The real thing is can we (perhaps with new advances in statistics?) find places where doctors at the end of all known treatment try things at random and see if anything works. For rare symptoms it often isn't possible to run a proper controlled study (note that I said symptoms and not disease - often we are not really sure what is going on). So if we can give doctors a list "here is what someone else tried and or treatment Y seems to make things worse that is a clue. Eventually we can say that we don't know why, but X seems like your best shot even if it we can never get to real statistical significance.

Doctors already have case studies which they (at least should) publish anytime they get someone who for whatever reason doesn't respond to standard treatment, or has something unknown. However it is hard for the next doctor to find any that might be relevant.