Also (and perhaps this is as I've only looked at the abstract) they don't seem to define a time scale over which the meaurements are taken.
Skimming the paper assessment times were anything from 12 weeks to 18 months! Doesn't than make a comparison dodgy?
ALso very different treatments they talk about a couple of trials with higher intensity exercise and say something about differences in the result here:
When examining this latter result in more detail,relatively large proportions dropping out of trial follow up, after receipt of GET, occurred in two trials that started GET at a relatively high intensity (70% estimated VO2 max or 70% of heart rate at individual anaerobic threshold), where 18% and 27% dropped out after GET compared to 10% and 15% after control interventions [7, 14]. The third trial with a relatively high proportion of trial drop-outs occurred after anaerobic GET (28% versus 19%)[
Is this saying if you actually get people to do get then the results are worse in terms of drop outs. What may be a key metric is the 'dose' of GET vs the damage/drop out rate.
With the dropout rate we also don't know the treatment profile for example if people stopped doing GET but filled out forms that isn't known or reported. To my mind as structural uncertainties are introduced into measurements then this should be in someway represented in the uncertainty of the results.