Tenecteplase Versus Alteplase: How the Noninferiority Trials Read

Stroke treatment is a race against injured brain tissue: when an artery is blocked, an eligible patient may receive an intravenous medicine that dissolves clot, undergo catheter thrombectomy, or receive both. Brain imaging and a rapid contraindication check are essential because the same symptoms can come from bleeding, and a clot-dissolving medicine can cause severe hemorrhage.

Alteplase built the modern evidence base for intravenous thrombolysis, and tenecteplase is a modified tissue plasminogen activator that can be given as a single bolus, while alteplase for stroke is delivered as a bolus followed by an infusion. The practical difference is attractive, but convenience alone cannot establish equal clinical outcomes. That question required randomized trials.

First recognize the emergency#

Sudden facial droop, arm or leg weakness, numbness, speech difficulty, vision loss, severe imbalance, or another abrupt neurologic deficit can be a stroke. Call emergency services immediately and record the last time the person was known to be well. Do not drive the person, give food, or wait for symptoms to resolve.

This article explains trial interpretation. It cannot tell you whether anyone is eligible. Clinicians must establish whether the event is ischemic, identify bleeding and other mimics, assess time and imaging, review recent surgery or trauma and anticoagulants, measure blood pressure and glucose, and consider thrombectomy. Treatment decisions occur in a monitored stroke system because serious bleeding can be fatal.

Why investigators chose noninferiority#

A superiority trial asks whether a new treatment produces better outcomes, and a noninferiority trial asks whether it preserves enough of an effective comparator's benefit that any possible loss is smaller than a prespecified margin.

That design is reasonable when the new option may offer other advantages, and tenecteplase's single-bolus administration can simplify preparation, avoid a one-hour infusion, and speed transfer to a thrombectomy center. Those benefits matter only if clinical effectiveness is not meaningfully worse. That is the whole point of the design.

Before enrollment, investigators choose an outcome, an effect scale, an analysis population, and a margin. Suppose the margin permits an absolute outcome difference as low as -5 percentage points. If the lower bound of the confidence interval is above -5, the trial can declare noninferiority even when the point estimate slightly favors the comparator, and a different margin could change the conclusion from the same data.

The margin should preserve a clinically acceptable part of the established treatment benefit and be justified from historical evidence. A wide margin makes noninferiority easier to show but tolerates more potential loss. The CONSORT extension therefore asks reports to state the margin and explain its clinical basis.

AcT: a pragmatic Canadian comparison#

The AcT trial randomized adults eligible for intravenous thrombolysis within 4.5 hours across Canadian stroke centers. It used an open-label, registry-linked design with blinded assessment of the main outcome, and the primary endpoint was a modified Rankin Scale score of 0 or 1 at 90 to 120 days, meaning no symptoms or no significant disability.

That outcome occurred in 296 of 802 participants assigned to tenecteplase, 36.9%, and 266 of 765 assigned to alteplase, 34.8%. The unadjusted risk difference was 2.1 percentage points, with a 95% confidence interval from -2.6 to 6.9, and the lower bound stayed above the prespecified -5-point margin, so the trial met its noninferiority criterion.

The result does not prove a 2.1-point benefit. The confidence interval crossed zero, and superiority was not the established conclusion. It says that, within the trial's precision, a clinically unacceptable loss larger than the selected margin was ruled out.

AcT's pragmatic enrollment and registry connection improve relevance to routine systems. Open-label treatment can affect workflow and cointerventions, though blinded outcome assessment reduces bias at the primary endpoint. The Canadian setting, protocols, and case mix still matter for transportability.

TRACE-2: similar question in a different population#

TRACE-2 randomized 1,430 patients in China who were eligible for standard intravenous thrombolysis but were ineligible for or declined endovascular thrombectomy. The primary outcome was again a Rankin score of 0 or 1 at 90 days.

In the modified intention-to-treat analysis, the endpoint occurred in 439 of 705 participants, 62%, after tenecteplase and 405 of 696, 58%, after alteplase. The risk ratio was 1.07, with a 95% confidence interval from 0.98 to 1.16. Its lower bound exceeded the prespecified noninferiority boundary, so the criterion was met.

The percentages are much higher than in AcT. That does not mean care was universally more effective in one country. Stroke severity, age, vessel occlusion, thrombectomy eligibility, enrollment, and outcome ascertainment can change the proportion reaching minimal disability. Randomization supports the within-trial comparison, not a simple cross-trial comparison of baseline outcomes.

Symptomatic intracranial hemorrhage within 36 hours occurred in 2% of each group. Mortality was 7% and 5%, with a wide confidence interval around the relative comparison, and the trial was not powered to prove equality for rare harms, so similar observed percentages do not remove bleeding risk.

ATTEST-2: use the whole disability scale#

A Rankin score threshold turns a seven-level disability scale into success or failure. Someone moving from severe to moderate disability may benefit without crossing the 0-or-1 threshold. ATTEST-2 therefore used an ordinal, or shift, analysis of the full 90-day Rankin distribution.

The common odds ratio was 1.07, with a 95% confidence interval from 0.90 to 1.27. The lower bound exceeded the prespecified odds-ratio noninferiority margin of 0.75. Tenecteplase was noninferior, but it was not superior.

Death occurred in 8% in both groups. Symptomatic intracerebral hemorrhage by the trial's SITS-MOST definition occurred in 2% in each group, while some broader bleeding measures differed numerically, and safety conclusions should use prespecified definitions and confidence intervals, not select whichever definition looks most reassuring or concerning.

An ordinal odds ratio also assumes a broadly common treatment relationship across cut points of the scale. Analysts should check that assumption and examine the actual outcome distribution. One summary measure cannot convey which disability transitions contributed most.

ORIGINAL: replication with another eligibility pattern#

The ORIGINAL trial enrolled patients in China within 4.5 hours and again compared the 0.25 mg/kg tenecteplase strategy with standard alteplase. A Rankin score of 0 or 1 at day 90 occurred in 532 of 732 participants, 72.7%, versus 515 of 733, 70.3%.

The adjusted risk ratio was 1.03, with a 95% confidence interval from 0.97 to 1.09. The lower bound was above the trial's noninferiority threshold of 0.937. Superiority was not established.

Consistent conclusions across several large trials are more persuasive than one study. Yet consistency refers to the noninferiority question, not identical event rates. The trials differed in geography, stroke severity, thrombectomy pathways, imaging, outcome model, margin, and analysis set.

Dose is part of the treatment, not a footnote#

The major supportive comparisons used tenecteplase 0.25 mg/kg, capped at 25 mg. That detail belongs in evidence interpretation because another dose cannot inherit the conclusion automatically.

NOR-TEST 2 part A tested 0.4 mg/kg in people with moderate or severe stroke; the trial stopped early after safety review, and the higher-dose strategy was associated with worse safety and functional results than alteplase. Earlier NOR-TEST participants often had minor stroke or stroke mimics, which limited what could be inferred about that dose in more severe disease.

The lesson is not that tenecteplase as a molecule is unsafe or that dose selection is arbitrary. It is that a noninferiority conclusion attaches to the exact regimen and population tested. Mixing 0.25 and 0.4 mg/kg studies without distinction can produce a misleading summary.

Intention-to-treat and per-protocol both matter#

In a superiority trial, analyzing participants as randomized usually protects against bias and often makes groups look more similar only when there is nonadherence, and in a noninferiority trial, anything that erases a real difference can make noninferiority too easy to claim.

For that reason you should inspect both intention-to-treat and per-protocol results. Agreement is reassuring. Major crossover, incorrect treatment, missing outcomes, or exclusion after randomization can weaken the conclusion even if a primary analysis crosses the margin.

TASTE illustrates the importance of the prespecified population. It enrolled patients selected by perfusion imaging. Its intention-to-treat analysis narrowly failed its noninferiority threshold, while a per-protocol analysis supported noninferiority. Recruitment ended early after other trials reported. The study does not overturn the larger evidence base, but it should not be rewritten as an unqualified positive result.

Convenience is a system outcome#

A bolus may reduce preparation steps, infusion-pump dependence, and transfer delays. Those are plausible advantages, especially when a patient must move rapidly to another center for thrombectomy. They should be measured through door-to-needle time, medication errors, transfer time, and clinical outcomes.

The alternative is not automatically slow. Experienced systems can deliver alteplase efficiently; a protocol change can also introduce new errors if stroke and heart-attack dosing are confused, staff are unfamiliar with vial preparation, or electronic ordering is unclear. Implementation therefore requires standardized ordering, pharmacist and nursing education, weight verification, independent checks, monitoring, and audit. A simpler administration route does not make a thrombolytic a simple medicine.

How guidelines and regulation fit the trials#

The FDA added adult acute ischemic stroke to the tenecteplase indication in 2025. The label specifies acute use, contraindications, bleeding warnings, and monitored administration. Approval is not a statement that every person with stroke symptoms should receive it.

The 2026 AHA/ASA acute ischemic stroke guideline recommends either tenecteplase at the evidence-supported stroke regimen or alteplase for eligible adults presenting within 4.5 hours of symptom onset or last known well. The guideline also addresses imaging-selected extended windows and thrombectomy, where the evidence and eligible populations differ.

Guidelines synthesize more than the four trials, including earlier studies, meta-analyses, harm data, feasibility, and implementation. They support a choice within an emergency protocol. They do not turn a population recommendation into advice you can act on yourself.

A disciplined noninferiority reading sequence#

When you meet a claim that one therapy is “as good as” another, ask:

1. Was the comparator already effective for the same population and outcome?
2. Was the noninferiority margin prespecified and clinically justified?
3. Which scale was used: risk difference, risk ratio, or odds ratio?
4. Did the full confidence interval stay on the acceptable side of the margin?
5. Did intention-to-treat and per-protocol analyses agree?
6. Were safety outcomes sufficiently frequent and precisely measured?
7. Does the claimed convenience translate into a reliable system benefit?

The tenecteplase evidence answers those questions well enough for current guidelines to offer either medicine in the standard window. It does not establish identical effects in every subgroup, a superior functional outcome, or freedom from hemorrhage.

References#

1. 2026 AHA/ASA acute ischemic stroke guideline
2. AcT trial
3. TRACE-2 trial
4. ATTEST-2 trial
5. ORIGINAL trial
6. TASTE trial
7. NOR-TEST 2 part A
8. FDA tenecteplase label
9. CONSORT noninferiority and equivalence extension

For your own health, talk with your clinician.*