Stroke treatment is a set of windows. Each closes at a different hour, each is opened or shut by a different scan, and each has a benefit and a harm that are known in numbers. This chapter gives all of them in absolute terms, then the order in which the questions are asked, so that "which treatment, if any" becomes a sequence rather than a guess. The first-minutes version, for the moment it happens, is the emergency chapter; the two agree on every figure.
1. Before any window: is it a bleed or a blockage?
About 85% of strokes are ischaemic (an artery blocked) and about 15% are haemorrhagic (an artery burst). Every treatment for one is harmful in the other, so the first CT scan decides which page of this chapter applies. Both UK and US guidance ask for that scan within an hour of arrival, and in the UK the National Clinical Guideline for Stroke says immediately on arrival for anyone who could be a reperfusion candidate. Where a large-vessel blockage is possible, the CT angiogram (dye in the arteries) is done at the same sitting; where the patient is outside the standard windows, perfusion imaging (which shows how much brain is still salvageable) is added.
2. Ischaemic stroke: the windows
Intravenous thrombolysis, 0 to 4.5 hours
A clot-dissolving drug into a vein. Tenecteplase 0.25 mg/kg to a maximum of 25 mg, as a single bolus, or alteplase 0.9 mg/kg over an hour. The two are equivalent on outcomes (in the AcT trial, 36.7% free of significant disability at 90 days on tenecteplase against 35.9% on alteplase, with symptomatic bleeding 3.2% against 3.1%), so the choice is practical: a five-second injection instead of an infusion pump. Both the 2026 American guideline and the UK guideline endorse either; NICE approved tenecteplase for the NHS in 2024.
| Treated within | Odds of being free of significant disability | What that means |
|---|---|---|
| 0–90 minutes | 2.55 (1.44–4.52) | The largest benefit. Almost nobody arrives in time |
| 91–180 minutes | 1.64 (1.12–2.40) | A third of the benefit already gone |
| 181–270 minutes | 1.34 (1.06–1.68) | Half of what it was. Roughly 14 people treated for one extra person free of disability |
| 271–360 minutes | 1.22 (0.92–1.61) | No longer statistically significant. The window closes at 4.5 hours |
The harm does not shrink with delay; only the benefit does. Large bleeding into the brain occurred in 5.2% of treated patients against 1.0% of untreated across every interval. Early deaths rise; deaths at three to six months do not. The reason to move fast is that every minute spends the benefit while the risk stays fixed.
Who it is for, in 2026. Anyone with a disabling deficit within 4.5 hours, regardless of how mild the stroke scale says it is, and without needing perfusion imaging first. That sentence is new in the American guideline and matches UK practice. For a non-disabling deficit (an isolated numb patch, say), trials found no benefit from thrombolysis, and both guidelines say give two antiplatelets instead. The main reasons a person cannot have it are a bleed on the scan, recent surgery or bleeding, a blood pressure above 185/110 that cannot be brought down, and taking an anticoagulant.
Thrombectomy, 0 to 6 hours
For a large-vessel occlusion in the front of the brain (the internal carotid or the first part of the middle cerebral artery), a catheter is threaded from the groin or wrist to the clot and the clot is pulled out. It is the most effective treatment in the whole of stroke medicine. Pooled across the first five trials (HERMES), 46% of treated patients were independent at 90 days against 26.5% with medical treatment alone, which works out at about one extra independent person for every 2.6 treated. Symptomatic bleeding was no higher. The UK guideline asks for it in anyone with a proximal occlusion and a disabling deficit (a stroke scale of 6 or more), and both guidelines say give the thrombolytic first if eligible and do not let it delay the procedure.
Thrombectomy, 6 to 24 hours
Two trials extended the window using imaging to find people whose brain was still salvageable. DAWN (6–24 hours, chosen by a mismatch between how severe the stroke looked and how small the dead core was): 49% independent against 13%. DEFUSE 3 (6–16 hours, chosen by perfusion imaging): 45% against 17%. Those are among the largest treatment effects in medicine, and they only apply to people who meet the imaging criteria. UK guidance: consider thrombectomy 6 to 24 hours after onset, including wake-up strokes, on a combination of the CT score and imaging mismatch. US 2026 guidance: the same, and the imaging can be plain CT and CT angiography where perfusion is not available.
Large strokes: the core is no longer the barrier it was
Until 2023, a large area of already-dead brain on the first scan excluded thrombectomy. Three trials changed that. SELECT2: 20% independent against 7%. ANGEL-ASPECT: 30% against 11.6%. RESCUE-Japan LIMIT showed the same. The absolute benefit is smaller than in a small-core stroke, but so is the alternative: without treatment most of these people die or are left bed-bound. The 2026 American guideline now recommends thrombectomy for selected large-core strokes; the UK guideline reached the same position in 2023.
The basilar artery
A blockage of the artery supplying the brainstem is the deadliest stroke there is and, until 2022, the one with the least evidence. ATTENTION (within 12 hours) and BAOCHE (6–24 hours) both found about 46% of patients had a good outcome with thrombectomy against 23–24% without, with more bleeding (around 5% against under 1%) and fewer deaths. The 2026 American guideline gives a strong recommendation up to 24 hours for a stroke scale of 10 or more. NICE's guideline, amended in 2025, now says consider it up to 24 hours with imaging showing salvageable tissue; the National Clinical Guideline said 12 hours in 2023.
Thrombolysis beyond 4.5 hours: where the evidence is still being sorted
Three situations, three answers.
| Situation | Trial | Result | UK | US 2026 |
|---|---|---|---|---|
| 4.5–9 hours, or waking with it, salvageable tissue on perfusion imaging | EXTEND (alteplase) | 35.4% free of disability against 29.5%; symptomatic bleeding 6.2% against 0.9%; no difference in death | Consider thrombolysis if imaging criteria are met (NCGS 2023) | "Reasonable" (class 2a) |
| Waking with it, MRI showing the stroke is fresh | WAKE-UP (alteplase) | 53.3% against 41.8%; symptomatic bleeding 2.0% against 0.4% | Consider, using the MRI mismatch | "Reasonable" (2a) |
| 4.5–24 hours, large-vessel occlusion, no thrombectomy available | TRACE-III (tenecteplase) | 33.0% free of disability against 24.2%; bleeding 3.0% against 0.8%; deaths unchanged at about 13%. Roughly 11 treated for one extra person free of disability; roughly 45 for one extra brain bleed | Not in UK guidance | "May be beneficial" (2b) |
| 4.5–24 hours, large-vessel occlusion, thrombectomy available | TIMELESS (tenecteplase) | No better than placebo. Symptomatic bleeding 3.2% against 2.3% | Not recommended | Not recommended; thrombectomy is the treatment |
The pattern is consistent: late thrombolysis helps when it is the only reperfusion on offer, and adds nothing when the clot is going to be removed anyway. HOPE, a fourth trial using alteplase in the late window, found more bleeding (3.8% against 0.5%) and no difference in deaths.
3. Haemorrhagic stroke: the first hour
There is no clot to remove; the treatment is to stop the bleed growing, and the evidence says the first hour is when that is decided. The UK guideline is specific and it is the more demanding of the two:
- Blood pressure. If systolic pressure is 150–220 and the bleed is within six hours of onset, lower it to 130–139 within one hour and hold it there for seven days (NCGS 2023). The American guideline says 130–140 is safe and may help, and warns against dropping below 130 or dropping fast.
- Anticoagulant reversal, within the hour. Warfarin: prothrombin complex concentrate and vitamin K. Dabigatran: idarucizumab. Apixaban, rivaroxaban, edoxaban: prothrombin complex concentrate in the UK; andexanet alfa is used in the US and is restricted in the UK to trials for this indication, because the ANNEXA-I trial found it controlled the bleed more often but caused more clots. This is why the medicine list matters at the scene.
- The bundle. INTERACT3 showed that doing these together — pressure, reversal, glucose and temperature control, all within an hour — improved outcomes in 7,000 patients. SSNAP now audits UK hospitals on whether reversal or blood pressure treatment happened within an hour of arrival.
- Surgery for a cerebellar bleed with pressure on the brainstem, for hydrocephalus, and for a minority of other large bleeds; most are managed without an operation. Tranexamic acid is not standard. A subarachnoid haemorrhage (a burst aneurysm on the surface of the brain) is a different disease with its own pathway: coiling or clipping of the aneurysm.
4. The decision sequence
This is the order in which a stroke team works through the windows. It is not a substitute for the team; it is what to expect them to have asked.
When was the person last known well?
Fix the time. It decides everything below.
CT now. Bleed or blockage?
Bleed: go to section 3, and the hour starts. Blockage or nothing visible: continue.
Is the deficit disabling, and are we inside 4.5 hours?
Yes and yes: thrombolysis now, tenecteplase or alteplase, unless contraindicated. Do not wait for the angiogram to give it. Non-disabling: two antiplatelets instead.
Is there a large-vessel occlusion on the angiogram?
Yes, inside 6 hours, disabling deficit: thrombectomy, at a centre that does it, with transfer if needed. The thrombolytic is given first if eligible and must not delay the transfer.
Outside 6 hours but inside 24, with an occlusion?
Perfusion imaging, or CT and angiography where perfusion is unavailable. Salvageable tissue: thrombectomy. Large core: still thrombectomy in selected patients. Basilar occlusion: thrombectomy to 24 hours.
Outside 4.5 hours and no thrombectomy is possible?
4.5–9 hours or wake-up with salvageable tissue: thrombolysis is reasonable in both countries. 9–24 hours with an occlusion: US guidance says it may help; UK guidance does not include it.
None of the above applies?
Aspirin, a stroke unit, and the 72-hour checklist. Most strokes end here, and the checklist is where most of the remaining benefit is.
5. What to ask when you are being asked to decide
The five questions in the emergency chapter still apply: bleed or blockage; benefit and harm in absolute numbers; what happens with nothing; how long you have to decide; and whether the clinician would offer it to their own relative. The numbers in this chapter are the ones a stroke team should be able to give you, and "about one in ten do better, about one in forty-five have a brain bleed" is a sentence, not a shrug.
6. Where UK and US guidance differ
- Late thrombolysis 9–24 hours without thrombectomy. US: may be beneficial. UK: not included. Both agree on 4.5–9 hours and wake-up with imaging.
- Basilar occlusion. US: strong recommendation to 24 hours with a stroke scale of 10 or more. UK: NICE now says consider to 24 hours; the 2023 clinical guideline said 12.
- Haemorrhage blood pressure. UK sets 130–139 within one hour for seven days; US says 130–140 is safe and may help. The UK wording is firmer.
- Reversal of the newer anticoagulants. US uses andexanet; UK uses prothrombin complex concentrate outside trials.
- Delivery. The US guideline endorses mobile stroke units (a CT scanner in the ambulance) and taking suspected large-vessel strokes straight to a thrombectomy centre. The UK has neither as policy; the equivalent is the pre-alert, the regional network and the transfer. The Sentinel audit's door-to-needle time in England was 56 minutes in 2024–25 and thrombectomy still reaches a fraction of the 10–15% of strokes estimated to be eligible. The guidance is close; the delivery is not.