Predictors of Functional Recovery in Individuals Exposed to Late-Stage Exercise Programs After Total Knee Replacement: A Secondary Analysis of a Randomized Controlled Trial.

Only 199 of the 240 randomised participants were analysed; 18 were missing the WOMAC survey and 23 the performance tests, and the reasons those people dropped out are not reported in this analysis. The sample size was fixed by who had complete data rather than by a power calculation, and the study was underpowered to find predictors in the community exercise group, where only 22 people improved. Follow-up was 6 months, which the authors themselves describe as relatively short. Participants were people willing to have their care decided by randomisation, which may limit how far the results generalise to other people after knee replacement. The link between higher baseline pain and recovery may be regression to the mean rather than a real mechanism, as the authors acknowledge. In the supervised physical therapy group, the medication findings could mean the opposite of what they look like: the authors note that regular NSAID or analgesic use may simply mark people whose pain was worse. Recovery was collapsed into a single measure built from six performance tests, which the authors say may hide differences between the exercise programmes on specific tasks. Rest in activities of daily living and home, nursing facility and inpatient rehabilitation were self-reported categories rather than verified care pathways.

The authors completed the ICMJE disclosure form and reported no conflicts of interest, and the funders played no role in the design, conduct or reporting of the study. The abstract was presented at the 2025 American Physical Therapy Association Combined Sections Meeting.

Do not tell patients that taking NSAIDs or as-needed painkillers will help them recover after knee replacement. In this study people already using those medicines were more likely to recover after supervised physical therapy, but the authors themselves say the same finding could mean those patients simply had worse pain, and the analysis cannot separate the two. Do not read higher baseline pain as a good sign either: the authors state that this link may be regression to the mean rather than a real effect.