To assess factors associated with physical activity (PA) in a large sample of primary care patients, 1,250 outpatients from 75 general practices were approached consecutively. Of these, 1,021 (81.7%) returned short forms of the international physical activity questionnaire (IPAQ) and the arthritis impact measurement scale (AIMS2-SF). In addition, the patient health questionnaire (PHQ-9) was used to assess concomitant depression. A stepwise multiple linear regression analysis with the IPAQ score as dependent variable was performed separately for knee and hip patients. The impact of osteoarthritis (OA) on patients with an affected knee (594/58.2%) was more severe, as reflected in higher scores in the "symptom", "lower body", and "affect" scales of the AIMS2-SF (p < 0.01 for all). These patients were also less active than patients with OA to the hip (p = 0.02 for IPAQ score). Main predictors of PA [change in p(F) for all factors < or =0.001] in knee OA patients were physical limitation to the lower limb (R(2) = 0.180), social network (R(2) = 0.121), pain (R(2) = 0.111), body mass index (R(2) = 0.041), and age (R(2) = 0.032). Predictors for OA at the hip (427/41.8%) differed slightly [change in p(F) for all factors < or =0.003): physical limitation to the lower limb (R(2) = 0.162), pain (R(2) = 0.131), PHQ-9 score (R(2) = 0.092), social network (R(2) = 0.078), and disease duration (R(2) = 0.043). Our findings suggest that factors associated with PA differ depending on the localization of the OA. Our results may help to tailor future interventions more appropriately. Further research is needed to determine whether these tailored interventions will result in increased PA.