Polygenic risk scores in predicting bleeding complications during oral anticoagulant therapy.

BACKGROUND: Currently available clinical tools have shown very limited ability to predict bleeding risk during oral anticoagulant (OAC) therapy. Polygenic risk scores (PRSs) have shown promise for improving risk prediction in cardiovascular diseases, but their clinical utility for predicting bleeding risk remains unknown. We evaluated whether PRSs for intracranial hemorrhage and gastrointestinal bleeding improve bleeding risk prediction in patients receiving OAC therapy. METHODS: We conducted a registry-based cohort study using genotype and linked health registry data from the FinnGen study. The study included 83,186 individuals receiving OAC therapy. PRSs for intracranial hemorrhage and gastrointestinal bleeding were constructed using the novel SBayesRC method and evaluated as continuous and categorical variables in Cox proportional hazards models. Predictive performance was assessed using Harrell's C-index and compared with the HAS-BLED score. RESULTS: During follow-up, 970 intracranial hemorrhages and 2074 gastrointestinal bleeding events occurred. Continuous PRSs were not associated with intracranial hemorrhage (adjusted hazard ratio [HR] per SD, 1.06; 95% CI, 0.99-1.13) or gastrointestinal bleeding (adjusted HR, 1.04; 95% CI, 0.99-1.09). Individuals in the lowest PRS category had a lower risk of gastrointestinal bleeding (adjusted HR, 0.67; 95% CI, 0.47-0.96), but this finding was not replicated using alternative PRS methods or GWAS data. Adding PRSs to the HAS-BLED score did not improve discrimination for intracranial hemorrhage (C-index, 0.62 vs. 0.62) or gastrointestinal bleeding (0.60 vs. 0.60). CONCLUSIONS: Bleeding PRSs provide no meaningful improvement in bleeding risk prediction among patients receiving OAC therapy and do not enhance risk stratification beyond established clinical risk factors.