Isn't there a huge difference between in vivo and in vitro, though? If it's in vitro, you can actually physically check that it's working, and then rerun it to confirm. If it's in vivo and you're doing an RCT it's much more prone to statistical noise and the other issues, which is why it's bad to try things out randomly and see what sticks. At a test power of 0.05, one in twenty studies will stick due to chance.