The Time Is Brain Review

Oedema reduction mediates thrombectomy benefit in large core stroke: secondary analysis of the TENSION trial

· DOI: 10.1093/esj/aakag055

stroke deep-dive thrombectomy large-core-stroke imaging

Reconstructed on October 14, 2026 for the missed October 8 publication slot. This is a replacement article, not a recovered original. Source selection: verified PMC look-back mode; the original paper selection was not recovered from the October 8 archived briefing.

Clinical Question (PICO)

Among patients with anterior-circulation ischaemic stroke and baseline ASPECTS 3–5 from the randomized TENSION trial who had analyzable baseline and follow-up CT, was endovascular thrombectomy (EVT) plus best medical treatment (BMT), versus BMT alone, associated with less CT-measured oedema progression and better 90-day disability—and how much of the EVT–outcome association was statistically linked to oedema change?[1]

Bottom Line

In this CT-based secondary analysis of 177 TENSION participants, EVT was associated with lower median change in net water uptake (ΔNWU) than BMT alone[1]. Mediation models estimated that 34% of EVT's association with 90-day modified Rankin Scale (mRS) was linked to ΔNWU, versus 14% linked to ΔASPECTS[1]. These are model-dependent, post-treatment mediation estimates—not proof that directly treating oedema would reproduce EVT's benefit, nor a validated treatment-selection rule[1].

Design

  • Trial type: Secondary analysis of the investigator-initiated, randomized, open-label, blinded-endpoint, two-arm TENSION trial (NCT03094715); this substudy analyzed a modified as-treated CT cohort, not the full randomized population[1].
  • N: 253 patients were in the parent trial; 177 with suitable baseline and approximately 24-hour follow-up non-contrast CT were included (91 BMT, 86 EVT as treated)[1]. The analysis excluded 76: 41 imaged primarily with MRI and 35 with unusable CT data or no evaluable follow-up CT[1].
  • Setting: TENSION was conducted in Europe and Canada; the paper does not give the substudy site count or enrollment dates[1].
  • Analysis: Ordinal regression for the 90-day mRS distribution, plus regression and mediation models for imaging changes[1]. Mediation models adjusted for selected baseline covariates; ΔNWU models also included follow-up-CT timing[1]. The mediator was measured after treatment[1].
  • Primary outcome: Ordinal mRS, all seven categories retained, at 90 (±14) days[1].

Population

Inclusion Criteria

  • Parent-trial participants with anterior-circulation ischaemic stroke and baseline ASPECTS 3–5 who had baseline and follow-up non-contrast CT suitable for ΔNWU and ΔASPECTS assessment[1].

Exclusion Criteria

  • No suitable CT data for this substudy; 41 participants were imaged primarily with MRI and 35 had unusable CT or no evaluable follow-up CT[1]. The complete parent-trial eligibility/exclusion list is not reproduced in this secondary-analysis article[1].

Baseline Characteristics (177 analyzed participants)

  • Median age 74 years (IQR 65–80 in the Results narrative; Table 1 gives 65–81); median NIHSS 18 (15–22); median baseline ASPECTS 4 (3–5)[1].
  • In the article's as-treated table, 84/177 were female (47.5%); Table 1 reports IV thrombolysis in 72/177 (40.7%), while the Results narrative says 73 (41%)[1].
  • Median baseline NWU was 16.1% (IQR 11.6–20.4) in Table 1; the Results narrative gives 16.4% (11.7–20.3)[1].

Interventions

  • EVT + BMT: Endovascular thrombectomy in addition to best medical treatment, as assigned/received in the parent trial[1].
  • BMT alone: Best medical treatment without EVT[1].
  • The secondary paper does not restate the complete procedural or medical-treatment protocols[1]. Follow-up CT was obtained at a mean 25 hours (median 24 hours) after admission; 90-day mRS assessment was performed by trained personnel blinded to allocation[1].

Outcomes

The table reports values as printed in the secondary analysis; ΔNWU is a median percentage-point change[1]. The article contains internal differences between the abstract, Results narrative and Table 1/2, retained below rather than silently reconciled[1].

Outcome BMT alone EVT + BMT Reported comparison
ΔNWU, median (IQR), percentage points 15.9 (9.7–21.7) 4.8 (−1.1 to 12.3) Table 1 P<.01; abstract says 15.9% vs 4.7%, P<.001[1]
ΔASPECTS, median (IQR) 2 (1–3) 2 (0–3) Table 1 P=.20; Results prose says P=.05[1]
90-day mRS, median (IQR) 6 (5–6) 5 (3–6) Table 1 P<.01[1]
Malignant infarction on follow-up CT 19 (21.1%) 18 (21.4%) Table 1 P=1.00; denominators are not displayed in this row[1]
EVT and 90-day mRS, unadjusted — Narrative cOR 0.39 (0.25–0.64), P<.001; Table 2 OR 0.30 (0.17–0.54), P<.01 Estimates differ between prose and table[1]
EVT and 90-day mRS, adjusted — Narrative acOR 0.34 (0.21–0.55), P<.001; Table 2 acOR 0.39 (0.19–0.79), P<.01 Adjusted estimates differ; adjustment descriptions also vary[1]
ΔNWU and worse 90-day mRS, unadjusted; per 1 percentage point — Narrative cOR 1.06 (1.03–1.09), P<.001; Table 2 OR 1.06 (1.03–1.10), P<.01 Estimates differ slightly between prose and table[1]
ΔNWU and worse 90-day mRS, adjusted; per 1 percentage point — Narrative acOR 1.07 (1.04–1.11), P<.001; Table 2 acOR 1.04 (1.01–1.08), P=.01 Adjusted estimates differ[1]
ΔASPECTS and worse 90-day mRS, unadjusted; per 1-point decline — Narrative cOR 1.48 (1.27–1.72), P<.001; Table 2 OR 1.55 (1.28–1.87), P<.01 Estimates differ between prose and table[1]
ΔASPECTS and worse 90-day mRS, adjusted; per 1-point decline — Narrative acOR 1.56 (1.33–1.83), P<.001; Table 2 acOR 1.66 (1.35–2.05), P<.01 Adjusted estimates differ[1]

Across the substudy, median follow-up NWU was 28.5% (IQR 19.2–36.6) in the abstract (Table 1 gives 28.6%); follow-up ASPECTS was 2 (1–3). Of the 85 patients who underwent EVT, 74 (87%) achieved mTICI 2b–3 reperfusion. Independent ambulation at 90 days (mRS 0–3) was reported in 35 participants (20%)[1].

The narrative and Table 2 do not present one reconciled set of ordinal estimates; the adjustment covariates also differ. In both, higher ΔNWU and a greater ΔASPECTS decline point toward worse 90-day mRS, while EVT points toward better outcome. These are associations in a selected as-treated cohort, not proof of mediation[1].

Mediation and other outcomes: EVT was associated with a 10.05-point lower ΔNWU (path a, P<.001)[1]. The ΔNWU mediation model estimated a total EVT effect of −0.93 mRS points and an indirect effect through ΔNWU of −0.32; the reported proportion associated with ΔNWU was 34%, versus 14% for ΔASPECTS[1]. A 5,000-resample bootstrap supported a more robust ΔNWU signal than ΔASPECTS[1]. In an exploratory malignant-infarction model, the indirect effect through ΔNWU was −0.09 (95% CI −0.17 to −0.03; P≈.007), while the direct effect was +0.10 (P=.16); this model result does not override the similar crude malignant-infarction proportions above[1].

Adverse Events / Safety: The article reports 9 symptomatic intracranial haemorrhages (5.3%) in the analyzed cohort; a denominator for that percentage is not stated alongside the count in the Results narrative[1]. It does not provide a separate mortality effect estimate in this substudy[1]. The absence of a fuller safety table here is not evidence of safety[1].

Figures

Graphical abstract of the TENSION secondary analysis linking thrombectomy, reduced oedema progression, and 90-day outcome
Graphical abstract. The authors summarize the 177-patient CT analysis and its model-estimated oedema-mediated association. The 34% figure is a mediation estimate, not an intervention effect of oedema treatment[1]. Source: Broocks et al., European Stroke Journal (2026), CC BY 4.0.
TENSION patient-inclusion flowchart showing the analyzed CT cohort and treatment allocation
Figure 1. Flow of the parent TENSION population into the CT substudy; 177 participants with analyzable baseline and follow-up CT formed the study cohort, and the figure also reports successful reperfusion among EVT-treated participants[1]. Source: Broocks et al., European Stroke Journal (2026), CC BY 4.0.
Four illustrative baseline and follow-up CT cases from EVT and BMT groups
Figure 2. Four illustrative cases (cases 1–2 EVT; cases 3–4 BMT) with baseline imaging and follow-up CT. The source caption reports ΔNWU of 2.0%, 1.5%, 20.4% and 19.6%, respectively; these examples are not group-level estimates[1]. Source: Broocks et al., European Stroke Journal (2026), CC BY 4.0.

Criticisms

  • CT eligibility removed 76 of 253 parent-trial participants; MRI-imaged patients were excluded, and the substudy used a modified as-treated analysis, leaving selection and generalizability concerns[1].
  • ΔNWU was measured after randomization[1]. Even with baseline adjustment and bootstrap sensitivity analysis, unmeasured mediator–outcome confounding and treatment-affected downstream care cannot be ruled out; the authors call the mediation hypothesis-generating[1].
  • Multiple comparisons were not adjusted[1]. The imaging biomarker requires specialized post-processing, and the paper says NWU is not a validated gatekeeping measure for EVT[1].
  • Several abstract/narrative/table values differ (including ΔNWU, ΔASPECTS and adjusted ordinal estimates); review the original tables and supplement before reusing a single value[1].

Funding

Supported by the EU Horizon 2020 research and innovation programme, Grant Agreement 754640[1]. The paper also reports author financial and professional disclosures; see the primary article for details[1].

The paper

  • Authors. Gabriel Broocks et al.
  • Title. Oedema reduction mediates thrombectomy benefit in large core stroke: secondary analysis of the TENSION trial.
  • Journal. European Stroke Journal.
  • Year. 2026.
  • DOI. 10.1093/esj/aakag055
  • PMCID / PMID. PMC13235715 / 42241572.
  • License. Creative Commons Attribution 4.0 International (CC BY 4.0).
  • Primary source. [1] https://www.ebi.ac.uk/europepmc/webservices/rest/PMC13235715/fullTextXML — Oedema reduction mediates thrombectomy benefit in large core stroke: secondary analysis of the TENSION trial.

Deep Dive — click to expand

What this is

This is a mechanistic secondary analysis nested in a randomized thrombectomy trial, asking whether reduced CT net water uptake helps explain the EVT–disability association in large-core stroke[1]. It connects treatment assignment, a post-treatment imaging mediator and 90-day mRS; it does not randomize oedema reduction itself[1].

1. Shadow Audit

The headline “34% mediated” can sound like one-third of EVT benefit is a proven oedema-treatment effect[1]. It is not[1]. The mediator was measured after treatment, the analysis was as-treated and CT-selected, and mediation depends on assumptions about mediator–outcome confounding that cannot be verified here[1]. The authors explicitly characterize the estimate as approximate and mechanistically informative, not a precise causal decomposition[1].

A second shadow is the mismatch between the biological story and the crude malignant-infarction counts: the table reports 19 (21.1%) with BMT and 18 (21.4%) with EVT, P=1.00, even though an exploratory mediation model estimates an indirect path[1]. This may motivate further study, but it is not evidence that the intervention reduced the observed event rate in this subgroup[1].

2. Inversion Engine

The causal interpretation would weaken if an independent, prospectively specified study with standardized serial CT found that ΔNWU no longer tracked functional outcome after accounting for treatment-affected care and other mediator–outcome confounders[1]. The present analysis cannot set a numeric clinical cutoff: it supplies no validated ΔNWU threshold for choosing treatment, and its 34% estimate is not an effect size for an oedema-targeted drug or procedure[1].

3. Second-Order Catalyst

The first change should be in research measurement, not the EVT pathway: collect reproducible baseline and follow-up NWU alongside extent measures and patient-centered outcomes, with prespecified handling of missing imaging and post-randomization factors[1]. The paper itself says the current workflow needs standardized, preferably automated validation and should not be used to select or deselect otherwise eligible patients[1].

4. Asymmetric Leverage

A selected cohort of 177 provides a useful randomized-trial context for a quantitative imaging signal, but the apparent 34% mediated proportion is much more dramatic than the raw comparison can justify[1]. The observed table-level ΔNWU medians were 15.9 percentage points with BMT and 4.8 with EVT; the outcome association is clinically interesting, while causal attribution to oedema remains model-sensitive[1].

5. Paradigm Destroyer

The paper challenges the idea that lesion extent alone captures all biologically relevant change after reperfusion[1]. It does not overturn treatment eligibility rules, prove that low-ASPECTS tissue is reversible, or establish that directly suppressing oedema improves disability[1]. A defensible update is to treat oedema dynamics as a candidate mechanism and research biomarker, not as a bedside gatekeeper[1].

MVP — Minimum Viable Proof

The minimum next proof is prospective replication using a prespecified CT-based NWU pipeline, transparent inclusion and intention-to-treat sensitivity analyses, and blinded 90-day functional assessment[1]. To claim that oedema itself is a therapeutic target, a separate intervention that changes oedema and improves patient-centered outcomes would be needed; this mediation analysis does not provide that test[1].

Best Combination

Pair the extent measure (ASPECTS) with a water-uptake trajectory (ΔNWU), while retaining ordinal mRS and safety outcomes as the clinical anchor[1]. Report the raw imaging distributions, model specification, missing-imaging flow and uncertainty together so the mechanistic signal cannot be mistaken for a treatment-selection rule[1].

Overvalue Warning

  1. Do not read 34% versus 14% as exact causal shares, or as evidence that oedema-directed treatment would recover 34% of EVT benefit[1].
  2. Do not use NWU to withhold EVT: this study establishes neither a validated threshold nor prospective clinical utility, and its adjusted estimates differ across the paper's narrative and tables[1].

Educational use notice. The Time Is Brain Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.