Objective: Preventive strategies against knee osteoarthritis (OA) require a knowledge of risk factors that influence the initiation of the disorder and its subsequent progression. This population-based longitudinal study was performed to address this issue.

Methods: Ninety-nine men and 255 women aged > or =55 years had baseline interviews and weight-bearing knee radiographs in 1990-1991. Repeat radiographs were obtained in 1995-1996 (mean followup duration 5.1 years, median age at followup 75.8 years). Risk factors assessed at baseline were tested for their association with incident and progressive radiographic knee OA by logistic regression.

Results: Rates of incidence and progression were 2.5% and 3.6% per year, respectively. After adjusting for age and sex, the risk of incident radiographic knee OA was significantly increased among subjects with higher baseline body mass index (odds ratio [OR] 18.3, 95% confidence interval [95% CI] 5.1-65.1, highest versus lowest third), previous knee injury (OR 4.8, 95% CI 1.0-24.1), and a history of regular sports participation (OR 3.2, 95% CI 1.1-9.1). Knee pain at baseline (OR 2.4, 95% CI 0.7-8.0) and Heberden's nodes (OR 2.0, 95% CI 0.7-5.7) were weakly associated with progression. Analyses based on individual radiographic features (osteophyte formation and joint space narrowing) supported differences in risk factors for either feature.

Conclusion: Most currently recognized risk factors for prevalent knee OA (obesity, knee injury, and physical activity) influence incidence more than radiographic progression. Furthermore, these factors might selectively influence osteophyte formation more than joint space narrowing. These findings are consistent with knee OA being initiated by joint injury, but with progression being a consequence of impaired intrinsic repair capacity.