Nine RCTs were included in the review (n=1,234). Study size ranged from 16 to 741 participants. All RCTs were randomised and reported sample size calculation and compliance. Allocation concealment was adequate in five RCTs.
Exercise was associated with significantly less pain than control for hip osteoarthritis overall (ES -0.38, 95% CI -0.68 to -0.08, p=0.01; n=1,234). Significant heterogeneity was present (I2=0.75).
Exercise was associated with significantly less pain than control in patients with hip as the index joint (ES -0.43, 95% CI -0.80 to -0.06, p=0.02; n=1,063). This was associated with significant heterogeneity (I2=76%).
These analyses were repeated excluding a study (n=741) with substantial methodological differences (exercise not taught by a health professional, rheumatologists randomised but patients not and adherence to exercise was low) that reported significantly less pain with control compared with exercise (ES 0.15, 95% CI 0.01 to 0.29, p=0.03). Exclusion of the study eliminated heterogeneity and maintained significance for both analyses.