More PACE trial data released

Rather telling @Barry - not much difference is there?

Imagine being shown those 4 graphs with no labels, and being told you could pick your treatment based on the results. Which would you choose?

There really isn’t a clear favourite jumping out for notice!! Perhaps opt for the GET option in the hope you could be that outlier!! But n=1 is no reason to pick one treatment over another of course!

Oh and interesting that the 3 biggest drops in distance walked were on the CBT arm. Wonder why?

 

The GET group did a little better on the 6mwt. One possibility is that this patient group learned to push through and did more effort to increase the amount walked. Does the borg data show this - that they were doing more effort in this group or were there no significant differences between groups?

 

I'll say again: Borg was *not* used to as a measure of RPE on the 6mWT. It was only used on the Step Test. So although it may seem reasonable to do so, I don't think you can make inferences about effort to increase the distance walked - because that was not measured.

I think the best you can say is that, in general (& across all groups), those who found the step test easier to perform at 52 weeks than at baseline, also managed to increase the distance walked in the 6mWT (and that's certainly not exclusively the case).
But we can't say anything about *why* that might be from these data.

 

I had a quick first-pass look at those who had zero entered for their 52w 6mwt (but non-zero at baseline), simply by looking at Excel "pages" of them. Each page is 26 rows, and GET was about 1 3/4 pages; the other arms were around 1 1/2. So I didn't pursue it further, given there didn't seem enough of a difference to be especially significant.

 

After the Matthees data came out, I looked at whether there were any differences between groups in terms of why pts might not have recorded a final WT.
The graph shows mean (SD) physical function at 52 weeks split by whether data on the 52-week walking test was missing or not.



Data on those who recorded PF at baseline and 52 weeks, with or without WT data at 52 weeks.

APT: 42 missing WT at 52 weeks (110 with, total 152), 27.6%
CBT: 23 missing (123 with, total 146), 15.7%
GET: 44 missing (109 with, total 153), 28.7%
SMC: 36 missing (115 with, total 151), 23.8%

 

OK - here's another perspective. This is missing WT data vs PF diff (difference in physical function between baseline and 52 weeks):



Anecdotally, it looks like they were deliberately trying to limit the losses from the CBT group. It would be very interesting to see the centre splits - because I suspect that Barts were trying to do the exact opposite. The thing we are all ignoring about PACE is that it was [also] a competition between GET and CBT, between Barts and Oxford [Edinburgh], between White's GET and Sharpe's CBT.

[eta clarifications]

 

I wonder how well the physical function ratings tally with actual function as would be measured by a 6 minute walking test?

I would have thought that the decision on whether to do the walk test should have been independent of trial arm. So the average and range of physical function scores for those choosing NOT to do the walk test “should” have been similar across the 4 treatment arms.

Yet that is not what we see. We see that the CBT group of those deciding NOT to take the test rated themselves higher than the others on physical function. I suppose that is not surprising as they have been coached to rate their scores high.... YET when it comes to the crunch these folk still decided not to do the walk test.

The APT and SMC groups look more alike, and perhaps this is more representative of how people electing NOT to do the walk test again, would rate themselves in a subjective questionnaire without being trained to report improvements.

The GET group was probably also influenced on how to respond, but perhaps not to such an extent as the CBT group.

So I wonder how much we can read into the other differences... ie that there is a bigger difference between the 2 sets of GET scores than for any of the other groups?

 

Ah! Interesting... .I hadn’t quite realised that.

 

Nor had I.

 

I have no idea whether that is actually true - it sort of seemed obvious - but maybe it isn't?

 

Interested in your use of the word "deciding". Is this simply that participants were expected to do the walk test, but of course with the option to decline? Or was it offered as a mutual decision between therapist and participant whether to do it or not? In the latter case the potential for selection bias would be huge.

 

Oh. The way you said it sounded like you knew.

 

What I seem to recall is that prototype PACE was a smaller study just comparing CBT and GET. Presumably the idea was to show one better than the other (although presuming may be dangerous). The final trial seems to have been designed in response to reviewers wanting a null comparator and pressure to include a patient preferred option (pacing). It was also I think larger. In a way all these changes were probably a mistake. A simple comparison could have been quite fair - that is as long as local centres were not trying to spin the results this way or that.

 

Oops!

 

That would be an interesting form of bias.

 

It makes sense considering both had already made their way into NICE guidelines. If either of them had been shown to be useless it would have been difficult to maintain the credibility of the process that recommended both. The point was not to show which is better, but that they are both "valid", since that had already been decided by the kabuki theater of the 2007 NICE review process. Same with the piss-poor Cochrane reviews and why the 2017 review was simply waived off initially.

So not so much a competition of ideas, they are both junk anyway, but a way to ensure evidence would be cherry-picked to show a minimal fake benefit to align with prior claims that both are "safe and effective", whatever that means, something that was explicitly acknowledged alongside the fact that they simply prefer results that make them look good.

Which yet again aligns with a confirmation exercise whose destination was pre-planned and guaranteed to deliver. And despite that, just about the weakest claim to evidence that could be pushed through a biased review process: "it's slightly better than nothing for a tiny % of carefully selected sample if you squint really, really hard from the right angle and don't ask too many questions about who those participants were or the meaning of words like 'fatigue', 'recovery', 'normal range' and 'science'".