Study Data ExtractionWe reclassified bleeding events in the selected studies according to our defined criteria to allow pooling of findings across studies based on standardized definitions of bleeding outcomes. This was necessary because no accepted standardized definition of bleeding for patients undergoing surgical procedures exists.15 We defined thromboembolic events as stroke, transient ischemic attack, systemic embolism, thrombosis of a mechanical heart valve, thrombosis of the cardiac chamber, deep-vein thrombosis or pulmonary embolism. We defined all-cause mortality as death from any cause. We resolved disagreements about data extraction by consensus and discussion with a fourth reviewer (JD).
For each study, 3 reviewers (AN, AA, SS) independently extracted data about study design, patient characteristics, perioperative interventions and the following clinical outcomes: major, clinically significant nonmajor and minor bleeding; thromboembolic events; re-intervention or re-operation; and all-cause mortality. We defined bleeding outcomes a priori. We defined major bleeding as clinically overt bleeding that was associated with at least 1 of the following: a > 2 g/dL decrease in hemoglobin, a transfusion of > 2 units of red blood cells, bleeding that was fatal, or bleeding that required another operation or reversal of anticoagulation. We defined clinically significant nonmajor bleeding as bleeding that was not major, but resulted in a visit to a medical facility or an unplanned procedure or intervention (e.g., suturing). We defined minor bleeding as bleeding that did not satisfy the criteria for major or clinically significant nonmajor bleeding.