In two cohorts of patients on oral anticoagulant therapy, routinely treated by the Leiden Thrombosis Service, the frequency of major bleeding complications was assessed during two years (1988 and 1991). With Poisson regression analysis the influence of the risk factors age, sex, target zone, achieved INR and type of coumarin derivative used were determined. Subsequently, a bleeding risk index was calculated, combining the results of the two cohorts. For various types of patients the relative risk of major bleeding complications was assessed. Age and achieved INR were the most important and consistent risk factors: rate ratio (RR) for age in 1988 1.46 per 10 years increase (95% confidence interval [CI] 1.20-1.78) and in 1991 1.57 per 10 years increase (95% CI 1.23-2.00); RR per unit increase in achieved INR in 1988 1.42 (95% CI 1.21-1.68) and in 1991 1.44 per unit increase in achieved INR (95% CI 1.18-1.74). Two methods were used to combine the results of 1988 and 1991. In the first method the mean bleeding risk index was calculated: In (incidence) [natural logarithm of the incidence rate of major bleeding] = -5.64+ 0.42 * age +0.26 * sex -0.29 * target range +0.36 * achieved INR -0.36 * coumarin type. In the second method only the consistent risk factors age and achieved INR were used: In (incidence) = -5.64 -0.42 * age +0.36 * achieved INR. These bleeding risk indexes can be used to assess the risk of major bleeding complications of individual patients and allow more individualized care by individual tailoring the desired anticoagulation.