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

Some questions that were going around my head today.

What sort of sample size would be needed to replicate this do we think? This has ~1500 people with ME, could a study be done with this (or fewer) people with ME and compared to the healthy controls in the biobank data?

Would it be possible to do a study with say, a subset of the DecodeME cohert, to see what the results were in a more clearly defined group and across different severities?

What are the minimum blood tests needed to compare?

Given most of us in DecodeME will have had common blood test results done is there anything that can be analysed with access to our NHS records as a starting point?

That would be nice! Presumably that would give people a clearer idea of what to focus on but also encourage and support research and funding?

 

I think you could try to replicate findings for individual markers with quite a small sample of maybe just 100 people from DecodeME. But there are issues with controls. The UK Biobank is OK to use as a source of healthy controls in terms of genes. But I think the age window would be different from that for DecodeME. So you might well not be comparing like with like. Possibly you could look at DecodeME cases aged 40-70, but the population being drawn from may be different.

Individual results for any given person do not tell us much because all these values are probably within the normal range. The meaningful result is the difference across a population/cohort. Results from routine care are probably not useful because all these tests are likely to be calibrated slightly differently in different labs.

There are lots of test that one could compare. It is not clear which of these are truly related to ME/CFS and in what way.

 

Interesting, thank you @Jonathan Edwards

 

I can’t remember what has been discussed on this thread already so I apologise if this is a repeat but a thought occurs to me: Samples were taken from the UK biobank. Given reported high rates of misdiagnosis by GPs, and the lack of requirement to have even been diagnosed by a doctor in order to be categorised as having ME/CFS for the biobank, it seems likely that there may be a high percentage of people in this study who are classed as having ME/CFS but do not meet any useful diagnostic criteria for the illness. What I’m wondering is whether a significant percentages of those cases may have the same undiagnosed conditions which are causing patients to wrongly think they have ME/CFS. Is it possible that it is the signals from conditions which are commonly misdiagnosed (and mis-self-diagnosed) as ME/CFS which are showing up in the data?

If it is possible, I hope* DecodeME may be able to tell us whether or not it is the case.

*Edit: changed wording.

 

Yes, and I have discussed this with Chris in detail. There is actually a problem the other way around too. If ME/CFS is underdiagnosed the diagnosed Biobank cases may not be representative of people with ME/CFS because they may have got their diagnosis for confounding reasons. One simple scenario is that if they also have another problem, whether or not diagnosed, that takes them to doctors frequently they may see more doctors and since only a proportion of doctors use the ME or CFS labels they will be more likely to be (correctly) diagnosed with ME/CFS.

And with a large sample you are almost guaranteed to pick up confounding associations of that sort which, paradoxically, would not show up on smaller studies.

DecodeME should help because the cohort ascertainment was better. There is still the same risk to a degree but I think it will be less. Also the risk is likely to be much less for gene associations I think. And of course the causal analysis is much easier because genes have to be upstream. In a way I see this study as a reassurance that one can demarcate how big an influence confounding factors are likely to have under what circumstances.