Although drug interactions with warfarin are an important cause of excessive anticoagulation, their impact on the risk of serious bleeding is unknown. We therefore performed a cohort study and a nested case-control analysis to determine the risk of serious bleeding in 4152 patients (aged 40-84 years) with nonvalvular atrial fibrillation (AF) taking long-term warfarin (> 3 months). The study population was drawn from the UK General Practice Research Database. More than half (58%) of eligible patients used potentially interacting drugs during continuous warfarin treatment. Among 45 identified cases of incident idiopathic bleeds (resulting in hospitalisation within 30 days or death within 7 days) and 143 matched controls, more cases than controls took > or = 1 potentially interacting drug within the preceding 30 days (62.2% vs. 35.7%) and used > 4 drugs (polypharmacy) within the preceding 90 days (80.0% vs. 66.4%). Conditional logistic regression analysis yielded an odds ratio (OR) of 3.4 (95% confidence interval [CI]: 1.4-8.5) for the risk of serious bleeding in patients treated with warfarin and > or = 1 drugs potentially increasing the effect of warfarin vs. warfarin alone adjusted for polypharmacy, diabetes, hypertension, heart failure, and thyroid disease; the adjusted OR for the combined use of warfarin and aspirin vs. warfarin alone was 4.5 (95% CI: 1.1-18.1). We conclude that concurrent use of potentially interacting drugs with warfarin is associated with a 3 to 4.5-fold increased risk of serious bleeding in long-term warfarin users.