But isn't this the same technology that is used in say Graves disease to test for TRAb? If for some magical reason Graves disease would stop existing, would these previously useful tests all of a sudden become "fake tests", simply because they don't make predictions anymore?
The technology is not in itself the issue. It is the nature of 'having an antibody'.
As I said, we all have antibodies that bind to everything. In Graves disease the tests show that patients' sera stick more antibody to a thyroid peroxidase substrate than healthy people (mostly, there are plenty of exceptions but the test is still skewed hugely to patients).
That might be due to patients having more antibody molecules that bind or ones that bind more strongly. In the body unless an antibody has a dissociation constant of about 10^-8 for its antigen it probably doesn't ever bind long enough to matter. But if the binding constant that determines whether or not antibodies stick to antigen in an ELISA well is different then you start measuring something very different. And binding constants are affected by antigen being immobilised on plastic. Moreover antibodies love binding to plastic, which is why an irrelevant protein like casein is used to 'block' that.
When all is done and dusted you have a test that gives you an optical density based on how much stain has been fixed to the antigen or the plastic which has a good chance of being more in patients who really have antibodies with a greater power to bind in vivo, but not guaranteed by any means. But the total amount of antibody stuck to the plate in any particular case is never mentioned because it depends mostly on the size of the plate, the dose of serum and other irrelevances.
So the test only gives a result with any meaning at all if you can consistently show two populations of sera show different binding when antigen is there. And as far as I am aware we don't have that for most of these receptors, if any. And some proteins probably pick up antibody for non-specific charge reasons. The list reasons for an artifactual result is almost endless.
So a test may be consistent in that the same sample gives the same result on several occasions compared to no antigen present, but as to whether it is 'accurate' that may be unanswerable. It doesn't measure amount as such. It may measure the wrong affinity, and so on.
As I said, people have run research programmes for several years on antibodies that turned out mostly to bind to plastic, or the blocking agent. Most research papers on antibody levels are junk. So the test is 'fake' in that it is presented as telling you something biological, when it may simply indicate some irrelevant physical chemistry.
