How big an effect? SF36 data
I'm a latecomer to the Rituximab party and don't have the energy to read the 370-odd preceding posts so apologies if this has been covered already
I think the SF36 data in the Fluge paper gives a useful insight on how big an effect Rituximab has:
van der Meer and colleagues criticised the fatigue scale used in the trial both because it was unvalidated and because it focused on change from baseline rather than absolute scores. They also say the effect on fatigue is small, pointing to the best change score of just under 4 (mean score for all particpants at 24-32 weeks), consistent with "slight improvement"
Fluge and colleagues pointed out that patients peaked at different times reducing the mean peak score, and that scores included non-responders as well as responders.
The SF36 physical function data (Table 4) can get round some of these problems. First, it's a widely used and validated (if imperfect scale), and measures total activity not change scores, both of which should keep van der Meer et al happy. Second, the authors reported
peak SF36 change, ie is based on the maximum change for each individual (I think) so should be more to Fluge et al's liking.
Baseline SF36 Physical Function score = 34 (Standard Deviation 6)
for Rituximab
Maximum change=39% (Rutiximab group vs 11% for control group)
Net maximum change = 28% (ie RXB change less control change)
Mean net maximum change=9.5 points (0-100 scale)
As far as I know, the maximum change is calculated per individual, so an improvement of 15 points is 50% for someone with a baseline score of 30 but only 25% for a patient with a baseline score of 60. However, to get a useable figure here I applied the mean maximum change to the mean baseline score.
For comparison, the PACE trial found a SF36 score change of 7.1 for CBT and 9.4 for GET. However, the enormous difference with PACE is that it was unblinded so the SF36 scores there might be biased by participants in the treatment groups (the 'therapeutic alliance' between patients and therapists was rated 'very high' so patients might unconsciously boost their scores to make their therapist look good). In the Fluge trial all particpants were blinded to which treatment they had.
So, 9.5 is a pretty good improvement, and in most trials that would cause a stir - but it isn't absolutely enormous either.
Of course, the maximum change figure includes non-responders as well as responders so if we pro-rata all the improvement to the responders (9 ex 13 with SF36 data), we get a score of
13.8 for responders, which begins to look pretty good.
However, trials generally don't report mean scores for responders and non-responders separately.
I don't claim this is mathematically rigorous but I think the numbers give probably give a reasonable feel for the effects of the study, using different data. The results appear to be good if not awesome.
Maybe I should add that I think the Fluge study is tremendously exciting. I don't know if rituximab will prove to be a magic bullet for CFS, even if they tweak the dosage, but I do think the study provides striking evidence of an immune system role in the illness. As the authors say:
the presented findings may have a major impact on the direction of biomedical research in CFS.