Of course, for actual efficacy, but you can tell if a treatment has a negative reaction or fails to effect the expected physiology in the expected way with a very small number of samples.

If everyone you give it to dies from reactions, or the expected effects seem to be happening on a physiological level, you don’t know much about efficacy but you do know a lot more about the eventual prospects for an actual clinical trial. It’s basically an extension of en-vidrio testing.