So in summary, it seems that Cochrane agreed with Robert Courtney on using the FINE trial unpublished data and the PACE trial selective reporting bias. But it's not clear if this will significantly alter the results. It seems mostly an illustration of Larun et al.'s bias in conducting this review.
The most important critique - leaving out all objectives outcomes - is not mentioned.
Vinks analysis showed that most of the GET-trials had objective outcomes. Some found improvements, but most of them didn't. So the conclusion of that would probably be that GET fails to show improvement on objective outcomes.
As I see it the main critique addressed in this document, is the fact that at follow-up, GET no longer showed a significant improvement on all outcomes except for sleep. So normally you would expect the review to write something like: "GET showed significant improvements post-treatment but no longer at follow-up". But instead of simply saying that, the Cochrane commentator suggests it is best to state that no reliable conclusion can be made of the follow-up data. So he/she wants the review to state:
“When exercise therapy was compared with 'passive control,' fatigue may be reduced at end of treatment (Analysis 1.1), however we are uncertain if there is any difference between groups at follow up”
So the outcomes at post-treatment are deemed reliable to mention, but the data at follow-up isn't. What the differences between the two?
- If we take fatigue as the main outcome, 453 patients in the intervention group and 387 patients in the control group gave data post-intervention. Data came from 7 studies.
- At follow-up there was data for 373 patients in the intervention group and 297 patients in the control group. Data came from 4 studies.
That doesn't seem like a large difference to me. It's confusing to only state the post-treatment outcomes and act like the follow-up doesn't provide reliable outcomes. There ain't much of a difference between the two. In both cases, the majority of patients came from three studies: Powell et al. 2001, the FINE trial (Wearden et al. 2010) and the PACE trial (White et al. 2011).
The interesting thing is that those three trial reported the follow-up data as their main outcome! So I think it's misleading, as the Cochrane commentator suggests, to state the post-treatment data and consider the follow-up data too unreliable to mention. That's contrary to how these trials were conducted and reported. And for obvious reasons: the response bias is probably much larger shortly after treatment. That's why well-conducted trials measure their main outcome some weeks/months after the end of treatment.