APOE ε4 and Risk of Intracranial Hemorrhage in Patients With Atrial Fibrillation Taking Apixaban

The APOE ε4 variant is causally linked to cerebral amyloid angiopathy and is a risk factor for intracranial hemorrhage (ICH) among warfarin-treated patients with atrial fibrillation. Nevertheless, its impact on those treated with apixaban remains unknown.

To test the hypothesis that APOE ε4 allele carriership is associated with an increased risk of ICH in patients with atrial fibrillation taking apixaban.

This cohort study involved data from the All of Us Research Program, a longitudinal, population-based study in the United States. Inclusion criteria were age older than 50 years, history of atrial fibrillation, and anticoagulation with apixaban. Participants with a history of ischemic stroke or ICH were excluded. Up to 3 years of follow-up data were available. Data were collected from 2017 to 2022 and analyzed from November 2023 to December 2024.

APOE ε2, ε3, and ε4 were ascertained using variants rs429358 and rs7412. APOE ε4 was modeled dichotomously (noncarriers [no alleles] vs carriers [1 or 2 alleles]).

Incident ICH, including any new intraparenchymal, subdural, or subarachnoid hemorrhage after initiation of apixaban therapy.

Of 413 477 All of Us participants, 2038 were eligible. Their mean (SD) age was 71 (9) years; 918 (45%) were female, 1120 (55%) were male, and 1710 (83%) had European ancestry. Among these participants, 483 (23.7%) were carriers of at least 1 APOE ε4 allele. After a median follow-up of 2.9 years, 26 participants sustained an ICH (cumulative incidence, 1.5%; 95% CI, 1.0%-2.2%), of whom 12 (cumulative incidence, 3.1%; 95% CI, 1.7%-5.3%) were carriers and 14 (cumulative incidence, 1%; 95% CI, 0.6%-1.7%) were noncarriers (P = .007). Multivariable Cox proportional hazard models confirmed this association: compared with noncarriership, APOE ε4 carriership was associated with a 3-fold increase in the risk of ICH (hazard ratio, 3.07; 95% CI, 1.42-6.65). APOE information improved the discrimination of risk prediction scores (C statistic of 0.74 and 0.68 for models with and without APOE, respectively; P = .03).

Further research is needed to evaluate whether cerebral amyloid angiopathy mediates the observed association and whether APOE e4 information improves clinical decision-making about anticoagulation therapy in patients with atrial fibrillation. The latter is important now that APOE information is used in clinical settings to guide antiamyloid treatment for Alzheimer disease and has been returned to millions of persons by direct-to-consumer genotyping companies.