OBJECTIVE: To compare, in real-life conditions, the retention rates of anti-tumor necrosis factor (anti-TNF) treatment (etanercept [ETN], adalimumab [ADA] and infliximab [IFX]) initiated as first-line biotherapy for rheumatoid arthritis (RA) and to evaluate, in case of failure, the switch to another anti-TNF or a non-anti-TNF biological. METHODS: Monocentric retrospective cohort including all patients with RA starting a first anti-TNF between 2001 and 2015. RESULTS: Among the 346 patients analyzed, 201 received ETN, 82 ADA and 63 IFX. The first anti-TNF was interrupted in 151 cases. The retention rates were 82.8%, 67.6%, 46.5%, 28.1% and 22.5% at 1, 2, 5, 10 and 15 years, respectively, with a median retention duration of 52.8 (18.9-136.2) months (ETN: 59.3 [19.1-NA), ADA: 79.9 [19.3-136.2] and IFX: 37.2 [17.5-134.5], P = 0.49). The predictive factors of discontinuation were active RA (Disease Activity Score of 28 joints - C-reactive protein [DAS28-CRP] hazards ratio [HR]: 1.22 [1.03-1.45]), inflammatory syndrome (erythrocyte sedimentation rate HR: 1.01 [1.0-1.02]; CRP HR: 1.00 [1.00-1.01]), absence of methotrexate treatment (HR: 0.60 [0.43-0.83]), and corticosteroid use (HR: 1.91 [1.31-2.78]). The patients who switched to another anti-TNF treatment had an inferior retention than those who switched to a non-anti-TNF treatment (HR: 0.39 [0.17-0.87], P = 0.02). CONCLUSION: In real life, there was no difference in retention among the three anti-TNF agents, and 25% of patients continued them at 15 years. After failure of an anti-TNF, the switch to a non-anti-TNF biotherapy showed better retention.