Cooko55 - for a non-scientist you ask good questions - so I...

Cooko55 - for a non-scientist you ask good questions - so I wouldnt worry about not being a 'expert'. I cant pretend to have all the answers (and am not going to try and defend departed Directors) - but some musings might be along these lines:

You ask why they didnt test on 20-30 unhealthy patients to better get the formulation right. I suppose the reason is is that they never saw this coming. No-one suspected that there would be a difference in exposure from sc dose between healthy and unhealthy patients. Whether they should have is harder to answer.

The fact this is coming out now however makes me a little bit suspious. Remember its only some patients who didnt get the right exposure. So why not show the results from those patients who did get the correct exposure - assuming thats a reasonable number. Even with the reduced N might still show a sig result if the effect is large enough. I appreciate this is all a bit messy - and after the fact - but as a 2a (dosing) study probably defendable - although it is out of sync with what was planned when the study was registered. But I'd bet NN are pondering just this question - if the exposure was right what would be the effect. And from this compare that effect to the effect size from other drugs out there for non-responders to mainline tx.

In terms of the results shown in the poster and the announcements - I know people have focussed purely on the non-statistically sig aspect. But any difference no how matter how small can be stat sig if the sample is large enough. Its the effect size thats actually most important. And theres an effect there but its not that great. And so because the effect is relatively modest it just took one thing to go wrong (exposure) to produce a ns result.

Interventions that have a very large effect size come out sig no matter how much you cock things up. The problem generally is that as medical science has advanced the easy stuff has all been done and new interventions produce more modest improvements (smaller effect sizes) - hence you get more unstable trial results because stuff invariably goes wrong somewhere. Its how wrong that becomes the issue.

As I said not trying to defend departed Directors - suppose just making the point that Murpheys Law doesnt stop when you do research. Its all part of the elegance and cruety of the RCT design that bias from stuff going wrong will never (rarely anyway) work in your favour.

Anyway all the best with it.

Southoz