Yes we want a diagnostic test but the whole point of that is to show that some clinical diagnoses were wrong - so you want a test that identifies something objective and which agrees with clinical diagnosis less than 95% of the time.
I would argue this just means such a test with less than 95% clinical agreement is more scientifically interesting, we can learn more from it, maybe much more. It might also mean its only a test for a subgroup, which is something we need to consider in fibro and ME. Coming from a background in neural networks, I can say that its common to find it detecting something that is not important, in other words association not causation, or even secondary consequences. So for example it might pick up a general marker that is found in many other diseases, but they did not use those in comparison groups for specificity analysis. Its a problem with the whole specificity issue.
With biomarker defined groups, rather than consensus guessing groups, you can then do further research. Maybe it will fail .. I expect the majority of such finds will fail. The point is you have a more focused group to do further testing with. So its scientifically valid at the level of a pilot study .. but as we know later, larger, and better designed trials often find these hypotheses are wrong.
Taking the line of reasoning with discordance would mean that a test which is 100% accurate, like one that exists now for ME, or 85% for a recent test, have little clinical or research utility. It might mean we could not have much hope in metabolomics, CPET, cytokine testing, and so on. Now I agree that we should not great hope in any of these, there is a long way to go, but you have to start somewhere. If its wrong and enough researchers are looking for the fatal flaws then they might rule it out. That happened with XMRV.
I would suspect that someone will use that fibro test with other similar diseases ... and there is a high chance they will find it positive, maybe in Lupus or something. That would then invalidate it as diagnostic, but it still might give clues as to pathology, which might require investigation. Indeed it might be an artefact of a too small cohort, and a larger cohort test will invalidate it.
The fact that the biomarker found is not obviously useful in identifying mechanism means that it needs more investigation, not that it is wrong.
Now a test with low clinician agreement, say at 60% or less, we might find something very interesting as to causal mechanisms, or it might just be an invalid test. So again it offers opportunity for investigation. It might also, as I already said, only be valid for a subgroup, which can then be separated from the rest in future studies.
The problems with fibro diagnostic biomarker discovery are a close parallel with ME biomarker diagnostic discovery. So I think this is useful exemplar of the issues we are going to have with future putative ME tests.
Now I would MUCH prefer we had an understanding of underlying causal mechanisms, with identification of subgroups or alternative diagnoses, and testing were based on that. However I see it as a chicken and egg thing ... how do you isolate causal mechanisms in potentially highly heterogeneous cohorts, with poor understanding of causal mechanisms, and multiple consensus definitions? Candidate biomarkers offer one way of doing that, even if flawed biomarkers are used initially. I am sure there are other ways, including those used in some of the current studies in the US and elsewhere, but which is the better path? Why?
There are other candidate markers for fibro I think, including possible SFPN, but they are also a long way from being fully validated or tested. In particular the problem with small fibers is that the testing is likely to have a high false negative rate.
Having said all that its painfully obvious that many tests have been put forward as useful, with one or two limited studies, then failed in the real world or in later studies. XMRV testing for example. Yet I would argue we learned a lot when we finally ruled out XMRV as a candidate cause for ME.
If you want to argue that this test will eventually prove to be wrong I would agree, to a high probability. That is the batting average, especially where causal mechanisms are still not understood. That does not mean it should not be investigated further. Proving it wrong, not just inferring it is wrong, is also valuable to us.
In general though, nearly all these theories and tests for things like ME have proven false over time, with a small selection still pending. This will continue until one (or more) are correct.
