The guidance in full
Reproduced from the primary documents so you can check the tool against its sources rather than trusting it.
Anticoagulant reversal in major trauma (NICE NG39 1.5.6–1.5.12)
- Rapidly reverse anticoagulation in patients who have major trauma with haemorrhage.
- Hospital trusts that admit patients with major trauma should have a protocol for the rapid identification of patients who are taking anticoagulants and the reversal of anticoagulation agents.
- Use prothrombin complex concentrate immediately in adults (16 or over) with major trauma who have active bleeding and need emergency reversal of a vitamin K antagonist.
- Do not use plasma to reverse a vitamin K antagonist in patients with major trauma.
- Consult a haematologist immediately for advice on adults (16 or over) who have active bleeding and need reversal of any anticoagulant agent other than a vitamin K antagonist.
- Consult a haematologist immediately for advice on children (under 16s) with major trauma who have active bleeding and may need reversal of any anticoagulant agent.
- Do not reverse anticoagulation in patients who do not have active or suspected bleeding.
NG39 also directs the reader to the MHRA safety advice on DOACs for a list of reversal agents.
Tranexamic acid and blood components in major trauma (NG39 1.5.4–1.5.5, 1.5.18–1.5.27)
- Use intravenous tranexamic acid as soon as possible in patients with major trauma and active or suspected active bleeding.
- Do not use intravenous tranexamic acid more than 3 hours after injury in patients with major trauma unless there is evidence of hyperfibrinolysis.
- For patients with active bleeding use a restrictive approach to volume resuscitation until definitive early control of bleeding has been achieved.
- In hospital settings do not use crystalloids for patients with active bleeding.
- For adults (16 or over) use a ratio of 1 unit of plasma to 1 unit of red blood cells to replace fluid volume. For children (under 16s) use a ratio of 1 part plasma to 1 part red blood cells, and base the volume on the child’s weight.
- For patients with active bleeding, start with a fixed-ratio protocol for blood components and change to a protocol guided by laboratory coagulation results at the earliest opportunity.
NICE noted in February 2016 that this was an off-label use of tranexamic acid.
Prothrombin complex concentrate (NICE NG24 1.12.3–1.12.6)
- Offer immediate prothrombin complex concentrate transfusions for the emergency reversal of warfarin anticoagulation in people with either: severe bleeding, or head injury with suspected intracerebral haemorrhage.
- Consider immediate prothrombin complex concentrate transfusions to reverse warfarin anticoagulation in people having emergency surgery, depending on the level of anticoagulation and the bleeding risk.
- Monitor the international normalised ratio (INR) to confirm that warfarin anticoagulation has been adequately reversed, and consider further prothrombin complex concentrate.
NG24 cross-refers to NICE’s stroke guideline for reversal in primary intracerebral haemorrhage, and to the MHRA safety advice on DOACs for reversing direct-acting oral anticoagulants.
Warfarin: the BSH thresholds, bleeding and not bleeding
- All hospitals managing patients on warfarin should stock a licensed four-factor PCC (1C).
- Emergency anticoagulation reversal in major bleeding should be with 25–50 U/kg four-factor PCC and 5 mg intravenous vitamin K (1B).
- Recombinant factor VIIa is not recommended for emergency anticoagulation reversal (1B).
- Fresh frozen plasma produces suboptimal anticoagulation reversal and should only be used if PCC is not available (1C).
- Anticoagulation reversal for non-major bleeding should be with 1–3 mg intravenous vitamin K (1B).
- Patients with an INR > 5.0 but who are not bleeding should have 1–2 doses of warfarin withheld and their maintenance dose reduced (1B). The cause of the elevated INR should be investigated (1C).
- Asymptomatic patients with an INR of ≥ 8.0 should receive 1–5 mg of oral vitamin K (1B). The INR should be rechecked the following day in case an additional dose of vitamin K is required.
- For surgery that requires reversal of warfarin and that can be delayed 6–12 h, the INR can be corrected by giving intravenous vitamin K. For surgery that cannot be delayed, the INR can be corrected by giving PCC and intravenous vitamin K. PCC should not be used to enable elective or non-urgent surgery (2C).
For acenocoumarol, phenindione and phenprocoumon the principles are the same, but further vitamin K should be considered for the agents with longer half-lives.
PCC dosing by INR — the licensed tables
Both UK four-factor PCCs use the same dose bands, expressed as IU of factor IX per kg:
| Pre-treatment INR | Dose (IU factor IX/kg) | Volume | Max single dose |
|---|---|---|---|
| 2.0 – 3.9 | 25 | 1 mL/kg | 2500 IU |
| 4.0 – 6.0 | 35 | 1.4 mL/kg | 3500 IU |
| > 6.0 | 50 | 2 mL/kg | 5000 IU |
Beriplex P/N: “The correction of the vitamin K antagonist-induced impairment of haemostasis is commonly reached approximately 30 minutes after the injection.” And: “Repeated dosing with Beriplex for patients requiring urgent reversal of vitamin K antagonist treatment is not supported by clinical data and therefore not recommended.”
Octaplex: correction persists approximately 6–8 hours, while co-administered vitamin K takes effect within 4–6 hours, “thus, repeated treatment with human prothrombin complex is not usually required when vitamin K has been administered”. Octaplex is given at 0.12 mL/kg/min (~3 units/kg/min), to a maximum of 8 mL/min.
The maximum single doses in both SmPCs are framed around patients over 100 kg. Note the mismatch worth knowing about: the SmPC tables start at INR 2.0 and cap the dose, whereas BSH quotes a flat 25–50 U/kg range — many trust protocols use a fixed 30 U/kg or a fixed 500/1000 IU-vial regimen instead. Use your trust’s protocol and the SmPC for the product your trust actually stocks.
Acute upper GI bleeding (NICE CG141 1.2.1–1.2.8)
- Transfuse patients with massive bleeding with blood, platelets and clotting factors in line with local protocols for managing massive bleeding.
- Base decisions on blood transfusion on the full clinical picture, recognising that over-transfusion may be as damaging as under-transfusion.
- Do not offer platelet transfusion to patients who are not actively bleeding and are haemodynamically stable.
- Offer platelet transfusion to patients who are actively bleeding and have a platelet count of less than 50 × 109/litre. Offer fresh frozen plasma to patients who are actively bleeding and have a prothrombin time (or INR) or APTT greater than 1.5 times normal. If fibrinogen remains less than 1.5 g/litre despite FFP, offer cryoprecipitate as well.
- Offer prothrombin complex concentrate to patients who are taking warfarin and actively bleeding.
- Treat patients who are taking warfarin and whose upper gastrointestinal bleeding has stopped in line with local warfarin protocols.
- Do not use recombinant factor VIIa except when all other methods have failed.
Also: Blatchford score at first assessment and full Rockall after endoscopy; consider early discharge if the pre-endoscopy Blatchford score is 0; endoscopy immediately after resuscitation in unstable patients with severe bleeding, and within 24 hours for everyone else. Do not offer acid suppression before endoscopy in suspected non-variceal bleeding; offer PPIs after endoscopy where there are stigmata of recent haemorrhage. For suspected variceal bleeding, offer terlipressin and prophylactic antibiotics at presentation. Continue low-dose aspirin for secondary prevention once haemostasis is achieved.
Andexanet alfa — what NICE recommends, and what it does not
“Andexanet alfa is recommended as an option for reversing anticoagulation from apixaban or rivaroxaban in adults with life-threatening or uncontrolled bleeding, only if: the bleed is in the gastrointestinal tract, and the company provides andexanet alfa according to the commercial arrangement.”
The committee’s reasoning is worth carrying in your head: there is no trial directly comparing andexanet alfa with PCC. An indirect comparison suggested improved survival in gastrointestinal bleeding or ICH, but lower survival for bleeds elsewhere, and the comparison is uncertain. The cost-effectiveness estimates were within NICE’s acceptable range for gastrointestinal bleeding but not for ICH or other sites — hence a recommendation that stops at the GI tract.
For intracranial haemorrhage there is no NICE recommendation to rely on: TA697 records that this part has been “updated and replaced by NICE’s technology appraisal guidance on andexanet alfa for reversing anticoagulation in people with intracranial haemorrhage (terminated appraisal)”.
Andexanet alfa is for apixaban and rivaroxaban. It is not the answer for edoxaban, dabigatran, or a vitamin K antagonist. The licensed indication is “adult patients”.
Dosing (Ondexxya SmPC 4.2)
| Regimen | Bolus | Infusion | Vials |
|---|---|---|---|
| Low | 400 mg at ~30 mg/min over 15 min | 4 mg/min for 120 min (480 mg) | 5 |
| High | 800 mg at ~30 mg/min over 30 min | 8 mg/min for 120 min (960 mg) | 9 |
Choose by the drug, its dose and the time since the last dose: apixaban 5 mg or less, or rivaroxaban 10 mg or less — low dose regardless of timing. Above those doses — high dose if the last dose was less than 8 hours ago, low dose at 8 hours or more.
Anticoagulant activity from a factor Xa inhibitor or LMWH is anticipated to return within 4 hours of the end of the infusion; antithrombotic therapy can be re-initiated as soon as medically indicated once the patient is stable and haemostasis is adequate.
Idarucizumab (Praxbind) — dose and restarting
“The recommended dose is 5 g idarucizumab (2 vials of 2.5 g/50 mL)”, “administered intravenously as two consecutive infusions over 5 to 10 minutes each or as a bolus injection.”
A second 5 g dose may be considered in patients with recurrent bleeding and prolonged clotting times, those at risk of life-threatening re-bleeding with prolonged clotting markers, or those needing a second emergency procedure with prolonged coagulation parameters.
“Dabigatran etexilate treatment can be re-initiated 24 hours after administration of idarucizumab, if the patient is clinically stable and adequate haemostasis has been achieved.” Other antithrombotic therapy such as LMWH can be started at any time once the patient is stable and haemostasis is adequate.
Praxbind is restricted to hospital use.
Heparin and LMWH — protamine, from the BSH guideline
Unfractionated heparin has a plasma half-life of 45–90 minutes at therapeutic intravenous doses, so stopping it plus general measures is often enough. Where it is not, protamine sulphate forms a stable, inactive salt with heparin.
Unfractionated heparin
Dose from the quantity of UFH given in the preceding 2 hours, assuming 1 mg protamine neutralises 80–100 units of UFH. BSH’s own worked examples: bleeding during an infusion of 1250 units/h requires 25 mg protamine; bleeding soon after a 5000-unit bolus requires 50 mg. Monitor the reversal effect with the APTT.
Low molecular weight heparin
Protamine reverses approximately 60% of LMWH. The BSH recommendations:
- LMWH given within 8 hours of the time reversal is needed: protamine sulphate 1 mg per 100 anti-Xa units of LMWH. If ineffective, consider a further 0.5 mg per 100 anti-Xa units (2C).
- LMWH given more than 8 hours before: consider smaller doses of protamine (2C).
- Consider rFVIIa only if life-threatening bleeding continues despite protamine and the timing suggests residual LMWH effect (2C).
Give protamine slower than 5 mg/min to minimise adverse reactions. Its half-life is about 7 minutes — shorter than UFH — so prolonged administration may be needed after subcutaneous heparin, where entry into the circulation is delayed. Severe allergic reactions including anaphylaxis, hypotension, bronchospasm and skin reactions occur in up to 10% of patients.
The APTT may be prolonged by LMWH but should not be used to assess the extent of its effect — that is the anti-Xa test.
Tranexamic acid in head injury (NICE NG232 1.3.17–1.3.18)
“For people with a head injury and a GCS score of 12 or less who are not thought to have active extracranial bleeding, consider:
- a 2 g intravenous bolus injection of tranexamic acid for people 16 and over
- a 15 mg/kg to 30 mg/kg (up to a maximum of 2 g) intravenous bolus injection of tranexamic acid for people under 16.
Give the tranexamic acid as soon as possible within 2 hours of the injury, in the pre-hospital or hospital setting and before imaging.” NICE noted in March 2023 that these were off-label uses.
Where there is suspected or confirmed extracranial bleeding as well, NG232 hands over to the major trauma guideline instead — which is the 3-hour window, not the 2-hour one.
Two trials that should change your reflexes
HALT-IT — tranexamic acid does not belong in GI bleeding
12,009 patients with significant upper or lower gastrointestinal bleeding, randomised to high-dose tranexamic acid (1 g loading dose then 3 g over 24 h) or placebo. Death due to bleeding within 5 days: 222/5956 (4%) with tranexamic acid versus 226/5981 (4%) with placebo, RR 0.99 (95% CI 0.82–1.18). Venous thromboembolic events were higher with tranexamic acid: 48/5952 (0.8%) versus 26/5977 (0.4%), RR 1.85 (95% CI 1.15–2.98). The authors’ conclusion: “tranexamic acid should not be used for the treatment of gastrointestinal bleeding outside the context of a randomised trial.”
This is the single most transferable error in this tool’s subject area: the trauma reflex of early tranexamic acid is not just unhelpful in GI bleeding, it carries a thrombotic cost.
PATCH — platelets are not the answer for antiplatelet-associated ICH
190 participants with spontaneous intracerebral haemorrhage while on antiplatelet therapy, randomised to platelet transfusion or standard care. The odds of death or dependence at 3 months were higher with platelet transfusion (adjusted common odds ratio 2.05, 95% CI 1.18–3.56; p=0.0114). The authors: “Platelet transfusion cannot be recommended for this indication in clinical practice.”
PATCH studied spontaneous ICH, not traumatic ICH or surgical bleeding, so it does not settle every case — but it removes any assumption that platelets are the obvious antidote to an antiplatelet.
General measures that apply to every case
From the BSH guideline’s table of non-pharmacological measures:
- Stop the antithrombotic drug.
- Document the timing and amount of the last drug dose, and any pre-existing renal or hepatic impairment.
- Estimate the half-life and the length of the functional defect the drug has caused.
- Assess the source of bleeding.
- Request full blood count, prothrombin time, APTT, thrombin time, fibrinogen concentration, creatinine concentration — and, if available, a specific assay for the drug’s effect.
- Correct haemodynamic compromise with intravenous fluids and red cell transfusion.
- Apply mechanical pressure if possible; use endoscopic, radiological or surgical measures.
Renal function matters more than it feels like it should: dabigatran is about 80% renally eliminated, with a half-life of roughly 13 hours normally and 22–35 hours when creatinine clearance is under 30 mL/min. In renal impairment, “wait for it to wear off” is a much longer plan than you think.
Related
CT head & C-spine rules · Drug monographs · Clinical scores · ED procedures · RSI planner
The decision tool tells you what to do. This chapter is about why — why the bleed decides the reversal and the number almost never does, why the trauma reflex to reach for tranexamic acid is the wrong one in a gastrointestinal bleed, why an alarming INR after a fall is usually not something to reverse at all, and why “reverse the DOAC” is a sentence that often has no correct ending. It is meant to be read before the shift, not during it.
A 78-year-old is brought in after a fall down four stairs. She takes apixaban for atrial fibrillation. She hit her head; GCS is 15, she is on no other drugs, observations are normal, and there is no external bleeding. A junior colleague asks the charge nurse to “get the andexanet ready” and sends off an urgent INR.
An hour later a different patient arrives: a 64-year-old on warfarin vomiting bright-red blood, INR 5.8, heart rate 118, systolic 92. The registrar reaches for the trauma major-haemorrhage pack and asks for tranexamic acid to be drawn up with the first unit of blood.
Two anticoagulated patients, two instincts, and both instincts are wrong. What should actually happen to each — and why does the same word, “reverse”, mean something different in each room?
The shape of it
Why, in one table
| The question | What decides it | Source |
|---|---|---|
| Reverse at all? | Active or suspected bleeding — not the number. No bleed, no reversal. | NG39 1.5.12 |
| Warfarin, major bleed | Four-factor PCC (25–50 U/kg by INR) and 5 mg IV vitamin K. Not plasma. | NG39 1.5.8–1.5.9 BSH |
| Warfarin, non-major bleed | 1–3 mg IV vitamin K. Not PCC. | BSH |
| Dabigatran, major bleed | Idarucizumab 5 g. A specific antidote exists. | SmPC |
| Apixaban / rivaroxaban, major bleed | Andexanet alfa — but NICE recommends it only for a GI bleed. | TA697 |
| Edoxaban, major bleed | No appraised antidote. Haematology; PCC off-label. | NG39 1.5.10 |
| Tranexamic acid | Yes in trauma and head injury, within the window. No in GI bleeding. | CRASH-2 / HALT-IT |
| Antiplatelet, bleeding | No antidote. Platelets do not reverse it and can harm. | PATCH |
| Raised INR, no bleeding | Withhold ± oral vitamin K by the BSH thresholds. This is not reversal. | BSH |
The bleed decides, not the number
The instinct that a high INR must be corrected is a laboratory instinct, not a clinical one. NICE is unusually blunt about it: NG39 1.5.12 reads, in full, “Do not reverse anticoagulation in patients who do not have active or suspected bleeding.” The whole logic of this module hangs off that sentence. Reversal is a treatment for bleeding, not a treatment for a number.
This matters because reversal is not free. Every patient in front of you is anticoagulated for a reason — a mechanical valve, atrial fibrillation with a high stroke risk, a recent pulmonary embolism — and that reason does not pause because they attended the ED. Reversing them exposes them to the thrombotic event the drug was preventing. Prothrombin complex concentrate in particular carries a thrombotic signal. So the decision to reverse is a decision to accept a thrombotic risk in exchange for controlling a haemorrhage, and if there is no haemorrhage, you are buying the risk and getting nothing for it.
The word that does the work in NG39 is “suspected”. A head injury in which intracranial haemorrhage has not yet been excluded is not “no bleeding” — it is suspected bleeding, and the moment suspicion is real, the calculus changes. NG24 1.12.3 tells you to offer immediate PCC for warfarin reversal in “head injury with suspected intracerebral haemorrhage”. The two recommendations are not in tension: you do not reverse a well patient with a raised INR and a trivial head injury, and you do reverse the patient in whom a bleed is genuinely suspected. What the anticoagulant should change, in the well patient, is your threshold for imaging and observation — not your reflex to reach for a reversal agent.
Anticoagulation is reversed for bleeding, suspected or actual. A frightening number in a patient who is not bleeding is a reason to think, not a reason to give PCC.
Warfarin: PCC not plasma, and vitamin K as well, not instead
For a major bleed on a vitamin K antagonist the answer is settled and consistent across every UK source: four-factor prothrombin complex concentrate plus intravenous vitamin K. NG39 1.5.8 says to use PCC immediately in adults with major trauma who are actively bleeding and need emergency VKA reversal. NG39 1.5.9 adds the negative: do not use plasma to reverse a VKA in major trauma. CG141 1.2.5 says the same for actively bleeding GI patients on warfarin, and NG24 1.12.3 covers head injury with suspected intracerebral haemorrhage. The BSH guideline gives the dose: 25–50 U/kg four-factor PCC with 5 mg IV vitamin K.
The reason PCC beats plasma is factor concentration. A VKA depletes the vitamin K-dependent factors (II, VII, IX, X). PCC replaces them in a small, fast, concentrated volume; plasma replaces them dilutely, slowly, and in a volume an already-bleeding, often elderly patient tolerates badly. The Sarode 2013 trial (Evidence tab) showed PCC corrected the INR to ≤1.3 within half an hour in 62% of patients versus 10% with plasma — a fivefold difference in the thing you actually want.
The part that is most often dropped is the vitamin K, and dropping it is a real error. PCC and vitamin K work on different clocks. The factors in PCC are consumed and cleared over hours; the Beriplex SmPC notes correction “commonly reached approximately 30 minutes after the injection” but does not support repeat dosing. Vitamin K takes 4–6 hours to act but then keeps the INR down by letting the liver make its own factors again. Give PCC without vitamin K and the INR creeps back up as the concentrate wears off. Give them together — PCC for the next few hours, vitamin K for the rest of the day — and the correction holds. This is why the recommendation is always “PCC and vitamin K”, never one or the other, and why the tool says to recheck the INR afterwards.
The INR-banded PCC dose (25 U/kg for INR 2–3.9, 35 for 4–6, 50 for >6) is a licensed table, but note the honest caveat the reference panel carries: many trusts use a flat weight-based or fixed-vial regimen instead. Use your trust’s protocol and the SmPC for the product your trust stocks — and never delay treating a life-threatening bleed while you wait for the INR to come back.
The DOACs: “reverse it” is a sentence that sometimes has no ending
The direct oral anticoagulants do not behave as a class when it comes to reversal, and the commonest mistake is to treat them as if they do.
Dabigatran is the easy one. It is a direct thrombin inhibitor with a specific antidote, idarucizumab (Praxbind) 5 g — two vials of 2.5 g, as two consecutive short infusions or a bolus. In the RE-VERSE AD cohort (Evidence tab) the median maximum reversal was 100%. Dabigatran is also about 80% renally cleared, so in renal impairment its half-life stretches from roughly 13 hours to 22–35 hours — “wait for it to wear off” is a much longer plan than it feels like, and a normal thrombin time and a normal APTT together are what make a significant drug level unlikely, not a normal thrombin time alone.
Apixaban and rivaroxaban are where the trap lies. There is a specific agent — andexanet alfa — but NICE (TA697) recommends it in one setting only: a gastrointestinal bleed, and only under the commercial arrangement. Not intracranial, not “other sites”. The committee’s reasoning is worth carrying: there is no trial comparing andexanet with PCC; an indirect comparison suggested better survival for GI and ICH bleeds but lower survival for bleeds elsewhere, and the cost-effectiveness only fell within range for the GI indication. For intracranial haemorrhage the separate appraisal was terminated — so there is no NICE recommendation to lean on there at all. Most UK major-haemorrhage protocols use four-factor PCC off-label for a life-threatening factor Xa inhibitor bleed outside the GI tract, on haematology advice — a discussion, not a protocol step.
Edoxaban is the dead end. It is a factor Xa inhibitor, but it sits outside TA697 entirely — andexanet is licensed and appraised for apixaban and rivaroxaban, not edoxaban. So for a major edoxaban bleed there is no appraised antidote at all: stop the drug, support, control the source, and call haematology.
Dabigatran → idarucizumab. Apixaban/rivaroxaban → andexanet, but only for GI. Edoxaban → no antidote. For any non-VKA agent in a bleeding trauma patient, NG39 1.5.10 requires immediate haematologist advice.
The raised INR that is not a bleed at all
A large share of “the INR is 7, should I reverse?” calls are about a patient who is not bleeding. Here the BSH over-anticoagulation thresholds apply, and none of them involves PCC:
- INR > 5.0, not bleeding: withhold 1–2 doses of warfarin, reduce the maintenance dose, and investigate why it rose. No vitamin K at this level.
- INR ≥ 8.0, not bleeding: 1–5 mg of oral vitamin K, and recheck the INR the next day. Oral, not IV — IV vitamin K is for bleeding or urgent surgical reversal.
The clinical work here is not the dose; it is the “why”. A raised INR with no dose change almost always has a cause — a new interacting drug (an antibiotic is the classic), intercurrent illness, alcohol, dietary change, or a dosing error. Find it, or it will happen again. And the commonest harm in this scenario is not under-treatment; it is an unowned follow-up: nobody arranges the recheck, nobody restarts the normal dose, and the patient drifts. Say explicitly who owns the recheck.
The exception that catches people is the anticoagulated head injury that has not yet been imaged. That is not this chapter’s patient. A raised INR plus a head injury with any suspicion of intracranial bleeding is “suspected bleeding” — image promptly (NG232), and reverse the moment a bleed is suspected on clinical or CT grounds.
Tranexamic acid: where it saves lives, where it takes them
Tranexamic acid is an antifibrinolytic, not a reversal agent — it does nothing to an anticoagulant — but it belongs in this module because the reflexes around it are so strongly learned in one setting and so wrong in another.
In trauma it works, and timing is everything. CRASH-2 (Evidence tab) showed a reduction in all-cause mortality, and its pre-specified time analysis is the reason NG39 draws a hard line at three hours: given within an hour of injury it cut death due to bleeding (RR 0.68), between one and three hours it still helped (RR 0.79), but given after three hours it increased death due to bleeding (RR 1.44). NG39 1.5.4–1.5.5 follows exactly this: give it as soon as possible in trauma with active or suspected bleeding, and do not give it beyond three hours unless there is evidence of hyperfibrinolysis. The dose NICE rests on is the CRASH-2 regimen — 1 g over 10 minutes then 1 g over 8 hours — though NG39 itself states no dose and most networks specify it locally.
In head injury there is a separate, tighter window. NG232 1.3.17 offers a 2 g IV bolus (15–30 mg/kg, max 2 g under 16) for GCS ≤12 with no active extracranial bleeding, within 2 hours of injury, before imaging. If there is also extracranial bleeding, NG232 hands over to the major trauma guideline — the 3-hour window, not the 2-hour one.
In gastrointestinal bleeding it does not belong, full stop. HALT-IT (Evidence tab) randomised over 12,000 GI-bleed patients and found no reduction in death from bleeding and more venous thromboembolism with tranexamic acid. If your trauma instinct just reached for the TXA as the blood arrives, this is the exact case where it is wrong — and it is the module’s headline warning for a reason.
Antiplatelets: no antidote, and platelets are not one
There is no reversal agent for aspirin, clopidogrel, ticagrelor or prasugrel. The intuitive move — transfuse platelets to “replace” the ones the drug disabled — is not supported and, in the one setting with a randomised trial, is harmful. PATCH (Evidence tab) randomised people with antiplatelet-associated spontaneous intracerebral haemorrhage to platelet transfusion or standard care and found the odds of death or dependence were higher with transfusion (adjusted common OR 2.05).
PATCH studied spontaneous ICH, not traumatic ICH or surgical bleeding, so it does not settle every case — but it removes the assumption that platelets are the obvious antidote. Transfuse platelets for genuine thrombocytopenia or as part of a major-haemorrhage protocol, not to “reverse” an antiplatelet. The real work is mechanical, endoscopic, radiological or surgical control, with pro-haemostatic adjuncts on haematology advice rather than reflexively.
The GI exception runs the other way: CG141 says to continue low-dose aspirin for secondary prevention once haemostasis is achieved, and to make continuation of clopidogrel a specialist-and-patient conversation — because the thrombotic risk of stopping is often the bigger danger. Stopping the antiplatelet is not automatically the safe choice.
Heparin and LMWH: protamine, and why it only half-works
Unfractionated heparin has a short plasma half-life (45–90 minutes at therapeutic IV doses), so stopping the infusion plus general measures is often enough on its own. Where it is not, protamine forms an inactive salt with heparin. The BSH dose is worked from the amount given in the preceding two hours (1 mg protamine per 80–100 units of UFH): bleeding during a 1250 units/h infusion needs about 25 mg, bleeding just after a 5000-unit bolus about 50 mg. Monitor with the APTT.
Low molecular weight heparin is the harder case, because protamine reverses only about 60% of it. Within 8 hours of the dose, give 1 mg protamine per 100 anti-Xa units (a further 0.5 mg/100 units if ineffective); beyond 8 hours, smaller doses. And the test trap: the APTT does not measure LMWH effect — that is the anti-Xa assay. Protamine itself is not benign: give it slower than 5 mg/min, and know that severe reactions including anaphylaxis occur in up to 10%. Its half-life (~7 minutes) is shorter than heparin’s, so after subcutaneous heparin — which keeps entering the circulation — repeated dosing may be needed.
The decision everyone forgets: restarting
Reversal feels like the end of the episode. It is not. The patient had a mechanical valve, or atrial fibrillation with a real stroke risk, or a recent clot — and that indication is still there the morning after. The restart decision is the one most often left unwritten, and an anticoagulant that is reversed and never restarted is its own source of harm.
You will rarely restart in the ED, and you should not have to. What the ED can do is make the decision owned: document the indication for anticoagulation, the drug and time of the last dose, the renal and hepatic function, and a clear statement of who is responsible for deciding when it restarts. General measures from the BSH table apply to every case regardless of the agent — stop the drug, send FBC/PT/APTT/thrombin time/fibrinogen/creatinine and a drug-specific assay if the lab offers one, control the source mechanically, and correct haemodynamic compromise with blood, not crystalloid.
Name, in the notes, who decides when the anticoagulant restarts. The commonest late harm here is not under-reversal — it is an unowned follow-up.
Ten pitfalls
01 Reversing a number, not a bleed
- The trap
- A high INR triggers PCC in a patient who is not bleeding.
- Why
- NG39 1.5.12 says do not reverse anticoagulation without active or suspected bleeding. Reversal buys thrombotic risk for no haemostatic gain.
- Instead
- Manage over-anticoagulation by the BSH thresholds; reserve reversal for a bleed.
02 Giving PCC without the vitamin K
- The trap
- PCC alone for a warfarin bleed; the INR creeps back up hours later.
- Why
- PCC corrects for hours; vitamin K keeps it corrected. They work on different clocks.
- Instead
- Give 5 mg IV vitamin K with the PCC, and recheck the INR.
03 Reaching for plasma to reverse warfarin
- The trap
- FFP used as the VKA reversal agent.
- Why
- NG39 1.5.9 says do not use plasma to reverse a VKA in major trauma; BSH calls FFP suboptimal, for use only if PCC is unavailable.
- Instead
- PCC. FFP in a bleed is coagulopathy replacement (CG141, PT/APTT >1.5×), not VKA reversal.
04 Tranexamic acid for a GI bleed
- The trap
- The trauma reflex fires as the blood arrives for a haematemesis.
- Why
- HALT-IT: no mortality benefit, more venous thromboembolism.
- Instead
- Do not give it. Endoscopy, PPI after scope, terlipressin + antibiotics if variceal.
05 Transfusing platelets to “reverse” an antiplatelet
- The trap
- Platelets given to counteract clopidogrel or aspirin in an ICH.
- Why
- PATCH found higher death or dependence with platelet transfusion in antiplatelet-associated spontaneous ICH.
- Instead
- Platelets only for thrombocytopenia or MHP; control the source; discuss with haematology and neurosurgery.
06 Assuming andexanet covers every factor Xa inhibitor bleed
- The trap
- “Andexanet for the DOAC bleed” — regardless of drug or site.
- Why
- TA697 recommends it only for apixaban/rivaroxaban GI bleeds; the ICH appraisal was terminated; edoxaban is outside it entirely.
- Instead
- Match agent to drug and site; for non-GI or edoxaban, haematology and off-label PCC.
07 Treating a suspected intracranial bleed as “no bleeding”
- The trap
- An anticoagulated head injury with a raised INR filed under “not bleeding, don’t reverse”.
- Why
- NG39’s trigger is active or suspected bleeding; a not-yet-excluded ICH is suspected bleeding.
- Instead
- Image promptly (NG232); reverse the moment a bleed is suspected clinically or on CT.
08 Using the APTT to gauge LMWH effect
- The trap
- A normal APTT read as “the LMWH has worn off”.
- Why
- The APTT does not reliably reflect LMWH; the anti-Xa assay does.
- Instead
- Use anti-Xa timing and levels; dose protamine by anti-Xa units, knowing it reverses ~60%.
09 Scaling an adult regimen onto a child
- The trap
- An adult PCC band or the 5 mg vitamin K dose applied to a bleeding child by weight.
- Why
- Every dose here is adult-derived; NG39 1.5.11 requires immediate haematologist advice for any anticoagulant in a bleeding child.
- Instead
- Call haematology first; take doses from the paediatric formulary and network protocol.
10 Reversing and never restarting
- The trap
- The bleed is controlled, the anticoagulant is stopped, and nobody owns the restart.
- Why
- The indication (valve, AF, VTE) survives the bleed; an un-restarted anticoagulant causes late harm.
- Instead
- Document the indication and name who decides when it restarts.
Seven red herrings
Not if there is no bleeding. INR ≥8 without bleeding is oral vitamin K and a recheck, not PCC.
Only apixaban/rivaroxaban, only a GI bleed, per TA697. Wrong drug or wrong site and andexanet is not the answer.
A normal thrombin time and APTT together make a high level unlikely; the thrombin time alone does not.
The APTT does not measure LMWH. Use the anti-Xa assay and the timing of the last dose.
PATCH says platelets worsen outcome in antiplatelet ICH. Platelets are for thrombocytopenia or MHP, not reversal.
In a GI bleed CG141 says continue low-dose aspirin once haemostasis is achieved; the thrombotic risk of stopping is real.
In GI bleeding TXA raised VTE with no benefit (HALT-IT); in trauma beyond 3 hours it increased death from bleeding (CRASH-2). Timing and setting decide.
The case, revisited
The 78-year-old on apixaban, well, minor head injury. She is not bleeding — she has a suspected-bleeding question, which is answered by imaging, not by andexanet. The right moves: no andexanet drawn up, no urgent INR chase (the INR does not measure apixaban), image the head promptly because she is anticoagulated (NG232), and observe. If the CT shows an intracranial bleed, then she has suspected/confirmed bleeding — and even then andexanet is not recommended for ICH (TA697’s ICH appraisal was terminated); that is a haematology conversation. The junior’s instinct treated the drug as the emergency; the head injury is the emergency, and it is still just a question until the scan answers it.
The 64-year-old on warfarin, haematemesis, INR 5.8, shocked. This is a major bleed on a VKA: four-factor PCC (INR band 4–6, 35 U/kg) plus 5 mg IV vitamin K, activate the major-haemorrhage protocol on physiology, and urgent endoscopy after resuscitation (CG141). The one thing not to do is the thing the registrar reached for: no tranexamic acid — HALT-IT showed it does not help GI bleeding and raises VTE. Continue low-dose aspirin only if she is also on it and once haemostasis is achieved. Same word, “reverse”, two rooms — and in each, the instinct that felt most like action was the one to resist.
How much of anticoagulation reversal rests on randomised evidence? More than most of trauma — but unevenly. The VKA and tranexamic-acid decisions have real trials behind them; the DOAC antidotes rest on single-arm cohorts with no comparator, which is exactly why NICE hedges on andexanet. What follows is the evidence behind the main decisions, with every figure taken from the published abstract.
Sarode 2013 — four-factor PCC vs plasma (RCT)
Phase IIIb, multicentre, open-label non-inferiority trial in non-surgical patients needing urgent VKA reversal for major bleeding; 202 in the intention-to-treat efficacy population (98 PCC, 104 plasma), median baseline INR 3.9 vs 3.6.
| Effective haemostasis (24 h) | 72.4% PCC vs 65.4% plasma — non-inferior (difference 7.1%, 95% CI −5.8 to 19.9) |
| Rapid INR reduction (≤1.3 at 0.5 h) | 62.2% PCC vs 9.6% plasma — superior (difference 52.6%, 95% CI 39.4 to 65.9) |
| Safety | Adverse events, thromboembolic events and deaths similar between groups |
Sarode R, et al. Circulation 2013;128(11):1234–43. PMID 23935011. Industry-funded (CSL Behring).
CRASH-2 (2010) and its time-to-treatment analysis (2011)
Randomised, placebo-controlled trial of tranexamic acid (1 g over 10 min then 1 g over 8 h) in adult trauma patients with, or at risk of, significant bleeding, within 8 h of injury.
| All-cause mortality | 14.5% vs 16.0% — RR 0.91 (95% CI 0.85–0.97; p=0.0035) |
| Death due to bleeding | 4.9% vs 5.7% — RR 0.85 (0.76–0.96; p=0.0077) |
| Given ≤1 h from injury | RR 0.68 (0.57–0.82) for death due to bleeding |
| Given 1–3 h | RR 0.79 (0.64–0.97) |
| Given >3 h | RR 1.44 (1.12–1.84) — increased death due to bleeding |
CRASH-2 collaborators. Lancet 2010;376(9734):23–32 (PMID 20554319); time analysis Lancet 2011;377(9771):1096–101 (PMID 21439633).
HALT-IT (2020) — tranexamic acid in GI bleeding
International randomised, double-blind, placebo-controlled trial of high-dose tranexamic acid (1 g loading then 3 g over 24 h) in significant upper or lower GI bleeding.
| Death due to bleeding (5 days) | 4% vs 4% — RR 0.99 (95% CI 0.82–1.18) |
| Venous thromboembolism | 0.8% vs 0.4% — RR 1.85 (1.15–2.98) |
| Arterial thromboembolism | 0.7% vs 0.8% — RR 0.92 (0.60–1.39) |
HALT-IT Trial Collaborators. Lancet 2020;395(10241):1927–36. PMID 32563378. CC BY 4.0.
PATCH (2016) — platelet transfusion in antiplatelet-associated ICH
Multicentre, open-label, masked-endpoint randomised trial (Netherlands, UK, France) in adults within 6 h of supratentorial spontaneous ICH, on antiplatelet therapy ≥7 days, GCS ≥8; 190 participants (97 platelet transfusion, 93 standard care).
| Death or dependence (mRS, 3 months) | Adjusted common OR 2.05 (95% CI 1.18–3.56; p=0.0114) — worse with transfusion |
| Serious adverse event in hospital | 42% vs 29% |
| In-hospital death | 24% vs 17% |
Baharoglu MI, et al. Lancet 2016;387(10038):2605–13. PMID 27178479.
ANNEXA-4 (2019) — andexanet for factor Xa inhibitor bleeding
Prospective single-arm study of 352 patients with acute major bleeding within 18 h of a factor Xa inhibitor (mean age 77; bleeding intracranial 64%, GI 26%). There was no control group.
| Anti-Xa reduction | Apixaban 149.7→11.1 ng/mL (92%); rivaroxaban 211.8→14.2 ng/mL (92%) |
| Excellent/good haemostasis (12 h) | 204 of 249 evaluable (82%) |
| 30-day death / thrombotic event | 14% / 10% |
| Key caveat | Anti-Xa reduction was not predictive of haemostatic efficacy overall |
Connolly SJ, et al. (ANNEXA-4). N Engl J Med 2019;380(14):1326–35. PMID 30730782. Industry-funded (Portola).
RE-VERSE AD (2017) — idarucizumab full cohort
Multicentre, prospective, open-label single-arm study of 5 g idarucizumab in 503 patients on dabigatran with uncontrolled bleeding (group A, 301) or needing an urgent procedure (group B, 202).
| Median maximum reversal (4 h) | 100% (95% CI 100–100) |
| Group A bleeding sites | GI 45.5%, intracranial 32.6%; median time to bleeding cessation 2.5 h |
| Group B periprocedural haemostasis | Normal in 93.4% |
| 90-day thrombotic / mortality | 6.3–7.4% / ~18.8% |
Pollack CV, et al. (RE-VERSE AD). N Engl J Med 2017;377(5):431–41. PMID 28693366.
Reading it critically
Where the evidence is strong and where it is not
- The VKA and TXA decisions are trial-grade. PCC-over-plasma (Sarode) and TXA-in-trauma-but-not-GI (CRASH-2, HALT-IT) rest on randomised trials with clinically meaningful endpoints. These are the firmest recommendations in the module.
- The DOAC antidotes are not. ANNEXA-4 and RE-VERSE AD are single-arm cohorts with no comparator and pharmacological (anti-Xa, % reversal) rather than patient-outcome primary endpoints. In ANNEXA-4 the anti-Xa reduction did not even predict haemostasis. This is why NICE recommends andexanet narrowly and why PCC remains the pragmatic off-label alternative.
- PATCH studied spontaneous, not traumatic, ICH. The harm signal is real but the population is not identical to a head-injured trauma patient; it tells you platelets are not a reflex antidote, not that they are forbidden in every scenario.
- Two of the six trials are industry-funded (Sarode by CSL Behring; ANNEXA-4 by Portola). That does not invalidate them, but it belongs in the appraisal.
- The timing findings are the most actionable. CRASH-2’s >3 h harm and HALT-IT’s VTE signal are the two results most likely to change what you actually do at 3 am.
Anticoagulation reversal is one of the few areas of resuscitation where the strongest evidence points at what not to do: don’t reverse without a bleed, don’t give plasma when PCC is available, don’t give tranexamic acid for a GI bleed, don’t transfuse platelets to reverse an antiplatelet. The positive actions for DOAC bleeds rest on weaker, uncontrolled evidence — so match the antidote to the drug and site where a trial or appraisal supports it, and treat everything else as a haematology conversation, not a protocol reflex.
What this is, and what it is not
The Decision tool implements NICE NG39, NG24, CG141, NG232 and TA697 with the BSH bleeding guideline, and quotes them verbatim in “The guidance in full” panels. The Teaching and Evidence tabs are different in kind: they are the author’s synthesis of those guidelines and the primary trials, written to explain the tool, not to extend it. Nothing in them changes what the tool tells you to do.
What this module deliberately does not do
- It is not a local major-haemorrhage protocol. It points to the physiological triggers and the reversal agents NICE and BSH name; your trust’s MHP is the operational document.
- It does not give paediatric doses. Every dose here is adult-derived; NG39 1.5.11 sends bleeding, anticoagulated children to immediate haematologist advice.
- It does not decide whether to restart anticoagulation, or when. It makes that decision explicit and owned; the restart itself is a ward or specialty decision.
- It is not a substitute for haematology advice in any non-VKA reversal — NG39 1.5.10 requires it, and this module says so.
Sources in full — decision tool
- NICE NG39. Major trauma: assessment and initial management (2016). Reversal, TXA and blood-component recommendations in section 1.5.
- NICE NG24. Blood transfusion (2015). PCC for emergency warfarin reversal, recommendations 1.12.3–1.12.6.
- NICE CG141. Acute upper gastrointestinal bleeding in over 16s (2012, updated). Transfusion, PCC and antiplatelet recommendations.
- NICE NG232. Head injury (2023). Tranexamic acid in head injury, recommendations 1.3.17–1.3.18.
- NICE TA697. Andexanet alfa for reversing anticoagulation from apixaban or rivaroxaban (2021, last updated 15 January 2025).
- BSH / BCSH. Makris M, et al. Guideline on the management of bleeding in patients on antithrombotic agents. Br J Haematol 2013;160(1):35–46 (reproducing Keeling et al 2011).
- SmPCs: Beriplex P/N (emc 6236), Octaplex (emc 6566), Praxbind (emc 5073), Ondexxya 200 mg (emc 10933).
Every quotation in the Decision tool is verbatim from the recommendation or SmPC named beside it; quotations in Teaching and Evidence are verbatim from the source cited beside them.
References — teaching and evidence
- National Institute for Health and Care Excellence. Major trauma: assessment and initial management (NG39). 2016.
- National Institute for Health and Care Excellence. Blood transfusion (NG24). 2015.
- National Institute for Health and Care Excellence. Acute upper gastrointestinal bleeding in over 16s: management (CG141). 2012.
- National Institute for Health and Care Excellence. Head injury: assessment and early management (NG232). 2023.
- National Institute for Health and Care Excellence. Andexanet alfa for reversing anticoagulation from apixaban or rivaroxaban (TA697). 2021, last updated 15 January 2025.
- Makris M, Van Veen JJ, Tait CR, Mumford AD, Laffan M. Guideline on the management of bleeding in patients on antithrombotic agents. Br J Haematol 2013;160(1):35–46. PMID 23116425.
- Sarode R, Milling TJ Jr, Refaai MA, et al. Efficacy and safety of a 4-factor prothrombin complex concentrate in patients on vitamin K antagonists presenting with major bleeding: a randomized, plasma-controlled, phase IIIb study. Circulation 2013;128(11):1234–43. PMID 23935011.
- CRASH-2 trial collaborators. Effects of tranexamic acid on death, vascular occlusive events, and blood transfusion in trauma patients with significant haemorrhage (CRASH-2). Lancet 2010;376(9734):23–32. PMID 20554319.
- CRASH-2 collaborators. The importance of early treatment with tranexamic acid in bleeding trauma patients: an exploratory analysis. Lancet 2011;377(9771):1096–101. PMID 21439633.
- HALT-IT Trial Collaborators. Effects of a high-dose 24-h infusion of tranexamic acid on death and thromboembolic events in patients with acute gastrointestinal bleeding (HALT-IT). Lancet 2020;395(10241):1927–36. PMID 32563378.
- Baharoglu MI, Cordonnier C, Al-Shahi Salman R, et al. Platelet transfusion versus standard care after acute stroke due to spontaneous cerebral haemorrhage associated with antiplatelet therapy (PATCH). Lancet 2016;387(10038):2605–13. PMID 27178479.
- Connolly SJ, Crowther M, Eikelboom JW, et al. Full study report of andexanet alfa for bleeding associated with factor Xa inhibitors (ANNEXA-4). N Engl J Med 2019;380(14):1326–35. PMID 30730782.
- Pollack CV Jr, Reilly PA, van Ryn J, et al. Idarucizumab for dabigatran reversal — full cohort analysis (RE-VERSE AD). N Engl J Med 2017;377(5):431–41. PMID 28693366.
Every trial figure on the Evidence tab is taken from the published PubMed abstract. Where a figure was not in the abstract, the page does not state it. No illustration was drawn for this page.