Objective. To evaluate the risk of serious bacterial infections associated with tumor necrosis factor α (TNFα) antagonists among rheumatoid arthritis (RA) patients. Methods. A retrospective cohort study of US RA patients enrolled in a large health care organization identified patients who received either TNFα antagonists or methotrexate (MTX). Administrative data were used to identify hospitalizations with possible bacterial infections; corresponding medical records were abstracted and reviewed by infectious disease specialists for evidence of definite infections. Proportional hazards models evaluated time-dependent infection risks associated with TNFα antagonists. Results. Hospital medical records with claims-identified suspected bacterial infections were abstracted (n = 187) among RA patients who received TNFα antagonists (n = 2,393; observation time 3,894 person-years) or MTX (n = 2,933; 4,846 person-years). Over a median followup time of 17 months, the rate of hospitalization with a confirmed bacterial infection was 2.7% among the patients treated with TNFα antagonists compared with 2.0% among the patients treated with MTX only. The multivariable-adjusted hazard ratio (HR) of infection among the patients who received TNFα antagonists was 1.9 (95% confidence interval [95% CI] 1.3-2.8) compared with patients who received MTX only. The incidence of infections was highest within 6 months after initiating TNFα antagonist therapy (2.9 versus 1.4 infections per 100 person-years; multivariable-adjusted HR 4.2, 95% CI 2.0-8.8). Conclusion. The multivariable-adjusted risk of hospitalization with a physician-confirmed definite bacterial infection was ∼2-fold higher overall and 4-fold higher in the first 6 months among patients receiving TNFα antagonists versus those receiving MTX alone. RA patients were at increased risk of serious infections, irrespective of the method used to define an infections outcome. Patients and physicians should vigilantly monitor for signs of infection when using TNFα antagonists, particularly shortly after treatment initiation. © 2007, American College of Rheumatology.