Background and Purpose: Clinical trials on stroke prevention in patients with atrial fibrillation have consistently shown clinical benefit from either warfarin or non–vitamin K antagonist oral anticoagulants (NOACs). NOAC-treated patients have consistently reported to be at lower risk for intracerebral hemorrhage (ICH) than warfarin-treated patients. The aims of this prospective, multicenter, multinational, unmatched, case-control study were (1) to investigate for risk factors that could predict ICH occurring in patients with atrial fibrillation during NOAC treatment and (2) to evaluate the role of CHA ₂ DS ₂ -VASc and HAS-BLED scores in the same setting. Methods: Cases were consecutive patients with atrial fibrillation who had ICH during NOAC treatment. Controls were consecutive patients with atrial fibrillation who did not have ICH during NOAC treatment. As within the CHA ₂ DS ₂ -VASc and HAS-BLED scores there are some risk factors in common, several multivariable logistic regression models were performed to identify independent prespecified predictors for ICH events. Results: Four hundred nineteen cases (mean age, 78.8±8.1 years) and 1526 controls (mean age, 76.0±10.3 years) were included in the study. From the different models performed, independent predictors of ICH were increasing age, concomitant use of antiplatelet agents, active malignancy, high risk of fall, hyperlipidemia, low clearance of creatinine, peripheral artery disease, and white matter changes. Low doses of NOACs (given according to label or not) and congestive heart failure were inversely associated with the risk of ICH. HAS-BLED and CHA ₂ DS ₂ -VASc scores performed poorly in predicting ICH with areas under the curves of 0.496 (95% CI, 0.468–0.525) and 0.530 (95% CI, 0.500–0.560), respectively. Conclusions: Several risk factors were associated to ICH in patients treated with NOACs for stroke prevention but not HAS-BLED and CHA ₂ DS ₂ -VASc scores.