A new Beginning..., page-2378

Lets see if my nose is right with this US second trial.

I think the results were very poor and RAP will realise that it would now be better off not to apply (and get rejected) for FDA approval.

The US second trial failed to meet the pre-defined end-points that had been set in the first US trial (superiority to 75% PPA+NPA) for any illness. RAP now claims the endpoints are arbitrary but they were set following a pre-submission FDA meeting (in 2016).

The argument for non-inferiority to the accuracy of clinical diagnosis is weak. RAP achieved 70% diagnostic accuracy for three illnesses and claims that 70% is better (or not worse than) than the agreement statistics (66%) between adjudicators in its trial. Perhaps but with one proviso.

There can’t be a scenario in which you drop below the accuracy of clinical diagnosis and get close to 50%. There were a high number of participants (15%) in RAPs second US trial who did not provide analysable data (only 1251 out of 1470). Model a worse-case scenario and assume all these participants would have been wrongly diagnosed by the app. RAPs algorithm accuracy estimates will drop below non-inferiority to clinical diagnosis and (most likely) are not statistically significantly better than chance (50%).

MEB tried exactly this same algorithm non-inferiority to clinical diagnosis argument to the FDA and failed. It too had a high problem with drop outs.

Finally the accuracy estimates presented by RAP include patient (parent) reported symptoms. What exactly does the analysis of cough sounds contribute over and above symptoms? No-one knows. The results presented to the market to date conceals what is supposed to be the whole point of the app (analysis of cough sounds). But the FDA is unlikely to approve an app with an interesting charade - (you have to cough into a smartphone).

We’ll see soon enough whether RAP decide discretion is the better part of valor (abandon the FDA process) or if I should be enrolling in RAPs course for retail investors to better recognise good trial results.