Giving clot busting drugs to stroke patients 4.5 hours or more after symptoms began improved recovery three months later, but came with a much higher rate of dangerous bleeding into the brain, according to a systematic review and meta-analysis of 14 randomized trials and 4,174 patients published in JAMA Network Open. The analysis, by researchers at Thomas Jefferson University Hospital and the University of Oxford, pooled trials in which patients were chosen by brain imaging rather than by the clock alone.
Key takeaways
- Late treatment made an excellent 90-day recovery about 22% more likely (risk ratio 1.22), roughly 6 to 8 extra good outcomes per 100 people treated.
- Symptomatic bleeding into the brain was about 2.4 times as common (risk ratio 2.44), roughly 1 extra bleed per 62 patients treated.
- Every trial selected patients with advanced brain imaging. The lesson is not that the clock matters less, it is that imaging shows who still has brain worth saving.
Why stroke treatment has a time window
Close to 9 in 10 strokes are ischemic, meaning a clot lodges in an artery feeding the brain and everything downstream of it stops getting oxygen. Brain tissue does not tolerate that for long. At the center of the blocked territory, cells begin dying within minutes.
Not all of it dies at once, though. Around that dying core sits a rim of tissue that is starving but still alive, kept barely fed by small side channels branching off neighboring arteries. Neurologists call it the penumbra, and it is the tissue treatment is actually fighting for. Reopen the artery and the penumbra can recover. Wait, and it converts to dead core, a little more with every passing minute.
Clot busting drugs, or thrombolytics, are how doctors reopen that artery without surgery. The original is tPA, short for tissue plasminogen activator and sold as alteplase; a newer version called tenecteplase is now widely used. It goes in through an IV and works by switching on the body’s own clot dissolving machinery. That machinery is not aimed at one spot. It thins the blood everywhere, which is why the main danger is bleeding, and the worst version of that is bleeding into the brain itself.
There is the trade behind the clock. As hours pass, less penumbra is left to rescue, so the possible benefit shrinks. The bleeding risk does not shrink along with it, and tissue that has been starved for hours is more fragile and bleeds more readily once blood flow returns. At some point the two lines cross and the drug does more harm than good.
The 4.5 hour mark is where the trials drew that line. The 1995 NINDS trial showed benefit when treatment started within 3 hours, and a European trial, ECASS III, extended that to 4.5 hours in 2008. It has been the rule ever since. But it is an average across thousands of patients, not a biological deadline inside any one person. Some people burn through their penumbra in 90 minutes. Others, with good side channel circulation, still have salvageable tissue many hours later, or wake up with symptoms and no idea when the stroke began. A clock cannot tell those two patients apart. A CT or MRI perfusion scan can, and that is exactly how the trials in this analysis picked who to treat.
What the study found
The team searched PubMed, Embase, and the Cochrane Central Register through March 3, 2026. Of the 4,174 patients pooled, 2,102 received intravenous thrombolysis and 2,072 got placebo or standard care. Nine of the 14 trials appeared in the past five years.
Patients given thrombolysis were about 22% more likely to reach an excellent functional outcome at 90 days, meaning a modified Rankin scale score of 0 or 1, which describes someone with no meaningful disability. The true benefit is very likely between 14% and 31% (95% CI 1.14 to 1.31). Using the broader bar of a good outcome, a Rankin score of 0 to 2, the gain was about 12% (risk ratio 1.12, very likely between 6% and 18%). In absolute terms the authors reported a number needed to treat of 12 to 16 for one additional favorable outcome, meaning about 6 to 8 more people per 100 treated end up meaningfully better off.
The harm ran the other way. Symptomatic intracerebral hemorrhage, bleeding into the brain bad enough to cause new symptoms, was about 2.4 times as common in the treated group (risk ratio 2.44). Here the uncertainty is wide: very likely somewhere between 45% higher and about four times higher (95% CI 1.45 to 4.09). The authors put this at a number needed to harm of 62, or about 1 extra symptomatic bleed per 62 people treated, roughly 16 per 1,000.
Deaths at 90 days did not clearly differ. The treated group was 13% higher on paper (risk ratio 1.13), but the range runs from 7% lower to 38% higher (95% CI 0.93 to 1.38), so the data cannot separate benefit from harm on mortality.
Dr. Kumar’s take
I am the person who gets called when one of these drugs causes a brain bleed. So I want to put the two numbers next to each other the way a family has to at 3 a.m.
Treat 100 people late and roughly 6 to 8 walk away meaningfully better than they otherwise would have. Treat those same 100 and about 1.6 have a symptomatic bleed into the brain they would not have had. That is close to a five to one ratio of help to harm. When I sit with a family, that is the sentence I use.
There is one detail most news stories will leave out, and it is the one that matters most. Every trial in this pool used advanced imaging to pick patients, and the authors say so directly, supporting extension of therapy “in appropriately imaging-selected patients.” The takeaway is not that the window is simply longer. The window was never really about time. It was about whether there is tissue left to save, and the clock was only ever a cheap proxy for that.
One caution. Pooling trials with different imaging rules and different definitions of a symptomatic bleed widens the real uncertainty beyond what the confidence intervals show. The bleeding range is already wide, running from a modest increase to a fourfold one, so the harm side is less settled than the benefit side.
What it means for you
If someone near you has stroke symptoms, none of this changes what you do: call emergency services immediately. Earlier is still better, because more brain is still alive earlier.
What it does change is what happens if you arrive late, or if you woke up with symptoms and nobody knows when they started. Those patients used to be turned away on the clock alone. This analysis supports scanning them instead, and asking the team one question: did the imaging show salvageable tissue?
Most of stroke risk is decided long before the ambulance. Blood pressure, atrial fibrillation, smoking, and diabetes do the heavy lifting, and the daily habits that protect an aging brain, including sleep, are worth more than any drug given in an emergency department.