: To examine whether physical activity (PA) was associated with fatigue, and quantify the extent of potential mediation through depressive symptoms or physical function (PF) on the relationship between PA and fatigue in symptomatic knee osteoarthritis (KOA).: This longitudinal study used data from the Multicenter Osteoarthritis Study (n = 484), comprising subjects aged ≥ 50 years. Baseline PA was quantified via an ankle-worn accelerometer. The outcome was fatigue, measured using a 0-10 rating scale at 2 year follow-up. Mediators included gait speed as a measure of PF and depressive symptoms at 2 year follow-up. Mediation analysis was carried out after adjustment for baseline confounders. Stratified analysis by baseline fatigue status [no/low (< 4) and high (≥ 4) fatigue] was performed.: A significant direct association was found between PA and fatigue at 2 years [unstandardized coefficient (B) = -0.054; 95% confidence interval (CI) -0.107, -0.002, p = 0.041]. The PA-fatigue relationship was not mediated by gait speed (B = -0.006; 95% CI -0.018, 0.001) or depressive symptoms (B = 0.009; 95% CI 0.009, 0.028). In the subgroup with high baseline fatigue, direct associations were found between PA and fatigue (gait speed model:, B = -0.107; 95% CI -0.212, -0.002, p = 0.046; depressive symptoms model: B = -0.110; 95% CI -0.120, -0.020, p = 0.017); but in the no/low baseline fatigue group, no significant association was found between PA and fatigue.: In the symptomatic KOA population, higher baseline PA was directly associated with reduced fatigue 2 years later, especially in those with high baseline fatigue. However, this relationship was not mediated by depressive symptoms or PF.