Factor Xa Inhibitor-Associated Hemorrhage After Andexanet Alfa's Withdrawal from the US Market
The one agent built specifically to reverse this bleed carried the only randomized data showing it also raises the recipient's own risk of a new clot — in the very organ everyone is trying to save. That finding is why it is no longer sold in the United States, and why tonight's decision is about what is left on the shelf.
W.B., a 79-year-old retired pharmacist, was found by his daughter on the kitchen floor of his apartment, confused and unable to move his left side, roughly an hour after she'd last spoken to him by phone. He takes apixaban for atrial fibrillation, diagnosed six years ago, along with metoprolol and a statin; he has no personal history of stroke, heart attack, or blood clot of any kind, and by his daughter's account has been meticulous about his medications "in the way only a pharmacist would be." His last confirmed apixaban dose was that morning, roughly six hours before he was found.
Noncontrast CT on arrival shows a moderate-sized right basal ganglia hemorrhage, roughly 25mL, without intraventricular extension. His NIHSS is 14 — dense left hemiparesis, left facial droop, mild dysarthria — and his blood pressure on arrival is 176/98. Given the timing of his last dose and his known apixaban use, reversal is being considered to limit further hematoma expansion, the single most modifiable driver of poor outcome in the first hours after an anticoagulant-associated bleed. Through most of the last decade the answer here would have been andexanet alfa, the only agent designed to sequester a factor Xa inhibitor directly rather than flood the system with replacement clotting factors. ANNEXA-I, the randomized trial that finally tested it against usual care — predominantly 4-factor PCC — did find better anti-Xa reduction and superior hemostatic efficacy, and was stopped early for that benefit. The same trial found thrombotic events in 10.3% of the andexanet arm against 5.6% with usual care, including an excess of ischemic strokes: harm in the very organ the drug exists to protect. On the FDA's own reanalysis those rates came out higher still, 14.6% against 6.9%, an advisory committee could not agree the 12-hour hemostatic endpoint justified them, and rather than convert an accelerated approval to a full one the manufacturer withdrew the product from the United States market in December 2025. So the shelf he is being treated from tonight holds essentially the comparator arm of that trial — an agent that has never itself been randomized against outcome in this condition, now standing as the default by subtraction rather than by evidence.
In the ED, choosing a reversal strategy
4-factor PCC, weight-based, and I'd like it started before he goes back to the scanner. I want to be clear about why, because it isn't that I think PCC is the better drug — andexanet alfa beat it on hemostatic efficacy in ANNEXA-I, and that trial stopped early on that result. It's that andexanet hasn't been commercially available in this country since December 2025. The FDA's reanalysis of ANNEXA-I put thrombotic events at 14.6% against 6.9% for usual care, the advisory committee wouldn't accept a 12-hour imaging endpoint as worth that, and the manufacturer pulled the accelerated approval application rather than defend it. It's still sold as Ondexxya in Europe and Japan. It is not on our shelf, and a 25mL bleed that is six hours old does not wait while we relitigate that.
You said it isn't that PCC is the better drug, and I want to hold you to that phrasing rather than let it slide past, because the chart is going to record what we gave and not why. PCC has never been randomized against outcome in factor Xa inhibitor-associated bleeding at all. Its standing here is inherited from vitamin K antagonist reversal, where replacing depleted factors II, VII, IX and X addresses the actual defect. Apixaban hasn't depleted anything — it's occupying factor Xa. We're adding substrate and hoping to outcompete an inhibitor that's still bound.
I'd still give it. There isn't a better option in the building, and ANNEXA-4 at least tells us hemostatic outcomes in this setting aren't uniformly dismal. But it should go in the note as the remaining option rather than the established one, and his daughter should hear it that way too. A default produced by a market withdrawal isn't the same thing as a default produced by evidence, and those two look identical six months later when someone reads this chart.
You're right that the distinction matters for the chart, and I'd sign that note as written. What I'd push back on is how much of this hour the question deserves. We have spent several minutes on the agent with the weakest outcome evidence in the room while his pressure has been sitting at 176/98.
INTERACT3 is the trial that actually moved functional outcome in intracerebral hemorrhage, and it did it with a bundle — early blood pressure control, glucose, temperature, and reversal where applicable — not with a reversal agent alone. ATACH-2 separately tells us not to chase 110-139 for its own sake. So: PCC now because it's what exists, anti-Xa level sent to confirm he actually has meaningful drug on board six hours out, pressure to 140-160, neurosurgery aware, repeat scan in six hours. The reversal agent is the part of this plan I'm least confident in and the part we've discussed most, and I think those two facts are related.
Agreed: weight-based 4-factor PCC started before repeat imaging, an anti-Xa level sent to confirm meaningful residual apixaban exposure at six hours, blood pressure brought to 140-160 rather than to ATACH-2's more aggressive range, and neurosurgery consulted on hematoma stability. Agreed too on how the note reads — PCC recorded as the agent that remains available rather than the agent shown to work.
Not agreed, and left standing rather than smoothed over: whether a center should attempt to obtain andexanet alfa through an international or expanded-access route for a patient like him. The neurointensivist thought the ANNEXA-I safety data made that effort hard to justify even if it were logistically possible; the pharmacologist thought a withdrawal driven by an unconvincing surrogate endpoint is not the same finding as a drug shown to be harmful, and that the two get conflated quickly. Neither position was overruled, and no attempt was made tonight.