Practice Signal
This review helps pharmacists decide whether and how to restart a DOAC after intracranial hemorrhage in atrial fibrillation, weighing the ischemic benefit in PRESTIGE-AF and COCROACH against the recurrent bleeding signal, ENRICH-AF's negative result, and where the bleed was.
In atrial fibrillation after intracerebral hemorrhage, every randomized trial points the same way: a DOAC prevents ischemic strokes and causes more brain bleeds. PRESTIGE-AF cut ischemic stroke from 8.60 to 0.83 per 100 patient-years but raised recurrent ICH from 0.82 to 5.00, and ENRICH-AF (ESC 2026) found no reduction in stroke or embolism with edoxaban in unselected patients. Restarting is a shared decision driven by the bleed's phenotype: most defensible after deep, hypertension-related ICH with blood pressure controlled, and hardest to justify after lobar ICH or suspected cerebral amyloid angiopathy.
Patients with atrial fibrillation who survive an intracerebral hemorrhage carry two competing risks: an ischemic stroke from the AF and another bleed in the brain. For years the restart decision rested on observational cohorts, which linked restarting anticoagulation to fewer ischemic events, but those patients were restarted because clinicians judged them safe. The first randomized data came in small pieces. SoSTART (UK, 203 patients) could not show that starting anticoagulation was noninferior for recurrent intracranial hemorrhage (8% vs 4%; adjusted HR 2.42, 95% CI 0.72-8.09), and 7 of the 8 recurrent bleeds in the start group were fatal. APACHE-AF (Netherlands, 101 patients after anticoagulant-associated ICH) found no difference in stroke or vascular death with apixaban (HR 1.05, 95% CI 0.48-2.31), with 4 recurrent ICHs on apixaban vs 1 without and the same number of ischemic strokes (6 in each arm).
COCROACH, a prospectively planned individual-patient meta-analysis of four trials (412 patients), sharpened the picture. Anticoagulation cut ischemic major cardiovascular events from 19% to 4% (HR 0.27, 95% CI 0.13-0.56), but its primary outcome of any stroke or cardiovascular death was not significantly reduced (14% vs 22%; HR 0.68, 95% CI 0.42-1.10), hemorrhagic events trended higher (7% vs 5%; HR 1.80, 95% CI 0.77-4.21), and death or dependence at one year did not differ. By 2023 the shape was clear: the ischemic benefit is real, the hemorrhagic cost is real, and the trials were too small to say which wins.
The 2022 AHA/ASA guideline for spontaneous ICH, written before the larger trials, leaves restarting anticoagulation to individualized judgment. It does stress investigating the cause of every ICH, because the cause changes prevention: in the pooled data it cites, recurrence ran about 7.4% per year after cerebral amyloid angiopathy-related ICH vs 1.1% per year after other ICH.
One subscription unlocks this Practice Signal, every other Signal, and every deep analysis we publish, plus PDF reports.