Pneumonia is a major cause of mortality and morbidity in rheumatoid arthritis (RA). This study was undertaken to determine the rate and predictors of hospitalization for pneumonia and the extent to which specific RA treatments increase pneumonia risk. RA patients (n = 16,788) were assessed semiannually for 3.5 years. Pneumonia was confirmed by medical records or detailed patient interview. Covariates included RA severity measures, diabetes, pulmonary disease, and myocardial infarction. Cox proportional hazards regression was used to determine the multivariable risk associated with RA treatments. After adjustment for covariates, prednisone use increased the risk of pneumonia hospitalization (hazard ratio [HR] 1.7 [95% confidence interval 1.5-2.0]), including a dose-related increase in risk (< or = 5 mg/day HR 1.4 [95% confidence interval 1.1-1.6], > 5-10 mg/day HR 2.1 [95% confidence interval 1.7-2.7], > 10 mg/day HR 2.3 [95% confidence interval 1.6-3.2]). Leflunomide also increased the risk (HR 1.2 [95% confidence interval 1.0-1.5]). HRs for etanercept (0.8 [95% confidence interval 0.6-110]) and sulfasalazine (0.7 [95% confidence interval 0.5-1.0]) did not reflect an increased risk of pneumonia. HRs for infliximab, adalimumab, and methotrexate were not significantly different from zero. There is a dose-related relationship between prednisone use and pneumonia risk in RA. No increase in risk was found for anti-tumor necrosis factor therapy or methotrexate. These data call into question the belief that low-dose prednisone is safe. Because corticosteroid use is common in RA, the results of this study suggest that prednisone exposure may have important public health consequences.