OBJECTIVE: To determine whether there is an association between osteoarthritis (OA) and incident social isolation using data from the European Project on OSteoArthritis (EPOSA) study. DESIGN: Prospective, observational study with 12 to 18 months of follow-up. SETTING: Community dwelling. PARTICIPANTS: Older people living in six European countries. MEASUREMENTS: Social isolation was assessed using the Lubben Social Network Scale and the Maastricht Social Participation Profile. Clinical OA of the hip, knee, and hand was assessed according to American College of Rheumatology criteria. Demographic characteristics, including age, sex, multijoint pain, and medical comorbidities, were assessed. RESULTS: Of the 1967 individuals with complete baseline and follow-up data, 382 (19%) were socially isolated and 1585 were nonsocially isolated at baseline; of these individuals, 222 (13.9%) experienced social isolation during follow-up. Using logistic regression analyses, after adjustment for age, sex, and country, four factors were significantly associated with incident social isolation: clinical OA, cognitive impairment, depression, and worse walking time. Compared to those without OA at any site or with only hand OA, clinical OA of the hip and/or knee, combined or not with hand OA, led to a 1.47 times increased risk of social isolation (95% confidence interval = 1.03-2.09). CONCLUSION: Clinical OA, present in one or two sites of the hip and knee, or in two or three sites of the hip, knee, and hand, increased the risk of social isolation, adjusting for cognitive impairment and depression and worse walking times. Clinicians should be aware that individuals with OA may be at greater risk of social isolation.