What this is

A post-hoc analysis of two major European LVO thrombectomy trials (MR CLEAN-MED and MR CLEAN-NO IV) finds that patients whose infarcts involved the deep perforator and insular territory were nearly three times more likely to have a worse 90-day modified Rankin Scale score than patients with cortex-only infarcts — even after adjusting for TICI reperfusion grade and bleeding complications. The primary 24-hour NIHSS outcome was negative, so the mRS finding is the load-bearing claim. This deep-dive takes the result apart.

2. Inversion Engine

What would have to be true for the opposite conclusion.

For the opposite conclusion — that perforator territory infarction does not independently drive 90-day outcome — to be the right one, the following would have to be true:

The TICI 2B-3 and Heidelberg 1c/2 effects must be the real signal. Both are independently significant in the multivariable model. If those variables are actually proxying for “harder thrombectomy, more retrieval passes, distal embolization,” then the perforator variable is just absorbing variance that belongs to a procedural-quality confounder. Procedural passes, clot burden score, and first-pass effect were not in the adjustment set.The “perforator and insular” group must be a misclassification. The paper collapses insula (a cortical structure supplied by MCA branches, not perforators) with perforator (lenticulostriate territory). If insular cortex is re-split into its own group, the “perforator effect” might vanish — leaving a banal “larger total infarct volume → worse mRS” finding.The 24-hour NIHSS null must be re-interpreted. The primary outcome was negative; the positive result is on the secondary outcome. In a frequentist post-hoc analysis with no pre-registered mRS hypothesis, the 2.94 OR with CI 1.73–4.98 needs to survive a multiple-comparisons adjustment across the three infarct-location contrasts. The paper does not apply one.The first assumption to break: “perforator territory” is a coherent, anatomically validated category on standard follow-up MRI. The authors admit that “neither direct angiographic techniques nor 7.0 Tesla magnetic resonance imaging (MRI) has the potential to reliably provide a complete, high-resolution image of perforator anatomy and potential collaterals.” On a 24-hour 1.5T or 3T DWI, the boundary between “deep perforator” and “juxtacortical insular M2 branch” is partly convention.

3. Second-Order Catalyst

If the result is real — who changes practice first.

Who changes practice first (months 0–6): Comprehensive stroke centers that already do post-thrombectomy MRI routinely. They will start retrospectively auditing their own perforator-infarct rate by operator, device (stent-retriever vs. aspiration), and number of passes, treating it as a new procedural quality metric. MR CLEAN-NO IV already showed no IV-tPA benefit; combined with this, the case for going direct to thrombectomy without IV-tPA gets a small additional nudge in the “perforator protection” direction, since tPA itself can lyse protective collaterals before thrombectomy.

Who changes practice last (12–24 months): Community hospitals without 24-hour MRI. They will not measure the outcome, so the finding does not change what they do — except via downstream guidelines.

What the field does in 12 months:

A pre-specified, prospective perforator-infarct endpoint appears in at least one ongoing LVO trial (most likely ANGEL-ACT, RESCUE BT2 follow-ons, or a POST-TNK/UK substudy).Industry-funded trials of intra-arterial adjunctive thrombolytics (POST-TNK and POST-UK are already published in JAMA 2025 and are cited here) gain a mechanistic justification they previously lacked: the target is the residual lenticulostriate thrombus, not the parent-artery clot.A guideline-writing committee (AHA/ESO) is asked whether successful reperfusion (TICI 2B-3) with concomitant perforator infarct should be re-classified from “successful” to “partially successful” in trial endpoints.## 4. Asymmetric Leverage Where is the asymmetric payoff?

Largest claim with most leverage, if true: “Perforator territory infarct — independent of infarct volume — roughly triples 90-day mRS disability.” If real, this reframes every thrombectomy outcome paper published since 2015, because TICI 2B-3 with a deep infarct is not equivalent to TICI 2B-3 without one. The leverage: it would change the primary endpoint of LVO trials (mRS shift conditional on perforator status) and re-price devices / techniques by their perforator-infarct rate.

Smallest claim that costs the most if wrong: The paper’s discussion implies that intra-arterial thrombolytics after thrombectomy (“combined thrombectomy and adjuvant thrombolysis for reperfusion of occluded perforating artery trunks”) should be studied more. POST-TNK and POST-UK are cited favorably. If the perforator-infarct → mRS association turns out to be a confounded proxy for procedural difficulty, then the entire intra-arterial-thrombolytic-after-thrombectomy research program (worth ~$200M+ in ongoing RCTs) is solving the wrong problem. Cost of being wrong: thousands of patients enrolled in trials with a flawed mechanistic premise.

Asymmetry: The upside of the large claim (re-doing thrombectomy endpoints) is enormous but slow. The downside of the small claim (over-investing in adjunctive IA thrombolysis on a flawed premise) is fast and irreversible — it has already redirected enrollment in two Phase III programs.

5. Paradigm Destroyer

What does this paper kill?

What this paper kills: Nothing fundamental. The pre-existing paradigm — recanalize the parent artery; residual small-vessel ischemia is unfortunate but tolerable — survives. The paper adds a quantitative penalty (3× odds of worse mRS) but does not invalidate the paradigm.

What this paper does dent: The implicit assumption that “successful reperfusion (TICI ≥ 2B)” is a sufficient endpoint. It is not. Two patients with identical TICI 2C and identical infarct volumes can have very different 90-day mRS if one has a basal-ganglia perforator infarct and the other does not. That is a real and previously under-quantified contribution.

What this paper does not kill:

The 90-day mRS as a primary endpoint in LVO trials.The parent-artery recanalization strategy.The validity of MR CLEAN-NO IV or MR CLEAN-MED conclusions on their original endpoints.## MVP — Minimum Viable Proof The cheapest test that would change evidentiary weight:

A pre-specified, prospective registry of 500 consecutive anterior-circulation LVO thrombectomy patients at ≥10 high-volume centers, with mandatory 24-hour MRI and blinded adjudication of perforator territory infarct by two neuroradiologists, plus 90-day mRS and 90-day Montreal Cognitive Assessment (MoCA). Cost: ~$2M, ~18 months. Single biomarker: perforator-territory DWI lesion volume (mL), adjudicated.

If this prospective registry reproduces the 2.94 OR (CI excluding 1) on 90-day mRS and shows a MoCA decrement of ≥2 points attributable to perforator volume, the paper graduates from “hypothesis-generating post-hoc” to “practice-informing.” If it fails to reproduce, the entire intra-arterial-adjunctive-thrombolytic research program needs to be re-scoped.

Best Combination

The single highest-leverage move based on the five frameworks:

Pre-register the perforator endpoint in an ongoing or planned LVO trial (ANGEL-ACT follow-on, or a successor to POST-TNK/POST-UK), with blinded core-lab adjudication of perforator-territory DWI on 24-hour MRI, and stratify randomization by infarct pattern at baseline.

This costs ~$500K incremental on a trial that already exists. It directly tests whether the 2.94 OR is causal or confounded, gives the IA-thrombolytic hypothesis the cleanest possible test, and produces a guideline-actionable answer in 2–3 years — before community practice ossifies around the unverified mechanistic story.

Overvalue Warning

Where the analysis is most likely to overstate the result.

Post-hoc subgroup. This is not a randomized comparison. The three infarct-location groups are defined by the outcome (post-treatment MRI), not by a pre-randomization variable. Reverse causation — that harder thrombectomies cause both more perforator infarcts and worse mRS via procedural injury — is not ruled out.Selection bias on MRI availability. Only 34% (397/1167) of trial participants had the required post-treatment MRI. Operators who suspected a perforator event were probably more likely to obtain one. The cohort is enriched for the lesion being studied.Collapsing “perforator” with “insular.” Insula is cortical MCA territory, not perforator. The headline-positive group is really “deep + insular” — any independent “perforator-only” effect may be smaller.Single-center / regional MRI reading. The paper does not state whether MRI adjudication was centralized or local. Local reads inflate effect sizes in imaging-defined subgroups.Funding. Unrestricted grants from Stryker, Medtronic, and Cerenovus (the three dominant thrombectomy device makers). The result does not threaten device revenue, but the discussion’s emphasis on adjunctive IA thrombolysis as a new market is a quiet conflict.Recency / novelty bias. This is the first large-n study to make this specific claim. First-of-kind results routinely overestimate effect sizes that shrink in replication (winner’s curse).The single bias to watch most closely: procedural-confounding. TICI 2B-3 and Heidelberg bleeding are in the model, but number of passes, total fluoroscopy time, and first-pass effect are not. A clinician who sees this paper and concludes “I need to protect the lenticulostriates” should first ask whether their procedural technique is already minimizing that risk.