A wide array of recommendations is available for the management of anticoagulation in patients with a prosthetic heart valve scheduled for non-cardiac surgery, ranging from avoidance of replacement anticoagulant therapy in all cases (excluding those with a recent thromboembolic event), to replacement anticoagulant therapy in all, without risk stratification. These guidelines are derived from only a few small- to medium-sized, non-randomized and often methodologically flawed studies conducted during the late 1970s, and applies mainly to caged-ball and caged-disc valves. Furthermore, extrapolation of the thromboembolic risk from data on patients not receiving oral anticoagulants at all is based on assumptions that are not necessarily valid. In this review, the direct and indirect evidence on which these guidelines are based is examined critically. Their applicability to the newer, less thrombogenic valve models is questionable. The need for further prospective, randomized studies is emphasized by the failure of existing studies to adjust properly for the main known or presumed thromboembolic risk factors, and their low statistical power to detect significant differences between protocols in an intention-to-treat manner. The evaluation of obstructive and non-obstructive thrombosis should serve as a secondary outcome measure in the assessment of anticoagulation management before non-cardiac surgery.