This is really weird to look at i.e. you remove the most successful result and it now becomes statistically significant:
I didn't get the same figures (actually a bigger difference) when I tried some basic calculations.
However, it is interesting to note that the O'Dowd control group is bigger than the O'Dowd CBT sample.
The lowest scores in both groups were in the O'Dowd study. This has the effect of bringing down the overall average for the control group i.e. artificially increasing apparent benefit for CBT. If one were to use a smaller control group (say 52, the same size as the CBT group), the overall average difference with CBT decreases quite a bit: it may or may not then be statistically significant (my average difference figures are different to start off with as I said so am not going to try to do it properly). Maybe not that important anyway.
A meta-analysis of 4 trials of CBT reporting changes in SF-36 physical function scores indicated no statistically significant difference between intervention and control groups (weighted mean difference, 10.42 [95% CI,3.86 to 24.69];I 2 = 79.6%, 4 trials) (Figure 2) (39, 42, 47, 56). However, physical function scores were higher for the intervention group when an outlier study (59) was removed in a sensitivity analysis (weighted mean difference, 6.02 [CI, 1.05 to 10.88];I 2 = 0.0%; 3 trials) (47, 56, 57).
I didn't get the same figures (actually a bigger difference) when I tried some basic calculations.
However, it is interesting to note that the O'Dowd control group is bigger than the O'Dowd CBT sample.
The lowest scores in both groups were in the O'Dowd study. This has the effect of bringing down the overall average for the control group i.e. artificially increasing apparent benefit for CBT. If one were to use a smaller control group (say 52, the same size as the CBT group), the overall average difference with CBT decreases quite a bit: it may or may not then be statistically significant (my average difference figures are different to start off with as I said so am not going to try to do it properly). Maybe not that important anyway.
Last edited: