Twelve studies concerning 10 trials were included.
Six RCTs satisfied at least 50% of the validity criteria. One trial did not satisfy any validity criteria. Methodological flaws included: lack of control for cointerventions; lack of intention-to-treat analysis; and insufficient information on methods used to allocate treatment, level of compliance, control for cointerventions, blinding of outcome assessment, eligibility criteria and description of interventions. Sample size and power of studies varied widely. Two studies had sufficient power (0.8 or more) to detect medium effects (ES = 0.5) and were of acceptable validity.
Most trials compared exercise therapy with either placebo therapy or no treatment.
Pain: 4 different outcome measures were used. Results in the two trials with acceptable validity and sufficient power were ES (hip or knee OA) = 0.58 (95% CI: 0.54, 0.62) and ES for the other trial (knee OA only) = 0.47 (95% CI: 0.44, 0.50) for aerobic exercise and ES = 0.31 (95% CI: 0.28, 0.34) for resistance exercise. In both trials, subjects had radiographic evidence of mild to moderate OA. Conflicting results were noted in the other 4 trials, with one favouring exercise, one reporting ES of borderline significance, and 2 low validity, low power studies reporting no treatment effect. Evidence indicates that exercise therapy has a small to moderate beneficial effect on pain in knee OA and to a lesser extent hip OA.
Self-reported disability: Three different outcome measures were used. Results in the two trials with acceptable validity and sufficient power were ES (hip or knee OA) = 0.26 (95% CI: 0.22, 0.30) and ES for the other trial (knee OA) = 0.41 (95% CI: 0.38, 0.44) for aerobic exercise and ES = 0.36 (95% CI: 0.33, 0.39) for resistance exercise. Results in the other 3 trials (all low validity and low power) were conflicting with two favouring exercise and one favouring the control therapy. Evidence indicates that exercise therapy has a small beneficial effect on self-reported disability in knee OA and to a lesser extent hip OA.
Observed disability in walking: Four different outcome assessments were used. Results in the two trials with acceptable validity and sufficient power were ES (hip or knee OA) = 0.28 (95% CI: 0.24,0.32) and ES for the other trial (knee OA) = 0.89 (95% CI: 0.85, 0.93) for aerobic exercise and ES = 0.31 (95% CI: 0.28, 0.34) for resistance exercise. Results in the other two studies were conflicting with one (acceptable validity, low power) favouring the control intervention and the other (low validity, low power) trial favouring exercise. Exercise has a small beneficial effect on walking performance.
Patient's global assessment of effect: Exercise has a medium to great beneficial effect according to one trial with acceptable validity and sufficient power reported ES (hip or knee OA) = 0.64 (95% CI: 0.60, 0.68). The other trial reporting this outcome (acceptable validity, low power) also favoured exercise.
Comparisons between different exercise therapy: None of the 4 trials comparing interventions had both acceptable validity and high power. No evidence was available in favour of one particular type of exercise programme.