Objective. Knee osteoarthritis (OA) is highly prevalent, especially in the elderly. Preventive strategies require a knowledge of risk factors that precede disease onset. The present study was conducted to determine the longitudinal risk factors for knee OA in an elderly population.Methods. A longitudinal study of knee OA involving members of the Framingham Study cohort was performed. Weight-bearing knee radiographs were obtained in 1983-1985 (baseline) index increased the risk of OA (OR = 1.6 per 5-unit increase, 95% CI 1.2-2.2), and weight change was directly correlated with the risk of OA (OR = 1.4 per 10-lb change in weight, 95% CI 1.1-1.8). Physical activity increased the risk of OA (for those in the highest quartile, OR = 3.3, 95% CI 1.4-7.5). Smokers had a lower risk than did nonsmokers (for those who smoked an average of 1 1 0 cigarettedday, OR = 0.4, 95% CI 0.2-0.8). Factors not associated with the risk of OA included chondrocalcinosis and a history of hand OA. Weight-related factors affected the risk of OA only in women.Conclusion. Elderly persons at high risk of developing radiographic knee OA included obese persons, nonsmokers, and those who were physically active. The direction of weight change correlated directly with the risk of developing OA.Osteoarthritis (OA) is the most common joint disease, especially in the elderly. However, partly because of the slow time course of disease development, no substantial longitudinal studies of risk factors for the disease have been performed. Knowledge of the risk factors for knee OA has been derived from crosssectional studies.Similar to many chronic diseases, the pathogenesis of OA is likely to be multifactorial. Risk factors consistently associated with the disease in cross-sectional studies include older age, female sex, and being overweight (1). In cross-sectional studies, the OA could develop first, leading a person to become sedentary and to gain weight. We have reported that higher weight in early life predisposes patients to knee OA (2), and that weight loss lowers the risk of developing symptomatic knee OA (3). Both results were based on a one-time assessment of OA occurrence (3). We (4) and others ( 5 )