PACE trial TSC and TMG minutes released

No, I think what they did was correct for a power analysis. Which is to estimate likely effect sizes based on previous studies and ensure sufficient numbers of patients to detect effects of that magnitude. Power analyses are about preventing the possibility of a type II error (missing an effect that's really there). The issue of power is different from that around whether an effect is clinically important.

 

I am pretty much with @Lucibee n this point. What matters, if you are claiming equipoise, is that the cognoscenti have significant doubt about whether a treatment is clinically useful and that you have reasons, either purely theoretical or from inconclusive previous studies, to think it might be.

Power studies are certainly to avoid type II errors but you want to avoid a type II error of clinical significance, not just a type II error that matches a rather measly result someone got before.

I have always been surprised by the size of PACE. When I did a four arm trial we reckoned we needed about 160 patients (and so recruited 160). For treatments to be much use for chronic disease you want the sort of level of benefit that gives you that sort of number. For things like prevention of cardiac problems with aspirin things are different because only a small proportion of recruits will ever get ill. I may be wrong but it seems to me that a trial that has to recruit 640 patients is looking for something hardly worth bothering with in the first place.

On the hand I think you can claim equipoise and still state which treatments you expect to do best. In a sense it is just as unethical to do a trial if you have no reason to expect something to do best. All that equipoise requires is that there is genuine doubt. That would mean for instance, that you have problems leaving patients on a placebo if you knew there was an proven alternative effective treatment they could be having during the period of the trial.

 

That pacing involved a therapist and visit to the clinic was an attempt to move it closer to CBT/GET in terms of placebo effect (not the best choice of words but you know what I mean). We have no idea how strong the placebo effect component was in each treatment arm. Ultimately the comparison of pacing and CBT/GET cannot be made with subjective measures because CBT/GET intentionally tries to introduce bias, while pacing does not. It's like comparing apples and oranges.

 

The adaptive pacing therapy manual for therapists in the PACE also made a complicated therapy out of the simple idea of pacing. It mentions "homework", "activity and fatigue diaries", "daily and weekly planning", "activity analysis sheets" and so on.

Pacing is in my view not clearly defined but I have never seen pacing described as such a complicated activity.

https://www.qmul.ac.uk/wolfson/media/wolfson/current-projects/1.apt-therapist-manual.pdf

 

This is one of life's little ironies.

One of the things held against ME sufferers is a supposed tendency to obsess about symptoms and to appear for appointments with reams of notes.

 

I was under the impression that larger sample sizes are not about establishing efficacy, they're about uncovering potential rare adverse events. You are correct that if it takes 160 patients per arm to achieve minimal statistical significance, the effect is unlikely to be clinically significant.

 

But if visiting the clinic causes exacerbation of symptoms, then you can't assess it properly. If they had been truly concerned about harms, they would have started from a completely different place. It was never a fair comparison, but I don't think they ever really cared about that.

 

That only really applies to phase 4 or supplementary phase 3 trials following proof of concept at phase 2 or 3. PACE is probably phase 2/3 in that sense.

 

For those who expressed an interest... (@Robert 1973 @Graham @Adrian @Trish @Esther12 @Luther Blissett @BruceInOz @Barry @Sasha @Jonathan Edwards )
I've now done a step test blog: https://lucibee.wordpress.com/2018/07/06/pace-trial-tiptoeing-around-the-step-test/

It's a bit rough and ready, so amendments and additions are likely.

 

I looked to see if anyone (of the dozens) involved in the trial had any expertise in exercise physiology - the only one I could find was on the Steering Committee (Prof Patrick Doherty - whose expertise is in cardiac rehabilitation) - and he obviously said nothing.

 

A bit OT: I've just seen a load of FOIs to the MRC I'd previously missed (or forgotten about) that seem a bit relevant to these minutes so I thought I'd post about them in here.

https://www.whatdotheyknow.com/request/when_were_the_mrc_informed_that#incoming-881123)

https://www.whatdotheyknow.com/request/cfs_pace_permission_to_drop_more

 

But was this a change to the management of the trial, or simply to the ways they intended to analyse it?

The removal of those two measures and the change to CFQ didn't require any change to be made to any data that was collected from patients, so may not constitute a material change in the trial itself. They were still working out their analysis plan way after the trial and the roles of the TSG/TMG had ended.

 

This comment from the PACE team is pretty funny to look back upon:

https://www.bmj.com/content/347/bmj.f5731/rr/664559

This being the BMJ, the PACE teams' misleading response to Kindlon was printed in their journal: https://www.bmj.com/content/347/bmj.f5963

Kindlon's response to them was not published in the journal: https://www.bmj.com/content/347/bmj.f5963/rr/667107

There were quite a lot of good responses to White's reply actually, and some funny stuff from Sean Lynch offering a semi defence of PACE, eg:

https://www.bmj.com/content/347/bmj.f5963/rr/671597

 

Do we know why Dieppe retired? He should have known there were problems from the outset.