Core Blood-Brain Barrier Disruption in Patients With Large Vessel Occlusion Authors: Richard Leigh, Nefeli Valyraki, Nae-Yuh Wang, Manar Abomulay, Adrien ter Schiphorst, Adrián Valls Carbó, María Hernández-Pérez, Frédérique Charbonneau, Caroline Arquizan, Anne Wacongne, Denis Sablot, Vincent Costalat, Canan Ozsancak, Clara Cohen, Marco Pasi, Grégoire Boulouis, Nicole Yuen, Jean-Philippe Desilles, Gregory W. Albers, Jean Marc Olivot, Pierre Seners Published: 2026 ABSTRACT BACKGROUND:Multimodal imaging has expanded treatment for patients with acute ischemic stroke with large vessel occlusion. Blood-brain barrier (BBB) disruption measured in the ischemic core is associated with hemorrhagic transformation. However, the associations between core BBB disruption (cBBBD) and baseline clinical/imaging variables, as well as 3-month outcome, have not been explored.METHODS:This is a retrospective multicenter analysis of consecutive anterior circulation patients with acute ischemic stroke with large vessel occlusion, presenting over a 4-year time period, who were transferred from a primary to a comprehensive stroke center for possible endovascular therapy, with magnetic resonance imaging that included perfusion-weighted imaging before transfer. Magnetic resonance imaging scans were processed using RAPID software to generate penumbral imaging variables. Perfusion-weighted images were processed to detect and quantify contrast leakage; cBBBD was calculated as the average of all leaky voxels in the ischemic core. Poor functional outcome was defined as a modified Rankin Scale score of >2 at 3 months. Linear regression was used except for the outcome, which used logistic regression, controlling for age, stroke severity, and baseline functional status.RESULTS:Out of 411 patients transferred for endovascular therapy, 291 were included in this analysis with a median age of 74 years; 49% were female patients. The median National Institutes of Health Stroke Scale score was 13, the mean core volume was 32.3 mL, and the mean cBBBD was 2.1%. 71% of patients underwent endovascular therapy. Admission National Institutes of Health Stroke Scale score (P<0.001) and glucose level (P=0.033) were independently correlated with cBBBD. All imaging variables correlated strongly with cBBBD (P<0.001). The strongest correlation was 0.50, observed between cBBBD and mismatch ratio (r2=0.254). Increasing cBBBD was independently associated with poor functional outcome (adjusted odds ratio, 1.16 [CI, 1.03–1.32]; P=0.019; n=279), indicating that for every 1% increase in cBBBD, the odds of having a poor functional outcome increase by 16%.CONCLUSIONS:In acute ischemic stroke with large vessel occlusion, disruption of the BBB in the core lesion is independently associated with clinical outcome. cBBBD represents a new imaging profile for acute stroke that may help guide treatments. Advances in endovascular therapy (EVT) have revolutionized the treatment of acute ischemic stroke with large vessel occlusion (AIS-LVO), achieving successful recanalization in over 80% of patients.1,2 Despite this, less than half of patients treated with EVT retain functional independence after stroke.2 Multimodal imaging has been used to identify patients most likely to benefit from revascularization.1,3,4 Traditionally, this approach has focused on identifying salvageable tissue with penumbral imaging.5 In practice, it is combined with postprocessing software and used to identify regions of ischemic penumbra and core infarct. Although this model is practically useful, it implies a homogeneity in the risk or injury to the underlying tissue regions. However, in reality, the injury to the underlying tissue, particularly in the region designated as core infarct, is heterogeneous. 6–8 Another way to examine tissue injury in ischemic and infarcted tissue is by measuring disruption of the blood-brain barrier (BBB). The BBB is a complex machinery of cells and cell structures that act together to form a physiological barrier regulating the transfer of cells and molecules between the cerebral vasculature and the brain parenchyma.9 Soon after the onset of cerebral ischemia, the BBB starts to become progressively more permeable, initially reflecting BBB dysfunction, but subsequently progressing to BBB injury, and eventually BBB rupture. The BBB can be imaged using dynamic susceptibility contrast (DSC) imaging, the same sequence used to generate perfusion-weighted imaging with magnetic resonance imaging (MRI).10–12 This approach has been used to estimate risk of hemorrhagic complications in patients treated with thrombolysis13 and thrombectomy.14,15 Using this approach, it has been shown that mild diffuse BBB disruption is reversible with reperfusion, whereas focal severe BBB disruption is associated with BBB rupture.16 More moderate BBB disruption may act as a mechanism for the development of subsequent vasogenic edema.17 For patients who present with AIS-LVO and are transferred for EVT, there is a unique opportunity to mitigate these deleterious effects before revascularization to minimize ischemia-induced injury with reperfusion and improve functional outcomes. Here, we look at the imaging variable core BBB disruption (cBBBD) in a population of patients with AIS-LVO who were subsequently transferred for thrombectomy. We evaluated which baseline, clinical, and imaging features were associated with cBBBD and explored how cBBBD related to subsequent functional outcome. Methods Our analysis was reported following the STROBE guidelines (Strengthening the Reporting of Observational Studies in Epidemiology; Supplemental Material).18 The data supporting the study findings are available from the corresponding author on reasonable request. Human Research Protocol The research was approved by the Rothschild Foundation Hospital review board (institutional review board no. 00012801, study number CE_20240326_1_PSS). According to the French legislation, each patient was informed of their participation in the study. However, the requirement for written informed consent was waived as this study only implied retrospective analysis of anonymized data collected as part of routine care. Study Population This is a retrospective analysis of a cohort of consecutive patients in France who presented to a primary stroke center (Millau, Nîmes, Orléans, or Perpignan), were diagnosed with an anterior circulation AIS-LVO based on MRI with DSC, and were transferred to a comprehensive stroke center (Montpellier or Tours) for possible thrombectomy. Inclusion periods were from August 2015 to December 2019 for Orléans, and from January 2020 to December 2023 for the other 3 primary stroke centers, the time period when MRI with perfusion-weighted imaging scans were routinely performed in these centers. Eligible patients received thrombolysis after their MRI scan before transfer, whenever recommended. After transfer, the large vessel occlusion status was reassessed either with noninvasive imaging (computed tomography [CT] angiography/magnetic resonance [MR] angiography) or by digital subtraction angiography to determine eligibility for thrombectomy. Patients were included in this analysis if they had an MRI with DSC that was successfully processed by RAPID software and the BBB image processing pipeline. Clinical variables routinely recorded in the acute stroke setting were collected. The 3-month modified Rankin Scale (mRS) score was assessed by a stroke neurologist or a research nurse. MRI Protocol MRI protocols varied across the 4 sites, but all included diffusion-weighted imaging, MR angiography, T2* or susceptibility weighted imaging, and DSC imaging. DSC images were collected before, during, and after administration of a weight-based dose of gadolinium contrast through a power injector. MRI Processing Diffusion-weighted imaging and DSC were processed using RAPID software to create, and measure the volume of, the following regions of interest: core infarct (based on an apparent diffusion coefficient threshold of <620×10−6 mm2/s), time-to-maximum of the residual function >6 seconds (Tmax6), and time-to-maximum of the residual function >10 seconds (Tmax10). Mismatch ratio was defined as the Tmax6 volume divided by the core volume, with an upper limit of 100 (to avoid infinite values). The hypoperfusion intensity ratio (HIR) was defined as the Tmax10 volume divided by the Tmax6 volume, with low HIR indicating milder hypoperfusion and better collaterals. DSC source images, after having been processed by RAPID software, were used to generate the BBB images. The recorded signal change during a DSC image acquisition is largely due to intravascular contrast. However, in the setting of BBB disruption, there is a competing signal moving in the opposite direction, which is proportional to the concentration of contrast that has leaked through the BBB and into the brain parenchyma. Once the regional differences in intravascular signal change are accounted for using an arrival time correction,11 the remaining effect on the signal that is due to BBB disruption can be isolated and expressed as a percent loss of signal due to contrast leakage on a voxel-by-voxel basis. Figure 1 shows an example of how different severities of BBB disruption affect the recorded signal and the corresponding BBB heatmap. BBB analysis was performed using MATLAB software. The effect of blood-brain barrier (BBB) disruption on the recorded signal and the corresponding heatmap. This figure demonstrates how disruption of the BBB affects the recorded signal on a dynamic susceptibility contrast magnetic resonance imaging (MRI) acquisition. Top left, BBB heatmap for the core infarct of a patient in the study. Bottom left, Color code indicating the severity of BBB disruption shown on the MRI. On the right, the average recorded signal is shown for each echelon of BBB disruption using a similar color code as the MRI. The dark blue line shows how the signal behaves in the absence of BBB disruption. Of note, the recorded signal has been converted to delta R (sometimes referred to as the gadolinium concentration curve) to reflect the change in signal from baseline. Additionally, an arrival time correction (ATC) has been applied to the curves before averaging them to remove the effect of regional differences in blood flow. The imaging variable, cBBBD, was defined as the average of all voxels in the core region of interest that demonstrated contrast leakage. In the absence of a core lesion (infinite mismatch), cBBBD was set to zero. BBB analysis was conducted by one author (R.L.) who was blinded to the clinical data. Statistical Analysis Characteristics of the study sample were described using mean and SD or median and 25th to 75th percentile (Q1–Q3) for continuous variables, as appropriate, and frequency and proportion, as percentages, for categorical variables. Univariable linear regression was used to study the association between cBBBD (as the dependent variable) and the following clinical baseline characteristics (as independent variables): age, sex, hypertension, diabetes, coronary artery disease, active tobacco use, prior stroke, statin use, admission systolic and diastolic blood pressure, serum glucose on admission, time from stroke onset to MRI (for patients with a witnessed onset), admission National Institutes of Health Stroke Scale (NIHSS) score, and prestroke mRS score. Univariable linear regression was also used to study the association between cBBBD (as the dependent variable) and the imaging variables: core volume, Tmax6 volume, Tmax10 volume, HIR, and mismatch ratio (as independent variables). These associations were further studied in a multivariable analysis, which included baseline characteristics selected as covariates through the purposeful selection approach described by Hosmer and Lemeshow.19 In anticipation of cBBBD to follow a skewed distribution, and likely also with excessive 0 values, we conducted sensitivity analyses using a generalized linear model approach employing a Tweedie distribution to evaluate the robustness of results from linear regression, although our sample size was large enough for the results to be valid without the normal distribution assumption for the outcome residuals being true. In addition, Spearman rank correlation was also performed for variables identified by linear regression as having a significant correlation with cBBBD. The binary variable poor functional outcome was defined as a 3-month mRS score >2. The association between cBBBD (as an independent variable) and poor functional outcome (dependent variable) was studied using logistic regression. Age, admission NIHSS score, and prestroke mRS score were then added to the model because these are established predictors of outcome. Imaging variables were not included as covariates due to their strong correlation with cBBBD. The relationship between cBBBD and poor functional outcome was also tested in 2 sensitivity analyses: (1) focusing on patients treated with thrombectomy, and (2) focusing on patients who achieved postthrombolysis or postthrombectomy modified Treatment in Cerebral Infarction score of 2b, 2c, or 3. Associations of cBBBD with the outcomes of 24-hour NIHSS score (linear regression) and death at 3 months (logistic regression) were also evaluated. A P<0.05 was considered statistically significant. Stata 19 and SAS 9.4 software were used to conduct the statistical analysis. Results Population Characteristics There were 291 patients with AIS-LVO in the registry who had adequate-quality MRI with DSC at the primary stroke center before transfer to the comprehensive stroke center for possible EVT (see Figure S1 for inclusion/exclusion details). The median age was 74 (Q1–Q3, 63–81), and 49% were female patients. The median NIHSS score was 13 (Q1–Q3, 7–19), the mean core volume was 32.3 (SD 45.2) mL, and the mean cBBBD was 2.11% (SD, 2.45). The median time from last known well to MRI for the 205 patients who had a witnessed onset was 129 (92–168) minutes. The occlusion sites were intracranial internal carotid artery in 53 (18%), first and second segments of the middle cerebral artery in 173 (59%) and 65 (22%), respectively. The median time from the MRI at the primary stroke center to arrival at the comprehensive stroke center was 195 (156–228) minutes. After transfer, 218 (75%) were taken for angiography, and 206 (71%) underwent thrombectomy. The reasons for not performing thrombectomy were recanalization or distal thrombus migration in 42, mild symptoms in 32, large core in 4, lack of mismatch in 5, and other reasons in 2. For those with a witnessed onset who were treated with thrombectomy, the median time from onset to EVT was 373 minutes (interquartile range, 320–458). For patients who were taken for EVT, the median time from MRI to recanalization or end of procedure (whichever came first) was 235 minutes (interquartile range, 192–282). Clinical characteristics of the population are summarized in Table 1. Table 1.Participant Characteristics (N=291) Participant Characteristics (N=291) Baseline Clinical Variables Associated With cBBBD In univariable analysis, cBBBD was associated with admission NIHSS score (β, 0.098 [CI 0.060–0.135]; r2=0.082; P<0.001; n=291) and glucose level (β, 0.180 [CI, 0.050–0.130]; r2=0.026; P=0.007; n=284) but none of the other variables tested (Table 2). Using Spearman rank correlation, NIHSS score (ρ=0.43; P<0.001) and glucose (ρ=0.15; P=0.012) were also significantly associated with cBBBD. In multivariable regression using purposeful selection, both serum glucose (P=0.038) and admission NIHSS score (P<0.001) were independently associated with cBBBD. Analyses using a generalized linear model approach with a Tweedie distribution for cBBBD showed similar results (Table S1). Table 2.Associations Between cBBBD and Predictors/Covariates in the Univariable Linear Regression Analysis Associations Between cBBBD and Predictors/Covariates in the Univariable Linear Regression Analysis Imaging Variables Associated With cBBBD In univariable linear regression, increasing cBBBD was associated with increasing core volume (P<0.001), Tmax6 volume (P<0.001), Tmax10 volume (P<0.001), higher HIR (P<0.001), and lower mismatch ratio (P<0.001). These associations are summarized numerically in Table 2 and visually in Figure 2. When testing these relationships with Spearman rank correlation, all correlations remained significant with P<0.001. In multivariable analysis, adding both serum glucose and admission NIHSS score to each of the models, glucose remained significant in all models, whereas admission NIHSS score was not significant in any model. Comparison of core blood-brain barrier (BBB) disruption with established imaging variables. Linear regression lines with 95% CIs are shown for the associations between core BBB disruption and the established imaging variables. A, Relationship with the volume measurements. B, Relationship with mismatch ratio (MMR) and hypoperfusion intensity ratio (HIR). MMR, which is capped at a maximum value of 100, is divided by 100 to normalize it to have the same range as HIR (0–1). Tmax6 indicates time-to-maximum of the residual function >6 seconds; and Tmax10, time-to-maximum of the residual function >10 seconds. cBBBD and Outcome Higher cBBBD was associated with an increased likelihood of a poor functional outcome (odds ratio [OR], 1.20 [CI 1.07–1.35]; P=0.002; n=279), indicating that for every 1% increase in cBBBD on arrival, the odds of a poor functional outcome at 3 months increase by 20%. Age (P<0.001), admission NIHSS score (P<0.001), and prestroke mRS score (P<0.001) were all also associated with a poor functional outcome. After adjusting for these variables in the model, cBBBD remained associated with poor functional outcome (OR, 1.16 [CI, 1.03–1.32]; P=0.019; n=279). When looking at the subgroup of 198 patients who underwent thrombectomy, the association between cBBBD and poor functional outcome was similar (OR, 1.18 [CI, 1.02–1.37]; P=0.023; n=198). In the subgroup of 216 patients who achieved reperfusion (either pre-EVT or post-EVT), the association was again similar (OR, 1.15 [CI, 1.02–1.29]; P=0.027; n=216). Increasing cBBBD was also associated with higher 24-hour NIHSS score (β, 1.021 [CI, 0.56–1.48]; r2=0.064; P<0.001; n=282) and an increased risk of death at 3 months (OR, 1.12 [CI, 1.00–1.25]; P=0.042; n=279), indicating the odds of dying increased by 12% for every 1% increase in cBBBD before transfer. Patterns of BBB Disruption To better understand what additional information cBBBD offers beyond the traditional imaging variables, we identified several cases from the data set that appeared similar based on their core/penumbra profile but differed based on the BBB profile. Figure 3 compares 2 patients who presented with a large core profile and an unfavorable mismatch. Despite having similar imaging features based on their RAPID profiles, their outcomes were different and more aligned with their BBB profiles. Additional examples can be found in the Supplemental Material. Figure S2 depicts a large core favorable mismatch, Figure S3 depicts a modest core favorable mismatch, and Figure S4 depicts a small core favorable mismatch. These selected cases demonstrated how cBBBD can vary despite similar profiles based on traditional imaging measures. Discussion Our results show that among patients with AIS-LVO referred for EVT, cBBBD detected on pretransfer clinical perfusion MR imaging was (1) common, varying in severity, and (2) associated with a deleterious impact on 3-month functional outcome for the whole population, those that were taken for thrombectomy, and those that achieved reperfusion. Taken together, these findings suggest that pretreatment core “leakage” could become an additional characteristic of the AIS-LVO imaging profile, providing a new biomarker for forecasting prognosis and targeting treatments aimed at improving the clinical response of patients to EVT. In this population of patients with AIS-LVO, who were scanned after a median delay of 2 hours and 15 minutes from witnessed onset, disruption of the BBB was detected in the core lesion of all patients; however, the severity varied and was associated with other baseline MR perfusion parameters. It was previously demonstrated that mild BBB disruption is reversible with reperfusion,16 and in the current study, we found that, on average, lower levels of BBB disruption were associated with more salvageable penumbra (higher mismatch ratio) and better collateral profile (lower HIR). It remains to be determined if mild, or absent, BBB disruption in the core infarct represents salvageable (or partially salvageable) tissue. Similarly, a recent study looking at areas of the penumbra that progress to be part of the final infarct volume found more severe BBB disruption in penumbral tissue that was not salvaged,20 suggesting BBB imaging could also be informative about tissue at risk. Future studies examining the topographical severity of critical hypoperfusion, diffusion restriction, and BBB disruption in the core and penumbra are warranted. It is also important to recognize that because the BBB disruption in this study was measured before transfer, there was a median delay of 6 hours before recanalization. Based on animal data, it is expected that the BBB will become more permeable with time in the setting of sustained hypoperfusion.9 We also anticipate that the relationship between time and cBBBD will vary from patient to patient based on a variety of factors. Although few patients in the current study had repeat MR imaging after transfer, future studies using data sets with serial MR imaging are planned to try to elucidate this relationship. The second main finding of this study is that pretransfer/pretreatment cBBBD is associated with a poor functional outcome at 3 months postdischarge, even in the subgroup of patients who achieved successful reperfusion acutely. Based on our prior work, we know that in patients with AIS-LVO who are successfully reperfused, there is an association between pretreatment BBB damage and posttreatment hemorrhagic transformation (HT)15 and that HT is associated with worse outcome.21 HT is one of the main drivers of poor clinical outcome in the current study as well, a relationship we will study more closely in a separate analysis. It has also been found that posttreatment BBB disruption is associated with cerebral edema17 and that the benefits of EVT are in part mediated by preserving BBB integrity.22 We, therefore, suspect that ischemia-reperfusion injury, in the form of edema and HT, may explain part of the relationship observed between baseline cBBBD and poor functional outcome. This may be particularly informative for patients with AIS-LVO presenting with a large core, because they are at increased risk of ischemia-reperfusion injury from edema23 or HT.24 This may also be true for patients presenting in a late window, as delay in reperfusion has been associated with worse ischemia-reperfusion injury.25 Finally, another interesting finding of our results was the association between increased cBBBD and elevated baseline glycemia, suggesting there may be direct BBB damage from hyperglycemia through various mechanisms, such as the thrombo-inflammatory response involving platelet and neutrophil–related mechanisms demonstrated in animal models.26 This hypothesis is supported by results of recent studies suggesting that inflammatory response in AIS-LVO contributes to ischemia-reperfusion injury,27 another possible mechanism linking cBBBD to outcome after stroke. The BBB imaging method in our study uses a standard clinical MRI sequence (DSC) that is already part of most acute stroke MRI protocols. As with conventional penumbral imaging, BBB imaging requires postprocessing both to create the BBB heat maps and to extract relevant metrics. The metric used in this study, cBBBD, is the average of all BBB disruption detected in the core. This metric captures a range of BBB disruption from absent to mild to moderate to severe. In previous studies, where the goal was to use BBB imaging to predict hemorrhagic complications, a different measure, mean permeability derangement,13–15 has been used. Mean permeability derangement is the average of all voxels in the core that are above a threshold. Mean permeability derangement is elevated when there are focal areas of severe BBB disruption, a pattern associated with hemorrhagic complications.16 Future studies may identify additional metrics from BBB imaging that have clinical relevance. BBB imaging could be incorporated into existing software platforms already in clinical use and provide BBB heat maps and metrics such as cBBBD and mean permeability derangement. However, determining the clinical utility of BBB imaging will require prospective studies. The operational definition of the leaky core model presented here does require the use of MRI. While MRI is widely available, its use in acute stroke triage varies, and CT remains the most common type of multimodal imaging used due to its speed and convenience. The development of penumbral imaging also began on MRI and was later adapted to work with CT perfusion imaging. CT perfusion has been used to detect clinically relevant pretreatment BBB disruption in patients with stroke,28–30 suggesting that a CT-based leaky core model may be possible, broadening the impact of the findings in the current study. We hope that this study can serve as a blueprint for developing a leaky core model to help guide acute stroke treatments. Whether on MRI or on CT, it is the accumulation of contrast in the parenchyma during perfusion imaging that defines the model. On MRI, the accumulation of contrast in the tissue causes a signal change that is opposite to that of intravascular contrast, making it easier to isolate. With CT, this is more challenging because the signal due to intravascular and extravascular contrast moves in the same direction, and identification of BBB disruption is based on retained signal rather than divergent signal (like with MR). However, despite these challenges, with technological advances and possibly with the assistance of artificial intelligence, a CT-based model is likely to be achieved. There are several limitations to this study; it is a retrospective analysis, thus the findings will need to be confirmed prospectively. The sample size was large; however, the population was heterogeneous, and not all patients were treated with EVT. All MRIs in this study were performed before the administration of thrombolysis, which itself can affect the BBB31; thus, we are unable to assess in this study how thrombolytic therapy may impact BBB integrity. Future research will look into this relationship using data from studies with serial imaging.32 Although the median core size in this study of 17 mL is typical for an AIS-LVO population,4 the clinical impact of cBBBD may be more relevant in a larger core population. This study used pretransfer MRI scans, providing a unique view into how the BBB may be targeted during transfer; however, BBB imaging acquired immediately before intervention may have different implications. Additional studies are needed. Conclusions In patients presenting with AIS-LVO who were imaged before transfer for EVT, disruption of the BBB in the core lesion was early, frequent, and independently associated with poor outcome. In addition to penumbral salvage, protection of the leaky core appears to be an important target for improving outcomes. Comparison of RAPID output and blood-brain barrier (BBB) output for 2 patients with a large core and no mismatch. The RAPID output and the BBB output for 2 patients presenting with a large core and an unfavorable mismatch are shown. Patient A presented with an internal carotid artery occlusion and a National Institutes of Health Stroke Scale (NIHSS) score of 27, had a core volume of 173 mL, and no mismatch (mismatch ratio=0.90). Patient B presented with a proximal M1 occlusion and an NIHSS score of 18, had a core of 177 mL, and no mismatch (mismatch ratio=0.89). Patient A has some scattered BBB disruption in the core infarct (outlined in black), most of which is minimal (green), and when averaged, gives a core BBB disruption (cBBBD) value of 3.1%. Patient B has extensive diffuse severe BBB disruption (red) throughout the core lesion that averages to a cBBBD of 18.8%. Although both patients had Treatment in Cerebral Infarction (TICI) 3 reperfusion, patient A achieved a good functional outcome, having a 3-month modified Rankin Scale (mRS) score of 2, whereas patient B had a poor outcome with an mRS score of 6. Article Information Presented in part at the European Stroke Organisation Conference, Maastricht, the Netherlands, May 6–8, 2026. Disclosures Dr Seners reports speaker fees from Boehringer Ingelheim and RAPIDAI. Dr Olivot reports speaker fees from Boehringer and Consulting from AbbVie. Dr Albers reports stock holdings in iSchemaView; compensation from Biogen, iSchemaView, and Genentech for consultant services. Dr Leigh reports a patent issued for System and method for blood-brain permeability imaging (BBPI), US Patent US10076263B2. Dr Abomulay reports grants from Agence Nationale de la Recherche. Dr Ter Schiphorst reports grants from the Fulbright Association, Stanford University, Rotary Foundation, and Philippe Foundation Inc. Dr Arquizan reports compensation from Novartis for consultant services; compensation from Medtronic Vascular Inc for other services; compensation from Amgen for other services; and compensation from Olea Medical for consultant services. Dr Boulouis reports compensation from Siemens for consultant services. Dr Costalat reports compensation from Penumbra Inc for consultant services; compensation from Balt USA, LLC for consultant services; compensation from Medtronic USA Inc for consultant services; compensation from Stryker Corporation for consultant services; compensation from MicroVention Inc for consultant services; and compensation from Johnson & Johnson Health Care Systems Inc for consultant services. The other authors report no conflicts. Supplemental Material Table S1 Figures S1–S4 STROBE Checklist Supplementary Material