Objective We evaluated C-reactive protein (C-RP), a quantitative marker of the acute phase response, as a potential biomarker of prevalent and incident osteoarthritis of the knee (OAK). Methods Serum C-reactive protein concentrations were characterized with ultrasensitive rate nephelometry in a population-based sample of 1025 women (318 African-American and 707 Caucasian) who are enrollees in a study of musculoskeletal conditions at the mid-life. Assignment of OAK was based on Kellgren-Lawrence (K–L) scores of 2 or more on radiographs. Prevalent OAK was based on the baseline (1996) score while the classification of incident OAK was based on a score of 2 or greater at the follow-up examination 2.5 years later amongst those with a baseline K–L scores of 0 or 1. Results At baseline, the prevalence of radiographic OAK was 12% in participants who were aged 27–53 years and 18% in the subgroup of women aged 40–53 years. The mean C-RP value was 2.31mg/L, with values ranging from below detection (0.3mg/L) to 47.4mg/L. Higher C-RP concentrations were associated with both prevalent and incident OAK ( P 0.0001, and P 0.0001, respectively). For each K–L score increase from 0 to 3, there was a significantly higher mean C-RP value. Compared to women without incident OAK, women who developed OAK in the 2.5-year follow-up had significantly higher baseline C-RP concentrations. Women with bilateral OAK had higher C-RP concentrations than women with unilateral OAK (6.65mg/L±0.56 vs 3.63mg/L±0.42, P 0.007). BMI was highly correlated with C-RP ( r =0.58) and obesity was an effect modifier with respect to OAK and C-RP concentrations. When stratified according presence or absence of OAK and obesity (BMI >30kg/m 2 ), mean C-RP values were: obesity and OAK, 6.3±0.4mg/L; obesity but not OAK, 4.3mg/L±0.2; no obesity but OAK, 1.7mg/L±0.8; and neither obesity nor OAK, 1.3mg/L±0.2mg/L. These stratum means were significantly different from each other, indicating a higher C-RP with OAK after accounting for obesity. Conclusion C-RP, as a measure of an acute phase response and inflammation, is highly associated with OAK; however, its high correlation with obesity limits its utility as an exclusive marker for OAK. Copyright 2002 OsteoArthritis Research Society International. Published by Elsevier Science Ltd. All rights reserved .