Objective
To evaluate and compare the use of oral glucocorticoids with three classes of bDMARDs in patients with rheumatoid arthritis (RA).

Methods
We included patients from 13 observational registries treated with a TNF-inhibitor, abatacept or tocilizumab and with available information on the use of oral glucocorticoids. The main outcome was oral glucocorticoid withdrawal. A McNemar test was used to analyse the change in the use of glucocorticoids after 1 year. Kaplan-Meier estimates and Cox regressions, adjusted for patient, treatment, and disease characteristics, were used to evaluate glucocorticoid discontinuation in patients with glucocorticoids at baseline. Because of heterogeneity, analyses were done by registers and pooled using random-effects meta-analysis.

Results
A total of 12,334 participants treated with TNF-inhibitors, 2100 with tocilizumab and 3229 with abatacept were included. At one-year, oral glucocorticoid use decreased in all treatment groups (odds ratio for stopping vs. starting of 2.19 [95% CI 1.58; 3.04] for TNF-inhibitors, 2.46 [1.39; 4.35] for tocilizumab; 1.73 [1.25; 2.21] for abatacept). Median time to glucocorticoid withdrawal was ≈2 years or more in most countries, with a gradual decrease over time. Compared to TNF-inhibitors, crude hazard ratios of glucocorticoid discontinuation were 0.65[0.48–0.87] for abatacept, and 1.04 [0.76–1.43] for tocilizumab, and adjusted hazard ratios were 1.1 [0.83–1.47] for abatacept, and 1.30 [0.96–1.78] for tocilizumab.

Conclusion
After initiation of a bDMARD, glucocorticoid use decreased similarly in all treatment groups. However, glucocorticoid withdrawal was much slower than advocated by current international guidelines. More effort should be devoted to glucocorticoid tapering when low disease activity is achieved.