PACE trial authors continue to ignore their own null effect - JHP by Mark Vink

[Barry53]

Note that the figure I posted also gives the results for the depression only group. Their results for CBT and GET look a little worse than the average. So I don't think for these two measures one can claim the results are down to people having depression which responded to CBT or GET.

But isn't it a non-trivial possibility that those whose only psychiatric issue was depression, might well have this as a consequence of their ME disabilities, and so be less likely to respond anyway? Without all the necessary data (and possibly the not trial run so as to disentangle such interactions within the available data), is there any way of knowing?

 

[Dolphin]

Despite receiving treatment deemed to be “effective” (White et al., 2017), and stating that at long-term follow-up “the benefits of CBT and GET were maintained some 2 years after treatment” (White et al., 2017), patients in all four treatment groups remained ill enough to re-enter the trial based on both subjective primary outcomes (Vink, 2017; White et al., 2011). There was no significant improvement on any of the trial’s objective measures, such as numbers returned to work or levels of fitness.

Given the was a statistically significant improvement on the main six minute walking test for graded exercise therapy, I would be uncomfortable quote in the last sentence.

What is the threshold for determining "statistically significant"? (Genuine non-sarky question, I don't know the answer). And does the GET improvement not have to be taken relative to the non-intervention SMC figure, meaning 1.3%? If so that sounds to me like statistically insignificant?

That 1.3% figure only relates to one way of defining improvement, the number of those who approved by 50% or more.

However there was a statistically significant improvement on the main six minute walking test results for graded exercise therapy, based on the mean. This was not in comparison to the baseline score but the statistically significant difference with the specialist medical care-only group:
379 m vs 348 m.
When baseline factors were taken account of, this is a difference of 35.3 metres.p=0·0002

 

[Barry53]

Thanks a lot to Dolphin for going through and highlighting possible issues with this paper, and also for posting about it here.

Agreed. I think it also demonstrate's the value of Mark Vink's contribution, given the combined benefit seen here of the article itself, @Dolphin 's critiquing of it, and the resulting discussion. Feels like how it should work.

 

[Dolphin]

Note that the figure I posted also gives the results for the depression only group. Their results for CBT and GET look a little worse than the average. So I don't think for these two measures one can claim the results are down to people having depression which responded to CBT or GET.

But isn't it a non-trivial possibility that those whose only psychiatric issue was depression, might well have this as a consequence of their ME disabilities, and so be less likely to respond anyway? Without all the necessary data (and possibly the not trial run so as to disentangle such interactions within the available data), is there any way of knowing?

I'm afraid you have lost me.

 

[Barry53]

I'm afraid you have lost me.

There were a group whose only psychiatric issue was depression, within a broader group who had psychiatric issues. I was just postulating that people with ME, given its very debilitating effects, might be quite dominant within the depression-only group. If so then although CBT and GET might typically have some success for people suffering with depression, the reason it might not have done so in this case could be because ME was the dominant issue, not their depression.

 

[Dolphin]

There were a group whose only psychiatric issue was depression, within a broader group who had psychiatric issues. I was just postulating that people with ME, given its very debilitating effects, might be quite dominant within the depression-only group. If so then although CBT and GET might typically have some success for people suffering with depression, the reason it might not have done so in this case could be because ME was the dominant issue, not their depression.

I have been thinking about this but I'm not sure I'm comfortable with it. The people who had no depression in the CBT and GET groups did slightly better on these two measures than those who do have depression. Are we saying that the depression group would contain a greater proportion of those with ME than the non-depression group. I suppose it is a possibility but I remain to be convinced.

 

[Dolphin]

Oops, wrong thread.

 

[Sean]

noticeable facial pallor is sometimes apparent and often precedes the onset of extreme tiredness in the patient.

[snip]

The hands and feet can be unusually cold and dependent rubor of the legs is often present when standing or sitting.

Unsolicited comments from people about me looking pale was one of the first externally confirmed signs for me, pre-diagnosis, that I wasn't imagining it.

I have also had the bizarre experience of my feet in particular, but also hands to a lesser extent, being distinctly cold to the touch, in the middle of a cloudless day at the peak of a blazing tropical summer! I don't mean a bit cool, I mean cold, such that when somebody else felt them they would be very surprised indeed. That is an indisputable physical sign.

 

[alex3619]

Without all the necessary data (and possibly the not trial run so as to disentangle such interactions within the available data), is there any way of knowing?

The typical depression instruments lead to an inference of depression. Its not a fact. The error rate is probably high. Depression is a symptom of a huge range of things, and that includes long term chronic illness. Depression has no objective diagnostic test. In many respects its in the same situation as CFS or ME. Yet its accepted, because traditionally its been accepted.

Depression is a real and sometimes very dangerous symptom. Its not a discrete clinical entity. Treating it as a discrete entity is a category mistake. I have very little doubt that some depression subgroups are discrete medical entities, but that leads into the whole heterogeneity debate.

In short, any reference to depression is about as reliable as any reference to pain or fatigue. Its the details that matter.

 

[alex3619]

That is an indisputable physical sign.

Metabolic or vascular, or both.

 

[Barry53]

I have been thinking about this but I'm not sure I'm comfortable with it. The people who had no depression in the CBT and GET groups did slightly better on these two measures than those who do have depression. Are we saying that the depression group would contain a greater proportion of those with ME than the non-depression group. I suppose it is a possibility but I remain to be convinced.

I'm not really making any claim because the simple truth is I don't know. I was more suggesting a possibility, that others more capable than I might better get to the bottom of. I just felt that buried in the data somewhere (data we may not have access to), might be the answer to whether a significant number of participants with ME might also have related depression - but I don't know. My wife has mild-to-moderate ME, and though it inevitably gets her down at times, she most certainly does not have clinical depression - thank heavens. But I don't know how typical that is of other similarly afflicted ME sufferers, and perhaps more importantly, I don't know how typical the PACE selection was or wasn't.

 

[alex3619]

If we took any of the typical questionnaires for depression, and removed all ME symptoms, what would be left?

 

[me/cfs 27931]

I'm not really making any claim because the simple truth is I don't know. I was more suggesting a possibility, that others more capable than I might better get to the bottom of. I just felt that buried in the data somewhere (data we may not have access to), might be the answer to whether a significant number of participants with ME might also have related depression - but I don't know. My wife has mild-to-moderate ME, and though it inevitably gets her down at times, she most certainly does not have clinical depression - thank heavens. But I don't know how typical that is of other similarly afflicted ME sufferers, and perhaps more importantly, I don't know how typical the PACE selection was or wasn't.

Of perhaps some relevance to the conversation, a recent study of UK ME/CFS biobank participants found:

"Unlike MS, [in ME/CFS] the severity of these [cognitive and sleep] symptoms seems to be unrelated to depression."

http://forums.phoenixrising.me/inde...ve-and-sleep-symptoms-in-me-cfs-and-ms.52448/

https://link.springer.com/article/10.1186/s12883-017-0896-0

 

[BruceInOz]

However there was a statistically significant improvement on the main six minute walking test results for graded exercise therapy, based on the mean. This was not in comparison to the baseline score but the statistically significant difference with the specialist medical care-only group:
379 m vs 348 m.
When baseline factors were taken account of, this is a difference of 35.3 metres.p=0·0002

Surely, given that approx 30% of data is missing in each group for the 6 minute walk, quoting a great p value and claiming statistical significance is meaningless? The 95% confidence interval they quote suggests they are 95% confident the "true" value (assuming normal distribution) of the 6mwt lies inside that range in the subgroup that completed the walk. They have no idea what the "true" value is for the complete group of (approx) 160 originally randomised.

 

[alex3619]

The 6MWT is highly problematic. If data were backed by actometers then we might be able to validate it. However the results were terrible for all groups. How do they rule out that one group might have been persuaded to push a little harder? How do we know this group, having decided to push harder, didn't rest up more? 6MWT is not very reliable. CPET would be better, and 2 day CPET better again. Indeed my understanding, with insufficient data being published, is that the step test they did didn't accord with improvement in fitness, which implies that improvement in distance had nothing to do with fitness. If the data shows otherwise then why was it not published?