All good Red bar. You’re a good advocate for RAP … and argue the substantive points.
A couple of things …
No not one of the illnesses was superior to 68.6 (look at lower bound of the 95% CI). You are actually saying non-inferior to 68.6.
The FDA will model the non-analysable; and run a worse-case scenario. This is to reflect the equivalent of an intent to treat analysis. A doctor doesn’t have the luxury of only diagnosing “analysable” people. To see how this modelling works in practice have a squiz at the IDx Nature paper (and the good editorial):
https://www.nature.com/articles/s41746-018-0040-6
RAP have not stated anywhere the results from the second US trial will be published in a medical journal.
The US investigators (MGH or elsewhere) have not publicly uttered one single word about the results. This is pretty unusual for "good" trial results.
You lost me on where or how you work out what the analysis of cough sounds adds to the predication. If someone was to say it is zero … where do you work out this is false.
In terms of this “good as clinical diagnosis accuracy” argument. The study wasn't designed to measure clinical diagnosis in routine practice.
Getting poor inter-rater reliability between your adjudicators simply suggests that you have failed to specify diagnostic criteria properly … or have over-worked, under-paid or rogue personnel. It isn’t actually a reason for lowering your bar from where you originally thought accuracy for clinical significance lay.
Particularly when the illness that was arguably best specified – pneumonia - using WHO PEP standards) had one of the worst outcomes.
Just as a complete aside (blokey) … I made one post on FFT before the failed trial results; sentiment none – not held. Hardly say “wrong” – more like it didn’t go the way I hoped it would.
https://hotcopper.com.au/threads/an...er-meets-biotech-perth.4363960/#post-35077886
But anyway it will be interesting if RAP stay the course arguing the results were good and have a crack at the FDA.
