Subtherapeutic anticoagulation levels increase both the risk and severity of thromboembolism. The aim of this study was to determine the cumulative incidence of subtherapeutic international normalised ratios (INRs) and to identify risk factors associated with a low INR. We performed a cohort study in 7,419 patients from a Dutch anticoagulation clinic. Patients who started a first treatment with oral anticoagulants between January 2000 and December 2005 and who were stably anticoagulated (4 consecutive INRs in the therapeutic range) were included. Within the cohort a nested case control study was performed to identify risk factors of subtherapeutic INRs and to determine how often a subtherapeutic INR is the result of medical interference in case of invasive procedures, hospital admissions, haemorrhage or overanticoagulation. In patients with a stable anticoagulation, the median time to a first low INR was 40 weeks. A subtherapeutic INR occurred twice as often in patients using acenocoumarol as in those using phenprocoumon (hazard ratio [HR] 2.1, 95% confidence interval [95%CI]:2.0 - 2.3) and was more common in patients with a high therapeutic range compared to a low therapeutic range (HR 1.8, 95%CI:1.5 - 2.2). Occurrence of a low INR also depended on indication for anticoagulant therapy, with the highest risk in patients who used anticoagulants as prophylaxis and the lowest risk in patients with mechanical heart valves. In 30% of cases the subtherapeutic INR was preceded by an event necessitating vitamin K or discontinuation of the anticoagulant drug.