To evaluate initial glucocorticoid (GC) therapy in patients with rheumatoid arthritis (RA). Six hundred sixty-nine patients with early RA were followed for 2 years in the multicenter "Course And Prognosis of Early Arthritis" cohort. Treatment was applied according to routine care. Assessments included disease activity (DAS28), disability Hannover Functional Status Questionnaire (FFbH), and treatment details. Mixed models, ANCOVA, and logistic regression models were used for statistical analysis. In total, 518 patients (77 %) received oral GCs at baseline; 20 % received a low dose (<7.5 mg prednisolone/day), 22 % received a moderate (7.5-19 mg), and 35 % received a high dose (≥20 mg). In a multivariate logistic regression analysis, higher DAS28 values (OR 1.3) were associated with the use of higher GC doses at baseline (p < 0.001). After adjusting for age, sex, and baseline DAS28 and DMARDs, the patients who started with high-dose GCs had a greater improvement in DAS28 (month 3) and FFbH (month 6, p < 0.001 each). At 2 years, the mean DAS28 remission rates and FFbH values were similar. In all GC groups, the mean dose was tapered to 4 mg/day within 6 months. The reported comorbidities were not increased in patients with high-dose GC therapy. Starting treatment with high-dose GCs led to a better clinical response within 3 to 6 months compared to starting patients on lower dosages. Irrespective of the starting approach, rheumatologists tapered GCs down to a low dose within 6 months. With this strategy, clinical outcomes at 2 years did not differ relevantly.