Background There is a discordance in classification of obesity when defined by body mass index (BMI) or waist circumference (WC). We aimed to examine whether categories of BMI- and WC-defined obesity are differentially associated with the risk of total knee arthroplasty for osteoarthritis. Methods 38,924 participants from the Melbourne Collaborative Cohort Study with BMI and WC measured at baseline (1990-1994) were included. Obesity status was defined as: not obese (non-obese BMI and non-obese WC); WC-defined obesity only (non-obese BMI and obese WC); BMI-defined obesity only (non-obese WC and obese BMI); and BMI- and WC-defined obesity. The incidence of total knee arthroplasty for osteoarthritis between January 2001 and December 2013 was determined by linking participant records to the National Joint Replacement Registry. Results Over 11.5±3.1 years follow-up, 1,875 participants underwent total knee arthroplasty for osteoarthritis. Participants with WC-defined obesity only (HR=1.79, 95%CI 1.51-2.53), BMI-defined obesity only (HR=2.39, 95%CI 2.02-2.84), and BMI- and WC-defined obesity (HR=3.14, 95%CI 2.82-3.49) had an increased risk of total knee arthroplasty compared with those who were not obese. Conclusions Individuals with either BMI- or WC-defined obesity should be targeted for prevention of knee osteoarthritis as both are significant predictors for severe osteoarthritis requiring a total knee arthroplasty.