Share
clock Created with Sketch.
01/06/17
01:07
Share
Originally posted by Southoz
↑
The debate around this announcement perhaps arises from a simple misunderstanding.
Previously RAP have collected cough data and developed a prediction model using a machine learning approach. These studies were not trials and were never registered as such.
The US study is a trial and is registered as one. The trial tests whether the prediction model works prospectively in the US setting.
The extent to which development of prediction models work in the real world is very dependent on the processes used to develop them. This development process is in fact vulnerable to investigator biases.
Clinical trials are simply a method that is used mainly in health to strip out, control or make transparent potential sources of bias in studies.
Levels of rigour are almost certainly going to be higher in a clinical trial than in a “study”. This is doubly so for clinical trials performed under the auspices of a regulatory authority.
Achieving these very high levels of rigour is costly and extremely difficult if researchers lack experience in conducting trials. For this reason RAP were wise to hire a CRO.
The rigour required for the FDA will exceed that for a top medical journal. RAP have published their previous studies in middling journals on the peripheral of mainstream health where levels of rigour are far less.
None of this suggests that there are problems with investigator integrity or honesty in RAPs previous studies. That would be pretty rare. But I would say that in a clinical trial the transparency of the processes act as a safeguard against someone doing the wrong thing. And this transparency is not usually present in studies.
For example I personally have come very close to trying to do the wrong thing in a single blind trial where I knew a very well deserving subject had drawn the short straw of being allocated to the control wait list group. There was no way for me to subvert the process and completely cover my tracks.
But anyway the first thing here is whether the prediction model works prospectively in the US setting. This is simply a test of whether a prediction model based on a machine based learning approach developed in the context of a non-health discipline (engineering / IT / technical) has much validity when put to the rigours of a clinical trial.
I would be cautiously hopefully optimistic about this ... because I think the FDA bar here is probably lower than what people think. A significant correlation to clinical diagnosis will probably get this across the line in the mHealth app kind of way.
But I am very negative about where RAP will end up. My prediction is Australia’s second mHealth train wreck (with ISN the first). The genetic DNA of a mHealth app determines its success. The genetic DNA here is machine learning / engineering / IT. In other words it is a solution searching for a problem – just as the ISN wheezeometer was.
But unfortunately you do need to actually experience a couple of mHealth trainwrecks before you start to get the hang of it.
In terms of making money. I would wait for the successful US trial results. At a market cap of $250m plus RAP will arguably be the worlds best share to short sell. You’ll know this will be true because the whole of HC will think exactly the opposite.
Expand
Shorting a biotech stock before FDA results........very unlikely, hence the shorting has been very minimal. You usually short a stocks that has a high probability that it will fall otherwise, if it doesn't and it spike big (which usually happens with positive FDA results) you need to cover big time.
Basically, the pps has been falling as there are no buying pressure and the "sellers" can do what they want for now.......
http://www.shortman.com.au/stock?q=rap