OBJECTIVE The purpose of this study was to better define the toxicity of low dose (less than or equal to 15 mg/d prednisone or equivalent) long-term (greater than 1 year) corticosteroids in the treatment of rheumatoid arthritis (RA). METHODS We examined an historical cohort of 112 RA patients on low dose (6.1 +/- 3.1 mg/d, mean +/- SD) long-term (6.2 +/- 4.6 years) prednisone (CS) and compared them to 112 matched RA patients not using prednisone (CO). CS were matched one-to-one with CO for sex (75% women), age (+/- 5 yrs), race (98% white), and duration of disease (+/- 5 yrs). Subjects were determined by review of unselected medical records from three distinct rheumatology practice settings. For CS, charts were abstracted from the date of prednisone start for predefined adverse events (AEs). RESULTS Ninety-two (92) AEs were noted in CS versus 31 in CO and included: fracture (CS:21 versus CO:8), serious infections (CS:14 versus CO:4), gastrointestinal (GI) bleed or ulcer (CS:11 versus CO:4), and cataracts (CS:17 versus CO:5). At time of first AE, CS prednisone average dose was 7.0 +/- 2.6 mg with a duration of 4.9 +/- 3.9 years. Stepwise multiple logistic regression analysis was used to create a model which included all clinically relevant variables and all parameters significantly different at the cohort inception. Prednisone average dose of greater than 10 to less than or equal to 15 mg/d correlated most strongly with the development of an AE (Odds Ratio (OR) = 32.3, 95% Confidence Interval (CI) 4.6, 220). Average prednisone 5 to 10 mg (OR = 4.5, 95% CI 2.1, 9.6), RA nodules (OR = 3.9, 95% CI 1.9, 8.0), and bony erosions (OR = 2.4, 95% CI 1.2, 4.7) also entered the final model. Kaplan Meier survival curves for the development of the first AE showed a dose-response relationship between prednisone and AE occurrence, independent of rheumatoid nodules. Subset analyses utilized a nested case control design for the development of three serious AEs: fractures, serious infections, and GI events. These analyses revealed possible relationships between prednisone use and the development of each specific AE (prednisone use OR: fracture 3.9, 95% CI 0.8, 18.1; infection 8.0, 95% CI 1.0, 64.0; and GI event 3.3, 95% CI 0.9, 12.1). CONCLUSIONS Although disease severity is an important confounding factor, low dose long-term prednisone use equal to or greater than 5 mg/d is correlated with the development of specific adverse events in a dose-dependent fashion.