I helped run a clinical trial for an antidepressant one time. It was a double-blind randomised within-subjects-replicated crossover design. All that control was

I helped run a clinical trial for an antidepressant one time. It was a double-blind randomised within-subjects-replicated crossover design. All that control was complicated, time consuming and very expensive. Only the hospital pharmacy knew whether participants were in treatment or placebo for a given session, and we never met the pharmacist, just received an orange vial. But we're pretty sure information still leaked, because drugs have side-effects. There's nothing to be done about this. Use an active placebo, you might say, but now you've added "get ethics approval to administer a nauseating placebo" to your 9-month-long ethics todo list. And then your placebo is worse than your active and the whole study is screwed because everyone spewed up in the MRI. Meatbags are tricky indeed.