CBT for CFS in Adults: Face to Face versus Telephone Treatment-A RCT (Burgess et al)

Dolphin

Senior Member
Messages
17,555
Likes
28,242
Cognitive Behaviour Therapy for Chronic Fatigue Syndrome in Adults: Face to Face versus Telephone Treatment - A Randomized Controlled Trial.

Behav Cogn Psychother. 2011 Sep 20:1-17. [Epub ahead of print]

Burgess M, Andiappan M, Chalder T.

Source
South London & Maudsley Trust, London, UK.

Abstract*

Background:

Previous research has shown that face to face cognitive behaviour therapy (CBT) is an effective treatment for chronic fatigue syndrome (CFS)/Myalgic Encephalomyelitis (ME).

However, some patients are unable to travel to the hospital for a number of reasons.

Aims:

The aim of this study was to assess whether face to face CBT was more effective than telephone CBT (with face to face assessment and discharge appointment) for patients with CFS.

Method:

Patients aged 18-65 were recruited from consecutive referrals to the Chronic Fatigue Syndrome (CFS) Research and Treatment Unit at The South London and Maudsley NHS Trust in London.

Participants were randomly allocated to either face to face CBT or telephone CBT by a departmental administrator.

Blinding of participants and care givers was inappropriate for this trial.

A parallel-groups randomised controlled trial was used to compare the two treatments.

The primary outcomes were physical functioning and fatigue.

Results:

Significant improvements in the primary outcomes of physical functioning and fatigue occurred and were maintained to one year follow-up after discharge from treatment.

Improvements in social adjustment and global outcome were noted and patient satisfaction was similar in both groups.

Conclusions:

Results from this study indicate that telephone CBT with two face to face appointments is a mild to moderately effective treatment for CFS and may be offered to patients where face to face treatment is not a viable option.

Despite these encouraging conclusions, dropout was relatively high and therapists should be aware of this potential problem.

PMID: 21929831 [PubMed - as supplied by publisher]
* I've given each sentence its own paragraph

Free full text on KCL site: http://www.kcl.ac.uk/innovation/groups/projects/cfs/publications/index.aspx
 

Dolphin

Senior Member
Messages
17,555
Likes
28,242
There were no objective outcome measures used.

Primary Outcomes used:
- Medical Outcomes Survey: short form (physical functioning subscale) (SF-36 PF)

- Chalder Fatigue Scale
Chalder Fatigue Scale (Chalder et al., 1993). This 11-item questionnaire measures the
severity of fatigue. Four response options range from less than usual to much more than
usual. Bimodal scoring gives a range of 011 and yields a cut-off for caseness or excessive
fatigue at 4 or over. This questionnaire has been used in previous CFS/ME intervention trials
and is reliable and valid (Cella and Chalder, 2010).
- they used bimodal scoring for this (i.e. 0-11). At baseline, the values for the two groups were very high, showing the problems with this method. Means: TCBT=10.41, CBT=10.06
They didn't give any figures for the numbers who weren't cases any more.

The data includes a lot of missing data. But generally they didn't carry forward missing values, instead presenting data for those who filled in the questionnaire.

Reading this, one might think TCBT didn't do as well as CBT:
Conclusions: Results from this study indicate that telephone CBT
with two face to face appointments is a mild to moderately effective treatment for CFS and
may be offered to patients where face to face treatment is not a viable option. Despite these
encouraging conclusions, dropout was relatively high and therapists should be aware of this
potential problem.
However, as the start of the discussion section says:
There were no significant differences on any primary outcomes between subjects in the two groups
 
Messages
1,383
Likes
362
Location
UK
To point out the obvious, there was no reference group (let alone a placebo-attention control) so we don't know how much they would have improved anyway - maybe neither mode of CBT worked? Did they have SMC as per PACE?
One could argue that it is difficult to accurately draw the conclusion that the interventions are effective as we did not find a difference between them and therefore change over time could be due to chance. However, evidence from previous trials, as already mentioned, have demonstrated the efficacy of face to face CBT for patients with CFS (Deale et al., 1997; Prins et al., 2001; Sharpe et al., 1996).
Both groups had smaller 12-month gains on SF36 than PACE (+19 from baseline), though they had been ill for about a year longer on average.
Telephone (with 55% drop outs apparently unaccounted for): +14
Face-to-face (34% drop outs): +12

Can't believe they are still pursuing the ME Association membership link as a predictor of outcome.
 
Messages
5,898
Likes
12,688
Location
South Australia
part of summary said:
It is interesting to note that participants who were working and did not have a job to return to were significantly more likely to drop out during treatment or followup than those who were not working. This issue is a complex one, potentially involving social reinforcers such as benefits (Bentall, Powell, Nye and Edwards, 2002) and significant others (Schmaling, Smith and Buchwald, 2000), making change on the patients part very difficult.
Blaming the patient never gets old!

Maybe those who lost their job had greater severity, had been out of work for longer etc.

Of course earlier, they said:
However, for physical functioning, participants with very low scores at baseline tended to drop out in the post treatment period; this suggests that job status and earlier values predicted the later dropout and hence in the random effects model, the analysis of which assumes only that the data are Missing At Random (MAR), these were included in the model to assess group differences.
The whole study is mostly meaningless though as there was no control group and no objective outcome measures.
 

Dolphin

Senior Member
Messages
17,555
Likes
28,242
A few more random observations (had a phone call which stopped me complete)

At present the two most widely accepted
operational criteria for CFS are the Oxford criteria (Sharpe et al., 1991) and the
Centres for Disease Control (CDC) Fukuda criteria (Fukuda et al., 1994).
Claiming the Oxford criteria are widely accepted isn't widely accepted.
But at least in this study, patients had to satisfy both.

A Cochrane review has shown that CBT improves fatigue
and physical functioning in about 40% of patients (Price, Mitchell, Tidy and Hunot,
2008).
This could give the impression that the treatment should get all the credit.
Controls also had significant improvements.

They also had to have had CFS for less than 10 years, as a longer
duration raised concerns over the accuracy of the diagnosis and the possibility that telephone
treatment would be inappropriate.
This is a new one on me.
I actually tend to have less confidence in the diagnosis at the start as sometimes people go on to be diagnosed with other conditions (either with more testing, new symptoms becoming apparent or whatever). After 10 years, I'd tend to be more confident of the diagnosis.
I think this may be an excuse for cherry-picking - the longer term patients might be harder to treat (they actually bring this point up later to an extent).

Patients were excluded from the trial
if they had any medical condition that may have accounted for their fatigue, had started
or changed medication within 3 months
to ensure stability of additional treatments, were
pregnant, had psychosis, drug abuse, a somatoform disorder or melancholic depression, a
subtype of major depression with specific features including anhedonia, severe weight loss,
psychomotor agitation or retardation, insomnia with early morning waking, and guilt.
(over-reaching?) While the bit in bold might be a better way of doing research, potentially it might bias towards people whose illness is more stable/less out of control/troublesome - it might bias towards better results than real life.

Once participants had consented and had been randomized, an initial detailed face to face
cognitive behavioural assessment took place with their assigned therapist. Participants in
the face to face group had their assessment split into two 1 1/2 hour appointments whereas
participants in the telephone group had one long appointment of up to 3 hours.
3 hours is quite a long appointment!
 

Esther12

Senior Member
Messages
13,774
Likes
28,356
The ten years things is odd. What new definition of CFS have they come up with?

My guess would be they'd try to spin it as 'more than 10 years and they must be really crazy. Those stubborn malcontents.'

Perhaps phrased more sensitively than that. I'm coming up for 10 years soon, whoop-dee-doo. What exciting quack-diagnosis will I get then? Fatigued Personality Type II?
 

Dolphin

Senior Member
Messages
17,555
Likes
28,242
More random observations

Self-rated global improvement was rated on a 6-item scale from very much better to very
much worse. Scores were dichotomized for the analysis by grouping together participants
who scored very much better and much better (1 or 2) with those who scored a little better
to very much worse (3 to 6).
This means one can't see how many people said they were worse - they've been merged together with "helped a little".
Not good "harms" reporting

Satisfaction with treatment was rated on a 7-point scale from
very satisfied to very dissatisfied. Scores were dichotomized for the analysis by grouping
together participants who were very satisfied or moderately satisfied (1 or 2) compared
respectively with those who were slightly satisfied to very dissatisfied (3 to 7).
---

I think it was you, Esther12, who pointed out to me that single item satisfaction questionnaires are problematic.
If anyone questions that, here's a reference:

Patient satisfaction with anaesthesia care: what is patient satisfaction, how should it be measured, and what is the evidence for assuring high patient satisfaction?

Best Pract Res Clin Anaesthesiol. 2006 Jun;20(2):331-46.

Heidegger T, Saal D, Nuebling M.
Source

Department of Anaesthesiology, Cantonal Hospital St Gallen, Rorschacherstrasse 95, 9007 St Gallen, Switzerland. thomas.heidegger@kssg.ch
Abstract

Patient satisfaction is a part of outcome quality. Many theories of satisfaction include patients' expectation. One definition of satisfaction is therefore the degree of congruence between expectation and accomplishment. The involvement of patients as well as experts is therefore an important step in the development of an instrument to measure patient satisfaction. Results of single-item ratings or overall satisfaction surveys are over-optimistic and do not represent the true indication of care. The construction of highly standardized (psychometric) questionnaires should include elements of content validity, criterion and construct validity, reliability and practicability. Based on the few available studies in anaesthesia, patient satisfaction is primarily determined by information and communication. There is great potential for improvement in this area. However, we do not know the best way to continuously improve patient satisfaction with anaesthesia care, or to what extent decisions should be shared between the anaesthetist and the patient.

Free at: http://201.245.103.106/archivos/sep 2011/ANESTESIA BASADA EN LA EVIDENCIA.pdf#page=103
---

(Not important at all)
Reasons for ineligibility of the 300 patients included: not meeting
Oxford and Fukuda criteria (37%), onset of over 10 years (20%), would not be able to attend
regular appointments, did not have a phone (14%) or were offered group CBT in the clinic (7%).
I'm not sure I recall them here or elsewhere saying why somebody would be offered group CBT in the same clinic instead? Anybody recall anything?

---
Baseline predictors of outcomes (same measure):

Fatigue: No
Also, there was no evidence for the predictive effect of baseline scores on later scores (p > .05).
(Given the other two measures are predictors, this may be because so many of the scores are likely to be 10 and 11)

Social adjustment scores (WSAS): Yes
As expected, the baseline social adjustment scores significantly (z = 6.57, p<.001) predicted the later scores at various time points.
Physical functioning: Yes
The baseline score significantly predicted (p<.001) the later physical function scores.
---
I don't understand how the following sentences fit together:
The considerable number of dropouts may question the findings of this study. However,
the post hoc power calculation showed a sample of 50 (25 in each group) had a mild to
moderate effect size of 0.35, which confirms that the results obtained are reliable. One may
have assumed that there would have been more dropouts from the face to face arm because of
the increased accessibility afforded by telephone sessions. However, this was not the case; a
variety of reasons were given by participants in both groups and accessibility was not one of
them.
I'm not sure I've seen talk of "post hoc power calculation" except perhaps in terms of what numbers they would have needed to get significance/what other studies might need.

I think it might be easiest to just read it that they're saying a Cohen's d value of d=0.35. They say this later:
The results from this study (0.35) suggest mild to moderate clinical effectiveness based on
estimations by Cohen (1988).
and the abstract says:
Results from this study indicate that telephone CBT with two face to face appointments is a mild to moderately effective treatment for CFS
which would fit in with a cohen's d value of 0.35.

I don't know what "post hoc" means in this context as I said e.g. did they just count people who completed and then made sure there was no extra weight to either group by averaging them out as 25 per intervention even if those weren't the final figures.

-----
Although a higher percentage of patients in the Deale et al. (1997) trial
reported feeling very much better or much better, satisfaction with treatment was similar in
both studies. It is important to consider why results from this trial were not as impressive as
our earlier RCT. Participants in this study had a longer duration of illness and were older,
factors that have been noted in other studies to have an association with symptom persistence
(Joyce, Hotopf and Wessely, 1997).
As I pointed out, they excluded people whose illness had lasted more than 10 years:
They also had to have had CFS for less than 10 years, as a longer
duration raised concerns over the accuracy of the diagnosis and the possibility that telephone
treatment would be inappropriate.
so this would have helped their results by this logic (but isn't mentioned in this way).

Later on they say:
One of our inclusion criteria was having CFS for less than 10 years as we
were concerned that a longer duration raised concerns over the accuracy of the diagnosis.
However, this may have meant that some patients were excluded who could have benefited.
A few paragraphs earlier, they said longer duration might lead to a worse outcome.
Now, they're saying the results could have been better if the people with a long illness duration (who they decided to exclude) the results might have been better!
(note: they don't go on to say, if they had been included the results might have been worse - they just say if they had been included, the results might have been better!)

---
It is of course possible that some of the participants who did not complete questionnaires had improved and
therefore decided to disengage with the service.
I don't know why they say this except that they are reaching.
It seems quite possible that some of the participants who did not complete questionnaires had disimproved and
therefore decided to disengage with the service.
 

Dolphin

Senior Member
Messages
17,555
Likes
28,242
The ten years things is odd. What new definition of CFS have they come up with?

My guess would be they'd try to spin it as 'more than 10 years and they must be really crazy. Those stubborn malcontents.' Perhaps phrased more sensitively than that.
Some form of that is my guess also - but they don't say it and it seems to me

I'm coming up for 10 years soon, whoop-dee-doo. What exciting quack-diagnosis will I get then? Fatigued Personality Type II?
Sorry to hear that. Coming up to 23 years myself.
 

biophile

Places I'd rather be.
Messages
8,977
Likes
16,579
Are the authors sloppy or lying?

Burgess & Andiappan & Chalder 2011 wrote: A Cochrane review has shown that CBT improves fatigue and physical functioning in about 40% of patients (Price, Mitchell, Tidy and Hunot, 2008).
This Cochrane (2008) review shows that 40% in the CBT group had a "clinical response" vs 26% in the usual care group. So CBT is apparently responsible for only 14% of patients in the CBT group to report improvement, not 40% (although one could argue that the remaining 26% in the CBT group may report greater improvement than the 26% in the usual care group, but Cochrane does not go into that and only reports group differences in fatigue of SMD -0.39, 95% CI -.0.60 to -0.19). Furthermore, measures of "clinical response" were just various fatigue scales, whereas differences in average physical functioning scores between the CBT group and usual care group were not significant at both post-treatment and short-term followup.

So what the Cochrane review actually suggests is, CBT may elicit a clinical response in fatigue for about 14% of patients, but without any significant improvement to physical function. Misrepresentation of the literature seems to occur frequently in biopsychosocial papers and shouldn't be allowed, I guess peer reviewers don't bother to do fact checking.

Burgess & Andiappan & Chalder 2011 wrote: We decided to omit a no treatment or placebo group because of the potential ethical objections of leaving patients untreated and the fact that we hypothesized that one approach, face to face CBT, would be superior to telephone CBT in reducing fatigue and improving physical functioning.
Oh please, "unethical" LOL, just admit you couldn't be bothered and/or lacked the resources and/or were too scared to compare the small improvements in your cherished CBT with natural course / usual care. Do I sound like a recalcitrant yet?
 
Messages
5,898
Likes
12,688
Location
South Australia
Burgess & Andiappan & Chalder 2011 wrote: We decided to omit a no treatment or placebo group because of the potential ethical objections of leaving patients untreated and the fact that we hypothesized that one approach, face to face CBT, would be superior to telephone CBT in reducing fatigue and improving physical functioning.
All of the language in the paper is speculative and reaching... Also, by misquoting the Cochrane review findings, they are insulting their readers.

Why do they push this double standard I wonder? If you want to prove pharmacological efficacy you must do so with a double blinded RCT. The equivalent to the Chalder study would be doing an open label study. Why are open label studies considered highly questionable, yet this study is not? Why do psychiatrists insist on such a double standard?
 
Messages
1,383
Likes
362
Location
UK
Agreed it's outrageous they should report the Cochrane fatigue finding so misleadingly, but then 40% sounds so much better than +14%. You have to question the quality of the peer review if either a) they are not familiar with the most important meta-analysis of CBT (the data even is in the abstract) or b) they are, and let the misrepresentation pass.

My recollection of Cochrane was that there was very little SF36 data (ie many studies didn't report on it) so the findings on that are rather circumspect, but I haven't reread it to check that.
 

biophile

Places I'd rather be.
Messages
8,977
Likes
16,579
Snow Leopard wrote: Also, by misquoting the Cochrane review findings, they are insulting their readers.
Indeed. Collins et al (inc. Crawley) did the same recently by citing the PACE Trial for this statement: "Evidence from a recent evidence trial of cognitive behavioural therapy and graded exercise therapy indicated a recovery rate of 30-40% one year after treatment." (http://www.biomedcentral.com/content/pdf/1472-6963-11-217.pdf)

Dubious definitions and misinterpretations of "normal" fatigue and "normal" physical function and positive clinical global impression aside, the same logic puts the SMC (standard medical care) de facto control group in the PACE Trial at a 15-25% "recovery" rate, leaving a true "recovery" rate (*cough* if you're aiming for the level of an 80 year old *cough*) of about 15% for CBT/GET, not 30-40%. Collins et al do not mention this and give the false impression that the 30-40% recovery rate is due to CBT/GET, just as Chalder et al did with Cochrane.

Oh well I guess that is their modus operandi? It irritates me that we are the supposed irrational extremists for having doubts about the cognitive behavioural approach while it is apparently OK for CBT/GET proponents to give misleading statements in medical journals. These aren't regular journalists here, these people are supposed to be world class (sic) experts in their field.

Snow Leopard wrote: If you want to prove pharmacological efficacy you must do so with a double blinded RCT. The equivalent to the Chalder study would be doing an open label study. Why are open label studies considered highly questionable, yet this study is not? Why do psychiatrists insist on such a double standard?
Yeah! They seem to think that because it is difficult to do such a study, then somehow that negates the problem.

oceanblue wrote: Agreed it's outrageous they should report the Cochrane fatigue finding so misleadingly, but then 40% sounds so much better than +14%. You have to question the quality of the peer review if either a) they are not familiar with the most important meta-analysis of CBT (the data even is in the abstract) or b) they are, and let the misrepresentation pass.
This seems to happen often, so either peer review in general is quite sloppy, or they are quite loose with interpretations of accuracy and let it slide, or they are purposely deceiving us. Not entirely related, but I have also wondered before, if these people believe CFS is mainly perpetuated by abnormal illness beliefs, then little white lies may be deemed acceptable to increase chances of "success" in CBT/GET.

oceanblue wrote: My recollection of Cochrane was that there was very little SF36 data (ie many studies didn't report on it) so the findings on that are rather circumspect, but I haven't reread it to check that.
Not as much data as for the fatigue measures, but still some. The physical function at post-treatment was derived from 4 studies (318 participants) and PF at short/medium-term followup was derived from 3 studies (275 participants). The more recent PACE results may be enough to swing the average into significance.
 
Messages
1,383
Likes
362
Location
UK
This seems to happen often, so either peer review in general is quite sloppy, or they are quite lose with interpretations of accuracy and let it slide, or they are purposely deceiving us. Not entirely related, but I have also wondered before, if these people believe CFS is mainly perpetuated by abnormal illness beliefs, then little white lies may be deemed acceptable to increase chances of "success" in CBT/GET.
Like the argument that they are doing it for our own good - it is just possible they actually believe that!

Not as much data as for the fatigue measures, but still some. The physical function at post-treatment was derived from 4 studies (318 participants) and PF at short-term followup was derived from 3 studies (275 participants). The more recent PACE results may be enough to swing the average into significance.
Thanks for digging that out. Given that PACE used a longer follow-up, and was generally a better-designed study (really), I tend to think the PACE results trump the limited meta-analysis. Unless the Cochrane SF36 studies all used proper attention-controls eg relaxation therapy.
 

alex3619

Senior Member
Messages
13,810
Likes
37,651
Location
Logan, Queensland, Australia
Oh well I guess that is their modus operandi? It irritates me that we are the supposed irrational extremists for having doubts about the cognitive behavioural approach while it is apparently OK for CBT/GET proponents to give misleading statements in medical journals. These aren't regular journalists here, these people are supposed to be world class (sic) experts in their field.
Hi biophile, I have been saying this is indeed their modus operandi for some time. Misquote, mistate, redefine then vacilate on their definitions - its about spin over substance. The PACE study is full of it. I was going to write a blog on confimation bias in ME research, then I thought no, an article or series of blogs on verificationism (institutionalized confirmation approach to science) and now I am wondering if my brain and resources are up to writing a book. This entire branch of medicine is an unsubstantiated theory - all of it. The biopsychosocial approach to ME is just an excellent exemplar of the field - full of deceptive claims, circular reasoning, deliberate ignoring of data that falsifies their theories, redefinition to solve issues as if redefining them changes anything much. Even with all their changes to established standards, all their false claims, they still fail to verify their models. I wonder if its not time to say the null hypothesis has been verified - psychosomatic medicine is disproved? I hope to look into that.

Verificationism has some merits, some value, but only under certain circumstances. Those circumstances do not exist in ME research, and I think I might be able to show almost everything they are doing violates basic scientific standards and ethics. I am looking into this.

Bye, Alex
 

biophile

Places I'd rather be.
Messages
8,977
Likes
16,579
onceablue wrote: Thanks for digging that out. Given that PACE used a longer follow-up, and was generally a better-designed study (really), I tend to think the PACE results trump the limited meta-analysis. Unless the Cochrane SF36 studies all used proper attention-controls eg relaxation therapy.
I had another quick look (http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD001027.pub2/pdf). The characteristics of control groups in studies which report physical function (either Karnofsky scale or PF/SF-36) seem mixed. Physical function was assessed at medium term followup too (I corrected my post above), here are the temporal definitions: "Outcomes were classified as post treatment, short term followup (1-6 months post-treatment), medium term follow-up (7-12 months post-treatment) and long term (longer than 12 months)."

Also I forgot to mention that while the 40% vs 26% clinical response figure is for post-treatment with fatigue as a primary measure, the difference becomes non-significant at short term follow up (3 studies, 353 participants). So in other words the effect seems brief, you couldn't ever tell that from Chalder's statement on the Cochrane review.

This is what the authors had to say:

"The benefits of CBT in sustaining clinical response and reduction of fatigue symptoms at short and medium term follow-up are inconclusive. The benefits of CBT in improving physical functioning and reducing depression, anxiety and psychological distress at post treatment and at follow-up are also uncertain. [...] Findings at follow-up were heterogeneous and inconsistent. [...] At follow-up, 1-7 months after treatment ended, people who had completed their course of CBT continued to have lower fatigue levels, but when including people who had dropped out of treatment, there was no difference between CBT and usual care."
The authors view the conclusions as "very preliminary findings" and state that there is a "surprising lack of high quality evidence" and note the unknown degree of response bias and poor reporting of adverse effects. Cochrane actually shows that the state of evidence for CBT in 2008 was very limited and relatively weak. Compare that to proponents who viewed CBT in 2008 as a wonderful and established treatment with mountains of unquestionable evidence. You're probably right that PACE was generally better conducted than these studies.

alex3619 wrote: Hi biophile, I have been saying this is indeed their modus operandi for some time. Misquote, mistate, redefine then vacilate on their definitions - its about spin over substance. The PACE study is full of it. I was going to write a blog on confimation bias in ME research, then I thought no, an article or series of blogs on verificationism (institutionalized confirmation approach to science) and now I am wondering if my brain and resources are up to writing a book. This entire branch of medicine is an unsubstantiated theory - all of it. The biopsychosocial approach to ME is just an excellent exemplar of the field - full of deceptive claims, circular reasoning, deliberate ignoring of data that falsifies their theories, redefinition to solve issues as if redefining them changes anything much. Even with all their changes to established standards, all their false claims, they still fail to verify their models. I wonder if its not time to say the null hypothesis has been verified - psychosomatic medicine is disproved? I hope to look into that. Verificationism has some merits, some value, but only under certain circumstances. Those circumstances do not exist in ME research, and I think I might be able to show almost everything they are doing violates basic scientific standards and ethics. I am looking into this. Bye, Alex
Hi Alex, I've come across many such things when reviewing the literature, so I think you are onto something important and I hope you manage to write something on it, I would be interested in reading it. The evidence doesn't support their cognitive behavioural hypothesis of CFS as a primary explanation for the symptoms and disability of ME, at most it may have limited explanatory power in a minority of patients. Even that may be generous as the nature of the correlations are unclear. I have come to see the gyroscope as a symbol of the cognitive behavioural approach to ME/CFS because it has kept itself upright due to ongoing spin and would soon crash without it.
 
Messages
1,383
Likes
362
Location
UK
I had another quick look (http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD001027.pub2/pdf). The characteristics of control groups in studies which report physical function (either Karnofsky scale or PF/SF-36) seem mixed. Physical function was assessed at medium term followup too (I corrected my post above)...

Also I forgot to mention that while the 40% vs 26% clinical response figure is for post-treatment with fatigue as a primary measure, the difference becomes non-significant at short term follow up (3 studies, 353 participants). So in other words the effect seems brief, you couldn't ever tell that from Chalder's statement on the Cochrane review.
...At follow-up, 1-7 months after treatment ended, people who had completed their course of CBT continued to have lower fatigue levels, but when including people who had dropped out of treatment, there was no difference between CBT and usual care."
The authors view the conclusions as "very preliminary findings" and state that there is a "surprising lack of high quality evidence" and note the unknown degree of response bias and poor reporting of adverse effects. Cochrane actually shows that the state of evidence for CBT in 2008 was very limited and relatively weak. Compare that to proponents who viewed CBT in 2008 as a wonderful and established treatment with mountains of unquestionable evidence. You're probably right that PACE was generally better conducted than these studies.
Thanks, that's very helpful, and more damning than I remembered. So, that really leaves us with PACE for evidence on CBT (&GET). It's a busted flush, basically.