You probably would have to have some kind of placebo control to figure it out what benefit/harm was actually due to the therapy vs. change over time. And then you'd have to control for things like what type of therapy you are getting CBT vs. Psychodynamic vs. DBT etc. And male vs. female therapist/client. And other client characteristics...like belief in the process, diagnosis, cooperation, intelligence, age. And other therapist traits...like age, experience, empathy, personality, etc.

Pills are much easier to research.