To investigate changes in the risk for serious infections (SIs) over time in Japanese rheumatoid arthritis (RA) patients treated with tumor necrosis factor inhibitors (TNFIs). This prospective cohort study included Japanese RA patients who began treatment with a TNFI from 2005 to 2007 (2005 group, n = 716, 634.2 patient years [PY]) and from 2008 to 2011 (2008 group, n = 352, 270.1 PY) at the time or after their enrollment in the registry of Japanese RA patients on biologics for long-term safety (REAL) database. Patients were observed for 12 months or until discontinuation of their initial TNFI in the REAL database. Drug discontinuation reasons and retention rates were analyzed. Incidence rates of serious adverse events (SAEs) were calculated with 95 % confidence intervals (CIs). The Cox proportional hazard model was applied to estimate the risk for SIs. The retention rate in the 2008 group was significantly lower than the 2005 group (p < 0.001). Discontinuation rates due to lack of efficacy or good control for the 2008 group were significantly higher than the 2005 group (p < 0.001). The crude incidence rate ratios comparing the 2008 group with the 2005 group for SAEs were 0.93 (95 % CI 0.65-1.34) and for SIs were 0.50 (0.24-1.03). The 2008 group had significantly lower risk for SIs than the 2005 group after adjusting for covariates (hazard ratio: 0.43 [0.20-0.93]). These results indicate significant decrease of the risk for SIs with TNFI treatment over time; this may be explained by evidence-based risk management of RA patients given TNFIs.