ANNEXA-I is a randomized trial of andexanet alfa in a narrow emergency: acute intracerebral hemorrhage in patients who had recently taken a factor Xa inhibitor. It found better control of hematoma expansion than usual care, but more thrombosis, especially ischemic stroke. It did not demonstrate a clear improvement in survival or functional disability at 30 days.
That result is both clinically important and incomplete. Stopping brain-bleed expansion is a plausible route to better outcomes, yet reversal also removes anticoagulant protection in people who often have a strong baseline reason for it. A trial can therefore improve a hemostatic endpoint while leaving the overall benefit-harm balance uncertain.
The regulatory setting also changed after publication. In December 2025, FDA said serious risks outweighed benefits, reported higher adjudicated thrombotic-event and thrombotic-death rates, and stated that US commercial sales would end after December 22, 2025. Carry that current safety communication with you into any historical discussion of the trial or the earlier label.
Why factor Xa inhibitor bleeding is difficult#
Apixaban, rivaroxaban, and edoxaban reduce clot formation by inhibiting activated factor X. They prevent stroke and treat or prevent venous thrombosis, but major bleeding can occur; intracerebral hemorrhage is especially dangerous because a modest increase in hematoma volume can injure tissue within a fixed skull.
Emergency management begins with stabilization, identifying the anticoagulant and last dose, imaging, blood-pressure management, neurosurgical and stroke assessment, and control of other contributors. Reversal is one part of a bundle. Time since the last dose, kidney and liver function, anti-factor Xa testing when rapidly available, bleed severity, and the reason for anticoagulation all matter.
Reversal creates an unavoidable tension. Less anticoagulant activity may help clot formation at the bleeding site. The same change may permit arterial or venous thrombosis. Some patients have atrial fibrillation, recent thrombosis, vascular disease, or immobility, so their untreated thrombotic risk is not trivial.
How andexanet differs from prothrombin complex concentrate#
Andexanet alfa is a modified recombinant factor Xa decoy. It binds certain direct and indirect factor Xa inhibitors, reducing their ability to inhibit native factor Xa, and earlier US accelerated approval covered patients treated with rivaroxaban or apixaban when reversal was needed for life-threatening or uncontrolled bleeding. The approval rested on anti-factor Xa reduction as a surrogate considered reasonably likely to predict benefit, with a postmarketing trial required.
Four-factor prothrombin complex concentrate, commonly called 4F-PCC, supplies vitamin K-dependent clotting factors. It was not designed as a molecular antidote to direct factor Xa inhibitors. In this use, it aims to enhance thrombin generation despite residual inhibitor. Its factor Xa inhibitor use and exact protocols have depended on guidelines, local policy, and evidence that is largely observational.
The strategies therefore cannot be compared only by whether a laboratory anti-factor Xa value falls. One acts as a decoy for inhibitor; the other increases clotting-factor substrate. Relevant outcomes include hematoma growth, rescue treatment, neurologic deterioration, surgery, thrombosis, disability, death, feasibility, and time to treatment.
The ANNEXA-I trial design#
ANNEXA-I was an international, randomized, open-label trial, and it enrolled adults with acute intracerebral hemorrhage who had taken a factor Xa inhibitor within 15 hours and presented within six hours after symptom onset or last known well. Important exclusions limited applicability to patients with very large hematomas, very low consciousness scores, planned surgery soon after enrollment, or other features set by the protocol.
Participants were assigned to andexanet or usual care. The usual-care arm allowed locally selected hemostatic treatment. In the published population, most control patients received prothrombin complex concentrate. This makes the trial highly relevant to PCC-based care, but it is not a pure andexanet-versus-one-standardized-PCC-product comparison.
The trial was open-label because the treatments and logistics were difficult to mask; image assessment and endpoint adjudication included blinded processes, which reduced some bias, but clinical co-interventions could still be influenced by knowledge of assignment.
The planned enrollment was larger, but the trial stopped early after a prespecified interim analysis met the efficacy stopping boundary. Early stopping was consistent with the protocol. It also means estimates can be less stable than results from the full planned sample, especially for safety and patient-centered outcomes.
What “hemostatic efficacy” meant#
The primary endpoint was a composite. A participant had hemostatic efficacy at 12 hours only if hematoma expansion stayed within a prespecified limit, the National Institutes of Health Stroke Scale had not worsened beyond a set amount, and no rescue therapy was given during the defined interval.
In the published report, 67.0% of patients assigned to andexanet met that composite, compared with 53.1% assigned to usual care. The adjusted difference was 13.4 percentage points. Anti-factor Xa activity fell much more with andexanet than with usual care.
The result supports a real biological and imaging effect. Do not let anyone translate it for you into “67% recovered” or “13% avoided disability.” The composite mainly captured short-term hematoma control plus neurologic stability and absence of rescue treatment. It was not the modified Rankin Scale, long-term independence, or survival.
Composite endpoints also need component review. A treatment can win the composite through one component that is more frequent or sensitive. Hematoma expansion was central to the difference. That is important because expansion predicts worse outcome, but prediction does not guarantee that changing it will improve every later outcome.
The thrombotic signal#
In the original publication, thrombotic events were more frequent in the andexanet group than in usual care, with ischemic stroke contributing substantially. The published report gave 10.3% versus 5.6% for thrombotic events and 6.5% versus 1.5% for ischemic stroke.
FDA's later safety review used additional adjudication and reported day-30 thrombosis in 14.6% with andexanet versus 6.9% with usual care. Thrombosis-related death was 2.5% versus 0.9%. Do not silently substitute those later figures into the original paper when you cite it; they reflect the regulator's subsequent assessment.
Possible causes include rapid neutralization of anticoagulation, the procoagulant condition of acute illness, baseline vascular risk, immobilization, and delay in restarting thromboprophylaxis or anticoagulation. The randomized comparison establishes that the treatment strategy increased events in this trial, even though no single mechanism explains every event.
What happened to disability and death#
The published trial did not show an appreciable difference between groups in the modified Rankin Scale at 30 days or in death by 30 days. It was not powered or completed to prove a modest functional or mortality benefit, so “no statistically clear difference” is not proof of exact equivalence.
Still, these outcomes are critical. A smaller hematoma is valuable because it may preserve function, but patients and families ultimately care about survival, cognition, communication, mobility, comfort, and the ability to live according to their goals. A larger and longer trial might have clarified whether the imaging benefit outweighed thrombotic harm. ANNEXA-I alone did not do so.
FDA's later conclusion changes current interpretation#
Andexxa received US accelerated approval in 2018. ANNEXA-I was meant to verify clinical benefit. FDA convened an advisory committee in November 2024 to discuss the results and safety. In December 2025, FDA stated that available serious risks, including increased thromboembolic events, were such that it considered risks to outweigh benefits.
FDA also said the manufacturer requested voluntary withdrawal of the biologics license for commercial reasons and confirmed an end to US sales after December 22, 2025. The communication did not turn ANNEXA-I into a negative trial on its primary endpoint. It shows that regulatory benefit-risk assessment uses more than whether a primary endpoint reached statistical significance.
Availability and regulatory status outside the United States can differ. A historical label, a guideline written before the safety communication, and a current formulary answer may therefore conflict on your desk, because they come from different dates and jurisdictions.
What the trial does not answer#
ANNEXA-I does not directly determine the best approach to gastrointestinal bleeding, traumatic bleeding outside the study criteria, urgent surgery without active bleeding, or hemorrhage linked to dabigatran or warfarin, and it also does not establish that any reversal is needed when little active anticoagulant remains.
The trial cannot cleanly compare andexanet with one uniform PCC regimen because usual care varied, and it cannot resolve how outcomes change with very early treatment, different anti-factor Xa levels, hematoma location, later anticoagulation restart, or current stroke-care bundles in every center.
Observational comparisons can add rare-event and real-world information, but treatment selection creates confounding. Sicker patients may receive one strategy preferentially. Randomization is the stronger design for causal comparison, while registries can show implementation, subgroups, and harms too uncommon for a trial.
A framework for reading reversal evidence#
First ask whether the trial population matches the emergency you are treating. Brain bleeding, gastrointestinal bleeding, trauma, and planned surgery have different mechanisms and endpoints. Next ask what comparator actually received. “Standard care” can conceal several treatments.
Then separate pharmacodynamic success from clinical benefit. Anti-factor Xa reduction shows target activity. Hematoma control is closer to clinical benefit. Disability, survival, thrombosis, and quality of life complete the balance.
Finally align the date. Product labeling, guidelines, trial publications, advisory discussions, safety communications, and commercial availability change on different schedules. Your decision today needs the current local protocol and regulatory setting, not a frozen summary from 2024.
Sources#
- ANNEXA-I randomized trial publication
- FDA December 2025 safety communication on Andexxa
- FDA Andexxa product and regulatory documents
- FDA November 2024 advisory committee materials
- AHA-ASA 2022 spontaneous intracerebral hemorrhage guideline update
- ACC-AHA-ACCP-HRS 2023 atrial fibrillation guideline
Suspected intracerebral hemorrhage or uncontrolled bleeding is an emergency.*
Questions and answers
What was the main result of ANNEXA-I?
Andexanet produced better hemostatic efficacy at 12 hours than usual care, driven largely by better hematoma control, and reduced anti-factor Xa activity more strongly. It also increased thrombotic events, including ischemic stroke.
Did ANNEXA-I prove that andexanet improves survival or disability?
No. The published trial did not show an appreciable 30-day difference in death or modified Rankin disability. The trial stopped early for its primary hemostatic endpoint and was not a definitive test of a modest patient-centered benefit.
Was ANNEXA-I a comparison only against four-factor PCC?
No. The comparator was usual care. Most control patients received prothrombin complex concentrate, making that strategy central to the comparison, but treatment was not one uniform PCC protocol for every participant.
Is andexanet currently sold in the United States?
FDA reported that US commercial sales would end after December 22, 2025 and that the manufacturer had requested voluntary withdrawal of the biologics license for commercial reasons. Current institutional and jurisdictional information must be checked separately.
Does this trial settle every factor Xa inhibitor bleeding decision?
No. It answers a specific question in selected acute intracerebral hemorrhage. Other bleed sites, different timing, little residual anticoagulant activity, urgent procedures, and patients outside the trial criteria require other evidence and emergency protocols.