The draft scope for the NICE guideline on ME/CFS is now out for consultation, June 2018

[Sasha]

Dear Adrian, when I obtained the raw data from the PACE Trial I wanted to analyse it in my own way without external influences directing my attention one way or another

I can understand wanting to come at the data with no preconceptions but having done so, there's every reason to look at what others have done in order to see if someone has already reported the same findings; has mentioned such findings but has critiqued them with arguments that you haven't thought of; or did other analyses that make your findings irrelevant or cast them in a different light.

If you don't do that, you have no way of knowing whether you're adding to the literature or muddying the waters.

 

[Graham]

@Peter Kemp , thanks for the graphs: I love looking at the way different analyses can be presented. What puzzles me though is what your horizontal axis represents. You'll find that in our "evaluatingpace" webpages we analysed the changes in fatigue levels in a different way.

 

[Peter Kemp]

Dear Adrian, did you look at the charts? They represent actual data, not how it was interpreted or reported, but how it actually is. It has nothing to do with reporting bias. I am disappointed (and in fact find it unlikely) that anyone involved in campaigning would be unaware of the repeatedly expounded problems caused by including 'fatigue' patients in ME/CFS research. In fact I consider it so unlikely that I think your question is vexatious. The charts show what I would expect to see in a mixed cohort of fatigue and bona fide M.E. patients. As I have already stated, there is no data to correlate with the alternative criteria for M.E. and Reeves et al CFS so these cannot be verified.

The argument many here are making is that such results are due to reporting bias rather than selection bias. This is supported by comparisons when looking at objective vs subjective outcomes. Also if you look at the correlation between changes in the subjective scores vs the 6mwt it is different between the CBT/GET and APT/SMC groups.

I can believe that getting some medical help will lead to symptom improvement (for example, some decent pain medication).

I don't understand your claims for selection bias. At the heart of this claim there is an assumption that CBT/GET will help some with fatigue but not others and then you need to give evidence for this split. What was it in the selection process that supported bias, what are the possible different mechanisms. Why did the figures for Oxford+London, Oxford+CDC, Oxford show roughly the same improvements.

 

[adambeyoncelowe]

Dear Adrian, did you look at the charts? They represent actual data, not how it was interpreted or reported, but how it actually is. It has nothing to do with reporting bias. I am disappointed (and in fact find it unlikely) that anyone involved in campaigning would be unaware of the repeatedly expounded problems caused by including 'fatigue' patients in ME/CFS research. In fact I consider it so unlikely that I think your question is vexatious. The charts show what I would expect to see in a mixed cohort of fatigue and bona fide M.E. patients. As I have already stated, there is no data to correlate with the alternative criteria for M.E. and Reeves et al CFS so these cannot be verified.

'Vexatious' is a rather politically charged term in this context. This is the second time in this thread you've accused others of having sinister motives. Might I suggest that you consider such feedback constructively, rather than as an attack on you?

Your numbers are certainly very interesting, but I think comparing your results with those of others would only make them stronger, not weaker.

 

[Trish]

Hi @Peter, I think I understand some of what you are saying.

It is clearly a problem in trials that purport to be of treatments for, or indeed biomedical studies of, ME patients, if a whole lot of patients with other fatiguing conditions are included. We all agree with that. PACE, if it had said it was studying ME as defined by PEM as the core symptom, would have been junk on the grounds that some patients did not have PEM.

However, PACE made it clear that what they were studying was people who fulfilled the Oxford criteria of 6 months disabling fatigue. They were not pretending it was a study of people with PEM, though it probably included some people with this core ME symptom. If I understand it correctly, the USA decisions on whether to include trials in their review of treatments excluded those that used Oxford criteria as not applying to ME/CFS.

The reanalysis using pre-trial approved criteria and long term follow up show that there was no clinically significant, and almost no statistically significant between group differences. The conclusion from this should be that for all people with chronic disabling fatigue of at least 6 months duration, CBT and GET are not effective treatments and should not be recommended, whether they have ME or not.

The trial fell down on many fronts, the key one being that subjective outcome measures in unblinded trials are not acceptable for all sorts of reasons.

What your data show is something well known - namely that all groups showed an improvement over the year of the trial on average on these 2 subjective measures. This could be partly because a lot of them had a temporary fatiguing condition that improved or resolved with time, or it could be that just getting the help with pain and sleep from the SMC provided for all groups helped to improve their subjective experience of their condition. But in any medical trail, it is the between group differences that matter, not the overall effect on all patients in the trial over time. So I don't understand how your analysis of this provides any new insights.

 

[Jonathan Edwards]

Dear @Peter Kemp,

I agree with a lot of what you say but these are issues we have discussed at length on the forums and I think are pretty much covered by what people have published. As Sasha says, if you want to publish the you need to cite other people who have published on the topic. There are lots of people with ME here who have been working on this as long as you have so it is hard to see why it should be difficult to find people to collaborate with.

I agree that it is likely that the recruitment to PACE could have favoured selecting people without typical ME that might have been more likely to improve. But since the improvement you mention occurs in all groups it cannot be evidence of being more likely to benefit from CBT and GET, but rather a likelihood of improving over the trial period in some sort of therapeutic context. The SMC arm subjects got pretty little as far as I can tell so the prima facie evidence is that they selected people who were rather likely to improve spontaneously.

Moreover, I am not yet convinced that the improvements are unusual for a trial like this. There are several reasons why people tend to improve in trials irrespective of treatments. One is regression to the mean for a relapsing remitting condition - where people get recruited at their worst and then return to average. Another is the placebo effect - which may not have operated much in the SMC group so might not fit. Then there is the 'reverse placebo' effect of people saying they are better when placed in the social role of being a therapy beneficiary. (I have recently emphasised this in my submission to the Scottish Parliament that they have published online.) There are probably other mechanisms. Major improvements across trials are common enough, especially if the trials are unblinded.

You are clearly wanting to hang your argument on the idea that evidence from PACE is weak because the 'wrong patients' were recruited. That is possible but I don't think we can prove it. Moreover, in real life nobody is a 'wrong patient'. Everyone deserves to be in a trial. For me the more persuasive argument is that as unblinded trials with subjective outcomes none of these trials of therapist-delivered treatments produce any valid data. There is no reliable evidence that anything in any of the treatment arms makes a real difference to any group of people.

We also have a problem in arguing in the context of NICE that maybe people with 'other fatigue' might benefit from therapist-delivered treatments. That is that NICE has to cover these people, for pragmatic reasons. So we end up with a suggestion that some groups might have CBT and GET and others not. That to me is the worst case situation because it takes us back to where we are now. In theory people for whom GET is unsuitable should not be having it even now. But they are. If the service providers are not that convinced that there are clear distinctions between patient groups we slip back into BPS mindset.

It is a pity that you have not been in on the debate longer because you have clearly thought about this a lot.

 

[Adrian]

Dear Adrian, did you look at the charts? They represent actual data, not how it was interpreted or reported, but how it actually is.

With any data you have and look at you need to carefully examine the sensors used to generate the data along with features in the collection. In this case the data is self reported fatigue and physical function which in itself is a function of a choice on what to report of perceived fatigue and perceived physical function which is assumed to be some function of actual fatigue (if this were a single quantity). This presents opportunities for reporting bias both to affect what people report as they like the therapist or feel pressured to support the given treatment as well as reporting biases in the way patients perceptions are changed (this is the aim of CBT and GET so this is a very likely and potentially large reporting bias).

Thus we can only conclude the data is dodgy and that what is measured does not represent actual fatigue and physical function and that the reporting mechanisms are affected by the trial.

It has nothing to do with reporting bias. I am disappointed (and in fact find it unlikely) that anyone involved in campaigning would be unaware of the repeatedly expounded problems caused by including 'fatigue' patients in ME/CFS research. In fact I consider it so unlikely that I think your question is vexatious. The charts show what I would expect to see in a mixed cohort of fatigue and bona fide M.E. patients. As I have already stated, there is no data to correlate with the alternative criteria for M.E. and Reeves et al CFS so these cannot be verifie

I know people talk about 'fatigue' patients but it doesn't make sense to my mind. The implication of what you are saying is that there are 'real ME' patients who CBT/GET don't work with but other CF patients that it works on. From this assumption I could only conclude that CBT and GET should be offered to patients with fatigue (but once other known conditions have been filtered out) but not those who have 'real ME'. This leads to the question of what is 'real ME' as opposed to CF - yes there are definitions and PEM may be an indicator but definitions are arbitrary or best guesses they are not based on any understanding of mechanism.

It is unclear to me that there are really two different diseases 'real ME' and CF and I have never seen any solid evidence to back up the theory that there is.

 

[Barry]

When the PACE trial arms are assessed as groups, a tiny difference is found for GET and CBT.

That is the nub of it surely. PACE purported to show them effective, and instead proved them ineffective. NICE CG53 has been perpetuated based on the former, and the new guideline must take much more heed of the latter. Do we need to confuse things beyond this.

 

[Trish]

@Peter Kemp , thanks for the graphs: I love looking at the way different analyses can be presented. What puzzles me though is what your horizontal axis represents. You'll find that in our "evaluatingpace" webpages we analysed the changes in fatigue levels in a different way.

Looks to me like the horizontal axis represents cumulative frequency, which would, I think more commonly be on the vertical axis.

 

[Barry]

They represent actual data, not how it was interpreted or reported, but how it actually is. It has nothing to do with reporting bias.

I don't understand this. The "actual data" is in fact data that was acquired from self-reports, so any reporting bias will inescapably be intrinsic to it. "How it actually is" is simply how it was actually reported, incorporating any reporting bias that affected that data.

 

[Peter Kemp]

Dear Sasha, I accept what you say, the problem is finding the time and energy. I had no intention of contributing to a substantial stakeholder response to NICE. I never even meant to write one for the Lyme guideline or all the BMJ rapid responses I have written since. But the situation is just too bad to ignore. I have had M.E. for over 40 years and for the past 3-4 years have just wanted to stop and have some peace. But it is for kids, like Graham's son, that I keep going. I was lucky, I was 21 years old before my life as I knew it came to an end. Yes, 'lucky'. I had some fun, got some memories and even all these years later they give me something good to look back upon. I breaks my heart when I hear about children that get M.E. or Lyme disease because I have heard of too many that have been ill right through their childhood and into adulthood up to the present. This cannot be allowed to go on. Parents do not bring children into the world to suffer - God knows how they cope.

A few weeks ago I visited a very dear friend on her 87th birthday. Since she retired and up until 2 years ago, she was a 24/7 carer for her adult daughter with M.E. Then my friend's whole body just packed-up and she is now confined to bed downstairs, whilst her daughter is confined to bed upstairs - where she has been stuck for 17 years now. Whilst I was with my friend, one of the carers came in with a message from the daughter, 'Do you know if there are any new treatments, please let me know?' I am sick to my heart with that question because there was nothing I could say except that I would have a look around and let her know. The only tiny consolation is that like me, she got a childhood and young adulthood, knows what it is like to have a job-career, romance, holidays hopes and dreams that could come true... So it is really for kids that I am motivated to keep going beyond my limits and keep doing what I can, even though illness has deprived me of every having any of my own. It is for them and what lies ahead for them, that gives me the strength to withstand the strain and indeed the ignorant and frustrated people that the battered remnants of a fighting spirit always seems to attract (or perhaps it's just me).

So excuse that I have not done everything that perhaps I should and excuse the moan.

Best Wishes,
Peter

I can understand wanting to come at the data with no preconceptions but having done so, there's every reason to look at what others have done in order to see if someone has already reported the same findings; has mentioned such findings but has critiqued them with arguments that you haven't thought of; or did other analyses that make your findings irrelevant or cast them in a different light.

If you don't do that, you have no way of knowing whether you're adding to the literature or muddying the waters.

 

[Peter Kemp]

Hi Trish and Graham, the horizontal of the chart is the count of individual participants. As I remarked, each diamond represents an individual participant. I think what made it confusing is that all the participants were sorted left to right with those with the lowest amount of change on the left and highest on the right. If you just look at one diamond, its colour tells you which trial arm that participant was in, then look at the left-hand scale, that tells you how much their ratings changed with the outcome measure. You will see that there were around 15% who deteriorated, a whole lot who improved a bit, and around 10 to 15% who improved significantly (some of these, spectacularly). You can also see the small difference between CBT and GET compared to APT and CBT.

Looks to me like the horizontal axis represents cumulative frequency, which would, I think more commonly be on the vertical axis.

 

[Sasha]

Dear Sasha, I accept what you say, the problem is finding the time and energy.

I can understand being short on time and energy. I hope that some of the comments on this thread will be helpful.

(Sorry for my short reply to your long post - having a bit of a rough day!)

 

[Peter Kemp]

Maybe, maybe not. If it is accepted as given that the PACE trial disproves GET and CBT all well and good. I do not get the impression that it is accepted, but please correct me if I am wrong. As I have explained, if the PACE trial was withdrawn completely, it would not guarantee that GET and CBT will not be in the next NICE guideline, and in fact NICE might be obliged to include them. It is all in my earlier post.

That is the nub of it surely. PACE purported to show them effective, and instead proved them ineffective. NICE CG53 has been perpetuated based on the former, and the new guideline must take much more heed of the latter. Do we need to confuse things beyond this.

 

[Trish]

Dear @Peter Kemp, thank you for sharing your personal experience with ME. From one long term ME sufferer to another, I share your pain. I'm pleased you have joined us here - I hope you will stick around and be part of our community.

As you have seen, we have lots of other people here like you who want to make a difference, and who have found that working together and learning from each other can make us make that difference more effectively. I only discovered ME forums a couple of years ago and it has changed my life immeasurably for the better.

It can be hard to hear from others who have been studying the PACE trial in great detail that our attempts at analysis of the data need to be examined carefully and shared before we can be sure they are helpful. I admit to having a go at my own analysis too (I was a Maths teacher), but I soon found that there were others among us with much more experience of both clinical trials and statistical analysis than me, and I was happy to leave the detailed analysis to them. I found both the special issue of the Journal of Health Psychology on the PACE trial, and the Wilshire et al reanalysis paper very thorough and enlightening. Much better than I could have done. Do read them - they really are excellent.

Thanks for the explanation of the graphs. I did get that. So the scale along the bottom represents the number of people with that score difference or less, otherwise known as the cumulative frequency.

 

[Peter Kemp]

Adrian, I agree with most of what you write and have written along these same lines myself. But your conclusion is not completely sound to me because it implies that subjective measures are unreliable per se. There are some huge discrepancies between the POM's and the walk test data, and other discrepancies like participant's CFQ and SF-36 scores contradicting each other. How much of these and other changes can be put down to placebo, optimism, deference, gratitude etc, I am not sure about. As I already mentioned, nearly half GET/CBT participants rated improvement which meets the 'minimum detectable change' thresholds for the measures. These ratings are from one year, and months after their therapy sessions had finished. I guess, weeks or months after most had met their therapist. In my experience deference does not hang around very long - especially for clients who did not gain any practical benefit from a therapy. So to me, whilst bias can explain the level of improvement for a lot of participants, it is inadequate to explain all changes.

Those registering significant improvement cannot be ignored. This is what I have got from looking at individual data - which I expect is what you have done also. Some of the changes would be nothing short of spectacular if those participants were bona fide M.E. or CFS patients. As I remarked in my bit about NNT's, 7 is quite good. But if the authors had settled for 15 or 20 they could produce astounding figures for improvement, with only the control group to rescue the situation for patients. So I would say that the control group is vital to our case, but so is selection bias.

Frankly, I am rather stunned to find that an admin's approach is that 'chronic fatigue' is a legitimate selection criteria for ME/CFS research. Perhaps I misunderstood your remarks on this. If this represents the group's position, I'm going back to Kansas.

With any data you have and look at you need to carefully examine the sensors used to generate the data along with features in the collection. In this case the data is self reported fatigue and physical function which in itself is a function of a choice on what to report of perceived fatigue and perceived physical function which is assumed to be some function of actual fatigue (if this were a single quantity). This presents opportunities for reporting bias both to affect what people report as they like the therapist or feel pressured to support the given treatment as well as reporting biases in the way patients perceptions are changed (this is the aim of CBT and GET so this is a very likely and potentially large reporting bias).

Thus we can only conclude the data is dodgy and that what is measured does not represent actual fatigue and physical function and that the reporting mechanisms are affected by the trial.

 

[Adrian]

Frankly, I am rather stunned to find that an admin's approach is that 'chronic fatigue' is a legitimate selection criteria for ME/CFS research. Perhaps I misunderstood your remarks on this. If this represents the group's position, I'm going back to Kansas.

This group doesn't have a position and I would certainly not view my position as the position of the forum. One function of the forum is to debate the evidence in detail and hence gain a better understanding (by critiquing different positions).

In terms of selection criteria. If you take something like PACE they had a broad selection criteria and claim via a subgroup analysis that benefits apply to various subgroups. Your analysis suggests that trials should be run on the basis of chronic fatigue symptoms because you claim that there are groups who will benefit from CBT and GET. You say these aren't people with ME but if there are other groups who just have CF then don't they also deserve research?

My position is that the PACE results don't show any obvious benefit and the methodology is poor and subjective measures unreliable. Hence I would say there is no obvious other group that would benefit and that PACE has failed to show anyone benefits from CBT or GET.

I think there are issues around potential selection bias particularly around harms but that is not the most important issue with PACE or other CBT/GET trials.

In my experience deference does not hang around very long - especially for clients who did not gain any practical benefit from a therapy. So to me, whilst bias can explain the level of improvement for a lot of participants, it is inadequate to explain all changes.

I guess that is where we differ. I don't think it is around deference, I think it is around how abilities are interpreted. If you repeatably tell people they will get better if they ignore symptoms and believe they will can do stuff then they are likely to move the threshold between 'limited a lot' and 'limited a little'. Something like the Sf36 is not well designed so with middle values a small movement in ability (or perceived ability, threshold change) can lead to quite large changes because of the way the questions are clustered. Add expectations that therapists push onto patients and I think it is easy to see why the improvements are marked.

Its important not to be taken in by the abstraction of the questionnaires as scales but go back to bits and bytes and say what is actually happening, what is being measured rather than accept their (and others) stories.

Given time (as in the Long term follow up) we see no differences between the groups which suggests the effect does wear off. I suspect that they knew that the effect wears off because they set up booster sessions before one of the measurement points and some time after the rest of the sessions.

Those registering significant improvement cannot be ignored.

From what I remember of the 6mwt data improvements were fairly evenly distributed with the biggest improvement (and the only one approaching any form of recovery) being within the APT group.

The problem is if the measurement system you are using measures too many factors you simply cannot associate a change to one particular factor. Where there is ambiguity you cannot make solid conclusions as the PACE team do.

I think the literature looking at reporting bias does reflect the types of sizes seen in PACE and this is one of the points raised by Wilshire in her JHP paper.

 

[Trish]

Frankly, I am rather stunned to find that an admin's approach is that 'chronic fatigue' is a legitimate selection criteria for ME/CFS research. Perhaps I misunderstood your remarks on this. If this represents the group's position, I'm going back to Kansas.

I think you have misunderstood, Peter. None of us think that fatigue studies are suitable for ME/CFS when defined properly to include PEM. International ME experts (apart from the BPS lot) are clear that the Oxford definition should be retired as a research definition.

But PACE, FINE and other trials that use this definition exist and are going to be considered by NICE. We therefore need to make all the arguments possible to demonstrate that these trials do not show the treatments are effective for anybody.

As @Jonathan Edwards says, the NICE guidelines are there to provide information on diagnosis and treatment that doctors in the UK should use.
Whether people with chronic disabling fatigue have ME or not, the answer is the same - CBT and GET don't work and shouldn't be offered to anybody.

 

[Rick Sanchez]

Adrian, I agree with most of what you write and have written along these same lines myself. But your conclusion is not completely sound to me because it implies that subjective measures are unreliable per se. There are some huge discrepancies between the POM's and the walk test data, and other discrepancies like participant's CFQ and SF-36 scores contradicting each other. How much of these and other changes can be put down to placebo, optimism, deference, gratitude etc, I am not sure about. As I already mentioned, nearly half GET/CBT participants rated improvement which meets the 'minimum detectable change' thresholds for the measures. These ratings are from one year, and months after their therapy sessions had finished. I guess, weeks or months after most had met their therapist. In my experience deference does not hang around very long - especially for clients who did not gain any practical benefit from a therapy. So to me, whilst bias can explain the level of improvement for a lot of participants, it is inadequate to explain all changes.

Those registering significant improvement cannot be ignored. This is what I have got from looking at individual data - which I expect is what you have done also. Some of the changes would be nothing short of spectacular if those participants were bona fide M.E. or CFS patients. As I remarked in my bit about NNT's, 7 is quite good. But if the authors had settled for 15 or 20 they could produce astounding figures for improvement, with only the control group to rescue the situation for patients. So I would say that the control group is vital to our case, but so is selection bias.

Frankly, I am rather stunned to find that an admin's approach is that 'chronic fatigue' is a legitimate selection criteria for ME/CFS research. Perhaps I misunderstood your remarks on this. If this represents the group's position, I'm going back to Kansas.

The complaint is not that subjective measures are unreliable per se, but that the combination of subjective measures and an unblinded trial are problematic.

Regarding significant improvement in PACE. The general improvement in the PACE trial for CBT / GET compared to the other arms was insignificant enough, that if the authors had just followed their original protocol they would have had a null result. I do not consider this to be a significant improvement.

Secondly. In addition to the problem of subjective measures and an unblinded trial. You don't only have to take the placebo and the therapist effect into account. The CBT and GET manuals explicitly stated that it was essential that patients were told to ignore negative symptoms, that patients should be told how effective the treatments given were, and how strong the scientific evidence for treatment was. The CBT and GET treatments are basically built to change the patients perception about their own health and illness. To me, this makes the exclusive use of subjective measurements as primary outcome measures seem like insanity. one of the things that surprises me the most about the results of the PACE trial is that the improvements in self reported outcomes weren't in fact higher in the CBT / GET arms than they were. Something that probably surprised the authors themselves.

 

[Peter Kemp]

Dear Jonathan,

Thank you for your remarks, I also agree with a lot of what yourself and others have said on this topic, though I find it interesting that I seem to be coming from a perspective where our views do not automatically meet.

I assure you that I am not hanging on to an argument, to me, the selection problem is actually a side-issue (which I happily ignored for years) compared to the numerous issues which show that the PACE trial report is unreliable and that GET and CBT are not treatments.

However, patient selection could become an issue and forewarned is forearmed as they say, but it has not worked here. Not only have my warnings gone unheeded, some people seem to be resistant to the idea. Perhaps this is because they believe that participant selection does not and will not matter and that might be true - I hope so.

I have only just learned that this group considers that 'fatigue' patients are legitimate participants in M.E. and CFS research, and that it is prepared to accept 'fatigue' patients in a NICE guideline for 'ME/CFS' patients. This is not something we are going to agree on. It is the VIRAS position that M.E. and critically defined CFS are distinct from chronic fatigue and that they can and should be differentiated for the purposes of diagnosis and treatment/management.

I think I am correct in thinking that yourself and your colleagues are tightly focussed on eliminating GET and CBT from the NICE guideline. That aim VIRAS certainly share, but not to the extent of watering-down diagnosis. The current NICE guideline at least limits who should be offered GET or CBT and NICE have listened to complaints of coercion. The current guideline also has PEM as an obligatory symptom for diagnosis - correctly IMO. If 'fatigue' patients are accommodated in the guideline there is a risk that PEM will be eliminated and only fatigue of a particular severity and duration will be required for diagnosis.

For reasons which I have set out, I am concerned that your group are placing great hopes in the influence on NICE of discrediting the PACE Trial. But I have my doubts about just how far that will go towards eliminating GET and CBT as 'treatments' for ME/CFS. Someone did recognise (sorry cannot remember who) using the PACE Trial data to prove that GET and CBT are not treatments - and therefore the theories behind them are wrong. That would be great. But there is also a risk that the PACE Trial was such an almighty cock-up that the data cannot be used to prove the opposite of what the authors claimed.

You seem to suggest that yourself, Jonathan: "the more persuasive argument is that as unblinded trials with subjective outcomes none of these trials of therapist-delivered treatments produce any valid data". If you are correct, then it would seem that the PACE Trial data cannot be used to disprove the author's theories.

Damned complicated stuff.

Good Luck with your goals,
Peter

Dear @Peter Kemp,

I agree with a lot of what you say but these are issues we have discussed at length on the forums and I think are pretty much covered by what people have published. As Sasha says, if you want to publish the you need to cite other people who have published on the topic. There are lots of people with ME here who have been working on this as long as you have so it is hard to see why it should be difficult to find people to collaborate with.

I agree that it is likely that the recruitment to PACE could have favoured selecting people without typical ME that might have been more likely to improve. But since the improvement you mention occurs in all groups it cannot be evidence of being more likely to benefit from CBT and GET, but rather a likelihood of improving over the trial period in some sort of therapeutic context. The SMC arm subjects got pretty little as far as I can tell so the prima facie evidence is that they selected people who were rather likely to improve spontaneously.

Moreover, I am not yet convinced that the improvements are unusual for a trial like this. There are several reasons why people tend to improve in trials irrespective of treatments. One is regression to the mean for a relapsing remitting condition - where people get recruited at their worst and then return to average. Another is the placebo effect - which may not have operated much in the SMC group so might not fit. Then there is the 'reverse placebo' effect of people saying they are better when placed in the social role of being a therapy beneficiary. (I have recently emphasised this in my submission to the Scottish Parliament that they have published online.) There are probably other mechanisms. Major improvements across trials are common enough, especially if the trials are unblinded.

You are clearly wanting to hang your argument on the idea that evidence from PACE is weak because the 'wrong patients' were recruited. That is possible but I don't think we can prove it. Moreover, in real life nobody is a 'wrong patient'. Everyone deserves to be in a trial. For me the more persuasive argument is that as unblinded trials with subjective outcomes none of these trials of therapist-delivered treatments produce any valid data. There is no reliable evidence that anything in any of the treatment arms makes a real difference to any group of people.

We also have a problem in arguing in the context of NICE that maybe people with 'other fatigue' might benefit from therapist-delivered treatments. That is that NICE has to cover these people, for pragmatic reasons. So we end up with a suggestion that some groups might have CBT and GET and others not. That to me is the worst case situation because it takes us back to where we are now. In theory people for whom GET is unsuitable should not be having it even now. But they are. If the service providers are not that convinced that there are clear distinctions between patient groups we slip back into BPS mindset.

It is a pity that you have not been in on the debate longer because you have clearly thought about this a lot.