Background:Many trials investigated the beneficial effect of physical activity (PA) on physical function (PF) in people with osteoarthritis (OA), but factors involved in this relationship are poorly understood. Considering the link between OA and obesity and obesity-related disorders, body composition (BC) could be one of these factors.Objectives:To examine the relationships between baseline components of PA and 5-year PF scores, considering BC variables measured at 3 years as potential mediators in theses associations (Figure).Methods:We used data from the KHOALA cohort, a French population-based multicenter cohort of 878 patients with symptomatic knee and/or hip OA, aged between 40 and 75 years old. Baseline PA intensity (Metabolic Equivalent of Task, MET), frequency (times/week), duration (hours/week) and type (weight-bearing or not) were assessed by the Modifiable Activity Questionnaire. PF was measured with the WOMAC questionnaire at 5 years (higher scores = greater functional limitations).Skeletal muscle mass (grams) and fat mass (grams) were measured by dual X-ray absorptiometry (DXA) in 358 patients at 3 years. Fat mass index (kg/m2), appendicular fat mass (kg), % of fat mass, lean mass index (kg/m2), appendicular muscle mass (kg), skeletal muscle mass index (kg/m2or %) were calculated based on DXA data. Sarcopenia was defined according to the FNIH Sarcopenia Project recommendations.A causal mediation analysis was used to highlight the mediating role of BC variables. Bivariate analyses (multiple linear and logistic regressions) were performed to select the variables of interest. Separate generalized linear models were used to describe the relationships between PA components, PF and selected BC variables. Unadjusted and adjusted for baseline confounders (age, gender, number of comorbidities, disease duration, mental health and vitality scores) models were ran.Results:A 1-MET increase in baseline PA intensity was significantly associated with an improvement in PF at 5 years (-3 points). Weight-bearing PA was also significantly associated with better PF scores (-5 points).A 1-MET-increase in PA intensity at baseline was associated with a subsequent decrease at 3 years in fat mass index (-0.86 k/m2), an increase in skeletal muscle mass index (≥ 6%), and a decrease in % of fat mass (-2%). Non-weight-bearing PA was significantly associated with a decrease in fat mass index (-2.5 kg/m2).A 1-point increase in PF score was associated with a reduction in skeletal muscle mass index (calculated from body mass index, -0.3%) and an increase in skeletal muscle mass index (calculated from height, +3 kg/m2). The presence of sarcopenia was significantly associated with a degradation of PF (+7 points).Crude analyses indicated that 20.4% of the effect of baseline PA intensity on PF scores at 5 years was mediated by skeletal muscle mass index (calculated from height), 23.2% by fat mass index and 26.6% by % of fat mass. Similarly, 19.3% of the effect of baseline PA type on PF scores at 5 years was mediated by fat mass index and 15.1% by % of fat mass. After adjustment, we found no longer evidence of a mediating role of BC variables in these associations.Conclusion:We found significant associations between a 1-MET increase in PA intensity, weight-bearing PA at baseline and improvement in PF at 5 years, without any mediating role of BC variables. Further studies are needed to better understand the factors involved in these associations, especially psychosocial variables.Disclosure of Interests:Maud Wieczorek: None declared, Christine Rotonda: None declared, Jérémie SELLAM: None declared, Francis Guillemin Grant/research support from: Francis Guillemin received a grant from Expanscience paid to his institution., Anne-Christine Rat: None declared