Management of patients with a mechanical heart valve or atrial fibrillation who require temporary interruption of warfarin before an elective surgical or other invasive procedure can present a clinical challenge. Although interruption of warfarin is thought to expose patients to a low risk of arterial thromboembolism, such events can have devastating consequences: valve thrombosis is associated with a 15% mortality rate, and an embolic stroke is associated with a 70% rate of major neurologic deficit or death. Attempting to mitigate this risk by administering a short-acting anticoagulant, typically low-molecular-weight heparin (LMWH), shortly before or after surgery as “bridging anticoagulation” can expose patients to serious bleeding complications. To walk this fine line between the twin risks of thromboembolism and bleeding requires an approach to perioperative anticoagulation that is sensible and, when possible, based on evidence. A sensible approach entails first estimating a patient’s risk of thromboembolism, which will help determine whether bridging can be justified on the basis of empiric considerations (as clinical trials addressing this issue have not yet been done). After estimating the patient’s risk, the clinician should next consider the risk of bleeding associated with the surgery and tailor postoperative anticoagulation to minimize this risk. An assessment of postoperative hemostasis is essential, as it can modify reinitiation of postoperative anticoagulation. Finally, after anticoagulant treatment is resumed, ongoing vigilance for bleeding and thromboembolism is needed, especially during the first 1 to 2 weeks after surgery when most of these adverse events occur. An approach to perioperative anticoagulation based on evidence is not as well developed as evidence-based approaches in other areas of antithrombotic therapy, though progress is being made. In recent years, studies (including the one by Wysokinski et al in this issue of Mayo Clinic Proceedings) have addressed the question of “How do we bridge?” What have we learned from these studies? First, administering bridging anticoagulation with subcutaneous LMWH can be done safely in an outpatient setting. This approach has been shown to be userfriendly—more than 85% of patients are able to self-administer LMWH—and more cost-effective than hospitalizing patients to administer intravenous heparin before and after a procedure. Second, if bridging anticoagulation is administered, the final preoperative dose of LMWH should be half the total daily amount and should be given approximately 24 hours before surgery. Administering LMWH in a twice-daily regimen, with the last dose given on the evening before surgery, or as a oncedaily regimen, with a full day’s dose given the day before surgery, results in a residual anticoagulant effect (elevated anti–factor Xa level) at the time of surgery in more than 80% of patients. Although this lingering impairment in hemostasis might not cause clinically relevant bleeding in patients undergoing minor procedures, such as cardiac catheterization or endoscopy, it could be important for patients having surgery associated with a high bleeding risk or receiving spinal anesthetic. Third, prior studies have improved our ability to stratify patients according to their risk of thromboembolism and their risk of bleeding so that bridging anticoagulation, if used, can be administered more judiciously to minimize risks for both types of events. In particular, we are better able to classify surgery and procedure types according to the risk of bleeding. Surgeries associated with high bleeding risks include major orthopedic surgery, cardiac and major vascular surgery, and urologic and neurosurgical procedures. In addition, we have identified procedures that, on the surface, do not seem to confer a high bleeding risk but can lead to serious bleeding with aggressive anticoagulation. Such procedures include prostate and kidney biopsy, pacemaker or defibrillator implantation, and resection of colonic polyps, especially sessile polyps larger than 2 cm in diameter. Further studies, however, are needed to better stratify patient risk of thromboembolism and bleeding. What these studies have not addressed is the overarching question of “Should we bridge?” Considerable cost and inconvenience are associated with bridging therapy; more importantly, its efficacy in preventing thromboembolism is uncertain, and it is possible that it could be doing more harm than good. Although bridging could minimize potentially devastating thromboembolic outcomes, the harm from such treatment could go beyond that of causing postoperative bleeding, which some clinicians consider benign and self-limiting. Postoperative bleeding can delay resumption of warfarin anticoagulation for 1