'Consumer-Contested Evidence: Why the ME/CFS Exercise Dispute Matters So Much' PLOS Blog post by Hilda Bastian

"It's medical science, not religion." Well that's Butterworth told.

https://twitter.com/user/status/1094937607409143808

 

3 cheers for Trevor Butterworth

 

Damn... what a weirdo this guy is.

Wat?

 

However, I would add that Sharpe is (sort of) correct - it is not a trial's job to test hypotheses - it is to test treatments.

That's the same mistake Horton made: "So they were really stepping back and comparing two philosophies, not just two treatments, two philosophies of what chronic fatigue syndrome was."

But it is also the mistake made in the mediation analysis (which Sharpe led), which makes the assumption that the underlying hypotheses are correct when they are not.

So he is trying to have it both ways.

 

Sharpe: I don't have a model.

Reality:



Literally says "illness model" in his own paper that he keeps telling people to read. The dishonesty of this man is limitless.

 

Reality check: those hypotheses are untestable without a mind-reading machine. Plus that's not even close to how deconditioning works and there's no such thing as fluctuating deconditioning.

 

The PACE authors' hypothesis was that GET and CBT would outperform SMC and APT.

The authors had published about GET and CBT, researched its effectiveness in RCT's and promoted their use as effective treatments for ME/CFS before starting with the PACE-trial. None of the authors had published about APT, provided a theoretical model or researched its effectiveness. In fact, some PACE-researchers such as Peter White had been a vocal critic of pacing. Researchers who had studies pacing, described that the PACE-trial mischaracterized it.

All this suggest that the PACE-authors were biased at the start of the trial. They didn't test hypotheses directly but GET and CBT were based on the etiological models they had proposed and promoted in the past. So in all, I think these are arguments we can make.

 

Indeed. But that's the only hypothesis they can test - that one treatment is better than another treatment. They can't say anything about the underlying models.

I can hypothesise that day will follow night, and can test that hypothesis, but that does not confirm my other hypothesis that this is caused by the blackbird singing at 4am.

 

Circular reasonning
https://twitter.com/user/status/1094964018106765314

 

Oh the irony of it!

Would that constitute a Freudian slip?

 

It is difficult, as the PACE team flit between presenting PACE as purely pragmatic, and something more. I think that it's generally best to give them leeway on this, and criticise them in areas where they have less leeway to wriggle out of things, but then there are things like this, where they talk about "PACE which is more explanatory in its design":



(From appendix 7 of the PACE trial ID and appendices folder: "Knowledge of other trials".)

 

CBT/GET can be effective while the underlying hypothesis is wrong. But if GET/CBT turn out to be ineffective that discredits the model the PACE-researchers have devoted their careers to. If day doesn't follow night, then your hypothesis about the blackbird is wrong. So in a way this was a test of their models and they've failed.

Their mediation analysis showed that fear avoidance and damage beliefs were reduced in the GET/CBT groups compared to the APT/SMC groups. Yet according to their protocol-specified outcomes, there was no clinically significant improvement in one group versus the other. Both patients and independent assessors said the GET/CBT therapies adhered to the manual, where the underlying model of GET/CBT was described with instructions on how to use them in practice. So that's a problem for the GET/CBT models they've proposed in earlier publications.

So I think we can say that (1) the PACE-authors were biased towards GET/CBT and the underlying fear-avoidance hypothesis in ME/CFS (2) that the trial did put this model to the test (3) that the model was not supported by the evidence.

I agree with Esther that this is not the most important point to make, just wanted to say that I think it is valid.

 

It might be worth revisiting Michael Sharpes Guardian interview for quotes;
https://www.s4me.info/threads/the-g...ublic-play-in-science.6474/page-3#post-117883

 

https://twitter.com/user/status/1095029437723262976

 

"And there are important and legitimate issues here – not just people who don’t agree with the results."

Suggests awareness of Sharpe and Co's constant response to PACE critics, that they simply object because they do not like the results.

ETA: Added 'object' which people previously had to guess was what I meant!

 

It is contradictory to say that PACE provides no evidence that the treatment works because placebo effects were not controlled for, while also saying it provides evidence that the underlying model is incorrect. Either we know what the change in outcomes meant or we don't. There are good reasons to assume that bias was significant, but we don't know if all the change was due to bias or only a part of it.

Therefore there are good reasons to believe that inducing bias in participants doesn't work because objective outcomes were so poor. Whether a hypothetical genuine change in illness beliefs would work is unclear.

If the PACE authors wanted to save their illness model, they would have to admit that bias significantly confounds results and that PACE was poorly designed.

PS: I probably took the opposite view on this a while ago. It's hard to wrap one's mind around it. That the illness model is not well defined doesn't help either. Do we know if the PACE authors believe that mere bias (placebo effect) is enough to cause improvement in health?

 

I don't see the contradiction.

Not having a control condition makes it easier for GET/CBT to obtain statistically significant effects, because of the extra care and therapist contact, the patient receives. Despite that advantage and despite other favorable circumstances (such as using the Oxford criteria, the Chalder fatigue scale, the biased newsletters), these treatments were unable to show a clinically significant effect.

So it's a bit like they were given a head start and still weren't able to win the race.

If they had won the race (showed clinically significant improvements as specified by the protocol) we would not be certain if they were the fastest because of the head start (the bias). But it doesn't work the other way around: you can't say: 'oh they were the fastest, but they didn't win the race because they were given a head start.' That doesn't make sense. All the biases in the PACE-trial that patients and scientists have documented were like a head start, they made it easier for GET/CBT to obtain clinically significant improvements on fatigue and physical functioning.

So it doesn't really matter that we don't know how much placebo effects came into play in this trial. If there were no placebo effects then GET/CBT failed to show clinically significant improvements as specified by the protocol. If there were strong placebo effects then GET/CBT definitely failed to show clinically significant improvements as specified by the protocol.

 

I'm trying to determine whether it is accurate to say that PACE provides evidence that the illness model behind CBT/GET is incorrect.

My conclusion is that it provides evidence that inducing biased responses does not work. Whether that invalidates the illness model depends on whether the illness model predicts that biased responses are sufficient to cause improvement in health or not.

 

Testing a theoretical model isn't easy. But I guess one tests a hypothesis by looking at predictions of that hypothesis that are testable.

The psychosomatic model proposed by White/Sharpe/Wessely/Chalder says disability in ME/CFS is due to deconditioning and false illness beliefs such as avoidance behavior and that this is reversible for example with GET or CBT. The model predicts that in the circumstances set up in the PACE-trial, GET and CBT would show clinically significant improvements as specified by the protocol. The fact that didn't happen is a serious blow to their model and etiology of ME/CFS. It makes it much more unlikely that, this model corresponds with reality.

I don't see how the possibility of bias in the PACE-trial would undermine such a conclusion, because all bias that was identified would make it easier for GET/CBT to show clinically significant improvements as specified by the protocol. So IMO, it doesn't really matter what the model says about inducing biased responses.

EDIT: You are however correct, that we don't know if patients actually changed their fear-avoidance beliefs because such changes could be due to response bias. So we're less certain in claiming that changing ME/CFS patients fear-avoidance beliefs doesn't work because it didn't result in a clinically significant improvement. Perhaps you meant that, apologies if I misunderstood.