How to select trial participants with ME/CFS?

Split thread

I think the problem is that nobody is knowledgeable about how we should best select patients. We think there is something called ME/CFS but we don't know if that is the best way to divide people up.

It is a bit like saying we think there is a breed of dog called a Yorkshire Collie and then everyone who thinks they have a Yorkshire Collie describing their dog. We still don't actually know whether such a breed exists.

In biological science terms I think it is better to pick something you think you might be able to explain - like delayed and prolonged worsening after exercise. You would then want to see how widely that varied and if there was some stand-out delay time try and relate that to some biochemistry.

 

Exertion? (I know exertion is impossible to define as we realised with the PEM factsheet, but exercise is too narrow).

 

I take your point but @paulendat is faced with 1,000 self-diagnosed people on the Internet. How can he apply a filter so that he gets the people he needs?

Could he, for example, use certain questions from DecodeME? Then his findings could be linked with DecodeME's genetic findings via subgrouping, potentially.

(@paulendat, DecodeME is a 20,000-strong GWAS of people with a health practitioner's diagnosis of ME/CFS in the UK and will soon report its results. Other researchers can use its database.)

 

That depends on what he needs. That depends on what clinical features he hopes to explain with a lab finding. That can have pretty little to do with diagnostic categories used in the clinic. A researcher cannot assume that they are studying 'a disease' that is known to be some homogeneous process. That is fatal. It leads to people studying 'osteoarthritis' which we know isn't a single homogeneous process and is a complete waste of time.

To be honest nobody should be investigating self-diagnosed people from the internet because there will be all sorts of confounding factors biasing the population mix. That is why CureME set up a population based recruitment process.

 

Those are all very good points but I'm hoping we can still help @paulendat find a way forward. Any thoughts?

 

In case it is of use, the questionnaire that DecodeME used to screen participants is available from this page, https://www.decodeme.org.uk/documents/

 

But I feel perhaps patients who are mild/moderate who can exercise would have a more distinctive definition of delayed PEM? A pwME who doesn't experience many symptoms overall, but experiences a horrible 3-4 day delayed PEM recovery period (or longer recovery depending on how much they overdid), and then once recovered returns to baseline.

For me, it's distinctive and follows the exact same pattern for the last 25+ years.

 

As long as you say that that’s the population you want to study, it’s fine.

I might have mistakenly interpreted it as saying that PEM is caused by exercise. Which I know JE does not think.

 

Thank you! This will help for sure.

 

That’s a very important consideration, @Trish

The cost of exercising with PEM extends far beyond the symptoms you might get. And if there is no evidence or reason to think that exercise increases the amount of things you can do, it simply might not be worth it at all.

 

One of the important points that Chris Pointings "404k estimated ME/CFS patients" points out is that only about 100k patients have a diagnosis. 75% of all the expected patients don't have a diagnosis yet and that is a huge deal for research, does excluding 3/4s of patients make sense? More to the point given how mixed the diagnosis is across the country is your research now pulling from particular regions more and excluding minorities? So while many might malign self diagnosis there is a clear problem here and research can not rely on doctors correctly identifying the condition. We know that there are people with CFS diagnoses who only have/had fatigue and a lot of people who meet NICE/CCC/IOM/International who don't because there GP doesn't believe in the disease. A large number of GP offices don't seem to diagnose the disease. Many only have a few and potentially those diagnoses came from a private GP and got added to the patients record and have nothing to do with the local NHS because its clearly not normal in that office (would love for that to be investigated more).

At the moment ME/CFS is a research clinical diagnosis based on a particular set of symptoms that there is no way to reliably independently test on a patient well, certainly not without potentially harming the patient and making them worse. You could enhance that with pedometers, sleep studies and maybe neurological testing in mild and some moderate patients who can do that without impact. CPETs and most other tests involve extra exertion and can not become routine. If the tests have a risk of causing PEM they have a risk of permanently worsening the patient and that should not be something that becomes a routine risk for research nor diagnosis. When people tell us they feel tired and fatigued and that it varies a lot and they are pushing through and they get bad headaches and memory issues etc we should just believe them as we do for a myriad of other conditions with no biomarker.

 

I have mentioned this before but I think research criteria for ME/CFS should move beyond clinical diagnostic criteria.

I can best describe this by the process required. First take a medic diagnosed group of ME/CFS patients and make a hypothesis or impose a further stricter condition on symptoms. Test to see if ANYONE in the group matches the hypothesis then select those as a subtype group characterised by the test or criterion imposed and a starting point for further experiments to discover mechanisms for just that group. This is in a way what covid has done in causing longcovid, shown us a clear subtype and even that includes different kinds of patients.

By defining subgroups via the experimental results they produce, we can make cohorts more homogenous though the population the results will be relevant to will be reduced. Its a trade off but it is impossible to make progress otherwise. If we repeat the process enough times we should be able to discover all or at least a majority of the subtypes and understand better what is going on for each.

Currently broad criteria mean we have a very mixed bag of ME/CFS patients, heterogenous, which means experimental results are likely to be white noise, awash with conflicting signals. By defining test cohorts more carefully we should be able to make progress with each subtype.

 

I'm not saying that everyone who is mild/moderate can exercise. I agree that PEM is cumulative and can develop into a new less functional baseline, so pacing, prioritizing/managing our energy limits is crucial.

I am referring to studying people like myself who are mild/moderate who have a distinctive delayed pattern after exercise, and by exercise I mean able to get out and do some sort of activity like as walking, mild swimming et that raises our heart rate and avoiding delayed PEM by staying within our energy limits. Those of us who don't deteriorate over time. At what point does the benefit and energy threshold start and end when we are no longer functional (delayed PEM)? What changes are occurring? Study those who have figured it out over the years and don't lose their baseline as a result.

I feel it would be more difficult to study immediate PEM. Is the pathophysiology of immediate and delayed PEM related or distinct?

 

I think there is a problem with this that has to be faced, though. As a rheumatologist I came across people who were sure they had lupus or rheumatoid or ank spond and reported the classical symptoms of the condition concerned. But they didn't have that disease. We also have good evidence for people being convinced that they have the typical symptoms for 'hypermobile Ehlers Danlos syndrome' or 'mast cell activation syndrome' when they probably don't in a meaningful sense have either.

This may not be the biggest problem. The bigger problems may be doctors using ME/CFS inappropriately - both too often and not often enough. But self-reporting through the internet based on predefined terminology seems to me potentially tricky. When electing symptoms doctors usually deliberately avoid trying to get answers in predefined terminology. We tend not to ask 'do you have unrefreshing sleep?'.

There is also the problem with self-selected internet using patients that they form an atypical sample for all sorts of other reasons. What may be the good news though is that as more studies are done we may be able to get a measure of just how troublesome these problems are and allow for them.

 

Something I’ve been tossing around as a potential dissertation idea is to start with a broad group of individuals who have been diagnosed with or suspect ME/CFS and doing unbiased baseline multi-omic clustering (from home visits) coupled with in-depth interviews and subjective categorization.

If there is a clear-enough molecular signature (not a biomarker, but a detectable pattern of co-variation among hundreds of analytes) it wouldn’t matter that some of the individuals don’t “actually” have ME because theoretically the algorithm would assign those participants to a different group.

This interview would be carried out by an expert clinician or someone very familiar with the disease and would involve not just cataloguing symptoms, but determining typical time scales, day-to-day variation, etc. Based on all that, the interviewer would then be asked to make a subjective categorization of all the patients.

The interview portion can be repeated with multiple interviewers to determine the degree of consensus in patient categorizations—I expect that there will be differences introduced by personal bias, but theoretically it would be possible to then do clustering on that set of subjective categorizations (i.e. pick out what the consensus patterns are, if any).

My thought is that the multi-omic clustering would be blinded from the interview process, and the final analysis would involve seeing whether predetermined multi-omic clusters match with predetermined assigned clusters based on subjective reports.

The biggest challenge in the multi-omic side would be whether there’s a strong enough signal at all at baseline. A colleague of mine had some success determining multi-omic subtypes in a LC study, so I’m hopeful in that regard.

Doing all baseline comparisons would eliminate the problem of different tolerance to exertion challenges (e.g. CPET) initially. If it is possible to determine subtypes at baseline, you could then repeat profiling after exertion within each subtype to see if there is a consistent pattern post-exertion despite inconsistencies in what individuals could manage for the actual challenge.

Obviously none of this procedure is impervious to bias—however, I think that will continue to be a problem inherent to ME research so long as there’s no biomarker. If reproducible (and that’s a big if), what this provides is the opportunity to call relatively homogenous sub-populations in any future study, so that we can know if an intervention is more likely to work (or if any other finding is more likely to show up) for subtype A, B, or C as previously defined.

And even if this doesn’t yield any reproducible groups at all, from either the multi-omics or interview side, I think that is still useful information—if done well enough, it could provide a definitive answer on whether there is merit to the idea of subtypes at all.

 

As soon as you say that you are excluding all severe people and probably a lot of moderate. It’s a problem many studies have had, the barriers of entry for people. And something that DecodeME (and the CureME team) did a lot of work to remove, but it was still a lot of work for a lot of patients and we couldn’t all give all the detail we wanted.

I absolutely understand as a researcher the detail you are looking for. But it’s something to be aware of as it’s cropped up a lot before. I think there may be ways of doing it without interviews. But the talk of them and especially repetition is a huge red flag for many.

 

Yes I had this thought—it could be adapted for severe (though possibly not very severe) by being an interview over text or through email link with one question at a time. I am thinking that participants would receive an initial set of prompts, which they could answer at their own pace and level of detail, with the possibility of follow-up questions from interviewers (that are curated to avoid repetition and which the participant can decline if unable to do).

You wouldn’t be able to get the same level of detail as from mild patients, but the goal is subjective categorization anyways—it doesn’t have to be perfect to be meaningful

 

It does sound complicated, and a lot will depend on what it is you're really trying to get at. It seems to be sort of about subgroups, but also about diagnostic challenges and the issue of bias in both patients and clinicians.

I'm not sure what the simple version of the question is ... but I realise it's only one idea, and you've only sketched it out in rough anyway!


[ETA: cross posted with your reply to someone else]