Variability in effect sizes of exercise therapy for knee osteoarthritis depending on comparator interventions

•The content of comparator groups in knee OA exercise therapy RCTs varied greatly.•Effect estimates of exercise for knee OA varied depending on comparator interventions.•Participant and intervention characteristics did not impact the effect estimates Systematic reviews of exercise therapy for knee osteoarthritis (OA) have largely ignored the variability in comparator interventions. To assess how effect estimates of exercise therapy for knee OA as reported in randomized controlled trials vary depending on the comparator interventions. We followed the Cochrane Handbook and PRISMA guidance to conduct and report this meta-epidemiological study. Randomised controlled trials (RCTs) were identified from systematic reviews published in 2015 or later and reference lists of included studies. Exercise therapy RCTs testing interventions that adhered to the American College of Sports Medicine (ACSM) guidelines compared to any comparator intervention in people with knee OA and reporting outcomes of knee pain, physical function and/or quadriceps strength at the end of intervention were included. Thirty-five RCTs with 2412 participants were included. Comparator interventions included no intervention, non-ACSM compliant exercise therapy, education/self-management, and passive modalities. For pain, standardized mean difference (SMD) for ACSM compliant exercise therapy compared to passive modalities was 1.76 (95% CI 0.49, 3.04), no intervention 0.93 (95% CI 0.50; 1.36), education/self-management 0.27 (95% CI 0.07, 0.47), and non-ACSM compliant exercise therapy 0.09 (95% CI -0.06, 0.23). For physical function, SMD for ACSM compliant exercise therapy compared to passive modalities was 1.29 (95% CI 0.41, 2.17), no intervention 0.76 (95% CI 0.15, 1.36), non-ACSM compliant exercise therapy 0.25 (95% CI -0.00, 0.51) and education/self-management 0.21 (95% CI -0.14, 0.55). For quadriceps strength, SMD for ACSM compliant exercise therapy compared to no intervention was 0.69 (95% CI 0.42, 0.96), non-ACSM compliant exercise therapy 0.23 (95% CI -0.01, 0.46), education/self-management -0.02 (95% CI -0.45, 0.42) and passive modalities 0.80 (95% CI -0.10, 1.71). The effect of exercise therapy for knee OA varies significantly depending on the comparator intervention. This variability should be assessed routinely in systematic reviews.