s / Osteoarthritis and Cartilage 22 (2014) S57–S489 S386 685 CLINICAL RECOGNITION OF SYMPTOMATIC MIDFOOT OSTEOARTHRITIS: FINDINGS FROM THE CLINICAL ASSESSMENT STUDY OF THE FOOT M.J. Thomas y, E. Roddy y, T. Rathod y, A. Moore y, H.B. Menz yz, G. Peat y. yKeele Univ., Staffordshire, United Kingdom; z La Trobe Univ., Melbourne, Australia Purpose: Osteoarthritis (OA) is a common yet poorly understood cause of disabling foot pain. In the absence of radiographic confirmation of OA, clinical diagnosis in primary care is inhibited by lack of evidence informing clinical examination. This study aimed to determine whether the presence of symptomatic midfoot OA (SMOA) can be clinically identified in older adults with midfoot pain presenting to primary care. Methods: A diagnostic model using brief clinical assessments was developed using cross-sectional data from 274 adults aged 50 years who had self-reported midfoot pain in the last month and attended a research assessment clinic between 2010-2011. All clinical assessment data were collected by trained physiotherapy or podiatry assessors adhering to a standardised, quality-controlled protocol. Presence of radiographic midfoot OA in at least one of four scored joints (1st and 2nd cuneo-metatarsal joint, navicular-first cuneiform joint, and talonavicular joint) was ascertained by a single reader using a validated atlas and scoring system, and who was blinded to the clinical assessment data. Radiographic OA was defined as a score of 2 for osteophytes or joint space narrowing on either weight-bearing dorso-plantar or lateral views. SMOAwas defined as co-occuring radiographic OA and midfoot pain. One foot per participant was entered into the analysis. The selection of predictor variables was based on known associations with OA or mechanically-driven putative links to SMOA. Significant predictor variables (p<0.25 from likelihood ratio tests) from univariable analyses were simultaneously entered into a multivariable logistic regression model and backward elimination (p1⁄40.05) was performed. The Hosmer-Lemeshow statistic assessed the calibration of the refitted model and the area under the curve (AUC) evaluated discrimination. Histograms visually summarised discrimination. Internal validation of the model was performed using 1000 bias-corrected bootstrap samples with replacement. Results: 274 participants without inflammatory disease comprised 125 men and 149 women (mean age 65 yrs, SD 9). Of these 155 had midfoot pain and 119 had SMOA. 16 univariable analyses identified 9 significant predictors and no collinearity was observed. In addition to forceentered variables (age, gender, body mass index (BMI)), only two independent predictors of SMOA were retained in the multivariable analysis: (i) reduced ankle dorsiflexion with the knee flexed and (ii) absence of a midfoot exostosis. Based on the strength of univariable association, the Foot Posture Index, subtalar inversion and ankle dorsiflexionwith the knee extended appeared tooweak to contribute to the final model, whereas the removal of the Arch Index and foot lengthcorrected navicular height was due to the stronger influence of age explaining these relationships. The final fitted model was well calibrated (p1⁄40.79) but discrimination was poor (AUC, 0.69; 95%CI: 0.62, 0.75). Bootstrapping revealed a small degree of overfitting. The use of categorical predictor variables in continuous form did not identify any other predictors, nor did it improve model performance. Conclusions: Brief clinical assessments offer only marginal improvement to age, gender and BMI for identifying SMOA. Milder severity in a Table: Associations between baseline MRI sum scores and changes of pain and physi