PURPOSE The practicing dentist must frequently advise on the risks involved with dental extractions in the patient taking an anticoagulant. This study assessed the risk of bleeding in a large heterogeneous cohort of patients on warfarin treated by practitioners in training (dental students and junior staff in a teaching hospital). MATERIALS AND METHODS This was a retrospective case-and-control study of 439 patients on warfarin (1,022 extractions) and 439 matched controls (1,049 extractions). Patients with an international normalized ratio (INR) lower than 2.2 had no specific measures, those with an INR 2.2 to 4 received suturing and tranexamic acid mouthwash, and those with an INR higher than 4 did not undergo extraction. Bayesian methods were used to estimate posterior probabilities of bleeding. RESULTS Of cases, 63% were men, 25% were older than 80 years, 40% had an INR lower than 2.2, and 9% had an INR higher than 3. Nine cases bled 0 to 10 days postoperatively, with 1 requiring admission and transfusion. Significant predictors of bleeding were INR and number of extractions (P < .001 for the 2 comparisons). There were no events of bleeding in controls or cases with an INR lower than 2.2 (95% credible interval [CrI] for difference, -0.7 to 1.6). The posterior mean of bleeding was 1% (CrI, 0.1-2.6) for an INR lower than 2.2, 2.3% (CrI, 0.9-4.5) for an INR of 2.2 to 3, and 8.4% (CrI, 3.5-15) for an INR higher than 3. CONCLUSION Unselected patients taking an anticoagulant with an INR lower than 2.2 had a similar risk of bleeding as control patients. The risk was approximately 1 in 40 in those with an INR of 2.2 to 3, whereas the risk in patients with an INR higher than 3 was approximately 1 in 11.