Abstract 2634: Warfarin INR-Response Curves in Subgroups of Patients with Nonvalvular Atrial Fibrillation: Should Target Anticoagulation Intensity be Adjusted for Patient Characteristics?: The ATRIA Study
Warfarin is highly effective in preventing thromboembolism (TE) in patients with atrial fibrillation (AF) but raises the risk of intracranial hemorrhage (ICH). Some guidelines suggest adjusting INR targets according to patient features. We used a graphical approach to examine whether history of stroke, age, or CHADS2 stroke risk score influenced the optimal INR range. We used a case-control analysis nested within our AF cohort of 13,599 individuals to generate INR-response curves. Clinical data and warfarin use were obtained from clinical, pharmacy and laboratory databases. TE (primarily stroke) and ICH were identified from ICD-9 codes and validated by chart review. Cases were compared to 4 randomly selected controls matched on all established AF stroke risk factors. Admission INR was used for cases; INR matched to case calendar date was used for controls. Odds ratios are reported relative to INR 2.0 –2.5. Curves were smoothed using cubic splines. Among 9,217 patients contributing 33,497 person-years on warfarin we identified 397 TEs and 164 ICHs. Overall, the odds of TE increased sharply at INR values <2.0, while the odds of ICH increased sharply at INR values >3.5. A similar pattern was observed for patients with and without prior stroke (Figure 1⇓), for patients stratified by age (<75 vs. >=75) and by CHADS2 score (0 –2 vs. 3– 6), (figures not shown). Our “U-shaped” curves indicate that the standard INR target of 2.0 –3.0 maximizes TE prevention while minimizing risk of ICH. This general pattern of INR-response was similar across patient subgroups arguing against adjusting INR targets for the patient characteristics we tested.