Tirofiban for Branch Atheromatous Disease–Related Stroke: The STRATEGY Randomized Clinical Trial

Clinical Question (PICO)

In adults aged 18–80 years with MRI-defined branch atheromatous disease (BAD)–related ischemic stroke, NIHSS ≤10, parent artery stenosis <70%, and no acute reperfusion treatment, does intravenous tirofiban started within 48 hours, added to aspirin, reduce early neurological deterioration (END) within 7 days or new stroke within 90 days compared with placebo plus aspirin?

Prespecified subgroup categories reported in Figure 3 were age, sex, hypertension, diabetes, dyslipidemia, parent artery stenosis, infarct diameter, anterior versus posterior circulation, and baseline NIHSS. These subgroup estimates are exploratory: the investigators state that interaction hypotheses were not prespecified and no multiplicity correction was applied.

Bottom Line

Tirofiban plus aspirin did not significantly reduce the primary composite: 79 events with tirofiban versus 89 with placebo, corresponding to reported Kaplan–Meier risks of 17.1% versus 19.6% (HR 0.88; 95% CI 0.65–1.19; P=.39). Moderate or severe bleeding was uncommon, with 1 versus 0 events; this does not establish safety equivalence. A favorable disability result and a stenosis-subgroup signal warrant confirmation rather than overturning the negative primary outcome.

Design

Trial type: Investigator-led, multicenter, randomized, double-blind, placebo-controlled superiority trial; aspirin was open-label.
N: 970 randomized and included in intention-to-treat analyses.
- Tirofiban plus aspirin: 486.
- Placebo plus aspirin: 484.
Randomization: Approximately 1:1, center-stratified blocks of 6, using an interactive web-response system and an independently generated sequence.
Masking: Identical tirofiban/placebo appearance and packaging; participants, investigators, outcome evaluators, and the clinical-events committee were blinded to assignment.
Setting: 38 hospitals in China.
Enrollment: November 15, 2022–November 19, 2024.
Follow-up: Scheduled through 90 days, not a reported mean follow-up. One tirofiban participant was lost to 90-day follow-up.
Analysis: Intention-to-treat; Cox models with center as a random effect for time-to-first-event outcomes, Kaplan–Meier cumulative risks, and last-contact censoring. Generalized linear models estimated secondary relative risks. Per-protocol analyses included 471 versus 468 participants and were reported as consistent with the primary analysis.
Primary outcome: END within 7 days or a new ischemic or hemorrhagic stroke within 90 days. END required a sustained ≥30-minute worsening of ≥2 NIHSS points or ≥1 point in motor function or consciousness, excluding intracranial hemorrhage and non-stroke causes of deterioration.
Power assumptions: Placebo risk 20% versus tirofiban risk 13.3%, 80% power, and two-sided α=.05. The trial was designed around a substantially larger effect than the observed separation.

Population

Inclusion Criteria

Age 18–80 years; acute ischemic stroke within 48 hours of symptom onset.
A single penetrating-artery territory infarction, with <70% ipsilateral parent artery stenosis.
Diffusion-weighted MRI lesion diameter <30 mm, with at least one of: involvement of ≥2 transverse slices; maximum diameter ≥15 mm; or a ventral pontine lesion reaching the surface without crossing the midline.
NIHSS ≤10 and NIHSS consciousness item 1a ≤1.

Exclusion Criteria

Intravenous thrombolysis or endovascular intervention after stroke onset.
Prior intracranial hemorrhage, intracranial tumor, vascular malformation, or aneurysm.
Preexisting disability, defined as prestroke mRS ≥2.
Dual antiplatelet therapy within 14 days before randomization, except concurrent aspirin and nonloaded clopidogrel initiated after stroke onset.
These are the principal exclusions reported in the main article, not an exhaustive reproduction of the supplementary eligibility list.

Baseline Characteristics (overall unless specified)

Median age 63 years (IQR 56–70); 62.6% male and 37.4% female.
97.3% Han Chinese.
Median NIHSS 3 (IQR 2–4); 60.9% had NIHSS ≤3.
Anterior circulation infarction 58.4%; posterior circulation 41.6%.
39.1% randomized within 24 hours; 60.9% between 24 and 48 hours.
Approximately 86%–87% of participants with available vascular imaging had parent artery stenosis <30%.
Parent artery stenosis data were missing for 13 tirofiban and 19 placebo participants.

Interventions

Tirofiban plus aspirin: Intravenous tirofiban 0.4 μg/kg/min for 30 minutes, followed by 0.1 μg/kg/min for 24 hours. Aspirin 300 mg on the randomization day, then 100 mg daily through day 90.
Placebo plus aspirin: Matched placebo infusion on the same schedule, plus the identical aspirin regimen.
What was not tested: Tirofiban versus a contemporary oral dual-antiplatelet regimen, treatment added to thrombolysis or thrombectomy, a longer tirofiban infusion, or a strategy restricted to patients with documented ongoing deterioration.

Outcomes

Primary Outcome — END within 7 days or new stroke within 90 days:

Overall: 79/486 versus 89/484 events; Kaplan–Meier risks 17.1% versus 19.6%; HR 0.88 (95% CI 0.65–1.19), P=.39.
END alone within 7 days: 71/486 (14.6%) versus 78/484 (16.1%); RR 0.91 (95% CI 0.55–1.48), P=.70.
New stroke within 90 days: 30 versus 35 events; Kaplan–Meier risks 7.0% versus 8.5%; HR 0.85 (95% CI 0.52–1.39), P=.52.

The primary composite and other specified time-to-event percentages are Kaplan–Meier estimates, not the event count divided by the randomized denominator. Other percentages below are raw proportions unless stated otherwise.

Prespecified subgroup categories, exploratory estimates — tirofiban versus placebo; HRs, 95% CIs, and nominal P values transcribed from Figure 3:

Category	Stratum	HR (95% CI)	Within-stratum P	Interaction P for category
Age	<60 years	0.79 (0.45–1.39)	.42	.61
Age	≥60 years	0.92 (0.64–1.32)	.64	—
Sex	Male	0.92 (0.63–1.35)	.68	.71
Sex	Female	0.82 (0.50–1.36)	.45	—
Hypertension	Yes	0.81 (0.55–1.19)	.28	.45
Hypertension	No	1.04 (0.63–1.71)	.87	—
Diabetes	Yes	1.03 (0.61–1.75)	.90	.47
Diabetes	No	0.81 (0.56–1.17)	.25	—
Dyslipidemia	Yes	1.08 (0.24–4.81)	.93	.90
Dyslipidemia	No	0.87 (0.64–1.18)	.37	—
Parent artery stenosis	0% to <30%	1.00 (0.72–1.40)	.99	.02
Parent artery stenosis	≥30% to <70%	0.37 (0.17–0.81)	.01	—
Maximum DWI infarct diameter	<15 mm	0.78 (0.52–1.16)	.22	.39
Maximum DWI infarct diameter	≥15 to <30 mm	1.02 (0.64–1.62)	.94	—
Location	Anterior circulation	0.84 (0.55–1.27)	.40	.62
Location	Posterior circulation	0.96 (0.61–1.49)	.84	—
NIHSS	≤3	0.80 (0.52–1.26)	.80*	.51
NIHSS	4–10	0.99 (0.65–1.50)	.95	—

*The NIHSS ≤3 row prints P=.80; that reported value is preserved rather than silently recalculated. It should be checked against the underlying analysis before secondary use. In the ≥30% to <70% stenosis stratum, event counts were 9/66 versus 21/60. Figure 3 prints Kaplan–Meier risks of 13.6% versus 51.0%; the placebo raw proportion is 35.0%, a different quantity. Neither those cumulative-risk estimates nor the nominal interaction P=.02 establish a validated treatment-selection rule. The authors explicitly characterize the stenosis finding as exploratory, without prespecified interaction hypotheses.

Secondary Outcomes:

END or new stroke within 24 hours: 12.6% versus 14.5%; RR 0.87 (95% CI 0.57–1.31), P=.50.
END or new stroke within 7 days: 14.6% versus 16.3%; RR 0.90 (95% CI 0.54–1.47), P=.66.
Composite vascular events within 90 days: 36 versus 38 events; Kaplan–Meier risks 8.2% versus 9.1%; HR 0.94 (95% CI 0.60–1.49), P=.80.
Disability or death, mRS 2–6 at 90 days: 144/485 (29.7%) versus 171/484 (35.3%); RR 0.84 (95% CI 0.73–0.97), nominal P=.02. This was a prespecified secondary outcome with no multiplicity adjustment, not a positive primary outcome.
Post-hoc ordinal mRS analysis: Common OR 0.82 (95% CI 0.66–1.04); the CI crossed the null. This analysis was post hoc and does not independently confirm a functional benefit.
Neurological improvement: Defined as ≥4-point NIHSS improvement or NIHSS 0–1. At 24 hours, RR 1.04 (95% CI 0.85–1.26), P=.72; at 7 days, RR 1.01 (0.91–1.13), P=.83; at 90 days, RR 1.02 (0.93–1.11), P=.69.
Quality of life at 90 days: Mean EQ-5D-5L 0.88 versus 0.87; β 0.01 (95% CI −0.01 to 0.03), P=.51. One tirofiban participant lacked both 90-day mRS and EQ-5D-5L data.

Adverse Events / Safety:

GUSTO moderate or severe bleeding: 1/486 (0.2%) versus 0/484; P>.99. A hazard ratio was not estimable/reported.
Intracranial hemorrhage: 2/486 (0.4%) versus 0/484; P=.50. The two tirofiban events comprised one symptomatic and one asymptomatic hemorrhage.
Vascular death: 2 versus 3; HR 0.67 (95% CI 0.11–3.99), P=.66.
All-cause death: 3 versus 4; HR 0.75 (95% CI 0.17–3.35), P=.71.
Any adverse event: 24.3% versus 19.8%; serious adverse events 3.5% versus 2.9%.
Repeat CT was not mandatory in clinically stable participants, so asymptomatic hemorrhage may have been underdetected. The small number of major bleeds limits what the trial can establish about rare harms.

Figures

*Figure 1.* Flowchart of Patient Enrollment and Randomization. The intention-to-treat analysis encompassed patients with all patients with randomization. AE indicates adverse event; MRI, magnetic resonance imaging; mRS, modified Rankin Scale.

*Figure 2.* Kaplan-Meier Curves of the Cumulative Incidence of Early Neurologic Deterioration or New Stroke (Primary Efficacy End Point) Within 90 Days. The inset shows the same data on an enlarged y-axis. HR indicates hazard ratio.

*Figure 3.* Forest Plots of the Effects of Tirofiban With Aspirin as Compared With Placebo With Aspirin for the Primary Efficacy End Point in Prespecified Subgroups. The event rates of early neurologic deterioration or new stroke within 90 days are Kaplan-Meier estimates of the percentage of patients with events. The trial was not powered to allow definite conclusions based on the results of the subgroup analyses. The size of each box is proportional to the number of patients in the subgroup. Data on the degree of parent vessel stenosis were missing for 13 patients in tirofiban-aspirin group and for 19 patients in the placebo-aspirin group. DWI indicates diffusion-weighted imaging; HR, hazard ratio; NIHSS, National Institutes of Health Stroke Scale.

Criticisms

Negative superiority is not equivalence. The primary HR confidence interval includes both a clinically meaningful reduction and an increase in hazard; a modest benefit remains possible, but unproven.
Secondary-outcome rescue is hazardous. The favorable dichotomized mRS result was not multiplicity-adjusted, while the post-hoc full-distribution analysis did not exclude no effect.
The stenosis subgroup is small. Only 66 versus 60 participants had ≥30% to <70% stenosis, with missing stenosis measurements in another 32 participants overall. Its attractive effect estimate requires external confirmation, not selective prescribing from a forest plot.
Imaging phenotype is not pathology. MRI enrollment criteria improve consistency but may still mix true parent-artery BAD with lacunar mechanisms. Most participants had little luminal stenosis; luminal measurements alone do not establish culprit plaque characteristics.
Timing and duration remain unresolved. Most participants entered after 24 hours, while deterioration clustered within the first 72 hours of onset; the tested regimen was a 30-minute loading infusion followed by 24 hours of maintenance. Earlier or longer administration is a new hypothesis, not a proven explanation of the result.
Comparator and setting limit transportability. This was aspirin-based treatment in predominantly Han Chinese patients with mild stroke, without reperfusion therapy or substantial prestroke disability. The results do not establish comparative effectiveness against oral dual antiplatelet therapy, or applicability to older adults above 80, severe stroke, or post-thrombectomy care.
Minimal follow-up loss does not eliminate other uncertainty. Only one participant lacked 90-day assessment, but imaging missingness, rare safety events, and the mixed END/new-stroke composite remain relevant.

Funding

Funding came from the National Natural Science Foundation of China (82425101); Noncommunicable Chronic Diseases–National Science and Technology Major Project (2023ZD0504800, 2023ZD0504801, 2023ZD0504802, 2023ZD0504803, 2023ZD0504804); Beijing Scholar Program (120); Beijing Municipal Science & Technology Commission (Z231100004823036); and Capital’s Funds for Health Improvement and Research (2022-2-2045).

Aspirin was purchased from Bayer Schering Pharma. Grand Pharma (China) supplied tirofiban and placebo. The article states that funders had no role in design, data collection, analysis, interpretation, or manuscript preparation; investigators conducted the trial and the national neurological research center performed the analyses. Bath and Turc disclosed outside advisory, consulting, lecture, or nonfinancial relationships; Yilong Wang disclosed the study-related grants. Registration: NCT05310968.

The paper

Authors. Yicong Wang, Xiaoling Liao, Shuo Feng, et al.; STRATEGY Investigators.
Title. Tirofiban for Branch Atheromatous Disease–Related Stroke: The STRATEGY Randomized Clinical Trial.
Journal. JAMA Neurology. 83(10):956–964.
Year. 2026; published online August 31, 2026, with the October 2026 issue. This is a look-back review, not a claim of publication today.
DOI. 10.1001/jamaneurol.2026.2855
PMCID. PMC13531381
PMID. 42671865

Deep Dive — click to expand

What this is

STRATEGY addresses a familiar ward problem: a relatively mild perforator stroke that worsens after admission. Its 2026 report tests a plausible platelet-based solution in a deliberately MRI-defined population, rather than importing evidence from large-vessel occlusion or thrombectomy. The blinded, 970-patient experiment did not establish that adding this short tirofiban regimen to aspirin prevents early worsening or subsequent stroke. The commentary below is interpretation of this trial, not a statement of current guideline recommendations.

1. Shadow Audit

The seductive finding is not the primary one. Disability or death was 29.7% versus 35.3%, nominal P=.02, and the stenosis subgroup produced HR 0.37. Either could dominate a presentation while pushing the actual primary result—HR 0.88, P=.39—into the background. That would reverse the trial's evidentiary hierarchy. The favorable disability dichotomy also sits beside a post-hoc ordinal mRS CI crossing the null and neutral neurological-improvement and quality-of-life findings.

There is a quieter mechanistic problem: most participants had <30% parent artery stenosis, and 60.9% were randomized after 24 hours. A label suggesting atherosclerosis does not prove a platelet-responsive lesion at the moment treatment begins. The study cannot distinguish a genuinely ineffective treatment from an effective treatment diluted across heterogeneous mechanisms or administered outside its optimal window. Those are alternatives to test, not excuses to declare efficacy.

2. Inversion Engine

To conclude that this is a broadly effective preventive regimen, a confirmatory study would have to move beyond a nominally favorable point estimate and exclude HR 1. The observed primary HR CI of 0.65–1.19 spans a 35% lower to a 19% higher hazard; it does not identify which side is true. The sample-size calculation targeted risks of 20% versus 13.3%, a 6.7-percentage-point separation, whereas the reported primary Kaplan–Meier risks separated by 2.5 points.

The opposite overstatement—“tirofiban cannot help BAD”—also fails. A clinically relevant reduction remains inside the interval. If the narrower hypothesis is benefit in ≥30% to <70% stenosis, it needs a prospectively specified interaction, reproducible imaging classification, adequate numbers, and a patient-centered outcome. The present 126-person subgroup and unadjusted interaction P=.02 cannot do that work alone.

3. Second-Order Catalyst

The first useful change is an evidence-review change, not a new infusion order. A stroke unit already using empiric tirofiban for broadly labeled BAD should explicitly separate routine use from research or exceptional individualized decisions and review its local protocol through the usual clinical governance process. This paper does not demonstrate that an infusion should replace reperfusion assessment or that aspirin alone is the preferred regimen for every patient in current practice.

The research sites best placed to move first are centers that can rapidly document lesion geometry, parent artery stenosis, and onset-to-treatment time. Their next protocol can distinguish treatment of established deterioration from prevention before deterioration, and can test whether plaque phenotype or timing actually modifies response. Starting earlier or treating longer remains experimental here; the trial did not randomize either choice.

4. Asymmetric Leverage

The potentially large payoff is finding a reliably responsive subgroup without exposing the much larger unselected population to an unproven infusion. But the subgroup signal is precisely where statistical fragility is greatest. The ≥30% stenosis estimate rests on 9 versus 21 events, not on an independently replicated treatment rule.

For scale only, the primary Kaplan–Meier risk difference of 2.5 percentage points corresponds to an illustrative reciprocal of 40 treated per event avoided if that difference were causal and reproducible. It is not an established number needed to treat because the primary effect remains uncertain. The secondary disability difference of 5.6 points gives an equally illustrative reciprocal of about 18, but carries the additional multiplicity problem. Neither should be placed in a protocol as a proven benefit. The asymmetric practical gain today is avoiding certainty that the data did not buy.

5. Paradigm Destroyer

The reflex this paper should retire is: “The perforator stroke may worsen, therefore more intensive platelet inhibition must help.” Biological plausibility earned a trial; it did not earn a positive result.

A defensible two-sentence addition to a local evidence summary is: “STRATEGY did not establish prevention of early neurological deterioration or new stroke with a short tirofiban infusion added to aspirin in MRI-defined BAD without reperfusion therapy. Do not treat its secondary disability result or exploratory stenosis subgroup as an established indication; retain standard acute-stroke assessment and require explicit evidence review before changing local treatment policy.”

MVP — Minimum Viable Proof

For broad adoption of this specific strategy, the minimum proof would be a prospectively defined, adequately powered randomized comparison showing a clinically meaningful patient-centered benefit, with uncertainty excluding no benefit and enough safety events to judge hemorrhage. For an imaging-selected strategy, that proof must apply specifically to a reproducibly defined stenosis/plaque subgroup rather than to a favorable slice of a negative primary analysis.

STRATEGY supplies neither proof yet. It supplies a well-defined negative primary trial and testable timing, mechanism, and subgroup hypotheses.

Best Combination

Read STRATEGY alongside the distinct clinical questions in its cited literature, not as a referendum on all intensive antiplatelet therapy. The paper cites CHANCE and THALES subgroup work on oral dual-antiplatelet treatment and atherosclerotic phenotype, and RESCUE BT on tirofiban before thrombectomy. Different drugs, background treatments, vascular mechanisms, and reperfusion settings prevent a simple transfer of their conclusions into BAD.

The useful combination is an evidence map: MRI-defined, non-reperfused perforator stroke in STRATEGY; oral-agent strategies in their own eligible populations; and endovascular-treatment studies in large-vessel occlusion. Its references identify the CHANCE intracranial-stenosis analysis, THALES atherosclerotic-origin analysis, and RESCUE BT trial. These are contextual references verified in STRATEGY's bibliography, not a new independent systematic review or a claim of current guideline endorsement.

Overvalue Warning

Do not promote the secondary or subgroup finding into the headline conclusion. Disability P=.02 and the stenosis interaction P=.02 are nominal, unadjusted signals; the main efficacy comparison was negative and the trial was not powered to settle subgroup benefit.
Do not translate scarce bleeding into demonstrated safety, or this comparator into a universal treatment choice. One versus zero major bleeding events leaves important uncertainty; aspirin-based care in this selected non-reperfused cohort does not answer tirofiban versus oral dual therapy, post-thrombolysis treatment, or post-thrombectomy management.

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.