- Question
- Do direct oral anticoagulants (DOACs) reduce the risk of ischaemic stroke without substantially increasing the risk of recurrent intracerebral haemorrhage in survivors of intracerebral haemorrhage with atrial fibrillation?
- Design
- Multicentre, open-label, phase 3 randomised (1:1) trial of 319 patients at 75 hospitals in six European countries, with only the events adjudication committee masked and a median follow-up of 1.4 years.
- Population
- Adults with spontaneous intracerebral haemorrhage, atrial fibrillation with an indication for anticoagulation, and a modified Rankin Scale score of 4 or less; median age was 79 years.
- Intervention
- A DOAC (dabigatran, apixaban, rivaroxaban or edoxaban), with the choice and dose left to the investigator.
- Control
- No anticoagulation; investigators used their judgment to prescribe an antiplatelet drug or no such therapy.
- Primary outcome
- Two coprimary endpoints: first ischaemic stroke (superiority) and first recurrent intracerebral haemorrhage (non-inferiority, with a hazard ratio margin below 1.735), assessed in the intention-to-treat population.
- Results
- First ischaemic stroke was less frequent with DOACs (HR 0.05; 95% CI 0.01-0.36), at a rate of 0.83 versus 8.60 per 100 patient-years. For recurrent intracerebral haemorrhage the non-inferiority margin was not met (HR 10.89; 90% CI 1.95-60.72), with rates of 5.00 versus 0.82 per 100 patient-years.
- Safety
- Serious adverse events occurred in 70 patients (44%) in the DOAC group and 89 (55%) in the no anticoagulant group; 16 (10%) and 21 (13%) patients died, respectively.
- Conclusion
- DOACs effectively prevented ischaemic stroke in these patients, but part of the benefit was offset by a substantially higher risk of recurrent intracerebral haemorrhage.
- Why it matters
- It showed that anticoagulating after intracerebral haemorrhage involves a real trade-off between ischaemic protection and bleeding risk, and underscored the need for further evidence and safer alternatives for selected patients.
Card generated automatically with AI from the abstract of the original publication and the trial registry; every figure was checked against those sources. It may contain errors: verify the data at the source before using it for clinical decisions.
Generated on October 14, 2026