This is a finding from a paper from 2004. However I think it is a
useful/important finding and thus thought I'd highlight it. I don't
think it has been highlighted - it is not clear from the report
itself. Tom]
I have just read the Cochrane Review:
Exercise therapy for chronic fatigue syndrome. Edmonds M, McGuire H,
Price J. Cochrane Database Syst Rev. 2004;(3):CD003200. Review. PMID:
15266475
This can be downloaded for free at:
http://www.cochrane.org/reviews/en/ab003200.html (press full text
button)
This is a review of research studies.
In the section, "Effects of interventions: Exercise Therapy vs.
Control (treatment as usual or relaxation+flexibility), it says:
"D - Drop-out: Drop-out was more common among exercise therapy
participants (23/161) than among control participants (13/154) (RR
1.73, 95% CIs 0.92 to 3.24,), although the difference was not
significant."
However if one looks at the graph, it approaches significance.
Analysis 1.3. Comparison 1 Exercise Therapy vs Control (treatment as
usual or relaxation + flexibility), Outcome 3 Acceptability of
Treatment.
Test for overall effect: Z = 1.71 (P = 0.088)
The normal test of significance is p<0.05. Authors will often then
refer to p values where 0.10<=p<0.05 as a trend. (occasionally, p<0.1
may even be defined as a significant result e.g. De Lange 2008 (Grey
Matter Volume, CBT and CFS study)
This can be referred to.
The fact that the authors don't refer to the p value in the text makes
one wonder whether the wording was influenced by, for example, the
people they thank.
Ellen Goudsmit also previously noticed another oddity about the
Chalder Fatigue Scale in this review i.e. they do some analyses on the
Chalder Fatigue Scale omitting the Powell (2001) study as it used the
bimodal scoring (0-11). The paper says this scale is a 14 question
version. However some of the studies used the 11-item version. These
are combined. One wonders whether none of the people involved or
helping spotted that.
Tom Kindlon