Edaravone dexborneol versus placebo on functional outcomes in patients with acute ischaemic stroke undergoing endovascular thrombectomy (TASTE-2): randomised controlled trial

Clinical Question (PICO)

Among adults aged 18–80 years with acute ischaemic stroke from anterior-circulation large-vessel occlusion, NIHSS 6–25 and ASPECTS 6–10, planned for endovascular thrombectomy within 24 hours, does intravenous edaravone dexborneol started before thrombectomy and continued for 10–14 days, versus matched placebo plus usual stroke care, improve 90-day functional independence (modified Rankin Scale [mRS] 0–2)? Prespecified subgroup analyses included clinical–ASPECTS mismatch profiles.

Bottom Line

In this investigator-initiated Chinese trial, edaravone dexborneol was associated with a borderline 5.4-percentage-point increase in 90-day functional independence (55.0% vs 49.6%; RR 1.11, 95% CI 1.00–1.23; P=0.05), with similar serious adverse-event rates. The overall effect was not corroborated by the ordinal mRS or other secondary efficacy outcomes; the larger mismatch-subgroup signal needs confirmation.

Design

Trial type: Investigator-initiated, multicentre, double-blind, randomised, placebo-controlled superiority trial.
N: 1,362 randomised.
- Edaravone dexborneol: 690.
- Placebo: 672.
Randomization: Central computer-generated permuted fixed-size blocks, 1:1; not stratified. Participants, investigators and outcome assessors were masked.
Setting: 106 hospitals in mainland China; participating centres had performed at least 50 thrombectomies in the preceding year.
Enrollment: 18 March 2022–17 February 2023 (Results section).
Mean follow-up: 90 days; outcomes assessed at 14 days/discharge and 90 days.
Analysis: Modified intention-to-treat efficacy population (1,360; excluding one lost to follow-up in each arm); logistic regression with risk ratios and risk differences. Multiple-imputation and covariate-adjusted sensitivity analyses were reported; safety used the full safety population.
Primary outcome: Functional independence, mRS 0–2 at 90 days. Primary safety outcome: serious adverse events during the study.

Population

Inclusion Criteria

Age 18–80 years; acute ischaemic stroke; NIHSS 6–25; ASPECTS 6–10.
Anterior-circulation occlusion at the intracranial internal carotid artery, T bifurcation, or M1 segment, with thrombectomy planned within 24 hours.
For 6–24-hour presentations, imaging eligibility was assessed against DEFUSE-3 or DAWN criteria.
Alteplase bridging therapy was allowed.

Exclusion Criteria

Intracranial haemorrhagic disease or allergy to study-drug components/excipients.
Systolic BP ≥180 mm Hg or diastolic BP ≥110 mm Hg despite antihypertensive treatment.
ALT or AST >3 times the upper limit of normal; chronic kidney disease, creatinine >1.5 times the upper limit of normal, or eGFR <60 mL/min.
A thrombolytic other than alteplase before thrombectomy.

Baseline Characteristics (overall)

Median age 67 years (IQR 57–73); 64.0% male; 98% Han ethnicity.
Median NIHSS 15 (IQR 12–19); median ASPECTS 9.
Successful reperfusion (eTICI 2B–3) in approximately 95% of treated patients.

Interventions

Edaravone dexborneol: Intravenous 37.5 mg (edaravone 30 mg plus (+)-dexborneol 7.5 mg) over 30 minutes, twice daily for 10–14 consecutive days. First dose was given before thrombectomy and no later than femoral-artery puncture.
Placebo: Visually matched intravenous placebo on the same schedule. Thrombectomy and guideline-based stroke-unit care were provided in both groups; other cytoprotective drugs were prohibited.

Outcomes

Primary Outcome (mRS 0–2 at 90 days):

Modified intention-to-treat: 379/689 (55.0%) with edaravone dexborneol vs 333/671 (49.6%) with placebo; RR 1.11 (95% CI 1.00–1.23), risk difference 5.4 percentage points (95% CI 0.1–10.7), P=0.05. The authors reported an NNT of 19. Multiple-imputation sensitivity analysis was similar (RR 1.11, 95% CI 1.01–1.21; risk difference 5.5 points, 95% CI 0.2–10.8; P=0.04).
Prespecified subgroup signal: In the clinical–ASPECTS mismatch subgroup defined as NIHSS ≥10 with ASPECTS ≥9, or NIHSS ≥20 with ASPECTS ≥7, functional independence was 178/321 (55.5%) vs 134/312 (42.9%); RR 1.29 (95% CI 1.10–1.52), risk difference 13.0 points (95% CI 5.6–20.3), P for interaction=0.003. The report also describes an interaction by baseline ASPECTS (8–10 vs 6–7). These are subgroup findings, not a separately powered primary comparison; the mismatch definition is clinical–ASPECTS, not a direct perfusion-imaging measurement of penumbra. Figures 3–4 show the prespecified subgroup analyses.

Secondary Outcomes:

No significant ordinal mRS shift (P=0.38); mRS 0–1 occurred in 40.2% vs 38.3% (RR 1.05, 95% CI 0.92–1.20; P=0.47).
Recurrent stroke: 1.5% vs 1.2% (RR 1.22, 95% CI 0.48–3.06; P=0.68). Mean NIHSS change at 10–14 days was −7.1 vs −7.0 (P=0.81); NIHSS reduction ≥4 points by day 10 was 73.1% vs 74.8% (RR 0.98, 95% CI 0.92–1.04; P=0.49).

Adverse Events / Safety:

Serious adverse events: 27.2% vs 25.7% (RR 1.06, 95% CI 0.89–1.26; P=0.53).
All-cause mortality at 90 days: 16.5% in each arm (RR 1.00, 95% CI 0.79–1.27; P>0.99).
Symptomatic intracranial haemorrhage within 36 hours: 5.3% vs 6.5% (RR 0.81, 95% CI 0.52–1.24; P=0.33).

Figures

Figure 1. Trial profile. EVT=endovascular thrombectomy.

Source: PMC PMC12776454 — wanc086850.f1.jpg. Click image to expand.

Distribution of 90-day modified Rankin Scale scores — *Figure 2.* Modified Rankin Scale score at 90 days. Scores on modified Rankin Scale are shown for patients in the edaravone dexborneol group (n=689) and placebo group (n=671). Scores range from 0 to 6, with 0 indicating no symptoms; 1, no clinically significant disability; 2, slight disability (able to handle own affairs without assistance but unable to carry out all previous activities); 3, moderate disability (requiring some help but able to walk unassisted); 4, moderately severe disability (unable to attend body needs and unable to walk); and 5, severe disability (requiring constant nursing care and attention).

Prespecified subgroup relative risks for the primary outcome — *Figure 3.* Primary outcome by prespecified subgroups (relative risks). Significantly greater likelihood of achieving functional independence with edaravone dexborneol, compared with placebo, was observed in predefined subgroups of ASPECTS and mismatch (defined as NIHSS score ≥10 and ASPECTS ≥9 or NIHSS score ≥20 and ASPECTS ≥7), with P for interaction <0.05. An interactive version of this graphic and downloadable data are available.

Prespecified subgroup risk differences for the primary outcome — *Figure 4.* Primary outcome by prespecified subgroups (risk differences). Significantly greater likelihood of achieving functional independence with edaravone dexborneol, compared with placebo, was observed in predefined subgroups of ASPECTS and mismatch (defined as NIHSS score ≥10 and ASPECTS ≥9 or NIHSS score ≥20 and ASPECTS ≥7), with P for interaction <0.05. An interactive version of this graphic and downloadable data are available.

Criticisms

The primary result is borderline (P=0.05; the risk-difference CI begins at 0.1 percentage points). Ordinal mRS, mRS 0–1, NIHSS change, recurrent stroke, mortality and the primary safety outcome did not show significant between-group differences.
The mismatch subgroup was defined from NIHSS and ASPECTS thresholds; it should not be re-described as a perfusion-imaging-confirmed penumbra subgroup. The authors tested multiple prespecified subgroups and mismatch definitions, so the interaction signal remains in need of prospective replication.
Generalisability is limited by a single-country population (98% Han), high-volume EVT centres and an approximately 95% eTICI 2B–3 reperfusion rate. Only 1,362 of 14,233 screened patients were enrolled (9.6%), largely reflecting treatment eligibility; the study also reports that a random centre effect was omitted because of convergence issues.
The sample-size calculation anticipated a 20% relative improvement, while the observed relative increase was about 11%, leaving the trial with less power for the smaller observed effect. The abstract gives an enrollment end date of May 2023, whereas the Results section reports 17 February 2023.

Funding

Supported by Simcere Pharmaceutical Group, China’s Ministry of Science and Technology, the National Natural Science Foundation of China, the CAMS Innovation Fund for Medical Sciences, and the Beijing Municipal Committee of Science and Technology. Simcere supplied the study drug and placebo; the authors state that funders had no role in study design, data collection, analysis or interpretation, writing, or the decision to submit. Registered at ClinicalTrials.gov as NCT05249920.

The paper

Authors. Wang C, et al.
Title. Edaravone dexborneol versus placebo on functional outcomes in patients with acute ischaemic stroke undergoing endovascular thrombectomy (TASTE-2): randomised controlled trial.
Journal. The BMJ.
Year. 2026.
DOI. 10.1136/bmj-2025-086850
PMCID. PMC12776454

Deep Dive — click to expand

What this is

TASTE-2 asks whether a cytoprotectant can add benefit after reperfusion has become routine for anterior-circulation large-vessel occlusion. The overall result is a small, borderline increase in 90-day mRS 0–2 after edaravone dexborneol plus thrombectomy; the more compelling 13-point difference is confined to a prespecified clinical–ASPECTS mismatch subgroup and is not yet a practice-changing estimate.

1. Shadow Audit

The headline “more likely to achieve functional independence” is accurate but can sound firmer than the data. The overall RR was 1.11 (95% CI 1.00–1.23; P=0.05), with a risk difference of 5.4 points (95% CI 0.1–10.7). The ordinal mRS distribution was not significantly different (P=0.38), nor were mRS 0–1 (P=0.47), NIHSS improvement, mortality, or serious adverse events. The primary signal therefore rests on one dichotomized threshold rather than a consistent shift across disability outcomes.

The subgroup language also needs precision: “mismatch” here is a clinical–ASPECTS rule (NIHSS plus ASPECTS), not direct measurement of salvageable tissue with perfusion imaging. It may identify a plausible treatment-responsive phenotype, but it does not prove a penumbra-freezing mechanism. The article reports a 13.0-point subgroup difference (95% CI 5.6–20.3; interaction P=0.003), within a trial that evaluated multiple prespecified subgroups and mismatch definitions.

2. Inversion Engine

The inferential threshold is no benefit: risk difference 0.0. The observed estimate is +5.4 percentage points, but the lower 95% confidence limit is only +0.1 points. A confirmatory estimate at or below zero—or a confidence interval that crosses zero—would erase the claim of overall superiority. Conversely, a reproducible benefit near the observed 5-point range, without added serious harm, would make the broad EVT population clinically relevant; the mismatch estimate should be tested separately rather than assumed to apply to everyone.

3. Second-Order Catalyst

The first change should be to research enrollment, not the routine thrombectomy medication order set. High-volume EVT centres can test the clinical–ASPECTS phenotype prospectively, use blinded central imaging review, and predefine both ordinal mRS and functional-independence endpoints. The signal is most actionable for trial designers now; the article does not establish routine edaravone dexborneol as standard adjunctive care.

4. Asymmetric Leverage

The broad-population estimate is modest—about five additional patients functionally independent per 100 treated at the point estimate—but could matter if it holds across the large denominator of thrombectomy practice. The mismatch subgroup’s reported 13-point difference is a larger possible payoff in a smaller, selected group. The asymmetry favors a focused confirmatory trial: a modest true effect could scale, while a false subgroup signal would otherwise expose many patients to an unproven infusion and add cost and workflow.

5. Paradigm Destroyer

This paper challenges the reflex that successful reperfusion is the end of acute treatment: a biologic adjunct may still be worth testing after the artery is reopened. It does not justify replacing that reflex with routine cytoprotection. Tomorrow’s protocol should add a trial-enrollment prompt for eligible anterior-circulation EVT patients with the prespecified clinical–ASPECTS profile, while keeping the drug off the standard order set outside a study.

MVP — Minimum Viable Proof

A second, independent randomized trial should reproduce a clinically meaningful increase in 90-day functional independence—approximately the 5.4-percentage-point overall estimate, with its confidence interval excluding no benefit—without an increase in serious adverse events or mortality. To change bedside practice on the subgroup claim, the 13-point mismatch effect must also survive prospective selection and independent validation.

Best Combination

TASTE-2 extends the edaravone-dexborneol question into the thrombectomy era; the earlier TASTE trial studied 1,165 patients who did not receive reperfusion therapy within 48 hours and reported mRS 0–2 in 67.18% versus 58.97% (OR 1.42, 95% CI 1.12–1.81). That earlier signal supports biological and clinical plausibility but is not a direct replication because its population and comparator differed. Read together, the trials justify testing cytoprotection as an adjunct to reperfusion—not substituting it for thrombectomy or extrapolating the earlier effect size to EVT patients.

Overvalue Warning

Do not convert “clinical–ASPECTS mismatch” into proof that perfusion-imaged penumbra was rescued; the subgroup rule was based on NIHSS and ASPECTS thresholds.
Do not present the reported NNT of 19 as settled efficacy. It is the point estimate from a borderline primary result, while disability-shift and other secondary outcomes were neutral and the population was selected from high-volume Chinese centres.

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.