Changes in treadmill performance deviated substantially from a normal distribution, and log transformation did not adequately yield an appropriate distribution in most studies. Thus, in accordance with each study’s Statistical Analysis Plan, the meta-analyses of the change in treadmill performance over baseline (PLC vs placebo) were performed using the non-parametric Cochran–Mantel–Haenszel test on the modified ridits, which is equivalent to the van Elteren test.