Background Structural joint pathology on MRI has been found in knees without radiographic evidence of osteoarthritis (OA). We have previously shown that structural abnormalities on MRI may be precursors of disease and associated with the detection of incident radiographic knee OA (ROA) up to 2 years later, and in some circumstances up to seven years later. The prognostic value of structural abnormalities on MRI for knee OA, however, is unknown. Objectives Estimate the probability of incident ROA over 10 years of follow-up, according to structural abnormalities detected on baseline MRIs. Methods A subcohort of participants (862 knees, one knee/person) from the Osteoarthritis Initiative (OAI) with at least one knee at risk of developing ROA (i.e., Kellgren-Lawrence (KL) 0,1 at baseline) was randomly selected. MRI-detected structural features of the knee were assessed by expert readers using the MRI Osteoarthritis Knee Score (MOAKS). Participants underwent bilateral posteroanterior fixed-flexion weight-bearing knee radiography at baseline and annually through year 4, and then every two years until year 10. Radiographs were centrally read for KL grade, with ROA defined as KL≥2. Survival was estimated with the Turnbull non-parametric maximum likelihood estimator, a generalization of Kaplan Meier curves for interval censored data. Survival curves were compared using a log rank permutation test available in the R package interval. Hazard ratios were estimated using Cox models fit using the R package icenReg, designed for interval censored data. Features were considered one at a time, with no adjustments. Results Knees with any of the following structural abnormalities had higher estimated probabilities for the development of radiographic OA within two, four and 10 years of follow-up: effusion-synovitis, Hoffa-synovitis, bone marrow lesions in the medial compartment and whole knee, surface area and full thickness cartilage damage in the medial compartment and whole knee, and medial meniscal extrusion (Table 1). Differences in the probability for the development of ROA within two, four, and 10 years of follow-up were significant (pointwise 95% confidence intervals did not include 0), between knees without the abnormality and knees in the most severe category for all features, with one exception at the two-year follow-up. All of these abnormalities were significantly associated with time to ROA over 10 years of follow-up. Conclusion Our results demonstrate the prognostic value of effusion-synovitis, Hoffa-synovitis, bone marrow lesions, cartilage damage, and meniscal extrusion for the development of ROA up to 10 years later. Future directions include developing a predictive model that incorporates multiple features. Disclosure of Interests C. Kent Kwoh Grant/research support from: EMD Serono and Abbvie, Consultant for: EMD Serono, Fidia, Regulus, GSK, Taiwan Lipopsome Company, Inc, Kolon Tissue Gene, Express Scripts, Frank Roemer Shareholder of: Boston Imaging Core Lab (BICL), LLC., Leena Sharma: None declared, Erin Ashbeck Consultant for: EMD Serono, Chengcheng Hu: None declared, Ali Guermazi Shareholder of: Boston Imaging Core Lab (BICL), LLC., Consultant for: Pfizer, AstraZeneca, MerckSerono, TissueGene, Galapagos and Roche