME (>1.2),4 and in patients with AIS with LVO who can benefit from EVT.
EXTEND-IA5 and DEFUSE 32 (>1.8), but equivalent to
the value proposed by other investigators (>2.6)21 who rec-
ommended a larger mismatch ratio for better detecting the Acknowledgment
benefit of reperfusion therapies. Optimal thresholds to dis- Open Access Funding provided by Universita degli Studi
criminate good outcome did not substantially change when di Firenze within the CRUI-CARE Agreement.
the subset of patients presenting in the late time window
were examined. Taken together, these data suggest that the Author Contributions
possibility of achieving a favorable outcome increases in the Conception and design of study: A.M. and
presence of a small ischemic core volume, a significant mis- E.F. Acquisition and analysis of data: all authors. Drafting
match between the extent of core and hypoperfusion, and a manuscript and figures: all authors.
large, but not oversized, ischemic penumbra volume. More-
over, this study proposes a new CTP target mismatch based Potential Conflict of Interest
on specific threshold values of Tmax alone for identifying
T.-Y.L. has a licensing agreement with GE Healthcare for the
critically hypoperfused tissue (Tmax > 9.5 seconds) and
CT Perfusion software. Dr. Goyal participated to ESCAPE,
ischemic core (Tmax > 16 seconds).
REVASCAT, and SWIFT PRIME trials. Dr. Demchuk con-
This study is not without limitations. First, as this
tributed to ESCAPE and REVASCAT trials and Dr. Menon
study was based on retrospective analysis, our findings
was involved in the ESCAPE trial.
require prospective validation. Second, this was a non-
randomized study consisting of a selected population in
which eligibility for EVT was decided by local stroke team
and data from patients not receiving EVT were lacking. References
1. Campbell BCV. Optimal imaging at the primary stroke center. Stroke
Therefore, we cannot exclude the influence of unmeasured
2020;51:1932–1940.
confounders and of selection bias in our analysis. Third,
2. Albers GW, Marks MP, Kemp S, et al. Thrombectomy for stroke at
early presenters were more represented than late pre- 6 to 16 hours with selection by perfusion imaging. N Engl J Med
senters, making the 2 groups unbalanced. Fourth, these 2018;378:708–718.
results are conditional on the use of the GE CTP 4D 3. Nogueira RG, Jadhav AP, Haussen DC, et al. Thrombectomy 6 to
algorithm; corresponding values for Tmax may need to be 24 hours after stroke with a mismatch between deficit and infarct. N
Engl J Med 2018;378:11–21.
derived for other CTP algorithms.25,26 Fifth, a comparison
4. Saver JL, Goyal M, Bonafe A, et al. Stent-retriever thrombectomy
with RAPID thresholds as part of the DEFUSE 3 selection after intravenous t-PA vs. t-PA alone in stroke. N Engl J Med 2015;
criteria was not performed, and therefore we are unable to 372:2285–2295.
comment whether Tmax target mismatch is different to 5. Campbell BC, Mitchell PJ, Kleinig TJ, et al. Endovascular therapy for
DEFUSE 3 parameters based on Tmax-CBF mismatch for ischemic stroke with perfusion-imaging selection. N Engl J Med
2015;372:1009–1018.
differentiating good and poor radiological and clinical out-
comes. Sixth, although consistent with some previous tri- 6. Goyal M, Menon BK, van Zwam WH, et al. Endovascular thrombectomy
after large-vessel ischaemic stroke: a meta-analysis of individual patient
als and meta-analyses,2,27 the use of a dichotomized mRS data from five randomised trials. Lancet 2016;387:1723–1731.
may have reduced statistical power and neglected possible 7. d’Esterre CD, Boesen ME, Ahn SH, et al. Time-dependent computed
outcome shifts in the mRS range 3 to 6.28 Seventh, the tomographic perfusion thresholds for patients with acute ischemic
stroke. Stroke 2015;46:3390–3397.
different devices used in our study period and the evolu-
tion of EVT technologies therefore during this period29 8. Jauch EC, Saver JL, Adams HP Jr, et al. Guidelines for the early manage-
ment of patients with acute ischemic stroke: a guideline for healthcare
may have influenced the results of this analysis. Finally, professionals from the American Heart Association/American Stroke
the exclusion of patients with low ASPECTS scores did Association. Stroke 2013;44:870–947.
not allow us to verify the rates of patients with unfavor- 9. Powers WJ, Derdeyn CP, Biller J, et al. 2015 American Heart
able NCCT and favorable CTP profiles who achieved a Association/American Stroke Association focused update of the 2013
guidelines for the early management of patients with acute ischemic
good functional outcome that was reported to be about stroke regarding endovascular treatment: a guideline for healthcare
60% in a recent study.19 professionals from the American Heart Association/American Stroke
Association. Stroke 2015;46:3020–3035.
In conclusion, we demonstrate the potential of a
simple Tmax only target mismatch paradigm in predicting 10. Puetz V, Dzialowski I, Hill MD, et al. The Alberta stroke program early
CT score in clinical practice: what have we learned? Int J Stroke
radiological and clinical outcomes in patients with AIS 2009;4:354–364.
undergoing EVT both in the early and late time windows. 11. Tan IYL, Demchuk AM, Hopyan J, et al. CT angiography clot burden
Further prospective studies are warranted to clarify the score and collateral score: correlation with clinical and radiologic

June 2022 887

ANNALS of Neurology

outcomes in acute middle cerebral artery infarct. AJNR 20. Olivot JM, Albucher JF, Guenego A, et al. Mismatch profile influences
Am J Neuroradiol 2009;30:525–531. outcome after mechanical thrombectomy. Stroke 2021;52:232–240.

12. Zaidat OO, Yoo AJ, Khatri P, et al. Cerebral angiographic revascular- 21. Kakuda W, Lansberg MG, Thijs VN, et al. Optimal definition for
ization grading (CARG) collaborators; STIR revascularization working PWI/DWI mismatch in acute ischemic stroke patients. J Cereb Blood
group; STIR thrombolysis in cerebral infarction (TICI) task force. Rec- Flow Metab 2008;28:887–891.
ommendations on angiographic revascularization grading standards 22. d’Esterre CD, Trivedi A, Pordeli P, et al. Regional comparison of
for acute ischemic stroke: a consensus statement. Stroke 2013;44: multiphase computed tomographic angiography and computed
2650–2663. tomographic perfusion for prediction of tissue fate in ischemic
13. Hacke W, Kaste M, Fieschi C, et al. Randomised double-blind placebo- stroke. Stroke 2017;48:939–945.
controlled trial of thrombolytic therapy with intravenous alteplase in 23. Lansberg MG, Straka M, Kemp S, et al. MRI profile and response to
acute ischaemic stroke (ECASS II). Second European-Australasian acute endovascular reperfusion after stroke (DEFUSE 2): a prospective
stroke study investigators. Lancet 1998;352:1245–1251. cohort study. Lancet Neurol 2012;11:860–867.

14. Brott T, Marler JR, Olinger CP, et al. Measurements of acute cerebral 24. Mlynash M, Lansberg MG, De Silva DA, et al. Refining the definition
infarction: lesion size by computed tomography. Stroke 1989;20: of the malignant profile: insights from the DEFUSE-EPITHET pooled
871–875. data set. Stroke 2011;42:1270–1275.

15. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas 25. Kudo K, Sasaki M, Yamada K, et al. Differences in CT perfusion maps
under two or more correlated receiver operating characteristic cur- generated by different commercial software: quantitative analysis by
ves: a nonparametric approach. Biometrics [online serial] 1988;44: using identical source data of acute stroke patients. Radiology 2010;
837–845. 254:200–209.

26. Rava RA, Snyder KV, Mokin M, et al. Assessment of computed

16. Albers GW, Goyal M, Jahan R, et al. Ischemic core and hypo-
tomography perfusion software in predicting spatial location and vol-
perfusion volumes predict infarct size in SWIFT PRIME. Ann Neurol
ume of infarct in acute ischemic stroke patients: a comparison of
2016;79:76–89.
sphere, Vitrea, and RAPID. J Neurointerv Surg 2021;13:130–135.
17. Rao V, Christensen S, Yennu A, et al. Ischemic core and hypo-
27. Campbell BCV, Majoie CBLM, Albers GW, et al. Penumbral imaging
perfusion volumes correlate with infarct size 24 hours after randomi-
and functional outcome in patients with anterior circulation
zation in DEFUSE 3. Stroke 2019;50:626–631. ischaemic stroke treated with endovascular thrombectomy versus
18. Albers GW, Goyal M, Jahan R, et al. Relationships between imaging medical therapy: a meta-analysis of individual patient-level data. Lan-
assessments and outcomes in solitaire with the intention for cet Neurol 2019;18:46–55.
Thrombectomy as primary endovascular treatment for acute ischemic 28. Saver JL. Optimal end points for acute stroke therapy trials: best
stroke. Stroke 2015;46:2786–2794. ways to measure treatment effects of drugs and devices. Stroke
2011;42:2356–2362.
19. Sarraj A, Hassan AE, Grotta J, et al. Optimizing patient selection for
endovascular treatment in acute ischemic stroke (SELECT): a pro- 29. Shafie M, Yu W. Recanalization therapy for acute ischemic stroke
spective, multicenter cohort study of imaging selection. Ann Neurol with large vessel occlusion: where we are and what comes next?
2020;87:419–433. Transl Stroke Res 2021;12:369–381.

888 Volume 91, No. 6