Introduction In the Perioperative Anticoagulant Use for Surgery Evaluation (PAUSE) Study a simplified, standardized, pharmacokinetic based perioperative interruption scheme was stablished for patients with non-valvular atrial fibrillation taking direct oral anticoagulants (DOACs). The rate of bleeding and thromboembolic outcomes was low, yet better awareness of bleeding predictors will aid the informed decision making for both clinicians and patients. Thus, we aimed to define the factors associated with perioperative bleeding events in the PAUSE cohorts. Methods The standardized interruption scheme is depicted in figure 1. We analyzed bleeding as the composite of major and clinically relevant non-major bleeding (MB/CRNM) according to the ISTH criteria. Putative predictors of bleeding were prospectively collected during the PAUSE recruitment. For patients taking dabigatran the dilute thrombin time was measured from a pre-operative blood sample. A DOAC-calibrated anti-factor Xa level was assessed preoperatively for patients taking factor Xa inhibitors. Continuous numeric variables were reported as means with SD; frequencies were reported when appropriate. We used stratified logistic regression models for analysis. All statistical analyses were performed in R version 3.6.0. Results There were 3007 patients requiring perioperative DOAC interruption, distributed in the apixaban (A) (N = 1,257), dabigatran (D) (N = 668), and rivaroxaban (R) (N = 1,082) cohorts. Characteristics of the patients who developed MB/CRNM are reported in table 1. More than one third of included patients underwent a high bleeding risk procedure. The rates of MB/CRNM were 3.02% in group A, 2.84% in group D, and 4.16% of group R patients from the point of DOAC interruption to 30 days after the procedure. In the univariate analysis, surgery bleeding risk was significantly associated with MB/CRNM (low vs high, OR 0.56, 95% CI 0.36-0.86; p=0.008). Multivariate analysis stratified by region found hypertension (OR 1.81, 95%CI 1.07-3.07; p=0.027), prior bleed or bleed predisposition (OR 1.62, 95% CI 1.02-2.58; p=0.043) were significantly associated with MB/CRNM (table 2); whereas surgery bleeding risk was not significant after adjusting other covariates. Female gender was associated with lower MB/CRNM risk (This result seems to be mainly driven by the rate of CRNM, 2.21% male had CRNM and 1.57% female had CRNM, and the signal disappeared in the model for MB). [Hypertension (OR 3.93, 95%CI 1.40-11.07; p=0.010), active cancer (OR 2.30, 95% CI 1.10-4.81; p=0.026) were significantly associated with MB (table 3)] Aspirin use was more prevalent among males (13.78%) than females (7.75%), but this did not fully explain the gender effect. The model for MB/CRNM had an area under the curve (AUC) of 0.65 (standard error 0.03). Drug level analysis was available for 2541 (84.5%) patients. In the stratified analysis with the DOAC level (>50ng/dL vs £50ng/dL) as a single predictor, there was no significant association with MB/CRNM observed (OR 1.07, 95% CI 0.38-2.96; p=0.902). In the multivariate model, adding DOAC level to the model did not improve the AUC. Discussion A standardized interruption scheme results in low risk of bleeding for patients with atrial fibrillation interrupting DOACs for surgery/invasive procedures. The protocol-defined surgical bleed risk in PAUSE was valid, and selected to numerically higher bleeding rates, but it was not independent predictor of bleeding when the interruption scheme and covariates are accounted for. Hypertension was the only potentially modifiable predictor of perioperative MB/CRNM bleeding. Likely selecting for a frail population, active cancer and hypertension were predictors of MB events. In the exploratory analyses, residual levels did not improve the predictive model. The multivariate model had a low performance and we do not advocate modification of the PAUSE-proposed interruption schema based on these results. The PAUSE trial is the first large scale study enabled to investigate the factors governing perioperative bleed among DOAC-Anticoagulated patients. Nonetheless, we did not have power to analyze DOAC cohorts independently due to the paucity outcomes. Overall, the results strengthen the validity of the PAUSE-proposed DOAC interruption scheme for atrial fibrillation patients who need surgery. Disclosures Tafur: Recovery Force: Consultancy; Janssen: Other: Educational Grants, Research Funding; BMS: Research Funding; Idorsia: Research Funding; Daichi Sanyo: Research Funding; Stago: Research Funding; Doasense: Research Funding. Spyropoulos:Bayer: Consultancy; Ingelheim: Consultancy; Portola: Consultancy; Janssen: Research Funding; Boehringer INgelheim: Research Funding; Janssen: Consultancy. Douketis:Pfizer: Membership on an entity's Board of Directors or advisory committees; Sanofi: Membership on an entity's Board of Directors or advisory committees; Leo Pharma: Membership on an entity's Board of Directors or advisory committees; Bristol-Myers Squibb: Membership on an entity's Board of Directors or advisory committees; Portola: Other: Consultant ; Janssen: Other: Consultant ; The Merck Manual: Patents & Royalties; Up-to-Date: Patents & Royalties.