Objective To assess effectiveness of osteoarthritis interventions to promote long-term physical activity behaviour change.

Design A systematic review and meta-analysis. Protocol registration PROSPERO CRD4201300444 5 (http://www.crd.york.ac.uk/prospero/).

Study selection Randomised controlled trials (RCTs) comparing physical activity interventions with placebo, no/or minimal intervention in community-dwelling adults with symptomatic knee or hip osteoarthritis. Primary outcomes were change in physical activity or cardiopulmonary fitness after a minimum follow-up of 6 months.

Data extraction Outcomes were measures of physical activity (self-reported and objectively measured) and cardiovascular fitness. Standard mean differences between postintervention values were used to describe the effect sizes.

Results 27 984 titles were screened and 180 papers reviewed in full. Eleven RCTs satisfied inclusion criteria, total study population of 2741 participants, mean age 62.2. The commonest reasons for study exclusion were follow-up less than 6 months and no physical activity measures. The majority of included interventions implement an arthritis self-management programme targeting coping skills and self-efficacy. Seven studies used self-report measures, the pooled effect of these studies was small with significant heterogeneity between studies (SMD 0.22 with 95% CI −0.11 to 0.56, z=1.30 (p=0.19) I² statistic of 85%). Subgroup analysis of 6–12 month outcome reduced heterogeneity and increased intervention effect compared to control (SMD 0.53, 95% CI 0.41 to 0.65, z=8.84 (p<0.00001) I² of 66%).

Conclusions Arthritis self-management programmes achieve a small but significant improvement in physical activity in the short term. Effectiveness of intervention declines with extended follow-up beyond 12 months with no significant benefit compared to control. The small number of studies (11 RCTs) limited ability to define effective delivery methods. Investigation of behavioural lifestyle interventions for lower limb osteoarthritis populations would benefit from consensus on methodology and outcome reporting. This includes use of validated physical activity reporting tools and planning for long-term follow-up.