As part of some work I'm into I've been reviewing a bit of literature on anti-depressants recently. Holy fuck is that a mine field. Literally every study the patients feel better in a matter of days no matter what the treatment. Looking for a signal in the noise is a really heavy lift. So the two things they can look for are magnitude of change in symptoms and persistence of change in symptoms. Seems like the placebo effect is generally as strong as the effect of the best drugs, but the best drugs tend to make people better for longer. It's pretty wacky altogether.
Complicating the issue is the exquisitely individualized response of patients to psychotropic drugs. You can't do a study with an n of 1 so you end up averaging everybody, and then saying "look it beats placebo" and then leaving it to the psychiatrist to puzzle it out one pill at a time, ten minutes at a time. Our system is deeply imperfect, and I am utterly bereft of the expertise to suggest anything better.