ENFR
Home  /  Cases
Case bank

The gray zones, case by case

Fictional spoken referrals in the 3am voice — the cases where the guidelines go quiet.

Synthetic cases. These are fictional teaching vignettes, model-generated for testing clinical-reasoning systems. No real patients are described on these pages.

What this is

A bank of constructed referrals, written the way a referral actually arrives: spoken, compressed, incomplete, at three in the morning.

They are fiction — model-generated vignettes built for testing clinical-reasoning systems — and they are realistic precisely because the situations they describe are the ones every stroke physician recognises. No patient is described here; no patient material was used.

Why they exist

Gray zones only show themselves in concrete cases. A guideline can tell you that the evidence for thrombectomy in baseline-dependent patients is thin; it takes a 91-year-old with a dense M1 occlusion and a family in the relatives' room to show you what "thin" costs. The four cases below were chosen because each one walks straight into a territory where the randomised envelopes end — the same territories our in-press review of EVT contraindications works through.

The four cases

Case 1 of 4

Your stance is not stored, counted or sent — it stays in this browser tab. The real measurement is the gray-zone decision study.

Baseline dependency against the trial envelope Two bars of five segments. The upper bar has four segments filled, the baseline modified Rankin score of 4. The lower bar has one filled, the independent baseline the pivotal trials other than MR CLEAN asked for.

Case 02 — frailty and baseline dependency

as spoken

Got a tough one in bay six. 91-year-old male, sudden left-sided weakness witnessed by care home staff 3 hours ago. Baseline is poor, advanced vascular dementia, mRS 4, needs assist of two for all transfers. Current NIHSS is 18, completely neglected on the left, eyes turned to the right. Non-contrast CT shows loss of insular ribbon on the right. CTA shows a dense right M1 segment occlusion. Thrombolysis was withheld given his advanced baseline dependency and family preferences. The family is here now and wants to understand what else can be done for him.

Where would you stand

Why this one is hard

Every pivotal thrombectomy trial in the anterior circulation except MR CLEAN asked for functional independence before the stroke; a baseline mRS of 4 was excluded by design elsewhere and barely represented there, so there is no usable randomised estimate for this man. What makes the case hard is that the exclusion was never a finding — it was a choice — and age, taken alone, points the other way: in the HERMES pooled analysis the treatment effect was intact at 80 and beyond. The variable that actually carries this decision is not the year of birth but reserve, and reserve is precisely what the trials filtered out before they started counting.

Read the full case

Two vessel segments, two occlusions A wide proximal vessel segment rising from below with an occlusion mark across it, and a second, narrower and more distal segment above it carrying a second occlusion mark.

Case 05 — low NIHSS, tandem occlusion

as spoken

I have a 55-year-old male in ED, onset of mild left arm weakness and slurred speech 2 hours ago. Fully independent baseline, mRS 0. NIHSS is low, only 5, but CTA shows a complex lesion: there is a proximal right internal carotid artery occlusion at the bulb, and also a distal right M1 occlusion. Non-contrast CT is normal with an ASPECT score of 10. We have initiated intravenous thrombolysis given the early timeframe, and the infusion is running now. He has no medical contraindications. The concern is the tandem nature of the block despite the low clinical score.

Where would you stand

Why this one is hard

The referring physician has named the problem exactly: the deficit is mild and the anatomy is not. The pivotal thrombectomy trials were built around deficits well above this man's score — MR CLEAN's protocol admitted from NIHSS 2, yet the randomised record holds few patients like him — so the strongest evidence in stroke medicine barely reaches him. What the observational record adds is uncomfortable in both directions. Patients like him deteriorate early and often — in the largest minor-stroke LVO cohort, roughly one in eight worsened under thrombolysis alone, and a more proximal occlusion predicted it — and a tandem carotid-plus-M1 lesion is close to the definition of proximal.

Read the full case

A severity score that will not hold still A line swinging up and down across an evening, starting at a score of 4 and ending at 16. 4 16

Case 06 — basilar, fluctuating, on a DOAC

as spoken

Got an unstable posterior circulation case. 70-year-old male, sudden onset diplopia and slurred speech 5 hours ago. Baseline mRS is 2 due to severe knee osteoarthritis. Since arrival, his deficits have been fluctuating wildly; NIHSS was 4 on admission, but is now 16 with bilateral weakness and deteriorating consciousness. CTA shows a high-grade stenosis or subtotal occlusion of the mid-basilar artery. Non-contrast CT is unremarkable. MRI is unavailable until morning. Thrombolysis was not given because he was on therapeutic apixaban for atrial fibrillation, last dose taken 6 hours ago. He is currently intubated for airway protection.

Where would you stand

Why this one is hard

Almost every number in this vignette is unstable, starting with the one the trials are anchored to. The positive basilar trials wrote their envelopes around measurable severity — ATTENTION enrolled at NIHSS 10 and above — and this man has been 4 and 16 within the same evening. Which of those is his score? The fluctuation that makes the case frightening also makes it unmeasurable, and now he is intubated and sedated, so the examination that would answer the question has been traded for his airway.

Read the full case

Fourteen hours, and a large core A clock face with its hands at fourteen hours and the elapsed sweep shaded, beside a hemisphere outline holding a shaded core. 14 h

Case 07 — large core, late window

as spoken

Sorry to wake you. We have an 80-year-old female, last seen well 14 hours ago. Found on her kitchen floor. Prestroke mRS was 0. Her NIHSS is 21, dense left hemiplegia and neglect. Imaging performed was CT, CTA, and perfusion. CTA shows a right proximal internal carotid artery occlusion. CT perfusion shows a large area of severely depressed cerebral blood flow, under thirty percent of normal, measuring approximately 90 millilitres in the right MCA territory. The area of delayed perfusion is only slightly larger, around 105 millilitres. No thrombolysis was given due to the prolonged timeframe. Vitals are currently stable.

Where would you stand

Why this one is hard

Five years ago this page would have been short: a 90 mL core was a closed door. Six randomised trials and their individual-patient meta-analysis reopened it — benefit across the large-core stratum, mortality lower with treatment, haemorrhage cost real but paid — and her core sits well under the 150 mL mark beyond which the estimates stop being interpretable. On tissue alone, she is inside the new evidence.

Read the full case

How these cases were made

The synthetic label above is not an assertion to be taken on trust; the cases come off a written recipe, and the recipe is checkable. Every case is built on the same attribute spine — age, sex, time last seen well, NIHSS, occlusion site and side, collateral score, ASPECTS, CT perfusion — the same attributes a stroke-imaging platform shows, so the vignette reads the way a referral arrives. Internal-consistency rules keep the fiction honest: ASPECTS and core volume must move together, the deficit must lateralise opposite the occlusion, the mismatch ratio must be the arithmetic of its own numbers, and a case with no perfusion imaging gives no perfusion numbers. Some cases deliberately drop exactly one attribute — an unwitnessed onset, a failed perfusion run — because real referrals do.

And the generator operates under one hard rule: it outputs the case only — never an eligibility verdict, never a hint at which trials apply. The judgement a case is meant to exercise is deliberately left out of it. That is what makes these vignettes usable for testing clinical reasoning rather than leading it.

Source: the case generator's own specification, 2026.

Pages are numbered by bank ID; future cases join mechanically.

The invitation

Send us the anonymised cases where the guidelines went quiet.

Break the demo — send a case

For stroke physicians and vascular neurologists only. This is not a route for personal medical questions; if you are a patient or a relative, do not use this address. The demo does not run anonymously, and it is not a clinical service. Every case is reviewed and approved by the administrator before it runs; responses may take several days. Never use this for a live clinical decision — for acute stroke, follow your local pathway.

Your case must contain no images and no patient-identifying data, and it must be retrospective — not an active patient awaiting a decision. A case may appear on this site anonymously, as a constructed teaching case; you are named only if you ask to be.

Send your case