ANNEXA-I Trial: Andexanet Alfa for Factor Xa Inhibitor-Associated Acute Intracerebral Hemorrhage
In adults with acute intracerebral hemorrhage on apixaban, rivaroxaban, or edoxaban within 15 hours of last dose (hematoma 0.5–60 mL, NIHSS ≤35, GCS ≥7), does andexanet alfa improve a composite hemostatic-efficacy surrogate at 12 hours compared with usual care (85.5% of which was 4F-PCC)? First randomized trial of an FXa-inhibitor antidote specifically in acute ICH.
Connolly SJ et al. (NEJM 2024;390(19):1745-1755) · doi:10.1056/NEJMoa2313040 · 530 patients
Population
Included
- Adults with acute intracerebral hemorrhage confirmed on CT or MRI within 2 hours before randomization
- Hematoma volume 0.5 to 60 mL on baseline imaging
- Most recent FXa inhibitor (apixaban, rivaroxaban, or edoxaban) dose within 15 hours before randomization
- NIHSS at most 35
- GCS at least 7 at time of consent
- Time from symptom onset to baseline scan within 6 hours (per amendment; originally 12 hours)
- Patients with an intracerebral hemorrhage who were receiving a factor Xa inhibitor, with the most recent dose taken within 15 hours before randomization
- After the protocol was amended, only patients with an acute intracerebral hemorrhage as the main bleeding event (not subdural or subarachnoid hemorrhage), with an estimated hematoma volume of 0.5 to 60 ml and a maximum score on the National Institutes of Health Stroke Scale (NIHSS) of 35 (scores range from 0 to 42, with higher scores indicating worse neurologic deficit), were eligible
- The hematoma needed to be evident in the cerebrum on a computed tomographic (CT) or magnetic resonance imaging (MRI) scan within 2 hours before randomization
- An amendment changed the eligibility criterion for the time from onset of bleeding symptoms to the baseline imaging scan from 12 hours or less to 6 hours or less
Excluded
- Hematoma volume greater than 60 mL
- Planned neurosurgical intervention within 12 hours of enrollment
- Thrombotic event within 2 weeks before randomization
- Prior andexanet exposure
- Subdural or subarachnoid hemorrhage as primary bleeding site (after protocol amendment)
- GCS below 7 or NIHSS above 35
- A Glasgow Coma Scale score of less than 7 at the time of consent (scores range from 3 to 15, with lower scores indicating poor neurologic function)
- A score of more than 35 on the NIHSS
- Surgery planned within 12 hours after enrollment
- A thrombotic event within 2 weeks before enrollment
- Patients who had previously received andexanet were not enrolled
Source: Connolly SJ et al., New England Journal of Medicine 2024;390(19):1745-1755· Retrieved 2026-06-09
Primary Outcome
Composite hemostatic efficacy at 12 h: hematoma volume change ≤35%, NIHSS rise <7, no rescue therapy 3–12 h
Hemostatic efficacy at 12 h
Composite was driven by hematoma-volume control; NIHSS-change and rescue-therapy components did not differ. Trade-off: ischemic stroke 6.5% vs 1.5% (+5.0 pp, NNH 20); any thrombotic event 10.3% vs 5.6% (P=0.048, NNH 22). No functional benefit (exploratory mRS 0–3 at 30 days: 28.0% vs 31.0%) and no mortality benefit (27.8% vs 25.5%, P=0.51). DSMB-halted early at pre-specified interim — effect size may be overestimated.
Study Arms
- Agent
- Andexanet alfa, a modified recombinant inactive form of human factor Xa that binds and sequesters factor Xa inhibitor molecules, rapidly reducing anti-factor Xa activity and restoring thrombin generation
- Dose
- A high-dose bolus or a low-dose bolus followed by a continuous infusion. The use of a high-dose or low-dose bolus was in accordance with the label approved by the Food and Drug Administration and was based on the type, the amount, and the timing of the most recent dose of the factor Xa inhibitor that was received. The andexanet formulation included 500 mg of mannitol per vial.
- Route
- IV
- Frequency
- Bolus over the course of 15 to 30 minutes followed by a continuous infusion over the course of 2 hours
- Duration
- Approximately 2 hours (bolus plus continuous infusion)
- Co-interventions
- Usual supportive care for acute intracerebral hemorrhage. Prothrombin complex concentrate was not given to the andexanet group as part of the assigned treatment
Dose/administration verbatim from Connolly NEJM 2024 p.1747 (Randomization and Treatments). 78.1% of andexanet patients received the low-dose regimen and 20.1% received high-dose. Anti-factor Xa activity fell by a median of 94.5% at the 1-to-2-hour nadir. Open-label assignment with blinded outcome adjudication.
- Agent
- Usual care determined by local physicians, with andexanet excluded. Prothrombin complex concentrate could be used
- Route
- Per local protocol
- Co-interventions
- Usual care was determined by local physicians at their discretion but excluded andexanet and could include prothrombin complex concentrate. 85.5% of usual-care patients received prothrombin complex concentrate within the first 3 hours (median dose 3000 IU; 92.4% received four-factor PCC)
Control = usual care excluding andexanet (Connolly NEJM 2024 p.1747). Randomization was stratified (after the first amendment) according to the intention to use prothrombin complex concentrate, and according to the time from onset of symptoms to baseline imaging (under 180 vs 180 minutes or more). Anti-factor Xa activity fell by a median of 26.9% with usual care.
Safety
30-day mortality
27.8%
25.5%
30-day mortality 73 of 263 (27.8%) andexanet vs 68 of 267 (25.5%) usual care. Adjusted difference 2.5 percentage points (95% CI -5.0 to 10.0), P=0.51. No mortality benefit; trial not powered to detect a mortality difference. Source: Connolly SJ et al., NEJM 2024, Table 3 p.1753.
Ischemic stroke within 30 days
6.5%
1.5%
Ischemic stroke within 30 days: 17 of 263 (6.5%) andexanet vs 4 of 267 (1.5%) usual care. Increase 5.0 percentage points (95% CI 1.5-8.8). NNH approximately 20 for ischemic stroke. Source: Connolly SJ et al., NEJM 2024, Table 3 p.1753. Note: hemorrhagicStroke field repurposed here to surface the dominant thrombotic-event component; this is an ISCHEMIC stroke signal in ANNEXA-I.
Any thrombotic event within 30 days
10.3%
5.6%
Any thrombotic event within 30 days: 27 of 263 (10.3%) andexanet vs 15 of 267 (5.6%) usual care. Adjusted increase 4.6 percentage points (95% CI 0.1-9.2), P=0.048. Assessed in the extended population (530 patients including those enrolled after database lock for the interim analysis). Source: Connolly SJ et al., NEJM 2024, Table 3 p.1753.
Trial Design
Type
- Multicenter multinational open-label RCT with blinded outcome adjudication
- 1:1 randomization stratified by intention to use PCC and (after amendment) time-to-baseline-scan
- Pre-specified interim analysis at 450 patients with significance threshold P<0.031
- DSMB-halted at interim for hemostatic efficacy (May 2023, 452 patients analyzed)
- Hierarchical testing to preserve family-wise alpha at 5%
Timeline
Enrolled June 2019 to May 2023; stopped early for efficacy at pre-specified interim analysis
N
530
Enrollment
530 patients in the extended safety population (263 andexanet / 267 usual care). Primary efficacy at pre-specified interim: 452 patients (224 andexanet / 228 usual care). 131 sites in 23 countries. Enrolled June 2019 to May 2023. Multicenter multinational open-label RCT with blinded outcome adjudication. Originally planned for 900 patients; DSMB halted at interim (May 2023) after hemostatic-efficacy threshold P<0.031 crossed. Published NEJM 2024.
ClinicalTrials.gov
NCT03661528Bedside Pearl
For acute intracerebral hemorrhage on an FXa inhibitor (apixaban, rivaroxaban, edoxaban) within 15 hours of last dose, andexanet improved a composite hemostatic surrogate (hematoma expansion, NIHSS, rescue therapy) at 12 h from 53.1% to 67.0% (NNT 8). Anti-FXa activity dropped 94.5% vs 26.9%. The cost: ischemic stroke 6.5% vs 1.5% (NNH 20) and any thrombotic event 10.3% vs 5.6% (NNH 22). No functional benefit or mortality benefit at 30 days. Quote both NNT and NNH when consenting; if patient has high cardioembolic baseline (atrial fibrillation, recent thrombotic event) the trade-off shifts. 4F-PCC remains the comparator that 85.5% of usual-care patients actually received.