Defining the Therapeutic Ceiling of Endovascular Thrombectomy in Large-Core Stroke — Beyond the Limits of ASPECTS

What this is

A Korean nationwide multicenter registry of 552 EVT-treated anterior circulation LVO patients (2022–2024) plus embedded target trial emulations asking whether the ASPECTS <6 label — which gates every major large-core EVT trial — actually identifies the right patients, and where the upper edge of EVT benefit truly sits. The headline answer is uncomfortable: ASPECTS-only large-core patients (low score but small volume) do about as well as small-core patients, while volumetrically confirmed large-core patients do badly regardless of ASPECTS, and EVT benefit on the ordinal 90-day mRS disappears above roughly 110 mL of DWI core. The ceiling is real, but it isn’t where ASPECTS says it is.

1. Shadow Audit

The largest unwritten finding is who is not in the large-core ceiling — the ASPECTS-only large-core group. Their poor-outcome proportion (~11.8%) is statistically indistinguishable from small-core patients (~11.7%), with no excess risk after adjustment. Practically, this means the ~30% of “large-core” trial enrollees whose ASPECTS <6 comes from scattered territorial lesions rather than true high-volume infarction are being pulled into a high-risk, low-benefit bucket every time we use ASPECTS as the gate. The paper frames this as a “large-core paradox” — the group labeled large core is partly composed of patients who biologically are not.

A second shadow: the 110 mL ceiling is on DWI volume, but the same patients in this cohort had a median CTP core of 18.6 mL and a median NCCT core of 3.1 mL. The “ceiling” is therefore a moving target depending on which modality and which segmentation algorithm you trust. Anyone porting 110 mL into a non-DWI workflow is borrowing a number from a different imaging dialect without translating it.

A third: the ceiling was derived in a population where median LKW-to-groin was 4.2 hours. At longer windows (12–24 hours) the salvageable penumbra shrinks even at lower volumes, and the ceiling likely moves down. The paper doesn’t stratify the 110 mL finding by treatment window — a real omission given how aggressively late-window EVT has expanded.

2. Inversion Engine

For the conclusion to flip — i.e., for ASPECTS to regain primary prognostic value and the volumetric ceiling to disappear — three things would have to be true at once:

1. ASPECTS would need to track volume more tightly. The current correlation is only moderate (the paper cites refs 14, 15). For ASPECTS <6 to be independently prognostic in multivariate models that include volume, ASPECTS would need to capture something volume does not — perhaps eloquent-cortex territory weighting that matters for quality of survival even when total volume is similar. That’s biologically defensible (a 50 mL stroke in the dominant M1 stem is different from 50 mL scattered), but the current data say it doesn’t move the needle on mRS 5–6.

2. The 110 mL ceiling would have to be artefactual. To invert, you’d need a subgroup where EVT still helps at 150+ mL of DWI core. ANGEL-ASPECT and RESCUE-Japan LIMIT both enrolled ASPECTS 3–5 patients with relatively high volumes, and subgroup analyses there did not show consistent benefit at the top of the volume range — which is consistent with, not contradictory to, the 110 mL finding.

3. The nonrandomized causal estimate would have to be biased. The cOR 0.51 (95% CI 0.27–0.96) in the ASPECTS-defined emulation is real, but the ceiling finding in the DWI emulation depends on a much smaller n=208 subgroup. To flip, you’d need a residual confounder that selectively wrecks outcomes in the ≥110 mL stratum — e.g., withdrawal-of-care bias, which is plausible but not addressed in the emulation.

3. Second-Order Catalyst

If 110 mL DWI is real, the protocol that moves first is the comprehensive stroke center’s auto-triage pipeline, not the small hospital’s CT scanner. Here’s why: to operationalize a 110 mL ceiling, you need volumetric measurement before groin puncture — meaning you need either RAPID/CTP or an equivalent automated DWI/NCCT volumetric tool available at the receiving hospital within minutes of arrival. The institutions that move first will be the ones who already run auto-ASPECTS and auto-CTP pipelines (Cleveland Clinic, Mass General, Karolinska, Mount Sinai, the Korean centers in this registry) and where the neurointerventionalist sees a quantitative core number alongside the ASPECTS, not in place of it.

Within 12–18 months: a multi-society consensus statement clarifying that ASPECTS is a trial-eligibility marker, not a biological ceiling marker; volumetric thresholds enter the AHA/ESO guideline appendices as “consider” language. Within 24–36 months: a randomized trial randomizing door-time triage to ASPECTS-only vs ASPECTS+volumetry in the ASPECTS 3–5 band, with non-inferiority on functional outcome and superiority on cost/length-of-stay.

The patient subgroup that moves first: the late-presenting (12–24 hour) LVO patient with ASPECTS <6 — exactly the population where the “futile recanalization” debate is hottest and where a 110 mL DWI cutoff gives everyone (interventionalist, family, ICU team) a defensible shared-decision number.

4. Asymmetric Leverage

The asymmetric payoff is on the ASPECTS-only large-core side, not the volumetric-large-core side. That’s the counter-intuitive read of the paper.

• The volumetric-large-core group (true big strokes, ASPECTS <6, DWI ≥50 mL): high bad-outcome rate, modest absolute EVT benefit. Standard gamble; routine EVT decisions.
• The ASPECTS-only large-core group (low ASPECTS but DWI <50 mL): the EVT community is currently treating these patients on the assumption they’re “large core.” If they’re biologically small-core, we’re a) offering thrombectomy that they would have done well with anyway (modest absolute benefit) and b) enrolling them into “large-core” trials and registries that dilute the treatment-effect estimate for the genuinely large-core subgroup. The leverage is to re-stratify by volume first, ASPECTS second, which re-allocates about a quarter of the current “large-core” denominator into the small-core bucket where the absolute benefit per treated patient is larger and cheaper to deliver.

The corollary: payers and systems will look at this paper and ask whether paying for automated volumetric software (RAPID, e-Stroke, JLK) at every thrombectomy-capable center pays for itself in fewer futile procedures. The asymmetric answer: yes, mostly by removing ASPECTS-only false-positives from the EVT pathway, not by changing who gets pulled back from EVT.

5. Paradigm Destroyer

This paper kills the reflex to read ASPECTS <6 as a binary ceiling for EVT. Tomorrow-morning protocol change in two sentences:

1. In any LVO patient being considered for EVT, ASPECTS <6 must be paired with an automated ischemic-core volume before groin puncture — DWI volume (≥50 mL large, ≥110 mL likely ceiling) or CTP rCBF<30% volume (≥70 mL large), whichever the receiving center runs in <10 minutes.
2. Patients with ASPECTS <6 but DWI core <50 mL (or CTP core <70 mL) should be treated as standard EVT candidates with routine prognostic counseling, not routed into “large-core” pathways, “futile recanalization” discussions, or aggressive hospice conversations before the procedure.

It also kills a subtler reflex: the assumption that the pivotal-trial ASPECTS <6 ceiling maps cleanly onto a biological ceiling. It doesn’t. ASPECTS is a pragmatic gate; volume is the substrate. Treating them as interchangeable has been quietly inflating the “large-core” denominator for a decade.

MVP — Minimum Viable Proof

If a single number from this paper had to move bedside practice tomorrow, it is this: In EVT candidates with ASPECTS <6, an automated DWI ischemic-core volume of approximately 110 mL or greater identifies a subgroup in whom the ordinal 90-day mRS benefit of thrombectomy is no longer detectable in this registry’s target trial emulation. Conversely, an ASPECTS <6 patient with DWI core below 50 mL is biologically a small-core patient and should be counseled, treated, and prognosticated as such.

Best Combination

Combine this finding with three prior threads:

1. RESCUE-Japan LIMIT, ANGEL-ASPECT, SELECT2, TENSION (2022–2024). Those trials established that EVT does help ASPECTS 3–5 large-core patients on average. This paper says: the average is hiding an ASPECTS-only subset (~25% of the trial denominators) that drove much of the apparent benefit, and a true volumetric-large-core subset (≥110 mL) where the benefit is gone. The trials’ ASPECTS-based conclusions survive, but their generalizability claims need to be re-anchored to volume.

2. DEFUSE-3, DAWN, EXTEND (late-window trials). These established penumbra-based selection (mismatch) for the 6–24 hour window. The 110 mL ceiling here was estimated at median LKW-to-groin of 4.2 hours; at late windows the ceiling likely drops because the penumbra shrinks. Combine: at 6–24 hours, the practical ceiling on DWI volume is probably 70–80 mL, not 110 mL. This paper doesn’t say that explicitly but it’s the natural extension.

3. The futile-recanalization literature in cancer-associated stroke (CACTUS nomogram, PMID 42760131). Cancer-related LVO has its own prognostic scoring; the volumetric-versus-topographic debate here likely generalizes there too — biology (volume) beats topography (ASPECTS) in patients where the biology is what drives outcome.

Overvalue Warning

1. Don’t overread 110 mL as a hard contraindication to EVT. This is an observational emulation, not a randomized comparison in the ≥110 mL subgroup. The 95% CI on the cOR for that stratum includes values consistent with modest benefit. Until an RCT randomizes DWI ≥110 mL patients to EVT vs BMT, this is a counseling number, not a refusal number. Withdrawal of care decisions should not be made on 110 mL alone.

2. Don’t overread ASPECTS-only large-core as “fine for EVT, ignore the ASPECTS.” The paper shows their adjusted outcomes are comparable to small core — but their unadjusted illness severity (the reason ASPECTS was low) is still real. They’re not low-risk; they’re correctly triaged. The point is they shouldn’t be lumped into the “large-core” denominator for trial design or shared decision-making, not that they should be treated as routine.