SAT0584 Clinical validation of two panels of biomarkers to predict symptomatic drugs response in knee oa

Background The prediction of drug responses based on the analysis of multiple clinical variables and omics data is mandatory for accomplishing the promise of precision medicine in rheumatology. Objectives Integrating clinical-radiological-analytical variables and proteomics data for predicting patient‘s response to different treatments, in order to optimise therapeutic outcomes in OA. Methods A panel of 10 serum proteins potentially useful to predict OA patient‘s response was qualified using ELISA Kits in the whole Multicentre Osteoarthritis interVEntion trial with Sysadoa (MOVES) cohort. Patients were classified as responders (R) and non-responders (NR), either to Chondroitin sulfate/Glucosamine hydrochloride (CS+GH; Droglican, Bioiberica, Spain) or Celecoxib, after 6 months of treatment. Logistic regression analyses, adjusted by confounder variables previously reported as significant in bivariate approaches, were used to analyse the contribution of the measured proteins to our prediction models of drug response in knee OA. Appropriate receiver-operating-characteristics (ROC) curves were also calculated. Results In the discovery phase of the study, two different panels of putative predictive biomarkers useful to stratify OA patients according to their unique protein profiles was identified by shotgun proteomics (n=80). In the verification phase , the panel of 6 proteins specific for Droglican treatment (APOA2, APOA4, APOH, ITIH1, C4BPa and ORM2), and the panel of 4 proteins specific for Celecoxib treatment (α2HS, SHBG, CD5L and TSP1) were verified in a larger cohort of OA patients (n=262 for CS+GH group and n=244 for Celecoxib group) by ELISA assays. In the qualification phase , the sensitivity and specificity of a panel of 4 validated proteins (ORM2, APOA2, ITIH1, and TSP1) were tested in blind in the whole MOVES cohort at baseline (n=506). In CS+GH group, only ORM2 levels were significantly lower in responders compared to non-responders (R: 192,8 ug/mL vs NR: 261,6 ug/mL; p=0,042), while no statistically significant differences were found in the Celecoxib group. Five clinical and two analytical parameters recorded at baseline significantly influence patients‘ response regardless of treatment. Notably, if we include ORM2 as covariate, we found a specific interaction between response to CS+GH and baseline protein levels (p=0,007) thus increasing the power of our prediction model (ROC up to 0,843). In the Celecoxib group, non parametric analysis showed increased levels only of TSP1 at baseline in responders compared to non-responders (R: 363,03±104,83 ng/mL vs NR: 331,95±92,59 ng/mL; p=0,041), while no statistically significant differences were found for this protein in the CS+GH group. When we include in the regression model 4 predictive variables (2 clinical and 2 analytical) and TSP1 as covariate, we found a specific interaction between response to Celecoxib and baseline protein levels (p=0,045) thus increasing the predictive power of this model (ROC up to 0,749). Conclusions Combining clinical and analytical parameters, we clinically validated (qualified) 2 panels of biomarkers that could efficiently predict OA patients‘ response to CS+GH with an accuracy of 84,3% or to Celecoxib with an accuracy of 74,9%. Disclosure of Interest V. Calamia: None declared, F. Picchi: None declared, I. Rego: None declared, M. Camacho: None declared, L. Gonzalez: None declared, P. Fernandez-Puente: None declared, M. Herrero Employee of: Bioiberica S.A.U., H. Martinez Employee of: Bioiberica S.A.U., J. Verges: None declared, C. Ruiz-Romero: None declared, F. J. Blanco: None declared