PACE trial graphs and gifs

Yes, but it does emphasise the effect of the scaredy-cat therapy called APT. In effect you have one group that is told "You've got CFS, here's some painkillers, now get on with it!" You have two groups that are told "You are making a big drama out of it: get real, it's not as bad as you say!": and you have one group that is told "Ooooh, you are ever so ill. Cut back on everything and take it easy!"

I think the animations reflect that. If you assume that each group is similarly mixed, with a range of reactions to being told what to do, the variations in patterns are fascinating.

 

I know this is not quite what this thread is about but there are a lot of interesting graphs in their analysis of the paper the PACE authors did on reported harms (which I think I've posted somewhere(?)
eta: https://www.sciencedirect.com/science/article/pii/S0022399914001883 )

Harm and the PACE trial by Neil McGregor
This article was first published on June 9, 2017 on The #MEAction Network Australia Facebook page. Neil also sent this to Australia’s Chief Medical Officer

"There is a significant anomaly within this data. The percentage with SERIOUS DETERIORATION should be equal to or less than those reporting a SERIOUS ADVERSE EVENT within the same group. This indicates a significant problem in reporting of the results."

"By 52 weeks most of the variation within the therapy responses was not significantly different from the normal expected disease variation. The data actually shows that the majority of subjects (>90%) actually had an increase in reporting of adverse events (Mean >3.27 events per subject). This increase in reporting of negative outcomes in the vast majority of CFS subjects should be combined with the fact that <10% of subjects actually reported a therapeutic benefit, and between 9 and 25% actually reported a deterioration in their physical functioning. Thus according to the definition of “HARM”, these therapies had a greater potential to be harmful than to be of potential benefit. This is supported by the 12-week patient assessment, which showed a greater proportion of subjects with negative outcomes compared with the expected normal symptom distribution patterns."

https://www.melbournebioanalytics.org/first-news-item/

 

I can see nearly no difference here.

 

I've been watching the animated GIFs a number of times, and the patterns are making me think that we really do need to get additional data about whether the patients were diagnosed with depression, CFS alone, or ME. It seems to me that in all groups the bulk of the patients show the same progression: the difference lies in a significant minority.

That's the trouble with statistics: in aiming to give an overall figure to a process, the vast variations are necessarily compressed into a handful of measures. Simplification is necessary, but we are primarily visual animals, and we are able to see patterns that are not easily gleaned from the data itself.

It seems to me that the behaviour of a significant minority is different in the 4 groups: by that I mean primarily in the way in which they fill in the questionnaires. The more cohesive increase in walking distance among the GET group suggests that their improved group average was not caused so much by patients walking further, but by fewer patients either not improving or actually walking less far – in other words their average was not dragged down so much (at 52 weeks their s.d. was the least of the groups).

Are the unusual ones in each group the most susceptible to suggestions proffered by therapists, or to the rejection felt by being in the SMC group? Could we identify a core group for which the therapies have no effect? Does a diagnosis of ME remove most of these "susceptible" folk, or are they a general feature of all groupings?

 

If you look at Figure 2 in the main PACE paper, the London ME subgroup come out best on the main outcome measures (tho probably not significantly so, and of course that might be biased by drop-outs) - so arguably, those with "ME" (as defined by their version of the London criteria) are more determined to seem better than everyone else and more suggestible.



It's a rubbish figure though. It would have been better to show the PF in the top row and CFQ in the bottom row. Also, they really should have said in each sub-figure how many patients are included in each group at each timepoint. Wish we had the data - then I'd be able to produce a much better comparison!

 

I'd forgotten all about that set of graphs! Thanks!