Efficacy and potential determinants of exercise therapy in knee and hip osteoarthritis: A systematic review and meta-analysis.

The trials varied widely in exercise type, duration, participants and what usual care meant, so the pooled result may not tell you which programme to use or how to dose it. Most trials could not blind participants or clinicians, and many had high risk of bias in allocation concealment, reporting and small sample size, which may inflate exercise effects. Missing data were common, and some standard deviations were imputed from other trials, so the estimates depend on incomplete reporting. Publication bias was suggested by Egger's test for pain, function and performance, and small study effects may have exaggerated effects, especially at early time points. Subgroup analyses were done at study level and were exploratory, so they cannot predict how one patient will respond. The benefit was time dependent: effects peaked around 2 months and often faded by 9 to 18 months, so a short course may not be enough.

Funding came from Versus Arthritis, formerly Arthritis Research UK, the University of Malaya and the Ministry of Higher Education in Malaysia. The authors declared no competing interests.

Do not read this as proof that any single exercise programme will keep helping for years. The pooled benefit peaked early and then faded, and the subgroup patterns are study-level associations, not tested predictors for an individual patient.