Tenecteplase versus alteplase: what the trials show
Tenecteplase versus alteplase in ischemic stroke: what the major trials found, how non-inferiority margins work, and what the evidence cannot settle.
By Manouchehr Hessabi, MD, MPH
For decades, the clot-dissolving drug used for ischemic stroke was alteplase. In January 2026, the American Heart Association and American Stroke Association published a guideline that endorses either alteplase or a newer drug, tenecteplase, within 4.5 hours of symptom onset. That change rests on a small number of large randomized trials, and those trials share a design that is widely misread: they were non-inferiority trials.
This article walks through what the trials found, what "non-inferior" actually claims, why one trial of a higher dose was stopped early, and what the evidence still cannot settle. It is educational and not a substitute for personal medical advice. The choice of treatment in an individual stroke is made by the treating stroke team.
What is the difference between tenecteplase and alteplase?
A few terms first. An ischemic stroke is a stroke caused by a clot blocking an artery that supplies the brain. Thrombolysis is treatment with a drug that dissolves the clot. Both alteplase and tenecteplase are tissue plasminogen activators, drugs that trigger the body's own clot-dissolving system.
The practical difference is how they are given. In the Canadian AcT trial protocol, alteplase was given as a small bolus (a single rapid injection) followed by a 60-minute infusion of the remainder. The ORIGINAL trial report describes tenecteplase as a bioengineered variant of alteplase with greater fibrin specificity and a longer half-life, which allows it to be given as a single bolus. The doses mentioned below are trial protocol facts, reported so the results can be read correctly, not treatment guidance.
Two measurement tools recur throughout. The modified Rankin Scale (mRS) grades disability after stroke from 0 (no symptoms) to 6 (death); all of the trials below used a score of 0 or 1 at about three months as their main measure of a good outcome. The National Institutes of Health Stroke Scale (NIHSS) scores how severe a stroke is at presentation.
What does non-inferior actually mean?
Most people are familiar with superiority trials, which ask whether treatment A is better than treatment B. A non-inferiority trial asks a different question: is the new treatment not worse than the established one by more than a prespecified amount? That amount is called the non-inferiority margin, and it is written down before the trial begins.
The margin is the whole argument. A trial "meets non-inferiority" when the entire plausible range of the difference, usually the 95% confidence interval, stays on the acceptable side of the margin. Readers should always look for the margin first, because a result is only as reassuring as the margin it was tested against.
The tenecteplase trials wrote their margins in two different ways:
- An absolute margin. AcT required that the lower bound of the 95% confidence interval for the difference in the proportion of patients with a good outcome be above minus 5 percentage points. In plain words, the data had to rule out tenecteplase being worse by 5 points or more.
- A relative margin. TRACE-2 and ORIGINAL expressed the comparison as a risk ratio, the proportion with a good outcome on tenecteplase divided by the proportion on alteplase. The lower bound of its 95% confidence interval had to be above 0.937, which rules out tenecteplase producing good outcomes at a rate more than about 6% lower in relative terms than alteplase.
"Non-inferior" is not the same as "equivalent" or "the same." A trial can meet its margin while its point estimate leans slightly toward either drug. And a trial that fails to meet its margin has not, by that fact alone, proven the new drug worse; it has only failed to rule out an unacceptable difference.
What did the three large trials find?
Three large trials of tenecteplase at 0.25 mg/kg against alteplase at 0.9 mg/kg each met their prespecified margin.
| Trial | Setting | Patients analyzed | Good outcome (mRS 0 to 1), tenecteplase vs alteplase | Margin met? |
|---|---|---|---|---|
| AcT (2022) | 22 centers, Canada | 1,577 | 36.9% vs 34.8% | Yes |
| TRACE-2 (2023) | 53 centers, China | 1,401 | 62% vs 58% | Yes |
| ORIGINAL (2024) | 55 centers, China | 1,465 | 72.7% vs 70.3% | Yes |
AcT, published in The Lancet by Menon and colleagues (2022), was a pragmatic, registry-linked trial in which outcomes were assessed by blinded review. The unadjusted risk difference was 2.1 percentage points (95% CI minus 2.6 to 6.9), so the lower bound sat above the minus 5 point margin. Symptomatic intracerebral hemorrhage within 24 hours occurred in 3.4% of the tenecteplase group and 3.2% of the alteplase group, and death within 90 days in 15.3% and 15.4%.
TRACE-2, published in The Lancet by Wang and colleagues (2023), enrolled patients eligible for thrombolysis but ineligible for, or declining, endovascular thrombectomy (mechanical removal of the clot through a catheter). The risk ratio for a good outcome was 1.07 (95% CI 0.98 to 1.16). Symptomatic intracranial hemorrhage within 36 hours was 2% in both groups. Mortality at 90 days was 7% versus 5% (risk ratio 1.31, 95% CI 0.86 to 2.01), an interval wide enough to be compatible with no difference, some harm, or some benefit.
ORIGINAL, published in JAMA by Meng and colleagues (2024), found a risk ratio for a good outcome of 1.03 (95% CI 0.97 to 1.09). Symptomatic intracerebral hemorrhage occurred in 1.2% of each group, and 90-day mortality was 4.6% versus 5.8% (risk ratio 0.80, 95% CI 0.51 to 1.23). The published disclosures list two authors as employees of Boehringer Ingelheim (China).
The three trials report very different absolute rates of good outcomes, from about 35% to about 72%. That most likely reflects differences in who was enrolled and how severe their strokes were: the comparison that matters is within each trial, between its two arms.
Why was one trial stopped early?
"Tenecteplase" is not one intervention. The evidence attaches to a dose and a population.
The Norwegian NOR-TEST 2 trial, part A, reported in Lancet Neurology by Kvistad and colleagues (2022), tested a higher dose, 0.4 mg/kg, against standard alteplase in moderate or severe stroke (NIHSS 6 or more). Enrollment stopped after a per-protocol safety review found an imbalance in symptomatic intracranial hemorrhage that crossed the prespecified stopping criteria. Among patients analyzed, a good outcome occurred in 32% on tenecteplase versus 51% on alteplase, any intracranial hemorrhage in 21% versus 7%, and death at three months in 16% versus 5%. The authors concluded they could not show non-inferiority at that dose, and at the time of that report, part B of the trial was continuing with 0.25 mg/kg.
This is what stopping rules are for. They are written before a trial starts so that a safety signal triggers action rather than debate. The trade-off is that a stopped trial is usually small, 204 patients in this analysis, so its estimates are imprecise. It can show that a dose carried an unacceptable risk in that setting. It cannot, on its own, describe the dose's effect precisely.
What changed in the 2026 guideline?
The 2026 AHA/ASA Guideline for the Early Management of Patients With Acute Ischemic Stroke, published online January 26, 2026 and in the August 2026 issue of Stroke, replaces the 2018 guideline and its 2019 update. Its summary lists new evidence on thrombolytic choice among the key updates. The American Heart Association's announcement states that the guideline endorses either tenecteplase or alteplase within 4.5 hours of symptom onset, and notes that a single dose of tenecteplase simplifies treatment compared with the 60-minute alteplase infusion. A published correction to the guideline also appears in the same issue.
In practical terms, a single injection means there is no hour-long infusion to manage, including when a patient is transferred between hospitals.
What the trials cannot tell us
- Open-label design. In all four trials, patients and treating clinicians knew which drug was given, which is hard to avoid when one drug is a bolus and the other an infusion. Outcome assessment was blinded, which protects the measurement of disability at three months. It does not protect against differences in care along the way.
- Analysis populations. AcT analyzed an intention-to-treat population. TRACE-2 used a modified intention-to-treat population of patients who received the allocated drug, and ORIGINAL a full analysis set. In non-inferiority trials this choice matters in an unusual way: anything that blurs the difference between arms, such as patients not receiving their assigned treatment, pushes results toward "no difference," which is exactly the conclusion the trial is testing.
- Populations. TRACE-2 excluded thrombectomy candidates, two of the three large trials were conducted in China, and NOR-TEST 2 used a different dose in more severe strokes. Results apply most directly to patients like those enrolled.
- Mortality. Point estimates for death leaned in opposite directions in TRACE-2 and ORIGINAL, with confidence intervals that include no difference in both. The honest reading is that these trials were not built to settle mortality on their own.
Where this sits in stroke outcomes research
The tenecteplase story is a clear case study in how clinical evidence accumulates: a margin fixed in advance, trials in different countries and health systems testing the same question, a safety signal at a different dose handled by a prespecified rule, and a guideline that weighs the whole body of evidence rather than any single result. It also shows why reading "non-inferior" correctly matters. The claim is narrow, specific, and only as strong as the margin behind it.
Readers interested in how recovery after stroke is measured and studied can explore the stroke outcomes research program on this site, along with the peer-reviewed publications behind it.