P dependence

I put this in a deck a while back - it’s remarkable how many of these decision errors are still built into the Development process in major pharma companies… Key is to be asking the question of what the process is designed to do...

I posted this back on LinkedIn at the same time as my friends at Exploristics posted this excellent piece on p values… P-value pitfalls in clinical trials

“P values can give a feeling of confidence that a study result is significant. However, does a low p value always mean what you think it does?”

Clearly this is an issue for phase III studies, when the regulators will be faced with a go-no go decision. However, I would argue that it is under-appreciated as a factor in earlier stage decisions, when a study ‘succeeding’ or failing on the hard p=0.05 measure dictates whether it progresses into further development. Many of the strategic errors above can be seen in this over-simplification…

Within a traditional safety-PoC-approval study process, so much is unexplored - however errors 3 and 4 above tend to lead to error 7 - there are so many paths that have no evidence at all.