Disclosure: Funding for this study was provided through the American College of Clinical Pharmacy Research Institute, the Sanofi-Aventis Thrombosis Investigator Development Award. Background Variation in warfarin anticoagulation intensity occurs frequently. Patients who spend time below their desired therapeutic range may be at increased risk for thromboembolism (TE). Previous studies report the risk of TE in relation to anticoagulation time in range over the entire study period or report a snapshot of anticoagulation status (INR level) in the setting of acute TE. However, available data fails to quantify the risk of TE for individual patients presenting with significant subtherapeutic INR excursions. The purpose of this study was to assess among patients with established, stable, therapeutic anticoagulation the proportion that experienced an anticoagulation-related adverse event (ARAE) within 90 days after a significant subtherapeutic INR excursion and calculate the hazard for an ARAE compared to a matched cohort of similar patients who did not have a subtherapeutic excursion. Methods This retrospective, longitudinal study was conducted among [15,000 patients who had been enrolled in a centralized anticoagulation service. Patients were included if they had been treated with warfarin between 1/1/1998 and 12/31/2005. Patients with at least 1 INR value C 0.5 units below their patient-specific therapeutic range (index INR), 2 previous INR results at least 2 weeks apart and within or above their therapeutic range were assigned to the Low INR cohort (LC). The LC was compared to a matched set of Therapeutic Cohort (TC) of patients treated with warfarin during the same period with a therapeutic index INR and no INR value C 0.2 units below the target range in the ensuing 90 days. Episodes of ARAE (TE, bleeding, and death) in the 90 days following the index INR were identified and validated by chart review. Proportions of the individual ARAEs were calculated and multivariate conditional Cox proportional hazards modeling was performed to estimate hazard ratios between the cohorts. Results A total of 1,596 patients in the TC were matched to 1,119 patients in the LC. Indication for anticoagulation was similar between cohorts (P [ .05): atrial fibrillation 45%, venous thromboembolism 35%, heart valve disorder 4%, and other 16%. Greater than 90% of patients in each cohort had a target INR of 2–3. Mean percent of INRs in range was higher for the LC in both the pre-period (LC 65%, TC 56%, P \ .001) and post index INR period (LC 66%, TC 51%, P \ .001). There were no differences in the proportions or adjusted hazard ratios (HR) of thromboembolism (LC 0.4%, TC 0.4%; HR = 1.7, 95% CI 0.2, 12.0), bleeding (LC 1.5%, TC 1.2%; HR = 2.2, 95% CI 0.9, 5.1), or death (LC 3.3%, TC 3.3%; HR forthcoming) between cohorts (all P [ .05). Conclusions Patients with stable, therapeutic anticoagulation experiencing and a significant subtherapeutic INR excursion at least 0.5 INR units below the patient-specific INR range have a low risk of subsequent thromboembolic events that is similar to that experienced in matched patients without such an