The coagulation "cascade" model accurately represents the mechanisms of the prothrombin time and activated partial thromboplastin time tests. However, these tests and the "cascade" model do not accurately reflect the risk of hemorrhage or thrombosis in vivo. In hepatic insufficiency, a balanced reduction in the levels of most of pro- and anticoagulant proteins produced in the liver does not impair thrombin generation until levels are quite low. However, the ability of the coagulation system to tolerate or recover from an insult is markedly impaired in liver disease. This allows the coagulation system to be more easily tipped into a state favoring either hemorrhage or thrombosis.