OBJECTIVE: To study the main comedications associated with major bleeding during anticoagulant therapy with coumarins in a non-selected population under everyday circumstances. METHODS: The study population for this retrospective cohort study inclued all new users of phenprocoumon or acenocoumarol aged 40-80 years, during the period 1992-2000 in the PHARMO Record Linkage System. All patients were followed until the last dispensing of phenprocoumon or acenocoumarol, the first bleeding complicationrequiring hospitalization, death, or the end of the study period. The number of days on coumarins alone and the number of days on coumarins in combination with several potentially interactive drugs during follow-up were determined for each patient. RESULTS: The inclusion criteria of this study were met by 19,935 new users of phenprocoumon or acenocoumarol. During follow-up, 552 patients were hospitalized for bleeding. Of all potentially interactive drugs started during anticoagulant therapy by at least50 patients and with at least five bleedings, antibacterial drugs were associated with a four to seven times increased risk of bleeding. Among non-steroidal anti-inflammatory drugs, naproxen had the highest relative risk. Antithrombotic salicylates and tramadol were associated with a three times increased risk of bleeding. CONCLUSION: Antibacterial drugs, non-steroidal anti-inflammatory drugs, antithrombotic salicylates and tramadol were the main potentially interactive drugs associated with major bleeding during anticoagulant therapy with coumarins under everyday circumstances.