It’s not always a placebo, either. It’s my understanding that in testing of things like pain killers the control is often something like a benzodiazepine just because the patient understands that they are likely to feel relaxed/sleepy and a true placebo would essentially unblind the study.