The Ever-Expanding Thrombolysis Window: From 4.5 Hours to 24 — Are We There Yet?

The foundational approval of intravenous alteplase for acute ischemic stroke rested on data from the NINDS trial (1995), which established a 3-hour therapeutic window. This was extended to 4.5 hours by ECASS III (2008), which demonstrated modest but significant benefit in carefully selected patients (mRS 0-1: 52.4% vs 45.2%, OR 1.34). For more than a decade, 4.5 hours remained the outer boundary of routine thrombolysis practice, accepted in guidelines and embedded in clinical reflexes worldwide.

But even as these time windows became doctrine, the paradigm was shifting beneath them. Advanced imaging could identify salvageable tissue hours—sometimes days—after symptom onset. Thrombectomy trials like DAWN and DEFUSE 3 proved that tissue-based selection could identify patients who benefited from intervention up to 24 hours from last known well. The question became inevitable: if mechanical thrombectomy works in extended windows with the right imaging criteria, why wouldn't thrombolysis?

The Transition to Tissue-Based Selection

The first significant departure from rigid time criteria came from trials addressing strokes of unknown onset, particularly wake-up strokes. WAKE-UP (2018) used DWI-FLAIR mismatch as a surrogate for tissue viability, enrolling 503 patients with ischemic lesions visible on diffusion-weighted imaging but not yet evident on FLAIR. The approach worked: favorable outcome (mRS 0-1) occurred in 53.3% of thrombolysis patients versus 41.8% in placebo (adjusted OR 1.61, 95% CI 1.09-2.36). Importantly, symptomatic intracranial hemorrhage rates remained low at 2.0% versus 0.4%, well within the acceptable risk threshold established in earlier trials.

EXTEND (2019) took the next logical step, using CT or MRI perfusion imaging to identify patients with mismatch between the ischemic core and hypoperfused penumbra in the 4.5 to 9-hour window. Among 225 patients, 35.4% of those receiving alteplase achieved mRS 0-1 compared to 29.5% in the placebo group (risk ratio 1.44, 95% CI 1.01-2.06). The signal was modest but present, and the safety profile was reassuring. These trials established proof of concept: penumbral imaging could extend the treatment window safely, and in appropriately selected patients, efficacy followed.

What WAKE-UP and EXTEND shared was methodological restraint. Their imaging criteria were conservative, enriching for patients with substantial salvageable tissue. Their sample sizes were moderate. And critically, both trials enrolled mixed populations—some with large vessel occlusions, many without. They suggested that late-window thrombolysis was feasible, but they left open a crucial question: in the era of near-universal thrombectomy access for LVO patients, who actually benefits from extended-window alteplase?

The Large Vessel Occlusion Paradox

Between 2024 and early 2025, three major trials tested thrombolysis in the 4.5 to 24-hour window using perfusion imaging selection. All three targeted populations enriched for large vessel occlusions. All three demonstrated acceptable safety. And all three struggled to show meaningful functional benefit.

TIMELESS (2024) randomized 458 patients with proximal occlusions and favorable perfusion profiles to tenecteplase versus placebo, with nearly all patients undergoing thrombectomy. The primary outcome—mRS 0-2 at 90 days—occurred in 56.6% of the tenecteplase group versus 54.5% in placebo, a difference that failed to reach statistical significance. Symptomatic hemorrhage rates were 3.2% versus 2.3%, numerically higher but not prohibitively so. The trial was safe. It simply didn't move the needle on outcomes.

TRACE-III (2024), conducted primarily in China, enrolled 502 patients using similar perfusion criteria in the same extended window. The investigators reported an efficacy signal: 33% achieved mRS 0-1 with thrombolysis versus 24.2% without (adjusted OR 1.37, 95% CI 1.04-1.81 p=0.03). The point estimate suggested benefit, but the confidence interval crossed unity. The trial's conclusion acknowledged this ambiguity—possible benefit, but not definitive.

CHABLIS-T II (2025) provided perhaps the most revealing data. Among 356 patients with LVO and perfusion mismatch treated between 4.5 and 24 hours, thrombolysis significantly improved angiographic recanalization rates before thrombectomy (48% vs 31%, p=0.003). Yet this radiographic success did not translate into better functional outcomes. The rate of mRS 0-2 was virtually identical: 48.2% versus 47.2%. Here was the paradox laid bare—thrombolysis was doing what it was supposed to do biologically, but in patients destined for thrombectomy, that biology was clinically redundant.

The interpretation was becoming clear. In LVO patients eligible for endovascular therapy, late-window thrombolysis might improve vessel patency, might facilitate thrombectomy, might even accelerate time to reperfusion—but none of this was sufficient to overcome the dominant effect of mechanical clot retrieval. One trial doesn't change guidelines. A body of evidence does. And by mid-2025, the body of evidence in LVO populations was consistently neutral.

Efficacy Emerges in Non-LVO Populations

The strategic pivot came with the recognition that non-LVO patients—those with distal occlusions, branch vessel occlusions, or extensive microvascular disease—represented a fundamentally different treatment scenario. These patients were ineligible for thrombectomy under most contemporary criteria. For them, thrombolysis in the extended window wasn't redundant; it was the only available reperfusion strategy.

OPTION (2026), published in JAMA, was designed with this logic in mind. The trial enrolled 568 patients presenting 4.5 to 24 hours from last known well, all with non-LVO strokes and CT perfusion evidence of mismatch (core <70 mL, mismatch ratio ≥1.2). Patients were randomized to alteplase versus standard care. The primary outcome—mRS 0-1 at 90 days, representing complete or near-complete recovery—occurred in 43.6% of the teneceteplase group compared to 34.2% in controls (adjusted risk ratio 1.28, 95% CI 1.09-1.51, p=0.003). The number needed to treat was 11. Symptomatic hemorrhage occurred in 2.8% versus 0%, a significant increase but within the range observed in earlier thrombolysis trials and far below rates that would negate the benefit.

OPTION worked because it asked the right question of the right population. Patients with non-LVO strokes and preserved penumbra represented an underserved group—symptomatic enough to present to the emergency department outside the conventional window, imaging-viable enough to harbor salvageable tissue, yet excluded from both standard thrombolysis (by time) and thrombectomy (by occlusion location). In this population, alteplase wasn't competing with a superior alternative. It was filling a therapeutic void.

Supporting evidence emerged from other directions. ROSE-TNK (2023), though primarily a safety study, enrolled 80 patients with non-LVO strokes in the 4.5 to 24-hour window and demonstrated hemorrhage rates comparable to historical controls. EXPECTS (2025), focusing on posterior circulation strokes—a notoriously difficult subgroup to study—showed a trend toward benefit in basilar artery branch occlusions treated with extended-window thrombolysis, though the sample was too small for definitive conclusions. HOPE (2025), an open label clinical trial of mixed LVO and non-LVO patients, found that the treatment effect was concentrated almost entirely in the non-LVO subgroup (interaction p=0.04).

The pattern was consistent. Efficacy followed patient selection. In populations where thrombolysis represented the primary or only reperfusion modality, extending the window with perfusion imaging made both biological and clinical sense. In populations where thrombectomy was available and guideline-recommended, it did not.

Implementation, Evidence Gaps, and the Path Forward

The OPTION results have catalyzed rapid adoption at experienced stroke centers. Many academic hospitals in the United States now maintain 24-hour thrombolysis protocols for non-LVO patients with favorable perfusion profiles, often using automated CT perfusion software (RAPID, Viz.ai, or similar platforms) to streamline imaging interpretation. But implementation is uneven, and the devils—as always—reside in the details.

First, perfusion imaging criteria remain heterogeneous across protocols. OPTION used a core volume threshold of 70 mL and a mismatch ratio of 1.2, but TIMELESS accepted cores up to 90 mL, and institutional protocols vary further. We lack head-to-head comparisons to know whether these thresholds are interchangeable or whether subtle differences in patient selection might explain variability in outcomes. The temptation to liberalize criteria at the bedside—treating a patient with a 75 mL core because "close enough"—must be resisted until we have data to support such flexibility.

Second, bleeding risk stratification remains imperfect. The OPTION hemorrhage rate of 2.8% was acceptable in aggregate, but subgroup analyses suggested higher risk in patients with more severe baseline strokes (NIHSS >15), extensive white matter disease, and those on antiplatelet therapy. These are precisely the patients most likely to be excluded from trials but encountered routinely in practice. Real-world registries will be essential to understanding whether trial-derived safety profiles generalize to broader populations.

Third, and perhaps most challenging, is the implementation gap between comprehensive stroke centers and community hospitals. OPTION-style protocols require immediate access to advanced perfusion imaging, ideally with automated processing, and expert interpretation of mismatch profiles. Many community hospitals lack this infrastructure. Telemedicine networks can bridge some of this gap, but logistical realities—transfer times, imaging compatibility, workflow integration—remain formidable. The question isn't just what works; it's what's implementable.

From a guideline perspective, the lag is predictable. The AHA/ASA last updated acute stroke guidelines in 2024, before OPTION was published. The European Stroke Organisation released extended-window recommendations in early 2025 based on TIMELESS and TRACE-III, offering a weak conditional recommendation for selected patients but acknowledging the evidence was inconclusive. OPTION changes this calculus fundamentally, but guideline committees move deliberately. Expect updated recommendations later, likely endorsing extended-window thrombolysis for non-LVO patients with perfusion mismatch as a Class IIa or IIb recommendation.

Regulatory pathways are equally uncertain. Alteplase remains approved in the 0-4.5 hour window; tenecteplase, increasingly used off-label, has no FDA approval for stroke in any time window in the United States. OPTION used alteplase, which simplifies regulatory interpretation, but many centers prefer tenecteplase for its bolus administration and pharmacologic profile. Whether OPTION's results can be extrapolated to tenecteplase—and whether regulatory agencies will accept such extrapolation—remains unresolved.

Looking forward, several questions demand answers. Can simplified imaging criteria (e.g., ASPECTS-based selection or DWI-FLAIR mismatch) substitute for perfusion imaging in settings where CTP is unavailable? Early feasibility studies are underway. Do patients with very mild deficits (NIHSS <6) benefit from extended-window treatment, or does the hemorrhage risk outweigh potential gains? OPTION excluded these patients, leaving the question open. And critically, what is the upper limit of the core volume that permits safe treatment? OPTION capped cores at 70 mL, but biological plausibility arguments suggest some patients with larger cores might still benefit if the penumbra is extensive.

To understand where we're going, we need to remember where we've been. Twenty years ago, thrombolysis beyond three hours was considered reckless. Fifteen years ago, 4.5 hours felt like pushing the boundary. Today, we're treating selected patients at 24 hours and seeing meaningful benefit—not because the drug has changed, but because our ability to identify the right patients has evolved. The window hasn't expanded uniformly; it has fragmented into population-specific windows, each defined by imaging physiology rather than clock time.

Bottom Line: Are we there yet?

Are we there yet? For non-LVO patients with demonstrable penumbral mismatch, the answer is increasingly yes—24 hours is a defensible and evidence-supported window. For LVO patients undergoing thrombectomy, extended-window thrombolysis remains an open question, with current data suggesting minimal added value. For patients in community hospitals without perfusion imaging, for those with atypical presentations, for subgroups underrepresented in trials—no, we are not there yet. We have tools that work, but we are still learning whom to treat, how to select them efficiently, and how to scale these advances beyond academic centers.

The arc of thrombolysis research has always been one of progressive refinement—narrower patient selection enabling broader time windows, better imaging justifying later treatment. OPTION represents the latest inflection point in that arc, not its terminus. Stroke neurology has moved from time-based dogma to tissue-based pragmatism. The next phase will be integration—translating trial evidence into bedside algorithms, reconciling advanced imaging with resource constraints, and ensuring that patients in all settings benefit from what we have learned. The evidence base is strong enough to guide practice. Implementation is the challenge that remains.