The antithrombotic potential of oral anticoagulants is undisputed as the frequency of recurrent thrombosis is high unless anticoagulant therapy is continued after hospital discharge. However, the relationship between potency and/or changes in anticoagulant therapy and frequency of complications remains unclear. Optimizing the clinical management of oral anticoagulation information obtained by databases may be advantageous in addition to meeting safety criteria, as described in the "Saarland Model." The Phoenix-database implicates an association between bleeding complications and the hypertensive elderly. From 1968 to 1993 most reports about cerebral/intraspinal bleedings occurred at prothrombin (PT)-values below 20% in the elder patients (>60 years of age) (12%; 367 reports). In the Saarland Model, 60 patients were followed from our department during a 3-year period. Our findings suggest neither a correlation of the range of PT values and the bleeding events nor an association with age or hypertension. It became obvious that "stable phases" of International Normalized Ratio (INR) [+/-15% change of 4 serial controls using nearly constant weekly oral anticoagulant dosages (+/-15%)] might be considered as a valid criterion of safety. At least the individual risk profile determines the patient's fate.