Thrombectomy improves functional outcome in large core stroke partly by preventing severe cerebral edema: a real-world mediation analysis

When a major brain artery becomes blocked, roughly a quarter of patients rapidly develop what stroke specialists call a large infarct core: an extensive region of already-dead or dying brain tissue visible on an early CT scan. For years, these patients were considered beyond rescue, and many centers declined to attempt mechanical clot removal. That changed with a wave of randomized trials demonstrating that endovascular thrombectomy, the catheter-based procedure that physically extracts the occluding clot, can still improve survival and independence even in this gravely injured population. Yet a stubborn paradox has lingered. Even after successful clot removal, death and severe disability remain common, and clinicians have disagreed about one of the most feared secondary complications: malignant cerebral edema, the progressive swelling of the infarcted hemisphere that can crush midline brain structures and culminate in fatal herniation.

Now a prospective real-world study from Da Nang Hospital in Vietnam, published in Frontiers in Neurology, offers one of the clearest quantitative answers to date on how much of thrombectomy’s benefit operates through preventing this catastrophic swelling. The research team, led by Thien Quang Le and Nhan Duc Le, with collaborators at Washington University, Oxford, Harvard-affiliated hospitals and Toronto Metropolitan University, enrolled 202 consecutive patients with anterior-circulation large-vessel occlusion, stroke severity of at least 6 on the National Institutes of Health Stroke Scale, an Alberta Stroke Program Early CT Score of 5 or below, and admission within 24 hours of onset. Crucially, patients were selected using only non-contrast CT and CT angiography, a simplified workflow deliberately chosen because perfusion imaging is unavailable in much of the world where the global stroke burden is heaviest.

The study’s primary endpoint was deliberately pragmatic. Rather than relying on quantitative edema biomarkers that require dedicated post-processing software, the team defined severe cerebral edema as clinically significant mass effect: a maximum midline shift of at least 5 millimeters on follow-up imaging within 72 hours. This threshold is not arbitrary. A shift of 5 millimeters or more is widely used as a trigger for intensified neurological monitoring, osmotherapy, and consideration of decompressive hemicraniectomy, and it can be measured reliably on routine CT at the bedside in any emergency department. Of the 202 patients, 119 underwent thrombectomy while 83 received medical management alone, and the two groups differed substantially at baseline, with the medically managed patients being older, more severely affected, presenting later, and carrying larger infarcts on admission CT.

To untangle this pronounced selection bias, the investigators applied inverse probability of treatment weighting with doubly robust adjustment, a statistical technique that reweights the observational cohort to mimic the balance of a randomized trial. After weighting, baseline differences in age, sex, admission severity, infarct extent, and time from onset collapsed to standardized mean differences below 0.03. The results were striking. Mass effect developed in 31.1 percent of thrombectomy patients compared with 51.8 percent of medically managed patients, an adjusted odds ratio of 0.28, meaning clot removal cut the odds of dangerous swelling by roughly three-quarters. The median maximum midline shift was 0 millimeters in the thrombectomy group versus 5 millimeters under medical management. Favorable outcome, defined as independence or mild disability at 90 days on the modified Rankin Scale, was achieved by 47.1 percent of thrombectomy patients versus 19.3 percent of the comparison group, with an adjusted odds ratio of 2.32. Ninety-day mortality was 30.3 percent versus 50.6 percent, and early neurological deterioration was also significantly reduced. Symptomatic intracranial hemorrhage, assessed by SITS-MOST criteria, did not differ between groups, addressing a central safety concern.

The study’s most consequential contribution, however, lies in its causal mediation analysis, a statistical framework that decomposes a total treatment effect into direct and indirect pathways. The team asked a precise question: how much of thrombectomy’s functional benefit flows through preventing severe edema? The answer was approximately 29.5 percent in the intention-to-treat comparison and 29.0 percent in an as-treated analysis comparing good reperfusion, defined as mTICI 2b or better, against poor reperfusion. The as-treated comparison functions as a natural experiment internal to the cohort, and there the indirect pathway reached formal statistical significance. Notably, the proportion mediated rose monotonically as the outcome threshold captured more severe disability, reaching 41.5 percent at a modified Rankin Scale score of 4 or worse. In other words, preventing malignant swelling is most prominent in averting the worst outcomes: severe disability and death.

These estimates converge remarkably with independent analyses using entirely different edema metrics. In the randomized TENSION trial, approximately 34 percent of thrombectomy’s improvement in 90-day outcome was associated with reduced net water uptake, a densitometric measure that directly quantifies tissue water content within the infarct, while only 14 percent was attributable to change in lesion extent. An analysis of the MAGIC registry found that incorporating malignant cerebral edema attenuated the reperfusion-outcome coefficient by roughly 32.5 percent. By contrast, a mediation analysis of the HERMES dataset had found that reduction of follow-up infarct volume explained only 12 percent of thrombectomy’s clinical benefit. The directional agreement across methodologically distinct cohorts suggests that edema attenuation is a generalizable pathway of reperfusion benefit, and one that accounts for a larger share of that benefit than final infarct volume reduction. Edema prevention and infarct limitation therefore appear to represent distinct, partly independent components of the treatment effect, with a substantial remainder operating through penumbral salvage, blood-brain barrier preservation, and metabolic restoration.

The findings also help reconcile an apparent contradiction in the literature. Secondary analyses from the SELECT2 and ANGEL-ASPECT trials had reported seemingly increased midline shift after thrombectomy in patients with very large cores, creating uncertainty about whether reperfusion might actually worsen swelling. The Da Nang team’s interpretation draws on the bimodal response hypothesis from the HERMES meta-analysis: reperfusion was not associated with greater edema overall but did exacerbate swelling in patients with extremely extensive established ischemia, while remaining beneficial in moderate-large cores with salvageable penumbra. Mechanistically, cerebral edema after large-vessel occlusion develops in overlapping stages. In the first hours, energy failure and impaired sodium-potassium ATPase activity drive cytotoxic edema, redistributing water into cells. As ionic gradients establish across an initially intact blood-brain barrier, additional water is drawn into the lesion, and only later, with progressive barrier breakdown, does a vasogenic component contribute substantially to mass displacement. Early recanalization, the authors argue, may interrupt this sodium-driven water influx while ionic edema still predominates, before vasogenic swelling becomes established. In severely injured tissue with advanced barrier disruption, however, reperfusion may transiently accentuate capillary leakage, which likely explains the early shift increases observed in some trial analyses that were no longer evident at later time points.

Beyond explaining benefit, the study delivers a practical tool for anticipation. Using ten pre-specified admission variables, the team built a parsimonious prognostic model for mass effect that retained just two parameters: lower baseline ASPECTS, reflecting more advanced ischemic injury on the admission CT, and higher admission plasma glucose, a marker of metabolic stress. Each one-point decrease in ASPECTS raised the odds of mass effect, as did each increment in glucose. On 500-iteration bootstrap internal validation, the two-variable model achieved an optimism-corrected area under the curve of 0.82 with good calibration, meaning that any emergency department, regardless of resources, can identify at the bedside which large-core stroke patients face the greatest risk of dangerous swelling. This supports individualized intensification of neurological monitoring, earlier repeat imaging, and prompt escalation to osmotherapy or decompressive hemicraniectomy in high-risk patients.

Subgroup analyses reinforced the coherence of the findings. The treatment effect remained directionally consistent across age, severity, collateral status, and ASPECTS strata, with formal interaction testing non-significant for all stratifying variables. In the ASPECTS 3-5 subgroup, thrombectomy was associated with substantially lower odds of mass effect and a higher likelihood of favorable outcome, concordant with the edema-mediated pathway mainly present in patients with salvageable tissue. In the ultra-low ASPECTS 0-2 stratum, point estimates were directionally consistent but markedly imprecise, a limitation the authors attribute to underpowering rather than evidence of effect modification. The benefits were obtained in patients selected without advanced perfusion imaging, supporting the pragmatic applicability of an NCCT-CTA pathway as thrombectomy capacity expands in resource-limited regions, where a substantial proportion of the global stroke burden is treated.

The authors are careful about the boundaries of their claims. This was a single-center, non-randomized study in which treatment allocation rested with treating clinicians, and residual confounding by indication cannot be excluded. Mediation analysis in observational data rests on identification assumptions, most importantly the absence of unmeasured confounding of the mediator-outcome relationship, and proportions mediated should be regarded as approximate, model-dependent indicators rather than precise causal quantities. The midline shift endpoint is a late and indirect surrogate whose expression depends on intracranial reserve, and earlier stages of edema captured by quantitative net water uptake were not measured. Still, the convergence of this bedside-accessible estimate with trial-based analyses using sophisticated quantitative biomarkers strengthens the interpretation that preventing malignant swelling is a real and clinically meaningful component of thrombectomy’s benefit. Whether pairing clot removal with standardized edema-targeted management, including optimized glucose and blood pressure control, or adjunctive anti-edema agents such as intravenous glibenclamide and the neuroprotective ApTOLL, further improves outcomes remains to be tested prospectively. For now, the message for clinicians is concrete: roughly one-third of thrombectomy’s power to rescue patients from the gravest strokes operates through keeping the injured brain from swelling to the point of self-destruction, and two routinely available admission numbers can flag who needs the closest watch.

Subject of Research: Thrombectomy improves functional outcome in large core stroke partly by preventing severe cerebral edema: a real-world mediation analysis

Article Title: Thrombectomy improves functional outcome in large core stroke partly by preventing severe cerebral edema: a real-world mediation analysis

Article References: Le, T. Q., Nguyen, S. V. D., Pham, T. P., Tran, T. V., Nguyen, T. Q., Yahanda, A. T., Dmytriw, A. A., Huynh, V. T., Duong, H. Q., Nguyen, T. T., & Le, N. D. (2026). Thrombectomy improves functional outcome in large core stroke partly by preventing severe cerebral edema: a real-world mediation analysis. Frontiers in Neurology, 17, Article 1936247. https://doi.org/10.3389/fneur.2026.1936247

Image Credits: AI Generated

DOI: 10.3389/fneur.2026.1936247

Keywords: Thrombectomy, improves, functional, outcome, large, core, stroke, partly, preventing, severe, cerebral, edema