Clinical PerspectiveWhat Is New?This study suggests that successful reperfusion is associated with a reduced malignant cerebral edema and with improved outcomes in patients with large‐core infarction with baseline Alberta Stroke Program Early Computed Tomography Scores of 3 to 5, partially through a reduction in brain edema formation, but not in those with baseline Alberta Stroke Program Early Computed Tomography Scores of 0 to 2.What Are the Clinical Implications?Clinically, our findings suggest that targeting reperfusion edema represents a potential strategy to improve the efficacy of endovascular thrombectomy.This study has important implications by providing valuable insights for identifying patients with baseline Alberta Stroke Program Early Computed Tomography Scores 0–2 who may derive maximal therapeutic benefits from targeted neuroprotective intervention, because successful recanalization failed to significantly decrease malignant cerebral edema incidence in this group.
Endovascular thrombectomy (EVT) demonstrates superior efficacy in improving functional outcomes compared with medical therapy alone in patients with large‐core infarction (LCI) with baseline Alberta Stroke Program EarlyComputed Tomography Scores (ASPECTS) of 3 to 51, 2, 3, 4; however, the benefit was limited for patients with ASPECTS of 0 to 2.5 Up to 18% to 51.0% of patients die or survive with significant disability despite treatment.6 There is increasing focus on identifying key factors that compromise the efficacy of EVT in patients with LCI.
Cerebral edema associated with LCI may significantly diminish the clinical benefits of EVT.7, 8 The risk of developing malignant infarction is maximized in patients with a large early ischemic core, which can lead to severe complications by progressive ischemic edema within 2 to 3 days after stroke onset, with mortality rates of up to 80%.9, 10 However, whether reperfusion can reduce or aggravate malignant cerebral edema (MCE) in patients with LCI remains conflicting. Experimental data and some human studies have suggested that reperfusion could exacerbate brain edema in the setting of large ischemic volumes.11, 12, 13 Conversely, some investigations have indicated that successful reperfusion may attenuate edema.14 Additionally, 2 recent mediation analyses further indicate that a significant portion of the therapeutic benefits of endovascular reperfusion stems from mitigating brain edema rather than solely from reducing final infarct volume in small to moderate‐sized ischemic infarctions.15, 16 Nevertheless, whether this mediating effect is also present in patients with LCI needs more research.
The success of EVT in LCI stroke has emphasized the need to elucidate the impact of reperfusion on cerebral edema dynamics, particularly in the context of MCE development. This analysis evaluated the association between reperfusion and MCE in participants from the MAGIC (Prospective Multicenter Registry on Early Management of Acute Ischemic Stroke) registry. Building on preliminary evidence, we hypothesized that the clinical benefits of reperfusion might be partially attributable to edema reduction.
Additionally, we explored whether successful recanalization’s protective effects vary with different baseline ASPECTS profiles.
METHODS
Study Population
This subanalysis used data from the MAGIC registry, a nationwide, prospective study (URL: http://www.chictr.org.cn; unique identifier: ChiCTR2100051664) conducted across 38 stroke centers in China (Nov 2021–Feb 2023), enrolling patients with LCI within 24 hours of last known well. LCI was defined as noncontrast‐enhanced CT (NCCT) 0 to 5 to enhance generalizability, supported by its strong correlation with advanced imaging defined large cores17 and guideline endorsement18 for EVT selection.5 Full methodology and primary outcomes have been published previously.19 The inclusion criteria were as follows: (1) aged ≥18 years; (2) confirmed LCI stroke identified on NCCT ASPECTS 0 to 5; (3) time interval from last known well to large vessel occlusion stroke confirmation within 24 hours; and (4) pre‐troke modified Rankin Scale (mRS) score <2. The exclusion criteria were as follows: (1) hemorrhagic transformation affecting imaging assessment, (2) absence of postoperative imaging or poor imaging quality, (3) loss to 90‐day follow‐up, and (4) patients with onset to puncture time >24 hours.
The flow chart for inclusion in the study population is displayed in Figure 1.
Ethical Approval
The MAGIC registry received ethical approval from the Institutional Review Board of Xinqiao Hospital, Army Medical University, with subsequent protocol endorsement by local institutional review boards at all participating sites. Written informed consent was obtained from all study participants or their legally authorized representatives.
Procedures and Data Collection
Baseline clinical characteristics, radiological profiles, procedural parameters, and clinical outcomes were systematically collected according to standardized criteria. Stroke cause was classified according to the TOAST criteria (Trial of ORG 10172 in Acute Stroke Treatment).20 Collateral status was graded using the American Society of Interventional and Therapeutic Neuroradiology/Society of Interventional Radiology collateral scale,21 which was further divided into poor collateral (score 0–2) and good collateral (score 3–4) statuses. Recanalization was confirmed by CT angiogram or magnetic resonance angiogram during hospitalization in the overall cohorts.
Successful reperfusion was defined as modified Treatment in Cerebral Ischemia classification 2b‐3 in the EVT arm.22 Functional outcome was assessed using the 90‐day mRS score.
Definition of MCE
All patients underwent at least 1 NCCT follow‐up scan within 72 hours after stroke, on which midline shift was assessed. In the EVT arm, an NCCT must be performed immediately after EVT, and at least 1 additional NCCT is required within 72 hours after stroke onset. Patients with MCE were defined as those with (1) ischemic infarcts demonstrating significant mass effect, involving >50% of the middle cerebral artery territory, with accompanying local cerebral edema evidenced by sulcal effacement and lateral ventricular compression and (2) midline shift ≥5 mm, measured at the septum pellucidum or pineal gland, with associated effacement of the basal cisterns.15 The imaging characteristics of all enrolled subjects were evaluated by 2 independent neurologists/interventionalists blinded to the clinical information.
Statistical Analysis
Baseline characteristics, treatment details, and outcomes were compared between groups with and without MCE using descriptive analyses. Categorical variables were analyzed using the chi‐square test, nonnormally distributed continuous variables with the Mann–Whitney U test, and normally distributed continuous variables with the Student’s t test.
The associations between recanalization/reperfusion and MCE were evaluated using binary logistic regression. Variables demonstrating marginal associations (P<0.10) in preliminary univariable analyses were subsequently incorporated into multivariable regression models using backward elimination for optimal variable selection.
Ordinal logistic regression (univariable and multivariable) was used to assess the relationship between MCE and 90‐day mRS outcomes. The effect of MCE was quantified as common odds ratios (ORs) and adjusted common ORs for worse functional outcomes.
To assess MCE as a mediator of the relationship between reperfusion status and functional outcome, mediation analysis was performed using the template described by Baron and Kenny23 and the method by Vanderweele and Vansteelandt,24 with MCE as the mediating variable, and age, intravenous thrombolysis, occlusion site, baseline National Institutes of Health Stroke Scale score (NIHSS), collateral status, and glucose levels as covariates. Although mediation analysis should be used cautiously in observational studies due to potential confounders affecting both the exposure and outcome, we employed a large sample size and, based on expert knowledge and existing literature,14 comprehensively measured and adjusted for all known exposure–mediator and mediator–outcome confounders as covariates in the statistical model. We also conducted a sensitivity analyze to enhance the robustness of the results.
The statistical significance of the mediation effect, with results reported as 95% CIs derived from bootstrap methods (1000).
For all analyses, a 2‐tailed value of P≤0.05 was considered statistically significant. IBM SPSS (version 26.0; IBM Corp, Armonk, NY) and R‐studio (version 4.3.3, R Foundation for Statistical Computing, Vienna, Austria) software were used for all statistical analyses.
Exploratory Analysis
Stratified analyses to explore the differential mediation effect of MCE progression on functional outcome among patients with different baseline ASPECTS (0–2 versus 3–5), with particular emphasis on identifying potential clinical implications of therapeutic heterogeneity. We quantified the percentage change in the regression coefficients before and after adjusting for MCE to estimate its mediating role.
Data Availability
Deidentified participant data and a corresponding data dictionary (including variable definitions, coding schemes, and protocol details) will be shared with qualified researchers upon formal request to the corresponding author, subject to institutional data transfer agreements and ethics committee approval.
RESULTS
Baseline Characteristics and Incidence of Malignant Cerebral Edema
Among the 750 patients enrolled in the MAGIC registry, 52 were excluded (30 patients with hemorrhagic transformations affecting imaging assessments, 13 patients without no postoperative imaging or poor imaging quality, 5 patients lost to 90‐day follow‐up, 4 patients with onset‐to‐puncture time more than 24 hours). Given the limited impact of 52 excluded patients on the overall sample size, we employed complete‐case analysis and did not perform data imputation. Finally, 698 patients (306 women [43.8%]; median age, 70 [interquartile range, 61–78] years; median, ASPECTS, 4 [interquartile range, 2–5]) were included in our analysis, of whom 36.8% (n=257) developed MCE.
The cohort stratified by ASPECTS severity demonstrated a significantly higher incidence of MCE in patients with ASPECTS 0 to 2 (54.3%, 125/230) versus ASPECTS 3 to 5 (28.2%, 132/468). Demographical, clinical features, and outcomes of patients with versus without MCE are shown in Table 1.
Patients with MCE had lower admission ASPECTS (median 3 versus 4), higher admission NIHSS score (median 19 versus 16), higher fasting blood glucose (median: 7.6 versus 6.8 mmol/L), lower proportions of recanalization (52.1% versus 60.1%), worse collateral status, higher internal carotid artery occlusion rates, and worse 90‐day mRS score (all P<0.05).
Predictors of Malignant Cerebral Edema
First, we evaluated the association between recanalization and MCE in the overall cohort. Recanalization was achieved in 57.2% of patients (399/698), and it was associated with lower MCE rates (33.6% versus 41.1%; P=0.041). In the multivariable models, recanalization independently predicted lower MCE rates (adjusted OR, 0.68 [95% CI, 0.47–0.99]; P=0.042), as shown in Figure 2A.
Lower baseline ASPECTS, internal carotid artery occlusion, and poor collateral status were significantly associated with MCE (all P<0.05).
We then evaluated the association between successful reperfusion and MCE in the EVT arm (n=498). Successful reperfusion occurred in 394/448 patients (88.0%). Those with successful reperfusion had fewer MCE formation compared with unsuccessful cases (33.5% versus 59.3%; P<0.001).
More details are presented in Tables S1 and S2. After adjusting for confounders, successful reperfusion reduced the risk of MCE by 66% (adjusted OR, 0.34 95% CI, 0.18–0.67; P=0.002). Other independent predictors associated with MCE included lower baseline ASPECTS, higher baseline NIHSS, higher fasting blood glucose, and internal carotid artery occlusion (Figure 2B, all P<0.05).
Malignant Cerebral Edema and Functional Outcomes
The presence of MCE was associated with a worse functional outcome (common OR, 4.37 [95% CI, 3.22–5.93]; P<001 and adjusted common OR, 3.32 [95% CI, 2.31–4.76]; P<0.001; Table 2). The distribution of mRS scores of 0 to 6 at 90 days according to the presence of MCE is displayed in Figure 3. In the multivariate analysis with adjustment for confounders, MCE was significantly associated with increased mortality at 90 days both in the overall cohorts and in the EVT arm (adjusted OR, 3.47 [95% CI, 2.31–5.21], P<0.001, adjusted OR, 4.61[95% CI, 2.73~7.78], P<0.001, respectively), Tables S3 and S4.
Mediation Analyses
Mediation analysis revealed that MCE partially explained the effect of reperfusion status on functional outcomes. When accounting for MCE mediation, recanalization’s effect decreased by 15% (Figure 4A), whereas that of reperfusion decreased by 32.5% (Figure 4B), indicating a stronger mediation role in the EVT group than in the broader patient population. In exploratory analysis, MCE mediated 11.5% (P=0.028) of recanalization benefit attenuation in ASPECTS 3–5 patients (Figure 4C).
Nonetheless, no mediation effect was observed in patients with ASPECTS 0 to 2 (recanalization → mRS score, β=−0.26, P=0.140; recanalization → MCE, β=−0.37, P=0.209), suggesting that recanalization failed to confer clinical benefits through MCE reduction in this subgroup.
Sensitivity Analysis
To address potential limitations of mediation analysis in an observational setting, we conducted a sensitivity analysis that excluded 244 patients with a borderline ASPECTS of 5. This yielded a final analytic cohort of 454 patients (197 women [43.4%]; median age, 70 [interquartile range, 63–78] years), of whom 204 (45.0%) developed MCE; more details were listed in Table S5. Patients with MCE were significantly less likely to achieve a favorable outcome (common OR, 3.97 [95% CI, 2.75–5.74]; P<0.001 and adjusted common OR, 3.01 [95% CI, 1.95–4.67]; P<0.001; see Table S6).
The subsequent mediation analysis revealed that MCE was a partial mediator, explaining 16.8% of the total effect of recanalization on functional outcomes (Figure 4D).
Subgroup Analysis
Given the distinct pathophysiological features of atrial fibrillation‐related cardioembolic strokes, we performed a subgroup analysis of 308 patients with atrial fibrillation. Among them, 116 patients (37.7%) developed MCE. In multivariable models, recanalization status, lower baseline ASPECTS, internal carotid artery occlusion, and elevated fasting blood glucose were significantly associated with MCE (all P < 0.05); more details were shown in Table S7.
These findings align with those observed in the overall cohort.
DISCUSSION
This MAGIC registry analysis establishes robust associations between endovascular recanalization/reperfusion and reduced MCE risk in patients with LCI. Notably, the neuroprotective benefits of reperfusion were mediated through edema attenuation in patients with preserved tissue integrity (ASPECTS 3–5), demonstrating a 2.2‐fold stronger indirect treatment effect in the EVT cohort (β=32.5% versus 15.0% overall). Additionally, the absence of this mediation pathway in patients with ASPECTS 0 to 2 suggests a critical threshold of ischemic injury beyond which microcirculatory failure becomes irreversible, negating edema‐modulated therapeutic gains.
Prior studies examining the association between reperfusion and the progression or alleviation of cerebral edema have yielded inconsistent findings. A real‐world cohort study of patients with low ASPECTS demonstrated that successful reperfusion was significantly correlated with attenuated cerebral edema progression, which may contribute to improved clinical outcomes.14 However, according to the HERMES (Highly Effective Reperfusion Using Multiple Endovascular Devices) meta‐analysis, reperfusion therapy led to greater midline shift in patients with very large infarct volumes (>130 mL), suggesting a potential detrimental effect on functional recovery.12 The secondary analysis of the ANGEL‐ASPECT (Endovascular Therapy in Acute Anterior Circulation Large Vessel Occlusive Patients With a Large Infarct Core) trial revealed transient elevation of mass effect at 24 hours post EVT in patients with LCI, resolving within 3 to 5 days.13 In our study, revascularization, regardless of reperfusion in the EVT group or recanalization in the overall cohort, uniformly associated with attenuated progression of MCE in patients with LCI. The discrepancy may be due to differences in the cerebral edema evaluation timepoints and the characteristics of the enrolled population.
The ANGEL‐ASPECT subgroup analysis assessed edema at 24 hours post procedure, whereas contemporary evidence indicates malignant edema typically peaks at 48 to 72 hours.10 This timing may underestimate peak edema severity in the medical management group. Besides, the ANGEL‐ASPECT analysis assessed the thrombectomy’s total treatment effect encompassing both failed and incomplete interventions, without dedicated analysis of successful vessel recanalization outcomes, which may underestimate edema rates in the medical management group while potentially overestimating the relative increase in cerebral edema proportion attributed to EVT. Despite variations in study outcomes, the collective evidence suggests that targeting reperfusion edema represents a potential strategy to enhance the effectiveness of EVT.
MCE has been well established as an independent prognostic indicator of poor functional recovery and elevated mortality risk in populations with acute ischemic stroke.10, 13, 15, 25 In this study, only 5.5% of patients with MCE achieved an mRS score ≤2 at 90 days, and 64.6% died. Current therapeutic options for MCE remain limited, making edema mitigation a critical treatment target. Our analysis demonstrated that successful reperfusion was independently associated with reduced MCE incidence and improved functional outcomes in patients with LCI.
However, significant heterogeneity was observed across baseline ASPECTS strata, with this treatment effect confined to the ASPECTS 3 to 5 subgroup (indirect effect β=−0.1, P=0.028) and absent in the ASPECTS 0 to 2 cohort. This ASPECTS‐stratified mediation pattern aligns with the emerging understanding of therapeutic windows in LCI: although patients with ASPECTS 3 to 5 retain modifiable edema pathophysiology amenable to combined reperfusion‐edema control strategies, ASPECTS 0 to 2 cases exhibit microcirculatory “point of no return” where neuroprotective interventions targeting mitochondrial rescue or inflammasome regulation may be required.6, 26, 27
Measuring the mediation effect of edema reduction on outcome, we observed a stronger effect (32.5% versus 13.5%) in the EVT arm compared with the overall cohort. The differential mediation effects suggest that EVT may modify the pathophysiological relationship between MCE and outcomes through rapid penumbra salvage in patients with LCI. Recently, a pooled, patient‐level analysis of 2 randomized controlled trials showed that the total effect of successful reperfusion on functional outcome was mediated through 3 distinct pathways: infarct growth alone accounted for 42% of the association, midline shift independently contributed 12%, and an additional 25% was explained by the mediation of midline shift reduction via infarct growth.16 Reperfusion obtained by EVT may hold greater advantages in rapidly salvaging the penumbra, thereby more effectively alleviating cerebral edema, thus further underscoring the necessity of endovascular therapy.
Our observations emphasize the importance of cerebral edema as a therapeutic target in EVT‐treated patients presenting with LCI stroke. Stroke severity (quantified by higher NIHSS scores and low ASPECTS), proximal arterial occlusion, higher admission glucose, and poor collateral circulation are associated with MCE.14, 23, 28 We noticed the impact of admission glucose levers on MCE formation was amplified in the EVT cohort, suggesting protocol‐driven glucose management should be prioritized in order to alleviate MCE in EVT‐treated patients. Although successful recanalization failed to significantly decrease MCE incidence in patients with ASPECTS 0 to 2, our findings provide valuable insights for identifying specific patient subgroups that may derive maximal therapeutic benefits from targeted neuroprotective intervention.
Future studies exploring the early initiation of antiedema drugs (such as glibenclamide),29 particularly in combination with multitarget neuroprotective strategies and timed with reperfusion interventions, may optimize therapeutic efficacy and functional recovery in patients with LCI.
It is important to note that the presented results were achieved using NCCT ASPECTS 0 to 5 to define LCI. As such, our findings are directly generalizable to widespread clinical settings where advanced imaging is not routinely available. The use of ASPECTS <6 as a criterion is supported by its established correlation with CT perfusion core volume >70 mL17‐both are guideline‐endorsed thresholds for LCI, as evidenced by published studies.1, 2, 3, 4, 5 Moreover, the proportion of patients with a favorable outcome in the MAGIC study’s EVT group was 36.9%,19 marginally lower than the 38% observed in the SELECT2 (A Randomized Controlled Trial to Optimize Patient’s Selection for Endovascular Treatment in Acute Ischemic Stroke) trial, where nearly all patients (98%) were evaluated using NCCT+CT perfusion.4
We also recognize the potential for misclassification bias with NCCT‐ASPECTS, which could lead to either overestimation (eg, chronic leukoaraiosis misinterpreted as acute infarct) or underestimation (eg, in hyperacute presentations).30 NCCT has lower sensitivity for early ischemia compared with CT perfusion or diffusion‐weighted imaging, which may have led to patient misclassification and resulted in an underestimation of the true treatment effect in a population with more rigorously defined LCI. Furthermore, the lack of detailed infarct volume parameters prevented a comprehensive investigation into the effect of successful reperfusion on functional outcomes through pathways involving multiple mediators and potential cross‐mediation.
Strengths and Limitations
The key highlights of this study are manifested through 4 principal strengths: the statistical power derived from a large sample size, enhanced external validity through a multicenter study design, technical insights provided by granular procedural parameters, and blinded end point adjudication of MCE employing standardized imaging criteria. However, several limitations should be acknowledged. First, the nonrandomized, retrospective design introduces unavoidable selection bias and fundamentally limits causal inference.
Furthermore, we acknowledge the strong possibility of residual confounding, particularly from inconsistent clinical management—including the use of hyperosmolar agents, glycemic and blood pressure protocols, and post‐EVT antithrombotic timing—all of which could significantly influence the development of MCE. Second, although variable selection incorporated clinical knowledge, the use of univariable screening may still have omitted some potential confounders, such as gender.31 Third, the exclusion of 52 patients—including patients who were unable to complete imaging due to unstable vital signs or had severe intracranial hemorrhage potentially concurrent with MCE—might have led to an underestimation of the overall MCE rate. Fourth, as only 20.2% of the patients underwent decompressive craniectomy, our study was not sufficiently powered to rigorously analyze the impact of decompressive hemicraniectomy on the outcomes of patients with MCE.
Last, findings derived from this Chinese cohort require cautious extrapolation across ethnic populations due to systematic disparities in stroke pathogenetic profiles.
CONCLUSIONS
In conclusion, our study suggests that successful reperfusion is associated with a reduced MCE and improved outcomes partially by reducing brain edema formation in patients with LCI with baseline ASPECTS 3 to 5. This study demonstrates that reperfusion confers multifaceted therapeutic advantages beyond salvaging ischemic brain tissue at risk of infarction in patients with LCI.
Sources of Funding
This work was supported by the Scientific research project of Wuhu Municipal Health Commission (WHWJ2023z003), Scientific Research Project for Middle‐aged and Young People of Wannan Medical College (WK2024ZQNZ63), Anhui Province Clinical Medicine Transformation Special Project (No. 202427b10020048) and Research Funds of Centre for Leading Medicine and Advanced Technologies of IHM (No. 2023IHM01053).
Disclosures
None.
Supporting information