Oral anticoagulant therapy has been shown to be effective for several indications. The optimal intensity of anticoagulation for each indication, however, is largely unknown. To determine this optimal intensity, randomised clinical trials are conducted in which two target levels of anticoagulation are compared. This approach is inefficient, since the choice of the target levels will be arbitrary. Moreover, the achieved intensity is not taken into account. We propose a method to determine the optimal achieved intensity of anticoagulation. This method can be applied within a clinical trial as an "efficacy-analysis", but also on data gathered in day-to-day patient care. In this method, INR-specific incidence rates of events, either thromboembolic or hemorrhagic, are calculated. The numerator of the incidence rate is based on data on the INR at the time of the event. The denominator consists of the person-time at each INR value, summed over all patients, and is calculated from all INR measurements of all patients during the follow-up interval. This INR-specific person-time is calculated with the assumption of a linear increase or decrease between two consecutive INR determinations. Since the incidence rates may be substratified on covariates, efficient assessment of the effects of other factors (e.g. age, sex, comedication) by multivariate regression analysis becomes possible. This method allows the determination of the optimal pharmacological effects of anticoagulation, which can form a rational starting point for choosing the target levels in subsequent clinical trials.