OBJECTIVES: To evaluate the risk of infection and associated healthcare costs in Rheumatoid Arthritis (RA) patients treated with anti-Tumor Necrosis Factor (anti-TNF) therapy in Alberta, Canada. METHODS: RA patients initiating anti-TNF therapy between January 2004 and March 2009 in Edmonton and Calgary were followed prospectively to identify treatment efficacy and adverse events. Clinical and self-reported data was linked with provincial healthcare administrative databases. Infections (any and severe) were identified by using ICD 9 and 10 diagnosis codes. We used Cox-regression to assess the risk of infection and  linear regression to assess the associated costs. RESULTS: The cohort consists of 1,086 patients (70% female, mean age of 54 years) with a mean follow-up of 2.3 years. Seventy percent of patients (n=764) reported an infection during follow-up, while 4% (n=42) suffered a severe infection. Compared to patients on their first anti-TNF (n=731), patients who switched to another anti-TNF (n=212), patients on DMARD (n=75), and patients switched from DMARD to anti-TNF (n=68) had similar Hazard Ratios (HR) (p>0.05) for both any and severe infection. Pre-existing lung disease (HR=1.98, p<0.001) and heart disease (HR=1.42, p=0.037) increased, while male sex (HR 0.79, p=0.005) decreased the risk of any infection. The risk of a severe infection was increased by underlying anemia (HR=3.20, p=0.018) and in those with longer disease duration (HR=1.03, p=0.032), but was reduced in patients with university-level education (HR=0.34, p=0.018), and osteoarthritis (HR=0.37, p=0.035). In linear regression, Log(cost) was significantly associated with higher baseline HAQ score and longer disease duration and in patients who required a switch between anti-TNF agents for inefficacy or adverse events. CONCLUSIONS: The risk of any or severe infection did not differ significantly between treatment groups. Some pre-existing diseases increased while being male and having university education decreased the infection risk. Healthcare cost variations between the treatment groups were small.