Upper limb motor recovery in chronic stroke-longitudinal aggregate analysis from control group outcomes.

Only 20 studies were pooled and the authors ran no risk-of-bias assessment, so the estimate rests on studies of possibly uneven quality. Most papers did not describe the exact content, dose or frequency of the exercise the control group received, so what sits behind the 0.9-point average is uncertain and subgroup analyses were not possible. People were included from 3 months after stroke, earlier than the usual 6-month definition of chronic stroke, so some of the change may be spontaneous recovery rather than activity. The analysis works from group averages, so it says nothing about which individual patients improved or why three groups got worse. In some studies the standard deviation was estimated from the interquartile range or the range rather than reported directly. Only the before and after scores were used; extra follow-up scores reported by some studies were deliberately left out. The effect size measure used does not account for the correlation between a person's own before and after scores, which the authors say could be addressed with a different measure.

The authors declare that no financial support was received for the research or its publication. No other competing interests are reported in the paper.

Do not read this as evidence that exercise is pointless after stroke. The analysis covers only the control arms of trials, where the exercise given was often not described, and the 0.9-point average could reflect spontaneous recovery, being observed in a study, or an unstructured training effect rather than a real treatment. The authors say themselves that it does not mean exercise is ineffective, only that unsupervised, non-targeted activity was not enough on its own to produce recovery.