More data has not automatically produced more confident decisions at the bedside. LiveTextbook is navigational guidance, not a medical device and not a diagnostic.
What it is
LiveTextbook is a curated corpus of acute-stroke evidence in which every claim carries its source. Put a decision to it and it returns what the literature holds for and against, with the trials, the guidelines and the figures attached, and the same thing tomorrow. At three in the morning with a referral in hand, that is the trials that enrolled a patient like this one, where they stopped enrolling, and the analyses that disagree. It surfaces the sources; the decision stays at the bedside.
Three worked examples
These three are the cases where that matters, because the literature does not converge. They are constructed teaching cases, written in the register in which a referral arrives; no patient is described here, and no patient material was used.
Case one — mild deficit, left M1, thrombolysis running
The referral. A 46-year-old man, last seen well 3.5 hours ago, NIHSS 4 with mild right arm weakness and expressive aphasia. CTA shows a left M1 occlusion with good collaterals (80%). NCCT ASPECTS 10. CT perfusion demonstrates a CBF<30% core of 4 mL, Tmax>6s volume of 52 mL, mismatch ratio 13.0. This is a proven low-NIHSS large-vessel occlusion. Intravenous thrombolysis is running.
The question it forces. Whether to add thrombectomy to thrombolysis already running, in a man whose deficit is mild and whose occlusion is not. The randomised evidence was built in patients several times more severe, and his aphasia may or may not meet the published definition of a disabling deficit.
What the published evidence holds. In MINOR-STROKE, 729 patients with NIHSS ≤5 and a large-vessel occlusion intended for thrombolysis alone, early neurological deterioration occurred in 12.1%, and 11.0% on validation, predicted independently by a more proximal occlusion and a longer thrombus. Alteplase recanalised 11% of intracranial ICA occlusions in INTERRSeCT, 22% of proximal M1, 46% of distal M1 and 37% of M2 across the whole cohort, and 45% in its minor-stroke subset.
MILD-MT (NCT06179017), presented at ESOC 2026 and not published in full, is the one randomised signal, and its point estimates stay off this page until they are peer-reviewed. My own published meta-analysis of eleven observational cohorts found no functional advantage over best medical management (excellent outcome RR 1.10, 0.93–1.31) with symptomatic haemorrhage more than tripled (RR 3.53, 2.35–5.31); a propensity-matched SITS-ISTR comparison of thrombolysis plus thrombectomy against thrombolysis alone gave an adjusted OR of 0.46 for mRS 0–1.
In a perfusion sub-study of the same population, bridging did worse than thrombolysis alone below a mismatch volume of 40 mL (OR 0.48, 0.33–0.71) and no worse above it; 40 mL is that cohort's median rather than a validated threshold, and here it is 48 mL.
The 2019 ESO–ESMINT guideline finds no evidence for an upper severity limit and asks that patients below NIHSS 6 be enrolled in trials; the position that a mild but disabling deficit warrants immediate thrombectomy was an expert vote inside that process, not its graded text. The definition of disabling belongs to PRISMS: a deficit that, if unchanged, would prevent basic activities of daily living or return to work.
MOSTE (NCT03796468), TRIMIS (NCT06778226) and ENDOLOW are registered and unreported. What this case shows is a population the pivotal trials were not built around — MR CLEAN's protocol admitted patients from NIHSS 2 with no severity restriction, yet the randomised record holds few patients like him — on which the observational data and the one randomised signal point in opposite directions.
Sources: Seners, JAMA Neurology 2021, doi:10.1001/jamaneurol.2020.4557; Menon, JAMA 2018, doi:10.1001/jama.2018.12498; Lau, Stroke 2022, doi:10.1161/STROKEAHA.120.030380; Wu, Stroke and Vascular Neurology 2026, doi:10.1136/svn-2025-004832; Safouris, Stroke 2023, doi:10.1161/STROKEAHA.123.043937; Schwarz, European Journal of Neurology 2023, doi:10.1111/ene.15722; Seners, Stroke 2022, doi:10.1161/STROKEAHA.122.039182; Turc, European Stroke Journal 2019, doi:10.1177/2396987319832140; Khatri, JAMA 2018, doi:10.1001/jama.2018.8496.
Case two — distal basilar occlusion at 81, wake-up onset
The referral. An 81-year-old found on waking, last known well nine hours earlier, NIHSS 11. CTA shows a distal basilar artery occlusion with good posterior communicating collaterals. pc-ASPECTS 8. No intravenous thrombolysis was given. Prestroke mRS 3.
The question it forces. Whether to open a distal basilar occlusion nine hours after last known well in a patient of 81 already dependent before the stroke. The severity and the imaging are what the basilar trials required; the age and the prestroke score are what they excluded.
What the published evidence holds. ATTENTION enrolled within 12 hours with NIHSS ≥10 and, at 80 or above, required a premorbid mRS of 0 and pc-ASPECTS ≥8. BAOCHE enrolled from 6 to 24 hours, capped age at 80 and prestroke mRS at 1, and reported mRS 0–3 at 90 days in 46% against 24%, adjusted rate ratio 1.81 (1.26–2.60), symptomatic haemorrhage 6% against 1%, mortality 31% against 42%. This patient is outside ATTENTION on prestroke mRS and outside BAOCHE on age and prestroke mRS, and the 2026 AHA/ASA guideline recommends thrombectomy within 24 hours for a baseline mRS of 0 or 1 with NIHSS ≥10 and pc-ASPECTS ≥6, which this patient also does not meet.
Whether the clot's position in the basilar artery changes the expected benefit is unresolved: ATTENTION reported no significant interaction by occlusion site, in a subgroup analysis not powered to detect one, and in BAOCHE distal occlusions were a minority whose subgroup estimate could not be computed. What is left bears on reserve rather than the year of birth, and comes from the wrong circulation. HERMES, pooling five anterior-circulation trials, found the effect still present at 80 and above (common OR 3.68, 1.95–6.92) — an argument about physiological reserve that says nothing about the basilar artery. In a single-centre, predominantly anterior-circulation cohort of thrombectomy-treated patients aged 80 and over, very poor outcome occurred in 79.4% of those frail on the Clinical Frailty Scale against 52.4% of the non-frail, one-year survival 35.3% against 65.6%. In the MR CLEAN Registry, also anterior circulation, prestroke-dependent patients reached a favourable outcome — mRS 0–2 or no worsening from the prestroke score — in 27% against 42%, with similar rates of symptomatic haemorrhage and stroke progression, and the association did not survive adjustment.
What this case shows is a decision every randomised entry criterion for basilar occlusion excludes the patient from, leaving anterior-circulation trials and observational cohorts to carry it.
Sources: Tao, New England Journal of Medicine 2022, doi:10.1056/NEJMoa2206317; Jovin, New England Journal of Medicine 2022, doi:10.1056/NEJMoa2207576; Prabhakaran, Stroke 2026, doi:10.1161/STR.0000000000000513; Goyal, Lancet 2016, doi:10.1016/S0140-6736(16)00163-X; Tiainen, Journal of Stroke and Cerebrovascular Diseases 2022, doi:10.1016/j.jstrokecerebrovasdis.2022.106816; Goldhoorn, Stroke 2018, doi:10.1161/STROKEAHA.118.022352.
Case three — large core, onset unknown
The referral. A 77-year-old woman, last-seen-well time is unknown after being found by her neighbour, NIHSS 17 with left hemiplegia, neglect, and gaze deviation. CTA shows a right ICA terminus occlusion with poor collaterals (30%). NCCT ASPECTS 5. CT perfusion demonstrates a CBF<30% core of 76 mL, Tmax>6s volume of 128 mL, mismatch ratio 1.7. No thrombolysis was given.
The question it forces. Whether to open a right ICA terminus occlusion with an ASPECTS of 5 and a 76 mL core when nobody can say how long it has been occluded. The large-core evidence is stratified by time from last known well, and the imaging describes the tissue without dating the occlusion.
What the published evidence holds. Six randomised trials — RESCUE-Japan LIMIT, SELECT2, ANGEL-ASPECT, TENSION, LASTE and TESLA, which did not meet its primary endpoint — tested thrombectomy below ASPECTS 6, and ATLAS, their individual-patient meta-analysis with central imaging adjudication, found an adjusted generalised odds ratio of 1.63 (1.42–1.88) across 1,886 patients, benefit consistent across subgroups. Mortality was lower with thrombectomy (adjusted RR 0.82, 0.70–0.97), and the haemorrhage cost is real: in RESCUE-Japan LIMIT any intracranial haemorrhage occurred in 58% against 31%. Above an estimated core of 150 mL the evidence is limited, point estimates still favouring thrombectomy with intervals too wide to interpret; the core here is 76 mL.
The AHA science advisory on large-core thrombectomy records "strong evidence of its benefit in patients who have good prestroke functional status (mRS 0–1) and substantial stroke severity (National Institutes of Health Stroke Scale score ≥6) with occlusion of the internal carotid artery or proximal middle cerebral artery and a large ischemic core (ASPECTS 3–5)", and that "Data are limited to support EVT treatment of patients with matched core/perfusion, advanced age, or large core (ASPECTS 0–2) beyond 6 hours from LKW." This ASPECTS is 5; the unknown interval is one of the variables the second sentence names.
A trial-level Bayesian meta-regression across the six found the absolute effect below 0.10 at about 10 hours from onset and below 0.03 at about 18 hours, the second an extrapolation the authors flag as such; with last known well unknown, this case cannot be placed on that curve at all.
A meta-analysis of the two large-core trials requiring perfusion imaging found benefit in intermediate mismatch (OR 2.77, 1.11–6.89) and not in low mismatch, below a ratio of 1.2 or 10 mL (OR 1.47, 0.44–4.94); a later three-trial meta-analysis found no interaction, and at a ratio of 1.7 with 52 mL this case sits above both thresholds, in a band neither meta-analysis defined prospectively.
What this case shows is evidence stratified along an interval this patient does not have.
Sources: Yoshimura, New England Journal of Medicine 2022, doi:10.1056/NEJMoa2118191; Sarraj, New England Journal of Medicine 2023, doi:10.1056/NEJMoa2214403; Huo, New England Journal of Medicine 2023, doi:10.1056/NEJMoa2213379; Bendszus, Lancet 2023, doi:10.1016/S0140-6736(23)02032-9; Costalat, New England Journal of Medicine 2024, doi:10.1056/NEJMoa2314063; Yoo, JAMA 2024, doi:10.1001/jama.2024.13933; Sarraj, Lancet 2026, doi:10.1016/S0140-6736(26)00876-7; Gonzalez, Stroke 2025, doi:10.1161/STR.0000000000000481; Gajewski, Annals of Indian Academy of Neurology 2026, doi:10.4103/aian.aian_72_26; Chen, American Journal of Neuroradiology 2025, doi:10.3174/ajnr.A8553.