Yeah, I get the impulse, and there are studies which try to partition subjects into “responders” and “non-responders”. The problem is if you don’t set out the p

Yeah, I get the impulse, and there are studies which try to partition subjects into “responders” and “non-responders”. The problem is if you don’t set out the partitions before you start the study then any ex post facto partitioning is fertile ground for p-hacking. The correct thing is to do the exploratory study, notice that some group seem to respond well, then run a new study “IF for the prevention of CVD in <specific subgroup>”. That’s a lot to ask, of course.