Everyone likes reading a well argued post Meercat. Me included.
You ask who conducts a trial for non-significance. I would just say just about everybody who is conducting a phase 2 exploratory trial.
From your statistics training you would have learnt P<0.05 – alpha – type one error (falsely rejecting the null hypothesis). Next step ... did you ever wonder why in a P2 trial with several closely related efficacy measures no-one ever corrects the P value for multiple testing?
Well no-one bothers because the p value is entirely arbitrary. I would say p values are rarely, if ever, used for decision making within a P1/2 drug development program. Different kettle of fish to think this applies to P3 and a confirmatory context.
In some ways the problem in this trial was around the other way. What if statistically significant results had been found; for example if the “outliers” as WP mentioned hadn’t been outliers. Well the power was low, and so the gereralisability of the finding would be poor.
NEU punters have just been snared on this problem. In Retts P2a statistically significant positive result is found in on a effaciacy measure (MBAci). Punters assume drug effects can’t simply disappear, and so in the P2b Retts trial with a higher dose, for longer duration, in a targeted population the results will be even stronger. In fact the MBA effect magically disappeared; it had extremely poor external validity all along.
But anyway theres an old joke.
A lab guy injects 10 mice with a new wonder drug. Then he cuts the head off the mice. Miraculously one survives. He rushes up to tell the statistician what has just happened. The statistician thinks about it says “oh right ... so not statistically significant then”.
There is element to that joke occurring here. Trouble is ... the company has so badly communicated the trial results no-one is capable of working out from the announcement whether “not significant” means “not statistically significant” or “not important”.
The seeds for this were actually sown in the interim results announcement. The company stripped out the results the error bars. Pretty unusual; as I posted at the time.
https://hotcopper.com.au/threads/an...ment-ahz-ax.2973503/page-137?post_id=19978403
The interim results were the opportunity to start re-orientating the market to the more nuanced view about P2 results. Instead the company either decided a problem delayed was a problem half solved. Or alternatively it was just too hard for them.
As a consequence, 6 months later, you end up with one of the worst presentations of P2 trial results I think I have seen. Presumably partly because of WPs experience with big pharma. Big pharma never have to sell their P2 results to investors because no-one cares about them. This and the lack of integration of the vaccine part of the business in the overall scheme of things for AHZ.
What do is another question. But yep ... the retail investor reaction to the results was entirely predictable. Lets get someone else to pay for this thing. Perhaps thats why the results were presented in this way.
But the rest of your post Meercat ... a nice blend of FA/TA and a pleasure to read.
Regards Southoz
(20min delay)