Why are patient pleas about subtyping and stratifying in studies seemingly falling on deaf ears?

[Liface]

It was heartening on Twitter today to to see so many patients as outraged as I am at the continued lack of subtyping and stratification in Long COVID and ME trials, exemplified by the presentations at the International ME/CFS Conference 2026.

But it seems that that patients have been shouting into the void for years, and how trials are run hasn't changed.

Why is this?

I have some ideas on how we might get through to researchers, but I would love to do some group brainstorming here: what do you all think the blockers are?

1. Researchers still don't realize subtyping matters (unlikely?)

2. They know, but believe heterogeneity averages out with enough sample size

3. Researchers believe it is too costly to collect subtype data (for example, believing that they need to collect additional biomarkers, even though surveying patients would also be more signal than noise)

4. Researchers afraid that breaking groups into subtypes at the end will be viewed as P hacking by reviewers

5. Self reported subtyping (POTS, MCAS, PEM-dominance) viewed as too noisy to stratify on

6. The people who can actually exert pressure on publication (journal editors, statistical reviewers, grant panels) don't know or don't care about subtyping

7. Lack of validated subtyping instruments (there's no "official" 10-question subtyping survey everyone agrees on)

8. Selection bias on our end - maybe well-stratified trials exist and we just notice the bad ones (unlikely)

Which of these do you all think are the biggest hurdles - or are there any I'm not seeing?

 

[EndME]

I think there's a few subtleties here and much of the outrage on Twitter I've seen in the past often came from basic misunderstandings of trial methodology. There seems to be a fundamental misunderstanding between the possible existence of different phenotypes in ME/CFS and between interpretation of trial results or studies.

There is a group of patients and researchers that believe because there was a population of responders in a clinical trial that there is subgroup in which one has efficacy. This clearly does not follow, especially when there is typically also a subgroup of responders to placebo, often these even dominate the size of the other group. This is generally speaking an issue of statistical power. Seeing responders in trials does not imply the existence of a subgroup in which efficacy exists!

Long-Covid trials being run without any thought behind them are clearly pointless and has been a problem along a majority of Long-Covid trials. There is probably a substantial difference between ME/CFS and Long-Covid in that respect. Running a trial on "Long-Covid" without further stratification is clearly pointless, for ME/CFS this needn't necessarily be the case.

It should be noted that suggestions of subtyping by things such as POTS and MCAS does not seem meaningful either, as has been thoroughly discussed. Similarly stratifying ME/CFS patients by say "GPCR-aabs" or "autoimmune markers" seems no less arbitrary them stratifying them by the colour of their hair. There is no indication of these being subtypes of anyway, the research in fact rather suggests the opposite.

Of course all possible data should be collected pre-trial and post-trial and analysed accordingly, but suggesting subtypes after having identified a subgroup of "responders" by such data has in the past not been meaningful and is typically only meaningful after a further trial. We've seen many researchers being led astray by such interpretations and much money wasted.

As an example: Fluge and Mella tend to do well in their studies, what do you think they should do differently?

Edit: Note I think it's very possible that there are different illnesses currently understood to be ME/CFS (but certainly not as many as often suggested) and I can appreciate a rate of misdiagnosis as well, but that doesn't lead me to having a crystal ball to identify which is which.

 

[V.R.T.]

Because we have no good evidence that there are a dozen+ subtypes of MECFS.

Under the MECFS syndrome heading there may be one disease or several but there is no good reason to stratify patients according to vague overlapping criteria.

This stuff is rolled out over and over again by people who strongly believe that failed drugs actually work. I think if a drug really works even in a subset, it will look like the cyclo and dara responses. And F&M are working on figuring out why some respond and some don't.

But the idea that this disease is totally heterogeneous and every single drug works in a subset is pushed by charlatans who give patients a bunch of off label drugs that have people reporting small benefits that are almost certainly from the natural fluctuation of the illness.

What matters most is measuring step counts. The only responses that matter will be accompanied by huge increases in step count in participants.

 

[Verity]

Unless I misunderstood, I think Audrey Ryback said recently in her interview with David Tuller that the evidence they’ve collected showing two age-of-onset peaks is good evidence we’re looking at either one clinical entity or two. That is the only type of possible subgroups we have any significant evidence at all for as far as I know. I’m not aware of any other feature you could subgroup based on that would amount to more than a guess.

For LC, it’s a totally different story, of course.

 

[BrightCandle]

No studies that have tried to classify people into sub groups has managed to find any useful way to put patients into different groups, there is far too much overlap in symptoms. When the NIH tried all it did was recreate mild/moderate/severe/vsevere. These subgroupings don't exist especially when you get into very large numbers of patients there is no way to cleanly cut them up.

About all you can potentially do is if you treatment targets some biological process that you can measure and only applies in some patients you can select for that, BC007 did that with GCPR antibodies. It hasn't historically made any difference to the outcome however.

I don't think its clear this can be done on symptoms, only potentially on some blood test to confirm the target and then after treatment to confirm it changed in the way expected. But so far when this has been done it hasn't changed the outcome. What we saw today was trials failing because no one really responded more than placebo once a control was in place, if a sub group got better this would definitely show up in the data and it just didn't and many of them had sufficient size for this to show up.

 

[Liface]

I think there's a few subtleties here and much of the outrage on Twitter I've seen in the past often came from basic misunderstandings of trial methodology. There seems to be a fundamental misunderstanding between the possible existence of different phenotypes in ME/CFS and between interpretation of trial results or studies.

There is a group of patients and researchers that believe because there was a population of responders in a clinical trial that there is subgroup in which one has efficacy. This clearly does not follow, especially when there is typically also a subgroup of responders to placebo, often these even dominate the size of the other group. This is generally speaking an issue of statistical power. Seeing responders in trials does not imply the existence of a subgroup in which efficacy exists!

Long-Covid trials being run without any thought behind them are clearly pointless and has been a problem along a majority of Long-Covid trials. There is probably a substantial difference between ME/CFS and Long-Covid in that respect. Running a trial on "Long-Covid" without further stratification is clearly pointless, for ME/CFS this needn't necessarily be the case.

It should be noted that suggestions of subtyping by things such as POTS and MCAS does not seem meaningful either, as has been thoroughly discussed. Similarly stratifying ME/CFS patients by say "GPCR-aabs" or "autoimmune markers" seems no less arbitrary them stratifying them by the colour of their hair. There is no indication of these being subtypes of anyway, the research in fact rather suggests the opposite.

Thought experiment: if a trial on Ivrabradine was done in ME, should the readout be done as one group, or should the trial be split into ME-POTS and ME-Non-POTS?

There are very clearly subtypes: I personally have ME, but not POTS, Mast Cell Activation Syndrome, GI issues, small fiber neuropathy, etc.

 

[Kitty]

Because we have no good evidence that there are a dozen+ subtypes of MECFS.

Yup, that's about it. You can only identify subgroups based on blood, tissue or genetic differences, or by their responses to a targeted drug treatment.

We have neither, so subtyping is meaningless.

 

[Liface]

Yup, that's about it. You can only identify subgroups based on blood, tissue or genetic differences, or by their responses to a targeted drug treatment.

We have neither, so subtyping is meaningless.

You can just ask people.

"Do you have POTS?"

Or you can proxy it by asking which which treatments they've they've responded to in the past.

Both of these offer more signal than what we have now.

Let us not let perfect be the enemy of good.

 

[Verity]

Thought experiment: if a trial on Ivrabradine was done in ME, should the read out be done as one group, or should the trial be split into ME-POTS and ME-Non-POTS?

There are very clearly subtypes: I personally have ME, but not POTS, Mast Cell Activation Syndrome, GI issues, small fiber neuropathy, etc.

I don’t think it’s at all clear that the people with these diagnoses have something meaningfully different from the rest of ME patients in the sense that it would affect trial results—unless the trial is actually for one of those conditions and not ME.

Quick edit: In fact, it’s not even clear to me that people with those diagnoses look more like each other than they look like other ME patients.

 

[BrokenBeaktheBrave]

Yup, that's about it. You can only identify subgroups based on blood, tissue or genetic differences, or by their responses to a targeted drug treatment.

Forgive me because I'm in no way scientifically-inclined and very brain-foggy today but if we can diagnose MECFS without "blood, tissue or genetic differences, or by their responses to a targeted drug treatment" then why can't we identify subgroups without them ie identify subgroups based on symptoms?

 

[EndME]

Thought experiment: if a trial on Ivrabradine was done in ME, should the read out be done as one group, or should the trial be split into ME-POTS and ME-Non-POTS?

There are very clearly subtypes: I personally have ME, but not POTS, Mast Cell Activation Syndrome, GI issues, small fiber neuropathy, etc.

There is strong evidence that POTS does not capture the nature of orthostatic intolerance in ME/CFS even if some myths get perpetuated (this is a good summary of the problems: https://mecfsscience.org/the-problems-with-pots/). So it should not be split into ME-POTS and ME-non-POTS as there is no indiation that ME-POTS even exists in a meaningful manner.

If someone specifically hypothesises that Ivabradine addresses the issue of orthostatic intolerance in ME/CFS then I see a split analysis into ME/CFS with orthostatic intolerance and without as being crucial and special attention in the analysis should be paid to HR changes and maybe some other markers, but that's about it. We won't get futher by perpetuating memes.

If I remember correctly there was an Ivabradine trial that suggested what I wrote above: the 30 BPM limit gets affected but orthostatic symptoms seem rather unaffected.

 

[Verity]

Forgive me because I'm in no way scientifically-inclined and very brain-foggy today but if we can diagnose MECFS without "blood, tissue or genetic differences, or by their responses to a targeted drug treatment" then why can't we identify subgroups without them ie identify subgroups based on symptoms?

Because we really have no idea which symptoms represent significant differences between patients, if any do at all. There is some variation in symptoms in basically every illness, but most of that variation doesn’t indicate different drug responses. We’d be completely guessing at what kind of variation does do that for ME.

 

[ryanc97]

There is no proven subtype technology and Amaticas subtyping will fail or not produce anything

 

[Liface]

I don’t think it’s at all clear that the people with these diagnoses have something meaningfully different from the rest of ME patients in the sense that it would affect trial results—unless the trial is actually for one of those conditions and not ME.

Quick edit: In fact, it’s not even clear to me that people with those diagnoses look more like each other than they look like other ME patients.

If this opinion holds, you would agree that an Ivrabradine trial would read out exactly the same in someone with ME-POTS as someone without?

 

[ryanc97]

Because we have no good evidence that there are a dozen+ subtypes of MECFS.

Under the MECFS syndrome heading there may be one disease or several but there is no good reason to stratify patients according to vague overlapping criteria.

This stuff is rolled out over and over again by people who strongly believe that failed drugs actually work. I think if a drug really works even in a subset, it will look like the cyclo and dara responses. And F&M are working on figuring out why some respond and some don't.

But the idea that this disease is totally heterogeneous and every single drug works in a subset is pushed by charlatans who give patients a bunch of off label drugs that have people reporting small benefits that are almost certainly from the natural fluctuation of the illness.

What matters most is measuring step counts. The only responses that matter will be accompanied by huge increases in step count in participants

Word

Amatica thinks subtyping is key because Jack is too gullible and believes every single Internet story out there without checking their background.

For example there is an anecdote on Baricitinib recently, but if you check the posters X history, he claimed the same thing about Rapamycin saving his life

 

[BrokenBeaktheBrave]

Because we really have no idea which symptoms represent significant differences between patients, if any do at all. There is some variation in symptoms in basically every illness, but most of that variation doesn’t indicate different drug responses. We’d be completely guessing at what kind of variation does do that for ME.

Thanks Verity.

 

[Liface]

If someone specifically hypothesises that Ivabradine addresses the issue of orthostatic intolerance in ME/CFS then I see a split analysis into ME/CFS with orthostatic intolerance and without as being crucial and special attention in the analysis should be paid to HR changes and maybe some other markers, but that's about it. We won't get futher by perpetuating memes.

If I remember correctly there was an Ivabradine trial that suggested what I wrote above: the 30 BPM limit gets affected but orthostatic symptoms seem rather unaffected.

Completely agreed. I think we're actually saying the same thing here, as I would define this as subtyping. And yet we're not even doing this.

 

[Kitty]

You can just ask people.

"Do you have POTS?"

But what does that tell you?

The clinical picture of ME/CFS or Long Covid is a good fit for people with POT and without it. That means it's probably no more useful for gaining insights into the disease than stratifying cohorts by whether or not they get hay fever or migraine.

 

[EndME]

Completely agreed. I think we're actually saying the same thing here, as I would define this as subtyping. And yet we're not even doing this.

I do think it is a bit more complicated than this. Orthostatic intolerance, which is different to POTS, is widely reported in the ME/CFS community, so there might be a justifcation for a targeted trial and sample size problems might be less prominent, moreover it is simply identified by symptom reporting and perhaps an objective measure such as hours spent upright and does not imply false pathology.

That is a bit different to some of the other things you and especially other people often mention online. For example for SFN, there is no good evidence that the prevalence in ME/CFS is higher in the general population so there isn't even reason to suggest it as a subtype. Perhaps certain neuropathic symptoms are more common in ME/CFS and deserve being addressed, but that is an entirely different problem which then has to be studied appropriately, not just by slapping nonsensical labels onto things that imply the wrong kind of thing.

People have to realise that what is often talked about as subtypes on the internet are typically simply post-hoc or bad-quality research memes.

There also seems to be a misunderstanding of what a subtype constitutes. There isn't evidence that GI-issues is a meaningful subtype of ME/CFS, just as there isn't a subtype of MS with GI-issues. A trial should try to delineate all possibilities and collect all possible meaningful data, but we won't get anywhere calling everything a subtype a priori.

I would re-ask the question: Is there anything F&M should a priori do differently in their trial?
That is a different issues to gathering all possible data and analysing this, but one can then not typically afterwards suggest efficacy based on this subtype analysis and this is the common problem we see online and the same sort of nonsense seemingly produced by Amatica. F&M would obviously not do this since they are smart but that won't stop people from saying "it's effective in a subgroup" independently of what the data says.

 

[Liface]

I think ya'll are getting hung up on subtypes. Let's forget about subtypes, or we will go back and forth about semantic category error debates all day.

New crux: trials should collect all possible data, within the limits of their funding, be it patient surveys, past drug responses, biomarkers, on-site testing, etc. and stratify results based on that data.