Abstract
Introduction:
Observational studies have found an increased risk of hemorrhagic transformation and worse functional outcomes in patients with higher systolic blood pressure variability (BPV). However, the time-varying behavior of BPV after endovascular thrombectomy (EVT) and its effects on functional outcome have not been well characterized.
Patients and methods:
We analyzed data from an international cohort of patients with large-vessel occlusion stroke who underwent EVT at 11 centers across North America, Europe, and Asia. Repeated time-stamped blood pressure data were recorded for the first 72 h after thrombectomy. Parameters of BPV were calculated in 12-h epochs using five established methodologies. Systolic BPV trajectories were generated using group-based trajectory modeling, which separates heterogeneous longitudinal data into groups with similar patterns.
Results:
Of the 2041 patients (age 69 ± 14, 51.4% male, NIHSS 15 ± 7, mean number of BP measurements 50 ± 28) included in our analysis, 1293 (63.4%) had a poor 90-day outcome (mRS ⩾ 3) or a poor discharge outcome (mRS ⩾ 3). We identified three distinct SBP trajectories: low (25%), moderate (64%), and high (11%). Compared to patients with low BPV, those in the highest trajectory group had a significantly greater risk of a poor functional outcome after adjusting for relevant confounders (OR 2.2; 95% CI 1.2–3.9; p = 0.008). In addition, patients with poor outcomes had significantly higher systolic BPV during the epochs that define the first 24 h after EVT (p < 0.001).
Discussion and conclusions:
Acute ischemic stroke patients demonstrate three unique systolic BPV trajectories that differ in their association with functional outcome. Further research is needed to rapidly identify individuals with high-risk BPV trajectories and to develop treatment strategies for targeting high BPV.
Keywords: Stroke, blood pressure variability, thrombectomy, brain ischemia
Subject terms Cerebrovascular disease/stroke, ischemic stroke, revascularization, blood pressure
Introduction
Endovascular thrombectomy (EVT) has become the standard of care for patients affected by large-vessel occlusion ischemic stroke. About half of patients do not achieve functional independence despite EVT, and it is critical to identify treatable factors that may reduce secondary brain injury and promote penumbral recovery. 1 The optimal hemodynamic management is an important aspect of peri-procedural care that remains unresolved.
The loss of cerebral autoregulation in the acute phase of ischemic stroke leaves patients vulnerable to fluctuations in systemic blood pressure (BP). 2 Previous observational studies have found an increased risk of hemorrhagic transformation, worse functional outcomes, and increased mortality in patients with high blood pressure variability (BPV) during the first 24 h after endovascular thrombectomy (EVT)3,4 Thus, the postprocedural reduction of BPV in EVT-treated patients appears to be a plausible management strategy for improving outcomes. However, the ideal candidates and the best timing for such intervention remain unknown. This study aimed to characterize the time-varying behavior of BPV during the first days after EVT and assess its effects on functional outcomes.
Methods
Study design
We analyzed individual-patient blood pressure data and clinical covariates from an international cohort of patients enrolled in a multi-center observational study. The inclusion criteria for the cohort study were patients greater than or equal to 18 years of age with acute anterior large-vessel occlusion stroke and who underwent EVT. The anonymized patient data were collected from 11 comprehensive stroke centers: eight in the United States, one in South Korea, one in Germany, and one in France. Five of the 11 centers maintained prospective stroke data registries, while the remaining centers identified and collected data retrospectively. Most patients (n = 2165, 84%) were treated between 2012 and 2019; one center (n = 236) included patients from as early as 2005. The study was approved by the Yale Institutional Review Board. Informed patient consent was waived, and data across all centers were collected under the approval of institutional Data Use Agreements. The reporting of this study conforms to the STROBE statement. 5 All stroke management decisions were made by the patients’ attending providers and the clinical care team. All data generated or analysed during this study are included in this published article (and its supplementary information files).
Study variables
All participating sites were asked to compile repeated time-stamped systolic blood pressure (SBP), diastolic blood pressure (DBP), and mean arterial pressure (MAP) data within the first 72 h post-EVT. We note that the method (intra-arterial catheter vs. non-invasive blood pressure cuff) and frequency of blood pressure data acquisition were determined by the clinical care protocol at each site. All centers also provided the following baseline characteristics and clinical information for each patient: age, sex, medical history, admission National Institutes of Health Stroke Scale (NIHSS), admission Alberta Stroke Program Early CT Score (ASPECTS), treatment with intravenous alteplase (IV tPA), site of intracranial vessel occlusion on CTA, time-to-reperfusion in minutes, and postprocedural Thrombolysis in Cerebral Infarction (TICI) score.
Study outcomes
The primary outcome assessed in this analysis was stroke-related disability measured using the modified Rankin Scale (mRS) at 90 days. We dichotomized mRS with 0–2 defining a good outcome and 3–6 defining a poor outcome. While 90-day mRS is the standard objective measure of stroke-based outcome, other measures, including the discharge destination and discharge mRS, are good individual predictors of mRS at 3 months.6,7 Thus, in patients without a recorded mRS at 90 days, discharge mRS was used instead to ultimately assess the primary outcome. In addition, we performed a sensitivity analysis of only those patients with completed 90-day mRS assessments. Outcome data were collected as part of routine care via telephone calls or in-person follow-up appointments at all sites. We excluded patients with missing discharge mRS and 90-day mRS data for the primary outcome. Secondary outcomes included mRS shift, mortality, radiographic hemorrhagic transformation (HT), and symptomatic intracranial hemorrhage (sICH). HT was evaluated on post-procedural CT imaging and graded using the European Cooperative Acute Stroke Study (ECASS) II classification. 8 SICH was defined as any hemorrhage on follow-up CT that was associated with ⩾4 points increase in the NIHSS or that led to death, and that was identified as the predominant cause of the neurologic deterioration. 9
Blood pressure variability
Each patient’s recorded blood pressure measurements were divided into 12-h epochs for up to 72 h after EVT. We calculated BPV for SBP, DBP, and MAP using five established statistical methodologies: standard deviation (SD), coefficient of variation (CV), average real variability (ARV), successive variation (SV), and residual SD (rSD). 3
Previous studies advocated for taking multiple approaches when evaluating BPV. These measures of BPV were selected because most of them consider the distribution and linear trends of blood pressure, reducing the dependency of BPV on a patient’s mean BP level. 10 To achieve consistent and accurate measurements of variability across all patients, we only calculated BPV in patients’ epochs containing more than 10 blood pressure measurements recorded during the given 12-h period. We focused on using SBP SD for BPV when performing the group-based trajectory modeling, tertile classification, and statistical methods in this analysis. MATLAB 9.1 (R2016b) was used for calculating all measures of BPV.
Group-based trajectory modeling
We employed group-based trajectory modeling (GBTM), a finite mixture model, to address some of the limitations presented when using assignment rules for patient tertiles based on an overall summary statistic. GBTM generated systolic BPV (defined using SD) trajectories that emerged from the epoch-wise BPV data itself; this statistical method separated the heterogeneous longitudinal data into clusters that followed similar progressions over time using maximum likelihood estimation.
The “traj” package in STATA (StataCorp) was used to model group-based trajectories summarized as polynomial functions of time. We fitted the epoch-wise systolic BPV data using a beta distribution, and the maximum order of the polynomial functions representing each trajectory was three (cubic). All model parameter estimates are trajectory-specific and vary freely across groups. The optimal shape and number of trajectories were determined using the Bayesian information criterion (BIC) and significance between trajectories (p < 0.05). Descriptive names for each trajectory were given based on their appearances.
We calculated the posterior probability of trajectory group membership for patients at each epoch, as individuals’ probability of membership is updated after each sequential observation and based on the total available data up to that point. Patients were ultimately assigned membership to the trajectory group for which they had the highest probability at their final epoch. These group memberships were then entered into adjusted logistic regression models to test the association of trajectory group membership with the primary outcome.11,12
Tertile classification
In addition to GBTM, we used patients’ overall BPV (defined as SBP SD over the entire recording period) to sort, divide, and assign patients into tertiles, each containing about a third of the cohort: lower tertile (low BPV), middle tertile (intermediate BPV), and upper tertile (high BPV).
Statistical analysis
Normally distributed continuous variables were summarized as mean with SD, non-normally distributed continuous variables as median with interquartile range (IQR), and categorical data as proportions. Univariate comparisons between tertile and group-based trajectory cohorts were tested using the one-way ANOVA for continuous variables, the χ2 test for binary variables, and the Kruskal-Wallis test for ordinal variables.
We examined the associations between trajectory group membership with functional outcome using binary logistic regression. We also performed an mRS shift analysis using ordinal logistic regression. Variables selected for our adjusted analyses were based on baseline covariates with a p-value < 0.05 in the univariate analyses and considerations a priori from previous studies. We adjusted for the following variables in our multivariate logistic regression models: patient age, sex, baseline NIHSS, preprocedural ASPECTS, admission MAP, time from last known normal to reperfusion (in minutes), postprocedural TICI score, and the overall number of per patient BP measurements. We calculated E-values to provide the minimum strength of association that an unmeasured confounder would need to have with both the exposure (BPV) and outcome, conditional on the measured covariates, to fully explain away the association. 13 A repeated-measures analysis of the six average BPV measurements in each epoch was conducted with a mixed-effects model (assuming autoregressive covariance structure) to assess the effect of time and trajectory group on the average BPV throughout the early post-thrombectomy period with and without adjustment for baseline NIHSS score and age. Statistical analyses were conducted using R 3.6.2 (The R Foundation), with statistical significance defined as p < 0.05.
Results
Subject characteristics
Of the 2566 EVT-treated patients enrolled in the multi-center observational study, we included 2041 patients in our analysis (mean age 69 ± 14, 51.4% male, mean NIHSS 15 ± 7). We excluded 390 patients because of missing 90-day mRS and missing discharge mRS outcome data, as well as 135 patients with insufficient blood pressure data. The baseline characteristics for the final cohort are reported in Table 1.
Table 1.
Baseline characteristics of the entire cohort and by group-based BPV (SBP SD) trajectories.
| Variable | Entire cohort (n = 2041) | Trajectory #1: “Low” (n = 512) | Trajectory #2: “Moderate” (n = 1312) | Trajectory #3: “High” (n = 217) | p-Value |
|---|---|---|---|---|---|
| Tertile, n (%) | <0.001 | ||||
| Lower | 680 (33.3) | 404 (78.9) | 276 (21.0) | 0 (0.0) | |
| Middle | 681 (33.4) | 82 (16.0) | 596 (45.4) | 3 (1.4) | |
| Upper | 680 (33.3) | 26 (5.1) | 440 (33.5) | 214 (98.6) | |
| Overall BPV (mean (SD)) | 14.61 (4.67) | 10.55 (3.04) | 14.79 (3.04) | 23.12 (3.99) | <0.001 |
| Overall BP mean (mean (SD)) | 128.92 (14.29) | 122.65 (13.71) | 130.31 (13.70) | 135.35 (14.14) | <0.001 |
| Overall number of patient BP readings (mean (SD)) | 49.99 (28.40) | 51.45 (28.05) | 49.12 (28.27) | 51.82 (29.85) | 0.175 |
| Age (mean (SD)) | 68.96 (14.44) | 63.50 (15.75) | 70.52 (13.56) | 72.44 (13.07) | <0.001 |
| Admission NIHSS (mean (SD)) | 14.88 (6.53) | 14.68 (6.51) | 14.88 (6.55) | 15.37 (6.45) | 0.424 |
| Hemorrhagic transformation, n (%) | 706 (37.6) | 180 (38.9) | 446 (36.8) | 80 (39.2) | 0.636 |
| Symptomatic ICH, n (%) | 147 (10.9) | 39 (10.6) | 99 (11.5) | 9 (7.3) | 0.358 |
| Gender, M (%) | 1047 (51.4) | 281 (55.0) | 664 (50.7) | 102 (47.0) | 0.101 |
| Admission ASPECTS (median [IQR]) | 8.00 [7.00, 9.00] | 8.00 [6.00, 9.00] | 8.00 [7.00, 9.00] | 8.00 [6.00, 9.00] | 0.109 |
| Admission SBP, mmHg (mean (SD)) | 141.88 (25.72) | 134.44 (22.07) | 142.90 (25.80) | 150.18 (28.31) | <0.001 |
| Admission DBP, mmHg (mean (SD)) | 81.88 (18.00) | 79.71 (15.53) | 82.00 (18.17) | 85.38 (20.75) | 0.011 |
| Admission MAP, mmHg (mean (SD)) | 101.71 (18.23) | 97.83 (15.94) | 102.14 (18.39) | 106.69 (20.00) | <0.001 |
| Administration of BP management medication during admission, n (%) | 500 (50.0) | 97 (42.0) | 333 (51.8) | 70 (55.6) | 0.016 |
| Race, n (%) | 0.066 | ||||
| American Indian/Alaskan Native | 1 (0.0) | 1 (0.2) | 0 (0.0) | 0 (0.0) | |
| Asian | 377 (18.5) | 102 (19.9) | 244 (18.6) | 31 (14.3) | |
| Black | 86 (4.2) | 18 (3.5) | 55 (4.2) | 13 (6.0) | |
| Hispanic | 1 (0.0) | 0 (0.0) | 1 (0.1) | 0 (0.0) | |
| Pacific-Islander | 1 (0.0) | 0 (0.0) | 1 (0.1) | 0 (0.0) | |
| White | 1184 (58.0) | 270 (52.7) | 777 (59.2) | 137 (63.1) | |
| Medical History, n (%) | |||||
| HTN | 1347 (66.4) | 271 (53.3) | 907 (69.5) | 169 (78.6) | <0.001 |
| Antihypertensive medication use | 247 (42.4) | 37 (31.1) | 181 (45.5) | 29 (44.6) | 0.019 |
| CAD/MI | 283 (16.5) | 53 (12.9) | 189 (16.8) | 41 (22.3) | 0.014 |
| HLD | 733 (36.3) | 174 (34.3) | 488 (37.6) | 71 (33.8) | 0.304 |
| CHF | 136 (12.4) | 37 (13.1) | 87 (12.2) | 12 (11.3) | 0.872 |
| Afib | 754 (37.4) | 149 (29.6) | 518 (39.9) | 87 (40.8) | <0.001 |
| DM | 496 (24.5) | 107 (21.0) | 325 (25.0) | 64 (29.8) | 0.035 |
| Past AIS | 233 (15.8) | 51 (13.3) | 160 (16.8) | 22 (16.1) | 0.296 |
| Smoking | 603 (31.9) | 140 (29.9) | 391 (32.0) | 72 (36.5) | 0.246 |
| Occlusion on CTA, n (%) | |||||
| ICA/tICA | 469 (25.7) | 111 (25.5) | 302 (25.2) | 56 (29.0) | 0.532 |
| ACA | 35 (2.0) | 11 (2.7) | 20 (1.8) | 4 (2.2) | 0.499 |
| M1 MCA | 1079 (59.1) | 273 (62.6) | 702 (58.6) | 104 (53.9) | 0.105 |
| M2 MCA | 370 (20.3) | 82 (18.8) | 247 (20.6) | 41 (21.2) | 0.674 |
| Core Infarct Volume, mL (median [IQR]) | 5.50 [0.00, 22.75] | 10.00 [0.00, 21.50] | 5.00 [0.00, 23.00] | 5.00 [0.00, 23.75] | 0.685 |
| Treated with IV-tPA, n (%) | 955 (52.0) | 252 (54.4) | 604 (51.6) | 99 (48.5) | 0.342 |
| Onset to IV-tPA bolus, min (mean (SD)) | 310.38 (325.06) | 284.42 (318.12) | 303.86 (293.34) | 438.28 (492.28) | 0.006 |
| Onset to admission, min (mean (SD)) | 154.82 (162.90) | 156.55 (208.62) | 155.78 (146.03) | 142.07 (62.71) | 0.91 |
| Onset to EVT (mean (SD)) | 355.36 (541.76) | 322.02 (288.90) | 364.67 (623.55) | 373.99 (417.56) | 0.359 |
| Onset to reperfusion, min (mean (SD)) | 438.87 (547.03) | 417.75 (396.48) | 441.82 (606.41) | 471.96 (453.02) | 0.498 |
| TICI, n (%) | 0.001 | ||||
| 0 | 147 (7.3) | 29 (5.7) | 87 (6.7) | 31 (14.6) | |
| 1 | 66 (3.3) | 20 (3.9) | 38 (2.9) | 8 (3.8) | |
| 2A | 148 (7.3) | 33 (6.5) | 93 (7.1) | 22 (10.3) | |
| 2B | 903 (44.5) | 221 (43.4) | 598 (45.8) | 84 (39.4) | |
| 3 | 763 (37.6) | 206 (40.5) | 489 (37.5) | 68 (31.9) | |
| Good 3-month outcome (mRS 0–2), n (%) | 667 (39.7) | 187 (47.9) | 430 (38.7) | 50 (28.1) | <0.001 |
| Recanalization (TICI 2b–3), n (%) | 1666 (82.2) | 427 (83.9) | 1087 (83.3) | 152 (71.4) | <0.001 |
| Dead at 3-month outcome (mRS = 6), n (%) | 344 (20.5) | 60 (15.4) | 229 (20.6) | 55 (30.9) | <0.001 |
ACA: anterior cerebral artery; Afib: atrial fibrillation; AIS: acute ischemic stroke; CAD: coronary artery disease; CHF: congestive heart failure; CTA: computed tomography angiography; DM: diabetes mellitus; HLD: hyperlipidemia; HTN: hypertension; ICA/tICA: internal carotid artery/terminal internal carotid artery; MCA: middle cerebral artery; MI: myocardial infarction.
Blood pressure variability
The analysis included a total of 102,030 time-stamped blood pressure measurements. The mean number of recorded BP measurements per patient was 50 ± 28. In each epoch, the average number of BP measurements was 19 ± 11 for 0–12 h, 15 ± 8 for 12–24 h, 14 ± 7 for 24–36 h, 14 ± 6 for 36–48 h, 14 ± 6 for 48–60 h, and 14 ± 6 for 60–72 h.
Trajectories modeling and blood pressure variability tertiles
We implemented GBTM because capturing BPV trend information over time was not possible using the tertile classification approach. As detailed above, GBTM predicts (1) the trajectory of each epoch-wise BPV group, (2) the form of each trajectory, (3) the probability for each patient being assigned membership to a particular group, and (4) the group for which they have the highest probability, irrespective of their relation to the primary outcome.
We found that three distinct systolic BPV trajectories (p < 0.05) best fit the data using GBTM: low, moderate, and high (Figure 1(a)). The mean probability of final trajectory membership for patients was 75%. The “low” trajectory included 25% of the patients, the “moderate” trajectory had the largest percentage of patients at 64%, and patients in the persistently “high” trajectory comprised 11% of patients. BPV was significantly increased in the early phase after thrombectomy (the first 12 h for patients in the low and moderate trajectory group and 24 h for patients in the high trajectory group) and then stabilized for the remainder of the monitoring period (Figure 1(a)). Differences in patient characteristics by trajectory group were observed for age, hypertension, atrial fibrillation, coronary artery disease/myocardial infarction, diabetes mellitus, admission MAP, administration of antihypertensive medication during admission, and degree of recanalization status (Table 1).
Figure 1.
(a–c) Time series of epoch-wise BPV (SBP SD) in EVT-treated patients. Three unique systolic BPV trajectories using GBTM were identified over the first 72 h after EVT. The timepoint at zero corresponds to recanalization time or, for non-recanalized subjects, the end of EVT (a). Overall, BPV (defined as SBP SD over the entire recording period) was used to divide patients into tertiles, each containing about a third of the cohort (b). Time series of epoch-wise systolic BPV in patients with good and poor outcomes, using 90-day mRS or then discharge mRS (c).
We also divided the cohort into distinct systolic BPV tertiles (p < 0.05), each containing a third of the cohort: upper, middle, and lower (Figure 1(b)). Demographics and baseline characteristics by tertile are shown in Supplemental Table 1. Differences were observed in the distribution of age, sex, race, hypertension, atrial fibrillation, diabetes mellitus, admission MAP, administration of IV tPA bolus and antihypertensive medication during admission, and recanalization status. The mean values for overall SBP and systolic BPV parameters by tertile are shown in Supplemental Table 2.
Blood pressure variability and functional outcome
Seven hundred forty-eight patients (36.6%) had a favorable 90-day outcome or favorable discharge outcome if 3-month data was not available (mRS ⩽ 2). Patients with good outcome were significantly younger (64 vs 72 years; p < 0.001), had lower admission MAP (99 vs 103 mmHg; p < 0.001), more frequently received IV tPA (57% vs 49%; p = 0.003), and had a lower admission NIHSS (12 vs 16; p < 0.001). They were also less likely to have comorbidities, including hypertension, atrial fibrillation, chronic heart failure, diabetes mellitus, and a previous acute ischemic stroke. Lastly, the distribution of admission ASPECTS (p < 0.001), core infarct volume (0.009), time to reperfusion (p = 0.002), and degree of reperfusion (p < 0.001) differed between patients with good and poor outcome.
All measures of overall systolic BPV were significantly higher in patients with poor versus good outcomes (Table 2). Figure 1(c) shows a time series of epoch-wise systolic BPV (SBP SD) in patients with good and poor outcomes. After adjusting for age and admission NIHSS, BPV was significantly higher during the epochs that define the first 24 h after EVT among patients with unfavorable outcome (mean difference in SBP SD 1.8 mmHg, p < 0.001, and 1.1 mmHg, p < 0.001 for epochs 1 and 2, respectively). Compared to patients in the “low” trajectory, those in the “high” trajectory had significantly greater odds of a poor functional outcome after adjusting for age, sex, admission NIHSS, admission ASPECTS, admission MAP, onset to reperfusion (in minutes), postprocedural TICI score, and the overall number of BP measurements (odds ratio (OR) 2.2; 95% confidence interval (CI) 1.2–3.9; p = 0.008; Figure 2(a)). The sensitivity analysis of patients with 90-day outcomes (n = 1678), showed a similar association and effect size after adjusting for the same confounders (OR 2.1; (95% CI 1.2–3.9), Supplemental Table 3). In addition, the same direction of effect, albeit a weaker association, was observed when dividing patients into tertiles of BPV (OR 1.40 for upper vs lower tertile; 95% CI 0.97–2.00; p = 0.070; Figure 2(b)).
Table 2.
BPV and BP mean compared between EVT-treated patients with good (mRS, 0 and 2) versus poor outcome (mRS, 3–6) at 3 months (or discharge if 90-day mRS was not available).
| Variable (mean (SD)) | Entire cohort (n = 2041) | Good outcome (mRS 0–2; n = 748) | Poor outcome (mRS 3–6; n = 1293) | p-Value |
|---|---|---|---|---|
| Overall SBP mean | 128.92 (14.29) | 125.24 (13.26) | 131.05 (14.44) | <0.001 |
| Overall number of SBP measurements | 49.99 (28.40) | 47.71 (27.01) | 51.30 (29.10) | 0.006 |
| BPV = SBP SD | ||||
| Overall | 14.61 (4.67) | 13.51 (4.20) | 15.25 (4.81) | <0.001 |
| Epoch #1: 0–12 h | 13.36 (5.37) | 12.21 (4.78) | 14.01 (5.57) | <0.001 |
| Epoch #2: 12–24 h | 11.91 (4.70) | 11.21 (4.34) | 12.31 (4.86) | <0.001 |
| Epoch #3: 24–36 h | 11.39 (4.60) | 11.25 (4.69) | 11.46 (4.56) | 0.583 |
| Epoch #4: 36–48 h | 11.35 (4.70) | 10.93 (3.99) | 11.53 (4.96) | 0.164 |
| Epoch #5: 48–60 h | 11.30 (4.63) | 10.96 (3.92) | 11.40 (4.83) | 0.391 |
| Epoch #6: 60–72 h | 11.38 (4.58) | 11.35 (4.55) | 11.39 (4.60) | 0.95 |
| BPV = SBP CV | ||||
| Overall | 11.34 (3.46) | 10.77 (3.05) | 11.68 (3.63) | <0.001 |
| Epoch #1: 0–12 h | 10.44 (4.05) | 9.82 (3.65) | 10.79 (4.22) | <0.001 |
| Epoch #2: 12–24 h | 9.28 (3.54) | 9.03 (3.26) | 9.42 (3.68) | 0.026 |
| Epoch #3: 24–36 h | 8.90 (3.66) | 8.95 (3.43) | 8.88 (3.76) | 0.827 |
| Epoch #4: 36–48 h | 8.79 (3.67) | 8.66 (3.24) | 8.84 (3.84) | 0.601 |
| Epoch #5: 48–60 h | 8.70 (3.60) | 8.66 (2.87) | 8.71 (3.81) | 0.909 |
| Epoch #6: 60–72 h | 8.72 (3.40) | 9.01 (3.72) | 8.63 (3.30) | 0.356 |
| BPV = SBP SV | ||||
| Overall | 15.03 (4.93) | 14.06 (4.47) | 15.59 (5.10) | <0.001 |
| Epoch #1: 0–12 h | 14.66 (6.18) | 13.36 (5.40) | 15.41 (6.48) | <0.001 |
| Epoch #2: 12–24 h | 14.03 (5.89) | 13.29 (5.33) | 14.46 (6.15) | <0.001 |
| Epoch #3: 24–36 h | 13.70 (5.82) | 13.77 (5.99) | 13.67 (5.75) | 0.842 |
| Epoch #4: 36–48 h | 13.47 (5.62) | 13.10 (5.39) | 13.63 (5.71) | 0.307 |
| Epoch #5: 48–60 h | 13.47 (5.74) | 13.07 (4.82) | 13.60 (6.00) | 0.408 |
| Epoch #6: 60–72 h | 13.52 (5.86) | 13.08 (5.65) | 13.66 (5.93) | 0.419 |
| BPV = SBP ARV | ||||
| Overall | 11.42 (3.75) | 10.73 (3.38) | 11.82 (3.89) | <0.001 |
| Epoch #1: 0–12 h | 11.35 (4.78) | 10.36 (4.09) | 11.92 (5.05) | <0.001 |
| Epoch #2: 12–24 h | 11.12 (4.67) | 10.54 (4.31) | 11.45 (4.83) | <0.001 |
| Epoch #3: 24–36 h | 10.90 (4.61) | 10.90 (4.72) | 10.90 (4.57) | 0.983 |
| Epoch #4: 36–48 h | 10.76 (4.51) | 10.53 (4.34) | 10.86 (4.58) | 0.419 |
| Epoch #5: 48–60 h | 10.60 (4.39) | 10.34 (4.02) | 10.68 (4.50) | 0.488 |
| Epoch #6: 60–72 h | 10.73 (4.68) | 10.45 (4.67) | 10.81 (4.69) | 0.53 |
| BPV = SBP rSD | ||||
| Overall | 13.45 (4.18) | 12.46 (3.71) | 14.02 (4.33) | <0.001 |
| Epoch #1: 0–12 h | 12.09 (4.89) | 10.96 (4.35) | 12.73 (5.07) | <0.001 |
| Epoch #2: 12–24 h | 10.98 (4.42) | 10.41 (4.02) | 11.30 (4.60) | <0.001 |
| Epoch #3: 24–36 h | 10.62 (4.36) | 10.46 (4.40) | 10.69 (4.34) | 0.524 |
| Epoch #4: 36–48 h | 10.58 (4.52) | 10.12 (3.95) | 10.77 (4.73) | 0.111 |
| Epoch #5: 48–60 h | 10.44 (4.47) | 9.98 (3.53) | 10.59 (4.73) | 0.216 |
| Epoch #6: 60–72 h | 10.54 (4.30) | 10.59 (4.42) | 10.52 (4.27) | 0.901 |
Figure 2.
(a–d) Adjusted odds ratio of clinical and radiographic outcomes by BPV (SBP SD) trajectories. Adjusted odds ratios (midpoints) with 95% confidence interval (error bars) from the analysis of functional and radiographic outcomes by post-procedural 72-h BPV trajectory using SBP SD: (a) odds of poor functional outcome at 90 days or then at discharge, (b) odds of dead at 90 days, (c) odds of developing HT, and (d) odds of developing sICH. The double asterisk notation (**) above the respective error bars represents a statistically significant difference between the indicated group and the reference group. The “low” trajectory and lower tertile serve as the reference groups.
The distribution of mRS scores from patients in each trajectory group is shown in Figure 3. A greater proportion of patients in the “high” trajectory group had poor outcome than patients in the “moderate” and “low” trajectories. Moreover, patients in the “high” trajectory group had significantly more global disability, as indicated by an unfavorable shift in the distribution of mRS scores, than did patients in the “low” trajectory group (adjusted OR for a shift toward worse outcome 2.3; 95% CI 1.5–3.5; p < 0.001).
Figure 3.
Outcome of mRS distribution by group-based BPV (SBP SD) trajectory. The distribution of the modified Rankin Scale (mRS) at 90 days or then at discharge by systolic BPV trajectory group. Higher mRS scores and darker shades correspond to more unfavorable functional outcomes. The dashed lines demarcate good (mRS 0–2) versus poor (mRS 3–6) outcome classifications. Higher rates of poor outcome are observed in higher group-based BPV trajectory groups.
Blood pressure variability and hemorrhagic transformation
A hemorrhagic transformation occurred in 37.6% of patients (706/1879). No difference in the rate of HT was observed across trajectory groups in Figure 2(c) (p = 0.64, Table 3). The distribution of ECASS scores by BPV trajectory group is shown in Supplemental Figure 1. Symptomatic intracerebral hemorrhage was observed in 10.9% (147/1352). While there appears to be a trend toward an increased risk of sICH in the higher BPV trajectory groups (Figure 2(d)), the association was not statistically significant (p = 0.4, Table 3).
Table 3.
Group-based BPV (SBP SD) trajectories and clinical and radiographic outcomes.*
| Trajectory #1 (“Low”) | Trajectory #2 (“Moderate”) | Trajectory #3 (“High”) | |
|---|---|---|---|
| Outcome = 90-day mRS or then discharge mRS (binary, unfavorable outcome) | |||
| Participants | 512 | 1312 | 217 |
| Events | 288 | 840 | 165 |
| Event rate (%) | 56.2 | 64.0 | 76.0 |
| Group difference p-value | <0.001 | ||
| Original aOR ± 95% CI | 1.00 (ref) | 1.25 (0.88–1.78) | 2.19 (1.24–3.94) |
| Original p-value | 0.2090 | 0.0075 | |
| E-value for point estimate | 1.52 | 2.38 | |
| E-value for confidence interval | 1.00 | 1.73 | |
| Outcome = 90-day mRS or then discharge mRS (ordinal) | |||
| Participants | 512 | 1312 | 217 |
| Group difference p-value | <0.001 | ||
| Mean mRS (mean (SD)) | 2.92 (1.93) | 3.27 (1.94) | 4.01 (1.87) |
| Original aOR ± 95% CI | 1.00 (ref) | 1.28 (0.96–1.69) | 2.27 (1.48–3.49) |
| Original p-value | 0.0876 | 0.0002 | |
| E-value for point estimate | 1.52 | 2.38 | |
| E-value for confidence interval | 1.00 | 1.73 | |
| Outcome = Hemorrhagic Transformation (binary) | |||
| Participants | 463 | 1212 | 204 |
| Events | 180 | 446 | 80 |
| Event rate (%) | 38.9 | 36.8 | 39.2 |
| Group difference p-value | 0.636 | ||
| Original aOR ± 95% CI | 1.00 (ref) | 1.37 (0.97–1.95) | 1.44 (0.87–2.38) |
| Original p-value | 0.0756 | 0.1597 | |
| E-value for point estimate | 1.62 | 1.69 | |
| E-value for confidence interval | 1.00 | 1.00 | |
| Outcome = Symptomatic ICH (binary) | |||
| Participants | 368 | 861 | 123 |
| Events | 39 | 99 | 9 |
| Event rate (%) | 10.6 | 11.5 | 7.3 |
| Group difference p-value | 0.358 | ||
| Original aOR ± 95% CI | 1.00 (ref) | 1.22 (0.63–2.48) | 0.92 (0.27–2.70) |
| Original p-value | 0.5708 | 0.8847 | |
| E-value for point estimate | 1.74 | 1.39 | |
| E-value for confidence interval | 1.00 | 1.00 | |
| Outcome = 90-day Mortality (binary) | |||
| Participants | 390 | 1112 | 178 |
| Events | 60 | 229 | 55 |
| Event rate (%) | 15.4 | 20.6 | 30.9 |
| Group difference p-value | <0.001 | ||
| Original aOR ± 95% CI | 1.00 (ref) | 1.33 (0.80–2.28) | 2.18 (1.10–4.33) |
| Original p-value | 0.2866 | 0.0249 | |
| E-value for point estimate | 1.57 | 2.32 | |
| E-value for confidence interval | 1.00 | 1.28 | |
Analyses were adjusted for age, sex, admission NIHSS, admission ASPECTS, admission MAP, onset to reperfusion (in minutes), postprocedural TICI score, and the overall number of BP measurements.
Discussion
This analysis investigated the time-varying behavior of systolic BPV after EVT. We implemented a novel, data-driven approach to capture the temporal profiles of systolic BPV after EVT and segment our cohort into meaningful subgroups that follow similar trends over time. We demonstrated that patients can be separated into three distinct trajectories that differ in their relationship with functional outcome.
These results build on existing evidence showing a consistent association between high systolic BPV and unfavorable functional outcome.3,14–16 However, previous studies have not used longitudinal data methods to characterize changes in BPV over time and identify high-risk patients. Our analysis of individual patient BPV trajectories suggests the existence of a group of patients with persistently elevated BPV that was highest in the first 24 h after thrombectomy. Adjusting for significant confounders, these patients were about twice as likely to have severe disability or death at 90-day follow-up than those in the lowest trajectory group. This analysis supports the theory that increased BPV after stroke is harmful to the injured brain and impairs penumbral recovery. Under normal circumstances, the cerebral vasculature maintains a relatively constant cerebral blood flow across a wide range of blood pressures, a mechanism known as cerebral autoregulation. 17 This mechanism ensures that the cerebral blood flow matches the brain’s metabolic demands and protects it from hypo- or hyperperfusion. However, research indicates that the brain’s ability to autoregulate is impaired after stroke, leading to pressure passivity. 2 As a result, changes in systemic blood pressure are directly transmitted to the brain exposing the penumbra to deleterious fluctuations in systemic blood pressure. Thus, patients with the highest BPV after EVT appear most vulnerable. In a secondary analysis of the BEST study, a prospective multicenter cohort study of BP after thrombectomy, the investigators divided patients into tertiles based on their average BPV in the first 24 h after thrombectomy. 3 Adjusting for relevant confounders, the highest tertile of systolic BPV consistently predicted poor outcome (mRS 3–6). Given the relative stability of BPV trajectories throughout the early phase after large-vessel occlusion (LVO) stroke, a division into tertiles may provide a simple and clinically meaningful classification of patients into high- and low-risk groups.
There are several possible mechanisms through which high post-stroke BPV could increase the probability of a poor outcome. Prior studies have shown that cerebral ischemia triggers a robust inflammatory response, leading to a breakdown of the blood-brain barrier and an increased risk of reperfusion injury. 18 Theoretically, repeated surges in BP may exacerbate the risk of hemorrhagic transformation by promoting the rupture of blood vessels which have already been damaged by the ischemic insult. Two recent studies showed an increased risk of symptomatic intracerebral hemorrhage among patients with increased systolic BPV after successful mechanical thrombectomy.4,15 However, we did not find a significant association between BPV trajectory group and HT or sICH, which is consistent with other studies.3,14,16,19 While differences in methodology and heterogeneity of patients (e.g. reperfusion status) may have contributed to the conflicting results, our findings suggest that sICH and HT are not the primary mediators of the robust relationship between high BPV after stroke and unfavorable outcome. Aside from HT, BPV may lead to episodes of cerebral hypoperfusion with a risk of infarct progression, particularly among patients with incomplete reperfusion and cerebral microcirculatory alterations. In a study of patients with documented MCA occlusion, high BPV was independently associated with DWI lesion growth among patients with unsuccessful recanalization. 20 Interestingly, we observed significant differences in the TICI score across trajectory groups with the lowest rates of successful reperfusion (TICI 2b & 3) in the high BPV trajectory group. Chang et al. found a similar inverse relationship between BPV parameters and the degree of recanalization. 21 While reperfusion status itself is a significant predictor of functional outcome and, as such, was adjusted for in our regression models, its effect on the relationship between BPV and outcome remains complex and requires further definition. Several studies showed that the association between BPV and unfavorable outcomes appeared strongest among patients with incomplete recanalization. 14 However, others found the opposite effect.3,15,16 Increased blood pressure fluctuations can lead to secondary brain injury in patients with and without successful reperfusion through different pathophysiologic mechanisms.
Despite a plausible biological mechanism and consistent observational data, the association between increased BPV and worse outcome after ischemic stroke is not widely accepted as causal.
It has been hypothesized that injury to structures of the central autonomic network could result in sympathetic overactivity and higher BPV. However, a recent study found no association between infarct location and BPV. 22 Our study cannot definitively answer whether elevated BPV after thrombectomy is merely a marker of disease severity or if it contributes to the development of poor outcomes and thus presents a target for therapeutic intervention. There has not yet been a randomized controlled trial to therapeutically reduce BPV. Whether such lowering improves outcomes would need to be tested in a prospective clinical trial. Our findings may help characterize ideal candidates for future trials and define optimal timing for such intervention.
Our study has several important limitations. First, detecting subgroups in the cohort using GBTM may not be consistent with the underlying “true” population subgroups because of non-normality and sample fluctuations. Thus, a level of uncertainty should always be considered with the trajectory group assignment. Furthermore, groups should not be regarded as concrete because individuals do not belong to trajectory groups and, instead, are assigned to the group for which they have the highest probability of membership. 12 Second, as with all observational studies, the current study is subject to residual confounding, and missing data may bias the results in unpredictable ways. For example, we did not collect data on anesthetic management (conscious sedation vs. general anesthesia), which could have influenced BP and BPV beyond the endovascular procedure. Furthermore, our dataset does not specify antihypertensive drug class or medication administration timing. However, the observed odds ratio of 2.19 for poor outcome could only be explained away by an unmeasured confounder associated with both the exposure and the outcome by an odds ratio of 2.3-fold each, above and beyond the measured confounders. Weaker confounding could not do so. Third, 3-month functional outcomes were available for only 70% (1791/2566) of patients, and missing 3 months mRS was substituted with discharge mRS in 250 patients. Although patients with and without 3-months outcomes differed in several baseline characteristics (Supplemental Table 4), both groups showed similar BPV trajectories (Supplemental Figure 2), and a sensitivity analysis including only patients with available 3-months outcome showed a similar effect size between BPV trajectory group and outcome. Finally, while our large sample size and international cohort allowed for accurate trajectory generation and increased our finding’s generalizability, it also brings additional limitations. There are institutional variations in criteria for selecting patients for EVT, post-procedural BP recording and management, as well as post-procedural imaging protocols. Although non-invasive BP monitoring was the preferred modality of blood pressure measurement, some patients may have had invasive blood pressure recordings. This could have introduced a bias related to under damping. 23
In addition to demonstrating the reproducibility of our findings in other LVO stroke datasets, future research should focus on the early identification of patients with high-risk trajectories. Our study calculated BPV trajectories retrospectively from serial BP measurements collected during the early post-EVT period. Since trajectory group membership is not available until the data has been analyzed, our method cannot provide clinicians with real-time information to guide clinical decision-making. Thus, before we can test interventions to lower BPV, we need better tools to identify ideal candidates. One promising approach is the rapid assessment of BPV using spectral analysis of short-duration continuous BP recordings. 22 Machine learning models considering historical or intraprocedural hemodynamic data as well as clinical characteristics is another promising approach.
Conclusions
This multi-institutional cohort study of acute ischemic stroke patients provides a longitudinal characterization of systolic BPV during the first 72 h after EVT. Patients demonstrated three unique systolic BPV trajectories using GBTM that differed in their association with functional outcome. Further research is needed to identify individuals with high-risk BPV trajectories and develop treatment strategies for targeting high BPV.
Supplemental Material
Supplemental material, sj-pdf-1-eso-10.1177_23969873221106907 for Temporal profiles of systolic blood pressure variability and neurologic outcomes after endovascular thrombectomy by Ayush Prasad, Jessica Kobsa, Sreeja Kodali, David Bartolome, Liza Begunova, Darko Quispe-Orozco, Mudassir Farooqui, Cynthia Zevallos, Santiago Ortega-Gutiérrez, Mohammad Anadani, Eyad Almallouhi, Alejandro M Spiotta, James A Giles, Salah G Keyrouz, Joon-Tae Kim, Ilko L Maier, Jan Liman, Marios-Nikos Psychogios, Nolwenn Riou-Comte, Sébastien Richard, Benjamin Gory, Stacey Quintero Wolfe, Patrick A Brown, Kyle M Fargen, Eva A Mistry, Hiba Fakhri, Akshitkumar Mistry, Ka-Ho Wong, Fábio A Nascimento, Peter Kan, Adam de Havenon, Kevin N Sheth and Nils H Petersen in European Stroke Journal
Acknowledgments
We would like to acknowledge all participating investigators for sharing their data.
Footnotes
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Dr. Spiotta reports grants and personal fees from Penumbra, grants and personal fees from Stryker, grants from Medtronic, non-financial support from RAPID, personal fees from Terumo, and personal fees from Cerenovus outside the submitted work. Dr. Liman reports personal fees from Stryker outside the submitted work. Dr. de Havenon reports grants from AMAG and grants from Regeneron outside the submitted work. Dr. Ortega-Gutierrez reports grants from Stryker, IschemiaView, Viz.ai, and Siemens; personal fees from Medtronic and personal fees from Stryker outside the submitted work. Dr. Sheth reports grants from NIH, grants from AHA, grants from Hyperfine, grants from Biogen, grants from Zoll, and other support from Alva outside the submitted work.
Funding: The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the NINDS [K23NS110980].
Informed consent: Informed patient consent was waived, and data across all centers were collected under the approval of institutional Data Use Agreements.
Ethical approval: The Institutional Review Board of Yale University approved this study (IRB number: 2000025116).
Guarantor: NP
Contributorship: NP researched literature and conceived the study. AP, JK, and SK were involved in protocol development, gaining ethical approval, data collection, and analysis. AP and NP wrote the first draft of the manuscript. All authors reviewed and edited the manuscript and approved the final version of the manuscript
ORCID iDs: Sreeja Kodali 
https://orcid.org/0000-0002-7050-5165
Liza Begunova
https://orcid.org/0000-0003-0222-2347
Darko Quispe-Orozco
https://orcid.org/0000-0003-3627-5439
Mohammad Anadani
https://orcid.org/0000-0002-7813-2949
Joon-Tae Kim
https://orcid.org/0000-0003-4028-8339
Ilko L Maier
https://orcid.org/0000-0001-6988-8878
Jan Liman
https://orcid.org/0000-0002-7465-9655
Nolwenn Riou-Comte
https://orcid.org/0000-0002-9276-1581
Adam de Havenon
https://orcid.org/0000-0001-8178-8597
Nils H Petersen
https://orcid.org/0000-0001-9711-3340
Supplemental material: Supplemental material for this article is available online.
References
- 1. Goyal M, Menon BK, van Zwam WH, et al. Endovascular thrombectomy after large-vessel ischaemic stroke: a meta-analysis of individual patient data from five randomised trials. Lancet 2016; 387: 1723–1731. [DOI] [PubMed] [Google Scholar]
- 2. Petersen NH, Ortega-Gutierrez S, Reccius A, et al. Dynamic cerebral autoregulation is transiently impaired for one week after large-vessel acute ischemic stroke. Cerebrovasc Dis 2015; 39: 144–150. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Mistry EA, Mehta T, Mistry A, et al. Blood pressure variability and neurologic outcome after endovascular thrombectomy: a secondary analysis of the BEST study. Stroke 2020; 51: 511–518. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Kim TJ, Park H-K, Kim J-M, et al. Blood pressure variability and hemorrhagic transformation in patients with successful recanalization after endovascular recanalization therapy: A retrospective observational study. Ann Neurol 2019; 85: 574–581. [DOI] [PubMed] [Google Scholar]
- 5. von Elm E, Altman DG, Egger M, et al. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 2007; 370: 1453–1457. [DOI] [PubMed] [Google Scholar]
- 6. ElHabr AK, Katz JM, Wang J, et al. Predicting 90-day modified Rankin Scale score with discharge information in acute ischaemic stroke patients following treatment. BMJ Neurol Open 2021; 3: e000177. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Qureshi AI, Chaudhry SA, Sapkota BL, et al. Discharge destination as a surrogate for modified Rankin scale defined outcomes at 3- and 12-months poststroke among stroke survivors. Arch Phys Med Rehabil 2012; 93: 1408–1413.e1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Fiorelli M, Bastianello S, von Kummer R, et al. Hemorrhagic transformation within 36 hours of a cerebral infarct: relationships with early clinical deterioration and 3-month outcome in the European Cooperative Acute Stroke Study I (ECASS I) cohort. Stroke 1999; 30: 2280–2284. [DOI] [PubMed] [Google Scholar]
- 9. Hacke W, Kaste M, Bluhmki E, et al. Thrombolysis with alteplase 3 to 4.5 hours after acute ischemic stroke. N Engl J Med 2008; 359: 1317–1329. [DOI] [PubMed] [Google Scholar]
- 10. Manning LS, Rothwell PM, Potter JF, et al. Prognostic significance of short-term blood pressure variability in acute stroke: systematic review. Stroke 2015; 46: 2482–2490. [DOI] [PubMed] [Google Scholar]
- 11. Elmer J, Jones BL, Zadorozhny VI, et al. A novel methodological framework for multimodality, trajectory model-based prognostication. Resuscitation 2019; 137: 197–204. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Nagin DS, Odgers CL. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol 2010; 6: 109–138. [DOI] [PubMed] [Google Scholar]
- 13. VanderWeele TJ, Ding P. Sensitivity analysis in observational research: introducing the E-value. Ann Intern Med 2017; 167: 268–274. [DOI] [PubMed] [Google Scholar]
- 14. Bennett AE, Wilder MJ, McNally JS, et al. Increased blood pressure variability after endovascular thrombectomy for acute stroke is associated with worse clinical outcome. J Neurointerv Surg 2018; 10: 823–827. [DOI] [PubMed] [Google Scholar]
- 15. Huang X, Guo H, Yuan L, et al. Blood pressure variability and outcomes after mechanical thrombectomy based on the recanalization and collateral status. Ther Adv Neurol Disord 2021; 14: 1756286421997383. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Cho BH, Kim JT, Lee JS, et al. Associations of various blood pressure parameters with functional outcomes after endovascular thrombectomy in acute ischaemic stroke. Eur J Neurol 2019; 26: 1019–1027. [DOI] [PubMed] [Google Scholar]
- 17. Payne S. Cerebral autoregulation. Cham: Springer, 2016. [Google Scholar]
- 18. Khatri R, McKinney AM, Swenson B, et al. Blood-brain barrier, reperfusion injury, and hemorrhagic transformation in acute ischemic stroke. Neurology 2012; 79: S52–S57. [DOI] [PubMed] [Google Scholar]
- 19. Malhotra K, Goyal N, Katsanos AH, et al. Association of blood pressure with outcomes in acute stroke thrombectomy. Hypertension 2020; 75: 730–739. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Delgado-Mederos R, Ribo M, Rovira A, et al. Prognostic significance of blood pressure variability after thrombolysis in acute stroke. Neurology 2008; 71: 552–558. [DOI] [PubMed] [Google Scholar]
- 21. Chang JY, Jeon SB, Lee JH, et al. The relationship between blood pressure variability, recanalization degree, and clinical outcome in large vessel occlusive stroke after an intra-arterial thrombectomy. Cerebrovasc Dis 2019; 46: 279–286. [DOI] [PubMed] [Google Scholar]
- 22. Castro P, Ferreira F, Nguyen CK, et al. Rapid assessment of blood pressure variability and outcome after successful thrombectomy. Stroke 2021; 52: e531–e535. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Romagnoli S, Ricci Z, Quattrone D, et al. Accuracy of invasive arterial pressure monitoring in cardiovascular patients: an observational study. Crit Care 2014; 18: 644. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental material, sj-pdf-1-eso-10.1177_23969873221106907 for Temporal profiles of systolic blood pressure variability and neurologic outcomes after endovascular thrombectomy by Ayush Prasad, Jessica Kobsa, Sreeja Kodali, David Bartolome, Liza Begunova, Darko Quispe-Orozco, Mudassir Farooqui, Cynthia Zevallos, Santiago Ortega-Gutiérrez, Mohammad Anadani, Eyad Almallouhi, Alejandro M Spiotta, James A Giles, Salah G Keyrouz, Joon-Tae Kim, Ilko L Maier, Jan Liman, Marios-Nikos Psychogios, Nolwenn Riou-Comte, Sébastien Richard, Benjamin Gory, Stacey Quintero Wolfe, Patrick A Brown, Kyle M Fargen, Eva A Mistry, Hiba Fakhri, Akshitkumar Mistry, Ka-Ho Wong, Fábio A Nascimento, Peter Kan, Adam de Havenon, Kevin N Sheth and Nils H Petersen in European Stroke Journal