Preprint Replicated blood-based biomarkers for Myalgic Encephalomyelitis not explicable by inactivity, 2024, Beentjes, Ponting et al

No expert here, but I'm afraid these are mainly small differences inflated because of the enormous sample sizes. For blood traits, for example they report data of 1,455 patients and 131,303 controls. In such context, statistical significance does not mean much as even minor differences that are clinically meaningless would become statistically significant.
EDIT: I misread, most of the effect sizes are not that small - see quote below and comment from Simon M.

The authors write:

I would be interested in seeing which measurements had the largest standardised effect size. Figure 2A reports effect sizes but on the original scale which I find hard to interpret or compare. At first glance, the supplementary Excel files do not seem to provide standard deviations for patients and controls, so it might be difficult to calculate an effect size or get a sense of the overlap between groups.
EDIT: the data does include these standardized effect sizes in supplementary Table 11.

For measurements that show a significant difference, it would also be useful to see how many patients exceed a certain threshold as they report for the CRP level, as this puts things in a different perspective. There were only 0.4% more ME cases with a CRP > 10mg/L suggesting there isn't a substantial subgroup with highly abnormal levels. ME/CFS caseness was also mainly self-reported CFS without clinical confirmation to exclude other potential causes of the symptoms. So some patients might believe they have ME/CFS but in fact have another illness that could explain some abnormal values.

Still looks like a useful analysis though. For example if a new abnormality or biomarker is reported we can check here to see what the biobank data says about it.

 

I've not had the chance to read the paper in detail yet so perhaps I missed it but I noticed in the DAG that physical activity was modelled as a mediator but I didn't see mention of body mass index being a potential mediator.

The sorts of blood tests listed above are routinely performed on our psych admissions and it's not uncommon to see mildly elevated CRP levels, mildly elevated liver function tests and dyslipidaemia. I don't know whether this is due to overweight/obesity, psychiatric medications, the illnesses themselves, comorbidities, general ill-health (or something else entirely), but I think you're right that a psychiatric control group would be useful for a future study.

 

Supplementary Table 11: Cohen’s D estimates for BT, NMR, and proteomics

 

Aha, thanks!

 

For blood traits, these are the ones that had a Cohen's D > 0.2 for both males and females:

 

I'm no expert either, and I hope Chris Ponting will reply later.

But my take is slightly different to yours. Of course, big samples mean it’s easier to reach the statistical significance. And it would be extremely surprising to discover large effect sizes, given that there are no replicated blood biomarkers to date.

What I find striking is the quoted figure of 115 biomarkers replicate in males and females, with a Cohen‘s d of between 0.2 and 0.5 (small to scraping moderate), you wouldn’t expect any of these to replicate in the sample size used in almost all other ME studies.

instead, these results paint a picture of substantially different blood biology in ME patients (which, as the manuscript says, are unlikely to contain any severe patients and probably not many moderate ones), compared with healthy controls. And that these differences are not due to lower levels of inactivity amongst ME patients. so it’s a broadbrush view rather than homing in on a few biomarkers. at any rate, I was surprised to see such differences.

I don’t think it gets us any closer to finding a blood biomarker. But I don’t think that’s unexpected

.

 

@Chris Ponting

This looks interesting. I am about to spend a week speaking only Spanish so I may not be able to read through immediately!

My initial thoughts:

Nothing dramatic has turned up, but we knew that would be the case. And the point of the GWAS was that it had the power to force a causal interpretation on what otherwise may be a mess of epiphenomenal associations in studies like this one. But, there is still a realistic chance that, if analysis is careful, as it clearly is here, and several groups find particular molecules coming up repeatedly, using large cohorts, that we will get critical clues showing up. There are some candidates here.

I am getting older and although I can still handle complex arguments my attention span is dropping. I find the paper quite hard to read and I am not quite sure why but I have a few thoughts.

I was brought up to give the facts of the data as they are first and leave any interpretation either for later or to the reader. So in the abstract I want to know what at least some of the molecules are that are different. Whether they indicate inflammation or not is my business! The reason why this is important to me is that finding signs of 'inflammation', 'insulin resistance' and 'liver' means nothing much. Any sort of process might throw those up. But if I know that lipid profile and complement receptor 2 (CR2) have come up I immediately start ratcheting through a series of possible mechanistic inter-relations that might be unique to ME/CFS and for which other groups have raised clues.

Three things here seem to have come up before - complement proteins, lipids and leptin. Leptin is intriguing but I don't know much about it. Lipids get caught up in all sorts of regulatory processes and may be telling us something very important but they go a bit wide. Complement receptor 2 has a very special function, which has little or nothing to do with the inflammatory role of complement. It regulates further antibody production. Factors B and H are also intriguing because they feed the same path. Previously we heard of a C7 signal, which was more difficult to make sense of to me, but either way, the crucial thing lies in exactly which molecule is different because that can lead to a unique and unexpected pathway. That route that paid off in the autoimmune diseases in the end - teasing out highly specify pathways. C-reactive protein comes up in everything so a bit of signal is likely to be noise. CR2 doesn't

I appreciate that it is important to focus on the methodological aspects, even if the statistical methods are obscure to most of us. I do wonder though whether it would be possible to use less jargon. Why are molecules or cell features called 'traits'? What exactly is 'gene ontology' - I have never managed to remember what that means since ontology for me means something quite different.

 

This project looks great, @Chris Ponting, it's amazing the group is so committed that they're prepared to do this work after all their other work.

I always find papers hard to read because I can't grasp maths. It did strike me, though that inactivity looks a bit over-stated. There might be reasons focusing on it so much—maybe the results would be challenged if it weren't headed off properly—but going on the Walk Time figures, pwME weren't hugely less active than controls. A bit less, sure, but that's all.

Just a thought. I never understand the actual science, so I can't be of much help in explaining it better!

 

Yes, that jumped out at me as being an overstatement.

 

Wow! The contrast with BPS researchers with multiple COIs could hardly be greater. Huge thanks to everyone working on this.

And thanks to ME Research UK for funding the PhD.

Something I just discovered, which may be of interest to others: Opening the document in Adobe acrobat gives me the opinion to ask Adobe’s built in AI assistant questions about the paper.

 

I don't think it's an overstatement.

I think the point in the preprint is that you can show abnormalities in blood, ie there's something biologically awry. Whereas a psychosomatic explanation maintains that there is nothing wrong - it's an abnormal perception of normal, healthy (minor) body symptoms — then compounded by lack of activity. They might state that any biological effect is due to deconditioning. By showing the biological difference from people who are similarly deconditioned / non-active, that refutes this psychosomatic explanation.

I also don't think depression is a psychological illness (or psychosomatic). It's a biological disease that principally affects mood. But it also affects the cardiovascular and immune system — it's just that the mood disorder dominates the clinical picture acutely and is the potentially more immediately life-threatening. The cardiovascular issues come later. Of course they are often ascribed to "lifestyle factors" associated with depression rather than the underlying biology.

---

Association of Depression and Poor Mental Health With Cardiovascular Disease and Suboptimal Cardiovascular Health Among Young Adults in the United States (2023, Journal of the American Heart Association)

Depressive Symptoms and Mortality Among US Adults (2023, JAMA Network Open)

Association of Depression and Cardiovascular Disease (2023, The American Journal of Medicine)

Association of Symptoms of Depression With Cardiovascular Disease and Mortality in Low-, Middle-, and High-Income Countries (2020, JAMA Psychiatry)

The association of depression at any time to the risk of death following coronary artery disease diagnosis (2017, European Heart Journal - Quality of Care and Clinical Outcomes)

 

I think it would be better to say that it shows there is a physiological basis to the illness which is not associated with inactivity – or something along those lines.

BPS people are generally careful not to say that ME/CFS is psychosomatic. It’s not even clear what psychosomatic means. If one interprets it to include “physical illness caused or aggravated by mental factors,” then the data would not dispel that.

 

Thank you very much to you and your team @Chris Ponting it means a lot to know people are working so hard and passionately on our behalf!

 

In addition to the one point I made earlier in the thread, a few random thoughts from an (extremely quick) skim through:

Personally I would use ME/CFS rather than ME to align with NICE and to avoid the obvious criticism that there is no encephalomyelitis. Also useful to note that this is self-reported lifetime diagnosis of "chronic fatigue syndrome" so no indication that the patients would meet modern criteria for ME/CFS or that they have PEM.

"The study also controlled for potential confounders such as age, sex and physical inactivity" - I've no idea just how much information the UK Biobank can make available to researchers but diet & medication use (& perhaps even socioeconomic status & maybe even smoking & alcohol use) could also be potential confounders. Many ME/CFS patients will be taking various antidepressants, NSAIDs, pain modulators or herbal "supplements", many of which can lead to mild elevations of ALT & GGT. If I remember correctly beta-blockers, not uncommonly used by pwME, can lead to elevated triglyceride levels by altering lipid metabolism. Many pwME adhere to atypical diets either because they believe that they are healthier or in an attempt to alleviate gastrointestinal symptoms which might affect triglycerides & possibly ALT - also if I remember correctly leptin levels are affected by body fat & dietary protein intake affects alanine and urea levels. As mentioned earlier different comparison groups might be useful. There are some very interesting results in here (SOD3, complement proteins). Also I think Hutan pointed out a couple of typos - not that it matters but in my brief skim I also noticed "tryglycerides" and "apoliprotein" instead of apolipoprotein.

Hope to have the energy to actually read through this paper thoroughly at some point but it is wonderful to see interesting high-quality work being done and the fact that it was not funded - wow. Thanks so much to everyone who worked on this not only for your interest & dedication but also for your engagement with pwME.

 

I think we need a study of this against a control group that is also exposed to potentially myriad medications or other relevant factors.

This may be the most exciting part. Prospective targets for future investigation that appear to be gender agnostic might be huge for our field.

Your commitment to accountable science is inspiring.

 

I agree with a number of comments about the treatment of causation.

I don't think the study can discount a 'psychosomatic' account because psychosomatic can imply that thinking modifies peripheral body functions in such a way as to cause symptoms and that may be biochemically mediated. In fact, tha 'biosychosocial' view simply requires that psychosocial factors modulate the process, which might be by change in diet and so on.

More generally, I think the isolation of inactivity is an interesting approach but rather arbitrary, as others have said. Diet, drugs, social contact with people with viruses and all sorts may come in.

Perhaps the thing that worries me most is that the analysis seems to imply that there is some homogeneous 'ME' process throughout the cohort. There may be but there may not. The 'ME' cohort may include people with two completely different processes generating similar reports of symptoms. Being devil's advocate, one of those processes might be entirely 'psychosomatic'.

I would definitely stick with ME/CFS rather than ME, even if the defining of the cohort involves a diagnosis of 'ME' or 'CFS'.

As a final note, CRP is not necessarily an indictor of inflammation. It is an indicator of IL-6 activating hepatocytes. Even in rheumatoid arthritis, the paradigm high CRP disease, the generation of high CRP may be almost entirely due to activation of Kuppfer cells by immune complexes leading to local hepatic cytokine signalling and that is not inflammation for the bizarre technical reason that Kuppfer cells interact with complexes while sitting inside blood vessels rather than outside them, as all other macrophages do.

 

Yes you're right these aren't that small. I had misread and assumed that most statistically 'significant' findings would have very small effect sizes < 0.2 SD which doesn't seem the case. Apologies for the confusion (will EDIT my first post above).