OK, thank you.
Consider a pre-trial objective: "We will be testing this medication on patients who have already indicated that it helps with their affliction. Participants will be sorted into groups that get one of (a) a placebo, or (b) reduced, (c) increased, or (d) unchanged dosage relative to the dose they have reported to be effective enough relative to side effects." Such a trial could actually be useful, but good luck getting the epistemology sticklers to sign off on it.
Also, antidepressants characteristically have history-dependent effects. Often, suddenly going off one renders it ineffective if resumed. So good luck with the ethics.
Also: I would like to see the word "affliction" in more medical papers.
I hope you will find gruntling, in time.
Huh, that's a really good point (which I somehow missed in your original post).
That being said, it's incorrect that it's essential for RCT's to function, they'll still work, but with an attenuation based on the actual treatable population.
> Or, are you saying you know how to design the protocol for an RCT when you have no idea which, if any, of the participants actually suffer the condition your medication is supposed to treat?
Well I don't design said protocols, but I suspect the vaccine manufacturers may be able to point you in the right direction.
I think that your point about reliability of diagnosis is interesting, gonna need to think about it a bunch more, but in general, no diagnosis is without error (as most diagnoses are implemented as checklists, and there's some room for arbitrariness).
Just to note, if your point is correct then it's basically all mental-health, expectancy, placebo, psychotherapy RCT's are essentially worthless, which I suspect is not the right approach to take.
Does it matter if we can diagnose people effectively? Of course.
Does it make it impossible to run an RCT in the absence of effective diagnosis? Not impossible, but it definitely adds an interesting layer of complexity, as well as massively increasing the sample size required.