Cognitive behaviour therapy for chronic fatigue syndrome: Differences in treatment outcome between a

[A.B.]

Another paper by our favorite miracle workers:

Cognitive behaviour therapy for chronic fatigue syndrome: Differences in treatment outcome between a tertiary treatment centre in the United Kingdom and the Netherlands

• Treatment outcome of cognitive behaviour therapy (CBT) for chronic fatigue syndrome (CFS) was compared in two tertiary treatment centres in the Netherlands and the U.K.
• Effect sizes on fatigue severity and impairment differed between centres.
• Differences in patient characteristics could not explain variations in outcome.
• Differences in treatment protocols may be responsible for outcome differences.
• More attention should be paid to variation in treatment protocols in relation to outcome, to further develop and improve CBT for CFS.

Effect sizes differed between centres for fatigue (Cohen's D NL = 1.74, 95% CI = 1.52–1.95; UK = 0.99, CI = 0.73–1.25), physical functioning (NL = 0.99, CI = 0.81–1.18; UK = 0.33, CI = 0.08–0.58) and social functioning (NL = 1.47, CI = 1.26–1.69; UK = 0.61, CI = 0.35–0.86). Patients in the UK had worse physical functioning at baseline and there were minor demographic differences. These could not explain differences in centre outcome.

http://www.jpsychores.com/article/S0022-3999(16)30328-2/abstract

 

[Jonathan Edwards]

I think this emphasises the fact that nobody actually knows what CBT is. Is it just talking to somebody, or talking to somebody with a nice blue cardigan, or talking to somebody you wouldn't want to upset so you say you are better to keep them happy or talking to someone with something useful to say that you could not read in a book (and if so what the heck is it?). It would be useful to know because it it's just nice blue cardigans then it would save a lot of expensive 'psychology training'.

 

[Snow Leopard]

I know, perhaps it's due to the inherent uncertainty due to using subjective outcome measures with a therapy that is explicitly designed to change how people report subjective outcomes!

 

[Woolie]

The only good thing about this article is seeing some of our favourite PACE authors intertwined with some of the Dutch folks on an actual publication.

There goes that "independent assessment of PACE by researchers outside of the UK".

Hey, how about the pink background on the Journal of Psychosomatic Research? I think that really adds credibility, no?

 

[A.B.]

I'm more wondering how they conclude that differences in patient characteristis are not responsible for the difference in results. The UK patients are sicker. A rather obvious explanation would be that they are less likely to respond to placebo therapies such as CBT. That said, the effect sizes are very large. Miracle workers indeed.

 

[sarah darwins]

I can't seem to see where they say when they measured the 'effect'. Am I missing that?

 

[A.B.]

The effect size of 1.74 reported by Dutch clinics for fatigue means that about 95% of the control group fell below the average person in the intervention group.

An effect size of 0.99, reported by UK clinics for fatigue and by Dutch clinics for physical function means that about 84% of the control group fell below the average person in the intervention group.

Effect sizes of 0.2, 0.5, 0.8 were defined as "small", "medium", "large" by Cohen.

http://www.leeds.ac.uk/educol/documents/00002182.htm

 

[user9876]

The effect size of 1.74 reported by Dutch clinics for fatigue means that about 95% of the control group fell below the average person in the intervention group.

An effect size of 0.99, reported by UK clinics for fatigue and by Dutch clinics for physical function means that about 84% of the control group fell below the average person in the intervention group.

Effect sizes of 0.2, 0.5, 0.8 were defined as "small", "medium", "large" by Cohen.

http://www.leeds.ac.uk/educol/documents/00002182.htm

However I think the assumptions behind Cohen's test may not be met with the scales used.

 

[Jonathan Edwards]

The effect size of 1.74 reported by Dutch clinics for fatigue means that about 95% of the control group fell below the average person in the intervention group.

An effect size of 0.99, reported by UK clinics for fatigue and by Dutch clinics for physical function means that about 84% of the control group fell below the average person in the intervention group.

Effect sizes of 0.2, 0.5, 0.8 were defined as "small", "medium", "large" by Cohen.

http://www.leeds.ac.uk/educol/documents/00002182.htm

There weren't any controls were there?
'Effect' could just be regression to the mean here couldn't it?
I cannot see how any judgment about 'size of effect' can be made.

 

[user9876]

There weren't any controls were there?
'Effect' could just be regression to the mean here couldn't it?
I cannot see how any judgment about 'size of effect' can be made.

It may also depend on how the chose the standard deviation as I think that is not defined in the method.

 

[Sidereal]

Uncontrolled effect sizes were calculated for the CFQ, SF-36 physical functioning and the WSAS, using the within group Cohen's D [40]. The difference between the mean at pre- and post-treatment assessment was divided by a pooled standard deviation. Confidence intervals were calculated [41]. When post-treatment measurements were missing, but follow-up measurements were available, these were used. Followup was on average 3 months after post-assessment. To control for the use of different diagnostic criteria for CFS, effect sizes for the UK were also calculated for the subgroup of patients who met CDC criteria for CFS.

 

[Woolie]

Effect size is just the difference between the mean score before and after treatment, expressed as a standardised score. it means you can look at the value and get an instant idea of how big the effect was.

What @user9876 said was right:

Effect sizes of 0.2, 0.5, 0.8 were defined as "small", "medium", "large" by Cohen.

So the effect sizes in the Netherlands are massive.

I can't see the main paper, but there were no controls. So these are just measures of how much people who got treated improved on self reported fatigue, physical function and on a work and social adjustment questionnaire.

So you need to bear in mind that there could be huge "improvements" in people who didn't have CBT (there ceratinly were in the PACE control group). They may just be due to spontaneous improvement, or other stuff (like the mst severely affected people dropping out over time, which would make the average go up). You can't really make much of it without a control group.

It seems you don't need any of that annoying science stuff to publish in the Journal of Psychosomatic Research. I'll leave @Jonathan Edwards to chime in with his own theories there...

 

[Jonathan Edwards]

So the effect sizes in the Netherlands are massive.

But in the same sense that the difference between 25mm diameter waste pipes and inch diameter (25.4mm) waste pipes is 'massive', perhaps, because there is no overlap at all if the maximum variation is less than 0.1mm? The difference may truly be massive for a plumber because if he mixes them up the junctions will simply not fit. But for a child using the pipe as a didgeridoo the difference is indiscernible. Just damn lies and statistics it seems to me. How does one have a standard deviation of a pooled population one has decided, using this standard deviation, is not a single Gaussian distribution?

 

[Keith Geraghty]

Ive written two articles that speak about how subjective outcomes in CBT may come from a combination of therapy effect (someone listening), advice on stress management and sleep, and placebo --- rather than cognitive restructuring. I would really like to take my ideas forward with some kind of study to look more closely at what CBT actually involves and to ask patients about what it did for them. I am working on a proposal but think it would be hard to get funded, given the overwhelming populaity of CBT by NIHR (NHS) and so on.

What we need is some kind of trial of false CBT - this might be tricky to set up.

Jr of Psychosomatic Research does seem to be the 'go-to' journal for the PACE team.

 

[Woolie]

@Jonathan Edwards, I was looking forward to your theories on the Journal of Psychosomatic Research, actually. Always get a laugh at that!

The CI's are 95% confidence intervals. There are a relatively conservative measure of the uncertainty surrounding an estimate.The 95% CI represent the range (its upper and lower limits) in which the effect size would be likely to fall if you redid the study 100 times each time taking a new sample but applying the same methodology (actually 95 of the effect sizes should lie in that range).

They are always wide, because they aim to encompass almost the entire range where the effect size could be expected to lie. They are used in biology too, and they are large there too. Ignore these - they are not where the problem is.

 

[A.B.]

What we need is some kind of trial of false CBT - this might be tricky to set up.

Alternatively, real CBT for a problem that cannot improve with CBT, demonstrating how CBT can successfully create an illusion of improvement. Bonus points if the effect size is large. Extra bonus points if the CBT leads to people neglecting the problem due to having been brainwashed to view it as nonproblem or less serious than it is.

 

[Living Dead]

What we need is some kind of trial of false CBT - this might be tricky to set up.

My impression is that they are already using false CBT.

Original CBT for depression (which does not seem to be what is used for CFS) means to examine thoughts for logical errors, such as using "it always happens to me", when in fact it only happens 60% of the time.

Lecturing to patients that "you aren't sick, you are misinterpreting the signals from your body" doesn't really fit with the original description of CBT.

I don't know how much of either method is used in these studies, though.

 

[Woolie]

What we need is some kind of trial of false CBT - this might be tricky to set up.

@A.B.'s idea is nice. Another variation on this theme is to include an "organic" control (e.g. MS, lupus). If its CBT for CFS, the control group will probably exhibit similar improvements to the target group.

Trouble is, many CBT proponents will claim this as a win, because "every illness has a psychological component".

For some ideas for some good control conditions for CBT, see Lynch D, Laws KR and McKenna PJ. Cognitive behavioural therapy for major psychiatric disorder: does it really work? A meta-analytical review of well-controlled trials. Psychological Medicine 2010; 40: 9-24.

 

[A.B.]

Something like this maybe? Different interventions for asthma. The first graph shows self rated breathing, the second shows objectively measured breathing. Finding patients willing to believe that CBT could help with asthma might be difficult.

http://www.nejm.org/doi/full/10.1056/NEJMoa1103319#t=article

 

[Jonathan Edwards]

@Jonathan Edwards, I was looking forward to your theories on the Journal of Psychosomatic Research, actually. Always get a laugh at that!

The CI's are 95% confidence intervals. There are a relatively conservative measure of the uncertainty surrounding an estimate.The 95% CI represent the range (its upper and lower limits) in which the effect size would be likely to fall if you redid the study 100 times each time taking a new sample but applying the same methodology (actually 95 of the effect sizes should lie in that range).

They are always wide, because they aim to encompass almost the entire range where the effect size could be expected to lie. They are used in biology too, and they are large there too. Ignore these - they are not where the problem is.

Don't think I said anything about CI's actually! I agree that the problem is not to do with them.