← Back

DEFUSE 3

Thrombectomy for Stroke at 6 to 16 Hours with Selection by Perfusion Imaging

Share Share Copied! Compare

Year of Publication: 2018

Authors: Gregory W. Albers, Michael P. Marks, Stephanie Kemp, ..., for the DEFUSE 3 Investigators

Journal: New England Journal of Medicine

Citation: N Engl J Med 2018;378:708–18

Link: https://www.nejm.org/doi/full/10.1056/NEJMoa1713973

PDF: https://www.nejm.org/doi/pdf/10.1056/NEJMoa1713973

Clinical Question

Does endovascular thrombectomy improve outcomes compared to medical therapy alone in patients with anterior circulation large vessel occlusion 6–16 hours after last known well, selected by perfusion imaging?

Bottom Line

In patients with proximal anterior circulation occlusions and salvageable brain tissue on perfusion imaging, endovascular thrombectomy between 6–16 hours from last known well significantly improved functional outcomes and reduced mortality compared to medical therapy alone.

        Major Points
        Randomized controlled trial of thrombectomy vs. medical therapy in patients 6–16 hours after last known well with perfusion imaging evidence of salvageable brain tissue (ischemic core <70 mL, mismatch ratio ≥1.8, mismatch volume ≥15 mL assessed by RAPID software).
Eligible vessels: intracranial ICA or proximal MCA (M1 segment) on CTA or MRA. ICA occlusions were 35% of cohort, M1 MCA 65%.
Trial stopped early after 182 of planned 476 patients due to overwhelming efficacy at prespecified interim analysis.
Primary outcome (ordinal mRS shift at 90 days): OR 2.77 (95% CI 1.63–4.70, P<0.001) — thrombectomy shifted the entire mRS distribution favorably.
Functional independence (mRS 0–2) at 90 days: 45% vs 17% (RR 2.67, P<0.001, NNT ≈ 4).
Mortality: 14% vs 26% (P=0.05) — trend toward reduced mortality with thrombectomy.
No significant increase in sICH (7% vs 4%, P=0.75) or serious adverse events (43% vs 53%).
53% of patients had wake-up stroke; 75% selected by CT perfusion, 25% by MRI perfusion.
Median ischemic core volume was 9.4 mL (IQR 2.0–25.6); median perfusion lesion 114.7 mL. Median time from LKW to randomization ~10.9 hours.
Together with DAWN, led to AHA/ASA 2018 guideline update extending thrombectomy window to 16–24 hours for imaging-selected patients.

      

Design

Study Type: Randomized, open-label, blinded-endpoint controlled trial

Randomization: 1

Blinding: Blinded outcome assessment

Enrollment Period: May 2016 – May 2017

Follow-up Duration: 90 days

Centers: 38

Countries: United States

Sample Size: 182

Analysis: Ordinal logistic regression and stratified CMH test adjusted for stratification factors (age, NIHSS, core volume, time, site)

Inclusion Criteria

Age ≥18 years
Anterior circulation LVO (ICA or M1) on CTA/MRA
6–16 hours from last known well
Ischemic core <70 ml
Mismatch ratio ≥1.8 and mismatch volume ≥15 ml
mRS 0–2 before stroke

Exclusion Criteria

Large ischemic core ≥70 mL on RAPID automated perfusion analysis.
Mismatch ratio <1.8 or mismatch volume <15 mL.
Pre-stroke disability (mRS >2).
Intracranial hemorrhage on baseline imaging.
Posterior circulation occlusion (vertebral, basilar, PCA).
Isolated M2 or more distal MCA branch occlusion.
Known allergy to iodine contrast or history of contrast nephropathy.
Pregnancy or lactation.
Evidence of carotid dissection or high-grade cervical ICA stenosis requiring treatment.
Blood glucose <50 mg/dL.
Life expectancy <6 months from pre-existing conditions.
Participation in another interventional trial.

Baseline Characteristics

Mean Age: 70

Female %: 50

Comorbidities:

Other:
- IV tPA Use: 11
- Wake-Up Stroke: 53
- Witnessed Stroke Onset: 34

Symptom Onset <12h: No (all 6–16h; median ~10.5h)

NIHSS (Median): 16

ASPECTS (Median): 8

Ischemic Core Volume (Median): 9.4

Perfusion Lesion Volume (Median): 114.7

Imaging Modality:

CT Perfusion: 75
MRI Perfusion: 25

Occlusion Site:

ICA: 35
MCA (M1): 65

Arms

Field	Endovascular Therapy	Control
Intervention	Mechanical thrombectomy using any FDA-approved stent retriever or aspiration device (Solitaire, Trevo, Penumbra, etc.) at operator discretion, plus standard medical therapy per AHA/ASA guidelines including IV tPA if within window (only 11% received IV tPA due to late presentation). Goal: achieve mTICI 2b–3 reperfusion. Conscious sedation or general anesthesia per site protocol.	Standard medical therapy per AHA/ASA guidelines without endovascular intervention. Included IV tPA if within window, BP management, antithrombotics, and stroke unit care. Crossover to thrombectomy was not permitted.
Duration	Single procedure; target reperfusion within 16 hours of last known well	NA

Outcomes

Outcome	Type	Control	Intervention	HR / OR / RR	P-value
Ordinal shift in mRS at 90 days (0–6 scale), analyzed by ordinal logistic regression	Primary	Median mRS 4 (17% mRS 0–2)	Median mRS 3 (45% mRS 0–2)	28.00%	<0.001
Functional independence (mRS 0–2) at 90 days \| 95% CI: 1.60–4.48; NNT: 4	Secondary	17%	45%		<0.001
90-day mortality \| 95% CI: 0.30–1.02; NNT: 9; NNT_Note: Not statistically significant (P=0.05 exactly, 95% CI 0.30–1.02 crosses null). Trial stopped early and underpowered for this secondary endpoint.	Secondary	26%	14%		0.05
Symptomatic ICH	Adverse	4%	7%		0.75
Parenchymal hematoma type 2	Adverse	3%	9%		0.21
Serious Adverse Events	Adverse	53%	43%		0.18

Subgroup Analysis

Consistent benefit across all prespecified subgroups: age (<70 vs ≥70), sex, baseline NIHSS (≤17 vs >17), occlusion site (ICA vs M1 MCA), time from LKW (6–12h vs 12–16h), ASPECTS (≤7 vs >7), ischemic core volume (<10 mL vs ≥10 mL), imaging modality (CT perfusion vs MRI perfusion), and DAWN eligibility (met DAWN criteria vs did not). Treatment effect was present regardless of whether patients also met DAWN selection criteria, suggesting the perfusion mismatch approach captures a broader eligible population.

Criticisms

Stopped early after 182 of 476 planned patients — may overestimate treatment effect.
Open-label design may introduce bias despite blinded outcome assessment (PROBE design).
Generalizability limited to patients with access to RAPID perfusion imaging — not available at many hospitals, especially in lower-resource settings.
Only U.S. sites enrolled — limited international generalizability.
Small sample size (182) limits power for subgroup analyses.
Only 11% received IV tPA — different from early-window practice where >70% receive IV tPA before thrombectomy.
Posterior circulation excluded — cannot extrapolate to basilar artery occlusions in the late window.
RAPID software dependency — results may not be reproducible with other automated perfusion platforms.

Funding

National Institute of Neurological Disorders and Stroke (NINDS); RAPID software provided by iSchemaView

Based on: DEFUSE 3 (New England Journal of Medicine, 2018)