DAWN & DEFUSE 3 perfusion-based selection for mechanical thrombectomy 6–24 hours after last known well
For two decades, "time is brain" meant a hard 4.5-hour window for thrombolysis and a 6-hour window for thrombectomy. But roughly 25% of ischemic strokes are unwitnessed — wake-up strokes, unattended patients — leaving the true onset time unknown. The DAWN and DEFUSE 3 trials (2018) transformed stroke care by proving that carefully selected patients presenting 6–24 hours after last known well benefit enormously from thrombectomy, when advanced perfusion imaging shows tissue is still salvageable. Time is no longer the sole gatekeeper — tissue physiology is.
Historically, thrombectomy trials (MR CLEAN, ESCAPE, SWIFT PRIME, EXTEND-IA, REVASCAT — the 2015 "big five") enrolled patients within 6 hours of onset, using time as a crude proxy for tissue viability. This left a large population of eligible-seeming patients excluded purely because their exact onset time was unknown or exceeded 6 hours — even though many still harbored salvageable brain tissue.
The core insight driving DAWN and DEFUSE 3: infarct growth rate varies dramatically between patients depending on collateral circulation. A patient with robust leptomeningeal collaterals may have a small core and large penumbra at 20 hours; a patient with poor collaterals may have a large, unsalvageable core within 3 hours. Fixed time windows systematically both over-treat fast progressors and under-treat slow progressors.
Advanced imaging (CT perfusion or diffusion-perfusion MRI mismatch) directly visualizes the actual state of brain tissue rather than inferring it from a clock. This allows individualized, physiology-based patient selection — extending the population who can benefit from endovascular therapy well past the traditional 6-hour cutoff.
DAWN and DEFUSE 3 were stopped early for overwhelming efficacy by their data safety monitoring boards — the benefit of thrombectomy in carefully selected extended-window patients was so large it was considered unethical to continue randomizing patients to the medical-management-only arm.
Patient populations eligible for extended-window evaluation include:
• Wake-up strokes: patient was normal at bedtime, found with deficits on waking — last known well is bedtime, often 6–8 hours prior, but true onset could be anywhere in that interval • Unwitnessed strokes: found down at home, in a care facility, or in public with no witness to onset • Delayed presentation: known onset time but delayed access to care (rural transfer, symptom under-recognition, initial misdiagnosis) • Transferred patients: initial imaging at a non-thrombectomy-capable hospital, followed by transfer to a comprehensive stroke center — transit time alone can push a 5-hour-onset patient past the 6-hour mark
All of these patients were previously considered ineligible for reperfusion therapy under time-only protocols. The extended window criteria specifically target large-vessel occlusion (LVO) strokes — internal carotid artery or proximal (M1) middle cerebral artery occlusions — where the volume of tissue at risk is largest and the potential benefit of recanalization is greatest.
Perfusion imaging is the technology that makes extended-window selection possible. By measuring cerebral blood flow and blood volume at a voxel level, software algorithms partition the affected territory into two zones: the infarct core (tissue already dead, low cerebral blood volume) and the penumbra (tissue hypoperfused but still viable, at risk of infarction without reperfusion). This core-penumbra distinction is the physiological basis for both DAWN and DEFUSE 3 selection algorithms.
CT perfusion imaging tracks a bolus of iodinated contrast as it passes through brain tissue, generating time-density curves at each voxel. From these curves, software derives several parametric maps:
• Cerebral Blood Flow (CBF): rate of blood delivery per unit tissue • Cerebral Blood Volume (CBV): total blood volume per unit tissue • Mean Transit Time (MTT) and Tmax: time delay of contrast arrival, reflecting collateral-dependent flow delay
Automated platforms (RAPID — used in DAWN and DEFUSE 3; Viz.ai; e-Stroke) apply validated thresholds:
• Infarct core = tissue with relative CBF <30% of normal contralateral tissue (severely hypoperfused, presumed dead) • Penumbra/at-risk tissue = tissue with Tmax delay >6 seconds (hypoperfused but not yet infarcted) • Mismatch volume = penumbra volume minus core volume • Mismatch ratio = (core + penumbra) / core
Processing takes roughly 5–7 minutes from scan completion to a color-coded map delivered to the treating physician's phone or workstation — a critical efficiency gain that keeps door-to-puncture times short even as sophisticated analysis occurs.
DEFUSE 3 (and its predecessor DEFUSE) also permitted MRI-based selection using diffusion-perfusion mismatch:
• Diffusion-Weighted Imaging (DWI): restricted water diffusion in cytotoxic edema identifies the infarct core with high sensitivity within minutes of ischemia onset — the "core" lesion appears as a bright signal on DWI with corresponding low apparent diffusion coefficient (ADC) • Perfusion-Weighted Imaging (PWI): using either dynamic susceptibility contrast (gadolinium bolus) or arterial spin labeling (no contrast needed), PWI identifies the full territory of hypoperfusion (Tmax >6 sec), analogous to CTP • DWI-PWI mismatch = PWI lesion volume minus DWI lesion volume — this represents tissue at risk
MRI offers superior soft-tissue contrast and can better distinguish core from penumbra in some cases, plus it avoids iodinated contrast and ionizing radiation. However, CTP is faster to acquire, more widely available in emergency settings, and compatible with patients who have MRI contraindications (pacemakers, claustrophobia, agitation) — making CTP the dominant modality in most comprehensive stroke centers today.
The DAWN trial (DWI or CTP Assessment with Clinical Mismatch in the Triage of Wake-Up and Late Presenting Strokes, NEJM 2018) enrolled patients 6–24 hours from last known well using a "clinical-core mismatch" concept: a large neurological deficit (high NIHSS) paired with a surprisingly small infarct core implies a large volume of at-risk, salvageable tissue that clinical exam alone could not otherwise quantify.
DAWN used an age-stratified table pairing NIHSS score with maximum allowable infarct core volume (measured by RAPID software on CTP or MRI-DWI):
Group A — Age ≥80 years: • NIHSS ≥10 AND core volume <21 mL
Group B — Age <80 years: • NIHSS ≥10 AND core volume <31 mL • NIHSS ≥20 AND core volume 31–<51 mL
The logic: the higher the clinical deficit (NIHSS), the larger a core volume is still considered "mismatched" — because a severe deficit with even a moderate core implies a large penumbra is driving the clinical picture, so more tissue remains salvageable in absolute terms. Conversely, low NIHSS scores require very small cores to qualify, since a low deficit with a large core would imply little viable at-risk tissue remains.
This simulator uses a simplified single age-tier proxy of this logic for illustrative purposes: NIHSS ≥10 with core <21 mL, NIHSS ≥20 with core <31 mL, and an extended tier of NIHSS ≥20 with core <51 mL — approximating the qualitative shape of the DAWN decision matrix without reproducing the full age-stratified trial table.
DAWN's results were extraordinary: 49% of thrombectomy patients achieved functional independence (mRS 0–2) at 90 days versus just 13% with medical management alone — a 36 percentage-point absolute difference, among the largest treatment effects ever observed in a stroke trial.
The clinical-core mismatch concept exploits a simple physiological relationship: NIHSS score correlates with the total volume of brain tissue currently malfunctioning (core + penumbra + diaschisis), while the CTP/MRI core reflects only the tissue already infarcted. When these two numbers diverge — severe clinical deficit but small measured core — the gap must be explained by a large volume of penumbral tissue that is functionally impaired but not yet dead.
This approach has an elegant advantage: it requires no assumption about mismatch ratio or absolute penumbra volume from perfusion imaging alone — it cross-validates the imaging-derived core against an independent clinical measure. A patient can qualify for DAWN even without formal PWI/CTP quantification of penumbra, as long as core volume and NIHSS are known.
Age stratification exists because older patients tolerate smaller infarct cores due to reduced brain reserve and higher likelihood of poor outcome even with successful reperfusion of a similarly sized core — hence tighter (smaller) core cutoffs are required to justify intervention risk in patients ≥80.
DEFUSE 3 (Endovascular Therapy Following Imaging Evaluation for Ischemic Stroke, NEJM 2018) enrolled patients 6–16 hours from last known well using a purely perfusion-based mismatch profile — no clinical severity score required. Its three quantitative imaging criteria (mismatch ratio, absolute mismatch volume, and maximum core size) define what is now widely called a "target mismatch" profile, favorable for thrombectomy.
DEFUSE 3 target mismatch profile requires ALL THREE of the following, computed automatically by RAPID software from CTP or MRI DWI/PWI:
1. Initial infarct core volume <70 mL • Rationale: cores beyond 70 mL carry high hemorrhagic transformation risk after reperfusion and poor functional outcome regardless of recanalization success — similar to the ANGEL-ASPECT and SELECT2 large-core trials' rationale, but DEFUSE 3 used this as a hard exclusion rather than testing benefit in large cores
2. Ratio of perfusion lesion volume to core infarct volume ≥1.8 • This ratio quantifies relative mismatch — how many times larger the total at-risk territory (core + penumbra) is compared to the core alone • A ratio near 1.0 implies almost all hypoperfused tissue is already infarcted (little to salvage); a high ratio implies a large penumbra relative to a small core
3. Absolute volume of hypoperfused tissue potentially salvageable (mismatch volume) ≥15 mL • This guards against trivial ratios in very small lesions — a 2 mL core with a 4 mL penumbra gives a ratio of 3.0 but only 2 mL of absolute salvageable tissue, not clinically meaningful • The 15 mL absolute threshold ensures a clinically relevant volume of tissue is actually at stake
All three criteria must be satisfied simultaneously — the ratio and absolute thresholds work together to exclude both "too little to matter" and "too much already lost" scenarios.
DEFUSE 3 was also stopped early for efficacy: 45% of thrombectomy-treated patients achieved functional independence at 90 days versus 17% with medical management — and thrombectomy reduced mortality (14% vs 26%) without increasing symptomatic intracranial hemorrhage.
Though DAWN and DEFUSE 3 are often cited together, their selection logic differs meaningfully:
• DAWN incorporates clinical severity (NIHSS) as a core variable — it can select patients even when precise perfusion mismatch ratio isn't available, relying on the clinical-imaging discordance principle • DEFUSE 3 is purely imaging-based — no NIHSS threshold — relying entirely on quantitative perfusion mismatch, making it more reproducible across raters and imaging platforms but dependent on having reliable perfusion post-processing • DAWN's window extends to 24 hours; DEFUSE 3's trial window was capped at 16 hours (though many centers now apply DEFUSE 3-style criteria off-label out to 24 hours in practice, per AHA/ASA 2019 guideline updates) • A patient may meet one trial's criteria without meeting the other's — in contemporary practice, meeting EITHER DAWN or DEFUSE 3 criteria is generally considered sufficient to proceed with thrombectomy in the extended window, which is why both are incorporated together into the 2019 AHA/ASA acute stroke guidelines (Class I, Level of Evidence A)
The culmination of extended-window evaluation is a binary treatment decision informed by the mismatch analysis: if DAWN or DEFUSE 3 criteria are satisfied, mechanical thrombectomy proceeds regardless of exact time since onset, up to 24 hours. If neither criterion is met, the imaging indicates the infarct core has already consumed most of the at-risk tissue — reperfusion offers little functional benefit and disproportionately raises hemorrhagic transformation risk.
The practical extended-window decision algorithm, as implemented in comprehensive stroke centers:
1. Confirm large-vessel occlusion (ICA or M1-MCA) on CTA/MRA 2. Confirm last known well is 6–24 hours prior (or unwitnessed with a plausible window) 3. Obtain CTP or MRI DWI/PWI with automated core/penumbra quantification 4. Evaluate against DAWN criteria (NIHSS + age-stratified core threshold) 5. Evaluate against DEFUSE 3 criteria (mismatch ratio ≥1.8, absolute mismatch ≥15 mL, core <70 mL) 6. If EITHER criterion is met → proceed to thrombectomy 7. If NEITHER criterion is met → medical management (antiplatelet/anticoagulation per etiology, risk factor control) without endovascular intervention
This simulator's live metrics implement a simplified version of steps 4–6: adjusting time since last known well, core volume, and NIHSS score updates both mismatch determinations and the overall eligibility verdict in real time, illustrating how the same core volume can be "too much" at one clinical severity level yet acceptable at another, or how a fixed core becomes progressively less favorable as the assumed penumbra shrinks with time.
A useful clinical mental model: DAWN answers "does the clinical picture look worse than the visible damage?" while DEFUSE 3 answers "is there still meaningfully more at-risk tissue than dead tissue?" Meeting either question affirmatively justifies attempting reperfusion in the extended window.
Extended-window selection via DAWN/DEFUSE 3 criteria has important caveats:
• Both trials excluded patients with very large cores (≥70 mL to ≥51 mL depending on tier) — but subsequent large-core trials (SELECT2, ANGEL-ASPECT, RESCUE-Japan LIMIT, 2023) showed thrombectomy still benefits some large-core patients, suggesting the DAWN/DEFUSE 3 core ceilings may be overly conservative for a subset of patients • Automated perfusion software (RAPID, Viz.ai) can disagree with each other and with expert visual assessment by meaningful margins, especially in patients with poor collateral flow or motion artifact • The trials enrolled highly selected populations at expert centers — real-world generalizability, especially at centers with less perfusion imaging experience, may show smaller effect sizes • Time-based penumbra decay is not linear or uniform across patients — collateral status, blood pressure, and comorbidities modulate the actual rate of core growth, meaning two patients at the same clock-time since onset can have very different tissue fates • Posterior circulation (basilar artery) occlusions were not well represented in DAWN/DEFUSE 3, and extended-window selection criteria for basilar occlusion remain less standardized (see BASICS, BAOCHE, ATTENTION trials)
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| DAWN | 6–24 h from LKW | Clinical-core mismatch: NIHSS vs. age-stratified core volume cutoff (RAPID CTP/MRI) | 49% functional independence vs 13% control |
| DEFUSE 3 | 6–16 h from LKW | Perfusion-core mismatch: ratio ≥1.8, absolute mismatch ≥15 mL, core <70 mL | 45% functional independence vs 17% control |
| Combined practice (2019 AHA/ASA) | 6–24 h from LKW | Meeting either DAWN or DEFUSE 3 criteria qualifies for thrombectomy | Class I, Level of Evidence A recommendation |