Objective:To determine in a prospective population-based cohort study relationships between different measures of body mass and the incidence of severe knee and hip osteoarthritis defined as arthroplasty of knee or hip due to osteoarthritis.Materials and methods:Body mass index (BMI), waist circumference, waist–hip ratio (WHR), weight and percentage of body fat (BF%) were measured at baseline in 11 026 men and 16 934 women from the general population. The incidence of osteoarthritis over 11 years was monitored by linkage with the Swedish hospital discharge register.Results:471 individuals had knee osteoarthritis and 551 had hip osteoarthritis. After adjustment for age, sex, smoking and physical activity, the relative risks (RR) of knee osteoarthritis (fourth vs first quartile) were 8.1 (95% CI 5.3 to 12.4) for BMI, 6.7 (4.5 to 9.9) for waist circumference, 6.5 (4.6 to 9.43) for weight, 3.6 (2.6 to 5.0) for BF% and 2.2 (1.7 to 3.0) for WHR. Corresponding RR for hip osteoarthritis were 2.6 (2.0 to 3.4) for BMI, 3.0 (2.3 to 4.0) for weight, 2.5 (1.9 to 3.3) for waist, 1.3 (0.99 to 1.6) for WHR and 1.5 (1.2 to 2.0) for BF%.Conclusion:All measures of overweight were associated with the incidence of knee osteoarthritis, with the strongest relative risk gradient observed for BMI. The incidence of hip osteoarthritis showed smaller but significant differences between normal weight and obesity. Our results support a major link between overweight and biomechanics in increasing the risk of knee and hip osteoarthritis in men and women.