Various early arthritis clinics have provided extensive data on presentation, disease course and outcomes of early inflammatory arthritis (EIA). The present Review summarizes the epidemiological data from these early arthritis cohorts (EACs), which provide information about the frequency of, risk factors for, and outcomes of EIA and undifferentiated arthritis (UA). The studies demonstrate the large variation in selection criteria and outcome definitions in EACs, which demands careful interpretation of the results. The annual incidence of EIA ranges from 115 to 271 per 100,000 adults, and the incidence of UA ranges from 41 to 149 per 100,000 adults. Depending on the selection criteria used by the specific EACs, 13-54% of patients with UA will develop rheumatoid arthritis (RA) and in 21-87% UA will persist. Epidemiological data from the various EACs has enabled the development of prediction models for persistent and erosive arthritis. These data formed the basis of new classification criteria developed in 2010, which could enable earlier diagnosis and treatment. Future clinical research should focus on the role of imaging techniques in the early detection of synovitis and on the effect of early treatment on the outcomes of EIA.