It's based on the assumption that the human body is a complex organism made up of interacting subsystems, and that there is a distinction between testing the effect of a change on one subsystem and testing the effect of a change on the entire organism. That's a very abstract assumption, and easily applies in this case. If you can't see anything in common between the human body and a computer program, you're just not looking at an abstract enough level. You can start with "they both exist" and work your way towards the concrete from there.