| Intervention | Outcome | ARR % | NNT | Read it as |
|---|---|---|---|---|
| All stroke types | ||||
| Stroke unit careCochrane 2007 | Death / dependency | 3.6 | 28 | Highest population yield — no imaging, no procedure. Admit everyone. |
| Ischaemic stroke | ||||
| Thrombolysis <4.5hEmberson 2014 · TNK ≡ alteplase (AcT, TRACE-2) | mRS 0–1 | 5.2–9.8 | 10–19 | Earlier = larger ARR. TNK now an equal-footing alternative. Not for non-disabling minor stroke. |
| Thrombolysis 4.5–9h / WUSEXTEND 2019, pooled 2019 | mRS 0–1 | ~10 | 10 | Perfusion-mismatch only; core capped ~70mL. Enrolled cores were much smaller — mind haemorrhagic conversion. |
| Thrombectomy <6hMR CLEAN, ESCAPE, SWIFT-PRIME 2015 | mRS 0–2 | 20 | 5 | Functional independence. Proximal anterior LVO. |
| Thrombectomy 6–24hDAWN 2017, DEFUSE-3 2018 · large core: SELECT2, ANGEL-ASPECT | mRS 0–2 (CTP-sel) mRS 0–3 (large core) |
32 16 |
3.1 6 |
Best-selected, smallest NNT. Large-core pooled TESLA missed primary — not a homogeneous group. |
| AspirinIST, CAST 1997 | Recurrent stroke / death | 0.9 | 111 | Small per-patient, huge population — cheap and near-universal. Don't dismiss the big number. |
| HemicraniectomyVahedi 2007 pooled | Death | 49 | 2 | NNT 2 prevents death, not disability. Many survivors land at mRS 4 — say so when consenting. |
| Intracerebral haemorrhage | ||||
| Intensive BP lowering <140INTERACT2 (Anderson 2013); ATACH-2 2016 | Death / dependency | no sig. benefit |
Slide was wrong here. INTERACT2 primary endpoint not significant (OR 0.87, 0.75–1.01); benefit was the softer ordinal mRS shift only. ATACH-2: no benefit, ?renal harm. Lower smoothly, don't crash to <140. | |
| Anticoagulation reversalthe actual high-yield ICH move | Haematoma expansion | do it fast |
The intervention the original slide omitted. See block 6 for agents. | |