Endovascular Treatment for Patients Aged ≥80 Years with Large Ischemic Stroke — association is not age-invariant benefit

Clinical Question (PICO)

In patients aged ≥80 years with large-vessel-occlusion ischemic stroke and baseline noncontrast-CT ASPECTS ≤5, is endovascular treatment (EVT) plus standard medical treatment (SMT), compared with SMT alone, associated with less disability at 90 days, and at what hemorrhage and mortality cost? The investigators also examined age 80–84 versus ≥85 and other subgroups. These are exploratory secondary analyses, not verified prespecified randomized EVT comparisons.

Bottom Line

EVT was associated with a better 90-day modified Rankin Scale (mRS) distribution: adjusted common OR 2.83, 95% CI 1.48–5.70; P=0.003. Mortality was 50.9% versus 74.1%, but any intracerebral hemorrhage was more common; this nonrandomized comparison supports individualized assessment, not a causal guarantee of benefit or the claim that benefit is identical at every age.

Design

Study type: Pooled observational secondary analysis of the MAGIC registry and a selected subset of the MARVEL randomized trial; outcome assessors were blinded to treatment status. The EVT-versus-SMT comparison itself was not randomized.
N: 270 included patients.
- EVT plus SMT: 212.
- SMT alone: 58.
Randomization: None for the comparison reviewed here. MARVEL randomized an adjunctive methylprednisolone strategy among patients already undergoing EVT; borrowing its participants does not randomize EVT eligibility.
Setting: China. MAGIC recruited at 38 comprehensive stroke centers; MARVEL at 82. The number of unique centers contributing this selected pooled sample is not reported.
Enrollment: MAGIC, November 2021–February 2023; MARVEL, February 2022–June 2023. This secondary report was published September 28, 2026.
Follow-up: The principal endpoint was assessed at 90 days; mean follow-up was not reported.
Analysis: Multivariable ordinal logistic regression for mRS shift; binary logistic regression for dichotomous endpoints; linear regression for early NIHSS change. Adjustment and inverse probability of treatment weighting (IPTW) used hyperlipidemia, baseline NIHSS, baseline ASPECTS, last-seen-well-to-imaging time, and stroke etiology. This was not an intention-to-treat randomized EVT analysis.
Primary outcome: Distribution of 90-day mRS scores, from 0 (no symptoms) to 6 (death); common OR >1 favors a shift toward better outcomes.

Population

Inclusion Criteria

Patients aged ≥80 from MAGIC with large ischemic stroke, defined by ASPECTS ≤5.
Patients aged ≥80 from MARVEL with baseline ASPECTS ≤5; MARVEL's original ASPECTS eligibility was 3–10.
The included anterior-circulation occlusions were internal carotid artery and middle cerebral artery M1 or M2 occlusions.

Exclusion Criteria / Eligibility Boundaries

Age <80 and ASPECTS >5 were outside this secondary analysis.
Parent-study eligibility and exclusions governed enrollment, but this report refers readers to those protocols rather than fully reproducing them. Do not infer that all frail, premorbidly dependent, very-late-presenting, or medically unstable patients were eligible.
Posterior-circulation stroke is not represented in the reported occlusion categories.

Baseline Characteristics (overall)

Median age: 83 years (IQR 81–86); male: 40.7%.
Median NIHSS: 19 (IQR 16–22); median ASPECTS: 4 (IQR 3–5).
Atrial fibrillation: 61.9%; hypertension: 74.4%; intravenous thrombolysis: 34.8%.
EVT patients had higher baseline ASPECTS than SMT patients: median 4 versus 3, P=0.01. Hyperlipidemia was also more frequent: 19.3% versus 6.9%, P=0.04.
Among EVT recipients, 136 were aged 80–84 and 76 were ≥85.

Interventions

EVT plus SMT: Mechanical endovascular treatment alongside medical management. In this group, median last-seen-well-to-puncture time was 305 minutes (IQR 208–469), and to recanalization 405 minutes (IQR 296–582). These are observed times, not validated eligibility cutoffs. The secondary report does not specify one uniform device, procedural regimen, or medical-treatment dose.
SMT alone: Medical management without EVT. Intravenous thrombolysis was reported in 31.0% versus 35.8% of SMT and EVT patients, respectively.
Important cointervention: MARVEL contributed EVT-treated patients originally enrolled in an adjunctive methylprednisolone trial. Parent-study membership and adjunctive assignment are not among the five listed adjustment variables.

Outcomes

Primary Outcome — 90-day ordinal mRS

Median mRS was 6 (IQR 3–6) with EVT and 6 (IQR 5–6) with SMT. Table 2 reports adjusted common OR 2.83 (95% CI 1.48–5.70), P=0.003; IPTW common OR 3.11 (2.19–4.46), P<0.001. A distributional improvement does not mean that the median survivor becomes independent, or that the common OR is a risk ratio.

Exploratory subgroup analyses — Figure 3, as printed

Prespecification is not established in this secondary report. The figure supplies subgroup common ORs and 95% CIs, with interaction P values for each grouping; it does not supply individual subgroup P values. Those must not be invented.

Grouping	Subgroup common OR (95% CI)	Interaction P
Age	<85 years, necessarily 80–84 here: 4.98 (1.92–14.88); ≥85: 1.88 (0.74–5.08)	<0.001
Sex	Female: 3.86 (1.61–10.25); male: 1.80 (0.65–5.31)	0.626
Atrial fibrillation	No: 7.59 (2.25–32.06); yes: 2.05 (0.91–4.89)	0.090
Baseline NIHSS	≤17: 9.51 (2.67–46.08); >17: 1.53 (0.71–3.45)	0.005
Baseline ASPECTS	>2: 4.42 (1.98–10.76); ≤2: 1.11 (0.24–5.39)	0.067
Intravenous thrombolysis	No: 3.08 (1.38–7.33); yes: 3.32 (1.00–13.32)	0.528
Occlusion site	ICA: 2.00 (0.51–10.10); M1: 3.36 (1.37–8.91); M2: 5.24 (1.10–30.75)	0.802
Stroke etiology	Large-artery atherosclerosis: 4.71 (1.51–17.14); cardioembolism: 3.38 (1.37–9.26); unknown/other: 0.05 (<0.01–1.30)	0.084
Onset to imaging	≤360 minutes: 3.35 (1.48–8.23); >360: 3.47 (0.97–14.12)	0.233

Reporting inconsistency: Figure 3 prints an overall OR of 2.78 (1.45–5.59), with SMT n=57, whereas Table 2 uses 58 SMT patients and the headline OR of 2.83. Its age and NIHSS interactions also prevent a simple claim of age-invariant treatment effect. No reconciliation is supplied; the table above preserves the figure rather than silently harmonizing it.

Secondary Outcomes — conventional adjusted models unless specified

Outcome	EVT	SMT	Adjusted estimate (95% CI); P
mRS 0–2	40 (18.9%)	2 (3.4%)	OR 4.79 (1.48–24.47); 0.007
mRS 0–3	52 (printed as 25.0%)	3 (5.2%)	OR 4.46 (1.59–17.01); 0.003
mRS 0–4	83 (39.2%)	7 (12.1%)	OR 3.87 (1.75–9.66); 0.001
Pneumonia	160 (75.5%)	48 (82.8%)	OR 0.58 (0.26–1.19); 0.142
Early NIHSS change	Median 0 (IQR −9 to 2)	Median 0 (IQR −12 to 9)	β 1.15 (−2.62 to 4.92); 0.550

The printed mRS 0–3 percentage does not reconcile with 52/212 (calculated 24.5%); this review retains the original count and flags the percentage instead of treating it as corrected source data. Pneumonia reached significance only with IPTW: OR 0.49 (0.31–0.76), P=0.001.

Adverse Events / Safety

90-day mortality: 108/212 (50.9%) versus 43/58 (74.1%); adjusted OR 0.41 (0.20–0.81), P=0.009. The calculated crude difference is 23.2 percentage points lower with EVT, not a randomized causal risk reduction.
Symptomatic ICH within 48 hours: 43 (20.9%) versus 8 (13.8%); adjusted OR 1.78 (0.80–4.37), P=0.16. With IPTW: 1.61 (1.03–2.55), P=0.04. The safety conclusion is therefore model-sensitive, not proof of equal symptomatic-bleeding risk.
Any ICH: 81 (39.3%) versus 8 (13.8%); adjusted OR 3.65 (1.70–8.71), P=0.001; IPTW OR 3.75 (2.48–5.76), P<0.001.
The EVT hemorrhage percentages are compatible with a denominator of 206 rather than the full 212; the main table does not clearly reconcile that denominator. Counts and published percentages are shown without assuming missing scans were negative.
Age analyses answer different questions: Within EVT recipients alone, the age-group mRS comparison was not significant: adjusted common OR 0.91 (0.51–1.59), P=0.734. That is a prognostic comparison, not an EVT-versus-SMT interaction test. Continuous-age models reported no interaction for mRS 0–2, 0–3, or 0–4, but did report a mortality interaction, P=0.016.

Figures

90-day disability distribution by treatment and age — *Figure 1.* Modified Rankin scale score distribution at 90 days. The distribution of the modified Rankin scale scores at 90 days. ( A ) Overall patients; ( B ) Patients only in EVT group stratified by age (aged 80–84 vs aged ≥85). Abbreviations : EVT, endovascular treatment; SMT, standard medical treatment.

Predicted outcomes by age and treatment, including mortality interaction — *Figure 2.* Predicted probabilities of clinical outcomes at 90 days of the overall patients. Predicted probabilities with 95% CI by age: ( A – D ) Show mRS score 0–2, 0–3, 0–4 and mortality at 90 days, respectively, stratified by treatment arm (EVT vs SMT). Solid lines indicate estimated probabilities of outcomes; shaded areas represent 95% CIs. Abbreviations : CI, Confidence interval; EVT, Endovascular therapy; mRS, Modified Rankin Scale; SMT, Standard medical treatment.

Exploratory subgroup forest plot with age and NIHSS interactions — *Figure 3.* Subgroup analysis of the overall patients. The subgroup analysis was based on the adjusted common odds ratio (acOR) on the distribution of mRS score at 90 days. Subgroups were stratified by age (80–84 vs ≥85), sex (female vs male), history of atrial fibrillation (no vs yes), baseline NIHSS score (≤17 vs >17), baseline aspects (>2 vs ≤2), IVT (no vs yes), occlusion site (ICA vs M1 vs M2), stroke etiology (LAA vs cardioembolism vs unknown or others), and onset to imaging time (≤360 vs >360). The arrow indicates that the 95% CI was beyond the scale. Abbreviations : AF, atrial fibrillation; ASPECTS indicates Alberta Stroke Program Early Computed Tomography Score; ICA, intracranial carotid artery; IVT, intravenous thrombolysis; M1, the first segment of middle cerebral artery; M2, the second segment of the middle cerebral artery; NIHSS, National Institutes of Health Stroke Scale.

Criticisms

Not an EVT RCT: Pooling a registry with an adjunct-treatment RCT does not remove confounding by treatment selection. All MARVEL participants underwent EVT, so source-study effects can track treatment; neither source membership nor adjunctive assignment is listed in the adjustment model.
Age claims overreach the results: A nonsignificant age comparison inside the treated group cannot establish equal treatment benefit. Figure 3's age interaction and Figure 2's mortality interaction conflict with the abstract's unqualified age-consistency language. Sparse strata, multiple exploratory tests, and wide CIs preclude a new age or NIHSS exclusion rule.
Residual confounding is clinically material: Frailty, cognition, broader comorbidity, treatment-limitation decisions, and complications were incompletely captured or not available. ASPECTS already differed between groups before adjustment.
Small comparator and internal reporting discrepancies: Only 58 SMT patients were included. Figure 3 uses 57; the mRS 0–3 percentage and hemorrhage denominators also require clarification. Some subgroup estimates are extremely imprecise.
Missingness and weighting: One missing last-seen-well-to-imaging value was imputed using the article's incompletely described “sample imputation” approach. The main report does not clearly explain all outcome denominators or establish that weighted precision fully captures the uncertainty of treatment selection. IPTW cannot repair unmeasured confounding.
Generalizability: Selected patients from Chinese comprehensive stroke centers are not a representative sample of all octogenarians, especially those with severe premorbid disability or frailty. Neither low mortality nor independence is the typical observed outcome even after EVT.

Funding

Supported by the National Natural Science Foundation of China (82271349), National Science Fund for Distinguished Young Scholars of China (82425021), Noncommunicable Chronic Diseases–National Science and Technology Major Project (2024ZD0527905), National Natural Science Foundation of China Young Scientists Fund Project, Category C (82501597), and Graduate Research Innovation Project of Army Medical University (yjscx2026S04). A separate funder-role statement is not supplied in the report. Authors declared no conflicts of interest. Parent registrations: ChiCTR2100051664 (MAGIC) and ChiCTR2100051729 (MARVEL); these are Chinese registry identifiers, not NCT numbers.

The paper

Authors. Chawen Ding et al.
Title. Endovascular Treatment for Patients Aged ≥80 Years with Large Ischemic Stroke.
Journal. Clinical Interventions in Aging.
Year. 2026; volume 21, article 634303. Published September 28, 2026.
DOI. 10.2147/CIA.S634303
PMCID. PMC13632614
Figure license. © 2026 Ding et al.; Creative Commons Attribution–NonCommercial 4.0, as stated in the PMC article. Embedded original figures remain attributed to the authors and publication.

Deep Dive — click to expand

What this is

This September 2026 secondary analysis asks whether selected people aged at least 80 with large infarcts still show an association between thrombectomy and better recovery. Its headline shift OR is encouraging, but the most useful reading is not “age no longer matters.” It is that poor prognosis and possible treatment benefit can coexist—and that an observational comparison cannot tell us exactly which individual patient gains enough to justify the burden of treatment.

1. Shadow Audit

The favorable ordinal result sits beside a sobering absolute reality: 50.9% of EVT patients died by 90 days, and only 18.9% achieved mRS 0–2. Presenting the OR without those outcomes would conceal the actual counseling problem. Conversely, a median mRS of 6 in both arms conceals favorable shifts elsewhere in the distribution; neither the OR alone nor the median alone is an adequate bedside summary.

The more serious shadow is methodological. The report's reassuring age language is not consistent across its own analyses: an age interaction P<0.001 for ordinal mRS in Figure 3, a mortality interaction P=0.016 in Figure 2, and a nonsignificant age-group comparison among treated patients are three different findings. The last cannot cancel the first two. These discrepancies warrant clarification, not selective quotation of whichever result supports a preferred age policy.

2. Inversion Engine

For the conventional adjusted overall association to reverse direction, the common OR would have to cross from 2.83 through 1.0; its reported lower confidence bound is 1.48. That is evidence against a null association within the fitted model, not protection against selection bias outside it. A plausible competing account is that patients judged most salvageable were preferentially treated, while parent-study membership and unmeasured frailty shaped both access to EVT and survival.

The ≥85 subgroup already illustrates why the opposite conclusion remains statistically possible locally: its OR is 1.88 with CI 0.74–5.08, spanning unfavorable and favorable associations. The proper inversion test is not “did treated 85-year-olds differ significantly from treated 82-year-olds?” It is whether comparable patients of a given age would do better with EVT than without it. That counterfactual remains incompletely identified here.

3. Second-Order Catalyst

Editorial implication, not a guideline claim: The first institution likely to use this paper sensibly is a comprehensive stroke center auditing age-based refusals. Its stroke and neurointerventional teams can review whether age alone ended evaluation before imaging, baseline function, frailty, time, and patient preferences were considered. The paper's five-variable adjustment model also identifies a research need: prospective capture of the geriatric and treatment-selection variables it did not adequately measure.

The immediate change should be better assessment and documentation—not a mandate to intervene in every patient over 80, and not a new NIHSS ≤17 rule extracted from an exploratory forest plot. Any local protocol change still requires the center's established evidence and governance process.

4. Asymmetric Leverage

The crude independence contrast is 40/212 versus 2/58, a calculated difference of 15.4 percentage points. If a comparable causal gain were confirmed, it would be important even in a narrowly selected population. But converting that contrast into a treatment NNT would imply exchangeable groups that this design does not establish.

The low-cost asymmetric opportunity is to remove a nonclinical age-only barrier to expert review while retaining honest counseling. The high-cost mistake is to promise independence or survival from an observational average. Hemorrhage, severe disability, prolonged care needs, and the patient's valuation of survival with dependency belong in the same conversation as the potential functional gain.

5. Paradigm Destroyer

The reflex this paper challenges is “old plus large core equals automatically futile.” The reflex it does not justify replacing it with is “age has no effect, so thrombectomy is automatic.”

A two-sentence proposal for a locally reviewed pathway: “Do not terminate thrombectomy evaluation solely because the patient is aged ≥80; document imaging, baseline function, frailty, timing, and goals of care in the specialist assessment. Present both possible functional improvement and the high observed mortality and hemorrhage burden, without describing this cohort as randomized proof or imposing its exploratory subgroup thresholds.”

MVP — Minimum Viable Proof

The minimum defensible statement is: among these selected ≥80-year-old patients with ASPECTS ≤5, EVT was associated with a favorable 90-day mRS shift, adjusted common OR 2.83 (1.48–5.70), while 108/212 treated patients died. That is enough to challenge automatic futility language, but not enough to establish the causal absolute benefit or identify who should be treated. Greater confidence would require well-controlled, independently validated elderly-patient data with frailty, premorbid function, study source, cointerventions, and outcome missingness explicitly handled.

Best Combination

Place this report beside randomized large-core evidence rather than above it. The 2023 SELECT2 trial, identified in this paper's reference list and cross-checked against its PubMed abstract, randomized 352 patients and reported a generalized OR of 1.51 (95% CI 1.20–1.89) favoring thrombectomy for 90-day disability. SELECT2 provides stronger causal evidence in its enrolled population; the present analysis supplies a more specific but more confounded look at patients aged ≥80. The measures and populations are not interchangeable, and their ORs should not be treated as directly comparable magnitudes.

Use the randomized foundation to understand why large-core EVT is considered, then use this cohort to sharpen questions about older patients' prognosis, selection, and counseling. Do not numerically pool these heterogeneous estimates or double-count overlapping MAGIC-derived analyses as independent replication. No current guideline recommendation is being inferred from this secondary report.

Overvalue Warning

Do not read “no age difference among EVT recipients” as “equal EVT benefit at every age.” The comparison lacks an untreated counterfactual, the ≥85 estimate is imprecise, and the paper contains significant age-related interactions elsewhere.
Do not read the mortality contrast as a causal NNT or the bleeding P=0.16 as proof of safety. Treatment assignment was selected, safety denominators need clarification, and symptomatic-ICH significance changed under IPTW.

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.