Objective: The purpose was to evaluate the association between estimated joint stress from physical activity (PA) and hip/knee osteoarthritis (OA). Design: A nested case-control study was performed using data from the Aerobics Center Longitudinal Study. Participants without self-reported OA at baseline who attended the clinic between 1974 and 1993 and returned a follow-up questionnaire in 1990 or 1995 were eligible. Cases were those who reported a physician diagnosis of OA of the knee and/or hip at follow-up (N=415). A random sample of persons in the remaining cohort were classified as controls (N=1995). PA was measured at baseline by self-report and subjects were classified as 'moderate/high' or 'low' joint stress by PA type. Those reporting no PA were classified as sedentary with 'no' joint stress (the reference group). Men and women were analyzed separately. Stratified analysis and multiple logistic regression were used to assess the relationship between hip/knee OA and joint stress as predicted by PA. Results: After adjustment for age, body mass index, years of follow-up, and history of hip/knee joint injury, among men, there was no association between hip/knee OA and low joint stress while moderate/high joint stress was associated with reduced risk of hip/knee OA (adjusted odds ratio (OR)=0.62, 95% confidence interval (Cl)=0.43-0.89). Among women, both levels of joint stress were associated with reduced risk of hip/knee OA (OR=0.58, 95% Cl=0.34-0.99 for low and OR=0.24, 95% Cl=0.11-0.52 for moderate/high). Conclusions: PA may reduce the risk of hip/knee OA, especially among women. Further research should assess the combined effects of frequency, intensity, duration and joint stress level of PA on incidence of hip/knee OA. © 2002 OsteoArthritis Research Society International. Published by Elsevier Science Ltd. All rights reserved.