There's a lot of reversion to mean and the cyclic nature of bipolar sorts of depression to deal with, as well. Often there is likely no placebo "effect" that we can be certain of. Placebo doesn't have to have an effect, all you need is a base set of conditions where things might have improved, anyways.

We only know if a drug works if it shows better result than placebo. How do we know that placebo works, in a "who watches the watchers" sense, what is the placebo for placebo?