Background The emergence of new therapies for the treatment of rheumatoid arthritis (RA), the paucity of head-to-head studies, and the heterogeneous nature of responses to current biologics highlight the need for the identification of prognostic factors for treatment response and retention in clinical practice. Prognostic factors for patient retention have not been explored thoroughly despite data for abatacept and other biologics being available from national registries. Real-world data from the ACTION study may supplement the findings of randomized controlled trials and show how abatacept is used in clinical practice. The aim of this interim analysis was to identify prognostic factors for abatacept retention in patients with RA who received at least one prior biologic agent. Methods A large, international, non- interventional cohort of patients with moderate-to-severe RA who initiated intravenous abatacept in Canada and Europe (May 2008–January 2011) enrolled in the ACTION study. Potential prognostic factors for retention in this interim analysis (data cut-off February 2012; including patients from Canada, Germany, Greece, and Italy) were baseline demographics and disease characteristics, medical history, and previous and concomitant medication. Clinically relevant variables with p ≤ 0.20 in univariate analysis and no collinearity were entered into a Cox proportional hazards regression model, adjusted for clustered data. Variables with p ≤ 0.10 were retained in the final model (backward selection). Results The multivariate model included 834 patients. Anti-cyclic citrullinated peptide (CCP) antibody positivity (hazard ratio [95 % confidence interval]: 0.55 [0.40, 0.75], p < 0.001), failure of <2 prior anti-tumor necrosis factors (TNFs) (0.71 [0.56, 0.90], p = 0.005 versus ≥2 prior anti-TNFs), and cardiovascular comorbidity at abatacept initiation (0.48 [0.28, 0.83], p = 0.009) were associated with lower risk of abatacept discontinuation. Patients in Greece and Italy were less likely to discontinue abatacept than patients in Germany and Canada (Greece: 0.30 [0.16, 0.58]; Italy: 0.50 [0.33, 0.76]; Canada: 1.04 [0.78, 1.40], p < 0.001 versus Germany). Conclusions Real-world prognostic factors for abatacept retention include anti-CCP positivity and fewer prior anti-TNF failures. Differences in retention rates between countries may reflect differences in healthcare systems. The finding that abatacept has potential advantages in patients with cardiovascular comorbidities needs to be confirmed in further research.