The risk of recurrent venous thromboembolism (VTE) must be weighed against the risk of bleeding in deciding to keep patients on extended anticoagulation with vitamin K antagonists (VKAs). Most of the studies of risk of bleeding on VKAs are randomized controlled trials of highly selected patients followed for less than 1 year. We sought to determine the rate of bleeding in 'real world' patients on long-term anticoagulation with VKAs for VTE. We conducted a retrospective cohort study of patients monitored at our anticoagulation clinic who were treated with prolonged anticoagulation (>1 year) for secondary VTE prevention to assess the incidence of significant bleeding in this population. We found that most of our patients had serious comorbidities, including diabetes, cancer and solid-organ transplantation. The overall rate of bleeding was 10 episodes per 100 person-years, with major bleeding 5.2 episodes per 100 person-years. The rate of significant bleeding while on long-term warfarin may be higher than what is anticipated based on outcomes from closely controlled trials.