Objective: To understand the differences in prevalence and incidence estimates of osteoarthritis (OA), according to case definition, in knee, hip and hand joints. Method: A systematic review was carried out in PUBMED and SCOPUS databases comprising the date of publication period from January 1995 to February 2011. We attempted to summarise data on the incidence and prevalence of OA according to different methods of assessment: self-reported, radiographic and symptomatic OA (clinical plus radiographic). Prevalence estimates were combined through meta-analysis and between-study heterogeneity was quantified. Results: Seventy-two papers were reviewed (nine on incidence and 63 on prevalence). Higher OA prevalences are seen when radiographic OA definition was used for all age groups. Prevalence meta-analysis showed high heterogeneity between studies even in each specific joint and using the same OA definition. Although the knee is the most studied joint, the highest OA prevalence estimates were found in hand joints. OA of the knee tends to be more prevalent in women than in men independently of the OA definition used, but no gender differences were found in hip and hand OA. Insufficient data for incidence studies didn't allow us to make any comparison according to joint site or OA definition. Conclusions: Radiographic case definition of OA presented the highest prevalences. Within each joint site, self-reported and symptomatic OA definitions appear to present similar estimates. The high heterogeneity found in the studies limited further conclusions. (C) 2011 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.